Microsoft SQL Server 2000 Programming by Example

Copyright © 2001 by Que® Corporation

All rights reserved. No part of this book shall be reproduced, stored in a retrieval system, or
transmitted by any means, electronic, mechanical, photocopying, recording, or otherwise,
without written permission from the publisher. No patent liability is assumed with respect to
the use of the information contained herein. Although every precaution has been taken in the
preparation of this book, the publisher and author assume no responsibility for errors or
omissions. Nor is any liability assumed for damages resulting from the use of the information
contained herein.
Library of Congress Catalog Card Number: 00-111702
Printed in the United States of America
First Printing: April, 2001
04 03 02 01 4 3 2 1
Trademarks
All terms mentioned in this book that are known to be trademarks or service marks have been
appropriately capitalized. Que Corporation cannot attest to the accuracy of this information.
Use of a term in this book should not be regarded as affecting the validity of any trademark or
service mark.
Microsoft is a registered trademark of Microsoft Corporation.
SQL Server is a trademark of Microsoft Corporation.
Warning and Disclaimer
Every effort has been made to make this book as complete and as accurate as possible, but
no warranty or fitness is implied. The information provided is on an "as is" basis. The authors
and the publisher shall have neither liability nor responsibility to any person or entity with
respect to any loss or damages arising from the information contained in this book or from the
use of the CD or programs accompanying it.
Credits
Acquisitions Editor
Michelle Newcomb
Development Editor
Sean Dixon
Managing Editor
Thomas Hayes
Project Editor
Tonya Simpson
Copy Editor
Kay Hoskin
Indexer
Kelly Castell
Proofreader
Maribeth Echard
Technical Editor
Vincent Mayfield
Team Coordinator
Cindy Teeters
Media Developer
Michael Hunter
Interior Designer
Karen Ruggles
Cover Designer
Duane Rader
Page Layout
Brad Lenser
Liz Patterson

ii
 Dedication
This book is dedicated to Carlos and Yumaira, who taught me the key
elements for success in life: discipline and persistence.
— Carlos Eduardo Rojas
To Manuela, for being my wife, my confidant, my shelter, my partner, warmth
for my winters, and refreshing breeze for my summers.
— Fernando Guerrero

iii
 About the Authors

About the Authors

Carlos Eduardo Rojasis a program manager with Planeta Networks, an Internet company headquartered in
Coral Gables, Florida that provides broadband applications to Internet service providers in Ibero-America. He
specializes in the design of n-tier applications, database implementation, and administration of SQL Server
databases. Prior to this role, he was a consultant and trainer with Manapro in Caracas, Venezuela, where he
is originally from. Also, he has participated as a speaker in various TechNet conferences in Venezuela.
Carlos earned a B.S. degree in Computer Science from University Simón Bolívar, Caracas, Venezuela. He is
a Microsoft Certified Systems Engineer + Internet (MCSE+I), Microsoft Certified Database Administrator
(MCDBA), Microsoft Certified Sales Specialist (MCSS), and has been awarded with the MVP (Most Valuable
Professional) status on SQL Server. He is also a voting member and volunteer of PASS, the professional
association for SQL Server.
Carlos can be reached at carlos@sqlserverbyexample.com.
Fernando G. Guerrerois a principal technologist and SQL Server product consultant in QA, United Kingdom.
His main role involves IT training, course development, and internal mentoring.
He writes for SQL Server Magazine (http://www.sqlmag.com), presented a session on SQL Server 2000
at TechEd 2000 Europe, and has accepted to speak at PASS2001, TechEd Europe 2001, VBUG 2001, VBITS
2001, VSLive, and SQL2THEMAX conferences during the year 2001.
He is a Civil and Hydrologic Engineer with almost 20 years'experience in software development and design, in
many cases applied to engineering environments.
He holds seven Microsoft Professional Certifications, including MCSE+Internet, MCSD, MCDBA, MCT, and
has been awarded with the MVP (Most Valuable Professional) status on SQL Server. He is also a voting
member and volunteer of PASS, the professional association for SQL Server.
His professional experience covers six years (1981–1987) as a lecturer in the Valencia's Polytechnic
University (Spain, http://www.upv.es), where he was lecturing on surveying, photogrammetry, technical
drawing, and applied numerical methods in the Civil Engineering School and the Agricultural Engineering
School.
He built his own software company in Spain, TOU S.A., focused on desktop publishing and graphical tools for
the professional market, and was technical director in that company for four years (1987–1991).
Before joining QA (Nov. 1998), he spent eight years (1991–1998) as an international consultant for a
prestigious Spanish engineering firm (http://www.inypsa.es), living in different countries, designing,
developing, and implementing information systems for projects financed by the World Bank, the European
Union, and the European Investment Bank.
Fernando can be reached at fernan@sqlserverbyexample.com.

v
 Acknowledgments

Acknowledgments
First of all, I want to thank all my family, friends, and co-workers, all of whom made this book possible. I wish I
could name each and every one of you who contributed directly or indirectly to this book.
To my family, Carlos, Yumaira, Jesús Eduardo, and María Angélica. You were a great source of motivation for
this book. Also, a special gratitude to my uncle and aunt, José and Marí a; the CD you gave me helped me
relax during those endless nights writing the book.
To all my teachers, professors, and classmates from high school and university— I'll never forget those
wonderful years of my life and all the moments we shared. Special thanks to Juan Carlos Guzmán, who, since
then, has been a great source of support.
Thanks to all the teachers at the English Language Institute (ELI), University of Pittsburgh, for helping me
improve my writing skills, especially Stephanie Maietta-Pircio, Dorolyn Smith, Holly Stark, and Lois Wilson.
Also, thanks to Professor Janusz Szczypula from Carnegie Mellon University for all his help and support
during my time in Pittsburgh.
During my work experience, I've met a lot of exciting people who, in one way or the other, have helped me
grow in my professional career. Thanks to all of you in Database ACCESS, NetPeople, Manapro, and Planeta
Networks. Special thanks to José Alberto Nuñez, Carlos Guanchez, and María Dolores Nardi. Also, thanks to
the extraordinary Planeta Networks team, especially Rodolfo Tancredi, who always has been willing to guide
me since I began to work for Planeta Networks.
I want to express special gratitude to the group of SQL Server MVPs for honoring me as a new member. This
has been one of the most important awards in my professional career. In particular, thanks to Fernando
Guerrero, a great friend and very talented professional, for taking the time to review all the chapters of the
book.
Thanks to Holly Allender for making my dream of writing a book a reality. The editorial team at Que Publishing
also deserves my gratitude, in particular Michelle Newcomb for her patience and understanding throughout
the whole process of writing the book, and for her dedication and persistence to meet the deadlines. Thanks
to all the editorial team: Sean Dixon, Vincent Mayfield, Tonya Simpson, and Kay Hoskin.
Last, but not least, my most sincere thanks to those who believed in me, and to those who, by reading this
book, will inspire me for future publications.
Carlos Eduardo Rojas
January 2001
My life has had plenty of amazing opportunities and challenges, and I was very fortunate to meet amazing
people along the way. Each one of them has contributed to what I am now, and what I will be in the future. I
would like to pay tribute to them, as an honest gratitude gesture, for all the help that they generously gave me.
In chronological order:
To my father, Fernando, the intelligent, honest, and caring person, who gave me his love, dedication, and help
and taught me the importance of quality at work. To my mother, Maruja, who gave me her love, optimism, and
unbreakable happiness. They both are responsible for most of the good values that guide my life. They built a
great family, and I only hope to live long enough to enjoy their company. To my brothers and sisters: Ana,
Inmaculada, Carlos, Rocí o, and José Ignacio, I hope you know how important you are to me. I wish I could
spend more time with all of you.
To Professor Manuel Chueca, excellent professor and amazing human being, who gave me the opportunity to
teach on his team and helped me beyond any reasonable limit. To Dr. José Herráez, who generously gave
me his full support and friendship. To Dr. Luis Angel Alonso, who always helped me move forward. It was for
me an honor learning from all of you the joy of teaching. I miss you and the years we spent together.
To Tom Peters, whose books have inspired me for almost 15 years.
To Bernardo Cebolla and Vicente Cebolla, excellent friends and partners. We lived together an unforgettable
business and human experience, during the first years of the personal computing market. Our friendship will
remain forever.
Inypsa, one of the best Spanish engineering firms, gave me the opportunity to work on important international
projects around the world for almost eight years. I'd like to thank specially Juan Hernández, Casimiro del Pozo,
and José Luis Muñoz, for their trust, professionalism, and friendship. I wish you all the best.
During those years in Inypsa, I had the privilege of meeting excellent professionals and friends along the way.
To Reynaldo Barbosa, who continuously encouraged me and wisely advised me to join Inypsa; to Javier Gras,
excellent engineer and friend; to José María Pastor, my brother-in-law, dear friend, and excellent engineer; to
the amazing professionals and friends who worked with me during those years: Esther Pineda, Poernomo
Widrobo, Alvaro Chucatiny, Ludwing Céspedes, David Plaza, José Luis Sacristán, Samuel Sánchez, Oscar
Rocha, Ví ctor Hugo Durán, and Anil Pillai.

vii
Microsoft SQL Server 2000 Programming by Example

I want to thank Patrick Beasley, Jonathan Finch, Mike Shammas, and Barbara Savage for giving me the
opportunity to work in the best learning environment in the world: QA. I wish to thank Bill Walker for his
continuous support. To Patrick Beasley and Aaron Johal, you both offered me your hand from the very first
day, when I needed it most. Working for QA represents for me the possibility to learn from the greatest
training team you can ever imagine.
As a SQL Server MCT, I spend some time with other SQL Server MCTs in an amazing private newsgroup
where we share our experiences, fears, challenges, and achievements. Among these great MCTs, I would
want to express my gratitude to Dejan Sarka, one of the best SQL Server trainers of this galaxy and an
excellent and generous friend. I will always remember the excitement I felt when I finally met Itzik Ben-Gan.
Itzik is one of the most recognized SQL Server experts, an excellent friend, and the one who makes me work
at warp speed many Fridays with his puzzles. One sunny day at San Diego, Dejan and Itzik, by surprise,
introduced me to Kalen Delaney, and I felt like a novice actor meeting John Ford. I cannot forget other great
SQL Server MCTs, such as Ted Malone, Chris Randall, Robert Vieira, Tibor Karaszi, Victor Isakov, Aaron
Johal, and many others.
Last year I was honored with the SQL Server MVP award. My most sincere gratitude to Juan T. Llibre (ASP
and IIS MVP), Carlos Sánchez (Microsoft Spain), and Alberto Borbolla (VB MVP) for generously proposing me
as an MVP, and to the other SQL Server MVPs for accepting me on their team. I still cannot believe that I am
part of the amazing SQL Server MVP group. It is easy to feel small being surrounded by Bob Pfeiff, B.P.
Margolin, Brian Moran, Carlos Eduardo Rojas, Darren Green, Dejan Sarka, Gianluca Hotz, Itzik Ben-Gan,
Kalen Delaney, Michael Hotek, Neil Pike, Olivier Matrat, Ron Talmage, Roy Harvey, Russell Fields, Sharon
Dooley, Tibor Karaszi, Tony Rogerson, Trevor Dwyer, Umachandar Jayachandran, and Wayne Snyder.
Together we try to help the SQL Server community in different ways, mainly providing free user support in the
SQL Server newsgroups. We share ideas, wishes, and experiences in the most challenging newsgroup you
could imagine, together with a selected group of Microsoft support engineers and members of the SQL Server
developer team.
From the SQL Server group at Microsoft, I wish to thank Gert Drapers, Euan Garden, Lubor Kollar, Jim Gray,
Tom Barclay, Hal Berenson, Don Vilen, Adam Shapiro, Margo Crandall, Karl Dehmer, LeRoy Tutle, Rick
Byham, Shawn Aebi, Steve Dibbing, and Peter Kalbach. Their SQL Server courses, presentations, white
papers, classes, messages, and continuous support helped me understand this technology a little bit more
every day. And especially to Richard Waymire, the most knowledgeable SQL Server professional I ever met—
attending any of your speeches was a tremendous learning experience for me.
I wish to express my gratitude to the great professionals who made the Spanish SQL Server newsgroup one
of the best public SQL Server newsgroups. Among them: Antonio Soto, Carlos Paz, Carlos Sacristán, Claudio
Alabarce, Deman Thierry, Eladio Rincón, Emilio Bocau, Jesús López, Jorge López, Julián Valencia, Mariano
Melgar, Miguel Ángel Sanjuán, Miguel Egea, Norman Armas, Rodrigo Estrada, and Salvador Ramos.
I wish to thank Michelle Crocket, Kathy Blomstrom, Carol Martin, and the amazing technical edit team at SQL
Server Magazine for their continuous support. Writing for SQL Server Magazine is nothing but a pleasure
when you're surrounded by these great professionals.
To Carlos Rojas, the generous friend who gave me the opportunity to co-write this book, I will always thank
you for this great opportunity. Your continuous support to the SQL Server users'community, and especially to
the Spanish SQL Server newsgroup, proves your tremendous generosity and incredible knowledge level.
Writing a book like this would be impossible without the continuous help and support from the Que Publishing
editorial team: Vincent Mayfield, Sean Dixon, Kay Hoskin, Tonya Simpson, and especially Michelle Newcomb.
I am really impressed by their review process. However, if you still find any mistakes, or something you don't
like, in this book, you can blame only the authors.
Thanks to Ian Dolan, who helped me correct the style on my first chapters of this book.
Finally, I want to thank my wife, Manuela, and my daughters, Rocío, Marta, and Marina. They were extremely
supportive during these months that I've been working full time, day and night, writing this book. They came to
my room from time to time to give me a big smile, a kiss, a hug, and to tell me how much they love me,
despite the fact that I could not spend much time with them. Thank you for helping me fulfill this dream. Now
you can have your revenge— I promise to spend more quality time with you from today.
Fernando G. Guerrero
January 2001

viii
 Introduction

Introduction
The by Example Series

How does the by Example series make you a better programmer? The by Example series teaches
programming using the best method possible. After a concept is introduced, you'll see one or more examples
of that concept in use. The text acts as a mentor by figuratively looking over your shoulder and showing you
new ways to use the concepts you just learned. The examples are numerous. While the material is still fresh,
you see example after example demonstrating the way you use the material you just learned.
The philosophy of the by Example series is simple: The best way to teach computer programming is by using
multiple examples. Command descriptions, format syntax, and language references are not enough to teach a
newcomer a programming language. Only by looking at many examples in which new commands are
immediately used and by running sample programs can programming students get more than just a feel for
the language.

Who Should Use This Book?

Microsoft SQL Server 2000 Programming by Example is targeted toward people with previous experience in
any programming language.
As a database programming book, we expect you to have some background knowledge about logical
database design. Understanding how to define entities, attributes, and relationships between entities is
essential in producing any good database system. We will provide you with some comments about this
subject when required, but we will not go into deeper detail. If you feel uncomfortable about this subject, we
suggest that you read a general database design book first.
No prior experience in Transact-SQL is necessary; however, if you have experience working with the SQL
language, from any other database system, this book can be used as a reference in which you will find a lot of
useful examples that you can use to program applications in SQL Server.
If you do have experience with any previous version of SQL Server, you will find many examples that you can
use to practice the extended functionality of SQL Server 2000. However, this is not an upgrading book for
users of previous versions, so we do not assume any prior knowledge of previous versions.
If you are a Web developer, this book can teach you how to use SQL Server's new XML functionality to
access data from the Internet. If you are a SQL Server developer and you want to introduce yourself to the
new XML world, you can find in this book some useful examples on how to use this exciting new functionality.
Learning a new programming language is a mixture of theory and practice. We try to provide as many
examples as possible about every topic. We advise you to apply these new concepts as soon as possible in a
real scenario, because this is the best way to reinforce your learning effort. If you are not working in a
database design right now, create your own personal database to manage appointments, books, pictures, or
your personal music library.

This Book's Organization

This book provides you with the skills needed to develop and maintain SQL Server applications. Also, it
contains the enhancements introduced in SQL Server 2000.
We highly recommend that you go over all the examples in this book. They were designed to help you
understand each concept and feature of Transact-SQL. You can use Query Analyzer, which is explained in
Appendix B, "Using SQL Query Analyzer," to execute all examples presented in this book.
Commonly, there are some tasks that can be performed using Enterprise Manager instead of Transact-SQL.
Be aware that every task that you perform in Enterprise Manager translates to a set of instructions in
Transact-SQL executed behind the scenes. Because the purpose of this book is to teach you the Transact-
SQL language, examples are based in Transact-SQL and, in some specific cases, the way to perform the task
in Enterprise Manager is also explained.
Appendix A, "Using SQL Server Instances," shows you how to use one of the new features of SQL
Server 2000, multi-instance support. This appendix is useful to practice the distributed queries examples that
appear in Chapter 15, "Working with Heterogeneous Environments: Setting Up Linked Servers."
Chapter 6, "Optimizing Access to Data: Indexes," is an advanced chapter that teaches you how to
optimize access to databases using indexes efficiently. The information contained in this chapter, although

ix
Microsoft SQL Server 2000 Programming by Example

very important, is not essential to understand the next chapters. You can read this chapter when you feel
confident enough using SQL Server 2000 and you want to optimize your database.
As a programming book, we deliberately do not cover administration subjects. However, we do include
information about two important administrative subjects:

• Security— Understanding how SQL Server 2000 manages security is crucial to create a secure
database. In Chapter 1, "Relational Database Management Systems and SQL Server," we
explain this subject in detail from a programmer's point of view.
• Importing and exporting data— Because almost every programmer must import and export data from
time to time, we cover this subject in Chapter 14, "Transferring Data to and from SQL Server."

Chapter 16, "Working with XML Data in SQL Server 2000," is not included in this book's printed material. Due
to the late availability of the latest version of the SQL Server 2000 Web Release, we had to provide this
chapter in an online format only. You can download Chapter 16 from the http://www.mcp.com site
(http://www.mcp.com/que/byexample_que.cfm).
Unfortunately, some exciting new features are not covered in this book, such as

• SQL Server 2000 Administration

• Data Transformation Services Development— however, we cover basic creation of DTS packages
• English Query
• Full-Text Search
• Analysis Services

SQL Server 2000 is a vast product, and we decided to focus on what we consider to be the basic set of
features that every database developer should know about SQL Server 2000.
An evaluation version of SQL Server 2000 Enterprise Edition is included in this book. If you don't have SQL
Server installed on your machine, this evaluation version can be used to practice with all examples shown
throughout the book.
This book prepares you for one of the core exams of the Microsoft Certified Database Administrator (MCDBA)
certification: Exam 70-229 Designing and Implementing Databases with Microsoft SQL Server 2000
Enterprise Edition. This exam is also an elective of the Microsoft Certified Systems Engineer (MCSE)
certification. For details on this exam, you can visit Microsoft's Web site at
http://www.microsoft.com/trainingandservices/exams/examasearch.asp?PageID=70-229.

Conventions Used in This Book

Examples are identified by the icon shown at the left of this sentence.
Listing, code, Transact-SQL keywords, and object names appear in monospace font, such as

EXEC sp_help

Many examples contain output, either as warning and error messages, and result sets.
In those cases, you can identify the output by the icon shown at the left of this sentence.

x
 Introduction

We do not show the output for some examples if the output is obvious, irrelevant, or does not offer any benefit
to the reader. In general, we prefer to show the output, so you can check whether you executed the example
properly.
Note

Special notes augment the material you read in each chapter. These notes clarify concepts and
procedures.

Tip

You'll find numerous tips offering shortcuts, tested tricks, and solutions to common problems.

Caution

The cautions warn you about common problems and misconceptions when writing Transact-SQL
code. Reading the caution sections will save you time and trouble.

What's Next?

Microsoft SQL Server 2000 is a powerful tool, capable of managing big-scale databases fast and efficiently.
However, not even the most powerful hardware and the best relational database management system can
improve a poorly designed database application.
Learning the Transact-SQL language will help you create an efficient, versatile, and feature-rich database
application.
Please visit the by Example Web site for code examples or additional material associated with this book:
http://www.mcp.com/que/byexample_que.cfm
You can find comments, error logs, and additional code about this book on its own Web site:
http://www.sqlserverbyexample.com
You can contact the authors by email at
Carlos Eduardo Rojas: carlos@sqlserverbyexample.com
Fernando G. Guerrero: fernan@sqlserverbyexample.com
The public Microsoft newsgroups represent an amazing learning opportunity as well. You can find free support
from other SQL Server colleagues, SQL Server MVPs, and members of the Microsoft SQL Server group:

news://msnews.microsoft.com/microsoft.public.sqlserver.programming
news://msnews.microsoft.com/microsoft.public.sqlserver.ce
news://msnews.microsoft.com/microsoft.public.sqlserver.clients
news://msnews.microsoft.com/microsoft.public.sqlserver.clustering
news://msnews.microsoft.com/microsoft.public.sqlserver.connect
news://msnews.microsoft.com/microsoft.public.sqlserver.datamining
news://msnews.microsoft.com/microsoft.public.sqlserver.datawarehouse
news://msnews.microsoft.com/microsoft.public.sqlserver.dts
news://msnews.microsoft.com/microsoft.public.sqlserver.fulltext
news://msnews.microsoft.com/microsoft.public.sqlserver.mseq

xi
Microsoft SQL Server 2000 Programming by Example

news://msnews.microsoft.com/microsoft.public.sqlserver.odnc
news://msnews.microsoft.com/microsoft.public.sqlserver.olap
news://msnews.microsoft.com/microsoft.public.sqlserver.replication
news://msnews.microsoft.com/microsoft.public.sqlserver.security
news://msnews.microsoft.com/microsoft.public.sqlserver.server
news://msnews.microsoft.com/microsoft.public.sqlserver.setup
news://msnews.microsoft.com/microsoft.public.sqlserver.tools
news://msnews.microsoft.com/microsoft.public.sqlserver.xml
news://msnews.microsoft.com/microsoft.public.es.sqlserver
news://msnews.microsoft.com/microsoft.public.espanol.sqlserver.administracion
news://msnews.microsoft.com/microsoft.public.espanol.sqlserver.olap
news://msnews.microsoft.com/microsoft.public.fr.sqlserver
news://msnews.microsoft.com/microsoft.public.ae.arabic.sqlserver
news://msnews.microsoft.com/microsoft.public.arabic.sqlserver
news://msnews.microsoft.com/microsoft.public.de.sqlserver
news://msnews.microsoft.com/microsoft.public.il.hebrew.sqlserver
news://msnews.microsoft.com/microsoft.public.jp.sqlserver.server
Microsoft provides many white papers on SQL Server at the following address:
http://www.microsoft.com/sql/Index.htm
Go to Chapter 1 and start learning Microsoft SQL Server 2000 Programming by Example today!

xii
 Contents

Contents

About the Authors............................................................................................................................ v

Acknowledgments .........................................................................................................................vii
Introduction...................................................................................................................................... ix
The by Example Series ............................................................................................................. ix
Who Should Use This Book? ................................................................................................... ix
This Book's Organization .......................................................................................................... ix
Conventions Used in This Book................................................................................................ x
Contents.......................................................................................................................................xiii
Chapter 1. Relational Database Management Systems and SQL Server .............................1
Database Models ........................................................................................................................1
A Brief History of SQL Server....................................................................................................2
Basics of SQL Server Architecture ...........................................................................................4
Client/Server Applications Design ..........................................................................................41
Chapter 2. Elements of Transact-SQL .......................................................................................43
Data Definition Language (DDL).............................................................................................43
Data Manipulation Language (DML) ......................................................................................49
Data Control Language (DCL) ................................................................................................50
Data Types .................................................................................................................................54
Additional Elements ..................................................................................................................62
Programming Scripts and Batches .........................................................................................77
Chapter 3. Working with Tables and Views ..............................................................................81
Creating and Altering Tables ...................................................................................................81
Creating and Altering Views ....................................................................................................97
Chapter 4. Querying and Modifying Data ................................................................................109
Querying Data ..........................................................................................................................109
Modifying Data .........................................................................................................................140
Chapter 5. Querying Multiple Tables: JOIN s .........................................................................153
ANSI SQL-92 Syntax..............................................................................................................153
INNER JOIN............................................................................................................................156
OUTER JOINs .........................................................................................................................164
CROSS JOINs .........................................................................................................................171
Self Joins ..................................................................................................................................174
The UNION Operator ..............................................................................................................175
Chapter 6. Optimizing Access to Data: Indexes.....................................................................181
Introduction to Indexes ...........................................................................................................181
Benefits of Indexes..................................................................................................................182
How to Create Indexes ...........................................................................................................190
How SQL Server 2000 Stores Data .....................................................................................193
How SQL Server 2000 Modifies Data ..................................................................................195
Index Enhancements in SQL Server 2000..........................................................................195
Accessing Data Without Indexes: Table Scan....................................................................196
Types of Indexes .....................................................................................................................196
Covered Queries and Index Intersection.............................................................................206
Index Maintenance ..................................................................................................................209
Indexes on Computed Columns ............................................................................................219

xiii
Microsoft SQL Server 2000 Programming by Example

Indexed Views ..........................................................................................................................220

Index Tuning Wizard ...............................................................................................................223
Summary...................................................................................................................................227
Chapter 7. Enforcing Data Integrity..........................................................................................229
Types of Data Integrity ...........................................................................................................229
Enforcing Integrity: Constraints (Declarative Data Integrity) ............................................230
Chapter 8. Implementing Business Logic: Programming Stored Procedures ...................279
Benefits of Using Stored Procedures ...................................................................................279
Types of Stored Procedures ..................................................................................................280
Creating and Dropping Stored Procedures .........................................................................284
Using Parameters....................................................................................................................288
Altering Stored Procedure Definitions ..................................................................................291
The RETURN Statement ..........................................................................................................292
Executing Stored Procedures................................................................................................293
Stored Procedure Recompilation..........................................................................................301
Handling Errors ........................................................................................................................302
Nesting Stored Procedures....................................................................................................304
Application Security Using Stored Procedures...................................................................307
Chapter 9. Implementing Complex Processing Logic: Programming Triggers .................309
Benefits of Triggers .................................................................................................................309
Trigger Enhancements in SQL Server 2000 .......................................................................321
Inserted and Deleted Tables..........................................................................................321
Types of Triggers According to Their Order........................................................................328
Creating and Dropping Triggers............................................................................................336
Altering Trigger Definitions ....................................................................................................345
Disabling Triggers ...................................................................................................................346
Nesting Triggers ......................................................................................................................347
Recursive Triggers ..................................................................................................................354
Security Implications of Using Triggers ...............................................................................357
Enforcing Business Rules: Choosing Among INSTEAD of Triggers, Constraints, and
AFTER Triggers.......................................................................................................................358
Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF) ..........................361
Benefits of User-Defined Functions ......................................................................................361
Built-In User-Defined Functions ............................................................................................361
Types of User-Defined Functions According to Their Return Value ...............................368
Dropping User-Defined Functions ........................................................................................410
Preventing the Alteration of Dependent Objects:The SCHEMABINDING Option...........411
Deterministic and Nondeterministic Functions ...................................................................414
Altering User-Defined Functions Definition .........................................................................416
Security Implications of Using User-Defined Functions ....................................................416
Applying User-Defined Functions .........................................................................................417
Chapter 11. Using Complex Queries and Statements ..........................................................423
Subqueries ...............................................................................................................................423
Correlated Subqueries............................................................................................................441
Derived Tables.........................................................................................................................445
The CASE Function ................................................................................................................447
The COMPUTE Clause ..........................................................................................................451
The CUBE and ROLLUP Operators........................................................................................455
Using Hints ...............................................................................................................................459
Chapter 12. Row-Oriented Processing: Using Cursors ........................................................463

xiv
 Contents

Row-by-Row Versus Set-Oriented Processing ..................................................................463

Types of Cursors .....................................................................................................................470
Steps to Use Cursors..............................................................................................................477
Scope of Cursors.....................................................................................................................493
Using Cursors to Solve Multirow Actions in Triggers ........................................................498
Application Cursors .................................................................................................................499
Chapter 13. Maintaining Data Consistency: Transactions and Locks ................................503
Characteristics of Transactions (ACID) ...............................................................................503
Using Transactions .................................................................................................................503
Concurrency Problems ...........................................................................................................521
Isolation Levels ........................................................................................................................529
Types of Locks.........................................................................................................................535
A Serious Problem to Avoid: Deadlocks..............................................................................555
Chapter 14. Transferring Data to and from SQL Server .......................................................559
The Need for Transferring Data ............................................................................................559
Tools for Transferring Data Using SQL Server 2000 ........................................................560
The BULK INSERT Statement and bcp..............................................................................561
Using Data Transformation Services ...................................................................................579
The Copy Database Wizard ..................................................................................................596
Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers ....607
Distributed Queries .................................................................................................................607
Distributed Transactions ........................................................................................................637
Appendix A. Using SQL Server Instances ..............................................................................643
Installing SQL Server Instances............................................................................................643
Connecting to Instances.........................................................................................................653
System Functions Used in Multi-Instance Installations .....................................................657
Current Limitations ..................................................................................................................658
Appendix B. Using SQL Query Analyzer .................................................................................661
Installing SQL Query Analyzer ..............................................................................................661
The Query Analyzer Workplace ............................................................................................662
Managing User and Connection Options ............................................................................671
Defining and Using Templates for Query Analyzer............................................................679
Analyzing Queries ...................................................................................................................681
Working with the Transact-SQL Debugger .........................................................................685
Summary...................................................................................................................................687

xv
 Chapter 1. Relational Database Management Systems and SQL Server

Chapter 1. Relational Database Management Systems

and SQL Server
From the beginning of human history, knowledge has meant power. The success or failure of individuals,
companies, and countries depends on the amount and quality of knowledge they have about their
environment.
Knowledge is based on facts. In some cases, facts are made of abstract data, difficult to represent in precise
mathematical terms. However, the economic life of every company relies on precise data obtained from
external and internal sources.
Knowledge management is based on the ability to use this absolute data to interpret reality and arrive at
conclusions about how their environment reacts to specific conditions.
Data has value if it is accurate and comprehensive enough to serve business needs. However, the way the
data is stored and the mechanisms available to retrieve it are important factors to consider. Database
management systems provide reliable data storage systems and flexible data retrieval tools.
In this book, you learn how to develop database applications, using one of the latest and more powerful
database management systems: Microsoft SQL Server 2000.
This chapter teaches you the main concepts of Microsoft SQL Server 2000:

• Basic concepts about relational database systems

• SQL Server architecture and server components
• SQL Server client tools
• How to protect your data in SQL Server
• Basic principles about client/server application design and how SQL Server fits in this model

Database Models
To provide the required durability, data is stored in physical storage devices. These files are stored in different
logical formats depending on the database model selected by every particular database management system.
You can find many database models in the database market:

• Flat files
• Hierarchical
• Networked
• Relational
• Object
• Object-relational
• Document

Flat files, hierarchical, and networked models are mainly used in mainframes, whereas the other models have
been ported to client/server environments, based on personal computers. Discussing these database models
is out of the scope of this book. SQL Server 2000 implements the relational model, and the following section
teaches you the basic theory behind this popular database model.

The Relational Model

In the relational model, data is arranged in tables in which the physical location of every value is not
permanently predefined and is transparent to the data retrieval strategy. Every table is defined with a fixed set
of columns that map the entity attributes.
Data from different tables is related by logical links that are dynamically defined by the database application or
by the end user who sends a data request. Figure 1.1 shows a typical example of a relational database.

Figure 1.1. The relational model arranges data in tables with logical dynamically defined links.

1
Microsoft SQL Server 2000 Programming by Example

Users access data using an industry standard query language, called SQL. This means that the database
design focuses mainly on the data to store, producing a flexible database that can be used by many different
applications. This flexibility contrasts with the databases stored in a hierarchical or networked model, in which
the database structure was designed to solve a specific business problem.
As you can imagine, this flexibility represents more complexity for the database engine. This is the reason you
can expect better performance for the same hardware platform, using hierarchical or networked databases
rather than relational databases. However, the continuous improvements on relational database management
systems (RDBMS) is switching the database market to this technology, even at mainframe level, where a big
percentage of available data is still in hierarchical and networked format.
The relational model is based on the relational theory, in which modifications to the data are based on
relational algebra and relational calculus.
Dr. E. F. Codd, an IBM mathematician, published A Relational Model of Data for Large Shared Data Banks
(Communications of the ACM, Vol. 3, No. 6, June 1970). This document establishes the rules of the relational
databases. Many database vendors started to implement his theories soon after this publication. IBM DB2,
Oracle, Sybase, and Microsoft SQL Server are typical RDBMS products.
The language used to request information from RDBMS, SQL, is part of the ANSI standard since the ANSI
SQL 86 version. Many products are based on the ANSI SQL 89 and ANSI SQL 92 standards, but every
product offers different extensions. A new standard is available now, ANSI SQL 99, which expands the
traditional relational model nearer to the object-oriented model.
Microsoft SQL Server 2000 is based in a special SQL dialect, called Transact-SQL, which is an expanded
version of the ANSI SQL 92 standard.

A Brief History of SQL Server

Figure 1.2 shows the SQL Server timeline in which you can see the evolution of SQL Server compared with
the evolution of the Windows operating systems, the Intel and AMD processors, and the typical CPU speed
available at that time.

Figure 1.2. The SQL Server timeline.

2
 Chapter 1. Relational Database Management Systems and SQL Server

Microsoft SQL Server was born as a joint commercial venture between Sybase and Microsoft and signed on
March 27, 1987. It was developed for the 16-bit OS/2 platform, an operating system jointly developed between
IBM and Microsoft. SQL Server started its commercial life with the commercial support of Ashton-Tate, whose
mainstream database product, dBase, was the market leader at that time. Ashton-Tate/Microsoft SQL Server
1.0 arrived on the market in May 1989.
If you do not remember those remote years, personal computers were equipped with Intel I286 and I386
processors, and the I486 processor was just a newcomer. Personal computers had processors running at
12MHz and 16MHz and had a typical 20MB–40MB hard disk. Perhaps you do not remember those years
where RAM was measured in kilobytes.
During those first years, Sybase developed Microsoft SQL Server for OS/2 and Microsoft commercialized and
supported the product on this platform. Ashton-Tate abandoned the project soon after the first version.
Starting in 1992, Microsoft SQL Server 4.2 for OS/2 was released. In this case, it was a joint development
between Sybase and Microsoft, with active development collaboration from Microsoft. It was still a 16-bit
database product, running on a 16-bit operating system (OS/2 1.3). In that personal computer market, ruled
by 32-bit processors (Intel I386, I486, and AMD AM386) at more than 33MHz and running Windows 3.1,
working with a 16-bit backend database was not a good selling point.
About a year later, Microsoft SQL Server 4.2 for Windows NT was released. This product was the final point in
the joint development agreement between Sybase and Microsoft, although Sybase code remained in SQL
Server code for several more versions to come— up to version 6.5.
For Microsoft this was a no-way-back decision, and SQL Server has since been a Windows-only product. This
was the first 32-bit release and for many customers was, perhaps, just an excuse for buying the new Windows
NT 3.1 operating system.
Remember, 1995 was the year of Windows 95, the Pentium processor, and amazing 150MHz CPU speeds on
personal computers equipped with hard disks as big as 400MB or 600MB and with 4MB–8MB of RAM. That
year was the release of the new SQL Server 6.0 for Windows NT. It didn't have the new Windows 95 interface,
but it included new features, which made this product an important contender in the database market.
SQL Server 6.5 was the final version of the Sybase-based era. This version included client tools based on the
widely accepted Windows 95 interface, which runs on Windows NT 4.0.
Starting in 1999, a brand-new version came to the market: Microsoft SQL Server 7.0. It was a completely new
product, with exciting new tools, an enhanced database engine, and graphical user interface inherited from
the popular DaVinci Tools (or Visual Database Tools) already available in the Enterprise Edition of Microsoft
Visual Studio. New services, such as Full-Text Search, English Query, Data Transformation Services (DTS),
and OnLine Analytical Processing (OLAP), as well as a faster database engine, made this product a big
market success.

3
Microsoft SQL Server 2000 Programming by Example

This was a Microsoft-only database engine, in a nearly Microsoft-only personal computing market, where
processors ran at 200MHz–300MHz, hard disk sizes were already measured in gigabytes, and RAM available
in personal computers was more than 32MB.
The end of the second millennium was the birth of the newest SQL Server version which, to be in sync with all
the new Microsoft products of that year, was named Microsoft SQL Server 2000.
Changes have occurred since 1989. Corporations look at the PC market searching for powerful servers
equipped with Intel or AMD processors, running Windows 2000 and Windows 2000-compatible server
applications and services. The highest database transaction benchmarks are based on PC-like servers—
running Windows 2000— and the database market has changed forever.
The Microsoft SQL Server developer team does not stop here. They are already writing new versions that will
appear on the market soon, improving Internet support, providing support for new 64-bit Windows versions
and 64-bit processors, the new Windows XP operating system, new file systems, scaling up, out, and down,
and adding more functionality to this database management system .

Basics of SQL Server Architecture

SQL Server is a client/server relational database management system. Figure 1.3 shows the process that
every query must follow, from its origin as a SQL query in the client application running in the client computer,
to the final result set received by the client application.

Figure 1.3. Client access to a SQL Server database.

These steps are defined as follows:

1. The user selects an option in a client application. This option calls a function in the client application
that generates a query that is sent to SQL Server. The application uses a database access library to
send the query in a way SQL Server can understand.
2. The database library transforms the original request into a sequence of one or more Transact-SQL
statements to be sent to SQL Server. These statements are encapsulated in one or more Tabular
Data Stream (TDS) packets and passed to the database network library to be transferred to the server
computer.
3. The database network library uses the network library available in the client computer to repackage
the TDS packets as network protocol packets.
4. The network protocol packets are sent to the server computer network library across the network,
where they are unwrapped from their network protocol.

4
 Chapter 1. Relational Database Management Systems and SQL Server

5. The extracted TDS packets are sent to Open Data Services (ODS), where the original query is
extracted.
6. ODS sends the query to the relational engine, where the query is optimized and executed in
collaboration with the storage engine.
7. The relational engine creates a result set with the final data and sends it to ODS.
8. ODS builds one or more TDS packets to be sent to the client application, and sends them to the
server database network library.
9. The server database network library repackages the TDS packets as network protocol packets and
sends them across the network to the client computer.
10. The client computer receives the network protocol packets and forwards them to the network libraries
where the TDS packets are extracted.
11. The network library sends the TDS packets to the database access library, where these packets are
reassembled and exposed as a client result set to the client application.
12. The client application displays information contained in the result sets to the user.

These are some important points to consider about this process:

• The client application sends only a database request to the server computer.
• SQL Server sends only the final result set to the client application, saving network bandwidth.
• SQL Server receives the SQL request and sends result sets in return. SQL Server does not spend
server resources on user interaction.
• SQL Server is not responsible for the final output format; it is the responsibility of the client application.
The client application does not use client resources to solve low-level query solving and data access
processes.
• The client application can be designed independently of the database system used in the back end.
Data access operations are based in a high-level data access library, which can be easily changed to
connect to other types of database systems.
• The client application is not aware of the network protocol used to connect to the server, and this
protocol can be changed at any time, provided that the server and client share a common protocol .

When you install SQL Server 2000, you install two different sets of components:

• Server components are back-end services, responsible for data storage, data integrity, security,
concurrency, and so on.
• Client components are front-end applications, used by administrators, developers, and even end
users, to administer, develop, test, and use a SQL Server 2000 database system.

Server Components

What we call SQL Server 2000 is actually a collection of several Windows services:

• Microsoft SQL Server service (MSSQLServer)— The main service, responsible for data storage, data
integrity, consistency, concurrency, security, query processing, optimization, and execution.
• Microsoft SQL Server Agent (SQLServerAgent)— Responsible for scheduling jobs, managing alerts,
and Notifying operators. SQL Server Agent is an important service in SQL Server Administration
because so many administrative operations depend on it to be executed automatically at fixed
intervals— for example, backups, data consistency checks, rebuilding indexes, importing and
exporting data, replication, and so on.
• Microsoft Search— Provides full-text search capabilities to SQL Server, as well as to Microsoft
Exchange and Index Server.
• Microsoft SQL Server OLAP Service— Provides back-end support for Analysis Services.
• Microsoft Distributed Transaction Coordinator (MS-DTC)— Provides transaction support in multiserver
and heterogeneous environments.
• Server Network libraries— SQL Server can listen to several network libraries at the same time, waiting
for queries to answer, and use any of these libraries to send results to the client. The selected
database network library must have a compatible server network library to work with. SQL Server
2000 currently supports the following network libraries: TCP/IP Sockets, Named Pipes, Multiprotocol,
NWLink IPX/SPX, VIA ServerNET II SAN, VIA GigaNet SAN, Banyan VINES, and AppleTalk ADSP.

5
Microsoft SQL Server 2000 Programming by Example

Caution

Make sure that both client and server use the same network library or they will not be able to
communicate.

Tip

You do not need all these services to work with SQL Server 2000. Select only the services you
really need and you will save server resources.

Microsoft SQL Server Service contains different components that collaborate to provide back-end data
services. The three main components are

• Open Data Services— This component receives client requests from the network library and passes
them on to SQL Server. When SQL Server terminates the query process, it sends the result set to
ODS to be transferred through the network to the client application.
• Relational Engine— This component is responsible for parsing, optimizing, executing queries, and
enforcing security.
• Storage Engine— This component manages physical storage operations, such as data storage,
allocation and deallocation of data pages, transaction logging and recovery, database backups and
restoring, locking, and so on.

Note

Developers can extend the functionality of SQL Server by writing their own libraries based on ODS.
This is the basis of the extended stored procedures you see in Chapter 8, "Implementing
Business Logic: Programming Stored Procedures."

Client Tools

The preceding section discussed the different server components. In the client side, we can identify the
following components:

• The client application— This application, developed using any programming language, provides user
interaction, prepares requests to the database server, and shows query results to the user in a user-
friendly way.
• Database library— This library is responsible for translating application requests into specific
statements that the database server can understand.
• Client network libraries— This is the component that talks to the server network library to send and
receive TDS packets through the network.

6
 Chapter 1. Relational Database Management Systems and SQL Server

Figure 1.4 shows the different database libraries you can use to connect a client application to SQL Server
2000:

Figure 1.4. Data access options to connect to SQL Server from a client application.

• Direct HTTP access from an HTML page or an Active Server Page (ASP)— In this case, you use the
SQL ISAPI extension through a virtual directory in Internet Information Server, supporting direct XPath
queries through HTTP and XML input/output.
• Native SQL Server . DB-Library access— This is not the recommended solution because most of the
new SQL Server 2000 functionality is not exposed through DB -Library.
• Access to the ODBC API— Using any programming language, although C or C++ is recommended.
• Indirect access to the ODBC library through Remote Data Objects (RDO)— This solution provides an
object-oriented library that encapsulates database access objects. RDO is maintained for backward
compatibility with existing applications, but Microsoft recommends ActiveX Data Objects instead.
• Direct access to the OLE DB Provider library using C++ or C# is recommended— This solution
provides full access to SQL Server 2000 through a comprehensive data-access object model with
specific properties available in the native SQL Server OLE DB provi der.
• Indirect access to the OLE DB provider for SQL Server through ActiveX Data Objects (ADO) or
ADO.Net— ADO exposes an object model, as OLE DB does, easier to implement than direct access
to the OLE DB provider, and is suitable to any programming and scripting language, including any
version of Visual Basic and ASP.Net.
• Indirect access to the SQL Server ODBC driver through the OLE DB provider with or without the ADO
library— This solution is not recommended because of the extra steps involved in the data access,
unless specific ODBC functionality is required.
• Access to database metadata through ADOX and OLE DB— This is an alternative way to connect to
SQL Server, to send Data Definition Language (DML) statements, and metadata discovery. This
connectivity solution is not represented in the diagram because it is not a very common solution.
• Administrative access to SQL Server through SQL-DMO (Distributed Management Objects)— This is
the object model library that SQL Server Enterprise Manager uses to connect to SQL Server.
Developers can use all this functionality to build small administration applications, as subsets of what
Enterprise Manager can do.
• Windows Management Instrumentation (WMI)— WMI is a scalable Windows 2000 component,
common to other server applications, which exposes an object model to control and administer SQL
Server, as well as other server services and devices .

Note

7
Microsoft SQL Server 2000 Programming by Example

WMI install is not part of the SQL Server setup. You can install WMI support for SQL Server 2000
from the folder x86\OTHER\wmi on the SQL Server 2000 compact disc.

Currently, WMI access to SQL Server is based on SQL-DMO, but future releases might implement
it in a different way.

Tip

If you want to create a new application to connect to SQL Server 2000, write your application using
ADO and the native OLE DB provider for SQL Server. This will help the compatibility with the new
http://Microsoft.net development framework.

You can search for extra information on ADONET (ADO+) at the .NET Microsoft site:
http://www.microsoft.net

SQL Server 2000 includes some client applications you can use to administer and develop databases:

• Enterprise Manager
• Query Analyzer
• Profiler
• Upgrade Wizard
• Service Manager
• Command-line utilities

Caution

If you install SQL Server 2000 in the same computer as SQL Server 7.0, the version 7.0 client
utilities will be replaced with the new ones. This will give you extra benefits, but you might be
surprised at first by the different user interface.

Enterprise Manager

You can use Enterprise Manager to manage any SQL Server 2000 instance, including the default SQL Server
2000 instance, running locally or remotely. You also can use Enterprise Manager to manage any local or
remote SQL Server 7.0 installation. However, this version of Enterprise Manager is not compatible with SQL
Server 6.5.
Figure 1.5 shows the Enterprise Manager environment, similar to the well-known Windows Explorer interface,
in which you can identify different sections for every server:

Figure 1.5. SQL Server Enterprise Manager.

8
 Chapter 1. Relational Database Management Systems and SQL Server

• The SQL Server Administration Tree— This panel uses the TreeView control to display the structure of
every registered server. It displays different icons for every database object and shows context menus
for every object, according to the methods that can be applied to each specific object.
• The menu bar— In this menu bar, you can find the Action menu, which is equivalent to the object
context menu available from the tree; the View menu to specify how to display information about the
selected object in the tree; and the Tools menu to show general commands you can use in Enterprise
Manager to administer SQL Server.
• The taskbar— This is a dynamic list of icons that represents processes you can run in the current
context, as well as navigation keys through the tree.
• The information panel— This panel shows information in different formats, depending on the selected
object. Figure 1.5 shows the Taskpad, which, in this case, is an HTML page with information about
the Northwind database.

For every server, the Enterprise Manager tree shows the following sections (folders):

• Databases— This section includes the list of available databases in the connected server.
• Data Transformation Services— This section gives you access to the DTS Designer and the
Import/Export Wizard.
• Management— Enter this section to see the current activity of the connected server; to access SQL
Server Agent objects, alerts, jobs, and operators; to manage backups and database maintenance
plans; and to look at the SQL Server event log.
• Replication— This is where you can administer publications and subscriptions, if this server publishes
or subscribes to any database.
• Replication Monitor— This section is available only if Replication is installed on this server. In that
case, you can use this section to monitor and administer replication agents.
• Security— This section gives you access to the administration of SQL Server logins, server roles,
linked servers, and remote servers. Later in this chapter, in the "Security Model" section, you will
learn about SQL Server security and Chapter 15, "Working with Heterogeneous Environments:
Setting Up Linked Servers," covers linked and remote servers.
• Support Services— Access this section to administer other services, such as Distributed Transaction
Coordinator, Full-Text Search, and SQL Mail.
• Meta Data Services— This section gives you access to the Microsoft Repository.

Caution

9
Microsoft SQL Server 2000 Programming by Example

Do not confuse SQL Mail with SQLServerAgent Mail.

SQL Mail is a service that allows SQL Server users to use the mail-extended stored procedures, to
send and receive messages from Transact-SQL scripts, stored procedures, and triggers. SQL Mail
uses the MAPI profile defined for the MSSQLServer service account.

SQLServerAgent Mail is the feature that allows SQLServerAgent to send messages to operators by
email to notify job success, failure, or completion, and alerts notifications. SQLServerAgent Mail
uses the MAPI profile defined for the SQLServerAgent service account.

In many servers, both services use the same service account, so they use the same MAPI profile.
However, they are different services and they use email for different purposes and in different
circumstances.

SQL Server Enterprise Manager provides wizards to perform most of the standard administrative activities.
These wizards are available from the Tools menu and from the Taskpad, at server and database level. Figure
1.6 shows the Wizards list from the Taskpad. To access the Taskpad, select a server in the Enterprise
Manager tree, and in the View menu, select Taskpad. To show the wizards, click on the Wizard tab in the
Taskpad.

Figure 1.6. Access to the Wizards list from the Taskpad.

From SQL Server Enterprise Manager, you can design a database in a similar way to Visual Database Tools
(from Visual Studio Enterprise Edition). Figure 1.7 shows the Northwind database diagram. You can create
this diagram by opening the list of databases and opening the subtree for the Northwind database. There you
can right -click on Diagrams and select New Database Diagram. This menu will open the Create Database
Diagram Wizard that will lead you step-by-step through the creation of this diagram.

10
 Chapter 1. Relational Database Management Systems and SQL Server

Figure 1.7. The database diagram tool from Enterprise Manager.

Tip

If you run SQL Server in a network, you can install the SQL Server client tools, including Enterprise
Manager, in any workstation, without any server component. Then, you can register the servers
you have to administer in Enterprise Manager. This provides centralized administration, without
producing any overhead in the server, because the client administration tools runs in the client
computer, not in the server.

Query Analyzer

Query Analyzer is a client tool designed to send queries to SQL Server and to display results. This is not an
end-user application; instead, it is a developer tool used to manage databases and create database
applications through the use of Transact-SQL scripts.
Query Analyzer is the tool you will use throughout this book to practice the examples. Figure 1.8 shows
Query Analyzer after retrieving a query result set from SQL Server.

Figure 1.8. SQL Server Query Analyzer.

11
Microsoft SQL Server 2000 Programming by Example

Appendix B gives you a comprehensive coverage of Query Analyzer.

Profiler

SQL Server Profiler is a client tool that captures SQL Server activity and sends this activity to a file, database
table, or the screen, giving you a powerful analysis tool. For every trace you can

• Select which events to trace.

• Select which information to show for every event.
• Select how to group the events.
• Apply filters to include or exclude specific values, such as applications, logins, databases, hosts, and
so on.
• Save the trace in a trace file or a database table.

Using Profiler, you can

• Monitor real-time activity in SQL Server.

• Detect long-running queries.
• Trace locks and deadlocks.
• Summarize activity per database, user, host, and so on.
• Select which database objects are more heavily used to prioritize optimization decisions.
• Detect actual SQL Server activity from applications in which the source code is not available. You can
create a trace to look at what Enterprise Manager does when you administer SQL Server using
graphical user interface commands.
• Monitor database autogrowth or autoshrink.
• Perform security audits.

To start a trace, follow these instructions:

1. Open Profiler from the Microsoft SQL Server program group.

2. Choose File, New, Trace.
3. Select the server to monitor, and connect to the server.
4. The Trace properties form appears. There you can give a name to the trace and select the
SQLProfilerTSQL_Replay template from the drop-down list.

12
 Chapter 1. Relational Database Management Systems and SQL Server

5. Click Save to File and select a location for the trace file (using the standard .trc extension).
6. Leave the default values for the other fields and click Run.
7. Profiler will show you an empty window with several columns and only one row.
8. In the Tools menu, click Query Analyzer to open Query Analyzer.
9. In Query Analyzer, connect to the same server as in step 3.
10. In the query window type SELECT @@VERSION and press F5 to execute the query.
11. Go back to Profiler, and you will see many rows where before it was a single row trace. Scroll down
through the trace to the last row and you should see something similar to Figure 1.9.

Figure 1.9. SQL Server Profiler.

12. Stop the trace by choosing File, Stop Trace menu.

13. The instructions traced on Profiler include all the events selected by the trace template. Choose File,
Properties, and then click the Events tab to see the selected events.
14. Look at the Data Columns and Filters tabs in the Trace Properties window to see how this trace
template is defined.
15. Exit Profiler .

The Upgrade Wizard

The SQL Server Upgrade Wizard converts SQL Server 6.5 databases to the SQL Server 2000 format. You
can upgrade the entire server or selected databases. The upgrade process will transfer and convert the
database catalog, most of the server and database settings, and user data.
Note

After the wizard completes the upgrade, SQL Server 6.5 is still available. If you want to remove
SQL Server 6.5, you must uninstall it.

13
Microsoft SQL Server 2000 Programming by Example

Using the SQL Server Upgrade Wizard, you can upgrade a SQL Server 6.5 database to the default SQL
Server 2000 instance running in the same computer. In this case, you can use a tape device to avoid space
problems in the hard disk. However, it is more efficient to upgrade from one computer to another in the same
network.

Caution

To run the Upgrade Wizard, you must have already installed a default instance of SQL Server 2000
in your import server. If the default instance in the target computer is SQL Server 7.0, the Upgrade
Wizard available will be the one installed with SQL Server 7.0.

You can find the SQL Server Upgrade Wizard in the Microsoft SQL Server— Switch programs group. Figure
1.10 shows the main form of this wizard, right after the Wizard Welcome form.

Figure 1.10. The SQL Server Upgrade Wizard.

Note

The Tape option will be available only if you have a tape physically attached to the computer that is
running SQL Server.

Service Manager

14
 Chapter 1. Relational Database Management Systems and SQL Server

Using Service Manager, you can start, pause, and stop SQL Server services on any available SQL Server in
your network. Figure 1.11 shows the SQL Server Service Manager. There you can specify to autostart a
service whenever the operating system starts.

Figure 1.11. SQL Server Service Manager.

When you stop, the SQL Server Service, SQL Server

1. Disables new logins, excluding system administrators.

2. Performs a CHECKPOINT in every database to shorten recovery time the next time SQL Server starts.
Checkpoint is an internal process in SQL Server that ensures that every data modified in memory is
sent to disk.
3. Waits for all active statements and stored procedures to finish their work.
4. Shuts down.

Caution

Note that batches can be interrupted when you stop SQL Server, and if the batch was inside a
transaction, the transaction is automatically rolled back.

Chapter 13, "Maintaining Data Consistency: Transactions and Locks," teaches you how to
use transactions in SQL Server.

When you pause SQL Server, you only prevent new connections, but existing users can continue their work.
This gives you the opportunity to send a message to the connected users, so they can finish their work before
stopping SQL Server.

Command-Line Utilities

SQL Server 2000 setup installs several utilities that can be started from the command prompt. To use them,
you must open a command prompt window and type any of the following commands:

15
Microsoft SQL Server 2000 Programming by Example

• bcp—Bulk Copy Program. Use this utility to import or export data to and from SQL Server 2000.
Chapter 14, "Transferring Data to and from SQL Server," contains information on how to use
bcp.
• console—Displays backup and restore messages when the operation uses a tape device.
• dtsrun—This utility runs Data Transformation Packages from the command prompt.
• dtswiz—Use this utility to start the DTS Import/Export Wizard.
• isql—This is a query tool that uses DB -Library to connect to SQL Server. Use this tool to execute
scripts in SQL Server 2000 that do not require user interaction, such as administrative scripts. You
can send the output to a file.
• isqlw—Use this command to run SQL Query Analyzer.
• osql—This is a similar tool to isql, but it uses ODBC to connect to SQL Server.
• itwiz—Runs the Index Tuning Wizard, which will advise you about the best strategy to tune your
database. Chapter 6, "Optimizing Access to Data: Indexes," teaches you how to use the Index
Tuning Wizard.
• makepipe—This utility creates a pipe that helps you test the Named Pipes protocol with the
readpipe utility.
• readpipe—This utility reads from a pipe created using the makepipe utility.
• odbccmpt—Enables or disables the compatibility flag for ODBC applications, which solves some
compatibility problems related to the ODBC 3.7 drivers.
• odbcping—Tests connectivity to an ODBC data source.
• rebuildm—Rebuild Master utility. This utility rebuilds all the system databases.
• distrib—Configures and runs the Replication Distribution Agent.
• logread—Configures and runs the Replication Log Reader Agent.
• replmerg—Configures and runs the Replication Merge Agent.
• queueread—Configures and runs the Replication Queue Reader Agent.
• snapshot—Configures and runs the Replication Snapshot Agent.
• scm—Service Control Manager. This utility is the command-line version of the SQL Server Service
Manager, with extra functionality.
• sqlagent—Starts the SQLServerAgent service.
• sqldiag—Produces a full diagnostics report about SQL Server current environment and activity.
• sqlftwiz—Starts the Full-text Indexing Wizard.
• sqlmaint—Runs specific database maintenance tasks.
• sqlservr—Starts, stops, or pauses any instanceof SQL Server 2000.
• vswitch—Switches the default instance of SQL Server between SQL Server 6.5 and SQL Server
2000.

Caution

Setup does not install the makepipe, readpipe, or odbcping utilities. You can find them in the
x86\Binn directory from the distribution CD.

Database Components (Objects)

A SQL Server 2000 database contains different types of objects. Some objects contain user data, whereas
other objects are just definitions of objects, business rules declarations, and programs.
Data is arranged in tables and every field, identified by its name and data type, represents a different attribute.
Tables are the main database objects because you store your data in tables. You will learn how to create
tables in Chapter 3, "Working with Tables and Views."
Every individual value uses a specific data type. SQL Server provides a collection of data types, compatible
with the ANSI SQL 92 standard, but you can create your own user-defined data types, based on existing

16
 Chapter 1. Relational Database Management Systems and SQL Server

system supplied data types. Chapter 2, "Elements of Transact-SQL," teaches you how to use data types
and how to define and apply user-defined data types.
To guarantee data integrity and consistency, you can define constraints in the following manner:

• Primary key and unique constraints provide entity integrity, maintaining uniqueness in one or more
columns.
• Check and default constraints maintain domain integrity, checking for specific business rules to apply
to the inserted data.
• Foreign keys maintain referential integrity, maintaining links between related information in different
tables.

Chapter 7, "Enforcing Data Integrity," covers constraints in detail, as well as Rule and Default objects.
Complex queries can be defined as views, which can be reused in other queries, providing better readability
and easier maintenance. You learn about views in Chapter 3, "Working with Tables and Views."
To speed up access to data, you can create indexes on tables and views. Indexes store subsets of the
available information in an ordered way, as keys and pointers to the actual data, to provide fast access to the
data. Chapter 6, "Optimizing Access to Data: Indexes," discusses indexes in detail.
You can expand the SQL Server capabilities creating user-defined functions. These functions can be as
simple as a scalar function or as complex as a multistatement table-valued user-defined function. To know
more about user-defined functions, read Chapter 10, "Enhancing Business Logic: User-Defined
Functions (UDF)."
Complex processes can be defined as stored procedures. In this way, SQL Server can optimize the stored
procedure's execution plan on the first execution, and reuse this optimized execution plan for every further call.
Chapter 8, "Implementing Business Logic: Programming Stored Procedures," teaches you how to
create and use stored procedures.
You can define special stored procedures, called triggers, which, linked to a table, execute automatically
whenever you attempt any data modification to the table .

Security Model

SQL Server stores important data for your company, and you want to guarantee that every user can access
only the data she needs and with the appropriate access rights.
Ensuring a proper security policy in a SQL Server database is a task that starts in your IT environment. To
plan this security policy, you should find answers to the following questions:

• Who can access your company premises?

• Who can physically access your corporate servers?
• Who, and from where, can connect to your corporate network?
• Do you apply proper password policies in your network?
• Do you isolate sensitive servers in restricted networks?
• Do you follow adequate security auditing policies?

Your network is secure if you can identify and ensure

• What resources need shared access by nonadministrators

• Who can access shared resources
• Which users have access to a resource, from which places users can access this resource, and
during what period of time
• A password policy that prevents misuse of logins and passwords
• A proper audit policy to trace unauthorized access attempts to any resource, by tracing information
about failed access

In other words: To control access to important resources in your company, you need to identify the users who
access these resources, the date and time of each access, and the location from where each access is made.
SQL Server enforces security at different levels. Any data access, such as reading the unit price of a given
product, forces SQL Server to check data access security, following predefined steps, according to the SQL

17
Microsoft SQL Server 2000 Programming by Example

Server security model. Figure 1.12 shows how SQL Server checks for data access security, which is in
summary:

Figure 1.12. The SQL Server security model.

1. A user needs a valid login to gain access to SQL Server.

2. After a user has entered SQL Server, access to specific databases is controlled by the existence of a
valid user on the target database.
3. Users need specific permissions to execute specific statements at the database level.
4. Users need permissions per object and action.
5. Incomplete permissions to execute a given statement prevent the entire statement from being
executed.

According to the preceding comments, a login gives you access to SQL Server, and a user gives you access
to a specific database. You learn in more detail this important process in the following sections.
Note

Security management is an administrative task, usually out of the scope of database developers.
However, many common problems in database applications are related to security. As a database
developer, you will benefit from understanding the implications of database security. Having
security in mind helps you design a better database system that is more adequate to the business
requirements.

Authentication Modes

SQL Server 2000 is integrated with Windows, and it can use the authentication mode defined in your
Windows network. SQL Server 2000 can collaborate with Windows NT or Windows 2000 to authenticate this
user. In other cases, some users will access SQL Server from other networks, not members of any Windows
domain, yet you still need to provide them secure access to SQL Server. In these cases, SQL Server is the
only service responsible for user authentication.
SQL Server supports two authentication modes:

• Windows Authentication only, when only valid Windows users will have access to SQL Server.
• Mixed mode, when SQL Server accepts either Windows authentication or SQL Server authentication.

18
 Chapter 1. Relational Database Management Systems and SQL Server

To specify the authentication mode for your server, use Enterprise Manager, open the SQL Server properties
form, and select the required authentication mode in the Security tab, as you can see in Figure 1.13. After
you change the authentication mode, you must stop and restart SQL Server.

Figure 1.13. SQL Server authentication mode.

Windows Integrated Authentication

Usually, you have your SQL Server 2000 installed in a member server of a Windows domain. In this case,
every user needs a valid domain login and password to start a session on any computer member of this
domain.
Every domain user can be a member of any number of domain groups, and every domain group can belong to
one or more local groups on every specific server. In other words, one specific domain user can be
authenticated in a specific server by his or her direct or indirect membership in one or more local or domain
groups.
The permissions that any user has when trying to access any server resource, printer, shared folder or file, or
network application, is the combination of permissions applied to every group where this particular user has
membership; the only exception is no access, which cancels any possible permissions that this user has.
When a user tries to connect to SQL Server using Windows Integrated Authentication, it is not necessary to
supply the login name and password again. Windows has already checked this data and SQL Server does not
need to check it again. Windows informs SQL Server about the identity of this user and the windows
groups'membership.
SQL Server must check whether the user has a valid login defined on SQL Server for his or her own Windows
login, or for any windows group where this user has membership. If this search is successful, SQL Server
checks whether any of these valid SQL Server logins has denied access to SQL Server, in which case the

19
Microsoft SQL Server 2000 Programming by Example

user connection is rejected. If none of the valid logins for this Windows user has denied access to SQL Server,
the connection is established.
Windows authentication has an important advantage: You use Windows to control who can access your
network and how, why not use Windows to control who can access SQL Server and how? SQL is a networked
Windows application, after all.
Using this type of authentication, SQL Server doesn't store password information for Windows logins.
Caution

If you try to provide a login name and password to connect to SQL Server 2000 and this particular
server accepts only Windows Authentication, the connection will be rejected, even if the attempted
login was sa with the valid sa password.

Mixed (SQL and Windows) Authentication

In some environments, you can have users who are authenticated by Windows and users without Windows
credentials, or in other words, they don't have a Windows account. If this is the case, you should use Mixed
Authentication Mode. In this case, any user can con nect to SQL Server either by Windows Authentication or
by SQL Server Authentication.
Perhaps you want to have an extra security layer to access your databases and, even if Windows has
authenticated the user, you want to force the user to provide a valid login and password to connect to SQL
Server. In other cases, users access SQL Server from remote networks, perhaps from the Internet, and
Windows cannot authenticate them.
Caution

It is the user's responsibility to decide which credentials to use when trying to connect to SQL
Server. If the user selects Windows Authentication, the connection will be successful only if SQL
Server accepts his Windows credentials.

If the user selects SQL Server Authentication, the supplied login name and password must
correspond to a valid login name and password in SQL Server; otherwise, the connection will be
refused and SQL Server will not try to connect with the user's Windows credentials.

Connecting to SQL Server: Logins

To allow users access to SQL Server, you must create a login for them. When you install a new instance of
SQL Server, you have only the following logins:

• BUILTIN\Administrators—This is the login associated with the local Administrator group in the
local server where SQL Server is installed. Members of this group are considered SQL Server
administrators by default. You can remove this login.
• sa—This is the SQL Server system administrator login account used for SQL Server authentication.
This login cannot be removed, even if you select Windows Integrated Authentication only.
• YourDomain\SQLService—This is the login account for the SQL Server service account, if you
selected, as recommended, to use a domain account as a SQL Server service account (SQLService,
in this case).

Note

20
 Chapter 1. Relational Database Management Systems and SQL Server

The service account should be a member of the local Administrators account, and in that case it
already has a valid login as a member of the BUILTIN\Administrators group. However, it is
recommended that you maintain a separate login for this account, because it should not depend on
the existence of other logins.

You can add more logins using Enterprise Manager or using the sp_addlogin or sp_grantlogin system-
stored procedures, as in Listing 1.1.

To create new logins in SQL Server you have the following choices:

• Execute sp_addlogin to create a new login using SQL Server Authentication. In this case, you can
specify the password, default language, and default database for this login.
• Execute sp_grantlogin to grant access to SQL Server to an existing local user in the server that is
running SQL Server. In this case, the name of the login should have the format 'BUILTIN\User' .
• Execute sp_grantlogin to grant access to SQL Server to an existing local group in the server that
is running SQL Server. In this case, the name of the login should have the format
'BUILTIN\LocalGroup'. A typical example of this is the default 'BUILTIN\Administrators'
login created during setup.
• Execute sp_grantlogin to grant access to SQL Server to an existing domain user in a domain
trusted by the domain in which SQL Server is running. In this case, the name of the login should have
the format 'DomainName\User'. This is the case used by the service account login
'YourDomain\SQLService'.
• Execute sp_grantlogin to grant access to SQL Server to an existing domain global group in a
domain trusted by the domain where SQL Server is running. In this case, the name of the login should
have the format 'DomainName\GlobalGroup' .

Caution

Local users and groups are valid only in the computer in which they are created, so they cannot be
used to grant access to SQL Server in a different computer.

Note

To execute Listing 1.1, and the other examples in this chapter, you must log in to SQL Server
2000 from Query Analyzer using an administrator account.

Listing 1.1 Create Logins Using the sp_addlogin and sp_grantlogin System Stored Procedures

21
Microsoft SQL Server 2000 Programming by Example

-- Create a SQL Server login

-- using English as a default language
-- with Northwind as a default database

EXEC sp_addlogin
@loginame = 'Tim'
, @passwd = 'TimPassword'
, @defdb = 'Northwind'
, @deflanguage = 'English'

-- Create a SQL Server login

-- using Spanish as a default language
-- without password, and without default database

EXEC sp_addlogin
@loginame = 'Pedro'
, @deflanguage =Spanish'

-- Create a SQL Server login

-- for the local Guest Windows account

EXEC sp_grantlogin 'BUILTIN\Guest'

-- Create a SQL Server login

-- for the domain Guest account

EXEC sp_grantlogin 'YourDomain\Guest'

-- Create a SQL Server login

-- for the local group Users Windows account

EXEC sp_grantlogin 'BUILTIN\Users'

-- Create a SQL Server login

-- for the domain group Domain Users account

EXEC sp_grantlogin 'YourDomain\Domain Users'

New login created.

New login created.
Granted login access to 'BUILTIN\Guest'.
Granted login access to 'YourDomain\Guest'.
Granted login access to 'BUILTIN\Users'.
Granted login access to 'YourDomain\Domain Users'.

22
 Chapter 1. Relational Database Management Systems and SQL Server

Caution

If the server where SQL Server is installed is a domain controller, it does not have local users.
Therefore, the third example from Listing 1.1 will fail.

As you see in Listing 1.1, when you add a new SQL Server login, you can specify a password, a default
database, and default language for this particular login. When you add a Windows login using
sp_grantlogin, you cannot specify these options. However,

• The password cannot be specified because Windows will check the password when the user tries to
connect to Windows. SQL Server does not need to know this password.
• You can use the sp_defaultlanguage system stored procedure to mod ify the default language
that this particular login will use to communicate with SQL Server. This setting affects custom error
messages and date output formats. This procedure is valid for SQL Server and Windows logins.
• You can execute the sp_defaultdb system stored procedure to modify the default database for this
login. This procedure is valid for SQL Server and Windows logins.

Caution

Having a default database does not guarantee access to this database. To have access to a
database, you must be mapped to a valid user in that database.

The next section teaches you how to create database users.

To deny access to SQL Server to a particular login, you can use the following:

• EXECUTE sp_denylogin 'Domain\User' denies access to SQL Server to a domain user. The
login still exists, but nobody can connect to SQL Server using this login. If the login is a Windows
group, none of the members of this group will be able to connect to SQL Server— regardless of the
existence of other logins they might have access to.
• EXECUTE sp_revokelogin 'Domain\User' permanently removes this Windows login from SQL
Server. This does not guarantee that the users or members of this Windows group will not have
access to SQL Server, because they can still belong to one or more Windows groups with valid logins
in SQL Server.
• EXECUTE sp_droplogin 'SQLUser' denies access to SQL Server to a SQL Server login. In this
case, this login will be deleted from SQL Server permanently.

Note

You can change the password for an existing SQL Server login using the sp_password system
stored procedure.

23
Microsoft SQL Server 2000 Programming by Example

Using Databases: Users

After the user connects to SQL Server using a valid login, the connection is established, but there is not much
to do. To access real data, the user needs access to a database. This is achieved by creating a user on that
database.
When you set up SQL Server, the systems databases contain the following users:

• dbo—This is the database owner with full permissions, by default, on the entire database. This user
cannot be removed.
• guest—This is the default user for logins that don't have a specific user in a database. Because every
system database, Northwind, and Pubs databases have a guest user, any valid login can use these
databases directly. This user can be removed to guarantee authenticated access to a database only.

Caution

SQL Server uses the Model database as a template to create new databases. The Model database
does not have a Guest account; therefore, new databases will not have a guest user unless you
create it explicitly.

To create a user in the current database, you can use the sp_grantdbaccess stored procedure, as in
Listing 1.2. Each login can be mapped to a single user per database. Even if the login were related to a
Windows group, the mapped user in the database is considered a logically individual user.

Listing 1.2 Use sp_grantdbaccess to Grant Logins Access to a Database

USE Northwind
GO

-- Create a Tim user in the Northwind database

EXEC sp_grantdbaccess
@loginame = 'Tim'

-- Create a User in Northwind

-- for the local Guest login

EXEC sp_grantdbaccess
@loginame = 'BUILTIN\Guest'
, @name_in_db = 'LocalGuest'

-- Create a user in Northwind

-- for the domain Guest account

EXEC sp_grantdbaccess

24
 Chapter 1. Relational Database Management Systems and SQL Server

@loginame = 'YourDomain\Guest'
, @name_in_db = 'GlobalGuest'

-- Create a user in Northwind

-- for the local group Users Windows account

EXEC sp_grantdbaccess
@loginame = 'BUILTIN\Users'
, @name_in_db = 'LocalUsers'

-- Create a user in Northwind

-- for the domain group Domain Users account

EXEC sp_grantdbaccess
@loginame = 'YourDomain\Domain Users'
, @name_in_db = 'GlobalUsers'

Granted database access to 'Tim'.

Granted database access to 'BUILTIN\Guest'.
Granted database access to 'CallSQL\Guest'.
Granted database access to 'BUILTIN\Users'.
Granted database access to 'CallSQL\Domain Users'.
As you saw in Listing 1.2, users in a database do not have to have exactly the same names as the
corresponding logins in SQL Server. However, it is recommended that users and logins have the same names
to avoid maintenance problems .

Server and Database Roles

After a login has been granted access to a database, because it has a user defined on that database, it does
not have access to specific data yet. The user needs specific permissions to access data from database
objects.
Applying and managing to individual users can be very complex. SQL Server 2000 provides roles, at server
and database levels, to simplify permissions management:

• Fixed server roles— Group logins by general server permissions to simplify general logins
administration. These fixed server roles cannot be modified or dropped, not even their permissions.
You can only change their membership.
• Fixed database roles— Group users in a database by functionality, reducing permissions maintenance
overhead. Fixed database roles cannot be modified or dropped, not even their permissions. You can
only change their membership.
• User-defined database roles— Extend the functionality of fixed database roles, providing extra
flexibility on grouping users at the database level to apply specific permissions.

The most important fixed server role is sysadmin (System Administrators). Login members of the sysadmin
server role are not affected by SQL Server security at all.
Caution

Members of the sysadmin role have full permissions on every database object and every
database, and these permissions cannot be modified as long as the login is a member of the
sysadmin role. Therefore, select the members of the sysadmin role carefully.

25
Microsoft SQL Server 2000 Programming by Example

If you want to differentiate between users with only specific administrative privileges, you can use other fixed
server roles:

• Members of the serveradmin role can configure and stop or restart SQL server.
• Members of the setupadmin role can manage linked servers and specify which stored procedures
will automatically run when SQL Server starts.
• Members of the securityadmin role can manage logins, including changing passwords, and assign
permissions to create databases.
• Members of the processadmin role can stop processes running in SQL Server.
• Members of the dbcreator role can manage databases.
• Members of the diskadmin role can manage disk files and backup devices.
• Members of the bulkadmin role can execute BULK INSERT statements.

Note

Members of the sysadmin role have all the permissions listed for all other fixed server roles.

You can use system stored procedures to retrieve information about server roles and to manage server roles
membership, as in Listing 1.3.

Listing 1.3 Stored Procedures to Manage Server Roles

-- Get a list of the server roles

EXEC sp_helpsrvrole

-- get a list of permissions of a specific server role

EXEC sp_srvrolepermission setupadmin

-- Make Tim member of the sysadmin role

EXEC sp_addsrvrolemember
@loginame = 'Tim'
, @rolename = 'sysadmin'

-- Get a list of sysadmin role members

EXEC sp_helpsrvrolemember sysadmin

26
 Chapter 1. Relational Database Management Systems and SQL Server

-- Remove Tim from the sysadmin role

EXEC sp_dropsrvrolemember
@loginame = 'Tim'
, @rolename = 'sysadmin'

ServerRole Description
----------------------------------- -----------------------------------
sysadmin System Administrators
securityadmin Security Administrators
serveradmin Server Administrators
setupadmin Setup Administrators
processadmin Process Administrators
diskadmin Disk Administrators
dbcreator Database Creators
bulkadmin Bulk Insert Administrators

(8 row(s) affected)

ServerRole Permission
----------------------------------- -----------------------------------
setupadmin Add member to setupadmin
setupadmin Add/drop/configure linked servers
setupadmin Mark a stored procedure as startup

(3 row(s) affected)

'Tim'added to role 'sysadmin'.

ServerRole MemberName MemberSID

----------- ------------------------ -------------------------------------
sysadmin BUILTIN\Administrators 0x01020000000000052000000020020000
sysadmin YourDomain\SQLService 0x0105000000000005150000003FAD1462
FD43461E1525AF47EF030000
sysadmin distributor_admin 0xBAC6B1014B4F23408F6B0CEF54A0AB5E
sysadmin sa 0x01
sysadmin Tim 0x6E00C5CC4408ED47A33C5B210029109F

'Tim'dropped from role 'sysadmin'.

Note

MemberSID is the security ID, which could be generated by SQL Server for SQL Server logins or
the Windows Security ID directly.

27
Microsoft SQL Server 2000 Programming by Example

At database level, you can group users in database roles. Every database contains a fixed set of fixed
database roles:

• db_owner—Database owners or users who have, by default, granted permissions to perform any
action on every database object.
• db_accessadmin—Database access administrators manage users in the database.
• db_securityadmin—Database security administrators manage permissions.
• db_ddladmin—Database DDL administrators can execute any Data Definition Language statement.
• db_backupoperator—Database backup operators.
• db_datareader—Database users with permissions to read data from any table in the database.
• db_datawriter—Database users with permissions to modify data on any table in the database.
• db_denydatareader—Database users with denied permissions to read data from this database.
• db_denydatawriter—Database users with denied permissions to modify data from this database.

Caution

It is possible to be a member of the db_denydatareader and db_datawriter roles at the

same time. In this case, the user can modify data but cannot read it.

You can use system-stored procedures to retrieve information about fixed database roles, as in Listing 1.4.

Listing 1.4 Stored Procedures to Manage Fixed Database Roles

USE Northwind
GO

-- Get a list of the database roles

EXEC sp_helpdbfixedrole

-- Get a list of permissions of a specific database role

EXEC sp_dbfixedrolepermission db_ddladmin

-- Make Tim member of the db_owner role

EXEC sp_addrolemember
@rolename = 'db_owner'
, @membername = 'Tim'

-- Get a list of db_owner database role members

EXEC sp_helprolemember 'db_owner'

28
 Chapter 1. Relational Database Management Systems and SQL Server

-- Remove Tim from the db_owner role

EXEC sp_droprolemember
@rolename = 'db_owner'
, @membername = 'Tim'

DbFixedRole Description
--------------------- -----------------------------------
db_accessadmin DB Access Administrators
db_backupoperator DB Backup Operator
db_datareader DB Data Reader
db_datawriter DB Data Writer
db_ddladmin DB DDL Administrators
db_denydatareader DB Deny Data Reader
db_denydatawriter DB Deny Data Writer
db_owner DB Owners
db_securityadmin DB Security Administrators

(9 row(s) affected)

DbFixedRole Permission
--------------------- -----------------------------------
db_ddladmin All DDL but GRANT, REVOKE, DENY
db_ddladmin dbcc cleantable
db_ddladmin dbcc show_statistics
db_ddladmin dbcc showcontig
db_ddladmin REFERENCES permission on any table
db_ddladmin sp_changeobjectowner
db_ddladmin sp_fulltext_column
db_ddladmin sp_fulltext_table
db_ddladmin sp_recompile
db_ddladmin sp_rename
db_ddladmin sp_tableoption
db_ddladmin TRUNCATE TABLE
(12 row(s) affected)

'Tim'added to role 'db_owner'.

DbRole MemberName MemberSID

-------------- ------------ ------------------------------------
db_owner dbo 0x01
db_owner Tim 0x6E00C5CC4408ED47A33C5B210029109F

(2 row(s) affected)

'Tim'dropped from role 'db_owner'.

You can create your own database roles. This is useful in some specific circumstances:

• If you cannot create Windows groups to group your SQL Server users, you can create database roles
to group users according to the different set of permissions they must have.
• If you have Windows users and non-Windows users, you will have Windows logins and SQL Server
logins, and you want to apply permissions to them as a group, regardless of their login origin.

29
Microsoft SQL Server 2000 Programming by Example

Tip

Try to use Windows groups instead of user-defined database roles whenever possible. Windows
users are grouped by functionality, and they share common permissions to access different
resources at Windows level, printers, directories, and so on. It usually makes sense to maintain
them as a group in SQL Server, creating a single login for all of them and a single user on the
database they need. In this case, there is no need for a database role.

You can manage database roles with several stored procedures similar to the ones used in Listings 1.3 and
1.4. Listing 1.5 shows how to create a database role, how to manage membership of the role, and how to
remove the database role when it is no longer necessary .

Listing 1.5 Stored Proceduresto Manage User-Defined Database Roles

USE Northwind
GO

-- Add a new role

EXEC sp_addrole 'SalesAdmin'

-- Get a list of the database roles

EXEC sp_helprole

-- Make the SqlesAdmin role member of the

-- db_datareader database role

EXEC sp_addrolemember
@rolename = 'db_datareader'
, @membername = 'SalesAdmin'

-- Make Tim member of the SalesAdmin role

EXEC sp_addrolemember
@rolename = 'SalesAdmin'
, @membername = 'Tim'

-- Get a list of db_owner database role members

EXEC sp_helprolemember 'SalesAdmin'

-- Remove Tim from the db_owner role

EXEC sp_droprolemember
@rolename = 'SalesAdmin'
, @membername = 'Tim'

30
 Chapter 1. Relational Database Management Systems and SQL Server

-- Drop the SalesAdmin role

EXEC sp_droprole 'SalesAdmin'

New role added.

RoleName RoleId IsAppRole

------------------ ------ -----------
public 0 0
db_owner 16384 0
db_accessadmin 16385 0
db_securityadmin 16386 0
db_ddladmin 16387 0
db_backupoperator 16389 0
db_datareader 16390 0
db_datawriter 16391 0
db_denydatareader 16392 0
db_denydatawriter 16393 0
SalesAdmin 16400 0

(11 row(s) affected)

'SalesAdmin'added to role 'db_datareader'.

'Tim'added to role 'SalesAdmin'.

DbRole MemberName MemberSID

-------------- ------------ ------------------------------------
db_owner Tim 0x6E00C5CC4408ED47A33C5B210029109F

(1 row(s) affected)

'Tim'dropped from role 'SalesAdmin'.

Role dropped.
There is a special group on every SQL Server database; its name is public and every user and role in that
database belongs to public. The public role is very useful to define the default permissions that every user
has in the database .

Permissions

Now that you have users in a database and you grouped them by database roles, you can apply permissions
to them. These permissions can be

• Statement permissions— Permissions to execute specific Transact-SQL statements, to create

database objects, and to execute backups.
• Data access permissions— Permissions to read, delete, insert, update, or reference data, or
permissions to execute stored procedures and user-defined functions.

Regardless of the type of permission to apply, the user permission to perform any action on a specific
database object or statement can have any of these three states:

31
Microsoft SQL Server 2000 Programming by Example

• Neutral— No permissions. In this case, the user will have permissions to perform the required action,
depending on permissions applied to other roles or Windows groups the user might belong to.
• Denied— The user cannot perform the required action. It does not matter whether the user belongs to
other roles or groups with permissions to perform this action. Deny overrides everything else.
• Granted— In principle, the user has permissions to proceed with this action. However, the final
permissions depend on other existing permissions on the same action applied to other groups or roles
this par ticular user belongs to.

To decide which permissions a particular user has, SQL Server must combine permissions from the user with
permissions from the different roles and groups in which this user has membership. The only exception to this
is if any permission is denied to the user or any of the user's roles or groups, the final permission will be
denied .
Note

Remember, members of the sysadmin role are not affected by permissions.

Statement Permissions

You can define permissions for the following statements:

• BACKUP DATABASE—To execute full or differential database backups

• BACKUP LOG—To perform transaction log backups
• CREATE DATABASE—To create databases
• CREATE DEFAULT—To create independent DEFAULT objects
• CREATE FUNCTION—To create user-defined functions
• CREATE PROCEDURE—To create stored procedures
• CREATE RULE—To create independent RULE objects
• CREATE TABLE—To create tables
• CREATE VIEW—To create views

To grant statement permissions, you use the GRANT statement with the following syntax:

GRANT Statement TO Security_account

or

GRANT ALL TO Security_account

Security_account can be a database user or a user-defined database role.

To deny statement permissions, use the DENY statement with the following syntax:

DENY Statement TO Security_account

or

DENY ALL TO Security_account

To revoke previously granted or denied statement permission, use the REVOKE statement with the following
syntax:

32
 Chapter 1. Relational Database Management Systems and SQL Server

REVOKE Statement FROM Security_account

or

REVOKE ALL FROM Security_account

You can manage statement permissions in Enterprise Manager as well, in the Permissions tab available in the
Database Properties form, as shown in Figure 1.14.

Figure 1.14. Manage statement permissions from Enterprise Manager.

Caution

By default, when a user creates an object, the user becomes the owner of that object. However,
the dbo, or any member of the db_owner role, can take ownership of any existing object in the
database they own.

Tip

33
Microsoft SQL Server 2000 Programming by Example

It is not recommended to grant statement permissions to any user other than dbo. In this way, the
same user, dbo, will own every object and permissions management will be much easier.

It is not necessary to grant statement permissions to dbo because it has all these permissions by
default.

Note

DROP and ALTER statement permissions are granted only to the object owner and members of the
db_owner role.

RESTORE DATABASE and RESTORE LOG permissions are granted to members of the sysadmin
and dbcreator server roles. Note that members of the db_owner role do not have permissions to
restore the database they own.

Data Access Permissions

You can protect access to the data using the GRANT, DENY, and REVOKE statements you saw already in the
preceding section.
Depending on the object, you can manage permissions for different actions. Figure 1.15 shows the
permissions that can be applied to every different object .

Figure 1.15. Data access permissions available per database object.

34
 Chapter 1. Relational Database Management Systems and SQL Server

You can access object permissions in Enterprise Manager in three ways:

• Through the Object Permissions form, you can see the permissions applied to every user and role for
a particular object. Figure 1.16 shows you the Object Permissions form.

Figure 1.16. Data access permissions through the Object Permissions form.

35
Microsoft SQL Server 2000 Programming by Example

• Through the Database User Permissions form, where for every user you can see the permissions
applied on every object in the database. Figure 1.17 shows the User Permissions form.

Figure 1.17. Data access permissions through the Database User Permissions form.

36
 Chapter 1. Relational Database Management Systems and SQL Server

• Through the Database Role Permissions form, where for every user you can see the permissions
applied on every object in the database, as in Figure 1.18.

Figure 1.18. Data access permissions through the Database Role Permissions Form.

Some users have special data access permissions:

• Members of the sysadmin server role have full permissions on all objects and all databases available
in the server they manage. Even if you specifically deny permissions to these users, they can still
perform any action because SQL Server security does not affect SQL Server administrators.
• Members of the db_owner database role have, by default, full permissions on all objects from the
database they own. However, it is possible to deny permissions on specific actions and objects to
members of the db_owner role. In that case, DENY affects them, as with any other database user.
However, members of the db_owner role can revoke previously denied permissions and take
ownership of any object if necessary.
• Object owners have, by default, full permissions on the objects they own. However, the object owner,
members of the sysadmin server role, or members of the db_owner database role can deny
permissions on specific actions to the object owner. In that case, DENY affects the object owner, as
with any other database user. However, object owners can revoke previously denied permissions as
long as they are still owners of the target object.

Note

An object owner is not necessarily the database owner, and vice versa. However, it is
recommended that all objects in a database be owned by the user dbo.

37
Microsoft SQL Server 2000 Programming by Example

Application Security

In the preceding sections, you learned how to manage permissions to execute specific Data Manipulation
Language (DML) or Data Definition Language (DDL) statements, affecting database users and roles.
Managing permissions in this way can be complex as soon as the number of database objects and users
increases. Using SQL Server 2000, you can simplify permissions management using Application Security in
different ways:

• Creating stored procedures to retrieve and modify data— You can give users access to the data
through these stored procedures only, denying permissions to access directly the underlying tables.
Chapter 8 covers in detail how to implement applications security using stored procedures.
• Provide indirect access to the data through views— You can design your views to provide access to
particular columns only and to particular sets of rows, hiding the actual tables to the users. Chapter 3,
"Working with Tables and Views," covers how to implement applications security using views in
detail.
• Provide indirect access to the data through inline user-defined functions— Inline user-defined
functions are similar to views, but they expose parameters that give your users extra flexibility. You
learn in Chapter 10, "Enhancing Business Logic: User- Defined Functions (UDF)," how to
implement applications security using inline user-defined functions.
• Provide indirect read-only access to the data through multistatement table-valued user-defined
functions— These functions work similar to stored procedures, but they do not modify actual data.
Users can work with these user-defined functions as if they were tables. You can find examples of
table-valued functions in Chapter 10.

When a user invokes a stored procedure, view, or user-defined function, SQL Server checks first whether
every required object belongs to the same owner. In this case, SQL server will check permissions only in the
invoked object, and it does not check permissions on the underlying objects at all. However, if the ownership
chain is broken on any of the referenced objects, SQL Server must check permissions on every individual
object.
Chapters 3, 8, and 10 contain information and examples about implications of ownership chains on
application security.
Imagine now a different situation:
Your company has a sales application, called SalesInfo, to manage sales orders. Customers call your
company to place orders and you have a dedicated team of sales specialists who take orders on the phone
and enter them into your database using the SalesInfo application.
You have a call center with 100 computers and you have in total about 200 sales specialists who use these
computers in different shifts. Sales specialists sit every day at a different computer, and you give them an
individual domain account to log in to your network. Other employees can use the SalesInfo application
from their desktops or mobile computers.
Different employees might have different permissions in SQL Server, through different Windows groups and
database roles. However, the SalesInfo application must execute specific actions in the database,
regardless of the employee who is using the application. You want to make sure that if an employee has
access to this application, he or she will be able to use it at full functionality, regardless of the individual
database permissions the user might have.
To solve this problem, you could create a database role and assign permissions to this new role. However,
permissions on this database role will be merged with other permissions the user might have, resulting in
different actual permissions per employee.
SQL Server 2000 provides a solution to this problem through application roles. An application role is a special
type of database role that has no members and whose permissions are not combined with any user's
permissions.
You can create an application role using the sp_addapprole system stored procedure, providing a name for
the new application role and a password for its activation. To change the application role password, you can
execute the sp_setapprolepassword system stored procedure.
Applications can activate an application role using the sp_setapprole system stored procedure. User
permissions are disregarded when an application role is active.
Caution

38
 Chapter 1. Relational Database Management Systems and SQL Server

As we mentioned earlier in this chapter, every user and role belongs to the public role, including
application roles. After activating an application role, the only active permissions are the
permissions applied to the application role and to any group to which the application role belongs.

To drop an application role, you can use the sp_dropapprole system stored procedure. Listing 1.6
contains an example of how to create an application role, change its password, and activate the application
role. It also shows an example of how to drop the application role.

Listing 1.6 Application Role Management

USE Northwind
GO

-- Create an Application Role

EXEC sp_addapprole
@rolename = 'SalesInfo'
,@password = 'SIPassword'
GO

-- Change the Application Role password

EXEC sp_approlepassword
@rolename = 'SalesInfo'
, @newpwd = 'NewSIPassword'
GO

-- Drop the Application Role

-- Note: execute this code alone
-- to drop the SalesInfo application role

/*
EXEC sp_dropapprole 'SalesInfo'

*/

-- Activate the Appliction role

-- with encryption

EXEC sp_setapprole
@rolename = 'SalesInfo'
, @password = { Encrypt N 'NewSIPassword'}
, @encrypt = 'ODBC'

-- Activate the Appliction role

-- without encryption
-- Note: execute this conde instead of

39
Microsoft SQL Server 2000 Programming by Example

-- the preceding statement

-- if you do not want to use encryption

/*

EXEC sp_setapprole
@rolename = 'SalesInfo'
, @password = 'NewSIPassword'

*/

New application role added.

(1 row(s) affected)

The password for application role 'SalesInfo'has been changed.

The application role 'SalesInfo'is now active.

Applications can activate an application role at any time, providing the application role name and the specific
application role password. The general security process for the SalesInfo application could be as follows:

1. Every employee uses a personal windows login to connect to the network.

2. Create a Windows group called SalesInfoGroup to group all users of the SalesInfo application.
3. Create a login in SQL Server for the SalesInfoGroup Windows group. Provide the Northwind
database (or your particular target database) as a default database for this login.
4. Create the SalesInfoGroup user in the target database.
5. Create the SalesInfo application role in the target database as described in Listing 1.6.
6. The SalesInfo application can be designed to connect to SQL Server using Windows Integrated
security mode.
7. Every time the SalesInfo application connects to SQL Server, the application must call the stored
procedure sp_setapprole to activate the application role.
8. When the application role is active, user permissions are disregarded.

Caution

Application roles are defined in a single database and they can access this database only. If an
application role tries to access another database, it will use permissions available to the Guest
account, if the Guest account exists in the target database.

Tip

If you use application roles and you want to know which Windows user is actually connected, use
the SYSTEM_USER function. Even after an application role is active, SYSTEM_USER returns the
actual login.

40
 Chapter 1. Relational Database Management Systems and SQL Server

Client/Server Applications Design

You can design client/server applications to connect to SQL Server in many different ways. If you think about
the basic activities every application does, you can group these activities in three conceptual layers:

• Data— This layer deals with the actual data your application uses. Some data can be stored in a
relational database; some other data can be stored somewhere else, on any other format.
• Presentation— This layer is responsible for user interaction, accepting input from the user and
presenting results to the user.
• Business— This layer interacts with the data layer to access data, and with the presentation layer to
receive user requests and send results. This layer enforces business rules and executes business
processes.

You can design your application as a monolithic standalone client-side application, including these three
layers. This was the way many programs were created some years ago. If you deploy your application to
several users, every user works with a private set of data. If you need to apply any change to the
programming logic, you must rebuild and redeploy the application.
Trying to share data from different users forces you to separate the data layer from the other two and store
the data in a separate database server. You can still design your application to manage presentation and
business layers. In this case, changes on the business rules will force you to rebuild and redeploy the
application to every user.
You can use constraints, views, stored procedures, user-defined functions, and triggers to translate part of the
business layer to SQL Server. In this way, the database can be somehow a self-contained application with
data, and business rules and processes. This approach can give you great flexibility, because changes on any
business rules can be made in a central location: the SQL Server database.
In the previous examples, changes to the presentation layer force you to rebuild and redistribute your
application. However, you can base your presentation layer in Active Server Pages (ASP), running in Internet
Infor mation Server (IIS). In this case, the only thing that the users need is an Internet browser. Changes to
the presentation layer can be done easily in IIS, without any manual redistribution to the final users.
Using COM+ in Windows 2000 or Microsoft Transaction Server (MTS) in Windows NT 4.0, you can create
components that encapsulate some functionality from the business and data layers. The presentation layer
can use business components from COM+, and these business components can either use data components
in COM+ or use database objects directly in SQL Server.
Designing your application in this way, you can change your database system to a different database
structure or to a different database management system, if you need to, and the only changes to apply will be
in your COM+ components. The presentation layer will be unaware of the changes in the database system.
Figure 1.19 shows a typical schema of a multilayer, or multitier, application.

Figure 1.19. A multitier system.

41
Microsoft SQL Server 2000 Programming by Example

In Figure 1.19, you can identify the following components:

A. Data is stored in a SQL Server 2000 database and Exchange 2000 Server folders.
B. Data components running in COM+ encapsulate data access to other components and applications.
C. Business components expose complex data and business operations to other components.
D. Network applications provide user interaction to local users, based on business components from
COM+.
E. Active Server Pages, running on Internet Information Server, provide user interaction to local and
remote users, based on functionality provided by business components from COM+.
F. Local and remote users connect to IIS with an Internet browser to display the results of the Active
Server Pages.

What's Next?

This book is about Microsoft SQL Server 2000. In some cases, you will see some comments about previous
versions, but we wrote this book with SQL Server 2000 and for SQL Server 2000 database developers. This
book provides comprehensive information about the new features available in SQL Server 2000:

• In Appendix A, you can learn how to install multiple instances of SQL Server 2000, and Chapter 15
teaches you how to write distributed queries, spanning multiple servers and instances.
• Appendix B covers the increased functionality of the new Query Analyzer.
• You will find information about new datatypes and built-in functions in Chapters 2 and 3.
• Chapter 6 is about optimizing data access using indexes. It covers new features, such as indexed
views and indexes on computed columns.
• Cascade Declarative Referential Integrity is covered in Chapter 7.
• In Chapter 8, you will learn how to use the new Transact-SQL Debugger to test the execution of
stored procedures.
• Information on and examples of new trigger functionality, such as index order and INSTEAD OF
triggers, are available in Chapter 9.
• Chapter 10 provides comprehensive coverage of user-defined functions, which is one of the most
exciting features of this new SQL Server version.
• Chapter 14 teaches you how to use the new Copy Database Wizard and how to create your own
DTS packages.

42
 Chapter 2. Elements of Transact-SQL

Chapter 2. Elements of Transact-SQL

SQL, or Structured Query Language, is the language used to access and interact with a relational database. It
was created by IBM in the 1970s and first standardized by the American National Standards Institute (ANSI)
in 1989. ANSI released a new version of the standard in 1992, known as SQL-92 or SQL2.
Although all relational management systems, including SQL Server, are ANSI compliant at different levels,
they provide extensions to the SQL language that extend the language's functionality. However, be aware that
the code becomes less portable by using these extensions. Therefore, it's up to the programmer to comply or
not comply with the ANSI standard.
Transact-SQL is the SQL Server implementation of the ANSI SQL-92 standard. In addition to the ANSI SQL-
92 basic elements, Transact-SQL has extensions or enhancements that improve the capabilities of the
language. For example, Transact-SQL adds procedural control-of-flow elements, such as IF .. ELSE,
WHILE, BREAK, and CONTINUE.
Unlike procedural languages, Transact-SQL is a set-oriented database language (set-oriented means that it
processes groups of data at once). As such, it was designed to work efficiently with set operations, instead of
row-by-row operations. Thus, using Transact-SQL, you specify what you want to do with the whole set of data,
instead of indicating what to do with each piece of data, or in database terminology, each row.
This chapter teaches you the following elements of Transact-SQL:

• The Data Definition Language (DDL)

• The Data Manipulation Language (DML)
• The Data Control Language (DCL)
• Extensions of Transact-SQL, such as variables, operators, functions, control of flow statements, and
comments

Data Definition Language (DDL)

The Data DefinitionLanguage is used to create and manage databases and database objects, such as tables,
stored procedures, user-defined functions, triggers, views, defaults, indexes, rules, and statistics. Transact-
SQL provides a CREATEand a DROP statement for all these elements. Listing 2.1 shows the creation and
removal of a table in the Northwind database.

Listing 2.1 Creating a Table Using the CREATE TABLE Statement

USE Northwind

CREATE TABLE Employeedependents

(
dependentid INT IDENTITY(1,1),
lastname VARCHAR(20),
firstname VARCHAR(20),
)
GO

43
Microsoft SQL Server 2000 Programming by Example

The command(s) completed successfully.

In addition to CREATE and DROP statements, an ALTER statement is provided to modify the properties of some
of these objects (databases, tables, stored procedures, user-defined functions, triggers, and views). Listing
2.2 demonstrates how to add a new column to a table, using the ALTER TABLE statement.

Listing 2.2 Adding a New Column to the Employeedependent Table Using the ALTER TABLE
Statement

USE Northwind

ALTER TABLE Employeedependents ADD birthdate DATETIME

GO

The command(s) completed successfully.

In SQL Server, objects must be unique within users. This allows two users to own an object with the same
name. For example, consider two users in a certain database. Each one of these two users could own a table
with the same name. Therefore, in this case, there would be two tables with the same name in the database.
In the following example, users user1 and user2 successfully create a table with the same name (Tablex) in
the Northwind database.

1. Using Query Analyzer, connect to SQL Server with the sa login and execute the following code, which
creates both logins (login1 and login2 with blank password), adds users (user1 and user2) to
the Northwind database for these logins, and grants CREATE DATABASE permissions to these two
users:
2.
3.
4. USE Northwind
5.
6. EXEC sp_addlogin 'login1'
7. EXEC sp_addlogin 'login2'
8. EXEC sp_adduser 'login1','user1'
9. EXEC sp_adduser 'login2','user2'
10. GRANT CREATE TABLE TO user1

44
 Chapter 2. Elements of Transact-SQL

11. GRANT CREATE TABLE TO user2

12. GO
13. Using Query Analyzer, open another connection to SQL Server, but use the newly created login1
login with blank password, and execute the following code:
14.
15.
16. USE Northwind
17.
18. CREATE TABLE Tablex
19. (col1 INT)
20. GO
21. Using Query Analyzer, open a third connection to SQL Server using the newly created login2 login
with blank password, and execute the following code:
22.
23.
24. USE Northwind
25.
26. CREATE TABLE Tablex
27. (col1 INT)
28. GO
29. Finally, to check that both tables were successfully created, execute the following code from the first
connection (the one with the sa login):
30.
31.
32. USE Northwind
33.
34. PRINT 'user1'
35. SELECT * FROM user1.Tablex
36. PRINT 'user2'
37. SELECT * FROM user2.Tablex
38. GO

user1
col1
-----------
(0 row(s) affected)

user2
col1
-----------
(0 row(s) affected)
Notice that in the last piece of code, the name of the tables had to be qualified with the owner's name.
A fully qualified name has four parts:

Servername.databasename.owner.objectname
The first three parts can be omitted. Thus, if you specify only the object name, SQL Server uses the current
user, current database, and local server. The first part, the server name, must be specified when working with
distributed queries (queries that span servers). The second part, the database name, must be specified when
executing queries across databases. For example, Listing 2.3 shows a SELECT statement that is executed
from the Northwind database that queries a table in the Pubs database.

Listing 2.3 Using the Database Name to Fully Qualify Names

45
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT * FROM Pubs.dbo.Jobs

GO

job_id job_desc min_lvl max_lvl

------ -------------------------------------------------- ------- -------
1 New Hire - Job not specified 10 10
2 Chief Executive Officer 200 250
3 Business Operations Manager 175 225
4 Chief Financial Officer 175 250
5 Publisher 150 250
6 Managing Editor 140 225
7 Marketing Manager 120 200
8 Public Relations Manager 100 175
9 Acquisitions Manager 75 175
10 Productions Manager 75 165
11 Operations Manager 75 150
12 Editor 25 100
13 Sales Representative 25 100
14 Designer 25 100

(14 row(s) affected)

Finally, the third part or owner specifies the object owner. This is useful in cases when two or more users own
an object with the same name, like the previous example, in which both user1 and user2 own a table called
Tablex.

Rules for Identifiers

When creating databases or database objects, the name (object identifier) can have up to 128 characters and
116 characters for temporary objects (because SQL Server adds a suffix to the object's name).
A regular identifier is the one that complies with the following rules:

• The first character is either a letter, the at sign (@), the number sign (#), or the underscore character
(_).
• The identifier doesn't contain any spaces.
• The identifier is not a Transact-SQL reserved keyword.

Any identifier that doesn't comply with any of these rules is not considered a regular identifier and must be
enclosed in brackets. For example, Listing 2.4 shows the use of brackets when creating an object whose
name contains spaces (a delimited identifier).

46
 Chapter 2. Elements of Transact-SQL

Listing 2.4 Using Delimited Identifiers

USE Northwind

CREATE TABLE [Name contains spaces]

(
cola INT,
colb VARCHAR(20)
)
GO

The command(s) completed successfully.

There are some special considerations regarding identifiers:

• If the first character is #, it represents a local temporary object (either a local temporary table or a
local temporary stored procedure). Listing 2.5 shows the creation of a local temporary table called
#EmployeeBasicInfo.

Listing 2.5 Creating a Local Temporary Table

USE Northwind

CREATE TABLE #EmployeeBasicInfo

(
employeeid INT,
lastname VARCHAR(20),
firstname VARCHAR(20)
)
GO

47
Microsoft SQL Server 2000 Programming by Example

The command(s) completed successfully.

• If the first character is ##, it represents a global temporary object (either a global temporary table or a
global temporary stored procedure). Listing 2.6 shows the creation of a global temporary table called
##ProductBasicInfo.

Listing 2.6 Creating a Global Temporary Table

USE Northwind

CREATE TABLE ##ProductBasicInfo

(
productid INT,
productname VARCHAR(40)
)
GO

The command(s) completed successfully.

• If the first character is @, it represents a local variable. Because of this, you can't use the @ sign as
the first character of the name of any database object. The statement used to define a local variable is
the DECLARE statement. Listing 2.7 shows how to define a local variable (notice that the first
character of the variable's name is @).

Listing 2.7 Creating a Local Variable

48
 Chapter 2. Elements of Transact-SQL

USE Northwind

DECLARE @age INT

SET @age = 25
GO

The command(s) completed successfully.

Transact-SQL Programming Conventions

As a good programming practice, these are some conventions you can follow:

• Use uppercase characters for all reserved keywords.

• Capitalize all table names. In general, you should capitalize all objects that are collections.
• Use lowercase characters for all scalars, such as column names and variables.
• Keep names unique. In other words, try not to use the same name for more than one object.
• Regarding object ownership, the database owner (dbo) should be the owner of all objects in the
database because this makes database administration easier and prevents a broken ownership chain.
If, by any chance, the owner of a certain object must be changed, use the sp_changeobjectowner
system stored procedure. Moreover, if you want to change the database owner, use the
sp_changedbowner system stored procedure.

Data Manipulation Language (DML)

The Data Manipulation Language (DML) is the most commonly used component of Transact-SQL by
database developers. Basically, it is used to retrieve, insert, modify, and delete information from databases.
These four operations are performed through the commands that compose the DML, respectively:

• SELECT
• INSERT
• UPDATE
• DELETE

Therefore, any application or client who wants to interact with SQL Server to retrieve, insert, modify, or delete
information has to do it through one of these four elements of the Transact-SQL DML.
Listing 2.8 shows a basic example of each one of the four statements that comprise the Data manipulation
language (DML).

Listing 2.8 Using the DML to Interact with the Database

49
Microsoft SQL Server 2000 Programming by Example

USE Northwind

INSERT INTO Customers (customerid, companyname, contactname, contacttitle)

VALUES ('ACME1','ACME Publishing','Fernando','DBA')
GO

UPDATE Customers
SET contactname = 'Fernando Guerrero'
WHERE customerid = 'ACME1'
GO

SELECT customerid,companyname
FROM Customers
WHERE customerid = 'ACME1'
GO

DELETE Customers
WHERE customerid = 'ACME1'
GO

(1 row(s) affected)

(1 row(s) affected)

customerid companyname
---------- ----------------------------------------
ACME1 ACME Publishing

(1 row(s) affected)

(1 row(s) affected)

Data Control Language (DCL)

The Data Control Language is the subset of Transact-SQL used to manage security in databases. Specifically,
it is used to set permissions on database objects and statements. In general, after you create the database
and database objects (through DDL), you are ready to set up permissions using the statements provided by
the Data Control Language. The three statements that comprise the Data Control Language are

• GRANT—Used to grant access on an object or a statement to a user.

• DENY—Used to explicitly deny any permission on any object or statement. This always takes
precedence over any other permission inherited by role or group membership.

50
 Chapter 2. Elements of Transact-SQL

• REVOKE—Removes any entry in the permissions table (syspermissions) that either granted or
denied access on an object or a statement to a user. Hence, REVOKE is used to revert a previous
GRANT or DENY.

The syntax used for these statements varies depending on the kind of permissions you want to set— either on
an object or statement. The syntax used to set permissions on objects is

GRANT permission ON object TO user

DENY permission ON object TO user
REVOKE permission ON object TO user
Listing 2.9 shows how you can use the GRANT statement to allow user1 to see the contents of the
Categories table.

Listing 2.9 Using the GRANT Statement to Set Object Permissions

USE Northwind

GRANT SELECT ON Categories TO user1

GO

The command(s) completed successfully.

On the other hand, the syntax used to set permissions on statements is

GRANT statement TO user

DENY statement TO user
REVOKE statement TO user
Listing 2.10 shows an example of the GRANT statement, which gives user1 permission to create tables in
the Northwind database.

Listing 2.10 Using the GRANT Statement to Set Statement Permissions

USE Northwind

51
Microsoft SQL Server 2000 Programming by Example

GRANT CREATE TABLE TO user1

GO

The command(s) completed successfully.

Permissions can be set on either objects or statements. A database object can be a table, view, user-defined
function, stored procedure, or extended stored procedure. Thus, different permissions can be applied for each
kind of object. Table 2.1 lists permissions that apply for each database object. Notice that the three kinds of
user-defined functions have different sets of permissions that can be set.
Table 2.1. Permissions on Database Objects
Objects Permissions
Table, view, inline table valued function SELECT, INSERT, UPDATE, DELETE, REFERENCES
Scalar valued function EXECUTE, REFERENCES
Multistatement table valued function SELECT, REFERENCES
Stored procedure, extended stored procedure EXECUTE
All these kinds of permissions are very straightforward; they allow users to do what they say— SELECT,
INSERT, UPDATE, DELETE, and EXECUTE. Regarding the REFERENCES permission, to be able to create a
foreign key to a certain table, you need to have REFERENCES permissions on that table.
The second kind of permissions is statement permissions. Statements basically allow users to create objects
in the database and back up the database and the transaction log. These statements are BACKUP DATABASE,
BACKUP LOG, CREATE DEFAULT, CREATE FUNCTION, CREATE PROCEDURE, CREATE RULE, CREATE
TABLE, and CREATE VIEW. One statement in which permissions can be granted only in the MASTER
database is the CREATE DATABASE statement. Listing 2.11 demonstrates this fact; first, it creates a new
login (login3), then it adds a user (user3) in master for this new login, and then the CREATE DATABASE
permission is granted to user3. Notice that this permission is granted in the MASTER database.

Listing 2.11 Using the GRANT Statement to Set the CREATE DATABASE Statement Permission

USE Master

EXEC sp_addlogin 'login3'

EXEC sp_adduser 'login3','user3'
GRANT CREATE DATABASE TO user3
GO

52
 Chapter 2. Elements of Transact-SQL

New login created.

Granted database access to 'login3'.
Permissions are managed in the local database. In other words, you can set permissions on objects or
statements to users or roles only in the current database. If you want to set permissions on objects or
statements in some other database, you need to change the database context.
Tip

There is a way to set permissions (either on objects or statements) to all users in a database.
Because the PUBLIC database role contains all users and roles in a database, if you set
permissions to public, all users will inherit these permissions.

In Listing 2.12, the public database role in Northwind is allowed to create tables in the local database. As a
result, any user in Northwind will now be able to create tables.

Listing 2.12 Using the GRANT Statement to Set Permissions to Public

USE Northwind

GRANT CREATE TABLE TO public

GO

The command(s) completed successfully.

Security information is stored in the Syspermissions system table. The sp_helprotect system stored
procedure displays user permissions information on an object or statement and gets this information from
Syspermissions. sp_helprotect can receive the name of an object or a statement as a parameter and
return the security information associated with this object or statement. When used with no parameters, it

53
Microsoft SQL Server 2000 Programming by Example

returns permissions information for all objects and statements in the current database. Listing 2.13 shows
the security information of the CREATE TABLE statement in the Northwind database.

Listing 2.13 Listing Security Information Through the sp_helprotect System Stored Procedure

USE Northwind

EXEC sp_helprotect 'CREATE TABLE'

GO

Owner Object Grantee Grantor ProtectType Action Column

----- ------ ------------ ------- ----------- ------------ ------
. . public dbo Grant Create Table .
. . user1 dbo Grant Create Table .
. . user2 dbo Grant Create Table .

Data Types
Transact-SQL provides 27 different data types, which basically control the data that you can store in columns,
variables, parameters, and expressions.
SQL Server 2000 introduced three new data types: BIGINT, TABLE, and SQL_VARIANT. BIGINT uses 8
bytes to store integers, which is 4 bytes more than the INT data type, which uses only 4 bytes of storage size.
The TABLE data type behaves like a table, storing sets of rows. Be aware that the TABLE data type cannot be
used by columns in tables. And finally, the SQL_VARIANT data type allows you to store almost any value
(integer, character, numeric, and so on), except TEXT, NTEXT, IMAGE, TIMESTAMP, and SQL_VARIANT
values. The SQL_VARIANT data type is similar to the variant data type in Visual Basic.
The data types supplied by Transact-SQL are divided in these main categories: limited character, unlimited
character, binary, binary large objects, integers, approximate numeric, exact numeric, date and time, currency,
and other. Table 2.2 lists all these categories with their respective data types and a brief description of each
data type.
Table 2.2. Transact-SQL Data Types
Category Data Type Description
Limited CHAR Fixed-length character data (up to 8,000 characters)
character VARCHAR Variable-length character data (up to 8,000 characters)
NCHAR Fixed-length Unicode character data (up to 4,000 characters)
NVARCHAR Variable-length Unicode character data (up to 4,000 characters)
Unlimited TEXT Variable-length character data (up to 2,147,483,647 characters)
character NTEXT Variable-length Unicode character data (up to 1,073,741,823
characters)

54
 Chapter 2. Elements of Transact-SQL

Binary BINARY Fixed-length binary data (up to 8,000 bytes)

VARBINARY Variable-length binary data (up to 8,000 bytes)
Binary large IMAGE Variable-length binary data (up to 2,147,483,647 bytes)
objects
Integers BIGINT Integer from -2^63 to 2^63 - 1
INT Integer from -2,147,483,648 to 2,147,483,647
SMALLINT Integer from -32,768 to 32,767
TINYINT Integer from 0 to 255
BIT Binary integer with only two possible values (0 or 1)
Approximate REAL Approximate numbers with a precision between 1 and
numeric 7 (4 bytes of storage)
FLOAT Approximate numbers with a precision between 8 and
15 (8 bytes of storage)
Exact numeric DECIMAL Exact numbers that can use up to 17 bytes to store
data
NUMERIC Synonym of DECIMAL
Date and time DATETIME Date and time data from January 1, 1753 to
December 31, 9999. Time is accurate to the 1/300
of a second
SMALLDATETIME Date and time data from January 1, 1900 to June 6, 2079. Time is
accurate to the minute
Currency MONEY Currency data from -922,337,203,685,477.5808 to
922,337,203,685,477.5807
SMALLMONEY Currency data from -214,748.3648 to 214,748.3647
Other UNIQUEIDENTIFIER 16-byte GUID
TABLE Similar to a table database object, and used just
for temporary storage
SQL_VARIANT Can store any Transact-SQL data type, but TEXT,
NTEXT, IMAGE, TIMESTAMP, and itself
TIMESTAMP or 8-byte binary number that changes every time a column is inserted
ROWVERSION or updated
CURSOR Used only for variables and stored procedures output parameters
Note that Unicode data types (NCHAR, NVARCHAR, and NTEXT) use two bytes per character. This is why they
can store only half the space of the same non-Unicode data types (CHAR, VARCHAR, and TEXT). However,
because Unicode data uses two bytes per character, it can store data in any language. To specify Unicode
data, use the N prefix, as shown in Listing 2.14.

Listing 2.14 Specifying Unicode Data

USE Northwind

DECLARE @unicode_data NCHAR(20)

SET @unicode_data = N'This is unicode data'
GO

55
Microsoft SQL Server 2000 Programming by Example

The command(s) completed successfully.

The UNIQUEIDENTIFIER data type has an associated system function, NEWID, that generates new values
for these data types. Therefore, if you use the UNIQUEIDENTIFIER data type in a table, you can use the
NEWID system function as the column's default value, as Listing 2.15 shows.

Listing 2.15 Using the NEWID System Function

USE Northwind

CREATE TABLE Bigidentifiers

(
col1 UNIQUEIDENTIFIER DEFAULT NEWID()
)
GO

The command(s) completed successfully.

The new SQL_VARIANT data type can store up to 8,016 bytes of almost any base data type. Listing 2.16
shows how to use the SQL_VARIANT with character and integer data.

Listing 2.16 Using the SQL_VARIANT Data Type

USE Northwind

DECLARE @integer_data SQL_VARIANT, @char_data SQL_VARIANT

56
 Chapter 2. Elements of Transact-SQL

SET @integer_data = 845

SET @char_data = 'This is character data'
GO

The command(s) completed successfully.

The TIMESTAMP data type is not related at all to DATETIME or SMALLDATETIME. Moreover, you cannot
directly update a TIMESTAMP column because it updates itself when you insert or update a row that contains
a TIMESTAMP. Also, there can be only one TIMESTAMP column per table.

Creating Customized Data Types: User-Defined Data Types

Users can create their own data types using the data types provided by Transact-SQL as the base types. To
create user-defined data types, or UDDT, use the sp_addtype system stored procedure, and to drop them,
use sp_droptype. The basic syntax of the sp_addtype system stored procedure is

sp_addtype uddt_name, uddt_base_type, nullability

For example, suppose that you want to create a UDDT to store phone numbers that could be null. You can
define this user-defined data type using the CHAR data type as the base type with a length of 12, as shown in
Listing 2.17.

Listing 2.17 Creating User-Defined Data Types (UDDTs)

USE Northwind

EXEC sp_addtype phone_number,'CHAR(12)',NULL

GO

1 row(s) affected)

Type added.
Information about user-defined data types is stored in the systypes system table, which is located in all
databases. When created, the properties of user-defined data types can be displayed using the sp_help

57
Microsoft SQL Server 2000 Programming by Example

system stored procedure, which receives an object name as a parameter, which, in this case, would be the
name of the user-defined data type, as shown in Listing 2.18.

Listing 2.18 Using sp_help to Display UDDT's Properties

USE Northwind

EXEC sp_help phone_number

GO

Type_name Storage_type Length Prec Scale Nullable Default_name Rule_name

------------ ------------ ------ ---- ----- -------- ------------ ---------
phone_number char 12 12 NULL yes none none
Tip

UDDTs are stored in the database where they are created. However, if you want all your user
databases to have a set of predefined, user-defined data types when they're created, create these
UDDT in the model database. This is because every new database that is created in SQL Server is
a copy of the model database.

Listing 2.19 creates a UDDT called ssn in Model. This UDDT will be automatically transferred to every user
database that is created afterward.

Listing 2.19 Creating UDDTs in Model

USE Model

EXEC sp_addtype ssn,'CHAR(11)','NOT NULL'

58
 Chapter 2. Elements of Transact-SQL

GO

1 row(s) affected)

Type added.
UDDTs can be created also in Enterprise Manager. To accomplish this, right-click User Defined Data Types
(which is located inside the database folder) and then choose New User-Defined Data Type, which opens the
window shown in Figure 2.1.

Figure 2.1. Creating UDDTs in Enterprise Manager.

Data Type Selection Criteria

You should be very careful when choosing data types. Always make sure that the data type you're choosing is
the correct one and the length is appropriate, because it is very common to choose data types that are
oversized. For example, let's say that you choose VARCHAR(100) as the data type and length for a ZIP code
column. Although VARCHAR(100) is able to store this kind of data, you will waste a lot of space because ZIP
codes have only five characters. In a small table, this isn't a problem, but in big tables this can lead to
performance problems.
The same rule applies for integer data. Take a look at the maximum and minimum value of each integer data
type when choosing among them. This way, you avoid using a big data type when you could have used a
smaller one. For example, a very efficient way to store IP addresses in a table is to use four TINYINT
columns, because this data type can store integers from 0 to 255.
If the length is not specified when declaring character (CHAR, NCHAR, VARCHAR, and NVARCHAR) or binary
(BINARY and VARBINARY) data, SQL Server uses 1 as the length by default. Listing 2.20 shows the

59
Microsoft SQL Server 2000 Programming by Example

declaration of a variable in which you will be able to store just one character because the length was not
specified. Notice that although you don't get an error if you assign more than one character to the variable,
SQL Server stores only the first character.

Listing 2.20 Using the Default Length with Character Data

USE Northwind

DECLARE @onecharacter VARCHAR

SET @onecharacter = 'String'
SELECT @onecharacter
GO
----
S

(1 row(s) affected)
Be aware that fixed-length data types always use the length you defined. On the other hand, variable-length
data types use only the actual space that is being used by the value. For example, look at the table shown in
Listing 2.21. If you insert a row and the length of the lastname is 5, SQL Server will use just 5 bytes for the
storage of this value because the data type is VARCHAR. On the other hand, if the length of the firstname is
5, SQL Server has to use 20 bytes to store this value because the data type is CHAR. Therefore, when using
the CHAR data type, even if the length of the value is less than the length of the column, SQL Server uses the
length of the whole column to store this value.

Listing 2.21 Using Variable- and Fixed-Length Character Data

USE Northwind

CREATE TABLE Authors

(
lastname VARCHAR(20),
firstname CHAR(20)
)
GO

60
 Chapter 2. Elements of Transact-SQL

The command(s) completed successfully.

If you want to store data that can hold more than 8,000 bytes, use the TEXT, NTEXT, or IMAGE data types,
which can store up to 2GB. However, make sure that you really need to store more than 8,000 bytes, because
these data types use another set of statements (WRITETEXT, READTEXT, and UPDATETEXT).
Tip

You can use standard DML commands with TEXT, NTEXT, and IMAGE data, but only a portion of
the data can be accessed (using the SUBSTRING function, for example).

Be careful when you use approximate numeric data because, by definition, these data types (FLOAT and
REAL) store an approximation of the number. Therefore, they should not be used to perform exact
comparisons in WHERE clauses.

The TABLE data type cannot be used as a column data type when creating tables; thus, it is not possible to
have tables inside tables. Whenever possible, use the TABLE data type instead of temporary tables because
the first one is stored in memory, improving performance considerably. Usually, the TABLE data type is used
to store temporary result sets, as shown in Listing 2.22, in which a variable is created using the TABLE data
type and then two rows are inserted.

Listing 2.22 Using the TABLE Data Type

USE Northwind

DECLARE @Authors TABLE(lastname VARCHAR(20), firstname VARCHAR(20))

INSERT @Authors VALUES ('Guerrero','Fernando')
INSERT @Authors VALUES ('Rojas','Carlos')
SELECT * FROM @Authors
GO

61
Microsoft SQL Server 2000 Programming by Example

(1 row(s) affected)

(1 row(s) affected)

lastlame firstname
-------------------- --------------------
Guerrero Fernando
Rojas Carlos

(2 row(s) affected)
Although TIMESTAMP and ROWVERSION are synonyms, you should use ROWVERSION instead of TIMESTAMP,
because Microsoft could change the functionality of TIMESTAMP in the future to be compliant with the ANSI
SQL-92 standard, which states that the TIMESTAMP data type stores date and time data.

Additional Elements
In addition to DDL, DML, DCL, and data types, Transact-SQL has some additional elements or extensions
that make life easier for programmers and administrators, and also make Transact-SQL a more powerful
language. Be aware that these extensions are not ANSI-SQL standard; therefore, they are not portable. If you
are concerned about portability, you should avoid using any of these extensions.
SQL Server is not the only relational database management system that adds new elements to the standard
language; this is done by the majority of the commercial database systems today. If you want to check that
your code is compliant with the ANSI standard, use the SET FIPS_FLAGGER statement provided by SQL
Server, which receives as a parameter the level of compliance that you want to check: ENTRY,
INTERMEDIATE, or FULL. Listing 2.23 shows how this statement is used to check the compliance of a
query that contains the TOP clause, which is a Transact-SQL extension.

Listing 2.23 Usi ng the SET FIPS_FLAGGER Statement to Check for ANSI Compliance

USE Northwind

SET FIPS_FLAGGER 'FULL'

SELECT TOP 3 lastname

FROM Employees
ORDER BY hiredate

GO

62
 Chapter 2. Elements of Transact-SQL

FIPS Warning: Line 1 has the non-ANSI statement 'USE'.

FIPS Warning: Line 3 has the non-ANSI statement 'SET'.
FIPS Warning: Line 5 has the non-ANSI clause 'TOP'.
lastname
--------------------
Leverling
Davolio
Fuller
To deactivate the checking of the ANSI compliance (because it remains activated for the session), use the
same statement (SET FIPS_FLAGGER) with the OFF parameter; that is, SET FIPS_FLAGGER OFF.

Variables

In Transact-SQL, local variables are used in stored procedures, user-defined functions, triggers, and user
scripts. Variables are valid in the session that created them; for example, if a stored procedure creates a
variable, it is valid only during the execution of the stored procedure.
Variables are first declared, using the DECLARE statement and specifying the variables'name (which has to
begin with @) and data type. The syntax is

DECLARE @variable_name datatype

Then, a value is set to the variable using either SET or SELECT. When a variable is declared, its value is
initialized to NULL until a value is assigned. Listing 2.24 shows the creation of the @firstname variable,
which uses the VARCHAR data type with a length of 20. Next, its value is set using the SET statement, and
finally, its value is shown using the SELECT statement.

Listing 2.24 Using Variables in Transact-SQL

DECLARE @firstname VARCHAR(20)

SET @firstname = 'Maria'
SELECT @firstname
GO
--------------------

63
Microsoft SQL Server 2000 Programming by Example

Maria
You can also assign values to variables in a query. Using this approach, make sure that the query returns only
one row because, otherwise, you will get just one value in the variable. For example, Listing 2.25
demonstrates how to assign variables (@ln and @fn) in a query to the Employees table. This query stores
the value of the first and last name of the employee whose ID equals 1 in the @fn and @ln variables. Then, it
shows the value that was assigned to each one of these variables.

Listing 2.25 Assigning Values to Variables in Queries

USE Northwind

DECLARE @ln VARCHAR(20), @fn VARCHAR(20)

SELECT @ln = lastname, @fn = firstname

FROM Employees
WHERE employeeid = 1

SELECT @fn, @ln

GO
-------------------- --------------------

Nancy Davolio
System functions that begin with @@ used to be called global variables. In fact, they are system functions that
don't have any parameters, and they are not global variables because you cannot declare and assign a value
to them; they are managed by SQL Server instead. Table 2.3 lists some of these system functions and the
value they return.
Table 2.3. System Functions That Begin with @@
System Function Return Value
@@CONNECTIONS Number of connections to SQL Server since the service was started.
@@ERROR Error code of the last statement executed (if it succeeded, it returns 0).
@@IDENTITY Last identity value inserted in the current session.
@@MAX_CONNECTIONS Maximum number of connections allowed.

64
 Chapter 2. Elements of Transact-SQL

@@OPTIONS Information about set options in the current session.

@@ROWCOUNT Number of rows affected by the last statement executed.
@@SERVERNAME Name of the server where SQL Server is installed.
@@SPID ID of the current process.
@@VERSION Current version of SQL Server.
For example, Listing 2.26 shows how to use these system functions, specifically @@servername (the name
of this server is SQLBYEXAMPLE).

Listing 2.26 Using System Functions

SELECT @@servername
GO

-------------------------------

SQLBYEXAMPLE
Caution

There are no global variables in SQL Server. The @@ prefix is used just by SQL Server's system
functions. Although you can declare variables using the @@ prefix, they won't behave as global
variables; they will behave just as local ones.

Operators

Operators are used in Transact-SQL to deal with variables, scalars, and, in general, expressions. There are
different kinds of operators, and each kind is used to manipulate different kinds of data types.
The assignment operator is the equal sign (=). It is used to set values to variables, as shown in the prec eding
section (see Listings 2.24 and 2.25).
The arithmetic operators are + (addition), – (subtraction), * (multiplication), / (division), and % (modulo or
remainder of division). These operators are used to work with integers, approximate numeric, and exact
numeric. The + and – operators also behave as unary operators (positive and negative), which deal with only
one expression. In Listing 2.27, you can see an example of the use of the division and modulo operators
and the negative unary operator.

Listing 2.27 Using Arithmetic Operators

65
Microsoft SQL Server 2000 Programming by Example

SELECT 8/4
SELECT 9%4
SELECT -7
GO
-----------

-----------
1

-----------
-7
The comparison operators are = (equal to), <> (not equal to), < (less than), > (greater than), <= (less than or
equal to), and >= (greater than or equal to). Comparison operators are used to deal with any kind of data
types but TEXT, NTEXT, and IMAGE. Listing 2.28 shows an example that uses the less than or equal to
(<=) operator.

Listing 2.28 Using the Less Than or Equal to Operator

USE Northwind

SELECT employeeid, lastname, firstname

FROM Employees
WHERE employeeid <= 8
GO

66
 Chapter 2. Elements of Transact-SQL

employeeid lastname firstname

----------- -------------------- ----------
1 Davolio Nancy
2 Fuller Andrew
3 Leverling Janet
4 Peacock Margaret
5 Buchanan Steven
6 Suyama Michael
7 King Robert
8 Callahan Laura

(8 row(s) affected)
The logical operators are AND, OR, NOT, BETWEEN, IN, and LIKE. These operators check a condition and
evaluate to true or false. AND evaluates to true if all expressions are true. OR evaluates to true if any of the
expressions are true. NOT evaluates to false if the expression is true, and true if the expression is false.
Listing 2.29 shows how to use the AND logical operator in the WHERE clause of a SELECT statement.

Listing 2.29 Using the AND Logical Operator

USE Northwind

SELECT employeeid, lastname, firstname, city

FROM Employees
WHERE firstname='anne'AND city='london'
GO

employeeid lastname firstname city

----------- -------------------- ---------- ---------------
9 Dodsworth Anne London

(1 row(s) affected)

67
Microsoft SQL Server 2000 Programming by Example

BETWEEN is used to check for an inclusive range (in an inclusive range the limits are included), and its syntax
is

1st_expression BETWEEN 2nd_expression AND 3rd_expression.

The first expression, usually a column, is checked to see whether it falls within the range of the second
expression and the third expression (both included in the range). The syntax of BETWEEN is equivalent to
(using >= and <=):

(1st_expression >= 2nd expression) AND (1st_expression <= 3rd_expression)

Listing 2.30 shows an example of BETWEEN, which gets all the rows in the Employees table whose ID is
between 2 and 5 (including 2 and 5).

Listing 2.30 Checking for Ranges Using the BETWEEN Operator

USE Northwind

SELECT employeeid, firstname, lastname

FROM Employees
WHERE employeeid BETWEEN 2 AND 5
GO

employeeid firstname lastname

----------- ---------- --------------------
2 Andrew Fuller
3 Janet Leverling
4 Margaret Peacock
5 Steven Buchanan

(4 row(s) affected)
The IN operator is, at some level, similar to BETWEEN. Instead of a range, it checks whether the first
expression is contained in a list of expressions. The syntax is the following:

expression IN (expression_1, expression_2, ..., expression_n)

Listing 2.31 shows an example of IN, which gets all the rows from the Employees table whose ID is 2, 6, or
9.

Listing 2.31 Using the IN Logical Operator

68
 Chapter 2. Elements of Transact-SQL

USE Northwind

SELECT employeeid, firstname, lastname

FROM Employees
WHERE employeeid IN (2,6,9)
GO

employeeid firstname lastname

----------- ---------- --------------------
2 Andrew Fuller
6 Michael Suyama
9 Anne Dodsworth

(3 row(s) affected)
The LIKE operator is used to find patterns in strings (pattern matching). Typically, you will want to look for a
specific pattern in the values of the rows in a given table. The syntax of LIKE is (the expression is usually a
column)

expression LIKE pattern_to_find

You specify the pattern you are looking for by using wildcards. Transact-SQL provides four types of wildcards
you can use with the LIKE operator. The first wildcard, and the most commonly used, is the percentage
character (%), which is used to specify any string of any length (0 or more). For those of you who have
previously worked with DOS or Access, the percentage character (%) is similar to * in these environments.
Listing 2.32 shows three queries that use %: The first one gets all employees whose first name begins with a,
the second one gets all employees whose first name ends with e, and the last one gets all employees whose
last name contains the character sequence ae, no matter the position.

Listing 2.32 Using the Percentage (%) Wildcard with LIKE

USE Northwind

SELECT firstname, lastname

FROM Employees

69
Microsoft SQL Server 2000 Programming by Example

WHERE firstname LIKE 'a%'

SELECT firstname, lastname

FROM Employees
WHERE firstname LIKE '%e'

SELECT firstname, lastname

FROM Employees
WHERE firstname LIKE '%ae%'
GO

firstname lastname
---------- --------------------
Andrew Fuller
Anne Dodsworth

firstname lastname
---------- --------------------
Anne Dodsworth

firstname lastname
---------- --------------------
Michael Suyama
The second wildcard is the underscore character (_), which denotes any single character. The third wildcard is
used to search for a character within a range or a set, which is delimited by brackets. For example, [a–z]
denotes a range that contains all characters between a and z, and [abc] denotes a set that contains three
characters: a, b, and c. The last wildcard is a variation of the third one, in which you want to search for a
character not within a range or set. Listing 2.33 shows an example of each of these wildcards. The first
query gathers all the employees whose first name begins with any character, and the last four characters are
anet. The second one returns all employees whose first name begins with either j or s. The third query gets all
employees whose first name does not begin with the character a, m, j, s, l, or r.

Listing 2.33 Using Wildcards with LIKE

USE Northwind

SELECT firstname, lastname

FROM Employees
WHERE firstname LIKE '_anet'

SELECT firstname, lastname

FROM Employees
WHERE firstname LIKE '[js]%'

70
 Chapter 2. Elements of Transact-SQL

SELECT firstname, lastname

FROM Employees
WHERE firstname LIKE '[^amjslr]%'
GO

firstname lastname
---------- --------------------
Janet Leverling

firstname lastname
---------- --------------------
Janet Leverling
Steven Buchanan

firstname lastname
---------- --------------------
Nancy Davolio
The last operator is the plus sign (+), which is used to concatenate strings, as shown in Listing 2.34.

Listing 2.34 Using the String-Concatenation Operator

DECLARE @first VARCHAR(10), @second VARCHAR(10)

SET @first = 'SQL '
SET @second = 'Server'
SELECT @first + @second
GO
--------------------

SQL Server
Generally, + is used to concatenate columns when querying tables, as Listing 2.35 shows.

Listing 2.35 Using the String-Concatenation Operator to Concatenate Columns

71
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT firstname + ''+lastname

FROM Employees
WHERE employeeId = 1
GO
-------------------------------

Nancy Davolio

Control of Flow Statements

Transact-SQL provides statements that you can use to control the flow of the code in your scripts. The most
common statements are IF .. ELSE and WHILE, which are very common among modern programming
languages.
Note

Transact-SQL does not provide a FOR statement, like many other programming languages. Instead,
it provides a WHILE statement, which can expose basically the same functionality of the FOR
statement.

IF .. ELSE

The IF statement contains a condition that is evaluated by SQL Server; if it is true, the code right after IF is
executed, and if it is not true, the code right after ELSE is executed. Notice that the ELSE statement is optional.
If there is more than one statement to execute in IF or ELSE, these have to be delimited by the BEGIN and
END statements. For example, Listing 2.36 demonstrates the use of IF .. ELSE with multiple statements.
This example uses EXISTS, which evaluates to true if there is at least one row in the query (in this case, the
query is SELECT * FROM Shippers), or false if there are no rows in the query. Also, this example returns a
message to the client using the PRINT statement, which takes a string as a parameter (this is why the integer
must be converted to character).

Listing 2.36 Using IF .. ELSE Control of Flow Statements

72
 Chapter 2. Elements of Transact-SQL

USE Northwind

IF EXISTS (SELECT * FROM Shippers)

BEGIN
DECLARE @number_rows INT
SELECT @number_rows = count(*) FROM Shippers
PRINT 'There are '+ CAST(@number_rows AS VARCHAR(10))
+ 'rows in the Shippers table'
END
ELSE
PRINT 'This table does not contain any rows'

GO

There are 3 rows in the Shippers table

RETURN

RETURN is used to exit unconditionally from a script or stored procedure. When used in stored procedures,
RETURN receives a parameter, the return code. As a standard coding convention, a return code of 0 means
success, and any other number than 0 indicates than an error occurred.
Listing 2.37 shows an example of RETURN. Observe that the statement that is right after RETURN is not
executed.

Listing 2.37 Aborting Script Execution Using RETURN

PRINT 'First step'

RETURN 2
PRINT 'Second step (this is not executed)'
GO

73
Microsoft SQL Server 2000 Programming by Example

First step

WAITFOR

WAITFOR can be used in one of two ways. Using the first one, it instructs SQL Serve r to wait until a specific
time, and the syntax is

WAITFOR TIME 'time'

Using the second way, WAITFOR indicates SQL Server to wait a specific amount of time.

WAITFOR DELAY 'time'

Listing 2.38 shows an example of each of these two approaches. The first WAITFOR statement waits until
8:00 a.m., and the second one waits until 1 minute after 8:00 a.m.

Listing 2.38 Using WAITFOR

WAITFOR TIME '08:00:00'

PRINT getdate()
WAITFOR DELAY '00:01:00'
PRINT getdate()
GO

Jan 9 2001 8:00AM

J9n 9 2001 8:01AM

WHILE

WHILE iterates (executing some statements) until a certain condition is true. If there is more than one
statement to execute until the condition is true, enclose them between BEGIN and END.
Listing 2.39 shows an implementation of multiplication that uses WHILE to loop as many times as the
second number indicates.

74
 Chapter 2. Elements of Transact-SQL

Listing 2.39 Looping with WHILE Statements

DECLARE @a INT, @b INT, @result INT

SET @a = 3
SET @b = 4
SET @result = 0
WHILE @b > 0
BEGIN
SET @result = @result + @a
SET @b = @b - 1
END
SELECT @result
GO
-----------

12

(1 row(s) affected)

BREAK

BREAK is used inside a WHILE statement to exit unconditionally from the WHILE loop. When SQL Server finds
a BREAK inside a WHILE, it continues the execution with the instruction right after the END of the WHILE.

CONTINUE

CONTINUE is used inside a WHILE statement to transfer the execution to the beginning of the WHILE
statement, restarting the loop. Listing 2.40 demonstrates how CONTINUE and BREAK are used inside a
WHILE loop.

Listing 2.40 Using CONTINUE and BREAK

DECLARE @count INT

75
Microsoft SQL Server 2000 Programming by Example

SET @count = 0
WHILE @count < 10
BEGIN
IF @count = 3
BREAK
SET @count = @count + 1
PRINT 'This line is executed'
CONTINUE
PRINT 'This line is never executed'
END
GO

This line is executed

This line is executed
This line is executed

GOTO

GOTO directs SQL Server to continue the execution in the place where a label is defined. It is very useful for
error handling because you can define a generic error handler and then use GOTO to execute this error
handler in the code. Listing 2.41 shows how to alter the execution of a script using GOTO.

Listing 2.41 Altering the Execution Using GOTO

IF NOT EXISTS (SELECT * FROM Suppliers)

GOTO no_rows

IF NOT EXISTS (SELECT * FROM Employees)

GOTO no_rows

GOTO finished
no_rows:
PRINT 'An error occurred'

finished:
PRINT 'The program has finished its execution'

76
 Chapter 2. Elements of Transact-SQL

The program has finished its execution

Comments

Transact-SQL provides two ways to include comments inside code. Using the first way, you can include one-
line comments, which are specified using "--" (two dashes). In this case, anything that follows "--" in a
specific line is considered a comment and is not evaluated by SQL Server.
The other type of comments are multiline comments, which are delimited by "/*" and "*/". Anything between
these two delimiters is considered as a comment.
Listing 2.42 shows an example of both kinds of comments that can be used in Transact-SQL.

Listing 2.42 Inserting Comments in the Code

/*
This is an example of the use
of comments in Transact-SQL
*/
SELECT @@version -- this query shows the current version of SQL Server
GO

-------------------------------------------------------------
Microsoft SQL Server 2000 - 8.00.194 (Intel X86)
Aug 6 2000 00:57:48
Copyright (c) 1988-2000 Microsoft Corporation
Enterprise Edition on Windows NT 5.0 (Build 2195: )

(1 row(s) affected)

Programming Scripts and Batches

The main characteristic of a batch is that it is processed by SQL Server as a unit, similar to stored procedures.
A batch might contain one or more Transact-SQL instructions, and the last statement in a batch is the GO

77
Microsoft SQL Server 2000 Programming by Example

statement. However, some restrictions apply to batches. The statements CREATE DEFAULT, CREATE
PROCEDURE, CREATE RULE, CREATE TRIGGER, and CREATE VIEW cannot appear more than once in a
batch. If any of these statements appears in a batch, it must be the only statement. Hence, if you want to have
more than one of these statements in the code, add a GO statement right after each one of them (thus creating
multiple batches).
A script comprises one or more batches— each of them separated by a GO statement. Using scripts, you can
store the schema of your database (DDL that creates all database objects) in a file. Scripts can be generated
using the Generate scripts utility in the Enterprise Manager. To use this utility, go to the databases folder in
the Enterprise Manager, then right-click the database name, and choose All Tasks, Generate SQL Script. This
window appears in Figure 2.2.

Figure 2.2. Generating scripts in Enterprise Manager.

The GO Statement

The GO statement is used to separate batches. Even though it is not a Transact-SQL element; it is used just
by the SQL Server utilities. Actually, it can be changed in the Query Analyzer to any other word, if you go to
the Options menu and then to the Connections tab. This tab appears in Figure 2.3.

Figure 2.3. Changing the batch separator in the Query Analyzer.

78
 Chapter 2. Elements of Transact-SQL

What's Next?

You already know the basics of the Data Definition Language (DDL) and the data types that can be used in
Transact-SQL. In the next chapter, we will use these two elements to create tables and views. In the case of
tables, we will show you how to create different types of tables and to alter the definition of tables. Regarding
views, you will learn how to create, maintain, and manipulate data through views.

79
 Chapter 3. Working with Tables and Views

Chapter 3. Working with Tables and Views

A table is the basic unit of storage in a relational database. Tables and relationships (elements that link tables)
are the most important elements of the relational model, which was designed by E. F. Codd in 1970. A table is
composed of columns and a set of rows. First, a column represents an attribute of the entity described by the
table. For example, an employee table might have these columns: Social Security number (SSN), first name,
and last name. Second, a row, or a tuple, contains the actual data that is stored in a table. In the employee's
example, if there are 10 employees in the company, this table will contain 10 rows.
A database object similar to tables in the way it is queried is a view. A view, also called a virtual table, is
basically a predefined query stored in a database; every time the view is queried, SQL Server reads its
definition and uses this definition to access the underlying table. Views add a layer between applications and
tables because, through views, applications don't have to query tables directly.
In previous versions of SQL Server, a view never stored data. Now, using a new feature of SQL Server 2000
called indexed views, youcan create indexes on views (with some restrictions), and this translates into
permanent storage of the result set produced by the view.
This chapter teaches you the following:

• How to create and modify tables

• The types of tables available in SQL Server
• The advantages and usage of views
• How to use extended properties to store metadata (information that describes objects) in a database

Creating and Altering Tables

The first step in the databasedesign process is creating the entity relationship model, which is a conceptual
representation of the database. The entity relationship model is comprised of entities, attributes, and
relationships. An entity represents a real-world object, such as cars, employees, orders, students, courses,
and teachers. Each entity has characteristics,which are called attributes. For example, the entity called
employee has these attributes: Social Security number (SSN), last name, and first name. The last piece is the
relationship, which is a link between two or more tables. There are three types of relationships: one-to-one
(1:1), one-to-many (1:M), and many-to-many (M:N). For example, there's a one-to-many relationship between
employees and orders because one employee can place many orders, and an order can be placed by just
one employee.
The study of entity-relationship models is out of the scope of this book. However, it's important that you
understand that entity-relationship modeling is the core of database design, and that good applications cannot
be developed without a good database design.
After the entity-relationshipmodel is finished, the next step is to convert it into the database structure.
Specifically, a new table is created to represent each entity, and the table will have as many columns as
attributes in the entity. Also, a table is created to represent each many-t o-many relationship. The columns of
this table will be the primary keys of the tables involved in the many-to-many relationship.
Some CASE (computer-aided software engineering)tools out there transform an entity-relationship model into
a script that you can run against a database server to create the database schema.
Caution

Be careful when using the terms "relation" and "relationship," because they can lead to confusion.
A relationship links two or more tables in a database, whereas relationis a synonym of table in the
relational theory.

Caution

By definition, a set has no ordering. Thus, in a table, rows have no specific order because a table
contains a set of rows. For this reason, there is no concept of first or last row in a table.

81
Microsoft SQL Server 2000 Programming by Example

In SQL Server, the smallest unit of storage is the page. A page is 8KB in size and can store one or more rows;
however, a row cannot span pages. For this reason, a single row can store up to 8,060 bytes, unless you use
TEXT, NTEXT, or IMAGE data types, which are stored separately in their own pages.

Each group of eight pages formsanextent, which is the unit of storage of tables and indexes. There are two
types of extents: uniform and mixed. A uniform extent stores data for only a single object, whereas a mixed
extent stores multiple objects'data. In other words, a uniform extent is a set of eight pages (64KB) that stores
a single object's (table or index) data (see Figure 3.1). Refer to Chapter 6, "Optimizing Access to Data:
Indexes," for more information on data storage.

Figure 3.1. Data storage in SQL Server: pages and extents.

Types of Tables

In versions of SQL Serverprior to SQL Server 2000, there were only two types of tables:permanent and
temporary. The TABLE data type is a new feature of SQL Server 2000 that we can add to our set of tools.
Note

Generally, for simplicity, permanent tables are just called tables. In contrast, when referring to
temporary tables, the whole term is used (temporary table).

Permanent Tables

Permanent tables store the actual data in a database. These are the tables you create as a result of the
conversion of the entity relationship model to a database structure.
Permanent tables are storedin the database where they are created. There are system tables (sysobjects,
syscolumns, and sysconstraints) that keep track of the configuration information of tables such as
owner, creation date, name and type of each column, and constraints defined on the table, among others.
System tables are permanent tables that are created automatically by SQL Server at installation time. Their
name begins with sys—for example, sysobjects. The purpose of system tables is to store metadata of
databases, such as objects in the database (tables, views, stored procedures, extended stored procedures,
and user-defined functions), users, roles, permissions, data and log files, and so on. By default, users cannot
insert or modify the data in system tables. If you want to modify the data in system tables— which is not
recommended— use the sp_configure system stored procedure to enable the 'allow updates'
configuration option. Listing 3.1 shows how to enable this option.

Listing 3.1 Using sp_configure to Allow Modifications to System Tables

82
 Chapter 3. Working with Tables and Views

sp_configure 'allow updates', 1

GO
RECONFIGURE WITH OVERRIDE
GO
One of the most important system tables in every database is sysobjects, which stores information of
every object in a database. The object type is stored in the type column of sysobjects. Table 3.1 lists the
possible values of the type column of the sysobjects system table.
Table 3.1. Types of Objects in SQL Server
Value in the Type Column Type of Object
C Check constraint
D Default constraint
F Foreign key constraint
FN Scalar function
IF Inline table function
K Primary key or Unique constraint
P Stored procedure
R Rule
S System table
TF Multistatement table function
TR Trigger
U User table
V View
X Extended stored procedure
For example, if you want to list all system tables in the Northwind database, you can query sysobjects and
filter by the 'S' type, as shown in Listing 3.2.

Listing 3.2 Listing System Tables in the Northwind Database

USE Northwind

SELECT name,crdate
FROM sysobjects
WHERE type = 'S'

83
Microsoft SQL Server 2000 Programming by Example

name crdate
-------------------- ---------------------------
sysobjects 2000-04-18 01:51:58.910
sysindexes 2000-04-18 01:51:58.910
syscolumns 2000-04-18 01:51:58.910
systypes 2000-04-18 01:51:58.910
syscomments 2000-04-18 01:51:58.910
sysfiles1 2000-04-18 01:51:58.910
syspermissions 2000-04-18 01:51:58.910
sysusers 2000-04-18 01:51:58.910
sysproperties 2000-04-18 01:51:58.910
sysdepends 2000-04-18 01:51:58.910
sysreferences 2000-04-18 01:51:58.910
sysfulltextcatalogs 2000-04-18 01:51:58.910
sysindexkeys 2000-04-18 01:51:58.910
sysforeignkeys 2000-04-18 01:51:58.910
sysmembers 2000-04-18 01:51:58.910
sysprotects 2000-04-18 01:51:58.910
sysfulltextnotify 2000-04-18 01:51:58.910
sysfiles 2000-04-18 01:51:58.910
sysfilegroups 2000-04-18 01:51:58.910

(19 row(s) affected)

Temporary Tables

Temporary tables, like any other temporary object in SQL Server, are stored in tempdb and dropped
automatically by SQL Server if they are not dropped explicitly. This type of table is used as a temporary
working area for many purposes, such as multistep calculations and, also, to split up large queries.
There are two types of temporary tables: local and global. The name of local temporary tables begins with #,
whereas the name of global temporary tables begins with ##. Local temporary tables are available only in the
connection that created them, and when the connection is finished, the table is automatically dropped by SQL
Server, unless it is dropped explicitly using DROP TABLE. This type of table is very useful for applications that
run more than one instance of a process simultaneously, because each connection can have its own copy of
the temporary table, without interfering with the other connections executing the same code. For example, if
you create a stored procedure that uses temporary tables, every user or application running the stored
procedure will have its own copy of these temporary tables.
On the other hand, global temporary tables are available to all connections in SQL Server. Therefore, when a
connection creates a global temporary table, and other connections reference the table, they will be accessing
the same table. Global temporary tables last until the connection that created them finishes its execution.

Table Variables

In previous versions of SQL Server,temporary tables were the only way to store temporary data or result sets.
In SQL Server 2000, the TABLE data type can be used for this purpose. This new data type is more efficient
than temporary tables because it is stored in memory, whereas a temporary table is stored in tempdb.
Regarding scope, the TABLE data type is similar to local temporary tables, which have only local scope. As a
result, any variable that uses the TABLE data type is available only in the session where the variable is used,
and if this session calls a stored procedure, for example, table variables are not visible inside the stored
procedure, whereas temporary tables are visible.

84
 Chapter 3. Working with Tables and Views

To define a variable whose data typeis TABLE, use the DECLARE statement specifying TABLE as the data
type followed by the table structure. After declared, it is treated like any other table in SQL Server. Listing 3.3
shows how to declare a variable that uses the TABLE data type. This example also inserts a row in the table,
and then gets all rows on it.

Listing 3.3 Using the TABLE Data Type

DECLARE @employees TABLE (ssn INT, firstname VARCHAR(20), lastname VARCHAR(30))

INSERT @employees (ssn, firstname, lastname)

VALUES ('555555555','Rojas','Yumaira')

SELECT *
FROM @employees
(1 row(s) affected)

ssn firstname lastname

----------- -------------------- ------------------------------
555555555 Rojas Yumaira

(1 row(s) affected)
Table variables can also be used as the return value of user-defined functions. This will be covered in
Chapter 10, "Enhancing Business Logic: User-Defined Functions (UDF)."

Creating Tables

To create a table, you must specify the table's name, the columns that make up the table, and the data types
of the columns. You can create tables using a graphical interface in Enterprise Manager or using Transact-
SQL in Query Analyzer. The Transact-SQL statement used is CREATE TABLE,and the syntax is

CREATE TABLE Table_name (

column_1 data_type,
column_2 data_type,
.
.
column_n data_type
)
Listing 3.4 shows an example that creates the Drivers table, which contains three columns: license,
firstname, and make.

Listing 3.4 Creating Tables Using Transact-SQL

85
Microsoft SQL Server 2000 Programming by Example

USE Northwind

CREATE TABLE Drivers (

license VARCHAR(15),
firstname VARCHAR(30),
lastname VARCHAR(30)
)
In SQL Server, a table can have up to 1,024 columns and a database can contain up to 2,147,483,647 objects,
including tables. As you already know, the maximum size of a row in a table is 8,060 bytes. Nonetheless, you
can still create tables that have columns of variable data types (VARCHAR, NVARCHAR, and VARBINARY)
whose row size exceeds 8,060 bytes. In these cases, SQL Server creates the table, but it gives you a warning.
Be careful when creating these tables because you won't have all this space available. For example, suppose
you create a table that has two columns which use the VARCHAR data type, and the length is 5,000 bytes on
each column. You won't have 10,000 bytes available for each row— you will have only 8,060; therefore, if you
try to insert a row with more than 8,060 bytes, you will get an error. Listing 3.5 shows this example and the
warning message.

Listing 3.5 Creating Tables with a Row Size That Exceeds 8,060 Bytes

USE Northwind

CREATE TABLE Bigtable(

firstcolumn VARCHAR(5000),

secondcolumn VARCHAR(5000)
)
GO
Warning: The table 'Bigtable'has been created but its maximum row size
(10025) exceeds the maximum number of bytes per row (8060). INSERT or
UPDATE of a row in this table will fail if the resulting row length
exceeds 8060 bytes.
When a table is created, SQL Server automatically adds a new row to sysobjects (in the local database)
with the information about the newly created table. Also, a new row is added to syscolumns for each column
in the table. Be aware that dealing directly with system tables is not recommended, because their functionality

86
 Chapter 3. Working with Tables and Views

might change in future versions of SQL Server. The alternative is to use the INFORMATION_SCHEMA views
(INFORMATION_SCHEMA.COLUMNS and INFORMATION_SCHEMA.TABLES in this case), which, by the way,
are ANSI standard. Moreover, you can use the system stored procedure sp_help to show information of any
object in SQL Server, including tables.
Listing 3.6 demonstrates the use of sp_help to get a table's information. Notice that, for reasons of space,
two columns were removed from the output of sp_help (TrimTrailingBlanks and
FixedLenNullInSource).

Listing 3.6 Displaying a Table's Information

USE Northwind
GO
sp_help 'Drivers'

Name Owner Type Created_datetime

--------------- --------- ---------------- ---------------------------------
Drivers dbo user table 2000-11-16 02:10:31.320

Column_name Type Computed Length Prec Scale Nullable

------------ -------- --------- ------- ---- ----- ---------
license varchar no 15 yes
firstname varchar no 30 yes
lastname varchar no 30 yes

Identity Seed Increment Not For Replication

------------------------------- --------- ------------ -------------------
No identity column defined. NULL NULL NULL

RowGuidCol
--------------------------------------------
No rowguidcol column defined.
Data_located_on_filegroup
--------------------------------------------
PRIMARY

The object does not have any indexes.

No constraints have been defined for this object..

No foreign keys reference this table.

87
Microsoft SQL Server 2000 Programming by Example

No views with schema binding reference this table.

As already stated, another option to get an object's metadata is using INFORMATION_SCHEMAviews. Basically,
metadata is information about the object and its components. The advantages of using these views is that,
first, they're ANSI compatible (SQL-92 Standard), and second, they are independent of system tables, in the
way that if the functionality or schema of system tables changes in future versions, these views will still be
supported. Listing 3.7 shows how to get a table's metadata (the Drivers table) using two
INFORMATION_SCHEMA views: TABLES and COLUMNS.

Listing 3.7 Using INFORMATION_SCHEMA Views to Get Metadata

USE Northwind

SELECT *
FROM INFORMATION_SCHEMA.TABLES
WHERE table_name = 'Drivers'

SELECT *.
FROM INFORMATION_SCHEMA.COLUMNS
WHERE table_name = 'Drivers'

GO

TABLE_CATALOG TABLE_SCHEMA TABLE_NAME TABLE_TYPE

-------------- ------------- ----------- ----------
Northwind dbo Drivers BASE TABLE

(1 row(s) affected)

TABLE_CATALOG TABLE_SCHEMA TABLE_NAME COLUMN_NAME ORDINAL_POSITION DEFAULT

-------------- ------------- ----------- ------------ ---------------- -------
Northwind dbo Drivers license 1 NULL
Northwind dbo Drivers firstname 2 NULL
Northwind dbo Drivers lastname 3 NULL

(3 row(s) affected)
After a table is created, it must be populated. If you want to know how much space a table is using, use the
sp_spaceused system stored procedure. It takes the table's name as the parameter and displays the
number of rows and the used space. Listing 3.8 shows space information of the Employees table in the
Northwind database.

Listing 3.8 Displaying Space Information of Tables

88
 Chapter 3. Working with Tables and Views

USE Northwind
.
EXEC sp_spaceused 'Employees'
GO

name rows reserved data index_size unused

------------- ------- ---------- ---------- ------------- ------------------
Employees 9 320 KB 232 KB 48 KB 40 KB
Any object created by users, including tables, can be renamed using the sp_rename system stored
procedure, which has three arguments (the last one is optional). The first parameter is the old name of the
object, the second is the new name, and the third is the type of object to rename: column, database,
index, and userdatatype. When renaming tables the third parameter is not necessary, but it is required
to rename columns. When renaming columns, in the first parameter the name of the table that contains the
column must be specified, using the following syntax: 'table_name.column_name'.
User databases can be renamed if you specify database as the third parameter. Another system-stored
procedure, sp_renamedb, is used only to. rename user databases.
Listing 3.9 renames the Drivers table to Safedrivers, and then renames the column license to
licensenumber.

Listing 3.9 Renaming User Objects Using sp_rename

USE Northwind

EXEC sp_rename 'Drivers','Safedrivers'

GO

EXEC sp_rename 'Safedrivers.license','licensenumber','column'

GO

89
Microsoft SQL Server 2000 Programming by Example

Caution: Changing any part of an object name could break

scripts and stored procedures.
The object was renamed to 'Safedrivers'.
Caution: Changing any part of an object name could break
scripts and stored procedures.
The column was renamed to 'licensenumber'.
In tables, columns can allow null values, which represent unknown values. A null doesn't mean that the value
is either 0 for numbers or a zero-length string for characters, it just means unknown or undefined. The
nullability of a column is specified when the table is created and can be modified by altering the properties of
the column.
By default, Query Analyzer assumes that columns allow nulls if the nullability status is not specified when
tables are created using Transact-SQL. This is due to the fact that Query Analyzer automatically activates
ANSI settings every time a connection is made to SQL Server (the statement used for this setting is SET
ANSI_NULL_DFLT_ON).
In particular, when you create tables using the CREATE TABLE statement, the nullability status is specified
after the column data type. If the column allows null values, use the NULL keyword; otherwise, use NOT NULL.
For example, if you take a look at Listing 3.4, which creates the Drivers table, it doesn't specify the
nullability status for any of the columns, and this is why all of them allow null values (this can be seen in the
output of Listing 3.6).
Listing 3.10 illustrates how to explicitly specify the nullability status of columns when creating a table. It also
executes sp_help on this table to show whether each column is nullable.

Listing 3.10 Specifying Nullability Status When Creating Tables

USE Northwind

CREATE TABLE Cars (

serial VARCHAR(200) NOT NULL,
make VARCHAR(100) NOT NULL,
model VARCHAR(100) NOT NULL,
color VARCHAR(50) NULL
)

EXEC sp_help 'Cars'

GO

90
 Chapter 3. Working with Tables and Views

Name Owner Type Created_datetime

---------- --------- ---------------- ------------------------------
Cars dbo user table 2000-11-19 16:06:09.947

Column_name Type Computed Length Prec Scale Nullable

-------------- ----------- ----------- --------- ----- ----- ----------
serial varchar no 200 no
make varchar no 100 no
model varchar no 100 no
color varchar no 50 yes

Identity Seed Increment Not For Replication

------------------------------ -------- ----------- -------------------
No identity column defined. NULL NULL NULL

RowGuidCol
--------------------------------
No rowguidcol column defined.

Data_located_on_filegroup
--------------------------------
PRIMARY

The object does not have any indexes.

No constraints have been defined for this object.

No foreign keys reference this table.

No views with schema binding reference this table.
A table should always have a primary key, which is a column or set of columns that uniquely identifies each
row in a table. For example, the Social Security number can be the primary key in the Employees table
because there aren't two employees with the same Social Security number. Primary keys are part of the
integrity of tables (entity integrity), and this will be covered in Chapter 7, "Enforcing Data Integrity."
In case a table doesn't have aninherent primary key, the IDENTITY property can be used, which is basically a
number that auto-increments by itself and cannot be NULL. The IDENTITY property is similar to the
AUTONUMBER data type in Access. A seed and an increment can be specified with the IDENTITY column
when creating a table. If they're not specified, the default value is 1 for each of them.
Listing 3.11 creates the Dealers table, which contains an IDENTITY column, dealerid, with a seed of
10 and increment of 1.

Listing 3.11 Usingthe IDENTITY Property

91
Microsoft SQL Server 2000 Programming by Example

USE Northwind

CREATE TABLE Dealers(

dealerid INT IDENTITY(10,1) NOT NULL,
dealername VARCHAR(100) NOT NULL,
address VARCHAR(200) NULL,
city VARCHAR(100) NULL,
state CHAR(2) NULL,
zipcode VARCHAR(10) NULL
)
GO
Caution

The IDENTITY property used in tables is different from the IDENTITY function, which is used to
add an identity column to a table created by a SELECT INTO statement (the IDENTITY function is
covered in Chapter 4, "Querying and Modifying Data" ).

In SQL Server, only one IDENTITY column is allowed per table, and it cannot be updated. Also, because the
value of an IDENTITY column is automatically incremented every time a new row is inserted, you don't need
to specify this column when inserting data in the table. However, if you want to insert a specific value in an
IDENTITY column, use the SET IDENTITY_INSERT statement. The syntax of this statement is the following:

SET IDENTITY_INSERT Name_of_the_table { ON|OFF}

Every time you activate IDENTITY_INSERTin a table, don't forget to deactivate it (using the OFF keyword in
the SET IDENTITY_INSERT statement) after values are explicitly inserted in the IDENTITY column. Listing
3.12 shows how to insert data in the IDENTITY column of the Dealers table. In the first INSERT statement
the dealerid column is not specified, therefore, the next identity value is used (1 in this case). In the second
INSERT statement, a value of 8 is explicitly inserted in the dealerid column (notice the use of SET
IDENTITY_INSERT).

Listing 3.12 Inserting Data in IDENTITY Columns

USE Northwind

INSERT Dealers (dealername) VALUES ('Acme BMW')

92
 Chapter 3. Working with Tables and Views

GO

SET IDENTITY_INSERT Dealers ON

INSERT Dealers (dealerid,dealername) VALUES (18,'USA Toyota')
SET IDENTITY_INSERT Dealers OFF
GO

SELECT * FROM Dealers

GO

(1 row(s) affected)

(1 row(s) affected)

dealerid dealername address city state zipcode

----------- --------------- --------------- --------------- ----- ----------
10 Acme BMW NULL NULL NULL NULL
18 USA Toyota NULL NULL NULL NULL

(2 row(s) affected)
There are some systemfunctions related to IDENTITY columns, including IDENT_SEED, IDENT_INCR,
IDENT_CURRENT, SCOPE_IDENTITY, and @@IDENTITY. The first two functions, IDENT_SEED and
IDENT_INCR, return information about the increment and seed of an IDENTITY column in a specified table,
which is passed as the argument. Listing 3.13 demonstrates how these two functions are used.

Listing 3.13 Using IDENTITY System Functions (IDENT_SEED and IDENT_INCR)

USE Northwind

SELECT IDENT_SEED('Dealers'), IDENT_INCR('Dealers')

GO

----- -----

93
Microsoft SQL Server 2000 Programming by Example

10 1

(1 row(s) affected)
The last three functions, IDENT_CURRENT, SCOPE_IDENTITY, and @@IDENTITY, return last-generated
identity values. IDENT_CURRENT takes the name of a table as a parameter and returns the last identity value
inserted in this table. SCOPE_IDENTITY returns the last identity generated in the current scope, which can be
a stored procedure, trigger, user-defined function, or batch. Similarly, @@IDENTITY returns the last identity
value generated in the current session. The difference between these two functions is that @@IDENTITY is not
limited to the current scope; instead, it is limited to the current session, whereas SCOPE_IDENTITY is limited
to the current scope. A session might have one or more than one scope. Listing 3.14 shows how to call
these functions.

Listing 3.14 Using IDENT_CURRENT, SCOPE_IDENTITY, and @@IDENTITY

USE Northwind

SELECT IDENT_CURRENT('Dealers'), SCOPE_IDENTITY(), @@IDENTITY

GO
----- ----- -----

18 18 18

(1 row(s) affected)
Caution

@@IDENTITY behaves differently from any other system function that begins with @@. These
functions usually have serverwide scope. In contrast, @@IDENTITY is always associated with the
current session. To illustrate, if two users are connected to SQL Server (two different connections)
and these users insert a row in a table with an identity column, each one of them will get the value
they just inserted if they use @@IDENTITY.

Tip

Sometimes the IDENTITY property is not an option, because you might need to guar antee
uniqueness across tables, databases, or servers. For these cases, use the UNIQUEIDENTIFIER
(globally unique identifier, or GUID) data type, which is a 16-byte (128 bits) binary value.

94
 Chapter 3. Working with Tables and Views

After tables are created, you might want to change their owners. The sp_changeobjectowner system
stored procedure is used to change the owner of any object in SQL Server. This system-stored procedure is
very useful when you want to delete a user from a database and want to transfer all objects owned by this
user to another user in the database.

The statement used to drop user tables (system tables cannot be dropped) is DROP TABLEfollowed by the
name of the table. Be aware that if you create views or user-defined functions with the SCHEMABINDING
option and they reference tables, these objects (views or user-defined functions) have to be dropped first
before dropping the tables they reference.

Listing 3.15 illustrates how to drop a user table.

Listing 3.15 Using the DROP TABLE Statement

USE Northwind

DROP TABLE Dealers

GO
Notice that when you delete all rows from a table and it becomes an empty table, it still exists in the database;
therefore, it isn't dropped when it has no rows. want

Altering a Table's Definition

After tables are created,their structure can be modified using the ALTER TABLE statement.This statement can
be used to add columns, drop columns, change column properties (including data types), add constraints,
drop constraints, disable constraints and triggers, and re-enable constraints and triggers.
Caution

ALTER TABLE cannot be used if the compatibility level of the database is set to 65 (compatible
with SQL Server 6.5). To change this setting, use the sp_dbcmptlevel system stored procedure.

If you want to add a columnor columns to a table, use the following syntax:

ALTER TABLE Table_name ADD column_name data_type [NULL|NOT NULL]

Listing 3.16 adds a new column, mileage, to the Cars table.

95
Microsoft SQL Server 2000 Programming by Example

Listing 3.16 Adding a New Column to a Table

USE Northwind

ALTER TABLE Cars ADD mileage INT NULL

GO
To drop a column or columns, the following syntax is used:

ALTER TABLE Table_name DROP COLUMN column_name

For example, Listing 3. 17 drops the mileage and color columns of the Cars table.

Listing 3.17 DroppingColumns from a Table

USE Northwind

ALTER TABLE Cars DROP COLUMN mileage,color

GO
If you want to change the properties of a specific column, use this syntax:

ALTER TABLE Table_name ALTER COLUMN column_name new_data_type [NULL|NOT NULL]

Listing 3.18 changes the properties of three columns in the Safedrivers table. The first statement sets the
length of the licensenumber column to 30 and sets this column to not allow nulls. The second and the third
statements leave intact the data type and length of the firstname and lastname columns, but change the
nullability statusof these columns to NOT NULL.

Listing 3.18 Changing the Data Types and Nullability Status of Columns

USE Northwind

ALTER TABLE Safedrivers ALTER COLUMN licensenumber VARCHAR(30) NOT NULL

GO

ALTER TABLE Safedrivers ALTER COLUMN firstname VARCHAR(30) NOT NULL

GO

96
 Chapter 3. Working with Tables and Views

ALTER TABLE Safedrivers ALTER COLUMN lastname VARCHAR(30) NOT NULL

GO
Be aware that when adding a new column that doesn't allow NULL values, you must specify a default value for
all the rows in the table (using a DEFAULT constraint). The statement needed to add a new column with a
default constraint appears in Listing 3.19, which adds the state column (this column doesn't allow NULL
values) to the Safedrivers table, with a default value of FL.

Listing 3.19 Adding New Columns with a Constraint

USE Northwind

ALTER TABLE Safedrivers ADD state CHAR(2) NOT NULL

CONSTRAINT addstate DEFAULT 'FL'
GO
To drop a constraint:

ALTER TABLE Table_name DROP constraint_name

To disable a constraint (allowing data that normally would be rejected by the constraint):

ALTER TABLE Table_name NOCHECK CONSTRAINT constraint_name

Then, to re-enable the constraint:

ALTER TABLE Table_name CHECK CONSTRAINT constraint_name

To disable a trigger of a table (preventing the trigger from being executed):

ALTER TABLE Table_name DISABLE TRIGGER trigger_name

Then, to re-enable the trigger:

ALTER TABLE Table_name ENABLE TRIGGER trigger_name

Refer to Chapter 7, "Enforcing Data Integrity," for more information on constraints, and to Chapter 9,
"Implementing Complex Processing Logic: Programming Triggers," for more information on triggers.

Creating and Altering Views

A view is basicallya predefined query (a SELECT statement) that is stored in the database for later use.
Therefore, whenever you want to execute this predefined query again, you just have to query the view. The
tables that are referenced by the view are called base tables.
For example, suppose there is a complex query that involves many joins, and it is executed frequently by an
application. You can create a view that contains this query, and change the application to query this view
instead of executing the whole query.
Caution

Views are often called virtual tables. Be careful when using this terminology because some special
system tables in SQL Server are kept in memory; these are called virtual tables.

97
Microsoft SQL Server 2000 Programming by Example

Benefits of Views

Many benefits are associatedwith the use of views. Here's a summary of these benefits:

• Using views, users don't query the tables directly; therefore, you are creating a security layer between
users (or applications) and base tables. This, in turn, has another benefit: If the underlying database
schema changes, you don't have to change the application, just the views that access the tables.
• Views can be used to horizontally partition the data in a table. For example, suppose there is a table
with three columns, but some users are allowed to see only two of these three columns. You can
create a view that queries just the two columns they can see. Using this approach, these users will be
able to issue a SELECT * query against the view, which is not possible with the table.
• Information schema views can be used as an alternative way to deal directly with system tables. They
were introduced in SQL Server 7.0 as a method to provide information (metadata) about objects in
SQL Server. The benefit of using these views is that the functionality of system tables might change in
future releases of SQL Server, whereas these views'functionality will remain intact because they are
ANSI standard.
• Indexes can be created on views. This is a new feature of SQL Server 2000, which basically stores
the result set of a view in the database, or in other words, materializes the view. In general, the
advantage of indexed views is that this makes queries run faster, because SQL Server can take
advantage of the indexed view even if the view is not referenced in the query. When you create
indexes on views, SQL Server automatically updates the data of the index. Therefore, whenever the
data changes in the underlying tables, SQL Server updates the index.
• Another feature of SQL Server 2000 is the federated databases tech nology or distributed partitioned
views that are updatable. This is Microsoft's answer to the scale-out technology, in which the
database is spread across many servers, each server containing a subset of the whole data. This
technique is useful when you reach the point where scale-up (adding RAM, CPUs, and disks to the
database server) is not enough, and the database server cannot scale any more for whatever reason.
The trick is to create a view with the same name, in all the federated servers, that basically merges
the data in all these servers using UNION ALL statements. Then, when users access data, SQL
Server automatically takes the piece you need from the servers where it resides, making transparent
to users the fact that data is split in more than one server. The benefit of this new feature is that these
views are updatable, which allows applications to issue SELECT, INSERT, DELETE, and UPDATE
statements against these views, and SQL Server does the rest (queries or modifies the data in the
server where it resides).
• The last feature of SQL Server 2000 related to views is the introduction of instead-of triggers. In
previous versions of SQL Server, triggers could not be defined on views. Now, this new type of trigger
can be defined on views, which enhances tremendously the power of views in SQL Server. An
instead-of trigger, as its name indicates, executes the code of the trigger instead of the triggering
action (INSERT, UPDATE, or DELETE). This is covered in Chapter 9.

Creating and Dropping Views

Views are createdusing the CREATE VIEW statement. When you create a view, SQL Server checks that all
objects that the view references exist in the current database. Then, the code of the view is stored in the
syscomments system table, general information about the view is stored in sysobjects, and the columns
of the view are stored in syscolumns. A view can reference up to 1,024 columns.
The following is the basic syntax of CREATE VIEW:

CREATE VIEW View_name

AS
select_statement
Next, a SELECT statementis used to query the view. For example:

SELECT * FROM View_name

Listing 3.20 creates a view on the Customers table that gets all customers from Spain. Then, it contains a
SELECT statement to query this view.

98
 Chapter 3. Working with Tables and Views

Listing 3.20 Creating Views

USE Northwind
GO

CREATE VIEW Spaincustomers

AS
SELECT *
FROM Customers
WHERE country = 'Spain'
GO

SELECT * FROM Spaincustomers

GO
-- Partial results shown

CustomerID CompanyName ContactName

---------- ---------------------------------------- -----------------
BOLID Bólido Comidas preparadas Martín Sommer
FISSA FISSA Fabrica Inter. Salchichas S.A. Diego Roel
GALED Galería del gastrónomo Eduardo Saavedra
GODOS Godos Cocina Típica José Pedro Freyre
ROMEY Romero y tomillo Alejandra Camino

(5 row(s) affected)
Views can be nested 32 levels deep. In other words, a view can reference another view, and so on, up to 32
levels of nesting.
The system stored procedures that return a view's metadata are sp_help, which returns general information
about views; sp_helptext, which returns the definition of the view (if it's not encrypted); and sp_depends,
which displays object dependencies information, or, in other words, objects referenced by the view.Listing
3.21 shows how to use the first of these stored procedures, sp_help.

Listing 3.21 Displaying Views'Information

99
Microsoft SQL Server 2000 Programming by Example

USE Northwind

EXEC sp_help 'Spaincustomers'

GO

Name Owner Type Created_datetime

------------------- ---------- ---------- -----------------------------
Spaincustomers dbo view 2000-11-21 00:50:00.263

Column_name Type Computed Length Nullable

-------------- --------- --------- ------- --------
CustomerID nchar no 10 no
CompanyName nvarchar no 80 no
ContactName nvarchar no 60 yes
ContactTitle nvarchar no 60 yes
Address nvarchar no 120 yes
City nvarchar no 30 yes
Region nvarchar no 30 yes
PostalCode nvarchar no 20 yes
Country nvarchar no 30 yes
Phone nvarchar no 48 yes
Fax nvarchar no 48 yes

Identity Seed Increment Not For Replication

------------------------------- -------- ------------- -------------------
No identity column defined. NULL NULL NULL

RowGuidCol
-------------------------------------------------
No rowguidcol column defined.

No constraints have been defined for this object.

No foreign keys reference this table.
The object does not have any indexes.
Listing 3.22 shows an example of the sp_helptext system stored procedure that shows the definition of
the Spaincustomers view.

Listing 3.22 Displaying the View's Definition with sp_helptext

USE Northwind

EXEC sp_helptext 'Spaincustomers'

GO

100
 Chapter 3. Working with Tables and Views

Text
------------------------------------------
CREATE VIEW Spaincustomers
AS
SELECT *
FROM Customers
WHERE country = 'Spain'
Listing 3.23 demonstrates how to use sp_depends with views.

Listing 3.23 DisplayingObjects Referenced by the View

USE Northwind

EXEC sp_depends 'Spaincustomers'

GO

In the current database, the specified object references the following:

name type updated selected column
------------------------------ ---------------- ------- -------- -------------

dbo.Customers user table no yes Phone

dbo.Customers user table no yes Fax
dbo.Customers user table no yes Region
dbo.Customers user table no yes PostalCode
dbo.Customers user table no yes Country
dbo.Customers user table no yes ContactTitle
dbo.Customers user table no yes Address
dbo.Customers user table no yes City
dbo.Customers user table no yes CustomerID
dbo.Customers user table no yes CompanyName
dbo.Customers user table no yes ContactName

101
Microsoft SQL Server 2000 Programming by Example

Another way to get metadata information of views is using these INFORMATION_SCHEMA views:
INFORMATION_SCHEMA.TABLES and INFORMATION_SCHEMA.VIEWS. For example, Listing 3.24 uses
these two views to show informationof the Spaincustomers view.

Listing 3.24 Using INFORMATION_SCHEMA Views to Show Views'Metadata

USE Northwind

SELECT *
FROM INFORMATION_SCHEMA.TABLES
WHERE table_name = 'Spaincustomers'

SELECT *
FROM INFORMATION_SCHEMA.VIEWS
WHERE table_name = 'Spaincustomers'
GO

TABLE_CATALOG TABLE_SCHEMA TABLE_NAME TABLE_TYPE

-------------- ------------- --------------- ----------
Northwind dbo Spaincustomers VIEW

(1 row(s) affected)

TABLE_CATALOG TABLE_SCHEMA TABLE_NAME VIEW_DEFINITION

-------------- ------------- --------------- ---------------------------
Northwind dbo Spaincustomers CREATE VIEW Spaincustomers
AS
SELECT *
FROM Customers
WHERE country = 'Spain'

(1 row(s) affected)
Notice that when a view is created, if its definition contains a SELECT *,the query is expanded to get all the
columns referenced by the query to be able to store this information in syscolumns. In other words, SQL
Server translates the * into the actual list of columns. Sometimes, this can cause problems, because if you
create a view that queries all columns of a table (using a SELECT * query), and then add a new column to the
base table, this new column won't be included in the view. If you come across this problem, use the
sp_refreshview system-stored procedure to update the definition of the view in the system tables. This
stored procedure takes the name of the view as the parameter.
Listing 3.25 adds anew column to the Customers table, and then executes sp_depends to show that this
new column is not part of the view (because the SELECT * of this view was expanded when it was created).

102
 Chapter 3. Working with Tables and Views

Therefore, the output of sp_depends is exactly the same as the one of the preceding example (it doesn't
show the new column as part of the view).

Listing 3.25 Adding Columns to the Views'Base Tables

USE Northwind

ALTER TABLE Customers ADD Mobile VARCHAR(12) NULL

GO

sp_depends 'Spaincustomers'
GO

In the current database, the specified object references the following:

name type updated selected column
------------------------------ ---------------- ------- -------- -------------
dbo.Customers user table no yes Phone
dbo.Customers user table no yes Fax
dbo.Customers user table no yes Region
dbo.Customers user table no yes PostalCode
dbo.Customers user table no yes Country
dbo.Customers user table no yes ContactTitle
dbo.Customers user table no yes Address
dbo.Customers user table no yes City
dbo.Customers user table no yes CustomerID
dbo.Customers user table no yes CompanyName
dbo.Customers user table no yes ContactName
The next step uses sp_refreshview to update the viewdefinition. This is shown in Listing 3.26. Notice that
the new column (mobile) now appears in the output of sp_depends.

Listing 3.26 Using sp_refreshview

USE Northwind

103
Microsoft SQL Server 2000 Programming by Example

EXEC sp_refreshview 'Spaincustomers'

EXEC sp_depends 'Spaincustomers'

GO

In the current database, the specified object references the following:

name type updated selected Column
------------------------------ ---------------- ------- --------------------
dbo.Customers user table no yes Phone
dbo.Customers user table no yes Fax
-> dbo.Customers user table no yes Mobile
dbo.Customers user table no yes Region
dbo.Customers user table no yes PostalCode
dbo.Customers user table no yes Country
dbo.Customers user table no yes ContactTitle
dbo.Customers user table no yes Address
dbo.Customers user table no yes City
dbo.Customers user table no yes CustomerID
dbo.Customers user table no yes CompanyName
dbo.Customers user table no yes ContactName
The definition of a view can be encryptedin the syscomments system table using the WITH ENCRYPTION
option when creating the view. Using this option, the definition of the view cannot be seen by any user after
theview is created. If anybody tries to see the definition of the view, using sp_helptext for example, SQL
Server will display this message: "The object comments have been encrypted." Listing 3.27 creates a view
using the WITH ENCRYPTION option, and then tries to display its definition using sp_helptext.

Listing 3.27 Preventing Users from Reading the Code of Views

USE Northwind
GO

CREATE VIEW Mexicancustomers

WITH ENCRYPTION
AS
SELECT *
FROM Customers
WHERE country = 'Mexico'
GO

sp_helptext 'Mexicancustomers'
GO

104
 Chapter 3. Working with Tables and Views

The object comments have been encrypted.

Another feature of views is that you can create a virtual association between a view and the objects it
references. The advantage of this feature is that, when you activateit, any object that is referenced by the view
cannot be dropped. To create this virtual association, use the WITH SCHEMABINDING option when creating
the view.
The SCHEMABINDING option has two restrictions:

• The objects referenced by the view must also specify the owner— for example, 'dbo.Table_name'.
• The column list must be explicitly specified in the view; therefore, you cannot use SELECT *.

For example, Listing 3.28 creates a view (Toyotacars) that is schemabound to the Cars table. Note that
the SELECT statement contains the column list, and the name of the table specifies the object owner. Then,
SQL Server throws an error when trying to drop the base table (Cars).

Listing 3.28 Using the SCHEMABINDING Option

USE Northwind
GO

CREATE VIEW Toyotacars

WITH SCHEMABINDING
AS
SELECT serial, make, model
FROM dbo.Cars
WHERE make = 'Toyota'
GO

DROP TABLE Cars

GO

Server: Msg 3729, Level 16, State 1, Line 2

Cannot DROP TABLE 'Cars'because it is being referenced by object 'Toyotacars'.

105
Microsoft SQL Server 2000 Programming by Example

Normally, if you don't use SCHEMABINDING, objects referenced by views can be dropped, creating
inconsistencies in the database.
In general, a view cannot contain an ORDER BY clause. However, if you use the TOP 100 PERCENT clause
(which is covered in Chapter 4, "Query ing and Modifying Data" ) in the view, it is possible to add the
ORDER BY clause. For example, Listing 3.29 shows the creation of a view that contains an ORDER BY
clause.

Listing 3.29 Using the ORDER BY Clause in Views

USE Northwind
GO

CREATE VIEW Customersbyname

AS
SELECT TOP 100 PERCENT *
FROM Customers
ORDER BY contactname
GO
In general, you can modify data through views in the same way that you modify data in tables. However, some
restrictions need to be taken into con sideration when modifying data through views. Specifically, just one
table at a time can be updated when working through views. Thus, if a view references more than one table,
you cannot update data in all tables at once through the view. Also, data that is not modified through the view
must have a default value or accept nulls.
Regarding delet e operations, if you want to delete data from a certain table through a view, the view must
reference only one table (the table from which you want to delete data).
Listing 3.30 shows how to modify data stored in the Customers table through the Spaincustomers view.

Listing 3.30 Modifying Data Through Views

USE Northwind

UPDATE Spaincustomers
SET contactname = 'Maria Angelica Rojas',
contacttitle = 'Owner'
WHERE customerid = 'ROMEY'
GO
Sometimes, you want to make sure that when views are used for data modification, the new values of the
modified rows still belong to the result set of the view. To solve this problem, specify WITH CHECK OPTION
when creating the view. For example, if there's a view that lists all Brazilian customers and WITH CHECK
OPTION is specified, SQL Server throws an error when you try to change the country of any row through this
view, because the new value would not allow this row to be part of the view's result set. This example appears
in Listing 3.31.

106
 Chapter 3. Working with Tables and Views

Listing 3.31 Using WITH CHECK OPTION

USE Northwind
GO

CREATE VIEW Braziliancustomers

AS
SELECT *
FROM Customers
WHERE country = 'Brazil'
WITH CHECK OPTION
GO

UPDATE Braziliancustomers
SET country = 'USA'
WHERE customerid = 'WELLI'

GO

Server: Msg 550, Level 16, State 1, Line 1

The attempted insert or update failed because the target view either
specifies WITH CHECK OPTION or spans a view that specifies
WITH CHECK OPTION and one or more rows resulting from the operation
did not qualify under the CHECK OPTION constraint.
The statement has been terminated.
Similar to stored procedures, views can be used to enforce security in the database. Specifically, views can
be used to allow users to see only subsets of data. In some cases, views are better than assigning column
permissions in tables because, through a view, a user would be able to issue a SELECT * statement,whereas
with column permissions, the user wouldn't be allowed to issue a SELECT * against the table.
DROP VIEW is the statement used to drop the definition and permissions associated with a view. It can be
used to remove one or more views from the database, as shown in Listing 3.32.

Listing 3.32 Removing Views from the Database

107
Microsoft SQL Server 2000 Programming by Example

USE Northwind

DROP VIEW Customersbyname,Toyotacars

GO

Altering a View's Definition

The ALTER VIEW statementis used to change the definition or the options of views, keeping permissions
intact. If you drop the view, and then re-create it, permissions are lost.
You must specify the view's definition and options when using ALTER VIEW, because SQL Server overwrites
the old definition and options with the new ones specified in ALTER VIEW. Therefore, if you want to retain
functionality provided by options of the view, you must specify these options when altering the view.
For example, suppose that you want to alter the Mexicancustomers view to add the SCHEMABINDING
option. In this case, the ENCRYPTION option must be specified again if you want the view's definition to be
encrypted. Listing 3.33 shows this example.

Listing 3.33 Using ALTER VIEW

USE Northwind
GO

ALTER VIEW Mexicancustomers

WITH ENCRYPTION, SCHEMABINDING
AS
SELECT customerid, companyname, contactname
FROM dbo.Customers
WHERE country = 'Mexico'
GO
Tip

Be sure to store the definition of views in a safe place when you create views using the WITH
ENCRYPTION option, because after it is created, there is no way to show the definition. Therefore, if
you want to alter a view that is encrypted, you will need the source code.

What's Next?

You can create tables and views using the CREATE TABLE and the CREATE VIEW statements, respectively.
Furthermore, you can change their definition and properties by using the ALTER TABLE and ALTER VIEW
statements. After they're created, you can insert, modify, delete, and extract data in tables using the Data
Manipulation Language. This subset of the Transact-SQL language is covered in depth in the next chapter,
"Querying and Modifying Data."

108
 Chapter 4. Querying and Modifying Data

Chapter 4. Querying and Modifying Data

In Chapter 2, "Elements of Transact-SQL," you learned the basics of all the statements that make up the
Data Manipulation Language (DML), which are used to interact with the information stored in the database.
Moreover, these four elements of the DML (SELECT, INSERT, UPDATE, and DELETE ) are the core of
database programming.
This chapter teaches you the following:

• The components and syntax of the SELECT statement

• How to insert data in the database using the INSERT statement
• How to create a table and populate it on-the-fly from result sets
• How to modify data through the UPDATE statement
• How the data is removed using the DELETE statement

Querying Data

One of the main purposes of a database is to have a repository or a data storage system where information
can be extracted quickly. The SQL statement used to extract information from tables in a database is the
SELECT statement.

The SELECT Statement

The SELECT statement is used to extract information from the database or, in other words, to ask questions
(or queries) to the database.
The clauses or elements of the SELECT statement are FROM, WHERE, ORDER BY, GROUP BY, HAVING,
TOP, and INTO. The only element that is always required for queries is the FROM clause, which is used to
specify the table or view from which to extract information.
The basic syntax of SELECT is

SELECT column_list
FROM Table_name
When you issue a SELECT statement using this syntax, all rows are returned to the client because there are
no restrictions (the query doesn't have a WHERE clause).
The output of the SELECT statement is a result set that is composed of rows that come from one or more
tables or views (working with multiple tables at the same time using JOIN is covered in Chapter 5,
"Querying Multiple Tables: JOINs" ). If you want to get all columns of a table in a SELECT statement, use
the * wildcard instead of specifying the whole column list. However, if you want only certain columns to
appear in the output, these specific columns must be specified in the column list.
Listing 4.1 shows how to query a table using the * wildcard and using a column list. Notice that in both
cases the query returns all rows on the table without restrictions, but the second one shows only certain
columns.

Listing 4.1 Using a Basic SELECT Statement

USE Northwind

SELECT *

109
Microsoft SQL Server 2000 Programming by Example

FROM Shippers

SELECT ShipperID,CompanyName
FROM Shippers
GO

ShipperID CompanyName Phone

----------- ---------------------------------------- ------------------------
1 Speedy Express (503) 555-9831
2 United Package (503) 555-3199
3 Federal Shipping (503) 555-9931

(3 row(s) affected)

ShipperID CompanyName
----------- ----------------------------------------
1 Speedy Express
2 United Package
3 Federal Shipping

(3 row(s) affected)
Notice that the SELECT statement can be used by itself when printing constants or values of variables. Also,
SELECT is used to assign values to variables; similar to SET, which is used for the same purpose, but you
can assign only one variable for each SET statement. On the other hand, you can assign values to more than
one variable using one SELECT statement. In these cases (variable assignment and output), SELECT doesn't
need a FROM clause. Listing 4.2 demonstrates how to use SELECT and SET to assign values to variables,
and then it uses SELECT to show these values.

Listing 4.2 Using SELECT and SET to Assign Values to Variables and to Show These Values

DECLARE @firstname VARCHAR(10), @middlename VARCHAR(10), @lastname VARCHAR(10)

SET @firstname = 'Maria'
SELECT @middlename = 'Angelica', @lastname = 'Rojas'

SELECT @firstname, @middlename, @lastname

GO
---------- ---------- ----------

110
 Chapter 4. Querying and Modifying Data

Maria Angelica Rojas

In the column list of a SELECT statement, you also can include constants (or literals), which appear as new
columns in the result set. Furthermore, columns can be concatenated (using the + string concatenation
operator) to form a new column. These two techniques can be useful when populating tables using SELECT ..
INTO, to calculate values, and to build scripts dynamically.
Listing 4.3 contains two queries. The first one has a constant ('The name of the table is: ') and a
column (the name of the table that is extracted from the INFORMATION_SCHEMA.TABLES view). Notice that
in the output of the first query, the constant appears as the first column. The second query uses + to
concatenate two strings (a constant and a column) and generates one string (or a column resulting from the
concatenation). This query generates a script as output that can be used later.

Listing 4.3 Using Constants and Concatenating Strings in the Column List of SELECT

USE Northwind

SELECT 'The name of the table is: ', table_name

FROM INFORMATION_SCHEMA.TABLES
WHERE table_type = 'base table'

SELECT 'DROP TABLE '+ table_name

FROM INFORMATION_SCHEMA.TABLES
WHERE table_type = 'base table'
GO

-- partial results are shown

table_name
-------------------------- ----------------------------
The name of the table is: Cars
The name of the table is: Categories
The name of the table is: CategoriesBudget

(20 row(s) affected)

111
Microsoft SQL Server 2000 Programming by Example

-----------------------------------------

DROP TABLE Cars

DROP TABLE Categories
DROP TABLE CategoriesBudget

(20 row(s) affected)

When concatenating columns, make sure that the data type of the column is character. Otherwise, use
CONVERT or CAST to change it to character data to be able to use the concatenation operator. Listing 4.4
illustrates how to use CAST with a column whose data type is MONEY to change it to VARCHAR to be able to
concatenate it with other columns and constants.

Listing 4.4 Using String Concatenation and CAST in SELECT Statements

USE Northwind

SELECT 'The cost per unit of '+ productname + 'is '+

CAST(unitprice as VARCHAR(10))
FROM Products
GO

-- partial results are shown

-----------------------------------------------------------
The cost per unit of Chai is 18.00
The cost per unit of Chang is 19.00
The cost per unit of Aniseed Syrup is 10.00
The cost per unit of Chef Anton's Cajun Seasoning is 22.00
The cost per unit of Chef Anton's Gumbo Mix is 21.35
The DISTINCT clause is used to eliminate duplicates in a result set. For example, the Employees table has
more than one person with the same title. If you want to display all possible values of this column, you will get
repeated data, but if DISTINCT is used in the SELECT clause, only unique values will be listed. Listing 4.5
shows the difference between a query without DISTINCT and another one with it.

Listing 4.5 Using DISTINCT to Remove Duplicate Rows from a Result Set

112
 Chapter 4. Querying and Modifying Data

USE Northwind

SELECT title
FROM Employees

SELECT DISTINCT title

FROM Employees
GO

title
------------------------------
Sales Representative
Vice President, Sales
Sales Representative
Sales Representative
Sales Manager
Sales Representative
Sales Representative
Inside Sales Coordinator
Sales Representative

(9 row(s) affected)

title
------------------------------
Inside Sales Coordinator
Sales Manager
Sales Representative
Vice President, Sales

(4 row(s) affected)
In SELECT statements, the IDENTITYCOL keyword can be used instead of the name of an IDENTITY column.
For example, the Shippers table has a column with the IDENTITY property, shipperid. Therefore, when
referencing this column in a SELECT statement, you can use either its name or the IDENTITYCOL keyword,
which appears in Listing 4.6.

Listing 4.6 Using IDENTITYCOL Instead of the IDENTITY Column Name

113
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT shipperid
FROM Shippers

SELECT IDENTITYCOL
FROM Shippers
GO

shipperid
-----------
1
2
3

(3 row(s) affected)

ShipperID
-----------
1
2
3
.

(3 row(s) affected)

Column Aliases

You can use aliases to change default column names. Sometimes, assigning labels or aliases to the columns
in a SELECT statement can be beneficial because

• There is more than one column with the same name— This usually happens when you're working with
more than one table (using JOINs) and they have a column with the same name. In this case, it is
beneficial to use column aliases to differentiate between these columns.
• The column is the result of a calculation, becoming an expression— In these cases, SQL Server
doesn't assign a column name to these kinds of columns.

A column alias is specified using the following syntax:

column_name AS alias_name
The AS keyword is optional; therefore, the column name can be followed by the alias name. Also, the alias
name can contain up to 128 characters.
Caution

114
 Chapter 4. Querying and Modifying Data

The alias name must be enclosed in single quotation marks or brackets if it contains spaces.

Listing 4.7 shows how to use column aliases.

Listing 4.7 Using Column Aliases

USE Northwind

SELECT productname + '('+ quantityperunit + ')'as product_quantities,

unitsinstock + unitsonorder units
FROM Products
GO

partial results shown

product_quantities units
--------------------------------------------------------------- ------
Chai (10 boxes x 20 bags) 39
Chang (24 - 12 oz bottles) 57
Aniseed Syrup (12 - 550 ml bottles) 83
Chef Anton's Cajun Seasoning (48 - 6 oz jars) 53
Chef Anton's Gumbo Mix (36 boxes) 0
Grandma's Boysenberry Spread (12 - 8 oz jars) 120
Uncle Bob's Organic Dried Pears (12 - 1 lb pkgs.) 15
Northwoods Cranberry Sauce (12 - 12 oz jars) 6

The FROM Clause

You use the FROM clause to specify the tables or views involved in a query. In the case of multiple tables, the
type of JOIN and the JOIN condition are also specified in the FROM clause. Listing 4.8 shows a SELECT
statement that retrieves information from two tables, Territories and Region (Chapter 5 goes over
queries that involve multiple tables using JOINs).

Listing 4.8 Using the FROM Clause to Specify the Tables from Which Data Will Be Retrieved

115
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT Territories.territorydescription, Region.regiondescription

FROM Territories JOIN Region
ON Territories.regionid = Region.regionid
GO

partial results shown

territorydescription regiondescription
------------------------ ------------------
Westboro Eastern
Bedford Eastern
Georgetown Eastern
Boston Eastern
Cambridge Eastern
Braintree Eastern
Tables in other databases can be referenced in the FROM clause if you qualify them with the database name
and the owner (the last is optional). Further more, if you're working with linked servers (which are covered in
Chapter 15, "Working with Heterogeneous Environments: Setting Up Linked Servers" ), you can
access tables in those servers, but in this case you must qualify the table with the name of the server, and
then the database name (or catalog) and owner (or schema). Listing 4.9 illustrates this situation, retrieving
data stored in the Authors table in Pubs, from the Northwind database.

Listing 4.9 Using the FROM Clause to Specify Tables in Other Databases

USE Northwind

SELECT au_fname + ''+ au_lname AS name

FROM Pubs..Authors
GO

116
 Chapter 4. Querying and Modifying Data

name
----------------------------------
Abraham Bennet
Reginald Blotchet-Halls
Cheryl Carson
Michel DeFrance
Innes del Castillo
Ann Dull
Marjorie Green
Morningstar Greene
Burt Gringlesby
Sheryl Hunter
Livia Karsen
Charlene Locksley
Stearns MacFeather
Heather McBadden
Michael O'Leary
Sylvia Panteley
Albert Ringer
Anne Ringer
Meander Smith
Dean Straight
Dirk Stringer
Johnson White
Akiko Yokomoto

(23 row(s) affected)

A maximum of 256 tables can be referenced in a SELECT statement. If you have a query that requires
extracting information from more than 256 tables, use temporary tables or derived tables to store partial
results.

Table Aliases

You can use table aliases to make queries more readable, adding a label to a table (usually an identifier that
is shorter than the name of the table), and using this label to reference the table in the rest of the query.
Generally, table aliases are useful when you are writing queries that involve multiple tables (joining tables).
A table alias is specified using the following syntax (similar to column aliases):

Table_name AS alias_name
Notice that the AS keyword can be omitted. Therefore, specifying the table name followed by the alias name is
also valid. Listing 4.10 shows a query similar to the one shown in Listing 4.8, but this one uses table
aliases, which are used to qualify the columns in the column list and in the JOIN condition.

Listing 4.10 Using Table Aliases

117
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT T.territorydescription, R.regiondescription

FROM Territories T JOIN Region R
ON T.regionid = R.regionid
GO
partial results shown

territorydescription regiondescription
------------------------ ------------------
Westboro Eastern
Bedford Eastern
Georgetown Eastern
Boston Eastern
Cambridge Eastern
Braintree Eastern
Caution

If an alias is specified for a table, it must be used in the rest of the query— the name of the table
cannot be used.

The WHERE Clause

You have learned how to query a table (retrieving all rows) using the SELECT statement and the FROM clause.
Generally, you must restrict the number of rows that a query returns; therefore, only rows that meet certain
criteria or conditions will be part of the result set of the query. The WHERE clause restricts the result set of a
query based on a search condition. As a result, just the rows that meet the search condition will be returned
by the query. The syntax of a query that contains a WHERE clause is

SELECT column_list
FROM Table_name
WHERE conditions
Listing 4.11 shows a SELECT statement that retrieves the lastname, firstname, and hiredate of the
employees who live in Seattle.

Listing 4.11 Using the WHERE Clause to Restrict the Output of a Query

118
 Chapter 4. Querying and Modifying Data

USE Northwind

SELECT lastname, firstname, hiredate

FROM Employees
WHERE city = 'seattle'
GO

lastname firstname hiredate

-------------------- ---------- ---------------------------
Davolio Nancy 1992-05-01 00:00:00.000
Callahan Laura 1994-03-05 00:00:00.000

(2 row(s) affected)
In Transact-SQL, operators are used to work with expressions. Because a WHERE clause contains one or
more expressions to restrict the output of a query, all operators covered in Chapter 2 can be used in WHERE
clauses. These are LIKE, NOT LIKE, IN, NOT IN, BETWEEN, NOT BETWEEN, and comparison operators
(=, <>, <, >, <=, and >=). Listing 4.12 demonstrates how these operators are used in queries.

Listing 4.12 Using Operators in WHERE Clauses

USE Northwind

-- Returns all employees whose last name begins with 'b'

SELECT lastname, firstname
FROM Employees
WHERE lastname LIKE 'b%'

-- Returns all employees who don't live in Seattle, Redmond or Tacoma

SELECT lastname, firstname, city
FROM Employees
WHERE city NOT IN ('seattle','redmond','tacoma')

-- Returns all employees that were hired between 1/1/1993 and 12/31/1993
SELECT lastname, firstname, hiredate
FROM Employees

119
Microsoft SQL Server 2000 Programming by Example

WHERE hiredate BETWEEN '1993.1.1'AND '1993.12.31'

-- Returns all employees that live in any other city than London
SELECT lastname, firstname, city
FROM Employees
WHERE city <> 'london'.
GO

lastname firstname
-------------------- ----------
Buchanan Steven

(1 row(s) affected)

lastname firstname city

-------------------- ---------- ---------------
Leverling Janet Kirkland
Buchanan Steven London
NewFamily Michael London
King Robert London
Dodsworth Anne London

(5 row(s) affected)

lastname firstname hiredate

-------------------- ---------- -----------------------------
Peacock Margaret 1993-05-03 00:00:00.000
Buchanan Steven 1993-10-17 00:00:00.000
NewFamily Michael 1993-10-17 00:00:00.000

(3 row(s) affected)

lastname firstname city

-------------------- ---------- ---------------
Davolio Nancy Seattle
Fuller Andrew Tacoma
Leverling Janet Kirkland
Peacock Margaret Redmond
Callahan Laura Seattle.

(5 row(s) affected)
In a WHERE clause, many expressions can be combined using the AND and OR logical operators. Therefore:

• If AND is used, the rows returned by the query will be the ones that meet all the search conditions.
• On the other hand, if OR is used, the result set will contain the rows that meet any of the search
conditions.

An example of how these logical operators (AND and OR) are used in the WHERE clause that appears in
Listing 4.13.

Listing 4.13 Combining Expressions in the WHERE Clause Using AND and OR

120
 Chapter 4. Querying and Modifying Data

USE Northwind

-- Returns all employees whose last name begins with 'b'

-- and don't live in Seattle, Redmond or Tacoma
SELECT lastname, firstname, city
FROM Employees
WHERE lastname LIKE 'b%'
AND city NOT IN ('seattle','redmond','tacoma')

-- Returns all employees that either:

-- were hired between 1/1/1993 and 12/31/1993
-- or live in any other city than London
SELECT lastname, firstname, city, hiredate
FROM Employees
WHERE hiredate BETWEEN '1993.1.1'AND '1993.12.31'.
OR city <> 'london'
GO

lastname firstname city

-------------------- ---------- ---------------
Buchanan Steven London

(1 row(s) affected)

lastname firstname city hiredate

-------------------- ---------- --------------- -----------------------------
Davolio Nancy Seattle 1992-05-01 00:00:00.000
Fuller Andrew Tacoma 1992-08-14 00:00:00.000
Leverling Janet Kirkland 1992-04-01 00:00:00.000
Peacock Margaret Redmond 1993-05-03 00:00:00.000
Buchanan Steven London 1993-10-17 00:00:00.000
NewFamily Michael London 1993-10-17 00:00:00.000
Callahan Laura Seattle 1994-03-05 00:00:00.000

(7 row(s) affected)
When comparing DATETIME values in WHERE clauses, be aware that this data type stores both date and time.
Hence, if you want to compare just the date portion of the whole value, use the CONVERT function to get just
the portion you want. For example, if you need to retrieve all orders posted on 7/4/1996 no matter the time
they were posted, you can use the query shown in Listing 4.14.

Listing 4.14 Comparing DATETIME Values in WHERE Clauses

121
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT orderid, customerid, employeeid, orderdate

FROM Orders
WHERE CONVERT(VARCHAR(20),orderdate,102) = '1996.07.04'
GO

orderid customerid employeeid orderdate

----------- ---------- ----------- ----------------------------
10248 VINET 5 1996-07-04 00:00:00.000

(1 row(s) affected)
Caution

NULL values should be handled with care when comparing them in WHERE clauses. Specifically,
use IS NULL or IS NOT NULL, according to each case, to check for NULL values, and avoid
using comparison operators with NULL values— for example, 'column_name = NULL'—because
their behavior depends on the SET ANSI_NULLS setting.

Listing 4.15 shows an example of how IS NULL and IS NOT NULL are used within expressions in a
WHERE clause to search for NULL values.

Listing 4.15 Using IS NULL and IS NOT NULL to Make Comparisons with NULL Values

USE Northwind

-- Retrieves all suppliers whose region doesn't have a NULL value

SELECT companyname, contactname, region

122
 Chapter 4. Querying and Modifying Data

FROM Suppliers
WHERE region IS NOT NULL

-- Retrieves all suppliers whose region is NULL (or unknown)

SELECT companyname, contactname, region
FROM Suppliers
WHERE region IS NULL
GO

-- the output has been simplified

companyname contactname region

---------------------------------------- ------------------------------ ------
New Orleans Cajun Delights Shelley Burke LA
Grandma Kelly's Homestead Regina Murphy MI
Cooperativa de Quesos 'Las Cabras' Antonio del Valle Saavedra Asturia

(9 row(s) affected)
.
companyname contactname region
---------------------------------------- ------------------------------ ------
Exotic Liquids Charlotte Cooper NULL
Tokyo Traders Yoshi Nagase NULL
Mayumi's Mayumi Ohno NULL.

(21 row(s) affected)

Multiple expressions and the ISNULL function can be used in a WHERE clause as an elegant solution for
queries that contain optional search fields. For example, suppose you want to search for employees based on
city, title, or both. Two variables can be created to store the value of the city or title to search for (@city and
@title). If a variable equals NULL—for example, @city—this means that you are searching for a specific title
(which is stored in the @title variable). If both variables are NULL, you want to retrieve all rows in the table.
Usually, to solve this problem, you can validate each variable and create a query accordingly. These are the
possible cases:

• If just the city is used (@title equals NULL), build a query that just searches for employees who live
in this city.
• If just the title is used (@city equals NULL), build a query that just searches for employees whose title
is the one stored in the @title variable.
• If both values are used (city and title), build a query with two expressions in the WHERE clause, and
connect these two expressions with an AND operator.

Listing 4.16 shows how to code these three queries (each of them is based on the value of the @title and
@city variables). In this example, the @title variable is set to NULL and @city is set to London to retrieve
all employees who live in London.

Listing 4.16 Using Optional Search Fields with Different Queries

123
Microsoft SQL Server 2000 Programming by Example

USE Northwind

DECLARE @title VARCHAR(60), @city VARCHAR(30)

-- Setting @title to NULL and searching for all employees

-- who live in London

SET @title = NULL

SET @city = 'London'

IF @title IS NOT NULL AND @city IS NULL

SELECT lastname, firstname, title, city

FROM Employees
WHERE title = @title

IF @title IS NULL AND @city IS NOT NULL

SELECT lastname, firstname, title, city

FROM Employees
WHERE city = @city

IF @title IS NOT NULL AND @city IS NOT NULL

SELECT lastname, firstname, title, city

FROM Employees
WHERE city = @city
AND title = @title..

GO

lastname firstname title city

-------------------- ---------- ------------------------------ ---------------
Buchanan Steven Sales Manager London
NewFamily Michael Sales Representative London
King Robert Sales Representative London
Dodsworth Anne Sales Representative London

(4 row(s) affected)
However, as stated before, you can build just one query using the ISNULL function to validate each variable,
and one expression per variable; thus, solving the problem of the optional search fields with just a query and

124
 Chapter 4. Querying and Modifying Data

not using any IF statements. This query appears in Listing 4.17 (notice that it has the same output as the
preceding example).

Listing 4.17 Using Optional Search Fields Using One Query

USE Northwind
DECLARE @title VARCHAR(60), @city VARCHAR(30)

-- Setting @title to NULL and serarching for all employees

-- who live in London

SET @title = NULL

SET @city = 'London'

SELECT lastname, firstname, title, city

FROM Employees
WHERE city = ISNULL(@city,city)
AND title = ISNULL(@title,title) .
GO

lastname firstname title city

-------------------- ---------- ------------------------------ ---------------
Buchanan Steven Sales Manager London
NewFamily Michael Sales Representative London
King Robert Sales Representative London
Dodsworth Anne Sales Representative London.

(4 row(s) affected)
Caution

Be aware that IS NULL is different from the ISNULL function. The IS NULL clause is used to
make comparisons with NULL values, whereas ISNULL is a function that takes two arguments. If
the first one is NULL, it returns the second one; otherwise, the first argument is returned.

Data Aggregation and the GROUP BY Clause

125
Microsoft SQL Server 2000 Programming by Example

One of the benefits of the SQL language is that it enables you to generate summaries of the data stored in a
database. Sometimes, data as a whole might not make sense, but when summarized, it can be used for many
purposes.
Transact-SQL provides aggregate functions, which are used to generate summary values. Basically, they
return a single value based on a calculation on a set of values. Table 4.1 shows the most common aggregate
functions used in Transact-SQL.
Table 4.1. Transact-SQL's Aggregate Functions
Aggregate Description
Function
AVG Returns the average or arithmetic mean.
COUNT Returns the number of values (an INT data type). COUNT(*) can be used to return the
number of rows in a group (the group can be the whole table, obtaining the number of rows in
the table).
COUNT_BIG Similar to COUNT in that it returns the number of values, but it returns a BIGINT data type.
BIG_COUNT(*) can be used to return the number of rows in a group (the group can be the
whole table, obtaining the number of rows in the table).
MAX Returns the maximum value.
MIN Returns the minimum value.
SUM Returns the sum of the values. Takes only numbers as arguments.
Listing 4.18 indicates how these aggregate functions are used to return summary values based on the
values on the whole table.

Listing 4.18 Using Aggregate Functions

USE Northwind

-- Returns the average of unitsinstock

SELECT AVG(unitsinstock)
FROM Products

-- Returns the number of rows in the Employees table

SELECT COUNT(*)
FROM Employees

-- Returns the price of the most expensive product

SELECT MAX(unitprice)
FROM Products

-- Returns the birthdate of the oldest employee

SELECT MIN(birthdate)
FROM Employees

-- Returns the number of products in stock

SELECT SUM(unitsinstock)
FROM Products
GO

126
 Chapter 4. Querying and Modifying Data

-----------
40
-----------
9

---------------------
263.5000

--------------------------------
1937-09-19 00:00:00.000:

-----------
3119
The DISTINCT keyword can be used in any aggregate function to consider repeating values just once. For
example, to retrieve how many different titles the Employees table has, you can use the COUNT aggregate
function with the DISTINCT keyword, as shown in Listing 4.19. In this case, the DISTINCT keyword is
needed because more than one employee has the same title, and you want to count each title once to see
how many different titles are in this table.

Listing 4.19 Using DISTINCT in Aggregate Functions

USE Northwind

SELECT COUNT(DISTINCT title)

FROM Employees
GO

-----------
4
You use the GROUP BY clause to group rows in a result set, generating a summary row for each group of data.
All columns specified in SELECT must also be specified in GROUP BY. However, columns specified in the
GROUP BY clause don't have to be in the SELECT column list.

127
Microsoft SQL Server 2000 Programming by Example

To illustrate, Listing 4.20 shows an example that retrieves the number of employees per title. SQL Server
generates a row per each title (this is the column specified in the GROUP BY clause) and counts the number of
rows per title.

Listing 4.20 Using the GROUP BY Clause

USE Northwind

SELECT title, COUNT(*)

FROM Employees
GROUP BY title
GO
title
------------------------------ -----------
Inside Sales Coordinator 1
Sales Manager 1
Sales Representative 6
Vice President, Sales 1

(4 row(s) affected)
It might be necessary to generate a summary row for a table (just one row and not a row for each group). In
this case, because it is just one group (the whole table), use aggregate functions without the GROUP BY
clause, as previously shown in Listing 4.18. Moreover, you can use more than one aggregate function in the
same query. For example, to get the most recent date in which an order was placed, and the minimum orderid
in the Orders table, use the query shown in Listing 4.21.

Listing 4.21 Summarizing Data

USE Northwind

SELECT MAX(orderdate), MIN(orderid)

FROM orders
GO
--------------------------- -----------

128
 Chapter 4. Querying and Modifying Data

1998-05-06 00:00:00.000 10248

(1 row(s) affected)
If there's a WHERE clause in the query, it must be specified before the GROUP BY clause. SQL Server
evaluates the WHERE clause first, and then it generates the groups based on the columns specified in GROUP
BY. For example, to retrieve the number of customers in Spain and Venezuela, use the query shown in
Listing 4.22.

Listing 4.22 Restricting the Groups Generated by GROUP BY

USE Northwind

SELECT country, COUNT(*)

FROM Customers
WHERE country IN ('Spain','Venezuela')
GROUP BY country
GO

country
--------------- -----------
Spain 5
Venezuela 4

(2 row(s) affected)
Tip

As a new feature of SQL Server 2000, BIT columns can be used in a GROUP BY clause. This was
a limitation of GROUP BY in previous versions.

129
Microsoft SQL Server 2000 Programming by Example

The use of column aliases is recommended when working with aggregate functions, because when any
function is applied to a column, the result set doesn't show the original name of the column. Listing 4.23
shows an example of column aliases when using aggregate functions.

Listing 4.23 Using Column Aliases and Aggregate Functions

USE Northwind

SELECT country, COUNT(*) AS [number of customers]

FROM Customers
WHERE country IN ('Spain','Venezuela')
GROUP BY country
GO

country number of customers

--------------- -------------------
Spain 5
Venezuela 4

(2 row(s) affected)

The HAVING Clause

When using GROUP BY in a query to generate groups, you might want to set restrictions on these groups.
Specifically, the HAVING clause sets restrictions on the groups generated by GROUP BY. HAVING is similar to
WHERE in the sense that it restricts the output of the query, but HAVING is evaluated by SQL Server after the
groups are generated.
It's important to know that WHERE is evaluated first, then groups are generated (as a result of GROUP BY), and
finally, the HAVING clause is evaluated. Therefore, aggregate functions cannot be referenced in the WHERE
clause; they can be referenced only in the HAVING clause.
Listing 4.24 retrieves the number of customers of the countries that have more than five customers. This is
done by setting a restriction after the groups are generated (using a HAVING clause); hence, showing only the
countries that have more than five customers.

Listing 4.24 Setting Restrictions on the Groups Generated by GROUP BY Using HAVING

130
 Chapter 4. Querying and Modifying Data

USE Northwind

SELECT country, COUNT(*) AS [number of customers]

FROM Customers
GROUP BY country
HAVING COUNT(*) > 5
GO

country number of customers

--------------- -------------------
Brazil 9
France 11
Germany 11
UK 7
USA 16

(5 row(s) affected)
Similar to WHERE, multiple conditions can be specified in the HAVING clause, combining them with a logical
operator (OR or AND). Listing 4.25 shows how conditions can be combined in a HAVING clause.

Listing 4.25 Combining Conditions in a HAVING Clause

USE Northwind

SELECT country, COUNT(*) AS [number of customers]

FROM Customers
GROUP BY country
HAVING COUNT(*) > 5
AND COUNT(*) < 10
GO

131
Microsoft SQL Server 2000 Programming by Example

country number of customers

--------------- -------------------
Brazil 9
UK 7

(2 row(s) affected)

The ORDER BY Clause

A table comprises a set of rows, and a set, by definition, is unordered. Therefore, when retrieving data from
tables, SQL Server doesn't guarantee the order of the rows in the result set. This is because SQL Server
might optimize the query in a different way each time it is executed, depending on the data; resulting in a
different order of the rows each time the same query is executed. To guarantee a specific order in a result set,
use the ORDER BY clause. Listing 4.26 retrieves information from the Shippers table ordered by company
name in ascending order (this is the default in SQL Server).

Listing 4.26 Using ORDER BY to Guarantee the Order of Rows

USE Northwind

SELECT companyname, phone

FROM Shippers
ORDER BY companyname
GO

companyname phone
---------------------------------------- ------------------------
Federal Shipping (503) 555-9931
Speedy Express (503) 555-9831
United Package (503) 555-3199

(3 row(s) affected)

132
 Chapter 4. Querying and Modifying Data

You can include more than one column in the ORDER BY clause, and you also can specify how these values
will be sorted, either ascending (using the ASC keyword), which is the default, or descending (using the DESC
keyword). If more than one column is specified in the ORDER BY clause, SQL Server sorts the result set in the
order in which these columns appear (first, the first column, then the second column, and so on). Listing
4.27 shows how to specify multiple columns and how to order them (either ascending or descending) in the
ORDER BY clause.

Listing 4.27 Using Multiple Expressions in the ORDER BY Clause

USE Northwind

SELECT lastname, firstname

FROM Employees
ORDER BY lastname ASC, firstname DESC
GO

lastname firstname
-------------------- ----------
Buchanan Steven
Callahan Laura
Davolio Nancy
Dodsworth Anne
Fuller Andrew
King Robert
Leverling Janet
NewFamily Michael
Peacock Margaret

(9 row(s) affected)
Tip

As discussed in previous chapters, use TOP if you want to specify the ORDER BY clause when
creating a view.

The TOP N Clause

133
Microsoft SQL Server 2000 Programming by Example

TOP is used to limit the results of a query. It can be used in two ways: to retrieve the first N rows or to retrieve
the first N percent of the rows in the result set. The TOP clause must be used along with ORDER BY; otherwise,
SQL Server doesn't guarantee a specific ordering, and the TOP clause will be meaningless.
TOP returns the least significant values if they are sorted in ascending order. On the other hand, TOP retrieves
the most significant values if they are sorted in descending order. For example, to retrieve the most expensive
products, use a TOP clause and an ORDER BY clause sorting the unitprice column in descending order, as
shown in Listing 4.28.

Listing 4.28 Limiting the Output of a Query Using the TOP Clause

USE Northwind

SELECT TOP 10 productid, productname, unitprice

FROM Products
ORDER BY unitprice DESC

SELECT TOP 1 PERCENT productid, productname, unitprice

FROM Products
ORDER BY unitprice DESC
GO

productid productname unitprice

----------- ---------------------------------------- ------------
38 Côte de Blaye 263.5000
29 Thüringer Rostbratwurst 123.7900
9 Mishi Kobe Niku 97.0000
20 Sir Rodney's Marmalade 81.0000
18 Carnarvon Tigers 62.5000
59 Raclette Courdavault 55.0000
51 Manjimup Dried Apples 53.0000
62 Tarte au sucre 49.3000
43 Ipoh Coffee 46.0000
28 Rössle Sauerkraut 45.6000
(10 row(s) affected)

productid productname unitprice

----------- ---------------------------------------- -----------
38 Côte de Blaye 263.5000

(1 row(s) affected)
Caution

134
 Chapter 4. Querying and Modifying Data

If you're concerned about portability, be careful when using TOP because it is not ANSI standard.
Instead, it is a feature of Transact-SQL.

The argument of TOP is a positive integer in either case (percent or fixed number of rows).

Caution

The argument of the TOP clause must be an integer; it cannot be a variable. If you want to use a
variable, use dynamic queries (EXEC or sp_executesql).

In previous versions of SQL Server (6.5 and earlier), the only way to limit the result set of a query was by
using SET ROWCOUNT, which stops the processing of the query when it reaches the number of rows specified
by SET ROWCOUNT.

Be aware that TOP is more efficient than SET ROWCOUNT because TOP is evaluated at parse time, not at
execution time like SET ROWCOUNT. Another disadvantage of using SET ROWCOUNT is that it remains set
until you execute SET ROWCOUNT 0 to reset it to its original behavior (all rows are returned when executing a
query). When SET ROWCOUNT is enabled, it also affects modification operations (INSERT, UPDATE, and
DELETE). Listing 4.29 demonstrates the usage of SET ROWCOUNT (notice that the result set is equivalent to
the one shown in Listing 4.28).

Listing 4.29 Using SET ROWCOUNT

USE Northwind

-- Use SET ROWCOUNT 10 to limit the output of all queries to 10 rows

SET ROWCOUNT 10

SELECT productid, productname, unitprice

FROM Products
ORDER BY unitprice DESC

-- Use SET ROWCOUNT 0 to reset it to its original state (all rows are returned)
SET ROWCOUNT 0
GO

135
Microsoft SQL Server 2000 Programming by Example

productid productname unitprice

----------- ---------------------------------------- ----------------
38 Côte de Blaye 263.5000
29 Thüringer Rostbratwurst 123.7900
9 Mishi Kobe Niku 97.0000
20 Sir Rodney's Marmalade 81.0000
18 Carnarvon Tigers 62.5000
59 Raclette Courdavault 55.0000
51 Manjimup Dried Apples 53.0000
62 Tarte au sucre 49.3000
43 Ipoh Coffee 46.0000
28 Rössle Sauerkraut 45.6000

(10 row(s) affected)

Caution

If you use SET ROWCOUNT, don't forget to execute SET ROWCOUNT 0 to turn this setting off;
otherwise, it remains set during the connection, affecting all subsequent queries.

Use the WITH TIES keyword of the TOP clause when you want to include ties in the result set. If WITH TIES
is specified, the result set may contain more rows than the number of rows specified in the TOP clause
because all ties would be included. For example, Listing 4.30 shows a query that retrieves the top six units
in stock. Notice that seven rows are returned because there's a tie in the sixth position, and the query returns
all ties (two in this case).

Listing 4.30 Using WITH TIES in TOP Clauses

USE Northwind

SELECT TOP 6 WITH TIES productid, productname, unitsinstock

FROM Products
ORDER BY unitsinstock DESC
GO

136
 Chapter 4. Querying and Modifying Data

productid productname unitsinstock

----------- ---------------------------------------- ------------
75 Rhönbräu Klosterbier 125
40 Boston Crab Meat 123
6 Grandma's Boysenberry Spread 120
55 Pâté chinois 115
61 Sirop d'érable 113
33 Geitost 112
36 Inlagd Sill 112

(7 row(s) affected)

Using Dynamic Queries

In some situations, you might want to parameterize queries using variables to specify, for example, the table
to query. However, some elements cannot be specified dynamically in queries, such as the table name and
column names. In these specific cases, dynamic queries might be beneficial. Specifically, there are two ways
to execute dynamic queries: using EXEC (or EXECUTE), and using the sp_executesql system stored
procedure. These two ways are listed in Listing 4.31.
Caution

The string (a dynamic query) that is passed as an argument to sp_executesql must be a

Unicode string (to specify Unicode strings, use the N prefix when building the string).

Listing 4.31 Dynamically Generating and Executing Queries Using EXEC and sp_executesql

USE Northwind

DECLARE @tablename VARCHAR(20), @query NVARCHAR(100)

SET @tablename = 'Shippers'
SET @query = N'SELECT * FROM '+ @tablename

-- Executing the dynamic query using EXEC

EXEC (@query)

-- Executing the dynamic query using sp_executesql

137
Microsoft SQL Server 2000 Programming by Example

EXEC sp_executesql @query

GO

ShipperID CompanyName Phone

----------- ---------------------------------------- ------------------------
1 Speedy Express (503) 555-9831
2 United Package (503) 555-3199
3 Federal Shipping (503) 555-9931

ShipperID CompanyName Phone

----------- ---------------------------------------- ------------------------
1 Speedy Express (503) 555-9831
2 United Package (503) 555-3199
3 Federal Shipping (503) 555-9931

(3 row(s) affected)
The following are the disadvantages of using dynamic queries:

• The statements inside EXEC or sp_executesql are executed inside its own batch; therefore, these
statements cannot access variables declared in the outside batch.
• If the query to be executed by EXEC is not similar enough to a previously executed query due to
different format, values, or data types, SQL Server cannot reuse a previously executed query plan.
However, sp_executesql overcomes this limitation, allowing SQL Server to reuse the execution
plan of the query (because it can be cached in memory).

Tip

Use sp_executesql whenever possible when executing dynamic queries, because the plan has
a better chance of being reused.

Sometimes the dynamic query is very long and it becomes illegible. In these cases, you can use a variable to
store the entire string and then use this variable as the argument of EXEC or sp_executesql, as shown in
Listing 4.31. Also, you might want to insert carriage returns (using CHAR(13)) in the query to make it more
legible (in case you want to display it). Listing 4.32 indicates how to insert carriage returns in a dynamic
query.

Listing 4.32 Inserting Carriage Returns When Building Dynamic Queries

138
 Chapter 4. Querying and Modifying Data

USE Northwind

DECLARE @query NVARCHAR(100)

SET @query = N'SELECT * '+ CHAR(13)+ 'FROM Shippers'

-- To display the query (which has a carriage return)

SELECT @query

-- Executing the dynamic query

EXEC sp_executesql @query
GO

---------------------------

SELECT * FROM Shippers

(1 row(s) affected)

ShipperID CompanyName Phone

----------- ---------------------------------------- ------------------------
1 Speedy Express (503) 555-9831
2 United Package (503) 555-3199
3 Federal Shipping (503) 555-9931

(3 row(s) affected)
Caution

In SQL Server, EXECUTE can be used for three different purposes: to execute dynamic queries, to
execute stored procedures, and to assign execute permissions to users on stored procedures
(using GRANT, DENY, or REVOKE). The difference between executing a stored procedure and a
dynamic statement using EXECUTE is that the first one doesn't need to be enclosed in parentheses,
whereas the dynamic statement does.

There are some security issues when dynamic statements are executed inside a stored procedure. Usually, to
be able to execute a stored procedure, a user just needs to have EXECUTE permissions on the stored
procedure. However, if a dynamic query is used, the user also needs permissions on every object referenced
by the dynamic query. This is because the dynamic query is not parsed until the stored procedure is executed,
and SQL Server must check permissions on every object referenced by the dynamic query.

139
Microsoft SQL Server 2000 Programming by Example

Modifying Data
As you already know, SELECT is the element of the Data Manipulation Language (DML) that is used to extract
information from tables. The other elements of the DML are used to add, modify, and remove data from tables.
These elements are INSERT, UPDATE, and DELETE.

The INSERT Statement

INSERT is used to add new rows in a table. The following is the basic syntax to insert a row in a table:

INSERT INTO Table_name (column_1,column_2,..,column_n)

VALUES (value_1,value_2,..,value_n)
The order of the values to be inserted must be the same order of the columns specified in the column list. Also,
the INTO keyword can be omitted when using the INSERT statement, as shown in Listing 4.33.

Listing 4.33 Adding a New Row Using the INSERT Statement

USE Northwind

INSERT Territories (territoryid,territorydescription,regionid)

VALUES ('77777','Fort Lauderdale',4)
GO

(1 row(s) affected)
If you want to insert data in all columns of a table, the column list can be omitted, but keep in mind that the
values must be ordered in the same way that their respective columns appear in the table's definition (you can
see the order of the columns using the sp_help system stored procedure).
For example, Listing 4.34 inserts a row in the Territories table, omitting the column list.

Listing 4.34 Omitting the Column List When Inserting Data in All Columns of the Table

USE Northwind

140
 Chapter 4. Querying and Modifying Data

INSERT Territories VALUES ('88888','Miami',4)

GO

(1 row(s) affected)
SQL Server automatically handles IDENTITY columns by default. Therefore, when a row is inserted in a table
that has an IDENTITY column, you don't have to specify the IDENTITY column in the INSERT statement
because SQL Server provides a value automatically, as shown in Listing 4.35.

Listing 4.35 Inserting a Row in a Table with an IDENTITY Column

USE Northwind

INSERT Shippers (companyname, phone)

VALUES ('Super Fast Shipping','(503) 555-6493')
GO

(1 row(s) affected)
However, if you want to explicitly insert a value in an IDENTITY column, use SET IDENTITY_INSERT with
the table's name as an argument. This is demonstrated in Listing 4.36.

Listing 4.36 Inserting a Value in an IDENTITY Column

USE Northwind

141
Microsoft SQL Server 2000 Programming by Example

SET IDENTITY_INSERT Shippers ON

INSERT Shippers (shipperid,companyname, phone)
VALUES (20,'ACME Shipping','(503) 555-8888')
SET IDENTITY_INSERT Shippers OFF
GO

(1 row(s) affected)
There are two ways to insert NULL values in nullable columns: either explicitly (using the NULL keyword when
inserting data) or implicitly (the column is not referenced in the INSERT statement).
Similarly, there are two ways to use default values in INSERT statements: either explicitly (using the DEFAULT
keyword) or implicitly (if the column is not specified in the INSERT statement).
As a result, when columns are omitted in the column list of INSERT, SQL Server automatically provides a
default value (if one is defined on the column) or, if a default value is not defined and the column is nullable, a
NULL is used. On the other hand, if a column doesn't have a default value and it doesn't accept NULLs, a
value must be specified in the INSERT statement; otherwise, the INSERT operation will fail.
Listing 4.37 shows two equivalent INSERT statements. The first one uses the NULL and DEFAULT keywords,
whereas the second one omits these columns, producing the same result.

Listing 4.37 Omitting Specific Columns and Using the NULL and DEFAULT Keywords

USE Northwind

INSERT Products (productname,supplierid,categoryid,quantityperunit,

reorderlevel,discontinued)
VALUES ('Donut',NULL,NULLx,'6 pieces',DEFAULT,DEFAULT)

-- INSERT Products (productname,quantityperunit)

-- VALUES ('Donut','6 pieces')
GO

(1 row(s) affected)
Caution

142
 Chapter 4. Querying and Modifying Data

Keywords, such as NULL or DEFAULT, don't need to be enclosed in single quotation marks like
strings.

Furthermore, if you want to insert default values in all columns and NULL values in the nullable ones without a
default, use the following syntax (which also takes care of IDENTITY values):

INSERT Table_name DEFAULT VALUES

Be aware that to be able to use this syntax, all columns must meet at least one of these conditions:

• It must be an IDENTITY column.

• The column must have a default value defined on it.
• The column must be nullable.

Listing 4.38 shows an example of this syntax.

Listing 4.38 Inserting Default Values in All Columns

USE Northwind

INSERT Orders DEFAULT VALUES

GO

(1 row(s) affected)
INSERT may also be used to insert multiple rows in a table. This can be done through two approaches:

• Using a SELECT statement along with INSERT. In this case, the output of the SELECT statement is
inserted into the table. Listing 4.39 indicates how to insert multiple rows in a table using this
approach.

Listing 4.39 Inserting a SELECT Statement's Output into a Table

143
Microsoft SQL Server 2000 Programming by Example

USE Northwind

CREATE TABLE #employees_in_wa (

lastname NVARCHAR(40),
firstname NVARCHAR(20)
)

-- Inserting into the temporary table the last name

-- and first name of all employees from WA

INSERT #employees_in_wa
SELECT lastname,firstname
FROM Employees
WHERE region = 'WA'

SELECT * FROM #employees_in_wa

GO

(5 row(s) affected)

lastname firstname
---------------------------------------- --------------------
Davolio Nancy
Fuller Andrew
Leverling Janet
Peacock Margaret
Callahan Laura

(5 row(s) affected)

• Executing a stored procedure that has a SELECT statement on it, and inserting this output into a table.
Notice that this approach is similar to the first one; the only difference is that the SELECT statement is
wrapped into a stored procedure. Listing 4.40 shows how the output of a stored procedure is
inserted into a table.

Listing 4.40 Inserting the Output of a Stored Procedure into a Table

144
 Chapter 4. Querying and Modifying Data

USE Northwind
GO

CREATE PROC get_uk_employees

AS
SELECT lastname,firstname
FROM Employees
WHERE country = 'UK'
GO

CREATE TABLE #employees_in_uk (

lastname NVARCHAR(40),
firstname NVARCHAR(20)
)

-- Inserting into the temporary table the last name

-- and first name of all employees from UK

INSERT #employees_in_uk
EXEC get_uk_employees

SELECT * FROM #employees_in_uk

GO

(4 row(s) affected)

lastname firstname
---------------------------------------- --------------------
Buchanan Steven
NewFamily Michael
King Robert
Dodsworth Anne

(4 row(s) affected)

The DELETE Statement

You use DELETE to remove one or more rows permanently from a table. A DELETE statement may contain a
WHERE clause to restrict the rows to be deleted. The basic syntax of DELETE is

DELETE Table_name WHERE condition

If WHERE is not used in the DELETE statement, all rows are removed from the table. Listing 4.41 shows how
to delete specific rows from a table (using a WHERE clause).

145
Microsoft SQL Server 2000 Programming by Example

Listing 4.41 Deleting Rows from a Table

USE Northwind

DELETE Orders
WHERE customerid IS NULL
GO

(1 row(s) affected)
The TRUNCATE statement is also used to remove permanently all rows from a table. However, it has some
restrictions:

• The table cannot have foreign keys defined.

• TRUNCATE cannot contain a WHERE clause. Therefore, all rows in the table are removed.
• TRUNCATE reseeds the IDENTITY value of the table (if there is one).

Tip

TRUNCATE is faster than DELETE because SQL Server only logs the deallocation of pages, not the
removal of each row (like it does when dealing with DELETE statements).

Listing 4.42 illustrates the use of the TRUNCATE statement to remove all rows from a table.

Listing 4.42 Using the TRUNCATE Statement to Delete All Rows from a Table

USE Northwind
GO

146
 Chapter 4. Querying and Modifying Data

CREATE TABLE #shippers (

companyname NVARCHAR(20),
phone NVARCHAR(20)
)

INSERT #shippers
SELECT companyname,phone FROM Shippers

-- Using TRUNCATE to remove all rows from the #shippers table

TRUNCATE TABLE #shippers

SELECT * FROM #shippers

GO

(5 row(s) affected)

companyname phone
-------------------- --------------------

(0 row(s) affected)

The UPDATE Statement

The UPDATE statement sets new values in existing rows of a specific table. UPDATE modifies information in
just one table. Therefore, if you want to change data in some other table, use another UPDATE statement. The
basic syntax of UPDATE is

UPDATE Table_name
SET column_1 = new_value,
column_2 = new value,
.
.
column_n = new_value
WHERE condition
The new value of a column can be either a constant or an expression that may or may not contain the
previous value of the column.
A WHERE clause can be used to restrict the rows to be modified by the UPDATE statement.
Listing 4.43 shows an UPDATE statement that restricts the column to be modified with a WHERE clause. Also,
the new value of one of the columns, companyname, is based in the old value (concatenating the word
'Express' to the old value).

Listing 4.43 Using the UPDATE Statement

147
Microsoft SQL Server 2000 Programming by Example

USE Northwind

UPDATE Shippers
SET companyname = companyname + 'Express',
phone = '(305) 555 8888'
WHERE shipperid = 20
GO

(1 row(s) affected)
Using an UPDATE statement, the new values of columns can be stored in local variables when updating a
single row. This method is useful because you don't have to update the row first, and then use a SELECT
statement to get the new values. The following is the basic syntax:

UPDATE table
SET @variable = column = value
Listing 4.44 indicates how to store the new value of a column in a local variable.

Listing 4.44 Storing in Variables the New Values of Columns When Updating a Single Row

USE Northwind

DECLARE @availableunits SMALLINT

UPDATE Products
SET @availableunits = unitsinstock = unitsinstock + 20
WHERE productname = 'Chai'

SELECT @availableunits
GO
(1 row(s) affected)

------
59

148
 Chapter 4. Querying and Modifying Data

(1 row(s) affected)

The SELECT INTO Statement

SELECT INTO enables you to create a table on-the-fly and populate it using just one instruction. The new
table is populated with the output of the SELECT statement. SELECT INTO can be used to create either
permanent or temporary tables. Listing 4.45 shows the syntax of SELECT INTO.

Listing 4.45 Using SELECT INTO to Create and Populate Tables

USE Northwind

SELECT lastname, firstname

INTO #salesrep_employees
FROM employees
WHERE title = 'sales representative'

SELECT * FROM #salesrep_employees

GO
(6 row(s) affected)

lastname firstname
-------------------- ----------
Davolio Nancy
Leverling Janet
Peacock Margaret
NewFamily Michael
King Robert
Dodsworth Anne

(6 row(s) affected)

149
Microsoft SQL Server 2000 Programming by Example

Column alias es must be used for calculated columns. These aliases are the column names that SQL Server
will use when creating the new table specified in SELECT INTO. For example, Listing 4.46 uses an alias for
the first column, which is the result of the concatenation of two columns (firstname and lastname).

Listing 4.46 Using Column Aliases with SELECT INTO

USE Northwind

SELECT firstname + ''+ lastname AS fullname, country

INTO #employeescountry
FROM Employees
ORDER BY fullname

SELECT * FROM #employeescountry

GO

(9 row(s) affected)

fullname country
------------------------------- ---------------
Andrew Fuller USA
Anne Dodsworth UK
Janet Leverling USA
Laura Callahan USA
Margaret Peacock USA
Michael Suyama UK
Nancy Davolio USA
Robert King UK
Steven Buchanan UK

(9 row(s) affected)
The IDENTITY function is used to generate a column with consecutive numbers when using SELECT INTO.
Similar to the IDENTITY property, the IDENTITY function accepts three parameters: the data type, the seed
(or first number), and the increment (the last two are the arguments of the IDENTITY property). Listing 4.47
demonstrates how the IDENTITY function is used in SELECT INTO statements.

Listing 4.47 Using the IDENTITY function

150
 Chapter 4. Querying and Modifying Data

USE Northwind

SELECT IDENTITY(INT,1,1) as companyid, companyname

INTO #italiancompanies
FROM Customers
WHERE country = 'Italy'

SELECT * FROM #italiancompanies

GO

(3 row(s) affected)

companyid companyname
----------- ----------------------------------------
1 Franchi S.p.A.
2 Magazzini Alimentari Riuniti
3 Reggiani Caseifici

(3 row(s) affected)
In previous versions of SQL Server (7.0 and earlier), the SELECT INTO/ BULKCOPY database option had to
be set to TRUE if you wanted to use SELECT INTO to create permanent tables. In SQL Server 2000, the
SELECT INTO/ BULKCOPY and the TRUNC. LOG ON CHKPT. database options are no longer used. Now,
SQL Server provides three recovery models (SIMPLE, BULK LOGGED, and FULL), and basically, SELECT
INTO can be used with any of these models. For more information on recovery models, refer to Books Online.

What's Next?

You already know how to interact with single tables and extract data from them. In the next chapter, you will
learn how to extract data from multiple tables through JOINs, the different type of JOINs, and how to combine
the results of more than one query using the UNION operator.

151
 Chapter 5. Querying Multiple Tables: JOIN s

Chapter 5. Querying Multiple Tables: JOIN s

In previous chapters, you have been dealing with queries that involve just one table. Sometimes the data you
need to manipulate is spread across more than one table and, in this case, these tables must be combined or
joined to be able to retrieve all this data. Basically, a JOIN operation merges two or more tables into one
result set.
The capability to link or join tables and generate one result set from the data stored in many tables is one of
the most important characteristics of a relational database. Usually, tables are linked using foreign keys, and
these foreign key columns are used in JOIN operationsto combine tables and generate one result set. Notice
that tables don't necessarily need to have a foreign key defined to be able to join them.
Additionally, not only can JOIN be used in SELECT statements, but also in modification operations, as
UPDATE and DELETE. A DELETE operation can be based on information from more than one table if they are
joined in the FROM clause. The same rule applies for UPDATE operations.
This chapter teaches you the following:

• How to use the ANSI SQL-92 JOIN syntax

• How the different types of joins (inner joins, outer joins, cross joins, and self joins) work, and the
differences among them
• How to combine result sets from more than one query using the UNION operator

ANSI SQL-92 Syntax

In earlier versions of theSQL language (ANSI SQL-89), a JOIN operation was specified in the FROM clause
and the WHERE clause. Specifically, the tables involved in the JOIN operation were specified in the FROM
clause, and the JOIN conditions in the WHERE clause. For example, Listing 5.1 indicates how this old syntax
is used to specify an inner join. (This type of join will be explained in the next section.) As you can see in this
example, the two tables involved are listed in the FROM clauseseparated by a comma, and the columns that
link these two tables are specified in the WHERE clause (the JOIN condition).

Listing 5.1 Using the ANSI SQL-89 Syntax in Inner Joins

USE Northwind

SELECT *
FROM Products, Categories
WHERE Products.categoryid = Categories.categoryid
GO
In regard to outer joins, the *= and =* operators were used in the WHERE clause to specify LEFT OUTER
JOIN and RIGHT OUTER JOIN, respectively. This syntax appears in Listing 5.2, which performs a LEFT
OUTER JOINbetween Territories and Region.

Listing 5.2 Using the ANSI SQL-89 Syntax in Outer Joins

153
Microsoft SQL Server 2000 Programming by Example

USE Northwind

SELECT *
FROM Territories, Region
WHERE territories.regionid *= region.regionid
GO
Listing 5.3 shows a slight variation of the query shown previously in Listing 5.2. Notice that the WHERE
clause contains both the JOIN condition and an additional condition that forces SQL Server to show only rows
in which regionid equals 1.

Listing 5.3 Using the ANSI SQL-89 OUTER JOIN Syntax Along with Conditions

USE Northwind

SELECT *
FROM Territories, Region
WHERE territories.regionid *= region.regionid
AND region.regionid = 1
GO
As you might have noticed, using the old ANSI SQL-89's JOIN syntax, SQL Server may interpret a query in
an unexpected way, getting a totally different result set. This is because of the fact that SQL Server must
evaluate JOIN conditions and restrictions of the query together, because both elements are specified in the
WHERE clause. This issue was solved by the ANSI committee in the SQL-92 standard, which states that all
elements of a JOIN operation (tables involved in the JOIN and conditions of the JOIN) are specified in the
FROM clause, thus eliminating any ambiguity.
The ANSI SQL-92 standard introduced, among other things, five JOIN keywords: INNER JOIN, LEFT
OUTER JOIN, RIGHT OUTER JOIN, FULL OUTER JOIN, and CROSS JOIN. These keywords are used to
specify the type of join between two tables.
Listing 5.4 illustrates this new syntax for inner joins. Notice that this query is similar to the one shown in
Listing 5.1, but this uses the new JOIN syntax.

Listing 5.4 Using the ANSI SQL-92 JOIN Syntax

154
 Chapter 5. Querying Multiple Tables: JOIN s

USE Northwind

SELECT *
FROM Products INNER JOIN Categories
ON Products.categoryid = Categories.categoryid
GO
Through this new approach, SQL Server evaluates JOIN conditions first (which are specified in the FROM
clause), and then restrictions of the query (specified in the WHERE clause). In particular, the order of
processing of conditions in a join using the SQL ANSI-92 syntax is first, the conditions in the FROM clause are
evaluated, then the WHERE clause, and finally, the HAVING clause, if there is one. This process is used by
SQL Server to parse joins. However, SQL Server does not necessarily execute the FROM clause before the
WHERE clause.
Listing 5.5 shows a JOIN operation using the SQL ANSI-92 syntax, along with a restriction in the WHERE
clause.

Listing 5.5 Using Restrictions Along with a JOIN Operation Using the New Syntax

USE Northwind

SELECT *
FROM Territories LEFT OUTER JOIN Region
ON territories.regionid = region.regionid
WHERE region.regionid = 1
GO
The new JOIN syntax has a very powerful feature, which basically allows you to specify a condition of the
query in the FROM clause along with the JOIN condition. This is useful in cases in which we want this
condition to be evaluated before the JOIN operation, taking advantage of the order of processing of the query.
Listing 5.6 shows a query similar to the one illustrated in Listing 5.5, but different in that because a
condition is specified in the FROM clause, it is evaluated before the JOIN operation.

Listing 5.6 Using Restrictions in the FROM Clause

USE Northwind

SELECT *
FROM Territories LEFT OUTER JOIN Region
ON territories.regionid = region.regionid
AND region.regionid = 1
GO
Queries are easier to read using the new JOIN syntax because all components of the JOIN operation are
specified in the FROM clause. Also, the conditions of the query are specified in the WHERE clause, unless you

155
Microsoft SQL Server 2000 Programming by Example

want the condition to be evaluated before the JOIN operation. In that case, the condition must be specified in
the FROM clause, as shown in the previous listing.
Although SQL Server 2000 supports both JOIN syntaxes (ANSI SQL-89 and ANS I SQL-92), you should
change all queries to the new ANSI SQL-92 syntax because, in future releases, the SQL Server development
team might decide not to support the old syntax. Therefore, all applications you develop should use the new
ANSI SQL-92 JOIN syntax to avoid incompatibility problems in future versions of SQL Server.

INNER JOIN

In general, a JOIN operationcombines two or more tables, generating one result set from the information
stored in such tables. These tables should have similar columns, commonly foreign keys, which are the ones
used in JOIN operations to link tables. Also, as you might have noticed in previous examples, the columns
involved in a JOIN condition don't need to have the same name.
An INNER JOIN operation between two tables returns all common rows in these two tables. Specifically,
INNER JOIN evaluates the JOIN condition for each row in both tables and if this condition is met, the row is
included in the result set. For example, if you want to retrieve information about products and the name of the
supplier of each product, the Products table and the Suppliers table must be joined (through an INNER
JOIN), thus generating a result set with all the information needed (products and suppliers). Listing 5.7
shows the query that retrieves eachproduct and its supplier.

Listing 5.7 Using INNER JOIN to Retrieve Information from Two Tables

USE Northwind

SELECT productid, productname, companyname

FROM Products INNER JOIN Suppliers
ON Products.supplierid = Suppliers.supplierid
GO

-- Partial result

productid productname companyname

--------- ------------------------------- ----------------------------------
1 Chai Exotic Liquids
2 Chang Exotic Liquids
3 Aniseed Syrup Exotic Liquids
4 Chef Anton's Cajun Seasoning New Orleans Cajun Delights
5 Chef Anton's Gumbo Mix New Orleans Cajun Delights
6 Grandma's Boysenberry Spread Grandma Kelly's Homestead
7 Uncle Bob's Organic Dried Pears Grandma Kelly's Homestead
8 Northwoods Cranberry Sauce Grandma Kelly's Homestead

156
 Chapter 5. Querying Multiple Tables: JOIN s

9 Mishi Kobe Niku Tokyo Traders

10 Ikura Tokyo Traders
11 Queso Cabrales Cooperativa de Quesos 'Las Cabras'
12 Queso Manchego La Pastora Cooperativa de Quesos 'Las Cabras'
13 Konbu Mayumi's
14 Tofu Mayumi's
15 Genen Shouyu Mayumi's
Figure 5.1 shows a graphical representation of the INNER JOIN performed in Listing 5.7. In this figure, you
can see how an INNER JOIN is processed: For every row in the first table, SQL Server goes through the
second table trying to find a corresponding row based on the join column (supplierid in this case), and if a
row matches,it is returned in the result set.

Figure 5.1. Processing INNER JOIN operations.

Caution

Be aware that columns with NULL values don't match any values, because NULL is not equal to
anything. In particular, NULL is not equal to NULL.

Tip

To specify an INNER JOIN operation, you can use either JOIN or INNER JOIN (they're
equivalent).

157
Microsoft SQL Server 2000 Programming by Example

The columns specifiedin a JOIN condition don't necessarily need to have the same data type but, at least,
they have to be compatible. Basically, compatible means one of the following two things:

• Both columns have the same data type.

• If the columns have different data types, the data type of one column can be implicitly converted to the
data type of the other column.

For example, when two tables are joined and the JOIN condition has two columns with different data types,
SQL Server tries to perform an implicit conversion; otherwise, CAST or CONVERT must be used to perform an
explicit conversion. An example of this implicit conversion appears in Listing 5.8. Notice that the data types
of the columns in the JOIN condition are different (VARCHAR and INT), thus SQL Server performs an implicit
conversionto process the JOIN.

Listing 5.8 Performing an with Columns of Different Data Types in the JOIN Condition

USE Northwind

CREATE TABLE Parents (

parentid INT IDENTITY(1,1) PRIMARY KEY,
fullname VARCHAR(50),
relationship VARCHAR(50),
employeeid VARCHAR(10))
GO
SET SHOWPLAN_TEXT ON
GO

SELECT lastname, firstname, fullname

FROM employees JOIN Parents
ON employees.employeeid = parents.employeeid
GO

SET SHOWPLAN_TEXT OFF

GO

DROP TABLE Parents

GO

-- Notice that a convert operation is performed in the last line of the

-- execution plan

StmtText

158
 Chapter 5. Querying Multiple Tables: JOIN s

------------------------------------------------

SELECT lastname, firstname, fullname

FROM employees JOIN Parents
ON employees.employeeid = parents.employeeid

(1 row(s) affected)

StmtText
-------------------------------------------------------------------------
|--Nested Loops(Inner Join, OUTER REFERENCES:([Parents].[employeeid]))
|--Clustered Index Scan(OBJECT:([Northwind].[dbo].[Parents].
[PK__Parents__1F98B2C1]))
|--Clustered Index Seek(OBJECT:([Northwind].[dbo].[Employees].
[PK_Employees]),
SEEK:([Employees].[EmployeeID]=Convert([Parents].[employeeid]))
ORDERED FORWARD)

(3 row(s) affected)
The column list of a query that joins tables can reference any of the columns in any of these tables. There are
many different ways to show columns in the result set of a JOIN operation. Be aware that when more than
one table has a column with the same name, you must qualify the name of the column with the table's name—
for example, tablename.columnname. If a column name doesn't have a duplicate in any of the other joined
tables, it doesn't have to be qualified with the table name or alias. Listing 5.9 demonstrates how columns
with the same name are included in the result set of a JOIN.

Listing 5.9 Qualifying the Name of Columns in the Column List

USE Northwind

-- Notice that both tables that are being joined contain the regionid column
-- (this is the only column that has to be fully qualified in the query)

SELECT Region.regionid, territorydescription, regiondescription

FROM Territories JOIN Region
ON Territories.regionid = Region.regionid
ORDER BY Region.regionid
GO

-- Partial result

regionid territorydescription regiondescription

----------- ------------------------- -------------------------

159
Microsoft SQL Server 2000 Programming by Example

1 Westboro Eastern
1 Bedford Eastern
1 Georgetown Eastern
1 Boston Eastern
1 Cambridge Eastern
1 Braintree Eastern
1 Providence Eastern
1 Wilton Eastern
1 Morristown Eastern
1 Edison Eastern
1 New York Eastern
1 New York Eastern
1 Mellvile Eastern
1 Fairport Eastern
If you want to referenceall the columns in a table, this syntax can be used: tablename.*. If you specify only
* in the column list, all columns from all tables involved in the query will be returned. These two approaches
are shown in Listing 5.10.

Listing 5.10 Specifying All Columns from a Table in a Column List Using the * Keyword

USE Northwind

SELECT Territories.*
FROM Territories JOIN Region
ON Territories.regionid = Region.regionid

SELECT *
FROM Territories JOIN Region
ON Territories.regionid = Region.regionid
GO

-- Partial results

territoryid territorydescription regionid

--------------- ------------------------- -----------
01581 Westboro 1
01730 Bedford 1
01833 Georgetown 1
02116 Boston 1
02139 Cambridge 1
02184 Braintree 1
02903 Providence 1
03049 Hollis 3

160
 Chapter 5. Querying Multiple Tables: JOIN s

03801 Portsmouth 3
06897 Wilton 1
07960 Morristown 1
08837 Edison 1

TerritoryID TerritoryDescription RegionID RegionID RegionDescription

------------ --------------------- --------- --------- -----------------
01581 Westboro 1 1 Eastern
01730 Bedford 1 1 Eastern
01833 Georgetown 1 1 Eastern
02116 Boston 1 1 Eastern
02139 Cambridge 1 1 Eastern
02184 Braintree 1 1 Eastern
02903 Providence 1 1 Eastern
03049 Hollis 3 3 Northern
03801 Portsmouth 3 3 Northern
06897 Wilton 1 1 Eastern
07960 Morristown 1 1 Eastern
08837 Edison 1 1 Eastern
Table aliases can be used when referencingtables in JOIN operations to make queries easier to read.
However, make sure that if an alias is specified, every reference to the table uses the alias; otherwise (if the
name of the table is used and an alias was specified), you will get a syntax error. Listing 5.11 illustrates the
use of tablealiases in JOIN operations.

Listing 5.11 Using Table Aliases in JOIN Operations

USE Northwind

-- Notice that the aliases of the tables are used

-- in the column list and in the JOIN condition
SELECT P.productname, C.categoryname
FROM Products P JOIN Categories C
ON P.categoryid = C.categoryid
GO

-- Partial result

productname categoryname
---------------------------------------- ---------------
Chai Beverages
Chang Beverages
Aniseed Syrup Condiments
Chef Anton's Cajun Seasoning Condiments

161
Microsoft SQL Server 2000 Programming by Example

Chef Anton's Gumbo Mix Condiments

Grandma's Boysenberry Spread Condiments
Uncle Bob's Organic Dried Pears Produce
Northwoods Cranberry Sauce Condiments
Mishi Kobe Niku Meat/Poultry
Ikura Seafood
Queso Cabrales Dairy Products
Queso Manchego La Pastora Dairy Products
Tip

In general, the performance of a JOIN operation can be improved if the columns involved in the
JOIN condition are indexed.

A query can involve more thanone JOIN operation; therefore, more than two tables can be joined in a query to
generate one result set. In particular, to join more than two tables, specify two tables first, and then specify the
remaining tables one by one along with the JOIN condition. As stated earlier, not all columns from all tables
have to be specified in the column list; you just have to specify the ones you need.

For example, if you want to know all regions associated with employees in the Northwind database, you must
retrieve the territory for each employee first, and then retrieve the region of each territory. This is
accomplished by joining the Employees, Employeeterritories, Territories, and Region tables, as
shown in Listing 5.12.

Listing 5.12 Joining More Than Two Tables

USE Northwind

SELECT firstname, lastname, territorydescription, regiondescription

FROM Employees E JOIN Employeeterritories ET
ON E.employeeid = ET.employeeid
JOIN Territories T
ON ET.territoryid = T.territoryid
JOIN Region R
ON T.regionid = R.regionid
GO

162
 Chapter 5. Querying Multiple Tables: JOIN s

-- Partial result

firstname lastname territorydescription regiondescription

---------- -------------------- ----------------------- --------------------
Nancy Davolio Wilton Eastern
Nancy Davolio Neward Eastern
Andrew Fuller Westboro Eastern
Andrew Fuller Bedford Eastern
Andrew Fuller Georgetown Eastern
Andrew Fuller Boston Eastern
Andrew Fuller Cambridge Eastern
Andrew Fuller Braintree Eastern
Andrew Fuller Louisville Eastern
Janet Leverling Atlanta Southern
Janet Leverling Savannah Southern
Janet Leverling Orlando Southern
Janet Leverling Tampa Southern
Margaret Peacock Rockville Eastern
Margaret Peacock Greensboro Eastern
Margaret Peacock Cary Eastern
Steven Buchanan Providence Eastern
Steven Buchanan Morristown Eastern
Steven Buchanan Edison Eastern
Steven Buchanan New York Eastern
Steven Buchanan New York Eastern
Steven Buchanan Mellvile Eastern
Steven Buchanan Fairport Eastern
Internally, a JOIN that involves more than three tables works this way:

1. A result set is generated joining the first two tables.

2. This result set is joined with the third table and so on.
3. The columns specified in the result set are the ones shown in the result set of the JOIN operation.

A JOIN operation can also be used in UPDATE and DELETEstatements. Basically, this allows us to update or
delete rows based on information stored in many tables. For example, suppose that you have to increase the
price of all products of a certain supplier. In this case, you have to update the Products table and join it with
the Suppliers table because the name of the supplier is stored in the Suppliers table and not in the
Products table. Listing 5.13 increases by 5 dollars the unitprice of all products of the 'Exotic
Liquids' supplier.

Listing 5.13 Using JOIN in UPDATE Statements

USE Northwind

SELECT productid, unitprice, companyname

FROM Products P JOIN Suppliers S
ON P.supplierid = S.supplierid
WHERE companyname = 'Exotic Liquids'

UPDATE Products
SET unitprice = unitprice + 5

163
Microsoft SQL Server 2000 Programming by Example

FROM Products P JOIN Suppliers S

ON P.supplierid = S.supplierid
WHERE companyname = 'Exotic Liquids'

SELECT productid, unitprice, companyname

FROM Products P JOIN Suppliers S
ON P.supplierid = S.supplierid
WHERE companyname = 'Exotic Liquids'
GO

productid unitprice companyname

----------- --------------------- ----------------------------------------
1 18.0000 Exotic Liquids
2 19.0000 Exotic Liquids
3 10.0000 Exotic Liquids

(3 row(s) affected)

3 row(s) affected)

productid unitprice companyname

----------- --------------------- ----------------------------------------
1 23.0000 Exotic Liquids
2 24.0000 Exotic Liquids
3 15.0000 Exotic Liquids

(3 row(s) affected)
A JOIN operation can also be used in a DELETE statement in the same way as it is used in an UPDATE
statement.

OUTER JOINs
An OUTER JOIN operation returns all rows that match the JOIN condition, and it may also return some of the
rows that don't match, depending on the type of OUTER JOIN used. There are three types of OUTER JOIN:
RIGHT OUTER JOIN, LEFT OUTER JOIN, and FULL OUTER JOIN.
In INNER JOIN, the order of the tables in the query doesn't matter, whereas in OUTER JOIN, the order of
the tables in the query is important.
Tip

A LEFT OUTER JOIN can be translated into a RIGHT OUTER JOIN, and vice versa if you
change the order of the tables in the join.

An OUTER JOIN can be seen as the result of the union of an INNER JOIN and all unmatched rows in

164
 Chapter 5. Querying Multiple Tables: JOIN s

• The left table in the case of LEFT OUTER JOIN

• The right table in the case of RIGHT OUTER JOIN
• Or both in the case of FULL OUTER JOIN

RIGHT OUTER JOIN

A RIGHT OUTER JOIN operation returns all matching rows in both tables, and also rows in the right table that
don't have a corresponding row in the left table. In the result set of a RIGHT OUTER JOIN operation, the rows
that don't have a corresponding row in the left table contain a NULL value in all columns of the left table.
For example, imagine that you want to retrieve all regions with their respective territories and also the regions
that don't have a corresponding territory. To solve this problem, a RIGHT OUTER JOIN can be performed
between Territories and Region, thus preserving all rows in the Region table that don't have a
corresponding row in the Territories table. This query is shown in Listing 5.14, which also shows a
query that gets regions that don't have territories.
Notice that the last three rows of the result set are rows from the Region table that don't have a
corresponding row in the Territories table. This is the reason they have a NULL value in the first two
columns (which belong to the Territories table).

Listing 5.14 Using RIGHT OUTER JOIN

USE Northwind

INSERT Region VALUES (5,'Europe')

INSERT Region VALUES (6,'Latin America')
INSERT Region VALUES (7,'Asia')

-- Get regions with their respective territories

SELECT territoryid, territorydescription, R.regionid, regiondescription
FROM Territories T RIGHT OUTER JOIN Region R
ON T.regionid = R.regionid

-- Get regions that don't have territories

SELECT territoryid, territorydescription, R.regionid, regiondescription
FROM Territories T RIGHT OUTER JOIN Region R
ON T.regionid = R.regionid
WHERE territoryid IS NULL
GO

territoryid territorydescription regionid regiondescription

-------------------- ------------------------- ----------- ------------------
01581 Westboro 1 Eastern

165
Microsoft SQL Server 2000 Programming by Example

01730 Bedford 1 Eastern

01833 Georgetown 1 Eastern
02116 Boston 1 Eastern
02139 Cambridge 1 Eastern
02184 Braintree 1 Eastern
02903 Providence 1 Eastern
06897 Wilton 1 Eastern
07960 Morristown 1 Eastern
08837 Edison 1 Eastern
10019 New York 1 Eastern
10038 New York 1 Eastern
11747 Mellvile 1 Eastern
14450 Fairport 1 Eastern
19713 Neward 1 Eastern
20852 Rockville 1 Eastern
27403 Greensboro 1 Eastern
27511 Cary 1 Eastern
40222 Louisville 1 Eastern
60179 Hoffman Estates 2 Western
60601 Chicago 2 Western
80202 Denver 2 Western
80909 Colorado Springs 2 Western
85014 Phoenix 2 Western
85251 Scottsdale 2 Western
90405 Santa Monica 2 Western
94025 Menlo Park 2 Western
94105 San Francisco 2 Western
95008 Campbell 2 Western
95054 Santa Clara 2 Western
95060 Santa Cruz 2 Western
98004 Bellevue 2 Western
98052 Redmond 2 Western
98104 Seattle 2 Western
03049 Hollis 3 Northern
03801 Portsmouth 3 Northern
19428 Philadelphia 3 Northern
44122 Beachwood 3 Northern
45839 Findlay 3 Northern
48075 Southfield 3 Northern
48084 Troy 3 Northern
48304 Bloomfield Hills 3 Northern
53404 Racine 3 Northern
55113 Roseville 3 Northern
55439 Minneapolis 3 Northern
29202 Columbia 4 Southern
30346 Atlanta 4 Southern
31406 Savannah 4 Southern
32859 Orlando 4 Southern
33607 Tampa 4 Southern
72716 Bentonville 4 Southern
75234 Dallas 4 Southern
78759 Austin 4 Southern
NULL NULL 5 Europe
NULL NULL 6 Latin America
NULL NULL 7 Asia

(56 row(s) affected)

erritoryid territorydescription regionid regiondescription

------------- ---------------------- ----------- ------------------
NULL NULL 5 Europe

166
 Chapter 5. Querying Multiple Tables: JOIN s

NULL NULL 6 Latin America

NULL NULL 7 Asia

(3 row(s) affected)
Figure 5.2 shows a graphicalrepresentation of the RIGHT OUTER JOIN performed in Listing 5.14 (this
figure doesn't contain all data stored in the original tables). In this figure, you can see the rows of the right
table (Region) that don't have a corresponding row in the left table (Territories).

Figure 5.2. Processing RIGHT OUTER JOIN operations.

Tip

RIGHT OUTER JOIN is equivalent to RIGHT JOIN, so either one of them can be used to specify
a RIGHT OUTER JOIN operation.

Listing 5.15 shows a query that uses RIGHT JOIN instead of RIGHT OUTER JOIN. Notice that this query is
equivalent to the one shown in Listing 5.14.

Listing 5.15 Using RIGHT JOIN to Perform a RIGHT OUTER JOIN Operation

167
Microsoft SQL Server 2000 Programming by Example

USENorthwind

SELECT territoryid, territorydescription, R.regionid, regiondescription

FROM Territories T RIGHT JOIN Region R
ON T.regionid = R.regionid
GO

LEFT OUTER JOIN

In addition to the rows that match the JOIN condition, a LEFT OUTER JOIN returns the rows from the left
table that don't have a corresponding row in the right table.
In a LEFT OUTER JOIN operation, the unmatched rows of the result set have NULL values in the columns of
the right table.
Basically, a LEFT OUTER JOIN can be translated into a RIGHT OUTER JOIN if the order of the tables is
changed (the right table becomes the left and vice versa). This is why the order of the tables in an OUTER
JOIN operation is important.
Listing 5.16 shows a LEFT OUTER JOIN operation between Region and Territories. This query is
similar to the one shown in Listing 5.14, but the order of the tables was changed and also the type of JOIN,
thus generating the same result set seen earlier in Listing 5.14.

Listing 5.16 Using LEFT OUTER JOIN

USE Northwind

SELECT territoryid, territorydescription, R.regionid, regiondescription

FROM Region R LEFT OUTER JOIN Territories T
ON R.regionid = T.regionid
GO

territoryid territorydescription regionid regiondescription

-------------------- ------------------------- ----------- ------------------
01581 Westboro 1 Eastern
01730 Bedford 1 Eastern

168
 Chapter 5. Querying Multiple Tables: JOIN s

01833 Georgetown 1 Eastern

02116 Boston 1 Eastern
02139 Cambridge 1 Eastern
02184 Braintree 1 Eastern
02903 Providence 1 Eastern
06897 Wilton 1 Eastern
07960 Morristown 1 Eastern
08837 Edison 1 Eastern
10019 New York 1 Eastern
10038 New York 1 Eastern
11747 Mellvile 1 Eastern
14450 Fairport 1 Eastern
19713 Neward 1 Eastern
20852 Rockville 1 Eastern
27403 Greensboro 1 Eastern
27511 Cary 1 Eastern
40222 Louisville 1 Eastern
60179 Hoffman Estates 2 Western
60601 Chicago 2 Western
80202 Denver 2 Western
80909 Colorado Springs 2 Western
85014 Phoenix 2 Western
85251 Scottsdale 2 Western
90405 Santa Monica 2 Western
94025 Menlo Park 2 Western
94105 San Francisco 2 Western
95008 Campbell 2 Western
95054 Santa Clara 2 Western
95060 Santa Cruz 2 Western
98004 Bellevue 2 Western
98052 Redmond 2 Western
98104 Seattle 2 Western
03049 Hollis 3 Northern
03801 Portsmouth 3 Northern
19428 Philadelphia 3 Northern
44122 Beachwood 3 Northern
45839 Findlay 3 Northern
48075 Southfield 3 Northern
48084 Troy 3 Northern
48304 Bloomfield Hills 3 Northern
53404 Racine 3 Northern
55113 Roseville 3 Northern
55439 Minneapolis 3 Northern
29202 Columbia 4 Southern
30346 Atlanta 4 Southern
31406 Savannah 4 Southern
32859 Orlando 4 Southern
33607 Tampa 4 Southern
72716 Bentonville 4 Southern
75234 Dallas 4 Southern
78759 Austin 4 Southern
NULL NULL 5 Europe
NULL NULL 6 Latin America
NULL NULL 7 Asia

(56 row(s) affected)

Tip

LEFT OUTER JOIN and LEFT JOIN are interchangeable, so either one of them can be used to
specify a LEFT OUTER JOIN operation.

169
Microsoft SQL Server 2000 Programming by Example

FULL OUTER JOIN

A FULL OUTER JOINoperation returns

• All rows that match the JOIN condition.

• Rows from the left table that don't have a corresponding row in the right table. These rows have NULL
values in the columns of the right table.
• Rows from the right table that don't have a corresponding row in the left table. These rows have NULL
values in the columns of the left table.

Therefore, the result set of a FULL OUTER JOIN operation is like the intersection of the result sets generated
by LEFT OUTER JOIN and RIGHT OUTER JOIN.
For example, imagine that you want to know which suppliers are located in a country where a customer is
located. This is solved by performing an INNER JOIN between Suppliers and Customers on their
country column. If you also want to know the suppliers that are located in a country where there are no
customers, and vice versa (customers that are located in countries where there are no suppliers), you would
have to perform a FULL OUTER JOIN between Suppliers and Customers on their country column.
Listing 5.17 shows this FULL OUTER JOIN operation. Notice that NULL values in the result set indicate that
either there is no corresponding customer for a certain supplier in a country or, on the contrary, there is no
corresponding supplier for a specific customer in a country.

Listing 5.17 Using FULL OUTER JOIN

USE Northwind

SELECT S.companyname as suppliername, S.country as supcountry,

C.companyname as customername, C.country as cuscountry
FROM Suppliers S FULL OUTER JOIN Customers C
ON S.country = C.country
GO

-- Partial results

suppliername supcountry customername cuscountry

----------------------------- ----------- -------------------------- ----------
Heli Süßwaren GmbH & Co. KG Germany Alfreds Futterkiste Germany
NULL NULL Antonio Moreno Taquería Mexico
Exotic Liquids UK Around the Horn UK

170
 Chapter 5. Querying Multiple Tables: JOIN s

Specialty Biscuits, Ltd. UK Around the Horn UK

PB Knäckebröd AB Sweden Berglunds snabbköp Sweden
Svensk Sjöföda AB Sweden Berglunds snabbköp Sweden
NULL NULL Cactus Comidas para llevar Argentina
NULL NULL Centro comercial Moctezuma Mexico
NULL NULL Chop-suey Chinese Switzerland
NULL NULL Tortuga Restaurante Mexico
Refrescos Americanas LTDA Brazil Wellington Importadora Brazil
New England Seafood Cannery USA White Clover Markets USA
New Orleans Cajun Delights USA White Clover Markets USA
Grandma Kelly's Homestead USA White Clover Markets USA
Bigfoot Breweries USA White Clover Markets USA
Karkki Oy Finland Wilman Kala Finland
NULL NULL Wolski Zajazd Poland
Zaanse Snoepfabriek Netherlands NULL NULL
Leka Trading Singapore NULL NULL
Pavlova, Ltd. Australia NULL NULL
G'day, Mate Australia NULL NULL
Tokyo Traders Japan NULL NULL
Mayumi's Japan NULL NULL
Tip

FULL JOIN is equivalent to FULL OUTER JOIN; they can be used interchangeably.

CROSS JOINs
A CROSS JOIN generates a Cartesian product of the tables specified in the JOIN operation. In other words,
the result set of a CROSS JOIN operation contains every possible combination of rows of the tables involved
in the query. In particular, if there are n rows in the first table and m rows in the second table, the result set will
have n*m rows.
These are the two possible ways to specify a CROSS JOIN operation:

SELECT * FROM Table1, Table2

SELECT * FROM Table1 CROSS JOIN Table2
Listing 5.18 shows an example that generates a Cartesian product using CROSS JOIN. The tables involved
in the CROSS JOIN have 7 and 8 rows, respectively, and the result set has 56 rows (7*8).

Listing 5.18 Using CROSS JOIN to Generate a Cartesian Product

USE Northwind

SELECT *
FROM Region

SELECT categoryid, categoryname

FROM Categories

SELECT regionid, regiondescription, categoryid, categoryname

FROM region CROSS JOIN categories

171
Microsoft SQL Server 2000 Programming by Example

GO

RegionID RegionDescription
----------- --------------------------------------------------
1 Eastern
2 Western
3 Northern
4 Southern
5 Europe
6 Latin America
7 Asia

(7 row(s) affected)

categoryid categoryname
----------- ---------------
1 Beverages
2 Condiments
3 Confections
4 Dairy Products
5 Grains/Cereals
6 Meat/Poultry
7 Produce
8 Seafood

(8 row(s) affected)

regionid regiondescription categoryid categoryname

----------- -------------------- ----------- ---------------
1 Eastern 1 Beverages
1 Eastern 2 Condiments
1 Eastern 3 Confections
1 Eastern 4 Dairy Products
1 Eastern 5 Grains/Cereals
1 Eastern 6 Meat/Poultry
1 Eastern 7 Produce
1 Eastern 8 Seafood
2 Western 1 Beverages
2 Western 2 Condiments
2 Western 3 Confections
2 Western 4 Dairy Products
2 Western 5 Grains/Cereals
2 Western 6 Meat/Poultry
2 Western 7 Produce
2 Western 8 Seafood
3 Northern 1 Beverages
3 Northern 2 Condiments
3 Northern 3 Confections
3 Northern 4 Dairy Products
3 Northern 5 Grains/Cereals

172
 Chapter 5. Querying Multiple Tables: JOIN s

3 Northern 6 Meat/Poultry
3 Northern 7 Produce
3 Northern 8 Seafood
4 Southern 1 Beverages
4 Southern 2 Condiments
4 Southern 3 Confections
4 Southern 4 Dairy Products
4 Southern 5 Grains/Cereals
4 Southern 6 Meat/Poultry
4 Southern 7 Produce
4 Southern 8 Seafood
5 Europe 1 Beverages
5 Europe 2 Condiments
5 Europe 3 Confections
5 Europe 4 Dairy Products
5 Europe 5 Grains/Cereals
5 Europe 6 Meat/Poultry
5 Europe 7 Produce
5 Europe 8 Seafood
6 Latin America 1 Beverages
6 Latin America 2 Condiments
6 Latin America 3 Confections
6 Latin America 4 Dairy Products
6 Latin America 5 Grains/Cereals
6 Latin America 6 Meat/Poultry
6 Latin America 7 Produce
6 Latin America 8 Seafood
7 Asia 1 Beverages
7 Asia 2 Condiments
7 Asia 3 Confections
7 Asia 4 Dairy Products
7 Asia 5 Grains/Cereals
7 Asia 6 Meat/Poultry
7 Asia 7 Produce
7 Asia 8 Seafood

(56 row(s) affected)

A JOIN condition is not neededwhen using CROSS JOIN. However, a JOIN condition may be specified in the
WHERE clause, and in this case, the CROSS JOIN operation behaves like an INNER JOIN.
Listing 5.19 shows two equivalent queries. The first one performs a CROSS JOIN and has a join condition,
and the second one is an INNER JOIN operation. Both queries produce the same result set.

Listing 5.19 Using CROSS JOIN with a JOIN Condition

USENorthwind

SELECT territoryid, territorydescription, R.regionid, regiondescription

FROM Territories T CROSS JOIN Region R
WHERE T.regionid = R.regionid
AND R.regiondescription = 'Southern'

SELECT territoryid, territorydescription, R.regionid, regiondescription

173
Microsoft SQL Server 2000 Programming by Example

FROM Territories T INNER JOIN Region R

ON T.regionid = R.regionid
AND R.regiondescription = 'Southern'
GO

territoryid territorydescription regionid regiondescription

-------------------- ------------------------- ----------- ------------------
29202 Columbia 4 Southern
30346 Atlanta 4 Southern
31406 Savannah 4 Southern
32859 Orlando 4 Southern
33607 Tampa 4 Southern
72716 Bentonville 4 Southern
75234 Dallas 4 Southern
78759 Austin 4 Southern

(8 row(s) affected)

territoryid territorydescription regionid regiondescription

-------------------- ------------------------- ----------- ------------------
29202 Columbia 4 Southern
30346 Atlanta 4 Southern
31406 Savannah 4 Southern
32859 Orlando 4 Southern
33607 Tampa 4 Southern
72716 Bentonville 4 Southern
75234 Dallas 4 Southern
78759 Austin 4 Southern

(8 row(s) affected)
Usually, the purpose of a CROSS JOIN operation is to generate data for testing purposes because you can
generate a big result set from small tables. Furthermore, similar to the other JOIN operations, a CROSS JOIN
can involve more than two tables. In particular, the syntax used to perform a CROSS JOIN that involves three
tables is

SELECT *
FROM Table1 CROSS JOIN Table2
CROSS JOIN Table3
Caution

Be aware that a FULL OUTER JOIN operation is different from a CROSS JOIN operation. Usually,
a CROSS JOIN returns more rows because it returns every combination of rows in both tables.

Self Joins
A self join is a special type of join, in which a certain table is joined to itself. Basically, in a self join, two copies
of the same table are merged, generating a result set based on information stored in this table.

174
 Chapter 5. Querying Multiple Tables: JOIN s

Generally, self joins are used to represent hierarchies in a table. For example, the Employees table has a
column called reportsto, which has a foreign key pointing to the employeeid column in this table.
Therefore, if you want to retrieve the manager of any employee, the Employees table must be joined to itself.
Listing 5.20 demonstrates how to extract information from this hierarchy represented in the Employees
table, using a self join. Specifically, the query performs a self join to retrieve the name of Anne Dodsworth's
manager (her manager is also an employee).

Listing 5.20 Using a Self Join to Retrieve Hierarchy Information

USE Northwind

SELECT E1.employeeid, E1.firstname, E1.lastname,

E2.firstname as managerfirstname, E2.lastname as managerlastname
FROM Employees E1 JOIN Employees E2
ON E1.reportsto = E2.employeeid
WHERE E1.lastname = 'Dodsworth'
AND E1.firstname = 'Anne'
GO

employeeid firstname lastname managerfirstname managerlastname

----------- ---------- -------------------- ---------------- ----------------
9 Anne Dodsworth Steven Buchanan

(1 row(s) affected)
Caution

Notice that table aliases must be used when working with self joins to differentiate between the two
copies of the table.

The UNION Operator

UNION is an operator used to combine two or more SELECT statements and generate one result set. These
SELECT statements must meet some conditions, such as the following:

• They must have the same number of columns. You can work around this restriction if you use
constants in the SELECT statement with fewer columns, as shown in Listing 5.21.

Listing 5.21 Using a UNION Operator with Constants in the SELECT Statements

175
Microsoft SQL Server 2000 Programming by Example

-- A constant has to be used in the second SELECT statement

-- because Shippers doesn't have a contactname column

USE Northwind
SELECT companyname, contactname FROM Suppliers WHERE country = 'USA'
UNION
SELECT companyname, 'N/A'FROM Shippers
GO

companyname contactname
---------------------------------------- ------------------------------
New Orleans Cajun Delights Shelley Burke
Grandma Kelly's Homestead Regina Murphy
Bigfoot Breweries Cheryl Saylor
New England Seafood Cannery Robb Merchant
Speedy Express N/A
United Package N/A
Federal Shipping N/A

(7 row(s) affected)

• The column's data types must be compatible. In other words, the data types must be equivalent, can
be converted implicitly, or must be explicitly converted using either CAST or CONVERT. In the previous
example (Listing 5.21), the companyname column of both tables has the same data type
(NVARCHAR), and the other column (contactname) is compatible with the 'N/A' constant.

The column names of the result set in a UNION operation are taken from the column names of the first
SELECT statement of UNION. By default, UNION removes all duplicates from the result set. However, if you
want to keep duplicates in the result set, use the ALL keyword when performing a UNION operation. Listing
5.22 shows the difference between using UNION and UNION ALL.

Listing 5.22 Using the ALL Option in UNION Operators

USE Northwind

176
 Chapter 5. Querying Multiple Tables: JOIN s

SELECT city, country FROM Customers WHERE country = 'UK'

UNION
SELECT city, country FROM Suppliers WHERE country = 'UK'
SELECT city, country FROM Customers WHERE country = 'UK'
UNION ALL
SELECT city, country FROM Suppliers WHERE country = 'UK'
GO

city country
--------------- ---------------
Cowes UK
London UK
Manchester UK
(3 row(s) affected)

city country
--------------- ---------------
London UK
London UK
London UK
London UK
Cowes UK
London UK
London UK
London UK
Manchester UK

(9 row(s) affected)
The result set of a UNION operation can be ordered, but be aware that only one ORDER BY clause can be
specified when using UNION, and it must be specified in the last SELECT statement. Listing
5.23demonstrates how to use ORDER BY in a UNION operation.

Listing 5.23 Using an ORDER BY in a UNION Operation

USE Northwind

SELECT city, country FROM Customers WHERE country = 'UK'

UNION ALL
SELECT city, country FROM Suppliers WHERE country = 'UK'
ORDER BY city
GO

177
Microsoft SQL Server 2000 Programming by Example

city country
--------------- ---------------
Cowes UK
London UK
London UK
London UK
London UK
London UK
London UK
London UK
Manchester UK

(9 row(s) affected)
When using UNION, only the first SELECT statement can have an INTO keyword, which allows you to create
a table on-the-fly with the result set of the UNION operation. Listing 5.24 creates a temporary table that
stores the full name of employees and suppliers.

Listing 5.24 Creating a Table On-the-Fly with the Result of a UNION

USE Northwind

SELECT firstname + ''+ lastname as fullname

INTO #employeesandsuppliers
FROM Employees
UNION
SELECT contactname FROM Suppliers WHERE country = 'usa'
ORDER BY fullname

SELECT * FROM #employeesandsuppliers

GO
(13 row(s) affected)

fullname
-------------------------------

178
 Chapter 5. Querying Multiple Tables: JOIN s

Andrew Fuller
Anne Dodsworth
Cheryl Saylor
Janet Leverling
Laura Callahan
Margaret Peacock
Michael Suyama
Nancy Davolio
Regina Murphy
Robb Merchant
Robert King
Shelley Burke
Steven Buchanan

(13 row(s) affected)

What's Next?

You have been studying different ways to access data in databases. Until now, performance hasn't been an
issue. However, as you might have experienced, databases are constantly growing, and sometimes this can
hurt the performance of queries and applications that access the database. This performance degradation can
turn out to be a very serious problem, because we want to be able to issue queries against the database and
get the result set right away.
In general, indexes can be used to improve the performance of queries. The main characteristic of indexes is
that they speed up data retrieval, even when working with big tables. In the next chapter, you go through all
the steps and guidelines needed to create useful indexes that can improve the performance of queries and
applications.

179
 Chapter 6. Optimizing Access to Data: Indexes

Chapter 6. Optimizing Access to Data: Indexes

Perhaps the main reason to install a database system is to be able to search efficiently for data. Commercial
systems use vast amounts of data, and users have come to expect a reasonably short response time when
searching for information irrespective of how the search is carried out or the criteria used in the search.
Many other programming books and technical papers have already covered searching and sorting algorithms,
so it is not the purpose of this book to introduce any new theories on this subject.
To produce results in the quickest and most efficient way, SQL Server must have fast access to the data. It
does this by allowing every operation to have optimized access to any resource that it might need to use.
This chapter teaches you

• How to use indexes in everyday operations

• How SQL Server 2000 implements indexes
• How SQL Server 2000 accesses data from tables
• The differences between clustered and nonclustered indexes
• How to create, modify, and delete indexes
• How SQL Server 2000 accesses data stored as a clustered index
• How SQL Server 2000 accesses data stored as a heap
• How to create an index to cover a query
• What index fragmentation is and how to manage it
• How to use the Index Tuning Wizard

Introduction to Indexes
You are used to working with indexes in daily life. Many books have a table of contents. This table of contents
shows the book's outline, and in this outline, you can see on which page every topic starts. You can read just
the table of contents and search for a subject that interests you. After you find an interesting topic, you can
quickly go straight to the correct page number.
In this example, the book is organized according to the table of contents. The table of contents is not ordered
alphabetically but it is ordered by its outline. If you want to search for a specific topic, you have to search the
table of contents sequentially from the beginning to the end.
Every entry in the table of contents can be considered a key to the index, whereas the page number is no
more than a pointer to the physical page where the topic starts.
Most books have an indexat the end.The index is just a collection of keywords that are organized
alphabetically. These keywords represent different topics: For every keyword the index shows the page or
pages where this topic is covered. Because these keywords are alphabetically ordered, it is easy to search for
a specific topic without having to read the index sequentially.
An index key in SQL Server is like a keyword in the index of a book, and the page number of the index in the
book is the pointer to the physical page where the topic is covered.
Perhaps it is difficult to visualize a book as a table in a database, but a book is information that is stored in a
sequence of physical pages, which is not very different from the way SQL Server 2000 stores data, as you will
see later in this chapter.
Let's consider another familiar example. Every telephone company maintains a database of customers. This
database has, at least, the family name, first name, address, and telephone number of every customer.
Applications for new telephone lines are stored in order, but the telephone directory shows entries in
alphabetical order. If you know the customer's family name, you can easily search for their telephone number
by browsing the telephone directory. To help us carry out this search, every page in the telephone directory is
labeled with either the first or the last entry on the page, so you can read the labels and search for the entry
you want to find.
To search for a specific entry in the directory, you follow a process that could be called a binary search. SQL
Server uses binary search when it scans an index searching for specific information.
Having the data ordered by more than one attribute— that is, surname, first name, and address as in this
case— is what you do when you use a compos ite index. In this case, the combination of surname, first name,
and address is the index key. Telephone directory data is physically written in this order, and this is what you
call a clustered index in SQL Server. You will look at clustered indexes later in this chapter.
Because you often need to search for businesses offering a specific kind of service, some companies offer
special yellow pages books to help people search for specific businesses. The yellow pages order telephone
entries by activity. Entries are ordered alphabetically for every activity.

181
Microsoft SQL Server 2000 Programming by Example

The yellow pages show a typical example of a nonclustered index in SQL Server.
Searching for customers living in a specific street or area will mean having to sequentially read the standard
telephone book and extract, one by one, every entry that matches the searching criteria. Just as in the
preceding case, when SQL Server doesn't have a suitable index to execute a query, the only choice available
is to read the entire table first and then filter out every entry that matches the searching criteria.

Benefits of Indexes
SQL Server queries can benefit from the existence of suitable indexes in the following cases:

• Exact-match queries— When searching for rows with specific key values. These are queries with a
WHERE clause to restrict the query to specific values for every key column.
• Range queries— When solving queries that search for a range of values in a column.
• Filtering for values in the foreign key to solve a join operation— When using a JOIN predicate to
search for rows in a table based on keys from a second table.
• Hash and merge join operations— In some cases, having an index can speed up the execution of a
JOIN algorithm, because the data is exactly in the order that the JOIN algorithm uses.
• Covering a query— To avoid a full-table scan, when a narrow index contains all the required data.
• Avoiding duplicates— To check for the existence of suitable indexes in an INSERT or UPDATE
operation in an attempt to avoid duplicates.
• Sorting data— To produce an ordered output when using the ORDER BY clause.

In the following sections, you will see one example of each of these cases.
For every example, you can see the estimated execution plan before the query is executed. We do not
provide the result of the queries because we want you to pay attention to the execution plan, not to the final
result of the query.
If a query plan shows that it accesses a table with a table scan, SQL Server must read the entire table to
execute the query. This is not very efficient on big tables because SQL Server must read every data page.
If the query plan shows Index Scanas a table access method, SQL Server must read the entire index to solve
the query. If the index is a clustered index, this is equivalent to a table scan. If the index is a nonclustered
index, this strategy is usually more efficient than a table scan because the clustered index key usually is
shorter than the row in the table. Therefore, the number of index pages to read is smaller than the number of
data pages. We will discuss clustered and nonclustered indexes later in this chapter, in the "Clustered
Indexes" and the "Nonclustered Indexes" sections.
If you see in the query plan the words Index Seek, this usually represents good news for you, because SQL
Server can traverse the index searching for the required rows, following a binary search strategy, without
entirely reading the index.
Note

To show the estimated execution plan without running the query, you can use the menu Query -
Display Estimated Execution Time, the Ctrl+L keyboard shortcut, or click the Display Estimated
Execution Time icon in the toolbar.

Using Indexes on Point Queries

If you execute a query that searches for an exact match between fields and specific values, call this query a
Point Query. SQL Server might find it efficient to use an index to search for these values. Listing 6.1 shows
an example of this type of query, where you search for the product identified by ProductID = 10.
Figure 6.1 shows the query plan that SQL Server uses to execute the query of Listing 6.1. The query plan
showsClustered Index Seek, which means that SQL Server uses the index to seek for the requested value,
without reading the entire table or the entire index.

182
 Chapter 6. Optimizing Access to Data: Indexes

Figure 6.1. SQL Server can use an index to solve point queries, as in Listing 6.1.

Listing 6.1 SQL Server Uses an Index to Solve a Point Query

USE Northwind
GO
SELECT *
FROM Northwind.dbo.Products
WHERE ProductID = 10
Note

Readers can visit the book's Web site (http://www.sqlserverbyexample.com) to download a

more complex and bigger database to practice with more complex scripts that can show
differences on timing when they use indexes.

Using Indexes in Range Queries

A range query searches for data using a condition that specifies minimum and maximum values, such as
products with a stock level lower than the minimum essential level (UnitsInStock BETWEEN 0 AND 25).
Another example of a range query is a query that uses the LIKE operator, such as searching for customers
living in a specific urban area (Telephone LIKE '(321)%'). The filtering condition uses a range of
possible values (Telephone >= '(321)'AND Telephone < '(322)').
Listing 6.2 shows different cases of range queries using comparison operators, the BETWEEN operator, and
the LIKE operator. Figure 6.2 shows the execution plan for a range query using the BETWEEN operator. You
can identify that SQL Server uses an index on the PostalCode column because the query plan shows the
Index Seek icon.

183
Microsoft SQL Server 2000 Programming by Example

Figure 6.2. SQL Server can use an index to solve queries restricted to a range of values, as in Listing
6.2.

Execute each query from Listing 6.2 individually to study the estimated query execution plan for every query.
Note

You can select which part of a script to execute by marking the required block of statements in SQL
Query Analyzer with the mouse or the keyboard— in the same way you mark a block using any text
editor.

Listing 6.2 SQL Server Can Use Indexes When the Query Is Defined for a Range of Values

USE Northwind
GO

-- Combining > or >= with < or <=

SELECT *
FROM Northwind.dbo.Products
WHERE UnitPrice > 10
AND UnitPrice <= 20

-- Using the BETWEEN operator

SELECT *
FROM Northwind.dbo.Customers
WHERE PostalCode BETWEEN 'WX1'AND 'WXZZZZZ'

-- Which is equivalent to:

184
 Chapter 6. Optimizing Access to Data: Indexes

SELECT *
FROM Northwind.dbo.Customers
WHERE PostalCode >= 'WX1'
AND PostalCode <= 'WXZZZZZ'

-- Using the LIKE Operator with trailing wildcards

SELECT *
FROM Northwind.dbo.Customers
WHERE CompanyName LIKE 'Hungry%'

-- Which is equivalent to:

SELECT *
FROM Northwind.dbo.Customers
WHERE CompanyName >= 'Hungry'
AND CompanyName < 'HungrZ'

Using Indexes to Search for Foreign Key Values to Solve Join Operations

An example of this case is when SQL Server has to execute a join to retrieve data from two tables, such as
searching for orders containing products from a specific category.
Note

You can learn more about join algorithms in Chapter 5.]

Listing 6.3: shows the query to solve this example where, to produce the required information, the query
must join the Products and Order Details tables. You can see in Figure 6.3 that SQL Server uses an
index seek on the Products table to solve this join.

Figure 6.3. SQL Server uses an index to solve queries with a JOIN predicate and a WHERE clause.

Listing 6.3 SQL Server Uses an Index to Execute This Query That Joins Two Tables

185
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

SELECT Products.ProductID,
[Order Details].UnitPrice,
[Order details].Quantity
FROM Products
JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE Products.CategoryID = 1

Using Indexes to Speed Up the Execution of Hash and Merge JOIN Operations

If the columns used tojoin two tables have an index on every table that participates in the JOIN operation,
SQL Server can use the merge join algorithm. For example, if you join the Categories table and the
Products table by the CategoryID column, and there is an index on the CategoryID column in the
Categories table and another index on the CategoryID column in the Products table, SQL Server can
very efficiently use a merge join to connect both tables.
To execute the hash join algorithm, you do not need to have an index on the joining columns, but having an
index on the joining columns can speed up the process as well.
Note

You can learn more about join algorithms in Chapter 5.

Listing 6.4 shows a query where you join the Products table and the Order Details table by using the
ProductID column. The ProductID column has an index defined in the Products table and the Order
Details table; this is why SQL Server solves the query with an index scan on every table plus a hash join
operation. Figure 6.4 shows the estimated execution plan for this query.

Figure 6.4. SQL Server uses indexes to help in the hash-match/Inner join operation.

186
 Chapter 6. Optimizing Access to Data: Indexes

Listing 6.4 SQL Server Can Use Indexes to Join Tables

USE Northwind
GO

SELECT Products.ProductID,
[Order Details].UnitPrice,
[Order details].Quantity
FROM Products
JOIN [Order Details]]
ON Products.ProductID = [Order Details].ProductID

Using Indexes That Cover a Query

In some cases, you can find that an index contains all the information required to execute a query. For
example, if you want to produce a list of customers by name and you had an index on name, just reading the
index SQL Server provides is enough information to produce the required results.
In these cases, the index covers the query, and reading the index is more efficient than reading the table
because, usually, an index key is shorter than a table row.
Note

Later in this chapter, in the "Covered Queries and Index Intersection" section, covered
indexes are discussed in detail.

Listing 6.5 shows a query that can be executed using just an index defined on the CategoryID column of
the Products table. Figure 6.5 shows that SQL Server uses an index scan on the CategoryID index to
solve this query.

Figure 6.5. SQL Server uses an index to solve a query with no access required to data pages.

187
Microsoft SQL Server 2000 Programming by Example

Listing 6.5 SQL Server Can Use an Index to Avoid Access to Data Pages

USE Northwind
GO

SELECT DISTINCT CategoryID

FROM Products

Using an Index to Enforce Uniqueness

Every time you try to insert a new value in a column with a PRIMARY KEY or UNIQUE CONSTRAINT, SQL
Server must check to see whether the new value already exists. To speed up this process, SQL Server uses
the index created during the CONSTRAINT creation.
Note

Constraints are covered in Chapter 7.

Listing 6.6 inserts a new row in the Categories table. This table has a unique index on the CategoryID
column because this column has a FOREIGN KEY CONSTRAINT defined. Figure 6.6 shows that SQL Server
uses the index on the CategoryID column to solve the clustered index insert operation.

Figure 6.6. SQL Server uses an index to guarantee uniqueness in INSERT and UPDATE operations.

Listing 6.6 SQL Server Uses Unique Indexes to Check for Duplicate Values in INSERT and UPDATE
Operations

188
 Chapter 6. Optimizing Access to Data: Indexes

USE Northwind
GO

-- Execute this instruction first

SET IDENTITY_INSERT Categories ON

GO

-- Retrieve the Estimated Execution Plan of the following query

INSERT Categories (CategoryID, CategoryName, Description)

VALUES (9, 'Liquors', 'Whiskies, Brandies and other Spirits')
GO
-- Execute this instruction at the end

SET IDENTITY_INSERT Categories OFF

GO

Using Indexes to Help Produce Ordered Output

This is the more obvious use of indexes. If SQL Server can retrieve the data in order, it will not need to
reorder the data before it can display it.
Using the query from Listing 6.7 you can retrieve the ProductName and UnitPrice for the entire products
table, ordering the result by the ProductName column. SQL Server uses an index on ProductName to solve
the query, as you can see in Figure 6.7, because in this way it can retrieve the data already in
ProductName order.

Figure 6.7. SQL Server can use indexes when sorting data is required.

Listing 6.7 SQL Server Can Use Indexes to Produce an Ordered Output

189
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

SELECT ProductName, UnitPrice

FROM Products
ORDER BY ProductName ASC
If none of the indexes available matches the ordering criteria, SQL Server must execute a SORT process to
order the final result set.
Listing 6.8 shows a similar example as in Listing 6.7, but in this case we order by UnitPrice. Because
the UnitPrice column is not indexed, SQL Server mustexecute the SORT process, as shown in Figure 6.8.

Figure 6.8. SQL uses a SORT step to order data when no suitable index exists.

Listing 6.8 SQL Server Must Execute a SORT Process If None of the Indexes Available Matches the
Ordering Criteria

USE Northwind
GO

SELECT ProductName, UnitPrice

FROM Products
ORDER BY UnitPrice ASC

How to Create Indexes

190
 Chapter 6. Optimizing Access to Data: Indexes

To create an index on a table, you have to select on which columns to build the index, and then use the
CREATE INDEX statement. As an example, to create the index called ndx_Customers_City on the City
field of the Customers table, you must execute the statement of Listing 6.9.

Listing 6.9 Creating an Index Is Easy Using the CREATE INDEX Statement

USE Northwind
GO

CREATE INDEX ndx_Customers_City

ON Customers(City)
You can use Enterprise Manager to create indexes on tables using the Manage Indexes form. To open the
Manage Indexes form, you must display the list of tables in a database, select the table in which you want to
create the index, right-click with the mouse on the table name, or choose Action, All Tasks, Manage Indexes.
Figure 6.9 shows the ManageIndexes form.

Figure 6.9. Getting information on indexes available for every table or view is easy using the Manage
Indexes form.

You can select the databaseand the table in which to create the index by using the Database and Table/view
drop-down list boxes.

191
Microsoft SQL Server 2000 Programming by Example

To create an index, you can click the New command button, and the Create New Index form will appear.
Figure 6.10 shows the Create New Index form.

Figure 6.10. Creating indexes is easy using the Create New Index form.

Every index must have a name, and the name must be unique for the table or view where the index is defined.
Below the index name, you can see the list of columns you can make part of the index. Clicking the check
boxes on the left of the column name, you can specify which columns to include in the index. You also can
select whether the column must be included in ascending or descending order.
You can change the order of the columns in the index by using the Up and Down buttons.
Caution

The order of columns in an index is important: An index created on the (CompanyName, Phone)
columns, as in Listing 6.10, can be useful to search for telephone numbers of a certain company;
however, an index on (Phone, CompanyName) could be useful to search for the company where
a specific telephone number is installed.

Listing 6.10 You Can Create an Index on More Than One Column Using the CREATE INDEX Statement

192
 Chapter 6. Optimizing Access to Data: Indexes

USE Northwind
GO

CREATE INDEX ndx_Customers_Company_Phone

ON Customers(CompanyName, Phone)

CREATE INDEX ndx_Customers_Phone_Company

ON Customers(Phone, CompanyName)
After the index is created, nothing special happens. Our data still looks the same. However, having indexes
gives SQL Query Optimizer more ways to optimize a query— and it will use them as necessary.
Later in this chapter, in the "Types of Indexes" and "Index Maintenance" sections, you will learn about
more options that you can use to create different type ofindexes.

How SQL Server 2000 Stores Data

SQL Server stores the dataof every database in one or more data files. These files are stored directly on the
hard disk and are treated like other operating system files; however, SQL Server locks them as soon as it
opens a database.
It is inside the data files thatSQL Server stores tables and indexes, as well as a definition of everyobject in the
database. Properties of objects are stored in the database, too. Every database is almost self contained, so it
can be easily moved to another server if required.
Accessing the hard disk every timea user wants to retrieve some information, such as a product's price, is
inefficient. To read information from disk, the hard disk heads must physically move to access specific
information stored in a specific sector and track of the hard disk. The movement of heads on a hard disk can
be a slow process and can be carried out just a few times per second; however, after the hard disk heads are
in position, they can read the disk and access data very quickly.
Hard disks are designed to move blocks of data efficiently, not short sequences of bytes. SQL Server 2000
reads data in 8KB-page blocks, providing faster access to the data than reading individual fields or rows.
Every data page belongs to an individual table or index. Every table or index has a collection of pages that it
uses to store rows of data. Every row in a table must be contained in a single page, and because the page
contains some information required for internal SQL Server management,the maximum row size is 8,060
bytes.
If a table contains BLOB fields(Binary Large Objects such as image, text, or ntext), these fields are
stored in a separate collection of pages. That's why the limit of 8,060 bytes/row doesn't affect the BLOB fields.
Figure 6.11 shows the data page structure with the following sections:

Figure 6.11. This is the structure of a data page.

193
Microsoft SQL Server 2000 Programming by Example

• Page Header, where SQL Server stores information to help the management of this page. This
section has 96 bytes.
• Data block, where rows are stored. This section is limited to 8,060 bytes.
• Row offset, where SQL Server stores 2 bytes per row slot, as a pointer to the position where the row
starts inside the page.

Data pages, index pages, and text/image pages have the same general structure. The page header identifies
the type of data stored in the page and the data block structure is different for every type of page but, as a rule,
the general structure is the same.
SQL Server has other types of pages, which can be used in general allocation pages. They are used to keep
track of space used by every object and to keep control of unused space in every file. Because these general
allocation pages are defined as bitmaps, they provide very fast access to the information they hold. The study
of these pages is outside the scope of this book, but you can read about themin Books Online.

194
 Chapter 6. Optimizing Access to Data: Indexes

Because reading page by page is not very efficient, SQL Server arranges pages in groups of eight contiguous
pages. This group of eight contiguous pages is called an extent, and SQL Server tries to read and write data
extent by extent if possible. The allocation pages mentioned in the preceding paragraph track extents as a
unit of allocation.
SQL Server tries to access the hard disk one extent at a time, if possible, reading 64KB per extent. In addition,
because reading sequentially from a hard disk is always faster than reading randomly, SQL Server tries
whenever possible to optimize hard-disk access by reading several extents in sequence.
A mixed extent can contain data pages from different objects, but a uniform extent contains data from only
one object. The first eight pages of any object can be held in mixed extents, but as soon as the object needs
more than eight pages, SQL Server allocates uniform extents for it.
The preceding paragraph has explained how SQL Server stores and retrieves data. However, if you want to
retrieve only a few bytes of data, why does SQL Server read 8KB or 64KB of data?
SQL Server keeps data pages in a reserved space in RAM as a buffer, or data cache. Access to RAM is much
faster than access to a hard disk; so, keeping information in RAM provides a boost to performance to any
database application. Every time a process needs to access any data, it searches RAM to see whether the
data pages are already being held in cache. If the pages are not already in RAM, the process must request
the copy of the data pages from disk to RAM before trying to do any operation with them. If the data page is
modified by the process, the page will be marked as a dirty page becauseit does not contain the same
information as the copy available in the disk, and it will be copied to the disk by a separate process.

How SQL Server 2000 Modifies Data

To guarantee consistency and recoverability of the information, SQL Server considers every data modification
as part of a transaction. It will record transaction information, and every modification made as part of every
transaction, in a transaction log. In other words, SQL Server first records how the data will be modifi ed, and
then it modifies the data. The logical process could be explained as follows:

1. A data modification is requested.

2. SQL Server creates a query plan to execute the data modification request.
3. SQL Server starts a transaction and records it in the transaction log.
4. SQL Server writes every operation to apply to the data in the transaction log before modifying the data.
5. After the operations are registered in the transaction log, the affected data is modified in cache, and
the pages are marked as dirty. These changes are provisional. They will be permanent only if the
transaction completes successfully.
6. SQL Server applies steps 4 and 5 for every individual operation in the transaction.
7. If every operation was successful, SQL Server records a transaction commit in the transaction log,
and the changes are considered permanent.
8. If any operation in this transaction failed, SQL Server requests a rollback and will register in the
transaction log the operations required to roll back all the changes applied to the modified data pages.
The transaction is marked as rolled back in the transaction log.

We will cover transactions in Chapter 13, "Maintaining Data Consistency: Transactions and Locks,"
but as demonstrated in the preceding points, a transaction is a unit of work that must be completely committed
or completely rolled back.

Index Enhancements in SQL Server 2000

SQL Server 2000 providessome important index enhancements over previous versions:

• It is possible to create indexes in views. This feature can speed up the process of repetitive queries
using aggregate functions or multiple joins. This topic is covered in the section "Indexed Views,"
later in this chapter.
• It is possible to create indexes on computed columns, providing faster retrieval and sorting for data
calculated dynamically. This topic is cov ered in the "Indexes on Computed Columns" section,
later in this chapter.
• You can define the index on ascending or descending order. If you have a composite index, every
field can be ordered in ascending or descending order independently. Composite indexes are covered
in the "Clustered Indexes" and "Nonclustered Indexes" sections, later in this chapter.

195
Microsoft SQL Server 2000 Programming by Example

• It is possible to use the Tempdb database for index creation, providing better performance if Tempdb
is stored in a different physical disk subsystem.
• The index creation process can be executed in parallel, providing a faster execution than previous
versions.

Accessing Data Without Indexes: Table Scan

As explained before,table data is stored in pages, and pages are grouped in extents. Special pages called
IAM (Index Allocation Map) keep information about which extents are used by a particular table.
Information related to table definition is stored in different system tables:

• The name of the table, as well as the owner, creation date, and other general information, is stored in
the sysobjects system table. The sysobjects table has a field called ID, which provides a
unique object identification number for every object stored in a database. The ID field is used to relate
this object to extra information in other systems'tables. It is possible to retrieve the table ID calling the
OBJECT_ID system function.
• Column definitions are stored in the syscolumns system table.
• UNIQUE, PRIMARY KEY, and FOREIGN KEY constraints are treated as objects, which is why there
is an entry in the sysobjects system table to identify them.
• Other constraints, such as CHECK and DEFAULT constraints, are identified in sysobjects, too, but
their definition is stored in the syscomments table.
• The sysindexes table stores information about indexes'definitions. Every index is identified in the
sysindexes table by its name, its table ID, and a special value unique to every index on the same
table, called indid, with values from 0 to 255.

The sysindexes system table stores important information about how to start reading actual data or index
keys. Every table has at least one entry in the sysindexes table.
If a table doesn't have a clustered index, it is stored as a heap, because the rows are not in any specific order.
In this case, the table itself is defined in the sysindexes table as a row with indid = 0. The allocation
map of a table or index is stored in special pages, called the Index Allocation Map (IAM), and sysindexes
stores the address of the first IAM page of a specific table or index in the FirstIAM field.
IAM pages are special SQL Server pages that contain a sequence of bits, which is why they are called
bitmaps. Every bit indicates whether the correspondent extent in the file uses the page.
If SQL Server needs to retrieve data from a table that doesn't have any clustered index defined, it uses the
IAM pages to select which pages belong to the table and reads the pages in sequence to retrieve the
information.
Because the IAM pages reflect the physical order of the data pages, the reading process can be quite efficient.
However, using indexes can usually provide better performance than table scans.

Types of Indexes
SQL Server can create twogeneral types of indexes:

• Regular SQL Server indexes, which are created by Transact-SQL commands and maintained
automatically by SQL Server. Query Optimizer uses these indexes to speed the execution of queries.
• Full-text indexes, which are created by system-stored procedures, managed by the Microsoft Search
Service, and accessed by specific Transact-SQL extended commands. Query Optimizer doesn't use
full-text indexes automatically to optimize the execution of queries.

Full-text indexes are used by specific Transact-SQL functions. This type of index is out of the scope of this
book, but you can get information about them in Books Online.
This chapter covers in detail the regular SQL Server indexes because of their importance in query
optimization.
The information of the index is stored in pages in a similar way as the data pages. Index pages are organized
in a binary tree to speed up searching operations. These trees have three main sections:

196
 Chapter 6. Optimizing Access to Data: Indexes

• Root node— The root node is the page at the top of the tree, the begin ning of any search operation.
This page contains one entry per every page in the next level of the index.

Figure 6.12 shows the root node and its relationship with pages in the next level. Every entry in the
root node contains the first key of one page of the next level and the physical address of that page.
This physical address is a combination of the data file number and the page number.

Figure 6.12. SQL Server stores indexes in pages organized in a binary tree structure.

Every index has a root node. If the table is so small that it is possible to store all the index entries in a
single page, the index will have only one page— the root node.

• Leaf level— The base of an index is a collection of pages that contain one entry per every row in the
table. The actual composition of this level depends on the type of index, as you will see later. The
entries in the leaf level are in order.

Every index has a leaf level.

• Nonleaf level— If the root node doesn't have enough space to store one entry per every page in the
leaf level, it is necessary to create more pages, arranged in intermediate levels to maintain the link

197
Microsoft SQL Server 2000 Programming by Example

between the root node and the leaf level. These intermediate pages form the nonleaf level of the
index.

Not every index has a nonleaf level.

SQL Server tables can be organized in two ways:

• As a heap, where data is stored without any specific order.

• As a clustered index, where data is stored in order according to the order ofspecific key fields.

In a clustered index, the actual data pages are the leaf level of the index, because the data is already in order.

Clustered Indexes

As explained before with the telephone directory example, SQL Server can create clustered indexes, ordering
the actual data using the selected key columns as ordering criteria.
Figure 6.13 shows the schema of a clustered index. You can see in the schema that the leaf level of the
index is the actual data. Because the data is physically in order, there is no need for an extra layer with
ordered key values plus pointers to the physical rows.

Figure 6.13. A clustered index has a leaf level and a root node, plus more nonleaf levels if required.

Because a clustered index physically orders the actual data, it is not possible to create more than one
clustered index per table.
When you create a clustered index, perhaps the keys of the index are not unique in the table. If you created a
clustered nonunique index, SQL Server creates an extra hidden-integer field to uniquely identifyevery physical
row.
Caution

198
 Chapter 6. Optimizing Access to Data: Indexes

As you will see in Chapter 7, when you create a PRIMARY KEY, SQL Server creates a
CLUSTERED INDEX by default, unless you specify NONCLUSTERED in the PRIMARY KEY definition.
Thus, creating a PRIMARY KEY with the default settings prevents you from creating a new
clustered index.

Creating and Dropping Clustered Indexes

As Listing 6.11 shows, to create a clustered index you must specify CLUSTERED in the CREATE INDEX
statement.

Listing 6.11 You Must Specify the CLUSTERED Keyword to Create a Clustered Index

USE Northwind
GO

IF OBJECT_ID('OrderDetails') IS NOT NULL

DROP TABLE OrderDetails
GO

-- Create a new table with the same structure and data

-- as the Order Details table

SELECT *
INTO OrderDetails
FROM [Order Details]

-- Create a Clustered index on the new table

CREATE CLUSTERED INDEX C_ORDER_DETAILS_ORDER_PRODUCT

ON OrderDetails (OrderID, ProductID)
To drop a CLUSTERED INDEX, you have to execute a DROP INDEX statement, as in Listing 6.12. Note that
you must qualify the index name with the table name, because index names are not unique in the database.

Listing 6.12 You Must Execute the DROP INDEX Statement to Remove an Index from a Table

USE Northwind
GO

199
Microsoft SQL Server 2000 Programming by Example

DROP INDEX OrderDetails.C_ORDER_DETAILS_ORDER_PRODUCT

Note

If you did not execute the code from Listing 6.11 and you try to execute the code from Listing
6.12, you will get an error message because the OrderDetails table does not exist.

When you drop a clustered index, the leaf level is not removed; otherwise, the data itself would be removed.
Only the root node and the nonleaf levels will be deallocated.

Caution

You cannot specify the CLUSTERED keyword in the DROP INDEX statement.

Specify the UNIQUE keyword to declare a CLUSTERED INDEX as unique, as you can see in Listing 6.13.

Listing 6.13 You Must Specify the UNIQUE Keyword to Create a Unique Clustered Index

USE Northwind
GO

CREATE UNIQUE CLUSTERED INDEX UC_ORDER_DETAILS_ORDER_PRODUCT

ON OrderDetails (OrderID, ProductID)
Note

If you did not execute the code from Listing 6.11 and you try to execute the code from Listing
6.13, you will get an error message, because the OrderDetails table does not exist.

In SQL Server 2000, every column of the key in an index can be sorted using ascending or descending order.
Listing 6.14 shows an example of this new feature.

Listing 6.14 You Can Specify Descending or Ascending Order for Every Column in the Index Key

200
 Chapter 6. Optimizing Access to Data: Indexes

USE Northwind
GO

CREATE INDEX C_Products_Category_Price

ON Products (CategoryID ASC, UnitPrice DESC)

Accessing Data Through Clustered Indexes

If a table is stored as a clustered index and the Query Optimizer decides to use the clustered index to return
the result, SQL Server can access data in that table in different ways:

• As a Clustered Index Scan if the query doesn't restrict the data to be returned.
• As a Clustered Index Seekwhen the query is restricted to a certain number of rows.

Accessing Data Through a Clustered Index Scan

When a Clustered Index Scan is required, it is not guaranteed that the data will be returned in order, because
SQL Server could use the information stored in the IAM pages to access the data more efficiently than
navigating the index. If you need results in order, you must specify an ORDER BY clause in the query. This is
the case of the example of Figure 6.14.

Figure 6.14. SQL Server uses a Clustered Index Scan to execute unrestricted queries.

201
Microsoft SQL Server 2000 Programming by Example

Using a Clustered Index Seek to Execute Point Queries

SQL Server can use aClustered Index Seek to retrieve individual rows. Figure 6.15 shows an example.

Figure 6.15. SQL Server uses a Clustered Index Seek to execute restricted queries.

In this example, SQL Server navigates the index from the root node to the leaf level, applying a binary search
until reaching the required data.

Using a Clustered Index Seek to Execute Queries with a Range Search

Perhaps a more interesting use of a clustered index is to execute queries restricted to a range of values.
Figure 6.16 shows an example of a range search.

Figure 6.16. SQL Server uses a Clustered Index Seek to execute queries that contain a range search.

202
 Chapter 6. Optimizing Access to Data: Indexes

In this example, SQL Server navigates the index from the root node to the leaf level, searching for the lower
limit of the range, and then continues reading from the leaf level until it reaches the upper limit of the range.

Nonclustered Indexes

The leaf level of a nonclustered index contains one entry for every row in the table. Each entry in the index
contains the key columns and a pointer to identify the row in the table where this index entry points.
As explained earlier in this chapter, if the table doesn't have a clustered index, the pointer included on every
entry of the nonclustered index is a binary value. This binary value is a combination of the file number, page
number, and row slot number where the original row is stored. In this way, every entry in the index is
connected to the physical location of the row. If a row changes its physical location, the index pointer must be
modified. This process can produce some overhead on SQL Server.
If the table has a clustered index, the index entries don't point to the physical location of the rows. In this case,
the pointer is the clustered key of the corresponding row. You will see later in this chapter how to access data
using a nonclustered index when the data is stored as a clustered index.

Creating and Dropping Nonclustered Indexes

You can specifythe keyword NONCLUSTERED to create a nonclustered index using the CREATE INDEX
statement. This is the default option. Listing 6.15 shows how to create a NONCLUSTERED index.

Listing 6.15 You Can Specify the NONCLUSTERED Keyword to Create a Nonclustered Index

203
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

CREATE NONCLUSTERED INDEX C_ORDER_DETAILS_PRODUCT

ON [Order Details] (ProductID)
To drop a NONCLUSTERED INDEX,you have to execute a DROP INDEX statement, as in Listing 6.16.

Listing 6.16 You Must Execute the DROP INDEX Statement to Remove an Index from a Table

USE Northwind
GO

DROP INDEX [Order Details].C_ORDER_DETAILS_PRODUCT

When you drop a nonclustered index, the complete index is removed, including the leaf level, the root node,
and the nonleaf levels.
Caution

You cannot specify the NONCLUSTERED keyword in the DROP INDEX statement.

Specify the UNIQUE keyword todeclare a NONCLUSTERED INDEX as unique, as you can see in Listing 6.17.

Listing 6.17 You Must Specifythe UNIQUE Keyword to Create a Unique Nonclustered Index

USE Northwind
GO

IF OBJECT_ID('NewOrders') IS NOT NULL

DROP TABLE NewOrders
GO

SELECT *
INTO NewOrders
FROM Orders
GO

CREATE UNIQUE NONCLUSTERED INDEX UNC_ORDERS_ORDERID

204
 Chapter 6. Optimizing Access to Data: Indexes

ON NewOrders (OrderID)

Accessing Data Through Nonclustered Indexes

The way SQL Serverretrieves data through nonclustered indexes depends on the existence of a clustered
index on the table. As explained earlier in this chapter, the reason for this difference in behavior is to optimize
index maintenance, trying to avoid the continuous physical pointer modifications when data rows must be
moved because of reordering the clustered index.

Data Stored As a Heap

If table data is stored as a heap, the nonclustered index is a binary structure built on top of the actual data.
Figure 6.17 shows an example of index navigation.

Figure 6.17. SQL Server uses a Nonclustered Index Seek to search for rows within a search condition.

SQL Server searches for entries in the nonclustered index, starting from the root node and going down to the
leaf level of the index.
Every entry in the leaf level has a pointer to a physical row. SQL Server uses this pointer to access the page
where the row is located and reads it.

Data Stored As a Clustered Index

205
Microsoft SQL Server 2000 Programming by Example

If table data is stored as a clustered index, it is stored in the order specified by the clustered index definition.
The index is a binary structure built on top of the actual data, which in this case is the leaf level of the
clustered index.
Nonclustered indexes don't have a pointer to the physical position of the rows. They have as a pointer the
value of the clustered index key.
Figure 6.18 shows an example of nonclustered index navigation on top of a clustered index.

Figure 6.18. SQL Server uses a Nonclustered Index Seek to search for rows within a search condition,
and it must navigate the clustered index as well.

If SQL Server decides to use a Nonclustered Index Seek to execute a query, it must follow a process similar
to the one described in the preceding section, with an important difference: It also must navigate the clustered
index to arrive at the physical rows.
You have to consider the extra work when you use a nonclustered index on top of a clustered index. However,
if you consider the number of pages to read, you will consider this solution quite efficient.
The customers table could have 20,000 pages, so to execute a table scan, SQL Server will have to read
20,000 pages.
Indexes don't have many levels, often less than four. The right index image should be a pyramid with a base
of thousands of pages and only three or four pages high. Even if you have two pyramids to navigate, the
number of pages to read is still much smaller than reading through a full table scan.

Covered Queries and Index Intersection

Mentioned earlier in this chapter were different ways to access data, depending on the available indexes. Let's
consider the example of Listing 6.18. If you had an index on (City, CompanyName, ContactName),
this index has every field required to execute the query. In this case, it is not required to access the data
pages, resulting in a more efficient access method. Figures 6.19 and 6.20 show the query plan with and
without this index.

206
 Chapter 6. Optimizing Access to Data: Indexes

Figure 6.19. SQL Server uses the index on (City, CompanyName, ContactName) to cover the query.

Figure 6.20. SQL Server needs access to the data pages to execute the query if an index on (City,
CompanyName, ContactName) doesn't exist.

Listing 6.18 This Query Can Be Coveredby an Index on the Columns (City, CompanyName,
ContactName)

USE Northwind
GO

SELECT CompanyName, ContactName

FROM Customers
WHERE City = 'Madrid'
In these situations, you can say that the selected index covers the query.

207
Microsoft SQL Server 2000 Programming by Example

If you consider the example in Listing 6.19, you can see in the query execution shown in Figure 6.21 that
the index in (City, CompanyName, ContactName) still covers the query, even if you added a new field
called CustomerID.

Figure 6.21. SQL Server uses the index on (City, CompanyName, ContactName) to cover the query
even when you add the field CustomerID.

Listing 6.19 This Query Can Be Covered by an Index on the Columns (City, CompanyName,
ContactName)

USE Northwind
GO

SELECT CustomerID, CompanyName, ContactName

FROM Customers
WHERE City = 'Madrid'
The Customers table has a clustered index on the field CustomerID, and this field is used as a pointer for
every nonclustered index, such as the one you created on (City, CompanyName, ContactName). In
other words, every key entry in the index actually contains four fields: City, CompanyName, ContactName,
and CustomerID. That's why this index covers this query, too.
Trying to cover every query is almost impossible and requires too much space and maintenance cost. SQL
Server 2000 can combine indexes, if required, to execute a query, and this technique is called index
intersection. Listing 6.20 shows a query that selects three fields from the [Order Details] table and
applies two conditions to two of the selected fields. In this case, you could be tempted to create a composite
index on these three fields to cover the query, but if you had already an index on UnitPrice and another
index on OrderID, SQL Server can combine these two indexes to solve the query. Figure 6.22 shows the
query plan of this query.

Figure 6.22. SQL Server combines two indexes to solve the query.

208
 Chapter 6. Optimizing Access to Data: Indexes

Listing 6.20 This Query Can Be Solved by an Index on OrderID and Another Index on UnitPrice

USE Northwind
GO

SELECT OrderID, UnitPrice

FROM [Order details]
WHERE UnitPrice > 15
AND OrderID > 11000

Index Maintenance
SQL Server automatically maintains indexes on tables. Every INSERT, UPDATE, or DELETE operation forces
SQL Server to update the index information to keep it up to date. This maintenance produces some overhead
on these operations.
Any time you insert a new row and you had a clustered index, SQL Server must search for the correct page to
insert this new row. If the page is full, SQL Server decides the best way to insert this row, depending on the
following conditions:

• If the row has to be inserted at the end of the table and the last page doesn't have any free space,
SQL Server must allocate a new page for this new row.
• If the row has to be inserted into an existing page and there is enough free space in the page to
allocate this new row, the row will be inserted in any free space in the page, and the row-offset will be
reordered to reflect the new row order.
• If the row has to be inserted into an existing page and the page is full, SQL Server must split this page
into two. This is done by allocating a new page and transferring 50% of the existing rows to the new
page. After this process, SQL Server will evaluate on which one of these two pages the new row must
be inserted. The row-offset list must be reordered according to new row order.

Figure 6.23 shows the split-page process. This Page Split process produces some overhead for SQL Server
as well.

209
Microsoft SQL Server 2000 Programming by Example

Figure 6.23. To insert a new row into a full page, SQL Server must split the page.

The same process must be done to accommodate a new key into a leaf-level page of a nonclustered index.

Rebuilding Indexes

If you would like to change the index definition, you can use the CREATE INDEX statement with the
DROP_EXISTING option. Listing 6.21 shows an example where you want to convert a nonclustered index on
(OrderID, ProductID) on the [Order Details] table into a clustered index on the same fields.

210
 Chapter 6. Optimizing Access to Data: Indexes

Listing 6.21 You Can Modify Existing Indexes with the CREATE INDEX Statement and the
DROP_EXISTING Option

USE Northwind
GO

CREATE UNIQUE CLUSTERED INDEX UC_ORDER_DETAILS_ORDER_PRODUCT

ON OrderDetails (OrderID, ProductID)
WITH DROP EXISTING
Note

If you did not execute the code from Listing 6.11 and you try to execute the code from Listing
6.21, you will get an error message because the OrderDetails table does not exist.

In this case, other nonclustered indexes must be rebuilt because they must point to the clustered keys, and
not to the physical row locations.

If you had a table with aclustered index and several nonclustered indexes and you wanted to modify the
clustered index definition, you could drop the index and create it again. In this case, the nonclustered indexes
must be rebuilt after the clustered index is dropped, and they must be rebuilt again after the clustered index is
re-created. However, using the DROP_EXISTING option to rebuild the clustered index will save time, because
the nonclustered indexes will be rebuilt automatically just once, instead of twice, and only if you select
different key columns for the clustered index.

Tip

Create the clustered index before the nonclustered indexes. In this way, you can avoid rebuilding
the nonclustered indexes because of the creation of the clustered index.

When an index is rebuilt, the data is rearranged, so external fragmentation is eliminated and internal
fragmentation will be adjusted, as you'll see later in this chapter.

Another alternative to CREATE INDEX... WITH DROP EXISTING is to use the DBCC DBREINDEX.This
statement can rebuild all the indexes of a given table with a single command. This is the preferred way to
rebuild indexes if they are part of a constraint definition. Listing 6.22 shows the statement to rebuild all the
indexes of the [Order Details] table. In this case, indexes are rebuilt with the same definition they were
created.

211
Microsoft SQL Server 2000 Programming by Example

Listing 6.22 Use DBCC DBREINDEXto Rebuild All the Indexes of a Table

USE Northwind
GO

DBCC DBREINDEX ('[Order details]')

Index Fragmentation

Every time a page is split, the index suffers some fragmentation. If fragmentation were important, it would be
necessary to read more pages than normal to retrieve the same information. For read-only tables,
fragmentation must be as minimal as possible. However, for read/write tables it is better to have some
fragmentation to accommodate new rows without splitting too many pages.
You can adjust the fragmentation of an index using the FILLFACTOR option. FILLFACTOR expects a value
between 1 and 100, which specifies the percentage of the page that should be full at the time the index is
created. The actual percentage will be less than or equal to the specified fill factor.
Applying a FILLFACTOR will pack or expand the leaf-level pages to accommodate this new filling factor. For
nonleaf-level pages, there will be one free entry per page. Listings 6.23 and 6.24 show two examples of
rebuilding an index with a new FILLFACTOR. Listing 6.23 uses CREATE INDEX,and Listing 6.24 uses
DBCC DBREINDEX.

Listing 6.23 You Can Specify a New FILLFACTOR for an Existing Index with the CREATE INDEX
Statement and the FILLFACTOR Option

USE Northwind
GO

CREATE NONCLUSTERED INDEX OrderID

ON [Order Details] (OrderID)
WITH DROP_EXISTING, FILLFACTOR = 80

Listing 6.24 Use DBCC DBREINDEX to Rebuild All the Indexes of a Table with a Different FILLFACTOR

212
 Chapter 6. Optimizing Access to Data: Indexes

USE Northwind
GO

DBCC DBREINDEX ('[Order details]', '', 70)

Considering that fragmentation on nonleaf-level pages will be produced only when allocating and deallocating
new pages at leaf level, having a free entry per page should be considered normal. If you expected many new
rows and many new pages in the leaf level, you could be interceding to specify the PAD_INDEX option, which
will apply the FILLFACTOR value to nonleaf-level pages as well. Listing 6.25 shows how to apply this option
to one of the indexes of the [Order Details] table.

Listing 6.25 You Can Specify a New FILLFACTOR for an Existing Index with the CREATE INDEX
Statement and the FILLFACTOR Option, and Apply This FILLFACTOR to the Nonleaf-Level Pages with
the PAD_INDEX Option

USE Northwind
GO

CREATE NONCLUSTERED INDEX OrderID

ON [Order Details] (OrderID)
WITH DROP_EXISTING, FILLFACTOR = 80, PAD_INDEX
If you want to avoid fragmentation on data pages, you can build a clustered index and specify a FILLFACTOR
option with a value of 100. If there is already a clustered index on the table, you can rebuild the index and
specify a value of 100 for FILLFACTOR. Listing 6.26 shows how to pack the data pages on the Products
table by rebuilding the index PK_Products with a FILLFACTOR of 100.

Listing 6.26 Use DBCC DBREINDEX to Pack the Data Pages by Rebuilding the Clustered Index with a
FILLFACTOR of 100

USE Northwind
GO

DBCC DBREINDEX ('Products', PK_Products, 100)

If the clustered index is not required to provide a normal use of the table, and you want to pack data pages,
you can create a clustered index with FILLFACTOR 100 and drop the clustered index when it's created.

Index Statistics

SQL Query Optimizerselects the best strategy based on indexes available for every table to be used in the
query and for specific information about every index. For every index, Query Optimizer gets general
information from the sysindexes table about

213
Microsoft SQL Server 2000 Programming by Example

• The number of data pages in the index (the field dpages)

• The approximate number of rows (the field rowcnt)
• Density of the index (information included in the statblob field)
• Average length of the key (information included in the statblob field)

However, this information is not enough to predict whether the index is useful in a particular query.
Consider the Customers table. To filter customers who live in a specific city, having an index on the City
column can be useful. However, if 95% of your customers live in Toronto, the index on City will be useless
when searching for customers living in Toronto. In this case, a table scan will produce better results.
SQL Server maintains distribution statistics about every index. Statistic information is stored in the statblob
field of the sysindexes table.
SQL Server samples the data to select information organized in ranges of data. For every range, SQL Server
stores

• The number of rows where the key is in the specific range, excluding the maximum value of the range
• The maximum value of the range
• The number of rows wherethe value of the key is equal to the maximum value of the range
• The number of distinct key values in the range, excluding the maximum value of the range

SQL Server calculates the average density of every range as well, dividing number of rows by number of
distinct key values on every range.
Listing 6.27 shows how to use the DBCC SHOW_STATISTICSstatement to get statistics information about
the (Products.SupplierID) index.

Listing 6.27 Use DBCC SHOW_STATISTICS to Get Information About Index Statistics

USE Northwind
GO

DBCC SHOW_STATISTICS (Products, SupplierID)

Statistics for INDEX 'SupplierID'.

Updated Rows Rows Sampled Steps Density Average key length
------------------- ----- ------------- ------ -------- -----------------------
Oct 23 2000 5:16PM 77 77 20 3.3189032E-2 8.0
(1 row(s) affected)

All density Average Length Columns

----------------------- ------------------------ ------------------------------
3.4482758E-2 4.0 SupplierID
1.2987013E-2 8.0 SupplierID, ProductID

214
 Chapter 6. Optimizing Access to Data: Indexes

(2 row(s) affected)

RANGE_HI_KEY RANGE_ROWS EQ_ROWS DISTINCT_RANGE_ROWS AVG_RANGE_ROWS

------------ ---------- ------- ------------------- --------------
1 0.0 3.0 0 0.0
2 0.0 4.0 0 0.0
4 3.0 3.0 1 3.0
5 0.0 2.0 0 0.0
6 0.0 3.0 0 0.0
7 0.0 5.0 0 0.0
8 0.0 4.0 0 0.0
9 0.0 2.0 0 0.0
10 0.0 1.0 0 0.0
11 0.0 3.0 0 0.0
12 0.0 5.0 0 0.0
13 0.0 1.0 0 0.0
17 9.0 3.0 3 3.0
19 2.0 2.0 1 2.0
20 0.0 3.0 0 0.0
22 2.0 2.0 1 2.0
24 3.0 3.0 1 3.0
26 2.0 2.0 1 2.0
27 0.0 1.0 0 0.0
29 2.0 2.0 1 2.0

(20 row(s) affected)

DBCC execution completed. If DBCC printed error messages,

contact your system administrator.
SQL Server by defaultautomatically creates statistics when you create an index and automatically maintains
these statistics as you add rows to the base table. It is possible, but not advisable, to avoid automatic
statistics maintenance by setting an option at database level, as shown in Listing 6.28.

Listing 6.28 By Using sp_dboption, It Is Possible to Avoid Automatic Statistics Creation and
Maintenance for the Entire Database

USE Northwind
GO

-- Change the database setting to avoid

-- statistics creation and maintenance

EXEC sp_dboption 'Northwind', 'auto create statistics', 'false'

EXEC sp_dboption 'Northwind', 'auto update statistics', 'false'

-- To test the present settings

PRINT 'After changing to manual statistics maintenance'

EXEC sp_dboption 'Northwind', 'auto create statistics'

215
Microsoft SQL Server 2000 Programming by Example

EXEC sp_dboption 'Northwind', 'auto update statistics'

After changing to manual statistics maintenance

OptionName CurrentSetting
----------------------------------- --------------
auto create statistics off

OptionName CurrentSetting
----------------------------------- --------------
auto update statistics off
It is possible to create statistics on individual columns, or groups of columns, without creating an index. These
statistics can be helpful for Query Optimizer to select efficient query execution strategies. Listing 6.29 shows
different ways to create statistics.

Listing 6.29 It Is Possible to Create Statistics on Nonindexed Columns

USE Northwind
GO

-- To create statistics in an individual column

CREATE STATISTICS stProductsStock

ON Products(UnitsInStock)

-- To create statistics in a group of columns

CREATE STATISTICS stProductsStockOrder

ON Products(UnitsInStock, UnitsOnOrder)

-- To create single column statistics for all eligible columns

-- on all user tables in the current database

EXEC sp_createstats
To retrieve information aboutavailable statistics in a table, you can use the sp_helpstats system stored
procedure. Listing 6.30 shows an example of sp_helpstats execution.

Listing 6.30 Use the sp_helpstats System Stored Procedure to Get Information About Statistics

216
 Chapter 6. Optimizing Access to Data: Indexes

USE Northwind
GO

EXEC sp_helpstats
Products
statistics_name statistics_keys
------------------------------ ------------------------------

_WA_Sys_UnitPrice_07020F21 UnitPrice
QuantityPerUnit QuantityPerUnit
ReorderLevel ReorderLevel
stProductsStock UnitsInStock
stProductsStockOrder UnitsInStock, UnitsOnOrder
UnitsOnOrder UnitsOnOrder

Getting Information About Indexes

You can use the sp_help system stored procedure to retrieve general information about a table, including
the list of available indexes. Using the sp_helpindex system stored procedure you can get only the list of
indexes available for a specific table, as in the example included in Listing 6.31.

Listing 6.31 Use the sp_helpindex System Stored Procedure to Retrieve Information About Indexes
in a Table

USE Northwind
GO

EXEC sp_helpindex customers

index_name index_description index_keys

------------ ------------------------------------------------- ------------

217
Microsoft SQL Server 2000 Programming by Example

City nonclustered located on PRIMARY City

CompanyName nonclustered located on PRIMARY CompanyName
Contact nonclustered located on PRIMARY ContactName
PK_Customers clustered, unique, primary key located on PRIMARY CustomerID
PostalCode nonclustered located on PRIMARY PostalCode
Region nonclustered located on PRIMARY Region
To get specific information about individual index properties, you can use the INDEXPROPERTY system
function as demonstrated in Listing 6.32.

Listing 6.32 Use the INDEXPROPERTY Function to Retrieve Information About Any Index

USE Northwind
GO

-- To retrieve number of index levels

SELECT INDEXPROPERTY(OBJECT_ID('Products'), 'PK_Products', 'IndexDepth')

AS 'Index Levels'
-- To determine if the index is clustered

SELECT CASE
INDEXPROPERTY(OBJECT_ID('Products'), 'PK_Products', 'IsClustered')
WHEN 0 THEN 'No'
ELSE 'Yes'END as 'Is Clustered'

-- To determine if it is not a real index

-- So it cannot be used for data access
-- because it contains only statistics
SELECT CASE
INDEXPROPERTY(OBJECT_ID('Products'), 'PK_Products', 'IsStatistics')
WHEN 0 THEN 'No'
ELSE 'Yes'END as 'Is Statistics only'

-- To know if the index is unique

SELECT CASE
INDEXPROPERTY(OBJECT_ID('Products'), 'PK_Products', 'IsUnique')
WHEN 0 THEN 'No'
ELSE 'Yes'END as 'Is Unique'

218
 Chapter 6. Optimizing Access to Data: Indexes

Index Levels
------------
1

Is Clustered
------------
Yes

Is Statistics only
------------------
No

Is Unique
---------
Yes

Indexes on Computed Columns

In SQL Server 2000, you can create computed columns in a table. These columns don't use storage space,
and SQL Server maintains them automatically whenever the underlying data changes.
You can create a computed column SalePrice in the [Order Details] table to get the total sale value of
every row, considering the unit price and quantity. To speed up the process of searching, or sorting, for this
SalePrice column, you can create an index on this computed column, and Query Optimizer might use it, if
necessary. Listing 6.33 shows the complete process.

Listing 6.33 It Is Possible to Create Indexes on Computed Columns

USE Northwind
GO

-- Create the computed column SalePrice

ALTER TABLE [order details]
ADD SalePrice AS (UnitPrice * Quantity)

-- Create an index on SalePrice

CREATE INDEX ndxSale ON [order details] (SalePrice)

To create an index on a computed column, you have to check the following requirements:

219
Microsoft SQL Server 2000 Programming by Example

• The expression defining the computed column must be deterministic.An expression is deterministic if
it always produces the same results for the same arguments. Every function referenced in the
expression must be deterministic, too.

An example of a nondeterministic expression is (Month(GetDate()) because it uses a

nondeterministic function, GetDate, which changes every time you call it.

• The expression must be precise. To be precise, the expression cannot use the float or real data types,
or any combination of them, even if the final result uses a precise data type.

If you define the SalePrice computed column as (UnitPrice * Quantity * (1 -

Discount)), the expression is not precise because it uses an imprecise field: Discount. In this
case, you cannot create an index on this computed column.

• Because connection settings can affect results, the connection that creates the index in the computed
column, and every connection that modifies data affecting this index, must have settingsaccording to
the following list:
•
• SET ANSI_NULLS ON
• SET ANSI_PADDING ON
• SET ANSI_WARNINGS ON
• SET ARITHABORT ON
• SET CONCAT_NULS_YIELDS_NULL ON
• SET QUOTED_IDENTIFIER ON
• SET NUMERIC_ROUNDABORT OFF

Indexed Views
SQL Server 2000 can create indexes on views. This functionality is implemented by extensions in the CREATE
VIEW and CREATE INDEX statements. If a view is not indexed, it doesn't use any storage space. Whenever
you use the view in a Transact-SQL statement, SQL Server merges the view definition with the statement to
produce a single execution plan and it directly accesses the underlying tables on which the view is defined.
After a view is indexed, its index needsstorage space, as any standard index.
The process of creating an indexed view is as follows:

1. Create the view with SCHEMABINDING, which prevents modifications on the definition of referenced
objects.
2. Create a clustered index on the view to physically save the view results in a clustered index structure
where the leaf level will be the complete resultset of the view. The index key should be as short as
possible to provide good performance.
3. Create nonclustered indexes, if required.

After the creation of the clustered index, the view is stored like a clustered index for a table, but this
information is maintained automatically whenever data changes in the underlying tables.
If you reference the view, Query Optimizer will use the indexed view directly, and it will be unnecessary to
access the underlying tables.
SQL Server 2000 Enterprise Edition can use indexed views to optimize the execution of queries that don't
reference the indexed views explicitly, improving execution performance.
Listing 6.34 shows how to create an index on a view. Figure 6.24 shows how SQL Server uses the view
definition to access the base tables directly. Figure 6.25 shows that by having an indexed view, SQL Server
can avoid accessing the base tables, reducing the amount of IO required to execute the query.

Figure 6.24. When using nonindexed views, SQL Server accesses data directly from tables.

220
 Chapter 6. Optimizing Access to Data: Indexes

Figure 6.25. When using indexed views, SQL Server doesn't require access to the tables.

Listing 6.34 In SQL Server 2000, You Can Create Indexes on Views

USE Northwind
GO

-- Create the view

CREATE VIEW Customers_UK

WITH SCHEMABINDING
AS
SELECT CustomerID, CompanyName, ContactName, Phone
FROM dbo.Customers
WHERE Country = 'UK'

221
Microsoft SQL Server 2000 Programming by Example

-- Test how a normal query uses the view

SELECT CustomerID, CompanyName, ContactName, Phone

FROM Customers
WHERE Country = 'UK'
AND CompanyName like 'C%'

-- Create a clustered index on the view

CREATE UNIQUE CLUSTERED INDEX CustUK ON Customers_UK (CustomerID)

-- Create a nonclustered index on the CompanyName field on the view

CREATE NONCLUSTERED INDEX CustUKCompany ON Customers_UK (CompanyName)

-- Test how a normal query uses the view after indexing the view

SELECT CustomerID, CompanyName, ContactName, Phone

FROM Customers
WHERE Country = 'UK'
AND CompanyName like 'C%'
Not every view can be indexed because some requirements must be met. Some requirements affect the
definition of the view, and others the creation of the index. To create a view that can be indexed, it is
necessary to meet the following requirements:

• The ANSI_NULLS option must be set to ON to create the base tables and the view.
• The QUOTED_IDENTIFIER must be set to ON to create the view.
• The view must reference only base tables, from the same database and owner.
• The view must be created with the SCHEMABINDING option set to prevent changes on the underlying
objects. If the view uses a user-defined function, this must be created as well with SCHEMABINDING.
• To avoid ambiguity, objects must be referenced with two part names, owner, and object name. Three-
or four-part names are not allowed because the view cannot reference objects from other databases
or servers.
• All the expressions used in the view definition must be deterministic, and the expressions used in key
columns must be precise, as explained earlier, for indexes on computed columns.
• The view must specifically name all columns, because SELECT * is not allowed.
• You cannot use a column more than once in the SELECT clause, unless every time the column was
used as a part of a complex expression.
• It is not allowed to use subqueries, and that includes derived tables in the FROM clause.
• You cannot use Rowset functions, such as OPENROWSET, OPENQUERY, CONTAINSTABLE, or
FREETEXTTABLE.
• It cannot contain the UNION operator.
• It cannot contain outer or self-joins.
• It cannot contain the ORDER BY clause, and that includes the use of the TOP clause.
• You cannot use the DISTINCT keyword.
• The only aggregate functions allowed are COUNT_BIG and SUM. To use the SUM function, you must
select COUNT_BIG as well, and you must specify a GROUP BY clause.
• SUM cannot be used on a nullable column or expression.
• It cannot use full-text functions.
• COMPUTE or COMPUTE BY clauses are not allowed.
• The view cannot contain BLOB columns, as text, ntext, and image.

To create an index in a view, some more requirements must be met:

• Only the owner of the view can create indexes on the view.
• The connection settings must be the same as the settings for indexes on computed columns.

222
 Chapter 6. Optimizing Access to Data: Indexes

• If the view has a GROUP BY clause, only the columns specified in the GROUP BY clause can
participate in the index key.

Caution

To modify data that is used in an indexed view, you must set the seven connection settings in the
same way as for the index creation, or the operation will fail.

Note

Using the SCHEMABINDING option means you cannot alter or drop the base objects, and to do so,
you must drop the view to break the schema binding.

Index Tuning Wizard

Deciding which indexingstrategy to apply is not an easy task. Different queries can be optimized in different
ways using different indexes. To decide which is the best indexing strategy, it would be necessary to consider
statistically which strategy produces the best global performance.
The Index Tuning Wizard does just that. It uses a trace from SQL Profiler to analyze, propose, and apply, if
required, the best indexing strategy for the actual database workload.
With the integration of the Index Tuning Wizard in SQL Query Analyzer, it is possible to optimize a single
query or batch in Query Analyzer, without creating a trace with SQL Profiler. This can be considered as a
provisional solution, to speed up the process of one specific query. However, the best approach is still to use
a trace that is representative of the actual database workload.
The process of using the Index Tuning Wizard is almost the same in both cases; that's why you will learn how
to optimize a single query from Query Analyzer in this chapter. The query to optimize is represented in Listing
6.35.

Listing 6.35 You Can See How to Optimize the Following Query Using Index Tuning Wizard

USE Northwind
GO

SELECT OD.OrderID, O.OrderDate,

C.CompanyName, P.ProductName,
OD.UnitPrice, OD.Quantity, OD.Discount
FROM [Order Details] AS OD
JOIN [Orders] AS O
ON O.OrderID = OD.OrderID
JOIN [Products] AS P
ON P.ProductID = OD.ProductID
JOIN [Customers] AS C
ON C.CustomerID = O.CustomerID

223
Microsoft SQL Server 2000 Programming by Example

WHERE Country = 'UK'

Write the query in Query Analyzer and select the complete query with the mouse.
Open the menu Query— Index Tuning Wizard, or press Ctrl+I. The Index Tuning Wizard will start and it will
show the Welcome form. Click Next in this form and you will see the form shown in Figure 6.26.

Figure 6.26. The Index Tuning Wizard has different analysis modes.

In this form, you can decide whether you want to keep existing indexes; for the example, uncheck the check
box so you don't consider any index as fixed.
You can select whether the wizard can consider the creation of indexed view. Leave the check box checked.
Select Thorough Tuning Mode to get better results. Click Next.
To Specify Workload, leave the SQL Query Analyzer Selection option set and click Next. If you followed the
preceding instructions, the Index Tuning Wizard will be as shown in Figure 6.27.

Figure 6.27. The Index Tuning Wizard can analyze individual tables or a group of tables.

224
 Chapter 6. Optimizing Access to Data: Indexes

Select the Orders, Products, Customers, and Order Details tables to tune, and then click Next.
The Index Tuning Wizard starts analyzing and, after a few minutes, shows index recommendations. You can
review and select which recommendations are valid for you, according to your experience and the knowledge
of the actual data. Note that the Index Tuning Wizard estimates the relative performance improvement when
applying the new index strategy.
Click Next. Now either you can apply the changes directly or scheduled to a certain time, or you can script
these changes for further analysis. Select to save the script and provide a filename. You should receive a
script similar to that in Listing 6.36.
Click Finish to end the wizard.

Listing 6.36 These Are the Recommendations of the Index Tuning Wizard to Optimize the Query of
Listing 6.35

/* Created by: Index Tuning Wizard */

/* Date: 25/10/2000 */
/* Time: 23:36:33 */
/* Server Name: BYEXAMPLE */
/* Database Name: Northwind */
USE [Northwind]
go

SET QUOTED_IDENTIFIER ON
SET ARITHABORT ON
SET CONCAT_NULL_YIELDS_NULL ON
SET ANSI_NULLS ON
SET ANSI_PADDING ON

225
Microsoft SQL Server 2000 Programming by Example

SET ANSI_WARNINGS ON
SET NUMERIC_ROUNDABORT OFF
go

DECLARE @bErrors as bit

BEGIN TRANSACTION
SET @bErrors = 0

DROP INDEX [dbo].[Orders].[ShipPostalCode]

DROP INDEX [dbo].[Orders].[ShippedDate]
DROP INDEX [dbo].[Orders].[CustomersOrders]
DROP INDEX [dbo].[Orders].[OrderDate]
DROP INDEX [dbo].[Orders].[CustomerID]
DROP INDEX [dbo].[Orders].[ShippersOrders]
DROP INDEX [dbo].[Orders].[EmployeesOrders]
DROP INDEX [dbo].[Orders].[EmployeeID]
CREATE NONCLUSTERED INDEX [Orders7]
ON [dbo].[Orders] ([OrderID] ASC, [CustomerID] ASC, [OrderDate] ASC )

IF( @@error <> 0 ) SET @bErrors = 1

IF( @bErrors = 0 )
COMMIT TRANSACTION
ELSE
ROLLBACK TRANSACTION

BEGIN TRANSACTION
SET @bErrors = 0

DROP INDEX [dbo].[Order Details].[ProductID]

DROP INDEX [dbo].[Order Details].[orderID]
DROP INDEX [dbo].[Order Details].[price]
DROP INDEX [dbo].[Order Details].[OrdersOrder_Details]
DROP INDEX [dbo].[Order Details].[ProductsOrder_Details]
DROP INDEX [dbo].[Order Details].[ndxSale]

IF( @bErrors = 0 )
COMMIT TRANSACTION
ELSE
ROLLBACK TRANSACTION

BEGIN TRANSACTION
SET @bErrors = 0

DROP INDEX [dbo].[Customers].[Region]

DROP INDEX [dbo].[Customers].[CompanyName]
DROP INDEX [dbo].[Customers].[Contact]
DROP INDEX [dbo].[Customers].[ndx_Customers_City]
DROP INDEX [dbo].[Customers].[PostalCode]
DROP INDEX [dbo].[Customers].[City]

IF( @bErrors = 0 )
COMMIT TRANSACTION
ELSE
ROLLBACK TRANSACTION

BEGIN TRANSACTION
SET @bErrors = 0

DROP INDEX [dbo].[Products].[C_Products_Category_Price]

226
 Chapter 6. Optimizing Access to Data: Indexes

DROP INDEX [dbo].[Products].[CategoriesProducts]

DROP INDEX [dbo].[Products].[SuppliersProducts]
DROP INDEX [dbo].[Products].[CategoryID]
DROP INDEX [dbo].[Products].[ProductName]
DROP INDEX [dbo].[Products].[SupplierID]

IF( @bErrors = 0 )
COMMIT TRANSACTION
ELSE
ROLLBACK TRANSACTION

/* Statistics to support recommendations */

CREATE STATISTICS [hind_325576198_1A_2A_3A_4A_5A]

ON [dbo].[order details] ([OrderID], [ProductID], [UnitPice],
[Quantity], [Discount])
Now you can modify this script to meet other requirements and execute against the server to apply the
changes.

Summary
Understanding the way SQL Server 2000 stores, modifies, and retrieves data helps you to design databases
for optimal performance.
Which index strategy to apply has a big impact on the overall performance of the database. Different uses
require different indexing strategies.
Decision Support Systems (DSSs) are based on reading operations, which read many rows to produce a
report, usually with data aggregation. Online Transaction Processing Systems (OLTP) are based in fast
access to a small number of rows, producing modifications on the data, forcing index maintenance.
Many systems are a mixture of DSS and OLTP operations. SQL Profiler can help us to determine the actual
workload, and the Index Tuning Wizard can suggest an efficient strategy to apply.
New SQL Server 2000 features, such as indexes on computed columns and indexed views, could speed up
the execution of complex queries in many scenarios.

What's Next?

Indexes play an important role in some types of data integrity, and Chapter 7, "Enforcing Data Integrity,"
shows how SQL Server uses indexes to enforce primary key and unique constraints.
Having an adequate index strategy provides the basis for an efficient database system. The way SQL Server
produces query plans to execute stored procedures and triggers depends on the available indexes. Stored
procedures and triggers are covered in Chapter 8, "Implementing Business Logic: Programming
Stored Procedures," and Chapter 9, "Implementing Complex Processing Logic: Programming
Triggers."
User-defined functions benefit from indexes as well, because their query plan depends on the existence of
suitable indexes. In Chapter 10, "Enhancing Business Logic: User-Defined Functions (UDF)," you
learn how to define user-defined functions to solve business problems as a flexible alternative to using views
or stored procedures.

227
 Chapter 7. Enforcing Data Integrity

Chapter 7. Enforcing Data Integrity

Databases are as useful as the quality of the data they contain. The quality of the data is determined by many
different factors, and every phase in the life cycle of a database contributes to the ultimate quality of the data.
The logical database design, the physical implementation, the client applications, and the final user entering
the data in the database all have key roles in the final quality of the data.
Data integrity is an important factor that contributes to the overall quality of the data, and SQL Server 2000, as
a relational database management system, provides different ways to enforce data integrity. This chapter
teaches you

• Types of data integrity and how SQL Server helps you enforce them
• How to uniquely identify rows in a table using PRIMARY KEY and UNIQUE constraints
• How to validate values in new rows using CHECK constraints and RULE objects
• How to provide default values for columns using DEFAULT constraints and DEFAULT objects
• How to enforce referential integrity between tables using FOREIGN KEY constraints and how to use
cascade referential integrity
• Which constraint is appropriate in each case

Types of Data Integrity

Consider a commercial database in which you store information about products, customers, sales, and so on.
You can measure the integrity of the data contained in that database in different ways:

• Is the information related to one specific product stored in a consistent way, and is it easy to retrieve?
• Do you have different products with the same name or code?
• Can you identify your customers in a unique manner, even if they have the same name?
• Are there any sales that are not related to a specific customer?
• Do you have any sales of nonexistent products?
• Do you have a consistent price structure for your products?

To guarantee the integrity of the data contained in a database, you should ensure

• That every individual value conforms to specific business rules (domain integrity)
• That every object can be uniquely and unequivocally identified (entity integrity)
• That related data is properly connected (relational integrity)

Domain Integrity

Applying business rules to validate the data stored in the database enforces domain integrity. Your database
application has different ways to validate the data entered in a database, such as the following:

• The column data type restricts the values you can enter in this column. This prevents you from
entering textual descriptions in data columns or dates in price columns.
• The column length enforces the length of the data you can enter in a specific column.
• You can enforce the minimum and maximum length for any given value. For example, you can
determine that product codes should be at least five characters long and fewer than ten.
• You can restrict data that conforms to a specific format. This can be useful to validate ZIP or postal
codes or telephone numbers.
• It might be useful to restrict the valid range of values. You can limit the value to enter, as a date of
birth of a new bank customer, to dates between 1880-01-01 and today's date, to avoid possible
mistakes.
• The business meaning of a column might need to enforce that the values entered into the column
must be one of the possible values in a fixed list. Your sales application, for example, might classify
customers as individual customers, public institutions, or businesses.

229
Microsoft SQL Server 2000 Programming by Example

Entity Integrity

Every real object should be easily identified in a database. It is difficult to refer to a customer as "the customer
who lives in Seattle, has four children and 300 employees, is 40 years old, his first name is Peter, and his
telephone number ends with 345." For humans this data could be enough for searching our memory and
identifying a customer. However, for a computer program, such as SQL Server 2000, this way of customer
identification will force SQL Server to apply different conditions in sequence, one condition per attribute.
Perhaps it would be easy to identify every customer by a single unique value, such as 25634, stored in a
identification column, such as CustomerID. In this way, to search for a customer, SQL Server will have to
apply a simple condition: CustomerID = 25634.
This is especially important if you want to relate information from other entities, because every relationship
should be based on the simplest possible link.

Referential Integrity

Relational databases are called "relational" because the data units stored in them are linked to each other
through relationships:

• Customers have sales representatives who take care of them

• Customers place orders
• Orders have order details
• Every item in an order references a single product
• Products are organized by categories
• Products are stored in warehouses
• The products come from suppliers

You must make sure that all these links are well established, and that our data does not contain any orphan
data that is impossible to relate to the rest of the data.

User-Defined Integrity

In some situations, you are required to enforce complex integrity rules that are impossible to enforce by using
standard relational structures. In these situations, you can create stored procedures, triggers, user-defined
functions, or external components to achieve the extra functionality you require.

Enforcing Integrity: Constraints (Declarative Data Integrity)

SQL Server uses Transact-SQL structures to enforce data integrity. You can create these structures during
table creation or by altering the table definition after the creation of the table and even after data has been
inserted into the table.
To enforce entity integrity, SQL Server uses PRIMARY KEY and UNIQUE constraints, UNIQUE indexes, and
the IDENTITY property. UNIQUE indexes are covered in Chapter 6, "Optimizing Access to Data:
Indexes."
Note

The IDENTITY function is used to create an IDENTITY field in a table created by using the
SELECT INTO statement.

For domain integrity, SQL Server provides system-supplied and user-defined data types, CHECK constraints,
DEFAULT definitions, FOREIGN KEY constraints, NULL and NOT NULL definitions, and RULE and DEFAULT
objects. Data types were covered in Chapter 2, "Elements of Transact-SQL."

230
 Chapter 7. Enforcing Data Integrity

Note

DEFAULT definitions are called DEFAULT constraints as well. Because DEFAULT constraints don't
restrict the values to enter in a column but rather provide values for empty columns in INSERT
operations, SQL Server 2000 calls them properties, instead of constraints, reflecting their purpose
more accurately.

To enforce referential integrity, you can use FOREIGN KEY and CHECK constraints. Using complex structures,
such as stored procedures, triggers, and user-defined functions as part of constraint definitions, it is possible
to enforce complex business integrity rules.

Primary Keys

To enforce entity integrity in a given table, select the fields or combination of fields that uniquely identifies
every row.
Tables usually represent an entity, such as Products, and the primary key can be as simple as one single
field, which contains the unique identifiers for objects of that entity. You could consider the name of a product
as the unique identifier, but usually names are not unique in a table. That's why to uniquely identify every
product, you introduce a unique value in the field ProductID. This avoids ambiguity when referring to one
specific product.
Following a pure relational database design, you should identify which set of natural attributes uniquely
identifies every object of that entity. In some cases, this set will be a single attribute although, in most cases,
this set will be a collection of different attributes. In a pure relational design, you should define the PRIMARY
KEY on this set of attributes. However, you can create an artificial attribute, called a surrogate key, that
uniquely identifies every row, working as a simplification of the natural PRIMARY KEY.
Note

Whether to use a natural PRIMARY KEY or a surrogate artificial key as a PRIMARY KEY depends
on the implementation of the particular database product you use.

The recommendations in this chapter refer to SQL Server 2000. If you need to implement your
database on different database systems, we recommend you follow a more standard relational
approach.

Providing a new artificial integer column to be used as a primary key has some advantages. It is a short
value— only 4 bytes— and SQL Server uses this value very efficiently on searching operations and joining
tables through this field.

You can define the primary key constraint at column level, after the column definition, or at table level, as part
of the table definition. Another possibility is to create the table first and add the primary key constraint later,
using the ALTER TABLE statement.

Tip

231
Microsoft SQL Server 2000 Programming by Example

Providing a user-friendly name to the primary key constraints will help when referring to the
constraint in other statements and to identify the constraint after receiving a message from SQL
Server.

Because there is only a PRIMARY KEY constraint per table, a recommended naming standard for a
PRIMARY KEY can be PK_TableName.

You can use the code in Listing 7.1 to create a PRIMARY KEY in a single column of a table, using the
CREATE TABLE statement.

Listing 7.1 Define a PRIMARY KEY in a Single Column

-- Define a PRIMARY KEY in a single column

-- using the DEFAULT constraint name

CREATE TABLE NewRegions (

RegionID int NOT NULL
PRIMARY KEY NONCLUSTERED,
RegionDescription nchar (50) NOT NULL ,
)
GO

DROP TABLE NewRegions

GO

-- Define a PRIMARY KEY in a single column

-- specifying the constraint name

CREATE TABLE NewRegions (

RegionID int NOT NULL
CONSTRAINT PK_NewRegions
PRIMARY KEY NONCLUSTERED,
RegionDescription nchar (50) NOT NULL ,
)
GO

DROP TABLE NewRegions

GO

-- Define a PRIMARY KEY in a single column

-- specifying the constraint name
-- and defining the constraint at table level

CREATE TABLE NewRegions (

RegionID int NOT NULL,

232
 Chapter 7. Enforcing Data Integrity

RegionDescription nchar (50) NOT NULL ,

CONSTRAINT PK_NewRegions
PRIMARY KEY NONCLUSTERED (RegionID),
)
GO

DROP TABLE NewRegions

GO

-- Define a PRIMARY KEY in a single column

-- specifying the constraint name
-- and defining the constraint at table level
-- using the ALTER TABLE statement

CREATE TABLE NewRegions (

RegionID int NOT NULL,
RegionDescription nchar (50) NOT NULL)
GO
ALTER TABLE NewRegions
ADD CONSTRAINT PK_NewRegions
PRIMARY KEY NONCLUSTERED (RegionID)
GO

DROP TABLE NewRegions

GO
To define a primary key in a column, the column cannot accept nulls.
Because primary key values must be unique, SQL Server creates a UNIQUE index to help check whether new
values already exist in the column. Without an index, as explained in Chapter 6, SQL Server would have to
read every single row to determine the uniqueness of each new value. This index takes the same name as the
primary key constraint, and it cannot be removed without removing the constraint.
You can provide properties for the index, such as CLUSTERED, NONCLUSTERED, or FILLFACTOR, in the
constraint definition. Listing 7.2 shows how to declare the index of a PRIMARY KEY as NONCLUSTERED and
with a FILLFACTOR of 70%. Index properties were detailed in Chapter 6.

Listing 7.2 Define a PRIMARY KEY and Select Properties for Its UNIQUE INDEX

-- Define a PRIMARY KEY in a single column

-- and create the index a nonclustered
-- and 70% FillFactor

CREATE TABLE NewRegions (

RegionID int NOT NULL
PRIMARY KEY NONC
LUSTERED WITH FILLFACTOR = 70,
RegionDescription nchar (50) NOT NULL ,
)
GO

DROP TABLE NewRegions

GO
Caution

233
Microsoft SQL Server 2000 Programming by Example

It is important to understand that a constraint is a definition that enforces data validation, whereas
an index is a storage structure that speeds up the searching processes. A primary key is not an
index, but it uses an index for performance reasons.

It is possible to create a primary key in a group of columns. In this case, none of the columns in the primary
key can accept nulls.

Caution

You can use up to 16 columns in a PRIMARY KEY definition, as long as the total key size is less
than 900 bytes. It is advisable to keep the key size as short as possible.

Listing 7.3 shows how to create a PRIMARY KEY constraint on the combination of the ProductID and the
SaleID columns using two different versions: The first version creates the PRIMARY KEY directly in the
CREATE TABLE statement; the second version creates the PRIMARY KEY using the ALTER TABLE
statement, after the creation of the table.

Listing 7.3 Defining a Multicolumn PRIMARY KEY

-- Define a composite PRIMARY KEY in two columns

-- specifying the constraint name

CREATE TABLE ProductSales (

ProductID int NOT NULL,
SaleID int NOT NULL,
Quantity int NOT NULL,
Price money NOT NULL,
Description varchar(200) NULL,
CONSTRAINT PK_ProductSales
PRIMARY KEY NONCLUSTERED (ProductID, SaleID)
)
GO

DROP TABLE ProductSales

GO

-- Define a composite PRIMARY KEY in two columns

-- using the ALTER TABLE statement

234
 Chapter 7. Enforcing Data Integrity

CREATE TABLE ProductSales (

ProductID int NOT NULL,
SaleID int NOT NULL,
Quantity int NOT NULL,
Price money NOT NULL,
Description varchar(200) NULL)

ALTER TABLE ProductSales

ADD CONSTRAINT PK_ProductSales
PRIMARY KEY NONCLUSTERED (ProductID, SaleID)
GO

DROP TABLE ProductSales

GO
SQL Server can automatically provide values for a primary key defined in a single column in the following
ways:

• By using the IDENTITY property for the column to specify the seed and increment values.
• By using the uniqueidentifier data type combined with the NEWID function as a DEFAULT
definition to supply automatic GUID (Global Unique Identifier) values.
• By declaring a user-defined function as a DEFAULT definition, to provide unique values to the column.
• By declaring the column using a data type timestamp or rowversion. Although this is a technically
correct option, it is not a recommended solution because these values change whenever the row is
modified. Having a PRIMARY KEY defined in a timestamp column will make this constraint unsuitable
as a reference for FOREIGN KEY constraints defined on related tables.

Caution

For many relational database systems, following the ANSI SQL-92 standard, a timestamp column
holds the data and time of the latest modification of a row. SQL Server 2000 implements
timestamp values in a different way, and it is not a date and time value. This is the reason for the
new term: rowversion.

You should use the rowversion data type, instead of the timestamp, because future versions
of SQL Server could implement the timestamp data type in a different way, perhaps in the way
suggested in the ANSI SQL-92 standard.

Caution

If you use a rowversion or timestamp data type for a primary key column and you don't specify
NONCLUSTERED, the row will physically move every time the row changes because SQL Server
will change the value of the column automatically.

You can use Enterprise Manager to define a PRIMARY KEY in a table. To do it, right-click the table and select
Design Table to display the Design Table form.

235
Microsoft SQL Server 2000 Programming by Example

Figure 7.1 shows the Design Table form in which you can see the key icon on the PRIMARY KEY field. To
delete the PRIMARY KEY, click the Set Primary Key icon on the toolbar.

Figure 7.1. Use the Design Table form to edit the PRIMARY KEY.

To specify more properties about the PRIMARY KEY, you can open the Properties form by clicking the Table
and Index Properties icon on the toolbar. In the Indexes/Keys tab, you can modify or delete the PRIMARY KEY
definition, but you cannot create a PRIMARY KEY using this form.
Figure 7.2 shows the Properties form in which you can see how to change the FILLFACTOR or the
CLUSTERED property of the index associated to the PRIMARY KEY.

Figure 7.2. Use the index Properties form to change properties of a PRIMARY KEY index.

236
 Chapter 7. Enforcing Data Integrity

UNIQUE Constraints

You can create a PRIMARY KEY to enforce uniqueness in a field or group of fields, but you can have only one
PRIMARY KEY per table.
If you require enforcing uniqueness in other columns, you can create a UNIQUE constraint. For example, you
can have a PRIMARY KEY defined in the EmployeeID surrogate key in the Employees table, but you want to
enforce uniqueness on

• Social security number if all your employees have one

• National identification number in countries where this is the standard identification
• Passport number on overseas projects
• Driver's license number for your Drivers table.

Full Name is not usually a good candidate for a UNIQUE constraint, because you can have many employees
with the same name.
ProductName can be a good candidate, unless you have multiple products with the same name and different
size, color, or any other attribute. In this case, you either provide a name with the full description, including all
these attributes, or consider the combination of name and attributes as unique in your table.
A UNIQUE constraint is similar to a PRIMARY KEY, but you can have more than one UNIQUE constraint per
table. When you declare a UNIQUE constraint, SQL Server creates a UNIQUE index to speed up the process
of searching for duplicates. In this case, the index defaults to NONCLUSTERED, because you can have more
than one UNIQUE constraint but only one clustered index.
Note

237
Microsoft SQL Server 2000 Programming by Example

The number of UNIQUE constraints in a table is limited by the maximum number of indexes per
table, which is 249 nonclustered indexes plus one possible clustered index, as mentioned in
Chapter 6.

Contrary to PRIMARY KEY constraints, UNIQUE constraints can accept NULL values, but just one. If the
constraint is defined in a combination of fields, every field can accept NULL and have some NULL values on
them, as long as the combination of values is unique.

Suppose you declared a UNIQUE constraint in the combination of the fields (HouseNumber, HouseName,
Apartment, Address, City). In this case, you can have many rows where any of these fields are NULL,
but only one row can have all these fields NULL. In this example, the table can have different combinations of
NULL entries:

• Only one row with NULL in all these fields.

• Many rows with NULL value in HouseNumber, HouseName, Apartment, Address or City, as
long as the other fields are not NULL at the same time.
• Many rows with NOT NULL values in these fields.

If you want to enforce uniqueness in a column that accepts many NULL values, a UNIQUE constraint cannot
help you. In this case, you should use a trigger or a CHECK constraint with a user-defined function. You can
see examples of both strategies in Chapters 9 and 10.
Trying to create a UNIQUE constraint in a row or combination of rows with nonunique values will give you an
error message.
You can specify options for the UNIQUE index, in the same way as in the PRIMARY KEY definition. For more
index options, you can create a UNIQUE index instead of a UNIQUE constraint.
You can use the code of Listing 7. 4 to create a UNIQUE constraint in a single column of a table, using the
CREATE TABLE and the ALTER TABLE statements, using different syntax options.

Listing 7.4 Create UNIQUE Constraints

-- Define a UNIQUE constraint in a single column

-- using the default constraint name

CREATE TABLE NewRegions (

RegionID int NOT NULL
UNIQUE NONCLUSTERED,
RegionDescription nchar (50) NOT NULL ,
)
GO

DROP TABLE NewRegions

GO

238
 Chapter 7. Enforcing Data Integrity

-- Define a UNIQUE constraint in a single column

-- specifying the constraint name

CREATE TABLE NewRegions (

RegionID int NOT NULL
CONSTRAINT UC_NewRegions
UNIQUE NONCLUSTERED,
RegionDescription nchar (50) NOT NULL ,
)
GO
DROP TABLE NewRegions
GO

-- Define a UNIQUE constraint in a single column

-- specifying the constraint name
-- and defining the constraint at table level

CREATE TABLE NewRegions (

RegionID int NOT NULL,
RegionDescription nchar (50) NOT NULL ,
CONSTRAINT UC_NewRegions
UNIQUE NONCLUSTERED (RegionID),
)
GO

DROP TABLE NewRegions

GO

-- Define a UNIQUE constraint in a single column

-- specifying the constraint name
-- and defining the constraint at table level
-- using the ALTER TABLE statement

CREATE TABLE NewRegions (

RegionID int NOT NULL,
RegionDescription nchar (50) NOT NULL)

ALTER TABLE NewRegions

ADD CONSTRAINT UC_NewRegions
UNIQUE NONCLUSTERED (RegionID)
GO

DROP TABLE NewRegions

GO
As detailed earlier in this chapter for PRIMARY KEYs, you can provide properties for the index, such as
CLUSTERED, NONCLUSTERED, or FILLFACTOR, in the constraint definition. Listing 7.5 shows how to
declare the index of a UNIQUE constraint as nonclustered and with a FILLFACTOR of 70%. Index properties
were detailed in Chapter 6.

Listing 7.5 Create a UNIQUE Constraint and Define Properties of Its UNIQUE INDEX

-- Define a UNIQUE constraint in a single column

239
Microsoft SQL Server 2000 Programming by Example

-- and create the index a nonclustered

-- and 70% FillFactor
CREATE TABLE NewRegions (
RegionID int NOT NULL
UNIQUE NONCLUSTERED WITH FILLFACTOR = 70,
RegionDescription nchar (50) NOT NULL ,
)
GO

DROP TABLE NewRegions

GO
Listing 7.6 shows how to create a multicolumn UNIQUE constraint. A composite UNIQUE constraint is limited
to 16 columns and 900 bytes for the key size.

Listing 7.6 Creating a Multicolumn UNIQUE Constraint

-- Define a composite UNIQUE in two columns

-- specifying the constraint name

CREATE TABLE ProductSales (

ProductID int NOT NULL,
SaleID int NOT NULL,
Quantity int NOT NULL,
Price money NOT NULL,
Description varchar(200) NULL,
CONSTRAINT UC_ProductSales
UNIQUE NONCLUSTERED (ProductID, SaleID)
)
GO

DROP TABLE ProductSales

GO

-- Define a composite UNIQUE constraint in two columns

-- using the ALTER TABLE statement

CREATE TABLE ProductSales (

ProductID int NOT NULL,
SaleID int NOT NULL,
Quantity int NOT NULL,
Price money NOT NULL,
Description varchar(200) NULL)
ALTER TABLE ProductSales
ADD CONSTRAINT UC_ProductSales
UNIQUE NONCLUSTERED (ProductID, SaleID)
GO

DROP TABLE ProductSales

GO
You can provide default values for a UNIQUE field in the same way as for the PRIMARY KEY:

240
 Chapter 7. Enforcing Data Integrity

• Using the IDENTITY property for the column, specifying the seed and increment values.
• Using the uniqueidentifier data type combined with the NEWID function as a DEFAULT
constraint to supply automatic GUID (Global Unique Identifier) values.
• Declare a user-defined function as a DEFAULT constraint, to provide unique values to the column.
• Declare the column using a data type timestamp or rowversion. However, this option is useless
because timestamp values are unique, regardless of the existence of a UNIQUE constraint, and they
change whenever the row changes.

Caution

An IDENTITY property does not guarantee uniqueness on its column. SQL Server does not
guarantee uniqueness on IDENTITY columns unless you define a PRIMARY KEY constraint, a
UNIQUE constraint, or a UNIQUE index on that column.

You can use Enterprise Manager to define a UNIQUE constraint in a table. To do it, right-click the table and
select Design Table to display the Design Table form, and then click the Table and Index Properties icon on
the toolbar. In the Indexes/Keys tab, you can create, modify, or delete UNIQUE constraints.

In the Properties form, you have the choice to create a UNIQUE constraint or a UNIQUE index. Use UNIQUE
index if you want to provide extra functionality as Ignore Duplicate Key or Do Not Automatically Recompute
Statistics.

Figure 7.3 shows the Properties form in which you can see how to define Properties for a UNIQUE constraint.
Figure 7.4 shows a similar form to specify properties for the UNIQUE index associated to the UNIQUE
constraint.

Figure 7.3. Using the Properties form, you can define properties for a UNIQUE constraint.

241
Microsoft SQL Server 2000 Programming by Example

Figure 7.4. Using the Properties form, you can define properties for a UNIQUE index.

242
 Chapter 7. Enforcing Data Integrity

CHECK Constraints

A CHECK constraint defines a condition for one or more columns in a table on INSERT and UPDATE operations.
This condition can be defined by any expression that returns TRUE or FALSE. If the condition returns TRUE,
the operation continues, but if the condition returns FALSE, the operation is automatically rolled back.
It is possible to have many CHECK constraints per table, in which case they will be checked in creation order.
If one CHECK constraint fails, the operation is rolled back and no more CHECK constraints are tested. In these
situations, the client receives a single error message with information about the CHECK constraint that failed.
To define a CHECK constraint, you can use any expression as long as it references only columns in the same
table. A CHECK constraint cannot reference other rows in the same table or other tables, but it is valid to use a
user-defined function as part of the CHECK definition, and the user-defined function can use data from other
tables, databases, or servers. In Chapter 10, we'll discuss how to use user-defined functions in CHECK
constraints.
You can use a CHECK constraint to verify

• That the value has a valid format. For example, the Postcode should be a five-digit number.
• That the value is in a specific range of valid data. For example, the UnitPrice should be between 1
and 1,000, or greater than 0.
• That the value is not equal to any specific reserved value. For example, the name of a new product
shouldn't be an empty string ('').
• The result of a function applied to this value. For example, PaymentDue should be not later than 90
days from the SaleDate.

243
Microsoft SQL Server 2000 Programming by Example

You can define single column CHECK constraints during the creation of a table. Listing 7.7 shows how to
create a CHECK constraint on the NewEmployees table to ensure that the PostCode column accepts only
strings with five digits. The second example in Listing 7.7 creates a CHECK constraint that enforces positive
values for the UnitPrice column in the Products table.

Listing 7.7 Create a CHECK Constraint with the CREATE TABLE Statement

-- Define a CHECK constraint in a single column

-- using the default constraint name

CREATE TABLE NewEmployees (

EmployeeID int NOT NULL,
EmployeeName varchar(50) NOT NULL,
PostCode char(5) NOT NULL
CHECK (PostCode LIKE '[0-9][0-9][0-9][0-9][0-9]')
)
GO

DROP TABLE NewEmployees

GO
-- Define a CHECK constraint in a single column
-- specifying the constraint name

CREATE TABLE NewProducts (

ProductID int NOT NULL,
ProductName varchar(50) NOT NULL,
UnitPrice money NOT NULL
CONSTRAINT CC_Prod_UnitPrice
CHECK (UnitPrice > 0)
)
GO

DROP TABLE NewProducts

GO
You can create a CHECK constraint at table level in the CREATE TABLE statement. This is the only way to
define it if the CHECK constraint references more than one column. The expression can contain any number of
subexpressions using logical operators as long as the total expression evaluates to TRUE or FALSE.
Listing 7.8 contains three examples of CHECK constraints:

• The CC_Prod_Name CHECK constraint forces the ProductName column to accept only nonempty
strings.
• The CC_Order_DueDate CHECK constraint checks that the time between DueDate and SaleDate
is less than or equal to 90 days.
• The third example creates the CC_Order_DueDate CHECK constraint to check three conditions
simultaneously:
o The DueDate should be on or prior to 90 days after the SaleDate.
o The SaleDate cannot be a future date.
o The ShipmentMethod can have only three possible values: air ('A'), land ('L'), or sea
('S').

244
 Chapter 7. Enforcing Data Integrity

Listing 7.8 Create Multicolumn CHECK Constraints with the CREATE TABLE Statement

-- Define a CHECK constraint in a single column

-- specifying the constraint name
-- and defining the constraint at table level

CREATE TABLE NewProducts (

ProductID int NOT NULL,
ProductName varchar(50) NOT NULL,
UnitPrice money NOT NULL ,
CONSTRAINT CC_Prod_Name
CHECK (ProductName <> '')
)
GO

DROP TABLE NewProducts

GO

-- Define a CHECK constraint in a single column

-- specifying the constraint name
-- and defining the constraint at table level
-- as an expression

CREATE TABLE NewOrders (

OrderID int NOT NULL,
CustomerID int NOT NULL,
SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
CONSTRAINT CC_Order_DueDate
CHECK (DATEDIFF(day, SaleDate, DueDate) <= 90)
)
GO

DROP TABLE NewOrders

GO

-- Define a CHECK constraint in a single column

-- specifying the constraint name
-- and defining the constraint at table level
-- as a multiple expressions linked by
-- logical operators

CREATE TABLE NewOrders (

OrderID int NOT NULL,
CustomerID int NOT NULL,
SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
ShipmentMethod char(1) NOT NULL,
CONSTRAINT CC_Order_DueDate
CHECK
((DATEDIFF(day, SaleDate, DueDate) <= 90)
AND
(DATEDIFF(day, CURRENT_TIMESTAMP, SaleDate) <= 0)
AND (ShipmentMethod IN ('A', 'L', 'S'))
)
)
GO

DROP TABLE NewOrders

GO

245
Microsoft SQL Server 2000 Programming by Example

It is possible to create CHECK constraints for existing tables using the ALTER TABLE statement, as in Listing
7.9. In this case, you can specify whether it is necessary to check existing data.
The first example in Listing 7.9 creates three CHECK constraints on the NewOrders table: one CHECK
constraint for every condition used in the last example from Listing 7.8.
The second example in Listing 7.9 creates the same CHECK constraints as in the first example, but in this
case it specifies not to check existing data for the first and third CHECK constraints.
Tip

If you create a CHECK constraint as a sequence of multiple conditions, linked with the AND operator
only, break it into several single-condition CHECK constraints. The maintenance of these CHECK
constraints will be easier, and you will have more flexibility for enabling and disabling individual
conditions, if required.

Listing 7.9 Create CHECK Constraints with the ALTER TABLE Statement

-- Define multiple CHECK constraint

-- in existing tables specifying
-- the constraint name and defining
-- the constraint at table level
-- using the ALTER TABLE statement
-- checking existing data

CREATE TABLE NewOrders (

OrderID int NOT NULL,
CustomerID int NOT NULL,
SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
ShipmentMethod char(1) NOT NULL)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_DueDate
CHECK (DATEDIFF(day, SaleDate, DueDate) <= 90)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_SaleDate
CHECK (DATEDIFF(day, CURRENT_TIMESTAMP, SaleDate) <= 0)
ALTER TABLE NewOrders
ADD CONSTRAINT CC_Order_Shipment
CHECK (ShipmentMethod IN ('A', 'L', 'S'))

GO

DROP TABLE NewOrders

GO

-- Define multiple CHECK constraint

-- in existing tables specifying
-- the constraint name and defining
-- the constraint at table level
-- using the ALTER TABLE statement
-- checking existing data
-- only for one of the constraints

CREATE TABLE NewOrders (

246
 Chapter 7. Enforcing Data Integrity

OrderID int NOT NULL,

CustomerID int NOT NULL,
SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
ShipmentMethod char(1) NOT NULL)

ALTER TABLE NewOrders

WITH NOCHECK
ADD CONSTRAINT CC_Order_DueDate
CHECK (DATEDIFF(day, SaleDate, DueDate) <= 90)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_SaleDate
CHECK (DATEDIFF(day, CURRENT_TIMESTAMP,

ALTER TABLE NewOrders

WITH NOCHECK
ADD CONSTRAINT CC_Order_Shipment
CHECK (ShipmentMethod IN ('A', 'L', 'S'))

GO

DROP TABLE NewOrders

GO
To modify a CHECK constraint, you must drop the constraint and re-create it or use Enterprise Manager to do it.
To modify a CHECK constraint using Enterprise Manager, right-click the table and select Design Table to
display the Design Table form. Click the Table and Index Properties icon on the toolbar to display the
Properties form, and select the Check Constraints tab. Figure 7.5 shows the Check Constraints tab of the
Properties form. Using this form, you can

Figure 7.5. Use Enterprise Manager to modify a CHECK constraint.

247
Microsoft SQL Server 2000 Programming by Example

• Change the name of the CHECK constraint.

• Change the expression of the constraint.
• Specify whether you want to check existing data.
• Select whether you want to enforce this constraint when receiving data from replication (on by default).
• Enable the constraint for INSERT and UPDATE statements (on by default).

Caution

When you modify a CHECK constraint using Enterprise Manager, Check Existing Data is off by
default. However, when you create a new constraint using Transact-SQL, this option is on by
default.

Note

If you use Enterprise Manager to modify a CHECK constraint, Enterprise Manager will drop and re-
create the constraint for you, using the new settings.

248
 Chapter 7. Enforcing Data Integrity

Caution

Check constraints are evaluated when you insert data and when you try to update a column
referenced by a check constraint. However, if you update a column in a preexisting row, only this
column will be checked against CHECK constraints; the other columns will remain unchecked.

Checking constraints produces some overhead. You can disable a CHECK constraint for INSERT and UPDATE
operations if the data to insert is checked already with the same conditions. To disable a constraint for
INSERT and UPDATE operations, you can use the ALTER TABLE statement.

Listing 7.10 shows how to disable and reenable a CHECK constraint.

Listing 7.10 Use the ALTER TABLE Statement to Disable and Reenable Constraints

-- Create a CHECK constraint

-- in existing tables specifying
-- the constraint name and defining
-- the constraint at table level
-- using the ALTER TABLE statement
-- checking existing data

-- Create table

CREATE TABLE NewOrders (

OrderID int NOT NULL,
CustomerID int NOT NULL,
SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
ShipmentMethod char(1) NOT NULL)

-- Create constraint

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_DueDate
CHECK (DATEDIFF(day, SaleDate, DueDate) <= 90)

-- Disable constraint

ALTER TABLE NewOrders

NOCHECK CONSTRAINT CC_Order_DueDate

-- Reenable constraint
ALTER TABLE NewOrders

249
Microsoft SQL Server 2000 Programming by Example

CHECK CONSTRAINT CC_Order_DueDate

GO

DROP TABLE NewOrders

GO
Tip

Disable constraints before importing the data to speed up the importing process, and reenable
them after the data has been imported.

Replicated data has been checked already in the Publisher. Checking the same data again in the subscriber
is usually unnecessary. You can create a CHECK constraint that is disabled for replicated data specifying NOT
FOR REPLICATION after the CHECK keyword.

Listing 7.11 shows how to use the NOT FOR REPLICATION option in the CREATE TABLE and ALTER
TABLE statements.

Listing 7.11 Using the NOT FOR REPLICATION Option to Avoid Checking Replicated Data

-- Define a CHECK constraint in a single column

-- using the default constraint name
-- specifying NOT FOR REPLICATION

CREATE TABLE NewEmployees (

EmployeeID int NOT NULL,
EmployeeName varchar(50) NOT NULL,
PostCode char(5) NOT NULL
CHECK NOT FOR REPLICATION (PostCode LIKE '[0-9][0-9][0-9][0-9][0-9]')
)
GO

DROP TABLE NewEmployees

GO

-- Define multiple CHECK constraint

-- in existing tables specifying
-- the constraint name and defining
-- the constraint at table level
-- using the ALTER TABLE statement
-- checking existing data
-- specifying NOT FOR REPLICATION

CREATE TABLE NewOrders (

OrderID int NOT NULL,

250
 Chapter 7. Enforcing Data Integrity

CustomerID int NOT NULL,

SaleDate smalldatetime NOT NULL ,
DueDate smalldatetime NOT NULL,
ShipmentMethod char(1) NOT NULL)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_DueDate
CHECK NOT FOR REPLICATION
(DATEDIFF(day, SaleDate, DueDate) <= 90)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_SaleDate
CHECK NOT FOR REPLICATION
(DATEDIFF(day, CURRENT_TIMESTAMP, SaleDate) <= 0)

ALTER TABLE NewOrders

ADD CONSTRAINT CC_Order_Shipment
CHECK NOT FOR REPLICATION
(ShipmentMethod IN ('A', 'L', 'S'))

GO

DROP TABLE NewOrders

GO
Caution

If you disable a CHECK constraint to import data, test the data to see whether the check condition
can still be valid before reenabling the constraint.

You can remove a CHECK constraint by using the ALTER TABLE statement. Removing a table removes
constraints associated with the table.

Use the example in Listing 7.12 to create the NewProducts table, including the CC_Prod_UnitPrice
CHECK constraint, and drop the constraint using the ALTER TABLE statement.

Listing 7.12 Use the ALTER TABLE Statement to Remove CHECK Constraints

-- Define a CHECK constraint in a single column

-- specifying the constraint name

CREATE TABLE NewProducts (

ProductID int NOT NULL,
ProductName varchar(50) NOT NULL,
UnitPrice money NOT NULL
CONSTRAINT CC_Prod_UnitPrice
CHECK (UnitPrice > 0)
)
GO

-- Drop the constraint

-- Specifying its name

ALTER TABLE NewProducts

251
Microsoft SQL Server 2000 Programming by Example

DROP CONSTRAINT CC_Prod_UnitPrice

GO

DROP TABLE NewProducts

GO
Tip

Providing names to constraints makes it easier to drop, disable, and reenable them. Otherwise, you
must use the sp_helpconstraint system stored procedure to retrieve information about existing
constraints in a table.

DEFAULT Definitions

You can define a DEFAULT definition for columns to avoid repetitive data entry. If a column has a DEFAULT
definition, this value will be supplied if you don't provide a specific value for the column in the INSERT
statement. DEFAULT definitions are applied only in INSERT operations.
You can provide as a DEFAULT definition any expression that evaluates to a single scalar value with a data
type compatible with the data type of the column in which the DEFAULT definition is defined. This expression
can be

• A constant value
• Any system scalar function
• The result of a scalar user-defined function
• Any scalar expression made from any combination of the previous points, including mathematical
expressions

You can create a DEFAULT definition for a new column using the CREATE TABLE or ALTER TABLE
statements, defining the DEFAULT constraint at column level. To force the use of the default value, you can
omit the column, provide the DEFAULT keyword, or use the DEFAULT VALUES for the INSERT statement, as
detailed in Listing 7.13.
Caution

If you explicitly insert a NULL value into a column that accepts NULL, and the column has a
DEFAULT definition, the DEFAULT definition won't be applied.

Caution

You can have only one DEFAULT definition per column; otherwise, SQL Server will not know which
value to use. If you try to create more than one DEFAULT definition for a column, you will receive an
error message.

Listing 7.13 Create DEFAULT Definitions Using the CREATE TABLE Statement

252
 Chapter 7. Enforcing Data Integrity

-- Create the table NewCustomers

-- providing the value 'London'
-- as a column DEFAULT definition
-- for the City column,
-- a named DEFAULT constraint
-- for the CustomerName column,
-- The current date and time
-- for the CreaDate column
-- and the login name
-- for the CreaUser column

CREATE TABLE NewCustomers(

CustomerID int NOT NULL
IDENTITY(1,1)
PRIMARY KEY,
CustomerName varchar(30) NOT NULL
CONSTRAINT Def_CustName
DEFAULT 'To be entered',
City varchar(30)
DEFAULT 'London',
CreaDate smalldatetime
DEFAULT Getdate(),
CreaUser nvarchar(128)
DEFAULT System_User)
GO

-- Insert data into the NewCustomers table

-- Providing values for CustomerName
-- and City fields

INSERT NewCustomers (CustomerName, City)

VALUES ('MyComp corp.', 'New York')

-- Insert data into the NewCustomers table

-- Omitting to enter the City field

INSERT NewCustomers (CustomerName)

VALUES ('ACME Inc.')
SELECT *
FROM NewCustomers

-- Insert data into the NewCustomers table

-- Providing the default value
-- for the City field

INSERT NewCustomers (CustomerName, City)

VALUES ('NewDotCompany Ltd.', DEFAULT)

SELECT *
FROM NewCustomers

-- Insert data into the NewCustomers table

253
Microsoft SQL Server 2000 Programming by Example

-- Providing the default value

-- for every nonnull field

INSERT NewCustomers
DEFAULT VALUES

SELECT *
FROM NewCustomers

-- Drop the test table

DROP TABLE NewCustomers

CustomerID CustomerName City CreaDate CreaUser

----------- -------------------- --------- ---------------------- --------
1 MyComp corp. New York 2000-11-11 17:35:00 sa
2 ACME Inc. Lon+++don 2000-11-11 17:35:00 sa

CustomerID CustomerName City CreaDate CreaUser

----------- -------------------- --------- ---------------------- --------
1 MyComp corp. New York 2000-11-11 17:35:00 sa
2 ACME Inc. London 2000-11-11 17:35:00 sa
3 NewDotCompany Ltd. London 2000-11-11 17:35:00 sa

CustomerID CustomerName City CreaDate CreaUser

----------- -------------------- --------- ---------------------- --------
1 MyComp corp. New York 2000-11-11 17:35:00 sa
2 ACME Inc. London 2000-11-11 17:35:00 sa
3 NewDotCompany Ltd. London 2000-11-11 17:35:00 sa
4 To be entered London 2000-11-11 17:35:00 sa
Note

You can refer to the DEFAULT definitions as DEFAULT constraints or DEFAULT properties as well.
SQL Server 2000 Books Online refers to them as DEFAULT definitions, but earlier versions called
them DEFAULT constraints, and using SQL-DMO refer to them as DEFAULT properties.

You can add a DEFAULT definition to an existing column using the ALTER TABLE statement, as in Listing
7.14.

Listing 7.14 Create new DEFAULT Properties for New and Existing Columns with the ALTER TABLE
Statement

254
 Chapter 7. Enforcing Data Integrity

-- Create the table NewCustomers

-- providing the value 'London'
-- as a column DEFAULT definition
-- for the City column,
-- and a named DEFAULT constraint
-- for the CustomerName column

CREATE TABLE NewCustomers(

CustomerID int NOT NULL
IDENTITY(1,1)
PRIMARY KEY,
CustomerName varchar(30) NOT NULL
CONSTRAINT Def_CustName
DEFAULT 'To be entered',
City varchar(30)
DEFAULT 'London',
CreaDate smalldatetime)
GO

-- Use ALTER TABLE

-- to add a new column
-- with a DEFAULT definition

ALTER TABLE NewCustomers

ADD CreaUser nvarchar(128)
DEFAULT SYSTEM_USER

-- Use ALTER TABLE

-- to add a DEFAULT definition
-- to an existing column

ALTER TABLE NewCustomers

ADD CONSTRAINT Def_Cust_CreaDate
DEFAULT CURRENT_TIMESTAMP
FOR CreaDate
GO

-- Drop the test table

DROP TABLE NewCustomers

Note

You can use the system function CURRENT_TIMESTAMP as a synonym of GetDate(), and the
system function SYSTEM_USER as a synonym of SUSER_SNAME().

When you add a new column with a DEFAULT definition, the column gets the DEFAULT value automatically for
existing rows if

255
Microsoft SQL Server 2000 Programming by Example

• The column does not accept NULL

• The column accepts NULL but you specify WITH VALUES after the DEFAULT definition

Listing 7.15 shows how to add a column with a DEFAULT definition to a table with existing data, specifying
the WITH VALUES option.

Listing 7.15 Use the WITH VALUES Option to Provide a Default Value to a New Column on Existing
Rows

-- Create the table NewCustomers

-- providing the value 'London'
-- as a column DEFAULT definition
-- for the City column,
-- and a named DEFAULT constraint
-- for the CustomerName column

CREATE TABLE NewCustomers(

CustomerID int NOT NULL
IDENTITY(1,1)
PRIMARY KEY,
CustomerName varchar(30) NOT NULL
CONSTRAINT Def_CustName
DEFAULT 'To be entered',
City varchar(30)
DEFAULT 'London',
CreaDate smalldatetime
DEFAULT CURRENT_TIMESTAMP)
GO

-- Insert some data

INSERT NewCustomers (CustomerName, City)

VALUES ('MyComp corp.', 'New York')
INSERT NewCustomers (CustomerName)
VALUES ('ACME Inc.')

INSERT NewCustomers (CustomerName, City)

VALUES ('NewDotCompany Ltd.', DEFAULT)

INSERT NewCustomers
DEFAULT VALUES

SELECT *
FROM NewCustomers
GO

-- Use ALTER TABLE

-- to add a new column
-- with a DEFAULT definition
-- filling this new column
-- in existing rows

256
 Chapter 7. Enforcing Data Integrity

ALTER TABLE NewCustomers

ADD CreditLimit money
DEFAULT 1000.0
WITH VALUES

SELECT *
FROM NewCustomers

GO

-- Drop the test table

DROP TABLE NewCustomers

CustomerID CustomerName City CreaDate

----------- -------------------- ---------- -------------------
1 MyComp corp. New York 2000-11-11 17:35:00
2 ACME Inc. London 2000-11-11 17:35:00
3 NewDotCompany Ltd. London 2000-11-11 17:35:00
4 To be entered London 2000-11-11 17:35:00
CustomerID CustomerName City CreaDate CreditLimit
----------- -------------------- ---------- --------------------- -----------
1 MyComp corp. New York 2000-11-11 17:58:00 1000.0000
2 ACME Inc. London 2000-11-11 17:58:00 1000.0000
3 NewDotCompany Ltd. London 2000-11-11 17:58:00 1000.0000
4 To be entered London 2000-11-11 17:58:00 1000.0000

Foreign Keys

If you have two related tables— such as TableA and TableB—their relationship can be any of these three
types:

• One to one— Figure 7.6 shows the relationship between the Customers table and the
PublicCustomers table (this table does not exist in the Northwind database). Every row in the
PublicCustomers is related to a single row in the Customers table, but not every row in the
Customers table has a related row in the PublicCustomers table. As in this example, you can use
one-to-one relationships to expand a table creating a subtable to store information related only to
some specific rows.

Figure 7.6. A one-to-one relationship.

257
Microsoft SQL Server 2000 Programming by Example

• One to many— Figure 7.7 shows the relationship between the Products and [Order Details]
tables. For every product, you can have none, one, or many related rows in the [Order Details]
table. And every row in the [Order Details] table is related to only one row in the Products
table.

Figure 7.7. A one-to-many relationship.

• Many to many— Every employee can work in different territories, and in every territory you can have
many employees. Figure 7.8 shows the relationships between the Employees and the
Territories tables. As you can see in the diagram, the problem is solved by an intersection table,
EmployeesTerritories, that contains the keys to connect to both tables.

Figure 7.8. A many-to-many relationship.

258
 Chapter 7. Enforcing Data Integrity

To establish a relationship from one table, which can be called the "children" table, to another table, which can
be called the "parent" table, you must create in the children table a FOREIGN KEY constraint.
When you define a FOREIGN KEY in the children table, this constraint enforces some rules in both parent and
children tables:

• In the parent table, you cannot modify the related field on any row that has any related rows in the
children table. For example, you cannot change the product code, or sales in this product will be
invalid.
• You cannot delete a row in the parent table that has related rows in the children table. For example,
you cannot delete a product with sales, or the sales about this product will be orphans.
• You cannot insert new rows in the children table where the related field contains nonexisting values in
the related field on the parent table. For example, you cannot create a sale for a nonexisting product.
• You cannot modify the related field in the children table to a new value that does not exist in the
parent table. For example, you cannot change the product code in a sale for a nonexisting product
code.

After the FOREIGN KEY is created, SQL Server will check every statement involving the related tables to
prevent any violation of the relationship. In that case, the statement will be rejected and the data will not be
modified.
Caution

Because SQL Server checks constraints before modifying the data, if any constraint fails, the data
will never be modified. Therefore, this cancelled action will not fire any trigger you might have
defined.

259
Microsoft SQL Server 2000 Programming by Example

You can declare a FOREIGN KEY constraint either using the CREATE TABLE or the ALTER TABLE
statements. Listing 7.16 shows a complete example.

Listing 7.16 Create FOREIGN KEY Constraints with the CREATE TABLE or ALTER TABLE Statement

-- Create Customers and Orders tables

CREATE TABLE Customers(

CustomerID int
PRIMARY KEY,
CustomerName varchar(20) NOT NULL)

CREATE TABLE Orders(

OrderID int
IDENTITY(1,1)
PRIMARY KEY,
CustomerID int NOT NULL,
OrderDate smalldatetime NOT NULL
DEFAULT CURRENT_TIMESTAMP)
GO

-- Create the FOREIGN KEY constraint

ALTER TABLE Orders

ADD CONSTRAINT FK_Orders
FOREIGN KEY (CustomerID)
REFERENCES Customers (CustomerID)
GO

/*

-- Or you could define the Orders

-- table with References
-- without using the ALTER TABLE
-- statement to define the FOREIGN KEY
CREATE TABLE Orders(
OrderID int
IDENTITY(1,1)
PRIMARY KEY,
CustomerID int NOT NULL
REFERENCES Customers(CustomerID),
OrderDate smalldatetime NOT NULL
DEFAULT CURRENT_TIMESTAMP)
--GO

*/

-- Insert some Customers

INSERT Customers

260
 Chapter 7. Enforcing Data Integrity

VALUES (1, 'MyComp corp.')

INSERT Customers
VALUES (2, 'ACME Inc.')

INSERT Customers
VALUES (3, 'NewDotCompany Ltd.')

-- Insert some Orders

-- with the default Date

INSERT Orders (CustomerID)

VALUES (1)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (3)

INSERT Orders (CustomerID)

VALUES (3)

-- Try to update customer 3 PRIMARY KEY

-- to a different value

PRINT CHAR(10) + 'Trying to update a customer PK'+ CHAR(10)

UPDATE Customers
SET CustomerID = 30
WHERE CustomerID = 3

-- Try to insert a new order for a

-- nonexisting customer

PRINT CHAR(10) + 'Trying to insert an orphan Order'+ CHAR(10)

INSERT Orders (CustomerID)

VALUES (10)

GO

DROP TABLE Orders

DROP TABLE Customers

Trying to update a customer PK

Server: Msg 547, Level 16, State 1, Line 1

UPDATE statement conflicted with COLUMN REFERENCE constraint 'FK_Orders'. The
conflict
occurred in database 'ByExample', table 'Orders', column 'CustomerID'.
The statement has been terminated.

261
Microsoft SQL Server 2000 Programming by Example

Trying to insert an orphan Order

Server: Msg 547, Level 16, State 1, Line 1

INSERT statement conflicted with COLUMN FOREIGN KEY constraint 'FK_Orders'. The
conflict
occurred in database 'ByExample', table 'Customers', column 'CustomerID'.
The statement has been terminated.
Caution

Having a FOREIGN KEY defined in a column does not prevent this column from being NULL.
Therefore, it is possible to have orphan rows not related to any row in the parent table. To solve
this situation, you should declare the related columns as NOT NULL.

Note

It is required that the related field or group fields in the parent table must have defined a UNIQUE
index. Therefore, it is recommended to use the primary key as the related field.

You can use Enterprise Manager to create FOREIGN KEY constraints. The easiest way is by using the
Diagram tool. Figure 7.9 shows a diagram for the Northwind database, including the relationships between
tables.

Figure 7.9. To view and edit relationships between tables, you can use the Diagram tool in Enterprise
Manager.

262
 Chapter 7. Enforcing Data Integrity

To create a FOREIGN KEY in the Diagram tool, you can just drag a column from the children table and drop it
on the parent table.
Tip

In complex databases, create one diagram for every main table, showing only one or two levels of
relationship. In this way, it will be easier to manage. However, you can have a complete diagram as
well to get the full picture, if required.

Figure 7.10 shows how to create and modify FOREIGN KEY constraints using the Properties form of the
children table.

Figure 7.10. Using the Properties form, you can define relationships with related tables.

263
Microsoft SQL Server 2000 Programming by Example

You can prevent checking existing data when you create the FOREIGN KEY constraint by using the WITH
NOCHECK option.
To modify a FOREIGN KEY constraint, you must drop the constraint and re-create it.
Note

Remember that you can use ALTER TABLE... NOCHECK CONSTRAINT to disable constraint
checking and ALTER TABLE... CHECK CONSTRAINT to reenable the constraint.

To prevent the FOREIGN KEY constraint from checking replicated data, use the NOT FOR REPLICATION
predicate.

Caution

To create a FOREIGN KEY constraint that references a multicolumn UNIQUE index, PRIMARY KEY
or UNIQUE constraint, in the parent table, you must create the FOREIGN KEY in the group of
related fields.

264
 Chapter 7. Enforcing Data Integrity

FOREIGN KEY constraints are dropped automatically when the table is dropped. However, to drop a
constraint you can use the ALTER TABLE... DROP CONSTRAINT statement, specifying the name of the
constraint to drop.

Note

When you create a FOREIGN KEY constraint, SQL Server does not create any index on the
selected columns. However, the columns included in a FOREIGN KEY constraint are good
candidates for an index.

Cascading Operations: Cascaded Declarative Referential Integrity

The ANSI standard establishes four modes to solve changes that could break referential integrity definitions:

• RESTRICT or NO ACTION—Trying to delete or modify a row that is linked to some rows in a children
table produces an error message and the operation is rolled back. This is the default for SQL Server
2000.
• CASCADE—Modifications to a UNIQUE field are cascaded to every FOREIGN KEY that references the
field. Deleting a row forces a deletion of every related row where the FOREIGN KEY references the
deleted row. This mode is optional in SQL Server 2000.
• SET NULL—When the referential integrity is broken for some FOREIGN KEY values, because of an
update or delete operation in the parent table, those values are set to NULL. This mode is not
implemented in SQL Server 2000.
• SET DEFAULT—If the referential integrity is broken for some FOREIGN KEY values due to an update
or delete operation in the parent table, the FOREIGN KEY values are set to a default value. This mode
is not implemented in SQL Server 2000.

The standard mode of FOREIGN KEY constraints has been covered in a previous section. In the following
sections, you are going to see how to implement cascade operations in SQL Server 2000.
Nested cascade operations interact with nested triggers in the following way:

1. INSTEAD OF triggers, in the original table, execute first, usually with the execution of modifications on
individual tables, which you will consider the first level of update.
2. All cascade operations, from tables of the first level of update, execute without a predefined order.
The cascade operations continue through the related tables, eventually forcing further cascade
operations to be executed affecting other tables, until there are no more tables to apply cascade
operations.
3. AFTER triggers on the deepest level of the cascade operation fire first. On this level, the triggers of the
affected tables fire without predefined order. The execution of triggers moves one level up at a time,
firing every required trigger from affected tables. In this way, for a given table, SQL Server tries to fire
every AFTER trigger only once. The AFTER triggers on these levels fire only if the trigger is fired due to
the modification of one or more rows.
4. AFTER triggers on the first level of update fire regardless of the number of affected rows.

Caution

Because the interaction between constraints and triggers can be very complex, always document
your design and keep it simple.

265
Microsoft SQL Server 2000 Programming by Example

Note

You can use the Transact-SQL Debugger to debug triggers (INSTEAD OF and AFTER triggers).
You must create a stored procedure to do the first modification, and debug this procedure. The
debugger will jump from trigger to trigger if required.

Cascading Deletes

To define a FOREIGN KEY constraint as a cascaded DELETE action, you must use the ON DELETE CASCADE
in the REFERENCES clause of the FOREIGN KEY definition on the CREATE TABLE or ALTER TABLE
statements.
In Listing 7.17, you create the Customers table and the Orders table. You define a FOREIGN KEY
constraint between these two tables with the ON DELETE CASCADE option. Customer 2 has three related
rows in the Orders table. The operation that deleted customer 2 from the Customers table forces the
deletion of the three related rows in the Orders table.
Caution

Cascade operations can be nested from table to table producing potentially undesired results.

Listing 7.17 Cascade Deletion of Rows in Related Tables with FOREIGN KEY Constraints Defined As
ON DELETE CASCADE

-- Create Customers and Orders tables

CREATE TABLE Customers(

CustomerID int
PRIMARY KEY,
CustomerName varchar(20) NOT NULL)

CREATE TABLE Orders(

OrderID int
IDENTITY(1,1)
PRIMARY KEY,
CustomerID int NOT NULL,
OrderDate smalldatetime NOT NULL
DEFAULT CURRENT_TIMESTAMP)
GO

266
 Chapter 7. Enforcing Data Integrity

-- Create the FOREIGN KEY constraint

-- with CASCADE

ALTER TABLE Orders

ADD CONSTRAINT FK_Orders
FOREIGN KEY (CustomerID)
REFERENCES Customers (CustomerID)
ON DELETE CASCADE

GO

-- Insert some Customers

INSERT Customers
VALUES (1, 'MyComp corp.')

INSERT Customers
VALUES (2, 'ACME Inc.')

INSERT Customers
VALUES (3, 'NewDotCompany Ltd.')

-- Insert some Orders

-- with the default Date

INSERT Orders (CustomerID)

VALUES (1)

INSERT Orders (CustomerID)

VALUES (1)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (3)

INSERT Orders (CustomerID)

VALUES (3)

-- Show the data

PRINT CHAR(10) + 'Original Customers table'+ CHAR(10)

SELECT *
FROM Customers

PRINT CHAR(10) + 'Original Orders table'+ CHAR(10)

SELECT *
FROM Orders

GO
-- Delete Customer 2

267
Microsoft SQL Server 2000 Programming by Example

DELETE Customers
WHERE CustomerID = 2

PRINT CHAR(10) + 'Customers table after delete Customer 2'+ CHAR(10)

SELECT *
FROM Customers

PRINT CHAR(10) + 'Orders table after delete Customer 2'+ CHAR(10)

SELECT *
FROM Orders

GO

DROP TABLE Orders

DROP TABLE Customers

Original Customers table

CustomerID CustomerName
----------- --------------------
1 MyComp corp.
2 ACME Inc.
3 NewDotCompany Ltd.

Original Orders table

OrderID CustomerID OrderDate

----------- ----------- ------------------------------
1 1 2000-11-12 23:55:00
2 1 2000-11-12 23:55:00
3 2 2000-11-12 23:55:00
4 2 2000-11-12 23:55:00
5 2 2000-11-12 23:55:00
6 3 2000-11-12 23:55:00
7 3 2000-11-12 23:55:00

Customers table after delete Customer 2

CustomerID CustomerName
----------- --------------------
1 MyComp corp.
3 NewDotCompany Ltd.

Orders table after delete Customer 2

OrderID CustomerID OrderDate

----------- ----------- ------------------------------

268
 Chapter 7. Enforcing Data Integrity

1 1 2000-11-12 23:55:00
2 1 2000-11-12 23:55:00
6 3 2000-11-12 23:55:00
7 3 2000-11-12 23:55:00

Cascading Update s

To define a FOREIGN KEY constraint as a cascaded UPDATE action, you must use the ON UPDATE CASCADE
in the REFERENCES clause of the FOREIGN KEY definition on the CREATE TABLE or ALTER TABLE
statements.
Listing 7.18 is based in the same example as in Listing 7.17. You create the Customers table and the
Orders table. You define a FOREIGN KEY constraint between these two tables with the ON UPDATE
CASCADE option. You want to change the ID of customer 3 to 30. Customer 3 has two related rows in the
Orders table. The UPDATE operation changes the CustomerID from 3 to 30 in both tables automatically.
Caution

It is not recommended to change PRIMARY KEY values. This can produce identity integrity
problems in your applications.

Listing 7.18 Cascade Changes on Primary Keys to Related Foreign Keys with FOREIGN KEY
Constraints Defined As ON UPDATE CASCADE

-- Create Customers and Orders tables

CREATE TABLE Customers(

CustomerID int
PRIMARY KEY,
CustomerName varchar(20) NOT NULL)

CREATE TABLE Orders(

OrderID int
IDENTITY(1,1)
PRIMARY KEY,
CustomerID int NOT NULL,
OrderDate smalldatetime NOT NULL
DEFAULT CURRENT_TIMESTAMP)
GO
-- Create the FOREIGN KEY constraint
-- with CASCADE

ALTER TABLE Orders

ADD CONSTRAINT FK_Orders
FOREIGN KEY (CustomerID)
REFERENCES Customers (CustomerID)
ON DELETE CASCADE -- This is optional
ON UPDATE CASCADE

269
Microsoft SQL Server 2000 Programming by Example

GO

-- Insert some Customers

INSERT Customers
VALUES (1, 'MyComp corp.')

INSERT Customers
VALUES (2, 'ACME Inc.')

INSERT Customers
VALUES (3, 'NewDotCompany Ltd.')

-- Insert some Orders

-- with the default Date

INSERT Orders (CustomerID)

VALUES (1)

INSERT Orders (CustomerID)

VALUES (1)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (2)

INSERT Orders (CustomerID)

VALUES (3)

INSERT Orders (CustomerID)

VALUES (3)
-- Show the data

PRINT CHAR(10) + 'Original Customers table'+ CHAR(10)

SELECT *
FROM Customers

PRINT CHAR(10) + 'Original Orders table'+ CHAR(10)

SELECT *
FROM Orders

GO

-- Update Customer 3
-- Change CustomerID from 3 for 30

UPDATE Customers
SET CustomerID = 30
WHERE CustomerID = 3

PRINT CHAR(10) + 'Customers table after update Customer 3'+ CHAR(10)

SELECT *
FROM Customers

270
 Chapter 7. Enforcing Data Integrity

PRINT CHAR(10) + 'Orders table after update Customer 3'+ CHAR(10)

SELECT *
FROM Orders

GO

DROP TABLE Orders

DROP TABLE Customers

Original Customers table

CustomerID CustomerName
----------- --------------------
1 MyComp corp.
2 ACME Inc.
3 NewDotCompany Ltd.

Original Orders table

OrderID CustomerID OrderDate
----------- ----------- ------------------------------
1 1 2000-11-12 23:59:00
2 1 2000-11-12 23:59:00
3 2 2000-11-12 23:59:00
4 2 2000-11-12 23:59:00
5 2 2000-11-12 23:59:00
6 3 2000-11-12 23:59:00
7 3 2000-11-12 23:59:00

Customers table after update Customer 3

CustomerID CustomerName
----------- --------------------
1 MyComp corp.
2 ACME Inc.
30 NewDotCompany Ltd.

Orders table after update Customer 3

OrderID CustomerID OrderDate

----------- ----------- ------------------------------
1 1 2000-11-12 23:59:00
2 1 2000-11-12 23:59:00
3 2 2000-11-12 23:59:00
4 2 2000-11-12 23:59:00
5 2 2000-11-12 23:59:00
6 30 2000-11-12 23:59:00
7 30 2000-11-12 23:59:00

271
Microsoft SQL Server 2000 Programming by Example

Transact-SQL–Specific Integrity Structures

Transact-SQL language provides an alternative to the CHECK and DEFAULT constraints with the RULE and
DEFAULT objects. RULE and DEFAULT objects are not ANSI standard, so it is advisable to use constraints as
a general way to provide the same functionality.
One of the reasons to use these Transact-SQL objects is to create self- contained user-defined data types,
including not only the data type, but also the DEFAULT value and the RULE to check for domain integrity. If a
column uses one of these self-contained user-defined data types as a data type, this column will inherit the
DEFAULT definition and the RULE definition as well.
User-defined data types were covered in Chapter 2.
Using DEFAULT and RULE objects can help during the development process, if the same condition must be
applied to multiple columns. However, remember that they are not ANSI compliant.

DEFAULT Objects

DEFAULT objects are similar to the DEFAULT definition of a column, but you can create a DEFAULT object
independently of any column and bind it to specific columns or user-defined data types later.
To create a DEFAULT object, you use the CREATE DEFAULT statement, providing a unique name for the
DEFAULT object and defining the object as a constant, built-in function or any valid scalar expression.
To delete a DEFAULT object, you must use the DROP DEFAULT statement. You cannot drop a DEFAULT
object if it is used anywhere in your database.
Caution

The only way to modify a DEFAULT or RULE object definition is by dropping and re- creating the
object. Before dropping the object, you must unbind the object from any field and user-defined data
type.

To bind a DEFAULT object to a field or user-defined data type, you must use the sp_bindefault system
stored procedure, and the sp_unbindefault disconnects a bound DEFAULT object from a field or user-
defined data type. Only one DEFAULT definition or DEFAULT object can be defined per column; binding a new
DEFAULT object to a column overrides the existing one.

Note

DEFAULT and RULE objects are local to a database. Therefore, DEFAULT and RULE objects
created in the Master database can be used only in the Master database.

You can see a complete example of how to use DEFAULT objects in Listing 7.19.

Listing 7.19 Create Independent DEFAULT Objects and Bind Them Later to Any Field or User-Defined
Data Type

272
 Chapter 7. Enforcing Data Integrity

-- Create a DEFAULT object using a constant

CREATE DEFAULT NoSales

AS 0
GO

-- Create DEFAULT objects using expressions

-- based on built-in functions

CREATE DEFAULT ThisMonth

AS Month(CURRENT_TIMESTAMP)
GO
CREATE DEFAULT UserDB
AS SYSTEM_USER
+ '- '+ DB_NAME(DB_ID())
GO

-- Create two User-Defined Data Types

EXEC sp_addtype 'UDDTLoginDB', 'nvarchar(256)', 'NULL'

EXEC sp_addtype 'UDDTSales', 'money', 'NULL'

GO

-- Create a table to test the DEFAULT objects

-- and the User-Defined Data Types

CREATE TABLE TestDefaults(

ID int NOT NULL
IDENTITY(1,1)
PRIMARY KEY,
TotalSales money NULL,
SalesMonth tinyint NULL,
WhoWhere UDDTLoginDB)
GO

-- Insert a new empty row in the table

INSERT TestDefaults
DEFAULT VALUES

PRINT char(10) + 'No defaults defined'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Bind the NoSales DEFAULT object

-- to the TotalSales field in the TestDefaults table

EXEC sp_bindefault 'NoSales', 'TestDefaults.TotalSales'

GO

273
Microsoft SQL Server 2000 Programming by Example

-- Insert a new empty row in the table

INSERT TestDefaults
DEFAULT VALUES
PRINT CHAR(10) + 'Only DEFAULT on TotalSales defined'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Bind the ThisMonth DEFAULT object

-- to the SalesMonth field in the TestDefaults table

EXEC sp_bindefault 'ThisMonth', 'TestDefaults.SalesMonth'

GO

-- Insert a new empty row in the table

INSERT TestDefaults
DEFAULT VALUES

PRINT CHAR(10) + 'DEFAULT defined on TotalSales and SalesMonth'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Bind the UserDB DEFAULT object

-- to the UDDTLginDB User-Defined Data Type

EXEC sp_bindefault 'UserDB', 'UDDTLoginDB'

GO

-- Insert a new empty row in the table

INSERT TestDefaults
DEFAULT VALUES

PRINT CHAR(10) + 'DEFAULT defined on TotalSales, SalesMonth'

PRINT 'and the UDDTLoginDB User-Defined Data Type'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Add a new column to the TestDefaults table

-- Using the UDDTSales data type
ALTER TABLE TestDefaults
ADD ProjectedSales UDDTSales
GO

PRINT CHAR(10) + 'Add an empty field using the UDDTSales data type'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Bind the NoSales DEFAULT object

-- to the UDDTSales User-Defined Data Type

274
 Chapter 7. Enforcing Data Integrity

-- for future columns only

EXEC sp_bindefault 'NoSales', 'UDDTSales', 'futureonly'

GO

-- Insert a new empty row in the table

INSERT TestDefaults
DEFAULT VALUES

PRINT CHAR(10) + 'DEFAULT defined on UDDTSales data type as futureonly'

PRINT 'does not affect the existing fields using this UDDT'+ CHAR(10)

SELECT *
FROM TestDefaults
GO

-- Drop everything in order

-- Table first
-- UDDT next
-- DEFAULT last

DROP TABLE TestDefaults

EXEC sp_droptype 'UDDTSales'

EXEC sp_droptype 'UDDTLoginDB'

DROP DEFAULT NoSales

DROP DEFAULT ThisMonth

DROP DEFAULT UserDB

Type added.
Type added.

No defaults defined

ID TotalSales SalesMonth WhoWhere

------ ---------------- ---------- --------------
1 NULL NULL NULL

Default bound to column.

Only DEFAULT on TotalSales defined

ID TotalSales SalesMonth WhoWhere

------ ---------------- ---------- --------------
1 NULL NULL NULL
2 .0000 NULL NULL

275
Microsoft SQL Server 2000 Programming by Example

Default bound to column.

DEFAULT defined on TotalSales and SalesMonth

ID TotalSales SalesMonth WhoWhere

------ ---------------- ---------- --------------
1 NULL NULL NULL
2 .0000 NULL NULL
3 .0000 11 NULL

Default bound to data type.

The new default has been bound to columns(s) of the specified user data type.

DEFAULT defined on TotalSales, SalesMonth and the UDDTLoginDB User-Defined Data

Type

ID TotalSales SalesMonth WhoWhere

------ ---------------- ---------- --------------
1 NULL NULL NULL
2 .0000 NULL NULL
3 .0000 11 NULL
4 .0000 11 sa - ByExample

Add an empty field using the UDDTSales data type

ID TotalSales SalesMonth WhoWhere ProjectedSales
------ ---------------- ---------- ---------------- ---------------
1 NULL NULL NULL NULL
2 .0000 NULL NULL NULL
3 .0000 11 NULL NULL
4 .0000 11 sa - ByExample NULL

Default bound to data type.

DEFAULT defined on UDDTSales data type as future only does not affect the
existing fields
using this UDDT

ID TotalSales SalesMonth WhoWhere ProjectedSales

------ ---------------- ---------- ---------------- ---------------
1 NULL NULL NULL NULL
2 .0000 NULL NULL NULL
3 .0000 11 NULL NULL
4 .0000 11 sa - ByExample NULL
5 .0000 11 sa - ByExample NULL

Type has been dropped.

Type has been dropped.

Rule Objects

RULE objects are similar to CHECK constraints. However, you can create a RULE object independently of any
column and bind it later to specific columns or user-defined data types.
To create a RULE object, you use the CREATE RULE statement, providing a unique name for the RULE object
and defining the object as any expression that returns TRUE or FALSE.
Caution

You can't use user-defined functions as part of a DEFAULT or RULE object definition.

276
 Chapter 7. Enforcing Data Integrity

To delete a RULE object, you must use the DROP RULE statement. You cannot drop a RULE object if it is used
anywhere in your database.

To bind a RULE object to a field or user-defined data type, you must use the sp_bindrule system stored
procedure, and the sp_unbindrule disconnects a bound RULE object from a field or user-defined data type.

You can bind only one rule to a user-defined data type or a table field. However, a rule can coexist with one or
more CHECK constraints in a field; in this case, all the conditions will be checked. If you bind a new rule to a
field or user-defined data type, the old rule will be unbound automatically.

You can see an example of how to use RULE objects in Listing 7.20.

Listing 7.20 Create Independent RULE Objects and Bind Them Later to Any Field or User-Defined Data
Type

-- Define a Table to test RULE Creation

CREATE TABLE NewEmployees (

EmployeeID int NOT NULL,
EmployeeName varchar(50) NOT NULL,
PostCode char(5) NOT NULL )
GO

-- Create the RULE object

CREATE RULE RUPostCode

AS
(@PCode LIKE '[0-9][0-9][0-9][0-9][0-9]')
GO

-- Bind the RULE to the PostCode column

EXEC sp_bindrule 'RUPostCode', 'NewEmployees.PostCode'

GO

-- Insert data in the table to test the RULE

INSERT NewEmployees
VALUES (1, 'Paul', 'GL513')

INSERT NewEmployees
VALUES (2, 'Eladio', '01380')

SELECT *

277
Microsoft SQL Server 2000 Programming by Example

FROM NewEmployees

GO

DROP TABLE NewEmployees

GO

DROP RULE RUPostCode

GO

Rule bound to table column.

Server: Msg 513, Level 16, State 1, Line 1
A column insert or update conflicts with a rule imposed by a previous CREATE RULE
statement. The statement was terminated. The conflict occurred in database
'ByExample',
table 'NewEmployees', column 'PostCode'.
The statement has been terminated.
EmployeeID EmployeeName PostCode
----------- ------------------------------ --------
2 Eladio 01380
Note

The definition of the RULE object contains a variable. The name of this variable is not relevant; it
just represents the column to where the RULE object will be bound.

Note

Remember to keep it simple. Overengineering a database will produce execution overhead and a
difficult maintenance.

What's Next?

This chapter covered the creation and use of structures to enforce data integrity.
Chapter 8 covers the creation of stored procedures, where you can test the integrity of the data before
attempting any modification, having extra data control and access to more complex condition checking.
Chapter 9 covers triggers, which is another way to enforce data integrity. In that chapter, you will see how to
create triggers to enforce domain integrity and referential integrity.
User-defined functions are covered in Chapter 10. It is possible to use UDF as part of constraint definitions.
This new feature gives you tremendous flexibility in the definition of DEFAULT and CHECK constraints.

278
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

Chapter 8. Implementing Business Logic: Programming

Stored Procedures
A stored procedure is a database object that comprises one or more Transact-SQL statements. The main
difference between a stored procedure and a set of statements is that a stored procedure can be reused just
by calling its name. Therefore, if you want to rerun the code, you don't have to execute the whole set of
statements that compose the stored procedure one by one.
As a database developer, you will spend most of your time coding, fixing, and optimizing stored procedures
because they can be used for thousands of purposes. Not only can they be used to encapsulate business
logic for your applications, they also can be used for administrative purposes inside SQL Server.
This chapter teaches you the following:

• The benefits of using stored procedures

• The types of stored procedures in SQL Server
• The types of stored procedure parameters
• How to create, alter, and execute stored procedures
• How to handle errors in stored procedures
• Security considerations when working with stored procedures

Benefits of Using Stored Procedures

Usually, stored procedures are used to encapsulate or enforce business rules in your databases. For example,
if you have to do some calculations before inserting data in a table, you can embed this logic in a stored
procedure and then insert the data using this stored procedure. Similarly, if you don't want users to directly
access tables and any other objects, you can create stored procedures to access these objects and have
users use them, instead of manipulating objects directly. For example, Microsoft discourages users from
making direct modifications to system tables; however, SQL Server comes with system stored procedures to
manipulate system tables.
Caution

If you develop applications that modify system tables, you should stop doing this. Be advised that
in future releases of SQL Server, Microsoft won't allow users to modify system tables directly.

The following are the benefits and advantages of stored procedures:

• They are precompiled statements— An execution plan (or access plan) is created and stored in
memory the first time the stored procedure is run, and it is subsequently used each time you execute
the stored procedure, thus minimizing the time it takes to run. This is more efficient than executing
each statement separately, one by one, because SQL Server would have to generate an access plan
for each statement every time it is run.
• They optimize network traffic— You might say that stored procedures aren't related to network traffic
at all. However, when you execute a stored procedure that contains many statements, you just have
to call the stored procedure once, not each statement separately. In other words, the entire block of
code (the whole set of statements) doesn't need to be sent from the client to the server. For example,
if you create a stored procedure with 10 statements and execute it, you need to send only one
instruction to SQL Server instead of 10 separate instructions. This translates into fewer round trips to
SQL server, thus optimizing network traffic.
• They can be used as a security mechanism— In particular, if the owner of an object doesn't want to
give direct permissions to users on database objects, he can create stored procedures that
manipulate these objects, and then give execute permissions on these stored procedures to users.

279
Microsoft SQL Server 2000 Programming by Example

This way, users will be allowed only to execute these stored procedures, and they won't be able to
directly manipulate the objects that stored procedures reference. System stored procedures are an
example of this approach. SQL Server provides system stored procedures to prevent users from
dealing directly with system tables.
• They allow modular programming— You can encapsulate your business logic inside stored
procedures, and then just call them from applications. Therefore, all statements that make up a stored
procedure are executed as a whole in the server. Furthermore, you can embed conditional logic in a
stored procedure using any of the control of flow statements (IF...ELSE, WHILE) available in
Transact-SQL.
• They can be set to execute automatically when SQL Server starts— Any routine task that must be
executed whenever the SQL Server service starts can be programmed as a stored procedure and
then configured to run automatically using the sp_procoption system stored procedure.
• They can use parameters— This is one of the ways that stored procedures have to receive data from
and return it to the calling application. Parameters can be either input, which are similar to variables
passed by value, or output, which behave as variables passed by reference.

Types of Stored Procedures

In SQL Server, there are four types of stored procedures: system stored procedures, user-defined stored
procedures, temporary stored procedures, and extended stored procedures. System and extended stored
procedures are created automatically at installation time. The other types (user-defined, temporary) are the
ones users create explicitly.

System Stored Procedures

System stored procedures are created automatically in system databases when you install SQL Server. They
are basically a way to interact with system tables. Moreover, there is a system stored procedure for almost
any administrative task you perform in SQL server. Also, because Microsoft doesn't recommend dealing
directly with system tables, this is the preferred way to deal with them.
Every global system stored procedure's name has the sp_ prefix, and for this reason they can be executed
from any database. Listing 8.1 demonstrates this feature, calling the sp_helpdb system stored procedure
(which gives general information about databases) from the Northwind database.

Listing 8.1 Executing a System Stored Procedure (Which Is Stored in Master) from the Northwind
Database

USE Northwind
GO

sp_helpdb

280
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

--The output has been simplified

name db_size owner dbid created compatibility_level

------------ ------------- --------- ------ ----------- -------------------
master 12.19 MB sa 1 Aug 6 2000 80
model 1.13 MB sa 3 Aug 6 2000 80
msdb 13.50 MB sa 4 Aug 6 2000 80
Northwind 3.94 MB sa 6 Aug 6 2000 80
pubs 2.13 MB sa 5 Aug 6 2000 80
tempdb 8.50 MB sa 2 Jan 22 2001 80
Transact-SQL provides a system function, OBJECTPROPERTY,that is usedto check for a variety of object
properties. Specifically, the property 'IsMSShipped' checks whether an object is a system object. Thus, it
can be used to identify whether a stored procedure is a system stored procedure. This system function, like
many others in SQL Server, receives the object's ID as a parameter, which can be obtained using the
OBJECT_ID system function. The OBJECTPROPERTY function returns 0 if the property is true, or 1 if not.
Listing 8.2 shows the use of this property.

Listing 8.2 Using the OBJECTPROPERTY System Function to Check Whether an Object Was Created
During SQL Server Installation

USE Master
SELECT OBJECTPROPERTY(OBJECT_ID('sp_help'),'IsMSShipped')
GO
-----------
1

(1 row(s) affected)
Caution

Books Online states that 'IsMSShipped' returns 1 (true) for any object created in the SQL
Server installation process. This is not completely true, because 'IsMSShipped' returns 0 (false)
for any user object created when SQL Server was installed— for example,
Northwind.dbo.Shippers. Therefore, 'IsMSShipped' returns 1 for any system object
created at installation time. Notice that although Pubs and Northwind are created during the
installation process, they are not considered system databases.

User-Defined Stored Procedures

281
Microsoft SQL Server 2000 Programming by Example

You create user-defined stored procedures in SQL Server to implement business logic. Any task, no matter
how simple or complex, that comprises multiple statements and conditions can be programmed as a stored
procedure, and then the calling application just needs to execute the stored procedure, instead of executing
the whole set of statements separately.
User-defined stored procedures are created using the CREATE PROCEDURE statement, and then SQL Server
stores them in the current database.
Stored procedures'names, like any other object's name, must be unique within the database and unique to the
user who creates them (the owner). Hence, in a certain database, it is possible that two stored procedures
exist with the same name but with different owners.
Any stored procedure that is created in the master database with the sp_ prefix— for example,
sp_myprocedure—can be accessed from any other database. In general, when a stored procedure is
executed and its name has the sp_ prefix, SQL Server looks for it, first in the current database, and then, if it's
not found in the current database, SQL Server looks for it in the master database.
Caution

If you create a user-defined stored procedure in any database other than master, with the sp_
prefix on its name, and there is a stored procedure in master with the same name, the user-defined
stored procedure that resides in the user's database will be executed only when called from the
user database. This is because when SQL Server executes any stored procedure that contains the
sp_ prefix, SQL Server looks for it first in the current database, and then in master if it doesn't find
it in the current database. Be aware that Books Online incorrectly states that SQL Server looks for
it first in master and then in the current database.

For example, you can create a user-defined stored procedure in master, as Listing 8.3shows, and call it from
other databases.

Listing 8.3 Creation of a Stored Procedure, with the sp_ Prefix, in Master, and Execution in Pubs

USE Northwind
GO
CREATE PROCEDURE sp_showdatabasename
AS
SELECT 'Northwind'
GO

USE Master
GO
CREATE PROCEDURE sp_showdatabasename
AS
SELECT 'Master'
GO

-- When executed from Northwind, SQL Server executes

-- the one stored in Northwind
USE Northwind

282
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

EXEC sp_showdatabasename
GO
-- When executed from Pubs, SQL Server executes
-- the one stored in Master, because there isn't
-- a stored procedure called sp_showdatabasename
-- in the Pubs database
USE Pubs
EXEC sp_showdatabasename
GO
---------

Northwind

(1 row(s) affected)

------
Master

(1 row(s) affected)

Temporary Stored Procedures

These are stored procedures created by users and stored in the tempdb database. They are called temporary
because they are dropped automatically by SQL Server, unless you explicitly issue a DROP PROCEDURE
statement. Like any other temporaryobject in SQL Server, when creating temporary stored procedures, use
the # prefix for local and the ## prefix for global temporary stored procedures. Listing 8.4 shows the creation
of a temporary stored procedure. After executing the code shown in Listing 8.4 in Query Analyzer, expand
the Stored Procedures folder of tempdb in the Object Browser, and you will see the stored procedure
#getdatabasename listed. Then, close the current connection to SQL Server (close the window if you're
working in Query Analyzer), and refresh the Stored Procedures folder of tempdb; the table will be gone.

Listing 8.4 Creation of a Temporary Stored Procedure

CREATE PROC #getdatabasename

AS
SELECT db_name() AS database_name
GO
Basically, temporary stored procedures have the same functionality as user-defined stored procedures, with
one exception; they are dropped when the connection that creates them is finished.
Tip

283
Microsoft SQL Server 2000 Programming by Example

A temporary stored procedure, once created (and stored in tempdb automatically by SQL Server),
can be called from any database.

Extended Stored Procedures

Extended stored procedures are DLL programs written in C++ that extend the capabilities of SQL Server.
They are located in the master database. SQL Server has its own set of extended stored procedures whose
name begins with xp_, which are used mainly for administrative purposes. However, there are some
extended stored procedures that start with sp_ just to consider them as global— for example, sp_OACreate.
You can create your own extended stored procedure, coding a DLL using C++ and then adding it to SQL
Server as an extended stored procedure, using the sp_addextendedproc system stored procedure. Be
very careful when coding extended stored procedures (trap any kind of errors, deallocate memory, and so on)
because they run in the same memory space as SQL Server; thus, any error in an extended stored procedure
can crash SQL Server.

Creating and Dropping Stored Procedures

Stored procedures are created using the CREATE PROCEDURE statement or the equivalent statement CREATE
PROC. When a stored procedure is created, its properties are stored in the sysobjects system table, and its
definition (all the statements it contains) in the syscomments system table. A stored procedure is stored
in the current database; therefore, if you want to create a stored procedure in other databases, you have to
make the other database the current one before creating it (using the USE statement ).
After a stored procedure is created, you can view its parameters and definition using the sp_helptext
system stored procedure. You can view its properties using sp_help.
In Listing 8.5, you can see an example of the syntax used to create a stored procedure. Followed by the
creation, it shows the retrieval of the stored procedure's properties, using sp_help, and then its code, using
sp_helptext.

Listing 8.5 Creating a Stored Procedure and Retrieving Its Properties and Code

USE Northwind
GO

CREATE PROC dbo.getcurrenttime

AS
SELECT CURRENT_TIMESTAMP
GO
EXEC sp_help 'getcurrenttime'
EXEC sp_helptext 'getcurrenttime'
GO

284
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

Name Owner Type Created_datetime

----------------- ------- ------------------- -------------------------
getcurrenttime dbo stored procedure 2000-09-18 01:35:06.257

Text
----------------------------------
CREATE PROC getcurrenttime
AS
SELECT CURRENT_TIMESTAMP
There are three steps that SQL Server performs with stored procedures: parsing, name resolution, and
optimization.
SQL Server parses a stored procedure when it is created to check for correct syntax. Then, the stored
procedure's information is stored in sysobjects and syscomments.
The first time the stored procedure is executed, SQL Server checks that all the objects it references exist. This
is a feature of SQL Server called deferred name resolution, which allows you to create stored procedures that
reference objects that haven't been created yet. This is why this step is performed the first time the stored
procedure is executed, not when it is created.
In the last step, SQL Server finds an optimized execution plan, looking for the best way to execute each
statement inside the stored procedure. Then, an optimized execution plan is generated and stored in the
procedure cache, which is part of the memory that SQL Server allocates for its use (the other part of the
memory, the data cache, is used to store the data pages that SQL Server manipulates).
Figure 8.1 shows this three-step process (parse, name resolution, and optimization).

Figure 8.1. Creation and execution of stored procedures in SQL Server.

The execution plan of a stored procedure will remain in memory until SQL Server is stopped or when SQL
Server needs the memory allocated for the plan. Therefore, if the procedure cache becomes full, stored plans
are dropped to make space for new ones.
After the execution plan is created and stored in the procedure cache (memory), any time you execute the
stored procedure, SQL Server just needs to reuse the plan to manipulate the data. SQL Server shows this
cache information if you query the syscacheobjects system table. Be aware that syscacheobjects is a
virtual table, not a real one. The only purpose of this virtual table is to provide support for internal procedures
and DBCC commands, and the table is filled automatically with data when you use it. Specifically, you can
retrieve information about the procedure cache by querying this virtual table
(master.dbo.syscacheobjects).

285
Microsoft SQL Server 2000 Programming by Example

The process of generating a good access plan involves evaluating many factors, such as indexes and data in
tables. This is one of the reasons you should have good indexes on tables and views referenced by stored
procedures, and also keep statistics up to date, which is a database option that is set by default when you
create a database.
Tip

In a stored procedure, it's better to create the objects first (DDL), and then manipulate them (DML),
because this prevents the query processor from recompiling the stored procedure while it is
executed.

In SQL Server 6.5 and earlier, the only way to create an access plan was by using stored procedures. In
version 7.0 and later, the query processor can store execution plans in the procedure cache for all Transact-
SQL statements (including ad hoc queries). When reexecuting a Transact-SQL statement, if the query
processor detects that it can reuse the plan, it takes it from the procedure cache, optimizing the execution
time of the whole statement.

A feature of stored procedures, as mentioned earlier, is that they can be set to execute automatically when
the SQL Server service is started. Because they won't have any interaction with any application, they can't
have any input parameters. The stored procedure must be created by the system administrator in the master
database, and then the system stored procedure sp_procoption must be used to set it to execute when the
SQL Server service is started.

For example, suppose that you want to be able to know every time the SQL Server service was started. To
accomplish this, you can create a table in master to store the date and time when the SQL Server service has
been started, and then create a stored procedure that inserts a row in this table with the current date. Finally,
set this stored procedure to execute automatically whenever SQL Server is started. Listing 8.6 shows the
code needed to achieve these steps.

Listing 8.6 Using the sp_procoption System Stored Procedure

USE Master
GO

CREATE TABLE dbo.Sqlstatus (

lasttime DATETIME
)
GO

CREATE PROC dbo.insertsqlstatus

AS
INSERT Sqlstatus (lasttime)
VALUES (CURRENT_TIMESTAMP)
GO

286
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

EXEC sp_procoption 'insertsqlstatus','startup','true'

To test this example, follow the next steps:

1. Using Query Analyzer, connect to SQL Server as sa, or if using integrated authentication, with a
member of the System Administrators server role.
2. Run the code shown in Listing 8.6, which will create the Sqlstatus table and the
insertsqlstatus stored procedure, and then set this stored procedure to run automatically
whenever SQL Server is started.
3. Close any applications that might be using SQL Server (Query Analyzer, Enterprise Manager, and so
on).
4. Stop and restart SQL Server.
5. Connect to SQL Server using Query Analyzer, and issue a SELECT query against the Sqlstatus
table.

To verify that a stored procedure that was configured to execute automatically was successfully executed, you
can check the SQL Server error log. The error log will show the following message to indicate that the stored
procedure was executed successfully:

Launched startup procedure 'name_of_the_stored_procedure'

Tip

Another way to find the last time when SQL Server was started is by using the crdate column in
the sysdatabases system table in master. This column stores the creation date of the database,
and because tempdb is re-created every time the SQL Server service starts, you can get the last
time that SQL Server was started.

Some statements can't be included in a stored procedure's code. These statements are CREATE DEFAULT,
CREATE PROCEDURE, CREATE RULE, CREATE TRIGGER, and CREATE VIEW.

Stored procedures can be createdusing the WITH ENCRYPTION option, which encrypts the definition in the
syscomments system table; therefore, nobody can read the definition. If you try to see the code of a stored
procedure (using sp_helptext or any other method) and it has been encrypted, you will get this error

The object comments have been encrypted

Be cautious when you encrypt a stored procedure's definition, because you won't be able to display it again
unless you keep the original source code. Therefore, if you need to modify the definition of a stored procedure
that was created using the WITH ENCRYPTION option, you must use the original source code. It is always a
good idea to keep a copy of the original scripts that you used to generate the database schema.
Listing 8.7 creates the getcurrentuser stored procedure using the WITH ENCRYPTION option, and then
tries to show the code of the stored procedure using sp_helptext, without success.

Listing 8.7 Creation of a Stored Procedure Using the WITH ENCRYPTION Option

287
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

CREATE PROC dbo.getcurrentuser

WITH ENCRYPTION
AS
SELECT USER
GO

sp_helptext 'getcurrentuser'

The object comments have been encrypted.

Using Parameters
Like any function or procedure in any other programming language, stored procedures communicate with
applications or clients through parameters. The maximum number of parameters in a stored procedure is
2,100 (this was significantly increased from SQL Server 7, which had a maximum of 1,024 parameters per
stored procedure).
Caution

Be aware that Books Online incorrectly states in the "Maximum Capacity Specifications" section
that the maximum number of parameters in a stored procedure is 1,024.

When you develop stored procedures, to be able to access the value of a parameter inside the stored
procedure's body, you just have to specify the parameter's name (including the @ character).

Once created, information about stored procedures' parameters is stored in the syscolumns system table
(you already know that sysobjects stores general information and syscomments stores the code of the
stored procedure).

Parameters are defined right after the stored procedure's name when creating the stored procedure. The
parameter's name must have the @ character as the first character (like any variable in Transact-SQL). After
the name of the parameter, the data type must be specified, and then a default value, if there's one (the
default value is optional).

288
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

Listing 8.8 shows an example of the creation of a stored procedure (getemployeesbylastname) that
contains a parameter (@emplastname). This stored procedure gets the employees whose last name contains
the string indicated by the @emplastname parameter. Notice that, when creating stored procedures,
parameters are declared between the stored procedure's name and the AS keyword.

Listing 8.8 Creation of a Stored Procedure Using Parameters

USE Northwind
GO

CREATE PROC dbo.getemployeesbylastname

@emplastname VARCHAR(40)
AS
SELECT *
FROM Employees
WHERE lastname LIKE '%'+ @emplastname + '%'
GO
The default value of a parameter can be set to NULL. If a parameter doesn't have a default value, a value
must be supplied by the calling application when executing the stored procedure. On the other hand, if a
parameter has a default value, the calling application doesn't have to supply a value for this parameter if it
wants to use the default.
Listing 8.9 creates a stored procedure (getemployeesbylastname_default, a slight variation of the
stored procedure shown in Listing 8.8), which contains a parameter (@emplastname) with a default value.
Notice that the default value is specified just after the parameter's data type.

Listing 8.9 Creation of a Stored Procedure Using Default Parameters

USE Northwind
GO

CREATE PROC dbo.getemployeesbylastname_default

@emplastname VARCHAR(40) = 'a'
AS
SELECT *
FROM Employees
WHERE lastname LIKE '%'+ @emplastname + '%'
GO
There are two types of parameters, input and output:

• An input parameter is similar to a variable passed by value. Therefore, the stored procedure gets a
copy of the data and this doesn't affect the data outside the stored procedure. In other words, if you

289
Microsoft SQL Server 2000 Programming by Example

pass a variable as a parameter of a stored procedure, and the value of this variable is modified inside
the stored procedure, this doesn't change the value of the variable outside the stored procedure.
• An output parameter is like a variable passed by reference. Hence, because the stored procedure
gets a pointer to a variable, any changes made to it are reflected outside the scope of the stored
procedure. Using this type of parameter, a stored procedure can send values back to the calling
application. To take advantage of output parameters, and to distinguish them from input parameters,
the OUTPUT keyword must be specified when creating the stored procedure, and also when it is
executed.

Listing 8.10 shows the creation of a stored procedure (getemployeeaddress) that contains an input
parameter (@employeeid) and an output parameter (@employeeaddress). This stored procedure stores
the complete address of a given employee in the @employeeaddress output parameter. Notice that the
OUTPUT keyword must be specified when declaring output parameters.

Listing 8.10 Using Input and Output Parameters

USE Northwind
GO

CREATE PROC dbo.getemployeeaddress

@employeeid INT,
@employeeaddress NVARCHAR(120) OUTPUT
AS

SELECT @employeeaddress = address + '. '+ city + '. '+ region + '. '
postalcode + '. '+ country
FROM Employees
WHERE employeeid = @employeeid
GO
An advantage of using stored procedures is that they can return result sets, using SELECT statements in the
body of the stored procedure. However, one of the limitations of using parameters is that you can't use a
parameter to pass the name of a database object (table, column, or stored procedure) to the stored procedure.
For this purpose, you must build the query at runtime, generating a dynamic query (using EXEC or
sp_executesql). Notice that this is not a restriction of parameters; it is a restriction of the Data Definition
Language (DML).
To illustrate this idea, imagine that you want to create a stored procedure with one parameter, and this
parameter is the table you want to query. Listing 8.11 shows the code necessary to create this stored
procedure, using the EXEC statement.

Listing 8.11 Using Objects As Parameters and Building Queries at Runtime

USE Northwind

290
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

GO

CREATE PROC dbo.issuequery

@tablename NVARCHAR(256)
AS
DECLARE @query NVARCHAR(1000)
SET @query = 'SELECT * FROM '+ @tablename
EXEC (@query)
GO

Altering Stored Procedure Definitions

The code of a stored procedure can be modified using the ALTER PROCEDURE statement,or its equivalent
ALTER PROC. In SQL Server 6.5 and earlier, the only way to change a stored procedure's definition was to
drop and re-create it, but this approach has one drawback: Permissions and properties set on the stored
procedure are lost. Therefore, after re-creating the stored procedure, the database administrator had to set
permissions again.
Listing 8.12 modifies the definition of the stored procedure created in Listing 8.11. The new stored
procedure, in addition to the table's name, receives a column's name as a parameter.

Listing 8.12 Using ALTER TABLE to Modify the Code of a Stored Procedure

USE Northwind
GO

ALTER PROC dbo.issuequery

@tablename NVARCHAR(256),
@columname NVARCHAR(256)
AS
DECLARE @query NVARCHAR(1000)
SET @query = 'SELECT '+ @columname + 'FROM '+ @tablename
EXEC (@query)
GO
When you alter a stored procedure's definition (using the ALTER PROC statement):

• SQL Server keeps permissions intact on the stored procedure. As a result, any permissions set on the
stored procedure are kept after changing the stored procedure's code using ALTER TABLE.
• This doesn't affect any dependent objects (tables, triggers, or stored procedures). For example, if you
alter a stored procedure's definition and it references a table, the table isn't affected.
• This doesn't affect the property to run automatically when SQL Server starts, if this was previously set
using the sp_procoption system stored procedure. For example, if you alter the code of the stored
procedure created in Listing 8.6 (insertsqlstatus, which was set to run automatically whenever
SQL Server is started), SQL Server keeps this property intact.

In other words, if you either want to change the procedure's code without affecting permissions and properties,
or want to change the options of the stored procedure (WITH ENCRYPTION or WITH RECOMPILE), you can
use the ALTER PROCEDURE statement. However, notice that if you just need to change an option, you still
must specify the entire code of the stored procedure. Similarly, if you just have to change the code and
preserve the options,you also must specify the options.

291
Microsoft SQL Server 2000 Programming by Example

For example, if you want to encrypt the code shown in Listing 8.12, you would have to add the WITH
ENCRYPTION option to the definition of the stored procedure. Listing 8.13 shows you how to accomplish this,
and also shows that the code is in fact encrypted after executing this script.

Listing 8.13 Using ALTER TABLE to Modify the Code of a Stored Procedure

USE Northwind
GO

ALTER PROC dbo.issuequery

@tablename NVARCHAR(256),
@columname NVARCHAR(256)
WITH ENCRYPTION
AS
DECLARE @query NVARCHAR(1000)
SET @query = 'SELECT '+ @columname + 'FROM '+ @tablename
EXEC (@query)
GO

sp_helptext issuequery
GO

The object comments have been encrypted.

Notice that if you only want to add an option to the stored procedure's code (WITH ENCRYPTION, in the
previous example), you still have to specifythe entire code.

The RETURN Statement

The RETURN statement is used to exit unconditionally from a stored procedure. In other words, if SQL Server
reaches a RETURN statement when executing a stored procedure, it stops processing and returns the control
to the calling application.
The RETURN statement has one parameter,the return value,which is an integer that can be used to
communicate with the calling application. When creating a stored procedure, if you use a data type other than
integer for the return value, SQL Server allows you to create the stored procedure, but you will get an error
when it is executed.
The return value is 0 by default; therefore, if a stored procedure containsa RETURN statement without this
parameter, the return value will be 0. Therefore, it is equivalent to say RETURN 0 or RETURN. Similarly, if a
stored procedure doesn't have any return statement at all, the return value is 0.
In general, a return value of 0 indicates a successful completion of the stored procedure. Any return value
other than 0 usually indicates that there was an error in the execution of the stored procedure. The general

292
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

convention used in system stored procedures is 0 means success, and any other value indicates that an error
occurred.
Usually, the RETURN statement is very useful in the error-checking phase of the stored procedure; thus, if
there's any error that you want to trap in the calling application, the RETURN statement can be used to return
an error code.
Because you can use numbers other than 0 to return error codes to the calling application, if you want to have
customized error codes in your application, you can choose a number for each type of error, and then when
the application receives one of these error codes, it knows how to interpret them.
Listing 8.14 shows an example of a stored procedure (getemployee) that uses return values to indicate
whether a certain employeeid exists in the Employees table. Getemployee returns –1 if the employeeid
doesn't exist in the Employees table, and returns 0 if it does exist. Notice that the second RETURN statement
doesn't have the return value, so it's 0 (the default).

Listing 8.14 Using the RETURN Statement in Stored Procedures

USE Northwind
GO

CREATE PROC dbo.getemployee

@employeeid INT
AS
IF NOT EXISTS (SELECT * FROM Employees WHERE employeeid = @employeeid)
RETURN -1
ELSE
SELECT * FROM Employees WHERE employeeid = @employeeid

RETURN
GO
Caution

The return value is not a stored procedure's output parameter; these are different things. Database
developers sometimes confuse the return value with an output parameter. Keep in mind that a
stored procedure can have more than one output parameter but just one return value.

Executing Stored Procedures

There are a variety of ways to execute stored procedures. All depend on the calling application, language
used, and the programming interface (OLE-DB, ODBC, ADO, and so on). In Transact-SQL, the basic syntax
to execute a stored procedure is the following:

EXECUTE @return_value = procedure_name parameter_1,..,parameter_n

The EXECUTE statementmust be used if there's more than one instruction in the batch. Otherwise, if you want
to execute just the stored procedure and there are no more instructions in the batch, you can omit the
EXECUTE statement.
Tip

293
Microsoft SQL Server 2000 Programming by Example

If there's more than one instruction in the batch and the stored procedure is called in the first line of
the batch, you can omit the EXECUTE statement.

There are two ways to specify input parameters when executing a stored procedure:

• Use the name of the variables used in the parameter declaration of the stored procedure and their
value— With this approach, you can omit variables if you want to use their default values. Also, the
order of the parameters is not important. For example, Listing 8.15 creates a stored procedure that
inserts a row in the Customers table, and then executes it. Notice that all the parameters that don't
have a default value have to be specified.

Listing 8.15 Executing a Stored Procedure with Parameters

USE Northwind
GO

CREATE PROC dbo.InsertCustomer

@customerid NCHAR(10),
@companyname NVARCHAR(80),
@contactname NVARCHAR(60),
@contacttitle NVARCHAR(60) = 'Owner',
@address NVARCHAR(120) = NULL,
@city NVARCHAR(30) = 'Miami',
@region NVARCHAR(30) = 'FL',
@postalcode NVARCHAR(20) = '33178',
@country NVARCHAR(30) = 'USA',
@phone NVARCHAR(48) = NULL,
@fax NVARCHAR(48) = NULL
AS

INSERT INTO Customers (customerid,companyname,contactname,contacttitle,address,

city,region,postalcode,country,phone,fax)
VALUES (@customerid,@companyname,@contactname,@contacttitle,@address,@city,
@region,@postalcode,@country,@phone,@fax)
GO

InsertCustomer @customerid='MACMI',@contactname='Carlos Eduardo Rojas',

@companyname = 'Macmillan'
GO

• Use just the actual values that you want to pass to the stored procedure— With this method, the order
of the values is important. For this reason, values must be specified in the same order in which
variables appear in the parameter declaration section of the stored procedure. Also, default values
can be used, but they must be the last ones in the parameter declaration; otherwise, you would break
the sequence. Listing 8.16 shows the execution of the stored procedurecreated in Listing 8.15,
using this approach.

Listing 8.16 Another Wayto Pass Parameters When Calling a Stored Procedure

294
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

USE Northwind
GO

InsertCustomer 'QUEPU','QUE Publishing','Jesus Rojas'

GO

The values can also be passed as local variables used in the same batch in which the stored
procedure is being called. Listing 8.17 illustrates this variation.

Listing 8.17 Using Local Variables As Parameters When Calling a Stored Procedure

USE Northwind
GO

DECLARE @custid NCHAR(10),

@contname VARCHAR(60),
@compname VARCHAR(80)

SELECT @custid = 'SAMSP',

@contname = 'Maria Rojas',
@compname = 'Sams Publishing'

EXEC InsertCustomer @custid,@contname,@compname

GO
When output parameters are used in a stored procedure, the OUTPUT keyword must be specified again when
the stored procedure is executed. In addition to the OUTPUT keyword, a variable must be used to store the
value of the parameter after the stored procedure's execution. Listing 8.18 shows how to use output
parameters. It creates a stored procedure that gets customer information given its ID. Note that the code that
executes the stored procedure contains the OUTPUT keyword.

Listing 8.18 Using Output Parameters

USE Northwind
GO

295
Microsoft SQL Server 2000 Programming by Example

CREATE PROC dbo.getCustomerInfo

@customerid NCHAR(10),
@contact NVARCHAR(60) OUTPUT,
@company NVARCHAR(80) OUTPUT
AS
SELECT @contact = contactname,
@company = companyname
FROM Customers
WHERE customerid = @customerid
GO

DECLARE @customer_id NCHAR(10),@customer_name NVARCHAR(60),

@customer_company NVARCHAR(80)
SET @customer_id = 'SAMSP'

EXEC getCustomerInfo @customer_id, @customer_name OUTPUT,

@customer_company OUTPUT
SELECT @customer_name + '- '+ @customer_company
GO
-----------------------------------------------------------

Maria Rojas - Sams Publishing

(1 row(s) affected)
Caution

If the OUTPUT keyword is omitted when the stored procedure is executed, the parameter behaves
as an input parameter.

Listing 8.19 contains the execution of the same stored procedure executed in Listing 8.18, but 8.19 omits
the OUTPUT keyword in both variables. Notice that these values are lost after the stored procedure's execution.

Listing 8.19 Using Output Parameters Without the OUTPUT Keyword

USE Northwind
GO

296
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

DECLARE @customer_id NCHAR(10),@customer_name NVARCHAR(60),

@customer_company NVARCHAR(80)
SET @customer_id = 'SAMSP'
EXEC getCustomerInfo @customer_id, @customer_name, @customer_company
SELECT @customer_name + '- '+ @customer_company
GO
-------------------------------------------------------------

NULL

(1 row(s) affected)
If you want to process the return value of a stored procedure, you must store it in a variable when executing
the stored procedure. Listing 8.20executes the stored procedure created in Listing 8.14 (getemployee),
and demonstrates how to store the return value in a local variable for further processing.

Listing 8.20 Storing the Return Value of a Stored Procedure in a Variable

USE Northwind
GO

DECLARE @employeexists INT

EXEC @employeexists = getemployee 88
SELECT @employeexists
GO
-----------

-1
The result set returned byexecuting a stored procedure (if it contains a SELECT statement)can be inserted into
a table using the INSERT statement followed by the execution of the stored procedure. The data types of the
result set must be compatible with the ones of the table. Compatible means that the data types must be either
the same or they can be implicitly converted by SQL Server. Also, the number of columns of the stored
procedure's result set must match the table's definition. For example, if the stored procedure produces a result
set with three columns, you can't insert it in a table with two columns. In Listing 8.21, a stored procedure is

297
Microsoft SQL Server 2000 Programming by Example

created to getall the employees of a given country, and then a temporary table is created to store the result
set returned by the execution of the stored procedure.

Listing 8.21 Inserting in a Table a Result Set Returned by a Stored Procedure

USE Northwind
GO

CREATE PROC dbo.GetEmployeesCountry

@country NVARCHAR(30)
AS
SELECT employeeid,lastname,firstname
FROM Employees
WHERE country = @country
GO
CREATE TABLE #Employees_in_usa (
emp_id INT NOT NULL,
emp_lname NVARCHAR (20) NOT NULL,
emp_fname NVARCHAR (10) NOT NULL
)
GO

INSERT INTO #Employees_in_usa

EXEC GetEmployeesCountry 'USA'

SELECT * FROM #Employees_in_usa

GO

emp_id emp_lname emp_fname

----------- -------------------- ----------
1 Davolio Nancy
2 Fuller Andrew
3 Leverling Janet
4 Peacock Margaret
8 Callahan Laura

(5 row(s) affected)
A stored procedure canbe called from any database using fully qualified names. Notice that when calling a
stored procedure from any other database than the one where it resides (where it was created), you must fully
qualify the name of the stored procedure. This way, SQL Server knows where to look for the stored procedure.
If you call a stored procedure from the database where it resides, you just have to call it by its name. For
example, if you want to execute an extended stored procedure from any other database than master, you
must indicate that this stored procedure resides in master. Specifically, Listing 8.22 shows the execution of

298
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

xp_fixeddrives, which lists all the drives and space available, from the Northwind database. Notice that
the output you get can vary according to the number of drives available in your computer and the space
available on each one of them.

Listing 8.22 Using Fully Qualified Names to Call Stored Procedures

USE Northwind
GO

EXEC master..xp_fixeddrives

drive MB free
----- -----------
C 8315
D 8487
E 8316

(3 row(s) affected)

Using Query Analyzer's Object Browser to Execute Stored Procedures

In SQL 2000, the newobject browser (one of the best additions to the Query Analyzer) allows us to execute
stored procedures using a graphical interface. Using this method, you just have to enter the value of each
parameter using the GUI, and then Query Analyzer automatically generates the code necessary to execute
the stored procedure.
To execute a stored procedure using the object browser, follow these steps:

1. Open the Query Analyzer.

2. Connect to the server and choose the database.
3. Make sure that the object browser is open. If it is not open, choose Tools, Object Browser, or press F8.
4. In the object browser, expand a database, and then the stored procedures folder.
5. Right -click the stored procedure, and then click the Open option.
6. Query Analyzer opens the Execute Procedure window, in which you can enter the value of each
parameter.
7. Click the Execute button.

Caution

When entering any kind of string in the Execute Procedure window in the Query Analyzer, don't use
quotes to enclose the string.

299
Microsoft SQL Server 2000 Programming by Example

If you follow these steps to execute the stored procedure created in Listing 8.21 (GetEmployeesCountry),
you'll see the Execute Procedure window shown in Figure 8.2. Notice that the data type of the parameter is
NVARCHAR, and when typing the value of this parameter, you don't haveto use quotes to enclose it.

Figure 8.2. Executing stored procedures using the Execute Procedure option in Query Analyzer.

Figure 8.3 shows the results of the execution of the GetEmployeesCountry stored procedure. Query
Analyzer generated all the code needed to execute the stored procedure using the parametersyou provided in
the window.

Figure 8.3. Results of the execution of a stored procedure using the Execute Procedure option.

300
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

Stored Procedure Recompilation

As you already know, SQL Server creates an optimized execution plan, which is stored in memory, the first
time a stored procedure is executed. In general, you want SQL Server to reuse this execution plan for
subsequent executions of stored procedures. However, for diverse reasons, sometimes you might want to
force SQL Server to modify an execution plan. The reason might fall among one of these: The value of
parameters changed significantly, the objects referenced by the stored procedure changed in some way, the
data changed significantly or, last but not less important, indexes changed.
There are three ways to explicitly force SQL Server to generate another execution plan:

• Use the WITH RECOMPILE option when creating the stored procedure(CREATE PROC .. WITH
RECOMPILE)— With this approach, SQL Server doesn't cache the stored procedure's execution plan.
Instead, SQL Server compiles the stored procedure every time it is executed, generating a new
execution plan. Listing 8.23 illustrates a stored procedure created using the WITH RECOMPILE
option.

Listing 8.23 Creating a Stored procedure Using the WITH RECOMPILE Option

USE Northwind
GO

CREATE PROC dbo.GetEmployeesCountry2

@country NVARCHAR(30)
WITH RECOMPILE
AS
SELECT employeeid,lastname,firstname
FROM Employees
WHERE country = @country
GO

• At execution time, use the WITH RECOMPILE option (EXECUTE .. WITH RECOMPILE)— If this
method is used, SQL Server generates a new execution plan that is used in subsequent executions of
the stored procedure. Listing 8.24 shows the execution of a stored procedure, using the WITH
RECOMPILE option, forcing SQL Server to generate a new execution plan.

Listing 8.24 Using the WITH RECOMPILE Option in the Stored Procedure's Execution

USE Northwind
GO

EXEC GetEmployeesCountry 'USA'WITH RECOMPILE

301
Microsoft SQL Server 2000 Programming by Example

GO

• Use the sp_recompile system stored procedure— This is a slightly different way to recompile a
stored procedure. sp_recompile receives the name of an object as a parameter; the object can be
a stored procedure, table, view, or trigger. If a stored procedure's name or a trigger's name is used as
a parameter, this object (trigger or stored procedure) is recompiled the next time it is executed. On the
other hand, if the parameter specified is the name of a table or a view, any stored procedure that
references this table or view will be recompiled the next time it is executed. This is the preferred way
to request recompilation of all stored procedures (that reference a specific table or view) with just one
instruction. Listing 8.25 shows the use of the sp_recompile system stored procedure, forcing SQL
Server to recompile any stored procedure that references the Employees table in the Northwind
database.

Listing 8.25 Using sp_recompile to Force SQL Server to Generate a New Execution Plan for Every
Stored Procedure That References the Authors Table

USE Northwind
GO

sp_recompile 'Employees'
GO

Object 'Employees'was successfully marked for recompilation.

Handling Errors
An important element of any program you write is the error-checking section. During software development, it
is a good programming technique to check for errors in your code, and abort the execution of the program or
trap the error when it is found. If the program crashes, there's a high probability that it crashed because you
neglected to check or trap an error.
Transact-SQL provides twoelements that allow us to check and throw errors programmatically. These
elements are the @@ERROR parameterless(or niladic) system function and the RAISERROR statement.
The @@ERROR system function returns the error code (an integer different from 0) of the last statement
executed, if there was an error. On the other hand, if the last statement was executed successfully, @@ERROR
returns 0. Be aware that this value changes from one statement to the next one; hence, you must check this
value right after the statement is executed.
RAISERROR is used to explicitly throw an error. You can either use an ad hoc message or a message stored
in the Sysmessages system table (all SQL Server error messages are stored in Sysmessages). You can
add your own messages to this system table through the sp_addmessage system stored procedure, and to
delete messages, through sp_dropmessage. Notice that when creating a user-defined message with

302
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

sp_addmessage, you must specify a message ID greater than 50,001 (message IDs less than 50,000 are
reserved by SQL Server).
This is the syntax of RAISERROR:

RAISERROR (msg_id | msg_text, severity, state) WITH option

The first parameteris the message ID or the message text. If you specify a message ID, you need to have
previously created a user-defined message with sp_addmessage. If you want to use an ad hoc message,
the message text can have up to 400 characters. The second parameter is the severity level of the error,
which is a number between 0 and 25 (severity levels greater than 20 must be used by system administrators
for critical errors). If the severity level falls in the range of 0 through 10, it's considered an informational
message. Then, severity levels from 10 to 19 are used for trappable errors, and from 20 to 25 for critical errors
(which close the connection after the client receivesthe error message).
The third parameter, the state of the error, is an integer between 0 and 127 which, by the documentation, isn't
significant to SQL Server. Finally, there are two options (either can be used) in the last parameter, which are
optional:

• LOG—Stores the error information in the SQL Server error log and in the NT Application log. This
option must be specified when using severity levels higher than 19.
• NOWAIT—This option sends the error message immediately to the client application.

After executing RAISERROR, @@ERROR returns the value of the message IDof the error or, if you use an ad
hoc message, it will return 50,000.
Listing 8.26 demonstrates the use of sp_addmessage, @@ERROR, and RAISERROR.

Listing 8.26 Using @@ERROR and RAISERROR

USE Northwind
GO

sp_addmessage 50001,11,'An error occurred'

GO

CREATE PROC generateerror

AS
RAISERROR (50001,11,1) WITH LOG
SELECT @@ERROR
GO

generateerror
(1 row(s) affected)

303
Microsoft SQL Server 2000 Programming by Example

Server: Msg 50001, Level 11, State 1, Procedure generateerror, Line 4

An error occurred

-----------
50001

(1 row(s) affected)

Nesting Stored Procedures

Stored procedures can be nested up to 32 levels. Nested means that one stored procedure calls another, and
so on. If the nesting level exceeds 32, the execution of the whole group of stored procedures fails. SQL
Server provides a way to check the nesting level using the @@nestlevel system function.
When a stored procedure calls another stored procedure, the nesting level is incremented by one, and then
when the inner stored procedure finishes its execution, the nesting level decreases by one.
Listing 8.27 shows the creation of two stored procedures. The first one, CheckSupplier, returns –1 if a
given supplier name already exists in the Suppliers table, and 0 if it doesn't. The second one,
InsertSupplier, calls the CheckSupplier stored procedure to check whether the data is already stored
in the database, and if it's not, it's inserted. Observe in the output of Listing 8.27 that the nesting level is
incremented by one each time a stored procedure is called from another one.

Listing 8.27 Using the @@nestlevel System Function

USE Northwind
GO

CREATE PROC dbo.CheckSupplier

@supplier_name VARCHAR(40)
AS

PRINT '3) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))

IF EXISTS (SELECT * FROM Suppliers WHERE companyname = @supplier_name)

RETURN -1
ELSE
RETURN 0
GO

CREATE PROC dbo.InsertSupplier

@suppliername NVARCHAR(40),
@contactname NVARCHAR(30),
@contacttitle NVARCHAR(30)
AS
DECLARE @supplier_exists INT
PRINT '2) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
EXEC @supplier_exists = CheckSupplier @suppliername
PRINT '4) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
IF @supplier_exists = 0
INSERT INTO dbo.Suppliers (companyname,contactname,contacttitle)

304
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

VALUES (@suppliername,@contactname,@contacttitle)
ELSE
PRINT 'This supplier already exists in the database'
GO

PRINT '1) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))

EXEC InsertSupplier 'ACME','Fernando Guerrero','Owner'
PRINT '5) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
GO

The nesting level is 0

2) The nesting level is 1
3) The nesting level is 2
4) The nesting level is 1

(1 row(s) affected)

5) The nesting level is 0

When a main stored procedure calls other stored procedures, it must call them one after the other, and, as
you might already know, this is not considered nesting. Thus, because this is not considered nesting, you can
call more than 32 stored procedures from the main one. Listing 8.28 illustrates this situation. Pay close
attention to the nesting level in the output; the maximum value it reaches is 1 (because there's no nesting).

Listing 8.28 Using the @@nestlevel System Function

USE Northwind
GO

CREATE PROC dbo.ShowSecurityInfo

AS
EXEC sp_helpgroup
PRINT '2) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
EXEC sp_helpuser
PRINT '3) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
GO

PRINT '1) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))

EXEC ShowSecurityInfo
PRINT '4) The nesting level is '+ CAST(@@nestlevel AS VARCHAR(5))
GO

305
Microsoft SQL Server 2000 Programming by Example

1) The nesting level is 0

Group_name Group_id
------------------------------ --------
db_accessadmin 16385
db_backupoperator 16389
db_datareader 16390
db_datawriter 16391
db_ddladmin 16387
db_denydatareader 16392
db_denydatawriter 16393
db_owner 16384
db_securityadmin 16386
public 0

2) The nesting level is 1

UserName GroupName LoginName DefDBName UserID SID
---------- ---------------- --------- ------------ ------ --------
dbo db_owner sa master 1 0x01
guest public NULL NULL 2 0x00

3) The nesting level is 1

4) The nesting level is 0
An advantage of nesting is that when a stored procedure is called from another stored procedure, the inner
one can access all objects created by the outer stored procedure. For example, Listing 8.29 shows the
creation of two stored procedures, outerprocedure and innerprocedure, and as you can see,
innerprocedure can access the temporary table created by outerprocedure.

Listing 8.29 Accessing Objects Created by an Outer Stored Procedure, in an Inner Stored Procedure

USE Northwind
GO

CREATE PROC outerprocedure

AS
SELECT orderid, orderdate
INTO #Spain_orders
FROM Orders
WHERE Shipcountry = 'Spain'
AND Shipcity = 'Barcelona'
EXEC innerprocedure
GO

306
 Chapter 8. Implementing Business Logic: Programming Stored Procedures

CREATE PROC innerprocedure

AS
SELECT *
FROM #Spain_orders
GO

EXEC outerprocedure

orderid orderdate
----------- ------------------------------------------------------
10366 1996-11-28 00:00:00.000
10426 1997-01-27 00:00:00.000
10568 1997-06-13 00:00:00.000
10887 1998-02-13 00:00:00.000
10928 1998-03-05 00:00:00.000

(5 row(s) affected)

Application Security Using Stored Procedures

One of the advantages of stored procedures is that they can be used as a security mechanism to prevent
users from dealing directly with tables. The process is very straightforward: First, create the stored procedure,
and then assign execute permissions to the users on the stored procedure. Therefore, users don't need to
have permissions on every object referenced by the stored procedure. For example, if you create a stored
procedure that retrieves data from a certain table (using a SELECT query), you just have to grant execute
permissions on the stored procedure to the users, and then they will be able to run the stored procedure
(without having direct permissions on the table referenced by the stored procedure).
The first step SQL Server performs when a user executes a stored procedure is to check permissions on it. In
general, the user who is executing the stored procedure just has to have EXECUTE permissions onit. However,
there are two exceptions to this rule:

• If there's a dynamic query in the stored procedure (it contains either the EXECUTE statement or the
sp_executesql system stored procedure), the user executing it must have permissions on the
objects referenced by the dynamic query. In other words, SQL Server checks permissions on every
object referenced by the dynamic query. This is because if the stored procedure contains a dynamic
query at creation time, SQL Server doesn't know which objects are referenced by the dynamic query
until it is executed. To illustrate, if you create a stored procedure that accesses the Orders table
through a dynamic query, any user who executes the stored procedure, besides execute permissions,
has to have appropriate permissions in the Orders table.
• If the ownership chain is broken, SQL server checks permissions on each object with a different
owner, and only statements with sufficient permissions will be executed. This is why it's highly
recommended that the owner of a stored procedure owns all objects referenced by it, to avoid a
broken ownership chain.

For example, suppose there are three users in a database, Fernando, Carlos, and Michelle. Fernando owns a
table called Countries, and Carlos owns a table called Cities. This scenario appears in Figure 8.4.

Figure 8.4. Using ownership chains.

307
Microsoft SQL Server 2000 Programming by Example

Carlos grants SELECT permissions on the Cities table to Fernando. Then, Fernando creates a stored
procedure called citiesandcountries that accesses these two tables (Cities and Countries). After
creating the stored procedure, Fernando grants EXECUTE permissions to Michelle on the stored procedure,
and when Michelle executes it, she only gets the result set of the second query in the stored procedure. This
is because Michelle is accessing indirectly Carlos'table, and Carlos hasn't granted permissions on his table to
Michelle.
In this case, the ownership chain is broken because a stored procedure is accessing a table that has a
different owner than the stored procedure. In particular, SQL Server must check permissions on the Cities
table because this table's owner is not the same owner of the citiesandcountries stored procedure.
To solve this problem, Carlos would have to grant SELECT permissions to Michelle on his table. Notice that
Fernando doesn't have to grant SELECT permissions on his table to Michelle, because he's also the owner of
the stored procedure, which he already granted EXECUTE permissions to Michelle on, thus implicitly allowing
Michelle to access any object in the stored procedure owned by Fernando.
To summarize, if all objects referenced by a stored procedure belong to the stored procedure's owner, and
there aren't dynamic queries inside the stored procedure's definition, any user with just execute permissions
can successfully execute the stored procedure.

What's Next?

In this chapter, you learned the concepts that enable you to create and maintain stored procedures as efficient
and safe ways to access data. In the next chapter, we will go over a special kind of stored procedures,
triggers, which are executed automatically by SQL Server when modification operations take place.

308
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

Chapter 9. Implementing Complex Processing Logic:

Programming Triggers
You can include programmatic capabilities into your database using stored procedures, as described in
Chapter 8, "Implementing Business Logic: Programming Stored Procedures." However, tables are
considered as passive objects that accept modifications, and you must rely on your programs to build complex
business requirements.
You can incorporate complex business logic directly into your tables, defining special procedures that react to
specific actions automatically. These special procedures are called triggers. In this way, a table can be a self-
contained object with attributes (columns and properties), business rules (constraints), and application logic
(triggers).
This chapter teaches you the following:

• How to create triggers and how to specify whether they execute before or after the base object is
modified
• How to define a trigger to accomplish complex data modifications through views
• How to compare the before and after state of the object during the trigger execution, and detect which
columns have been modified
• How to write a trigger that detects whether the action affected only one row or multiple rows, and
reacts accordingly
• How to control nested and recursive execution of triggers
• How and when you can disable trigger execution
• How to select between triggers and constraints to enforce business rules

Benefits of Triggers

In Chapter 7, "Enforcing Data Integrity," you learned how to enforce business rules by defining
constraints. In some real case scenarios, you can find problems that cannot be solved by using constraints.
However, using the programmatic capabilities of triggers, you can create complex rules with endless
possibilities.
A trigger is a stored procedure that is called automatically whenever you execute the action to which the
trigger is defined. You cannot call a trigger directly, but you can execute the action that fires the trigger.
A trigger does not accept parameters and cannot call the RETURN statement to return a value, but it can return
results, as any other stored procedure, although it is not recommended. A trigger executes in the background
and it shouldn't return anything other than error messages, if required.
You can define several triggers for every INSERT, UPDATE, or DELETE action, or any combination of them.
To define a trigger to fire automatically whenever you insert or update data in the Products table, you can
use the simplified syntax from Listing 9.1. Later in this chapter, you will study the CREATE TRIGGER syntax
in detail.

Listing 9.1 Use the CREATE TRIGGER Statement to Define Triggers

CREATE TRIGGER TriggerName ON Products

AFTER INSERT, UPDATE
AS

-- Here you write your program logic

Caution

309
Microsoft SQL Server 2000 Programming by Example

You cannot define a trigger on a SELECT action because this action does not modify data.

When you have a trigger defined in a table, it does not matter how you modify the table. The trigger is always
fired whether you write DML statements directly, execute stored procedures, or use a database library from a
client application.

Tip

A trigger is defined in a table or view inside a database, but the code inside the trigger can refer to
objects in other databases, providing a unique way to implement referential integrity between
objects in different databases.

Using Triggers to Enforce Complex Domain Integrity

Think about an Employees table where you register employees'data. The field DriverNo stores the driver
license number, and because not every employee drives, this field must accept NULL. However, you want to
enforce uniqueness in this field because two employees cannot have the same driver license number. Let's
consider different ways to enforce this rule:

• You cannot create a UNIQUE constraint because you can have more than one employee without a
driver's license and, as you remember, UNIQUE constraints can accept only one NULL value.
• You cannot create a CHECK constraint to see whether you already have the same value in other rows
in the same table, because a CHECK constraint cannot access other rows in the same table.
• You cannot create a RULE object because these objects check for values in the selected field and
affected row only.
• You can create a stored procedure to insert data into the table, and inside the procedure you can
check for uniqueness of this value. In this case, you need two stored procedures, one to insert data
and another one to modify data. Because this business rule is enforced only through these two stored
procedures, you must make sure that your client applications insert and update employees'data only
through these stored procedures. System administrators can modify the base tables directly and
break this business rule.
• The solution could be a trigger that checks for uniqueness of the newly inserted value and either
accepts or rejects the modification. This trigger could be implemented as detailed in Listing 9.2

Listing 9.2 You Can Create a Trigger to Enforce Complex Uniqueness

USE Northwind
GO

310
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

-- Add the DriverNo fields to the Employees table

ALTER TABLE Employees

ADD DriverNo varchar(15) NULL
GO

-- Create the trigger to check

-- for uniqueness of the DriverNo field

CREATE TRIGGER isrEmployees

ON Employees
FOR INSERT, UPDATE
AS

IF UPDATE (DriverNo)
IF EXISTS (

SELECT DriverNo, COUNT(*)

FROM Employees
WHERE DriverNo IS NOT NULL
AND DriverNo IN
(SELECT DISTINCT DriverNo
FROM Inserted)
GROUP BY DriverNo
HAVING COUNT(*) > 1)

BEGIN

RAISERROR ('Driver license number not unique, INSERT aborted', 16, 1)

ROLLBACK TRA6N

END

-- This statement succeeds

UPDATE Employees
SET DriverNo = '74914173'
WHERE EmployeeID = 5

-- This statement fails

-- the driverno is not unique

UPDATE Employees
SET DriverNo = '74914173'
WHERE EmployeeID = 6

-- This statement fails

-- Trying to insert multiple repeated
-- DriverNo values

UPDATE Employees
SET DriverNo = '74914175'
WHERE EmployeeID BETWEEN 6 AND 10

-- reset to NULL succeeds because

-- NULL values are not considered
-- in this trigger

UPDATE Employees
SET DriverNo = NULL

311
Microsoft SQL Server 2000 Programming by Example

Server:Msg 50000, Level 16, State 1, Procedure isrEmployees, Line 16

Driver license number not unique, INSERT aborted
Triggers do not represent much overhead for SQL Server because their execution time depends mostly on
data access to other tables. However, it is not advisable to create complex logic inside triggers.
Triggers always run inside the transaction context of the action that fires them.
Caution

Be careful when using ROLLBACK TRAN inside a trigger, because it cancels the execution of the
batch that fired the trigger. The complete transaction is rolled back, including any other actions
performed during the transaction scope

Using Triggers to Maintain Denormalized Data

Let's consider a different example. You want to know the total sales for product categories. You want to know
this total for any given category at any time, and the results must reflect the actual data. You have several
different ways to do this.

Use Aggregrate Functions

Execute the query from Listing 9.3 every time you want to obtain this information, modifying the value for
CategoryName. This is not very efficient, because you must remember the complete syntax or save the
query in a script or template. Every time you execute this query, SQL Server must parse, compile, and
execute the query with the overhead associated to aggregate functions. If you rarely need to call this script,
this could be an acceptable solution.

Listing 9.3 Use Aggregate Functions to Get Summary Information

USE Northwind
GO

SELECT P.CategoryID, C.CategoryName,

SUM(OD.UnitPrice * Quantity * (1 - Discount)) as Total
FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C

312
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

ON C.CategoryID = P.CategoryID
WHERE CategoryName = 'Confections'
GROUP BY P.CategoryID, C.categoryName

CategoryID CategoryName Total

----------- --------------- ------------------------------
3 Confections 167357.22483158112

Create a View

Create the TotalSalesByCategory view, as in Listing 9.4, to produce this result and select from this view
filtering to the required CategoryName. This solution is flexible and provides good security control, but it
does not provide any improvements in speed because the view definition must be merged with the outer
query, and the final resulting query plan will be executed every time you use this view, with the overhead
inherent to the use of aggregate functions. This could be an adequate solution if the view is not going to be
executed frequently.

Listing 9.4 Create a View to Retrieve Summary Information.

USE Northwind
GO

-- Create the view

CREATE VIEW TotalSalesByCategory

AS
SELECT P.CategoryID, C.CategoryName,
SUM(OD.UnitPrice * Quantity * (1 - Discount)) as Total
FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.CategoryID = P.CategoryID
GROUP BY P.CategoryID, C.categoryName
GO

-- Use the view to search

-- for 'Confections'totals

SELECT *
FROM TotalSalesByCategory
WHERE CategoryName = 'Confections'

313
Microsoft SQL Server 2000 Programming by Example

Create a Stored Procedure

Create the stored procedure GetSalesByCategory, as in Listing 9.5, to produce the required results for a
given category. This solution can be efficient and faster than the previous ones, due to the reuse of the query
plan. If this procedure is not frequently used, compared to the insert and update operations on the underlying
tables, this solution is sufficiently efficient.

Listing 9.5 Create a Stored Procedure to Retrieve Summary Information

-- Create the procedure

CREATE PROCEDURE GetSalesByCategory

@CatName nvarchar(15)
AS
SELECT P.CategoryID, C.CategoryName,
SUM(OD.UnitPrice * Quantity * (1 - Discount)) as Total
FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.CategoryID = P.CategoryID
WHERE CategoryName = @CatName
GROUP BY P.CategoryID, C.categoryName
GO

-- Use the procedure to search

-- for 'Confections'totals

EXEC GetSalesByCategory 'Confections'

Note

The query plan of the last three examples is exactly the same, so it is the execution performance.
The latest method, using a stored procedure, is more efficient for repeated execution because it
avoids the need for parsing, optimizing, and compiling the query. The query plan can be reused
more easily than direct queries, and it produces lower network traffic .

Use Triggers

The fastest way to retrieve this information is to have the precomputed total in a new table
TotalCategoriesSales, and create triggers, as in Listing 9.6, to maintain this information automatically,
whenever you change data in the Order Details table. If your more frequent query is to get these totals, you
should provide this information as fast as possible. In this case, retrieving these values is highly efficient, but
modifying data is less efficient because triggers must run after every data modification.

Listing 9.6 Use Triggers to Maintain Summary Data

314
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

USE Northwind
GO

-- Create denormalized table

CREATE TABLE TotalCategoriesSales

(CategoryID int NOT NULL PRIMARY KEY,
CategoryName nvarchar(15) NOT NULL,
TotalSales money DEFAULT 0)

-- Synchronization Procedure
-- Including initial data population

CREATE PROCEDURE SyncTotalCategoriesSales

AS
TRUNCATE TABLE TotalcategoriesSales

INSERT TotalCategoriesSales
SELECT CategoryID, CategoryName, 0
FROM Categories

UPDATE TC
SET TotalSales =
(SELECT SUM(OD.UnitPrice * Quantity * (1 - Discount))
FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
WHERE P.CategoryID = TC.categoryID)
FROM TotalCategoriesSales TC
GO

-- Synchronize totals
-- We can run this procedure if necessary

EXEC SyncTotalcategoriesSales

GO

CREATE TRIGGER modOrderDetails

ON [Order Details]
AFTER INSERT, UPDATE, DELETE
AS

UPDATE TC
SET TotalSales = TotalSales
- D.UnitPrice * Quantity * (1 - Discount)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Deleted D
ON D.ProductID = P.productID
UPDATE TC

315
Microsoft SQL Server 2000 Programming by Example

SET TotalSales = TotalSales

+ I.UnitPrice * Quantity * (1 - Discount)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Inserted I
ON I.ProductID = P.productID
GO

-- Select initial values

PRINT CHAR(10) + 'Initial summary values'+ CHAR(10)

SELECT *
FROM TotalCategoriesSales

-- Insert a new Order

INSERT Orders (CustomerID)

SELECT 'ALFKI'

DECLARE @id int

SET @id = @@IDENTITY

-- Show the @@identity for later reference

SELECT @id AS OrderID

-- Sell 10 units of Ikura (cat 8) at $30.00 and 10% discount

PRINT CHAR(10) + 'Insert Ikura (cat 8) Order'+ CHAR(10)

INSERT [Order Details]

(orderID, ProductID, UnitPrice, Quantity, Discount)
VALUES (@id, 10, 30, 10, 0.1)

-- Sell 20 units of Tofu (cat 7) at $20.00 and 20% discount

PRINT CHAR(10) + 'Insert Tofu (cat 7) Order'+ CHAR(10)

INSERT [Order Details]

(orderID, ProductID, UnitPrice, Quantity, Discount)
VALUES (@id, 14, 20, 100, 0.2)
-- Sell 5 units of Queso Cabrales (cat 4) at $20.00 and no discount

PRINT CHAR(10) + 'Insert Queso Cabrales (cat 4) Order'+ CHAR(10)

INSERT [Order Details]

(orderID, ProductID, UnitPrice, Quantity, Discount)
VALUES (@id, 11, 20, 5, 0.0)

-- Test the new totals

SELECT *
FROM TotalCategoriesSales

-- Update quantity of the Queso Cabrales to 100

PRINT CHAR(10) + 'Increase Queso Cabrales (cat 4) to 100'+ CHAR(10)

316
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

UPDATE [Order Details]

SET Quantity = 100
WHERE OrderID = @id
AND ProductID = 11

-- Test the new totals

SELECT *
FROM TotalCategoriesSales

-- Remove Tofu from this order

PRINT CHAR(10) + 'Remove Tofu (cat 7) Order'+ CHAR(10)

DELETE [Order Details]

WHERE OrderID = @id
AND ProductID = 14

-- Test the new totals

SELECT *
FROM TotalCategoriesSales

-- remove the order completely

PRINT CHAR(10) + 'Remove Ikura (cat 8) Order'+ CHAR(10)

DELETE [Order Details]

WHERE OrderID = @id
AND ProductID = 10
PRINT CHAR(10) + 'Remove Queso Cabrales (cat 4) Order'+ CHAR(10)

DELETE [Order Details]

WHERE OrderID = @id
AND ProductID = 11

DELETE Orders
WHERE OrderID = @id

-- Test the new totals

SELECT *
FROM TotalCategoriesSales
Initial summary values

CategoryID CategoryName TotalSales

----------- --------------- ---------------------
1 Beverages 267868.1805
2 Condiments 106047.0850
3 Confections 167357.2248

317
Microsoft SQL Server 2000 Programming by Example

4 Dairy Products 234507.2813

5 Grains/Cereals 95744.5875
6 Meat/Poultry 163022.3591
7 Produce 99984.5781
8 Seafood 131261.7344

OrderID
-----------
11083

Insert Ikura (cat 8) Order

Insert Tofu (cat 7) Order

Insert Queso Cabrales (cat 4) Order

CategoryID CategoryName TotalSales

----------- --------------- ---------------------
1 Beverages 267868.1805
2 Condiments 106047.0850
3 Confections 167357.2248
4 Dairy Products 234607.2813
5 Grains/Cereals 95744.5875
6 Meat/Poultry 163022.3591
7 Produce 101584.5781
8 Seafood 131531.7344

Increase Queso Cabrales (cat 4) to 100

CategoryID CategoryName TotalSales

----------- --------------- ---------------------
1 Beverages 267868.1805
2 Condiments 106047.0850
3 Confections 167357.2248
4 Dairy Products 236507.2813
5 Grains/Cereals 95744.5875
6 Meat/Poultry 163022.3591
7 Produce 101584.5781
8 Seafood 131531.7344

Remove Tofu (cat 7) Order

CategoryID CategoryName TotalSales

----------- --------------- ---------------------
1 Beverages 267868.1805
2 Condiments 106047.0850
3 Confections 167357.2248
4 Dairy Products 236507.2813
5 Grains/Cereals 95744.5875
6 Meat/Poultry 163022.3591
7 Produce 99984.5781
8 Seafood 131531.7344

Remove Ikura (cat 8) Order

318
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

Remove Queso Cabrales (cat 4) Order

CategoryID CategoryName TotalSales

----------- --------------- ---------------------
1 Beverages 267868.1805
2 Condiments 106047.0850
3 Confections 167357.2248
4 Dairy Products 234507.2813
5 Grains/Cereals 95744.5875
6 Meat/Poultry 163022.3591
7 Produce 99984.5781
8 Seafood 131261.7344

Caution

The example from Listing 9.6 only works with single-row operations. Later in this chapter you will
see how to deal with multirow operation using the same example.

Note

To complete the example, you should create triggers in the Products and Categories tables as
well, because a product could be moved to a new category, and a category can change its name.

You could drop the CategoryName field from the TotalCategoriesSales table, but in that
case you have to join to the Categories table to filter the category by name. After you
denormalized your database to maintain summary values, why not denormalize one step further?
In this way, you need to access only one table to produce the required results.

Use Indexed Views

A different possibility would be to create an index on the SalesByCategory view, as in Listing 9.7. In this
case, no triggers are necessary to maintain this information, and the Query Optimizer can decide to use this
view to solve queries requesting data compatible with the data generated by this view, even if the queries do
not reference the view. This is the optimal solution for summary data retrieval, but it is not as efficient as the
solution based on triggers— if you expect many data updates.

Listing 9.7 Use Indexed Views to Retrieve Precomputed Summary Information

-- DROP the existing view

319
Microsoft SQL Server 2000 Programming by Example

IF OBJECT_ID('TotalSalesByCategory') IS NOT NULL

DROP VIEW TotalSalesByCategory
GO

-- Setting required to create an indexed view

SET ANSI_NULLS ON
SET ANSI_PADDING ON
SET ANSI_WARNINGS ON
SET ARITHABORT ON
SET CONCAT_NULL_YIELDS_NULL ON
SET QUOTED_IDENTIFIER ON
SET NUMERIC_ROUNDABORT OFF
GO

-- Create the view

-- With Schemabinding
-- and use two part names for objects
CREATE VIEW TotalSalesByCategory
WITH SCHEMABINDING
AS
SELECT P.CategoryID, C.CategoryName,
SUM(OD.UnitPrice * Quantity * (1 - Discount)) as Total,
COUNT_BIG(*) as CB
FROM dbo.[Order Details] OD
JOIN dbo.Products P
ON P.ProductID = OD.ProductID
JOIN dbo.Categories C
ON C.CategoryID = P.CategoryID
GROUP BY P.CategoryID, C.categoryName
GO

-- Create the index on the view

CREATE UNIQUE CLUSTERED INDEX ndx_CatTotals

ON TotalSalesByCategory (CategoryName)

-- Use the view to search

-- for 'Confections'totals
-- set SHOWPLAN_TEXT
-- to show that the view is used
-- instead of the base tables

SET SHOWPLAN_TEXT ON

SELECT CategoryID, CategoryName, Total

FROM TotalSalesByCategory (NOEXPAND)
WHERE CategoryName = 'Confections'

StmtText
---------
|--Clustered Index Scan(

320
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

OBJECT:([Northwind].[dbo].[TotalSalesByCategory].[ndx_CatTotals]),
WHERE:([TotalSalesByCategory].[CategoryName]='Confections'))

Trigger Enhancements in SQL Server 2000

SQL Server 2000 includes exciting new features for triggers. Some of these features are discussed later in
this chapter, but here we want to mention them briefly.
In previous versions, you were forced to use triggers to implement cascading operations. SQL Server 2000
provides cascade declarative referential integrity by FOREIGN KEY constraints with the ON DELETE
CASCADE and ON UPDATE CASCADE options, as explained in Chapter 7. This is the recommended way to
implement cascade operations.
SQL Server 2000 provides a new type of trigger: the INSTEAD OF trigger. These triggers are executed
instead of the firing actions, providing a mechanism to check for data changes before the data is actually
modified. Later in this chapter you will learn about the applications of these new triggers.
It is possible to create an INSTEAD OF trigger on a view, providing new ways to deal with data modifications
through views. Later in this chapter you will find some examples of this exciting feature.
As in version 7.0, you can create multiple triggers per action, but now you can specify which trigger will be the
first one to fire, providing some kind of initialization process for other triggers. You also can specify the last
trigger to execute, providing a cleansing mechanism to finish the triggering action.
The list of restricted statements to be executed inside triggers has been reduced from previous versions. A
very useful new possibility is the capability to create tables inside the trigger— mainly temporary tables— using
DML statements.
Note

In SQL Server 7.0, you could create a table inside a trigger only by using the SELECT INTO
statement.

Inside INSTEAD OF triggers, you can inspect the previous and after state of BLOB columns, because they
appear in the Inserted and Deleted tables.

Checking for changes in specific columns is now easier using the new COLUMNS_UPDATED function.

Inserted and Deleted Tables

Inside the trigger, you can check the previous and after column values for the affected rows by reading two
virtual tables: Inserted and Deleted. The Inserted and Deleted tables are read-only in-memory virtual
tables with the same columns as the base table.
The Inserted table contains the new inserted rows in a trigger defined for an INSERT statement or the new
values for modified rows in a trigger defined for an UPDATE statement. In a trigger defined for a DELETE
statement, the Inserted table exists, but it is empty.
The Deleted table contains the deleted rows in a trigger defined for a DELETE statement or the old values for
modified rows in a trigger defined for an UPDATE statement. In a trigger defined for an INSERT statement, the
Deleted table exists, but it is empty.
If the original action modified several rows at once, the trigger is fired just once and the Inserted and
Deleted tables are multirow tables containing the modified rows.
Inside a trigger, it is possible to modify the base table again. In this case, it is possible to join the base table to
the Inserted table to remodify only affected rows. Listing 9.8 shows an example of this possibility.

Listing 9.8 Inside a Trigger, You Can Remodify the Affected Rows

321
Microsoft SQL Server 2000 Programming by Example

-- Create a new table

CREATE TABLE NewCategories (

CategoryID int NOT NULL
IDENTITY(1,1) PRIMARY KEY,
CategoryName nvarchar(15) NOT NULL ,
Modified smalldatetime
DEFAULT CURRENT_TIMESTAMP,
Modifier nchar(128)
DEFAULT SYSTEM_USER)
GO

-- Populate the table with data

INSERT NewCategories
(CategoryName)
SELECT CategoryName
FROM Categories
GO

-- Create an AFTER UPDATE trigger

CREATE TRIGGER udt_NewCategories

ON NewCategories
AFTER UPDATE
AS
UPDATE NewCategories
SET Modified = CURRENT_TIMESTAMP,
Modifier = SYSTEM_USER
FROM NewCategories
JOIN Inserted
ON Inserted.CategoryID = NewCategories.CategoryID
GO

-- Show actual values

SELECT *
FROM NewCategories
WHERE CategoryID = 3
GO

PRINT CHAR(10) + 'Wait for 30 seconds'+ CHAR(10)

GO

WAITFOR DELAY '00:00:30'

-- Make sure Guest has permissions to modify

-- and read the NewCategories table

GRANT ALL ON NewCategories TO Guest

GO

-- Impersonate the user Guest

322
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

SETUSER 'Guest'
GO

-- Modify Category 3 Name

UPDATE NewCategories
SET CategoryName = 'Confectionery'
WHERE CategoryID = 3

-- Show new values

-- Where the trigger modified automatically
-- the Modified and Modifier columns

SELECT *
FROM NewCategories
WHERE CategoryID = 3
GO

-- Impersonate the previous user (dbo)

SETUSER
GO

-- Drop the demo table

-- it drops automatically the trigger

DROP TABLE NewCategories

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ -----------
3 Confections 2000-11-19 13:18:00 sa

Wait for 30 seconds

CategoryID CategoryName Modified Modifier
----------- --------------- ------------------------------ ------------
3 Confectionery 2000-11-19 13:19:00 Guest
In INSERT and UPDATE operations, the Inserted table contains the same information as the base table for
the affected rows.
In UPDATE operations, the Deleted table contains one row per every updated row in the table, but the
information is not common to the base table because the Deleted table contains the previous values and the
base table already contains the new values of the data.
In DELETE operations, the Deleted table does not contain any common rows with the base table, because
the rows already have been removed from the base table.
Listing 9.9 shows an example of the contents of the Inserted and Deleted tables in different situations.

Listing 9.9 You Can Use the Inserted and Deleted Tables Inside Any Kind of Trigger

323
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

-- Create a new table

CREATE TABLE NewCategories (

CategoryID int NOT NULL
IDENTITY(1,1) PRIMARY KEY,
CategoryName nvarchar(15) NOT NULL ,
Modified smalldatetime
DEFAULT CURRENT_TIMESTAMP,
Modifier nchar(128)
DEFAULT SYSTEM_USER)
GO

-- Create an AFTER UPDATE trigger

CREATE TRIGGER tr_NewCategories

ON NewCategories
AFTER UPDATE, INSERT, DELETE
AS

DECLARE @s varchar(80)
DECLARE @ni int, @nd int

SELECT @ni = COUNT(*)

FROM Inserted
SELECT @nd = COUNT(*)
FROM Deleted

SET @s = CASE
WHEN (@ni = @nd) AND @ni > 0
THEN 'AFTER UPDATE Trigger (compare Deleted and Inserted tables)'
WHEN @ni > @nd
THEN 'AFTER INSERT Trigger (look at the Inserted table)'
WHEN @ni < @nd
THEN 'AFTER DELETE Trigger (look at the Deleted table)'
ELSE 'TRIGGER fired by Null action (Inserted and Deleted are empty)'END

PRINT CHAR(10) + '#######################################################'

PRINT @s
PRINT '#######################################################'+ CHAR(10)

PRINT CHAR(10) + 'Inserted Table'+ CHAR(10)

SELECT *
FROM Inserted

PRINT CHAR(10) + 'Deleted Table'+ CHAR(10)

SELECT *
FROM Deleted
GO

324
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

-- Insert Operation

INSERT NewCategories (CategoryName)

VALUES ('Food')

-- Multiple Insert Operation

INSERT NewCategories (CategoryName)

SELECT CategoryName
FROM Categories

-- Delete one row

DELETE NewCategories
WHERE CategoryID = 7

-- Delete multiple rows

DELETE NewCategories
WHERE CategoryID BETWEEN 1 AND 4
-- Update one row

UPDATE NewCategories
SET CategoryName = 'Products'
WHERE CategoryID = 8

-- Update multiple rows

UPDATE NewCategories
SET CategoryName = UPPER(CategoryName)

-- Null operation because it does not affect any row

UPDATE NewCategories
SET CategoryName = UPPER(CategoryName)
WHERE CategoryID > 1000

-- Drop the demo table

-- it drops automatically the trigger

DROP TABLE NewCategories

###############################################################
AFTER INSERT Trigger (look at the Inserted table)
###############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

325
Microsoft SQL Server 2000 Programming by Example

1 Food 2000-11-19 13:43:00 sa

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

###############################################################
AFTER INSERT Trigger (look at the Inserted table)
###############################################################

Inserted Table
CategoryID CategoryName Modified Modifier
----------- --------------- ------------------------------ ------------
2 Beverages 2000-11-19 13:43:00 sa
3 Condiments 2000-11-19 13:43:00 sa
4 Confections 2000-11-19 13:43:00 sa
5 Dairy Products 2000-11-19 13:43:00 sa
6 Grains/Cereals 2000-11-19 13:43:00 sa
7 Meat/Poultry 2000-11-19 13:43:00 sa
8 Produce 2000-11-19 13:43:00 sa
9 Seafood 2000-11-19 13:43:00 sa

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

###############################################################
AFTER DELETE Trigger (look at the Deleted table)
###############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
7 Meat/Poultry 2000-11-19 13:43:00 sa

###############################################################
AFTER DELETE Trigger (look at the Deleted table)
###############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

Deleted Table

326
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
1 Food 2000-11-19 13:43:00 sa
2 Beverages 2000-11-19 13:43:00 sa
3 Condiments 2000-11-19 13:43:00 sa
4 Confections 2000-11-19 13:43:00 sa

###############################################################

AFTER UPDATE Trigger (compare Deleted and Inserted tables)

##############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
8 Products 2000-11-19 13:43:00 sa

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
8 Produce 2000-11-19 13:43:00 sa

###############################################################
AFTER UPDATE Trigger (compare Deleted and Inserted tables)
###############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
5 DAIRY PRODUCTS 2000-11-19 13:43:00 sa
6 GRAINS/CEREALS 2000-11-19 13:43:00 sa
8 PRODUCTS 2000-11-19 13:43:00 sa
9 SEAFOOD 2000-11-19 13:43:00 sa

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------
5 Dairy Products 2000-11-19 13:43:00 sa
6 Grains/Cereals 2000-11-19 13:43:00 sa
8 Products 2000-11-19 13:43:00 sa
9 Seafood 2000-11-19 13:43:00 sa

###############################################################
TRIGGER fired by Null action (Inserted and Deleted are empty)
###############################################################

Inserted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

327
Microsoft SQL Server 2000 Programming by Example

Deleted Table

CategoryID CategoryName Modified Modifier

----------- --------------- ------------------------------ ------------

Types of Triggers According to Their Order

SQL Server 2000 defines two different types of triggers:

• INSTEAD OF triggers that fire instead of the original action, before the base object is modified
• AFTER triggers that fire after the base table is modified

A table can have any number of AFTER triggers for every action, but only one INSTEAD OF trigger per action.
A view can have only INSTEAD OF triggers.

INSTEAD OFTriggers

INSTEAD OF triggers are fired automatically when you execute a statement by which the trigger is defined. In
this case, the trigger fires before the action is executed. However, inside an INSTEAD OF trigger, you have
access to the Inserted and Deleted virtual tables.
When the INSTEAD OF trigger starts its execution, the base table has not been modified yet. However, on
INSERT actions the Inserted table contains the default values for columns with DEFAULT definitions, and it
does not contain the values for the IDENTITY property.
Because the base table is not modified yet, constraints have not been checked yet. If inside the INSTEAD OF
trigger you don't send any action to the base table, the base table is not modified at all.
You can use INSTEAD OF triggers to check the data before the actual modification, in a more complex way
than with CHECK constraints. Depending on the results of your conditions, you can decide to apply the
requested changes or decide not to apply them at all.
If you want to modify data in a table, you can use the INSERT, UPDATE, and DELETE statements. For each
table or view, you can define only one INSTEAD OF trigger per each triggering action.

• An INSTEAD OF INSERT trigger allows you to cancel, partially or completely, an insert operation
without rolling back any transaction.
• An INSTEAD OF UPDATE trigger offers you the possibility of checking changes before they are
applied to the data. You can decide which changes to finally apply.
• An INSTEAD OF DELETE trigger allows you to cancel the deletion of specific rows, without rolling
back transactions.

You cannot create an INSTEAD OF trigger in a table where you have defined a FOREIGN KEY with CASCADE
operations.
Using SQL Server 2000, you can create INSTEAD OF triggers on views. In this way you can provide better
updating functionality to your views, as in the following examples:

• In a view that shows a FullName field defined as LastName + ', '+ FirstName, you cannot
execute an UPDATE statement to modify the FullName field directly, because it is a read-only
computed column. You can create an INSTEAD OF UPDATE trigger to allow for modifications in the
FullName field, resulting in changes in the FirstName and LastName fields. Listing 9.10 shows
an example of this case.

Listing 9.10 Use INSTEAD OF Triggers to Update Computed Columns.

328
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

USE Northwind
GO

-- Create a view based in the Employees table

CREATE VIEW EmployeesName

AS
SELECT ISNULL(EmployeeID, 0) AS EmployeeID,
LastName + ', '+ FirstName AS FullName
FROM Employees
GO

-- Show information related to the Employee 6

SELECT FullName
FROM EmployeesName
WHERE EmployeeID = 6

SELECT FirstName, LastName

FROM Employees
WHERE EmployeeID = 6
GO

-- Create INSTEAD OF INSERT trigger

CREATE TRIGGER isrEmployeesName

ON EmployeesName
INSTEAD OF INSERT
AS
INSERT Employees (LastName, FirstName)
SELECT LEFT(FullName, CHARINDEX(',', FullName) - 1),
LTRIM(RIGHT(FullName, LEN(FullName)
- CHARINDEX(',', FullName)))
FROM Inserted
GO

-- Create INSTEAD OF UPDATE trigger

CREATE TRIGGER udtEmployeesName

ON EmployeesName
INSTEAD OF UPDATE
AS
UPDATE Employees
SET LastName = LEFT(Inserted.FullName,
CHARINDEX(',', Inserted.FullName) - 1),
FirstName = LTRIM(RIGHT(Inserted.FullName,
LEN(Inserted.FullName)
- CHARINDEX(',', Inserted.FullName) ))
FROM Employees
JOIN Inserted
ON Inserted.EmployeeID = Employees.EmployeeID
WHERE Employees.EmployeeID = Inserted.EmployeeID
GO

329
Microsoft SQL Server 2000 Programming by Example

-- testing INSTEAD OF UPDATE Trigger

UPDATE EmployeesName
SET FullName = 'NewFamily, Michael'
WHERE EmployeeID = 6
GO
SELECT FullName
FROM EmployeesName
WHERE EmployeeID = 6

SELECT FirstName, LastName

FROM Employees
WHERE EmployeeID = 6
GO

-- testing INSTEAD OF INSERT Trigger

INSERT EmployeesName (EmployeeID, FullName)

VALUES (0, 'NewFamily, NewGuy')
GO

SELECT FullName
FROM EmployeesName
WHERE FullName LIKE 'New%'

SELECT FirstName, LastName

FROM Employees
WHERE LastName LIKE 'New%'
GO

-- deleting from this view

-- does not require an INSTEAD OF DELETE Trigger

DELETE EmployeesName
WHERE FullName = 'NewFamily, NewGuy'
GO

SELECT FullName
FROM EmployeesName
WHERE FullName LIKE 'New%'

SELECT FirstName, LastName

FROM Employees
WHERE LastName LIKE 'New%'
GO

DROP VIEW EmployeesName

GO
FullName
--------------------------------
Suyama, Michael

330
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

FirstName LastName
---------- --------------------
Michael Suyama

FullName
--------------------------------
NewFamily, Michael

FirstName LastName
---------- --------------------
Michael NewFamily

FullName
--------------------------------
NewFamily, Michael
NewFamily, NewGuy

FirstName LastName
---------- --------------------
Michael NewFamily
NewGuy NewFamily

FullName
--------------------------------
NewFamily, Michael

FirstName LastName
---------- --------------------
Michael NewFamily

• You can create a view to obtain the total company budget, based on the individual budget per
department. To increase the total budget, you can modify every department's individual budget.
However, if you had a view to summarize the total budget, it could be useful to adjust individual
budgets proportionally, by changing only the total budget field in the view.

Views based on aggregations are read-only, but you can create an INSTEAD OF UPDATE trigger on
the summary view to accept changes in the total budget, and distribute these changes proportionally
to every individual budget. Listing 9.11, applied to the Categories table, shows how to implement
this solution .

Listing 9.11 Use INSTEAD OF Triggers to Distribute Changes in Summary Data.

USE Northwind

GO

-- Modify the Categories table

331
Microsoft SQL Server 2000 Programming by Example

-- To add a new field to store

-- Budget value

ALTER TABLE Categories

ADD Budget money NULL
GO
-- Add a random budget per category

UPDATE Categories
SET Budget = ROUND(SIN(CategoryID)* 10000 + 20000, -3)
GO

-- Create a view to see the total budget

CREATE VIEW TotalBudget

AS
SELECT SUM(Budget) AS TBudget
FROM Categories
GO

-- Create an INSTEAD OF trigger

-- to UPDATE budgets proportionally

CREATE TRIGGER udtTotalBudget

ON TotalBudget
INSTEAD OF UPDATE
AS

-- Share the new budget proportionally

UPDATE Categories
SET Budget = ROUND(Budget *
(Inserted.TBudget /
Deleted.TBudget), -3)
FROM Inserted, Deleted

-- Adjust the difference to the biggest

budget

UPDATE Categories
SET Budget = Budget +
Inserted.TBudget -
(SELECT SUM(Budget)
FROM Categories)
FROM Inserted
WHERE Budget =
(SELECT MAX(Budget)
FROM categories)

GO

-- Test the view

PRINT CHAR(10) + 'Total Budget before the update'+ CHAR(10)

SELECT *
FROM TotalBudget

PRINT CHAR(10) + 'Individual Budgets before the update'+ CHAR(10)

SELECT CategoryID, Budget

332
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

FROM Categories

-- Increase the budget

UPDATE TotalBudget
SET TBudget = 200000

-- Test the update

PRINT CHAR(10) + 'Total Budget after the update'+ CHAR(10)

SELECT *
FROM TotalBudget

PRINT CHAR(10) + 'Individual Budgets after the update'+ CHAR(10)

SELECT CategoryID, Budget

FROM Categories

-- Remove the view and the Budget column

DROP VIEW TotalBudget

ALTER TABLE Categories

DROP COLUMN Budget
Total Budget before the update stored
procedures;triggers;distributing changes in summary data>distributing;changes
in summary data (triggers)>

TBudget
---------------------
174000.0000

Individual Budgets before the update

CategoryID Budget
----------- ---------------------
1 28000.0000
2 29000.0000
3 21000.0000
4 12000.0000
5 10000.0000
6 17000.0000
7 27000.0000
8 30000.0000

Total Budget after the update

TBudget

333
Microsoft SQL Server 2000 Programming by Example

---------------------
200000.0000

Individual Budgets after the update

CategoryID Budget
----------- ---------------------
1 32000.0000
2 33000.0000
3 24000.0000
4 14000.0000
5 11000.0000
6 20000.0000
7 31000.0000
8 35000.0000

• If a view is defined as a query that joins several tables, you can only execute an UPDATE statement to
modify through this view, if the statement affects only columns in a single table. In a view that joins
two tables, you can create an INSTEAD OF UPDATE trigger to apply changes to columns on both
tables simultaneously, as shown in Listing 9.12 .

Listing 9.12 Use INSTEAD OF Triggers to Update Views Defined on Multiple Tables

USE Northwind

GO

-- Create a new table to store the Budget

CREATE TABLE CategoriesBudget(

CategoryID int NOT NULL
PRIMARY KEY
REFERENCES Categories (CategoryID)
,Budget money NOT NULL
DEFAULT 0)
GO

-- Add a random budget per category

INSERT CategoriesBudget
SELECT CategoryID, ROUND(SIN(CategoryID)* 10000 + 20000, -3)
FROM Categories

GO

-- Create a view to see the name and budget

-- for every Category

CREATE VIEW CategoriesNameBudget

AS

334
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

SELECT C.CategoryID,
C.CategoryName,
CB.Budget
FROM Categories C
JOIN CategoriesBudget CB
ON C.categoryID = CB.categoryID

GO

-- Create an INSTEAD OF trigger

-- to UPDATE budgets proportionally stored
procedures;triggers;types>

CREATE TRIGGER udtCategoriesNameBudget

ON CategoriesNameBudget
INSTEAD OF UPDATE
AS

-- Update the Category Name

UPDATE C
SET C.CategoryName = I.CategoryName
FROM Inserted I
JOIN Categories C
ON C.CategoryID = I.CategoryID

-- Update the Budget

UPDATE CB
SET CB.Budget = I.Budget
FROM Inserted I
JOIN CategoriesBudget CB
ON CB.CategoryID = I.CategoryID

GO

-- Test the view

PRINT CHAR(10) + 'Names and Budgets before the update'+ CHAR(10)

SELECT *
FROM CategoriesNameBudget

-- Change name and budget of Category 7

UPDATE CategoriesNameBudget
SET Budget = 35000.0000,
CategoryName = 'Miscellaneous'
WHERE CategoryID = 7

-- Test the update

PRINT CHAR(10) + 'Names and Budgets after the update'+ CHAR(10)

SELECT *
FROM CategoriesNameBudget

-- Back Categories to normal

UPDATE Categories

335
Microsoft SQL Server 2000 Programming by Example

SET CategoryName = 'Produce'

WHERE CategoryID = 7

DROP VIEW CategoriesNameBudget

DROP TABLE CategoriesName

Names and Budgets before the update

CategoryID CategoryName Budget

----------- --------------- ---------------------
1 Beverages 28000.0000
2 Condiments 29000.0000
3 Confections 21000.0000
4 Dairy Products 12000.0000
5 Grains/Cereals 10000.0000
6 Meat/Poultry 17000.0000
7 Produce 27000.0000
8 Seafood 30000.0000

Names and Budgets after the update

CategoryID CategoryName Budget

----------- --------------- ---------------------
1 Beverages 28000.0000
2 Condiments 29000.0000
3 Confections 21000.0000
4 Dairy Products 12000.0000
5 Grains/Cereals 10000.0000
6 Meat/Poultry 17000.0000
7 Miscellaneous 35000.0000
8 Seafood 30000.0000

Caution

You cannot create an INSTEAD OF trigger in a view created using the WITH CHECK option.

After Triggers

AFTER triggers execute after the data has been modified. However, because you are still inside the
transaction, you can decide to cancel the action. Unlike INSTEAD OF triggers, you do not have to do anything
special to commit the transaction; it will be committed automatically unless you decide to roll it back.
You can create any number of AFTER triggers for every action on every table.

Creating and Dropping Triggers

336
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

To create a trigger, you can use the CREATE TRIGGER statement. In this statement, you must provide the
name of the trigger, the object where the trigger is defined, and the action or actions that will fire the trigger.
The CREATE TRIGGER statement must be executed as the only statement in a batch. The same restriction
applies to the CREATE PROCEDURE, CREATE VIEW, CREATE RULE, and CREATE FUNCTION statements.
You can use the WITH ENCRYPTION option to encrypt the trigger definition. In this way, you can prevent user
access to the trigger's internal code. Be careful when using this option, because there is no way to decrypt an
encrypted trigger. It is recommended that you save the CREATE TRIGGER script in a protected directory for
further review and modification.
Caution

If you encrypt a trigger, it cannot be copied using replication.

You can prevent the execution of the trigger when receiving replicated data by using the NOT FOR
REPLICATION option.

To define an AFTER trigger, you use either the FOR or the AFTER keywords. Use the INSTEAD OF keyword to
define an INSTEAD OF trigger.

Caution

Because SQL Server 2000 accepts AFTER triggers and INSTEAD OF triggers, you should start
using the AFTER keyword to create triggers that execute after the data is modified. The FOR
keyword is still valid, as a synonym of the AFTER keyword, but it is provided only for backward
compatibility.

Listing 9.13 shows some simplified examples of the CREATE TRIGGER statement.

Listing 9.13 Use the CREATE TRIGGER Statement to Create Triggers.

-- Create a trigger on the customer table

-- to track INSERT, UPDATE and DELETE actions

CREATE TRIGGER tr_Customers

ON Customers
AFTER INSERT, UPDATE, DELETE
AS
-- Your code here

337
Microsoft SQL Server 2000 Programming by Example

GO

-- Create a trigger on the customer table

-- to track INSERT, UPDATE and DELETE actions
-- Using the FOR keyword

CREATE TRIGGER tr_Customers

ON Customers
FOR INSERT, UPDATE, DELETE
AS
-- Your code here
GO
-- Create a trigger on the customer table
-- to track INSERT actions

CREATE TRIGGER isr_Customers

ON Customers
AFTER INSERT
AS
-- Your code here
GO

-- Create a trigger on the customer table

-- to track INSERT, UPDATE and DELETE actions
-- With encryption stored
procedures;triggers;dropping>

CREATE TRIGGER tr_Customers

ON Customers
WITH ENCRYPTION
AFTER INSERT, UPDATE, DELETE
AS
-- Your code here
GO

-- Create a trigger on the customer table

-- to track INSERT, UPDATE and DELETE actions
-- Preventing its execution when receiving
-- replicate data

CREATE TRIGGER tr_Customers

ON Customers
AFTER INSERT, UPDATE, DELETE
NOT FOR REPLICATION
AS
-- Your code here
GO

-- Create an INSTEAD OF trigger

-- on the NewCustomers view
-- to execute on attempts to modify data through the view

CREATE TRIGGER iotr_NewCustomers

ON NewCustomers
INSTEAD OF INSERT, UPDATE, DELETE
AS
-- Your code here
GO
-- Create an INSTEAD OF INSERT trigger
-- on the NewCustomers view
-- to execute on attempts to insert data through the view

338
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

CREATE TRIGGER iotr_NewCustomers

ON NewCustomers
INSTEAD OF INSERT
AS
-- Your code here
GO
It is not recommended to execute statements inside a trigger because they should be executed in the
background with minimum user interaction. This user interaction should be limited to error messages and
swarnings, which do not interfere with the results produced by the calling batch.
Tip

Execute the SET NOCOUNT ON statement at the beginning of the trigger code to avoid sending row
counter information for every statement that affects data.

Tip

If you want to use specific environment settings during the execution of a trigger, execute the
necessary SET statements at the beginning of the trigger code.

A trigger executes whenever the defined action is executed, regardless of the number of affected rows. To
avoid the unnecessary execution of any statement inside the trigger when there are no affected rows, use the
@@ROWCOUNT system function to test whether this value is bigger than 0.

When you create a trigger, SQL Server does not check for object names; it waits for execution time to check
existence of the referenced objects. This is called deferred name resolution, and it can give some flexibility to
the creation order of database objects.

Deciding the Order of Execution

You cannot decide on a specific order of execution for triggers defined against the same action, but you can
select which trigger will be the first one to execute and which trigger will be the last one to execute. You can
do this by using the sp_settriggerorder system-stored procedure, as in Listing 9.14.

Listing 9.14 You Can Change the Triggers' Order of Execution.

-- Create multiple triggers on the customer table

-- to track UPDATE actions

339
Microsoft SQL Server 2000 Programming by Example

CREATE TRIGGER tr1_Customers

ON Customers
AFTER UPDATE
AS
-- Your code here
PRINT 'This is the tr1 trigger'
GO

CREATE TRIGGER tr2_Customers

ON Customers
AFTER UPDATE
AS
-- Your code here
PRINT 'This is the tr2 trigger'
GO

CREATE TRIGGER tr3_Customers

ON Customers
AFTER UPDATE
AS
-- Your code here
PRINT 'This is the tr3 trigger'
GO

-- Test the order of execution

-- By using a MOCK operation

UPDATE Customers
SET ContactName = ContactName
GO

-- Specify the tr3 trigger as first trigger to execute

EXEC sp_settriggerorder 'tr3_Customers', 'FIRST', 'UPDATE'

-- Specify the tr2 trigger as last trigger to

execute

EXEC sp_settriggerorder 'tr2_Customers', 'LAST', 'UPDATE'

-- Specify the tr1 trigger as any order to execute

EXEC sp_settriggerorder 'tr1_Customers', 'NONE', 'UPDATE'

GO
-- Test the order of execution
-- By using a MOCK operation

PRINT CHAR(10) + 'After reordering'+ CHAR(10)

UPDATE Customers
SET ContactName = ContactName
Go

340
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

This is the tr1 trigger

This is the tr2 trigger
This is the tr3 trigger

After reordering

This is the tr3 trigger

This is the tr1 trigger
This is the tr2 trigger

Caution

Remember that INSTEAD OF triggers are always executed before the data is modified. Therefore,
they execute before any of the AFTER triggers.

Checking for Updates on Specific Columns

To check inside a trigger if a column has been updated, you can use the IF UPDATE() clause. This clause
evaluates to TRUE if the column has been updated.
To test for changes in multiple columns in a single statement, use the COLUMNS_UPDATED() function. This
function returns a bitmap with the update status of every column in the base table. In other words,
COLUMNS_UPDATED returns a sequence of bits, one bit for every column, and the bit is 1 if the column has
been updated or otherwise it is 0.
Listing 9.15 shows an example of these two functions.

Listing 9.15 Inside a Trigger You Can Check Which Columns Have Been Updated

CREATE TRIGGER tr_OrderDetails

ON [Order Details]
AFTER UPDATE
AS
-- Testing for changes to the PRIMARY KEY
IF UPDATE(OrderID)
BEGIN
PRINT 'Changes to the PRIMARY KEY are not allowed'
ROLLBACK TRAN

341
Microsoft SQL Server 2000 Programming by Example

END

-- Testing for changes on the 2nd, 3rd and 5th columns

IF ((COLUMNS_UPDATED() & (2 + 4 + 8)) > 0)
BEGIN
IF ((COLUMNS_UPDATED() & 2) = 2)
PRINT 'ProductID updated'
IF ((COLUMNS_UPDATED() & 4) = 4)
PRINT 'UnitPrice updated'
IF ((COLUMNS_UPDATED() & 8) = 8)
PRINT 'Quantity updated'
END
GO

PRINT CHAR(10) + 'Updating ProductID and UnitPrice'

UPDATE [Order Details]

SET ProductID = ProductID,
UnitPrice = UnitPrice

PRINT CHAR(10) + 'Updating Quantity only'

UPDATE [Order Details]

SET Quantity = Quantity

PRINT CHAR(10) + 'Updating OrderID'

UPDATE [Order Details]

SET OrderID = OrderID

Updating ProductID and UnitPrice

ProductID updated
UnitPrice updated

Updating Quantity only

Quantity updated

Updating OrderID

Changes to the PRIMARY KEY are not allowed

Multip le-Row Considerations

Keep in mind that a trigger can be fired by an action that modifies a single row or multiple rows in a single
statement.
If you define your trigger to work for single rows only, you should reject changes that affect multiple rows. In
this case, you can check whether the system function @@ROWCOUNT returns a value greater than 1.

342
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

You can define your trigger to deal only with multiple-row operations. In this case, you could use aggregate
functions or use cursors. None of these strategies is efficient for single-row operations.
The ideal situation would be to create a trigger with conditional logic to deal with either single-row or multiple-
row operations depending on the value returned by @@ROWCOUNT. Listing 9.16 shows a new version of the
example of Listing 9.6, optimized for both kinds of transactions.

Listing 9.16 You Can Use @@ROWCOUNT to Detect Multiple-Row Operations

-----------------------------
-- Create trigger for insert
-----------------------------

CREATE TRIGGER isrOrderDetails

ON [Order Details]
AFTER INSERT
AS
IF @@ROWCOUNT = 1
BEGIN

-- Single-row operation

UPDATE TC
SET TotalSales = TotalSales
+ I.UnitPrice * Quantity * (1 - Discount)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Inserted I
ON I.ProductID = P.productID

END
ELSE
BEGIN

-- Multi-row operation

UPDATE TC
SET TotalSales = TotalSales
+ (SELECT SUM(I.UnitPrice * Quantity * (1 - Discount))
FROM Inserted I
WHERE I.ProductID = P.productID)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID

END
GO

-----------------------------
-- Create trigger for delete
-----------------------------

343
Microsoft SQL Server 2000 Programming by Example

CREATE TRIGGER delOrderDetails

ON [Order Details]
AFTER DELETE
AS
IF @@ROWCOUNT = 1
BEGIN

-- Single-row operation

UPDATE TC
SET TotalSales = TotalSales
- D.UnitPrice * Quantity * (1 - Discount)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Deleted D
ON D.ProductID = P.productID

END
ELSE
BEGIN

-- Multi-row operation

UPDATE TC
SET TotalSales = TotalSales
- (SELECT SUM(D.UnitPrice * Quantity * (1 - Discount))
FROM Deleted D
WHERE D.ProductID = P.productID)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID

END
GO
-----------------------------
-- Create trigger for Update
-----------------------------

CREATE TRIGGER udtOrderDetails

ON [Order Details]
AFTER UPDATE
AS
IF @@ROWCOUNT = 1
BEGIN

-- Single-row operation

UPDATE TC
SET TotalSales = TotalSales
+ I.UnitPrice * I.Quantity * (1 - I.Discount)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Inserted I
ON I.ProductID = P.productID

UPDATE TC
SET TotalSales = TotalSales
- D.UnitPrice * D.Quantity * (1 - D.Discount)

344
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
JOIN Deleted D
ON D.ProductID = P.productID

END
ELSE
BEGIN

-- Multi-row operation

UPDATE TC
SET TotalSales = TotalSales
+ (SELECT SUM(I.UnitPrice * Quantity * (1 - Discount))
FROM Inserted I
WHERE I.ProductID = P.productID)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID
UPDATE TC
SET TotalSales = TotalSales
- (SELECT SUM(D.UnitPrice * Quantity * (1 - Discount))
FROM Deleted D
WHERE D.ProductID = P.productID)
FROM TotalCategoriesSales TC
JOIN Products P
ON P.CategoryID = TC.CategoryID

END
GO

Tip

As shown previously in Listing 9.16, you can easily define a trigger for AFTER UPDATE as a
sequence of the actions defined in the AFTER INSERT and AFTER DELETE triggers.

Altering Trigger Definitions

To modify the definition of a trigger, you can use the ALTER TRIGGER statement. In this case, the trigger will
take the new definition directly. Listing 9.17 shows how to execute the ALTER TRIGGER statement to modify
the tr_Employees trigger.
The syntax is identical to the CREATE TRIGGER statement. Moreover, because triggers are independent
objects, no objects are depending on them. They can be dropped and re-created any time, if necessary.
Caution

You can change the name of a trigger using the sp_rename stored procedure, but this does not
change the name of the trigger stored in the definition of the trigger in syscomments.

To rename a trigger, it is recommended to drop the trigger and re-create it with a different name.

345
Microsoft SQL Server 2000 Programming by Example

Listing 9.17 You Can Use the ALTER TRIGGER Statement to Modify a Trigger.

USE Northwind
GO

-- Create a trigger to restrict

-- modifications to the employees table
-- to the dbo

CREATE TRIGGER tr_Employees

ON Employees
AFTER UPDATE, INSERT, DELETE
AS
IF CURRENT_USER <> 'dbo'
BEGIN
RAISERROR ('Only Database Owners can modify Employees, transaction rolled back',
10, 1)
ROLLBACK TRAN
END
GO

-- Modify the trigger to restrict

-- modifications to the employees table
-- to the members of the db_owner role

ALTER TRIGGER tr_Employees

ON Employees
AFTER UPDATE, INSERT, DELETE
AS
IF IS_MEMBER('db_owner') <> 1
BEGIN
RAISERROR ('Only Database Owners can modify Employees, transaction rolled back',
10 ,1)
ROLLBACK TRAN
END
GO

Disabling Triggers
To prevent triggers from running when data arrives through replication, you can add the NOT FOR
REPLICATION option to the CREATE TRIGGER or ALTER TRIGGER statements. In this case, the trigger will
fire on direct modifications to the base table, but not from subscription actions.
Temporarily, you can disable a trigger to speed up some processes. To do so, you can use the ALTER TABLE
statement with the DISABLE TRIGGER option, as in Listing 9.18.

Listing 9.18 You Can Disable a Trigger

346
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

USE Northwind
GO

-- To disable a single trigger

ALTER TABLE Employees

DISABLE TRIGGER tr_Employees --, isr_Employees, udt_Employees

-- To disable several triggers from the same table

ALTER TABLE Employees
DISABLE TRIGGER tr_Employees, isr_Employees, udt_Employees

-- To disable all the triggers from a table

ALTER TABLE Employees

DISABLE TRIGGER ALL
To reenable the trigger, use the ALTER TABLE statement with the ENABLE TRIGGER option. Listing 9.19
shows how to reenable the triggers that were disabled in Listing 9.18.

Listing 9.19 You Can Reenable a Trigger

USE Northwind
GO

-- To enable a single trigger

ALTER TABLE Employees

ENABLE TRIGGER tr_Employees --, isr_Employees, udt_Employees

-- To enable several triggers from the same table

ALTER TABLE Employees

ENABLE TRIGGER tr_Employees, isr_Employees, udt_Employees

-- To enable all the triggers from a table

ALTER TABLE Employees

ENABLE TRIGGER ALL

Nesting Triggers

347
Microsoft SQL Server 2000 Programming by Example

A trigger can be defined to modify a table, which in turn can have a trigger defined to modify another table,
and so on. In this case, triggers force the execution of other triggers, and the execution stops when the last
action does not fire any more triggers.
Because triggers are a specialized form of stored procedures, you can nest trigger execution up to 32 levels.
Triggers, stored procedures, scalar user-defined functions, and multistatement table-valued functions share
this limit. If the execution of a sequence of nested triggers requires more than 32 levels, the execution is
aborted, the transaction is rolled back, and the execution of the batch is cancelled.
Nested triggers are enabled by default. You can change this option at server level by setting the "nested
triggers" option to 0, using the system stored procedure sp_configure.
You can read the system function @@NESTLEVEL to know how many levels of nesting you have during the
execution of a trigger, stored procedure, or user-defined function.
Note

In a nested trigger situation, all the triggers are running inside the same transaction. Therefore, any
errors inside any of the triggers will roll back the entire transaction.

Listing 9.20 shows an example where you define triggers to maintain sales totals at different levels.

1. You insert, update, or delete data in the Order Details table. This data modification forces the
execution of the AFTER UPDATE trigger.
2. The AFTER UPDATE trigger in the Order Details table updates the SaleTotal column in the
Orders table.
3. Because the SaleTotal column in the Orders table has been updated, the existing AFTER
UPDATE trigger in the Orders table runs automatically and updates the SaleTotal column in the
Employees table and the Customers table.

Listing 9.20 You Can Create Triggers That Can Be Nested in Sequence

USE Northwind
GO

-- Add the column SaleTotal to the

-- Orders table

ALTER TABLE Orders

ADD SaleTotal money NULL

-- Add the column SaleTotal to the

-- Employees table

ALTER TABLE Employees

ADD SaleTotal money NULL

-- Add the column SaleTotal to the

-- Customers table

348
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

ALTER TABLE Customers

ADD SaleTotal money NULL
GO

-- Initialize the data

UPDATE Orders
SET SaleTotal =
(SELECT SUM([Order Details].UnitPrice * Quantity * (1 - Discount))
FROM [Order Details]
WHERE [Order Details].OrderID = Orders.OrderID)

UPDATE Employees
SET SaleTotal =
(SELECT SUM(Orders.SaleTotal)
FROM Orders
WHERE Orders.EmployeeID = Employees.EmployeeID)

UPDATE Customers
SET SaleTotal =
(SELECT SUM(Orders.SaleTotal)
FROM Orders
WHERE Orders.CustomerID = Customers.CustomerID)

GO

-- Create nested triggers

CREATE TRIGGER isrTotalOrderDetails

ON [Order details]
AFTER INSERT, DELETE, UPDATE
AS

IF @@rowcount = 1

-- Single-row operation

UPDATE Orders
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Inserted
WHERE Inserted.OrderID = Orders.OrderID), 0)
- ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Deleted
WHERE Deleted.OrderID = Orders.OrderID), 0)
ELSE

-- Multi-row operation

UPDATE Orders
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Inserted
WHERE Inserted.OrderID = Orders.OrderID), 0)
- ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Deleted

349
Microsoft SQL Server 2000 Programming by Example

WHERE Deleted.OrderID = Orders.OrderID), 0)

GO

CREATE TRIGGER isrTotalOrders

ON Orders
AFTER INSERT, DELETE, UPDATE
AS

IF @@rowcount = 1
BEGIN
-- Single-row operation

UPDATE Employees
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SaleTotal
FROM Inserted
WHERE Inserted.EmployeeID = Employees.EmployeeID), 0)
- ISNULL(
(SELECT SaleTotal
FROM Deleted
WHERE Deleted.EmployeeID = Employees.EmployeeID), 0)

UPDATE Customers
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SaleTotal
FROM Inserted
WHERE Inserted.CustomerID = Customers.CustomerID), 0)
- ISNULL(
(SELECT SaleTotal
FROM Deleted
WHERE Deleted.CustomerID = Customers.CustomerID), 0)

END
ELSE
BEGIN

-- Multi-row operation

UPDATE Employees
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(SaleTotal)
FROM Inserted
WHERE Inserted.EmployeeID = Employees.EmployeeID), 0)
- ISNULL(
(SELECT SUM(SaleTotal)
FROM Deleted
WHERE Deleted.EmployeeID = Employees.EmployeeID), 0)

UPDATE Customers
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(SaleTotal)
FROM Inserted
WHERE Inserted.CustomerID = Customers.CustomerID), 0)
- ISNULL(

350
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

(SELECT SUM(SaleTotal)
FROM Deleted
WHERE Deleted.CustomerID = Customers.CustomerID), 0)

END
GO

-- Updating Order Details

-- and forcing the nested triggers
-- execution

update [order details]

set quantity = 100
where orderid = 10248
and productid = 11

-- Testing totals in Orders table

select CustomerID, EmployeeID, SaleTotal from orders

WHERE OrderID = 10248
SELECT SUM([Order Details].UnitPrice * Quantity * (1 - Discount))
FROM [Order Details]
WHERE OrderID = 10248

-- Testing totals in Employees

SELECT SaleTotal
FROM Employees
WHERE EmployeeID = 5

SELECT SUM(SaleTotal)
FROM Orders
WHERE EmployeeID = 5

-- Testing totals in Customers

SELECT SaleTotal
FROM Customers
WHERE CustomerID = 'VINET'

SELECT SUM(SaleTotal)
FROM Orders
WHERE CustomerID = 'VINET'
GO

-- Dropping triggers

DROP TRIGGER isrTotalOrderDetails

DROP TRIGGER isrTotalOrders

Analyzing the previous example, you can see that the data is updated only at the Order Details level, and
two nested triggers maintain the summary information in the tables Orders, Employees, and Customers.
You can solve the same problem without using nested triggers. Create three triggers in the Order Details
table: one trigger to update the SaleTotal column in the Orders table, a second trigger to update the
Employees table, and a third one to update the Customers table. You can see in Listing 9.21 how to
implement this solution (note, you must execute the code from Listing 9.20 before running the code from
Listing 9.21).
Tip

351
Microsoft SQL Server 2000 Programming by Example

Create one trigger per logical action, as in Listing 9.21, and avoid nested triggers. Your database
application will be more modular and the maintenance will be easier.

Listing 9.21 Every Table Can Have Multiple Triggers for Each Action

USE Northwind
GO

CREATE TRIGGER tr_OrderDetails_TotalOrders

ON [Order details]
AFTER INSERT, DELETE, UPDATE
AS

IF @@rowcount = 1

-- Single-row operation

UPDATE Orders
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Inserted
WHERE Inserted.OrderID = Orders.OrderID), 0)
- ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Deleted
WHERE Deleted.OrderID = Orders.OrderID), 0)

ELSE

-- Multi-row operation

UPDATE Orders
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Inserted
WHERE Inserted.OrderID = Orders.OrderID), 0)
- ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Deleted
WHERE Deleted.OrderID = Orders.OrderID), 0)

GO

CREATE TRIGGER tr_OrderDetails_TotalEmployees

ON [Order details]

352
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

AFTER INSERT, DELETE, UPDATE

AS

IF @@rowcount = 1

-- Single-row operation

UPDATE Employees
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Inserted
JOIN Orders
ON Inserted.OrderID = Orders.OrderID
WHERE Orders.EmployeeID = Employees.EmployeeID), 0)
- ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Deleted
JOIN Orders
ON Deleted.OrderID = Orders.OrderID
WHERE Orders.EmployeeID = Employees.EmployeeID), 0)

ELSE

-- Multi-row operation

UPDATE Employees
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Inserted
JOIN Orders
ON Inserted.OrderID = Orders.OrderID
WHERE Orders.EmployeeID = Employees.EmployeeID), 0)
- ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Deleted
JOIN Orders
ON Deleted.OrderID = Orders.OrderID
WHERE Orders.EmployeeID = Employees.EmployeeID), 0)

GO

CREATE TRIGGER tr_OrderDetails_TotalCustomers

ON [Order details]
AFTER INSERT, DELETE, UPDATE
AS

IF @@rowcount = 1

-- Single-row operation
UPDATE Customers
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT UnitPrice * Quantity * (1 - Discount)
FROM Inserted
JOIN Orders
ON Inserted.OrderID = Orders.OrderID
WHERE Orders.CustomerID = Customers.CustomerID), 0)
- ISNULL(

353
Microsoft SQL Server 2000 Programming by Example

(SELECT UnitPrice * Quantity * (1 - Discount)

FROM Deleted
JOIN Orders
ON Deleted.OrderID = Orders.OrderID
WHERE Orders.CustomerID = Customers.CustomerID), 0)

ELSE

-- Multi-row operation

UPDATE Customers
SET SaleTotal = SaleTotal
+ ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Inserted
JOIN Orders
ON Inserted.OrderID = Orders.OrderID
WHERE Orders.CustomerID = Customers.CustomerID), 0)
- ISNULL(
(SELECT SUM(UnitPrice * Quantity * (1 - Discount))
FROM Deleted
JOIN Orders
ON Deleted.OrderID = Orders.OrderID
WHERE Orders.CustomerID = Customers.CustomerID), 0)
GO

DROP TRIGGER tr_OrderDetails_TotalOrders

DROP TRIGGER tr_OrderDetails_TotalCustomers

DROP TRIGGER tr_OrderDetails_TotalEmployees

GO
Note

The examples in Listings 9.20 and 9.21 create a single trigger for the three actions: INSERT,
DELETE, and UPDATE. Creating individual triggers per action is more efficient, as in Listing 9.16,
from the execution point of view. I use this strategy here only to simplify the examples.

Recursive Triggers
If a trigger defined in the Products table modifies data in the Employees table, and the Employees table
has a trigger that in turn modifies the Products table, the trigger defined in the Products table will fire again.
This situation is called indirect recursion, because a single statement forces multiple executions of the same
trigger, through the execution of other triggers. This is a special case of nested triggers, and everything said
about it in the preceding section can be applied to this case.
In some scenarios, it is possible to have direct recursion, when a table has a trigger that modifies some data
in the table again. In this case, by default, SQL Server will not fire the trigger again, avoiding this direct
recursion.
To enable trigger recursion in a database you must set the 'recursive triggers' option to 'true' at
database level using the sp_dboption system stored procedure, or set the option RECURSIVE_TRIGGERS
ON in the ALTER DATABASE statement. Listing 9.22 shows both statements.

Listing 9.22 You Can Enable Recursive Triggers at Database Level Only

354
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

-- Enable Recursive triggers in Northwind

EXEC sp_dboption 'Northwind', 'recursive triggers', 'true'

-- Disable Recursive triggers in Northwind

ALTER DATABASE Northwind

SET RECURSIVE_TRIGGERS OFF
Consider the typical hierarchical table where you save cost and budget breakdown of a project cost control
system. Every row in this table has a single ID as primary key, but it refers to another row as a parent row,
excluding the root row: the project itself. Any change on Cost or Budget in a row has to be escalated to the
highest level, and you introduce costs only in rows with no children.
This strategy is very flexible, adjusting changes easily on the distribution of activities in the project. Listing
9.23 shows the code to implement this example .

Listing 9.23 Use Triggers to Maintain Hierarchical Data

-- Create the base table

CREATE TABLE CostBudgetControl (

ID int NOT NULL
PRIMARY KEY,
Name nvarchar(100) NOT NULL,
ParentID int NULL
REFERENCES CostBudgetControl(ID),
Cost money NOT NULL DEFAULT 0,
Budget money NOT NULL DEFAULT 0,
HasChildren bit DEFAULT 0)

-- Insert Cost Structure

-- Create a text file (Gas.txt)
-- with the following contents:
/*
1, Gas Pipeline Project, 0, 85601000.0000, 117500000.0000, 1
2, Engineering, 1, 800000.0000, 950000.0000, 1
3, Materials, 1, 23400000.0000, 28000000.0000, 1
4, Construction, 1, 61000000.0000, 88000000.0000, 1
5, Supervision, 1, 401000.0000, 550000.0000, 1
6, Line, 2, 300000.0000, 400000.0000, 0
7, Stations, 2, 500000.0000, 550000.0000, 0
8, Pipes, 3, 14500000.0000, 16000000.0000, 0
9, Machinery, 3, 8900000.0000, 12000000.0000, 0
10, Section A, 4, 31000000.0000, 47000000.0000, 1

355
Microsoft SQL Server 2000 Programming by Example

11, Section B, 4, 30000000.0000, 41000000.0000, 1

12, Welding, 5, 200000.0000, 250000.0000, 0
13, Civil, 5, 145000.0000, 200000.0000, 0
14, Buildings, 5, 56000.0000, 100000.0000, 0
15, Civil works, 10, 20000000.0000, 30000000.0000, 0
16, Civil works, 11, 18000000.0000, 25000000.0000, 0
17, Pipeline, 10, 11000000.0000, 17000000.0000, 0
18, Pipeline, 11, 12000000.0000, 16000000.0000, 0
*/

BULK INSERT Northwind.dbo.CostBudgetControl

FROM 'C:\Gas.txt'
WITH
(
FIELDTERMINATOR = ', ',
ROWTERMINATOR = '\n'
)
GO

UPDATE CostBudgetControl
SET ParentID = NULL
WHERE ID = 1
GO

-- Create the recursive trigger

CREATE TRIGGER udtCostBudget

ON CostBudgetControl
AFTER UPDATE
AS
IF @@rowcount>0
UPDATE CostBudgetControl
SET Cost = Cost
+ ISNULL((SELECT SUM(Cost)
FROM Inserted
WHERE Inserted.ParentID = CostBudgetControl.ID), 0)
- ISNULL((SELECT SUM(Cost)
FROM Deleted
WHERE Deleted.ParentID = CostBudgetControl.ID), 0),
Budget = Budget
+ ISNULL((SELECT SUM(Budget)
FROM Inserted
WHERE Inserted.ParentID = CostBudgetControl.ID), 0)
- ISNULL((SELECT SUM(Budget)
FROM Deleted
WHERE Deleted.ParentID = CostBudgetControl.ID), 0)
WHERE ID IN
(SELECT ParentID
FROM Inserted
UNION
SELECT ParentID
FROM Deleted)
GO

-- Enable Recursive triggers

ALTER DATABASE Northwind

SET RECURSIVE_TRIGGERS ON
GO

356
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

-- Total Cost and Budget

-- Before the update

SELECT Cost, Budget

FROM CostBudgetControl
WHERE ID = 1

-- Update some cost

UPDATE CostBudgetControl
SET Cost = 12500000.0000
WHERE ID = 17

-- Total Cost and Budget

-- After the update listings;triggers;mainataining hierarchical
data>

SELECT Cost, Budget

FROM CostBudgetControl
WHERE ID = 1

GO

DROP TABLE CostBudgetControl

Cost Budget
--------------------- ---------------------
85601000.0000 117500000.0000

Cost Budget
--------------------- ---------------------
87101000.0000 117500000.0000

Security Implications of Using Triggers

Only certain users can create triggers:

• The owner of the table on which the trigger has to be defined

• Members of the db_owner and db_ddladmin database roles
• Members of the sysadmin server role, because permissions don't affect them

The user who creates the trigger needs specific permissions to execute the statements defined in the code of
the trigger.
Caution

If any of the objects referenced in the trigger don't belong to the same owner, you can have a
broken ownership chain situation. To avoid this situation, it is recommended that dbo must be the
owner of all the objects in a database.

357
Microsoft SQL Server 2000 Programming by Example

Enforcing Business Rules: Choosing Among INSTEAD of Triggers,

Constraints, and AFTER Triggers
This is the final chapter that discusses techniques to enforce data integrity, and as a summary, you can
propose which ways are recommended to enforce data integrity:

• To uniquely identify every row, define a PRIMARY KEY constraint. This is one of the first rules to apply
to designing a normalized database. Searching for values contained in a PRIMARY KEY is fast
because there is a UNIQUE INDEX supporting the PRIMARY KEY.
• To enforce uniqueness of required values in a column or group of columns, other than the PRIMARY
KEY, define a UNIQUE constraint. This constraint does not produce much overhead because there is
a UNIQUE INDEX supporting this constraint.
• To enforce uniqueness of optional values (columns that accept NULL), create a TRIGGER. You can
test this uniqueness before the data modification with an INSTEAD OF trigger, or after the data
modification with an AFTER trigger.
• To validate entries in a column, according to a specific pattern, range, or format, create a CHECK
constraint.
• To validate values in a row, where values in different columns must satisfy specific conditions, create
one or more CHECK constraints. If you create one CHECK constraint per condition, you can later
disable specific conditions only, if required.
• To validate values in a column, among a list of possible values, create a look -up table (LUT) with the
required values and create a FOREIGN KEY constraint to reference the look-up table. You could
create a CHECK constraint instead, but using a LUT is more flexible.
• To restrict values in a column to the values contained in a column in a second table, create a
FOREIGN KEY constraint in the first table .
• To make sure that every entry in a column is related to the primary key of another table, without
exceptions, define the FOREIGN KEY column as NOT NULL.
• To restrict the values in a column to complex conditions involving other rows in the same table, create
a TRIGGER to check these conditions. As an alternative, create a CHECK constraint with a user-
defined function to check this complex condition.
• To restrict the values in a column to complex conditions involving other tables in the same or different
database, create a TRIGGER to check these conditions.
• To declare a column as required, specify NOT NULL in the column definition.
• To specify a default value for columns where no value is supplied in INSERT operations, declare a
DEFAULT property for the column.
• To declare a column as autonumeric, declare an IDENTITY property in the column and specify the
seed value and the increment.
• To declare a default value, which depends on values in other rows or tables, declare a DEFAULT
property for the column using a user-defined function as a default expression.
• To cascade changes on primary keys to related fields in other tables, declare a FOREIGN KEY with
the ON UPDATE CASCADE clause. Do not create triggers to perform this operation.
• To delete in cascade related rows when the row in the primary table is deleted, declare a FOREIGN
KEY with the ON DELETE CASCADE clause. Do not create triggers to perform this operation.
• To cascade complex operations to other tables to maintain denormalized data, create individual
triggers to execute this operation.
• To validate INSERT, UPDATE, or DELETE operations applied through a view, define an INSTEAD OF
trigger on the view.
• Do not use RULE objects unless you want to define self-contained user-defined data types. It is
recommended to declare CHECK constraints instead.
• Do not use DEFAULT objects unless you want to define self-contained user-defined data types. It is
recommended to declare DEFAULT definitions instead .

What's Next?

This chapter covered the creation and use of triggers as a way to enforce complex data integrity.

358
 Chapter 9. Implementing Complex Processing Logic: Programming Triggers

Chapter 10, "Enhancing Business Logic: User-Defined Functions (UDF)," covers user-defined
functions, which can be used as part of the trigger definition and as an alternative to triggers, providing extra
computing capabilities to CHECK constraints and DEFAULT definitions.
Chapter 12, "Row-Oriented Processing: Using Cursors," explains how to use cursors. This could be
useful in some triggers to deal with multiple-row actions.
Triggers always work inside a transaction, and Chapter 13, "Maintaining Data Consistency:
Transactions and Locks," covers specifically that: transaction and locks. There you can see the
implications of modifying data through triggers and how to increase concurrency, preventing undesired
blockings.

359
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

Chapter 10. Enhancing Business Logic: User-Defined

Functions (UDF)
Procedural languages are based mainly in the capability to create functions, encapsulate complex
programming functionality, and return a value as a result of the operation. Using SQL Server 2000, you can
define user-defined functions(UDF), which combine the functionality of stored procedures and views but
provide extended flexibility.
This chapter teaches you the following:

• What the built-in user-defined functions are and how to use them
• How to define user-defined functions that return a scalar value
• How to define user-defined functions that return a result set
• How to convert stored procedures and views into user-defined functions
• How to extend the functionality of constraints with user-defined functions

Benefits of User-Defined Functions

You learned in Chapter 8, "Implementing Business Logic: Programming Stored Procedures," how to
create stored procedures, which are similar to the way you create functions in other programming languages.
However, using stored procedures from Transact-SQL is not very flexible, because you can use them only
with the EXECUTE or INSERT...EXECUTE statements. If you have a stored procedure that returns a single
value, you cannot use this procedure inside an expression. If your procedure returns a result set, you cannot
use this procedure in the FROM clause of any Transact-SQL statement.
In Chapter 3, "Working with Tables and Views," you learned about views and how to use them
anywhere as a replacement for tables. However, when you define a view, you are limited to a single SELECT
statement. Unlike stored procedures, you cannot define parameters in a view.
Some user-defined functions are similar to views but they can be defined with more than one statement and
they accept parameters. You can call user-defined functions in the same way you execute stored procedures,
and you can use scalar user-defined functions as part of any expression anywhere in a Transact-SQL
statement where an expression is valid. Furthermore, you can use a user-defined function that returns a table
in the FROM clause of any Transact-SQL Data Manipulation Language (DML) statement.
User-defined functions have many benefits in common with stored procedures, as covered in Chapter 8.
However, user-defined functions have more useful benefits. They enable you to

• Use the result set returned by a stored procedure in the FROM clause of a query
• Join the results of two stored procedures, without using temporary tables to store intermediate results
• Use the result of a stored procedure in the IN operator
• Use a stored procedure as a subquery in the WHERE clause
• Create a view that cannot be solved with a single SELECT statement
• Create a view with parameters similar to the way Microsoft Access creates queries with parameters
• Extend the list of built-in functions with any financial function
• Create new mathematical functions for any special scientific database applications that you might
require

This chapter will help you discover how user-defined functions can help you solve these common
programming problems.

Built-In User-Defined Functions

SQL Server 2000 implements some system functions as built-in user-defined functions. Many of them are not
documented; Query Analyzer, Enterprise Manager, Profiler, Replication, and other client applications and
system processes use some of these built-in user-defined functions internally. These functions can be used
almost as any other user-defined function, but SQL Server itself implements them.

361
Microsoft SQL Server 2000 Programming by Example

You cannot change the definition of these built-in user-defined functions. In some cases, you cannot see their
definition using the sp_help or sp_helptext system stored procedures, and you cannot script them.
However, their definition is stored in the syscomments system table as any other user-defined function.
Caution

Microsoft does not guarantee that undocumented built-in user-defined functions will remain
unchanged in the future; however, we can use some of them as examples of what kind of
operations a user-defined function can do.

In some cases, built-in user-defined functions return a single scalar value, and all of them are undocumented:

• fn_CharIsWhiteSpace(@nchar) returns 1 if the variable @nchar only contains a space, a tab

character, a newline character, or carriage return character; it returns 0 otherwise.
• fn_MSSharedVersion(@len_minorversion) returns the major and minor version number of
SQL Server. @len_minorversion specifies how many digits to show for the minor version.
• fn_MsGenSqeScStr(@pstrin) returns the string @pstring, converting single quotes into two
single quotes so that you are able to concatenate this string with other strings to execute a dynamic
statement.
• fn_IsReplMergeAgent() returns 1 if the present process is executed by the Replication Merge
Agent.
• fn_GetPersistedServerNameCaseVariation(@servername) returns the server name of the
server specified in @servername with exactly the same case it uses in the sysservers system
table, regardless of the case used to call this function.
• fn_ReplGetBinary8LoDWord(@binary8_value) takes the lower four bytes from the
@binary8_value binary variable and converts them into an integer value.
• fn_ReplPrepadBinary8(@varbinary8_value) converts the varbinary(8) value stored in
@varbinary8_value into a fixed-length binary(8) value with leading zeros.
• fn_ReplMakeStringLiteral(@string) converts the value stored in the @string value into a
UNICODE string, including quotes, such as N'Hello', to be used in dynamically constructed
statements.
• fn_ReplQuoteName(@string) returns the value stored in @string en closed in square brackets.
You can use this function in dynamic execution to select object names that contain spaces or
keywords, such as [Order Details].
• fn_GenerateParameterPattern(@parameter) returns a pattern string you can use with the
LIKE operator to test for strings containing any case variation of the value stored in @parameter,
such as converting 'Hello' into '%[hH][eE][lL][lL][oO]%'. This is useful in case-sensitive
servers, databases, or columns.
• fn_UpdateParameterWithArgument, fn_SkipParameterArgument, and
fn_RemoveParameterWithArgument are internal functions, and their study is not the purpose of
this book.

Listing 10.1 shows some examples of scalar built-in, user-defined functions and the partial result of some of
them.

Listing 10.1 Using Undocumented Built-In User-Defined Functions

362
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

USE Northwind
GO

PRINT CHAR(10)
+ 'fn_chariswhitespace(CHAR(9))'
+ CHAR(10)

select fn_chariswhitespace(CHAR(9))
GO

PRINT CHAR(10)
+ 'fn_mssharedversion(1)'
+ CHAR(10)

select master.dbo.fn_mssharedversion(1)
GO

PRINT CHAR(10)
+ 'fn_replgetbinary8lodword(0x0304030401020102)'
+ CHAR(10)
select fn_replgetbinary8lodword(0x0304030401020102)
GO

PRINT CHAR(10)
+ 'fn_replmakestringliteral(@a)'
+ CHAR(10)

declare @a varchar(100)

set @a = 'peter is right'

select fn_replmakestringliteral(@a)
GO

PRINT CHAR(10)
+ 'fn_replprepadbinary8(123456890123)'
+ CHAR(10)

select fn_replprepadbinary8(123456890123)
GO

PRINT CHAR(10)
+ 'fn_replquotename("hello")'
+ CHAR(10)

select fn_replquotename('hello')

363
Microsoft SQL Server 2000 Programming by Example

fn_chariswhitespace(CHAR(9))

----
1

fn_mssharedversion(1)

----------
80

fn_replgetbinary8lodword(0x0304030401020102)

-----------
16908546

fn_replmakestringliteral(@a)
----------------------------------
N'peter is right'

fn_replprepadbinary8(123456890123)

------------------
0x0C0000010BA59ABE

fn_replquotename("hello")

----------------------------------
[hello]
In other cases, built-in, user-defined functions return a table. SQL Server documents some of them:

• fn_ListExtendedProperty produces a list of available extended properties for a given database

or database objects, such as database users, user-defined data types, tables, views, stored
procedures, user-defined functions, default objects, rule objects, columns of tables and views,
parameters of stored procedures and user-defined functions, indexes, constraints, and triggers.
• fn_HelpCollations returns a list of the available collations.
• fn_ServerSharedDrives returns a list of the drives shared by a clustered server.
• fn_VirtualServerNodes returns the list of server nodes, defining a virtual server in a clustering
server environment.
• fn_VirtualFileStats returns statistical I/O information about any file in a database, including
transaction log files.

Listing 10.2 shows some examples of how to use these table-valued, built-in, user-defined functions.
Note

364
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

As you can see in Listing 10.2, you must call some of the built-in user-defined functions with
double colons (::) to differentiate them from user-defined functions that are not built in and do not
use the dbo as owner. Most of the built-in user-defined functions have a system owner called
system_function_schema.

Listing 10.2 Table-Valued, Built-In, User-Defined Functions

USE Northwind
GO

PRINT CHAR(10)
+ 'fn_helpcollations'
+ CHAR(10)

select *
from ::fn_helpcollations()
WHERE name LIKE 'Cyrillic%'
GO

PRINT CHAR(10)
+ 'fn_listextendedproperty(NULL, NULL, NULL, NULL, NULL, NULL, NULL)'
+ CHAR(10)

select *
from ::fn_listextendedproperty(NULL, NULL, NULL, NULL, NULL, NULL, NULL)
GO

PRINT CHAR(10)
+ 'fn_MSFullText()'
+ CHAR(10)

select *
from master.dbo.fn_MSFullText()
GO
fn_helpcollations

name description
----------------------------- --------------------------------------
Cyrillic_General_BIN Cyrillic-General, binary sort

365
Microsoft SQL Server 2000 Programming by Example

Cyrillic_General_CI_AI Cyrillic-General, case-insensitive, accent-

insensitive,
kanatype-insensitive, width-insensitive
Cyrillic_General_CI_AI_WS Cyrillic-General, case-insensitive, accent-
insensitive,
kanatype-insensitive, width-sensitive
Cyrillic_General_CI_AI_KS Cyrillic-General, case-insensitive, accent-
insensitive,
kanatype-sensitive, width-insensitive
Cyrillic_General_CI_AI_KS_WS Cyrillic-General, case-insensitive, accent-
insensitive,
kanatype-sensitive, width-sensitive
Cyrillic_General_CI_AS Cyrillic-General, case-insensitive, accent-
sensitive,
kanatype-insensitive, width-insensitive
Cyrillic_General_CI_AS_WS Cyrillic-General, case-insensitive, accent-
sensitive,
kanatype-insensitive, width-sensitive
Cyrillic_General_CI_AS_KS Cyrillic-General, case-insensitive, accent-
sensitive,
kanatype-sensitive, width-insensitive
Cyrillic_General_CI_AS_KS_WS Cyrillic-General, case-insensitive, accent-
sensitive,
kanatype-sensitive, width-sensitive
Cyrillic_General_CS_AI Cyrillic-General, case-sensitive, accent-
insensitive,
kanatype-insensitive, width-insensitive
Cyrillic_General_CS_AI_WS Cyrillic-General, case-sensitive, accent-
insensitive,
kanatype-insensitive, width-sensitive
Cyrillic_General_CS_AI_KS Cyrillic-General, case-sensitive, accent-
insensitive,
kanatype-sensitive, width-insensitive
Cyrillic_General_CS_AI_KS_WS Cyrillic-General, case-sensitive, accent-
insensitive,
kanatype-sensitive, width-sensitive
Cyrillic_General_CS_AS Cyrillic-General, case-sensitive, accent-sensitive,
kanatype-insensitive, width-insensitive
Cyrillic_General_CS_AS_WS Cyrillic-General, case-sensitive, accent-sensitive,
kanatype-insensitive, width-sensitive
Cyrillic_General_CS_AS_KS Cyrillic-General, case-sensitive, accent-sensitive,
kanatype-sensitive, width-insensitive
Cyrillic_General_CS_AS_KS_WS Cyrillic-General, case-sensitive, accent-sensitive,
kanatype-sensitive, width-sensitive

fn_listextendedproperty(NULL, NULL, NULL, NULL, NULL, NULL, NULL)

objtype objname name value

-------- -------- ----- -------------------------

fn_MSFullText()

LCID
-----------
2052
1028
1043
2057
1033
1036
1031

366
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

1040
1041
1042
0
1053
Other built-in user-defined functions help you to manage user-defined traces from Transact-SQL. You use
system stored procedures to define traces and built-in user-defined functions to get information about them.

• fn_trace_getinfo shows information about a specific trace or all the traces defined.
• fn_trace_gettable opens a trace file from disk and shows its information in a table format.
• fn_trace_geteventinfo shows information about the events defined for an active trace.
• fn_tracegetfilterinfo shows the filters applied to a specific trace.

There is a table-valued, built-in, user-defined function— fn_dblog—that is not documented, but it can be very
useful in some cases. fn_dblog reads the information contained in the transaction log. This is an alternative
to the DBCC LOG statement, undocumented as well, and less flexible than fn_dblog. Listing 10.3 shows
an example of this function.

Listing 10.3 Use fn_dblog to Look at the Transaction Log

USE Northwind
GO

PRINT CHAR(10)
+ 'fn_log(NULL, NULL)'
+ CHAR(10)

select TOP 10
[Current LSN], Operation
from ::fn_dblog(NULL, NULL)
ORDER BY [Current LSN] DESC
GO
fn_log(NULL, NULL)

Current LSN Operation

---------------------- -------------------------
0000002e:00000010:0006 LOP_COMMIT_XACT
0000002e:00000010:0005 LOP_DELTA_SYSIND
0000002e:00000010:0004 LOP_MODIFY_ROW
0000002e:00000010:0003 LOP_SET_FREE_SPACE
0000002e:00000010:0002 LOP_MODIFY_ROW
0000002e:00000010:0001 LOP_MODIFY_ROW

367
Microsoft SQL Server 2000 Programming by Example

0000002d:000001e4:0001 LOP_BEGIN_XACT
0000002d:000001c3:0011 LOP_COMMIT_XACT
0000002d:000001c3:0010 LOP_DELTA_SYSIND
0000002d:000001c3:000f LOP_MODIFY_ROW
Note

If you want to see the definitions of these built-in user-defined functions, you have them in the
installation scripts. Using Query Analyzer, open the following files located in the INSTALL directory:
procsyst.sql, replcom.sql, replsys.sql, repltran.sql, and sqldmo.sql.

Types of User-Defined Functions According to Their Return Value

You can define a user-defined function with a single statement or with multiple statements, as you will see
later in this chapter in the "Creating and Dropping User-Defined Functions" section.
According to their return value, user-defined functions can be divided into three groups:

• Scalar functions that return a single scalar value.

• Table-valued functions that return a full result set, similar to a table.
• Inline user-defined functions are a special case of table-valued user-defined functions, but they are
limited to a single SELECT statement.

Scalar Functions

Scalar user-defined functions return a single value, and they can be used wherever an expression is accepted,
such as

• In the SELECT clause of a SELECT statement, as a part of an expression or as an individual column

• In the SET clause of an UPDATE statement, as a value to insert into a field of the table being updated
• In the FROM clause of any DML statement (SELECT, UPDATE, INSERT, DELETE), as a single-
column, single-row result set–derived table
• In the FROM clause of any DML statement, as part of the joining conditions in the ON clause
• In the WHERE clause or HAVING clause of any DML statement
• In the GROUP BY clause, as part of any grouping condition
• In the ORDER BY clause of any statement, as sorting criteria
• As a DEFAULT value for a column
• Inside a CHECK CONSTRAINT definition
• Inside a CASE expression
• In a PRINT statement, if the user-defined function returns a string
• As part of the condition of IF or WHILE statements
• As part of the definition of a compute column
• As a parameter to call a stored procedure or another user-defined function
• As a return value of a stored procedure, if the user-defined function returns an integer value
• As a return value of another scalar user-defined function

Scalar user-defined functions can be combined with other functions in an expression, as long as the data
types are compatible with the operation.
Tip

You can identify scalar user-defined functions because they return a scalar data type and their
definition is enclosed in a BEGIN...END block.

368
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

Creating Scalar User-Defined Functions

To create a scalar user-defined function, you must use the CREATE FUNCTION statement, as shown in
Listing 10.4.
The first example creates the MaxProductID function, which selects the maximum available ProductID
from the Products table. This function returns a scalar integer value.
The second example creates the WhoWhere function, which returns a scalar string with information about
which is the login of the connected user and from which machine the user is connected.
The third example returns the date of the latest executed process in SQL Server, which we can consider
today's date in general.
The fourth example is a bit more complex; it generates a random number, between 0.0 and 1.0, based on
the time of the latest statement executed in SQL Server. This function is similar to the system function Rand;
however, Rand cannot be used inside a user-defined function, whereas this PRand function can be used.

Listing 10.4 Creating Scalar User-Defined Functions

USE Northwind
GO

-- Returns the maximum ProductID from Products

CREATE FUNCTION dbo.MaxProductID

()
RETURNS int
AS
BEGIN
RETURN (
SELECT MAX(ProductID)
FROM dbo.Products
)
END
GO

-- Returns who and from where the query is executed

CREATE FUNCTION dbo.WhoWhere

()
RETURNS nvarchar(256)
AS
BEGIN
RETURN SYSTEM_USER
+ 'FROM '
+ APP_NAME()
END
GO

-- Returns the date of the latest executed statement

-- which is usually today

CREATE FUNCTION dbo.Today

()

369
Microsoft SQL Server 2000 Programming by Example

RETURNS smalldatetime
AS
BEGIN
DECLARE @sdt smalldatetime

SELECT @SDT = CONVERT(varchar(10), MAX(last_batch), 112)

FROM master.dbo.sysprocesses

RETURN @SDT

END
GO

-- Function that produces a non-predictable

-- Pseudo-Random Number

CREATE FUNCTION dbo.PRand

()
RETURNS float
AS
BEGIN
DECLARE @dt datetime
DECLARE @dts varchar(3)
DECLARE @t1 float
DECLARE @t2 float
DECLARE @r float
-- Obtain the time of latest executed statement

SET @dt = (
SELECT MAX(last_batch)
FROM master.dbo.sysprocesses
)

-- Select only the milliseconds

SET @dts = RIGHT(CONVERT(varchar(20), @dt, 114) , 3)

-- Scramble the digits

SET @t1 = CAST(SUBSTRING(@dts, 2, 1)

+ RIGHT(@dts, 1)
+ LEFT(@dts, 1) AS int)

-- Obtain the time of latest executed statement

SET @dt = (
SELECT MAX(last_batch)
FROM master.dbo.sysprocesses
)

-- Select only the milliseconds

SET @dts = RIGHT(CONVERT(varchar(20), @dt, 114) , 3)

-- Scramble the digits

SET @t2 = CAST(SUBSTRING(@dts, 2, 1)

+ RIGHT(@dts, 1)
+ LEFT(@dts, 1) AS int)

-- Select the random number

370
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

SET @r = '0'+ LEFT(RIGHT(CONVERT(varchar(40), @t1 * @t2 / pi(), 2), 21), 16)

-- Return the random number

RETURN @r
END
GO
You can identify several parts in the CREATE FUNCTION syntax for a scalar user-defined function:

• CREATE FUNCTION ownername functionname, where you can specify the owner of the function
and the name of the function.
• () is an empty list of parameters. We will discuss parameters in the next section of this chapter.
• RETURNS datatype, where you define the data type for the returned value as the function's result.
• AS BEGIN...END to mark the function definition body.
• The function definition body.

You can define the body of the function in a way similar to a stored procedure, declaring and using variables,
using control-of-flow statements, accessing data from other tables, and other databases and servers.
Caution

Remember that you cannot modify data in existing tables inside a user-defined function directly.
This includes the creation of temporary tables.

Because writing long user-defined functions can be complex, you can break down long functions into smaller
ones that can be reused more often. Listing 10.5 creates a new version of the PRand function, created in
Listing 10.4. This version uses a base function, called Get3Rand, to generate the scrambled three-digit
number. The NewPRand function uses the Get3Rand function to generate two values and combine them to
provide the new random number.

Listing 10.5 New Definition for the Random Function Using Other Base Functions

USE Northwind
GO

-- Create a base function to extract a three-digits

-- number based on the scrambled version of the
-- milliseconds information of the latest executed
-- statement in SQL Server

CREATE FUNCTION dbo.Get3Rand

()
RETURNS int
AS

371
Microsoft SQL Server 2000 Programming by Example

BEGIN
DECLARE @dt datetime
DECLARE @dts varchar(3)

SET @dt = (
SELECT MAX(last_batch)
FROM master.dbo.sysprocesses
)
SET @dts = RIGHT(CONVERT(varchar(20), @dt, 114) , 3)

RETURN CAST(SUBSTRING(@dts, 2, 1)
+ RIGHT(@dts, 1)
+ LEFT(@dts, 1) AS int)

END
GO

-- Create the new NewPRand Random function

-- based on the Get3Rand function

CREATE FUNCTION dbo.NewPRand

()
RETURNS float
AS
BEGIN
DECLARE @r float

SET @r = '0'+ LEFT(RIGHT(CONVERT(varchar(40),

dbo.Get3Rand() * dbo.Get3Rand() / pi(), 2), 21), 16)

RETURN (@r)
END
Tip

If your user-defined function requires using temporary tables, you can use table variables instead.
Table variables are defined inside the user-defined function and can be modified inside it.

Listing 10.6 uses table variables defined internally inside a user-defined function to store intermediate
results. This function calculates the medium UnitPrice for products stored in the Products table.

Note

The medium value is the central value of an ordered list of values. The medium does not have to
be equal to the average value.

Listing 10.6 Using Table Variables Inside a Scalar User-Defined Function

372
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

USE Northwind
GO

CREATE FUNCTION dbo.MediumProductUnitPrice

()
RETURNS money
AS
BEGIN

-- Create hosting table variable

DECLARE @t TABLE(
id int identity(1,1),
UnitPrice money)

-- Inserts the product prices in ascending order

INSERT INTO @t (UnitPrice)

SELECT UnitPrice
FROM Products
ORDER BY UnitPrice ASC

-- Selects the medium price

RETURN (
SELECT MAX(UnitPrice)
FROM @t
WHERE ID <=
(SELECT MAX(ID)
FROM @t) / 2
)

END

Using Parameters in User-Defined Functions

As you learned in Chapter 8, you can expand stored procedure capabilities by using parameters. You can
create user-defined functions with parameters, too.
The examples from the preceding section do not use any parameter, which is why their execution does not
depend on any value that the user might send. Most of the system-supplied mathematical functions accept
one or more parameters and return a scalar result according to the mathematical operation to execute.
Trigonometric functions use a number as a parameter and return a number as a result. String functions take
one or more parameters and return a string.
You can create user-defined functions to expand the collection of system-supplied functions, using
parameters. You must define a parameter list in the CREATE FUNCTION statement, after the function name.
The parameters list is enclosed in parentheses. You must provide a data type for every parameter and,
optionally, a default value.
Caution

373
Microsoft SQL Server 2000 Programming by Example

The parameter list can be empty if the function does not use any input parameters. In this case,
you must supply two parentheses () to specify an empty parameters list.

If you do not specify a parentheses-enclosed parameter list, or at least an empty one, you will get a
syntax error when trying to create the user-defined function.

Listing 10.7 shows some examples of user-defined functions using parameters.

The first function, TotalPrice, computes the total price of a specific sale. You must provide the quantity
sold, the unit price to apply, the agreed discount, and then the function returns the total price of the sale.

The second function in Listing 10.7, fn_FV, computes the future value of an annuity, the FV financial
formula, as described in Microsoft Excel and Microsoft Visual Basic.

The third and fourth functions provide an example of how to create a user-defined function to perform basic
encryption.

Caution

The intention of the SimpleEncrypt and SimpleDecrypt user-defined functions is only to show
how to define a function to modify a string. The encryption used in these functions is too simple to
be used in a production environment.

Listing 10.7 Some Scalar User-Defined Functions Using Parameters

USE Northwind
GO

-----------------------------------------------------------
-- Generic function to compute the total price of a sale
-- from the quantity, unitprice and discount
-----------------------------------------------------------

CREATE FUNCTION dbo.TotalPrice

(@Quantity float, @UnitPrice money, @Discount float = 0.0)
RETURNS money
AS

374
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

BEGIN
RETURN (@Quantity * @UnitPrice * (1.0 - @Discount))
END
GO

-----------------------------------------------------------
-- Compute the future value of an annuity based on
-- periodic fixed payments with a fixed interest rate
-- Parameters:
-- @rate: interest rate between payments
-- @nper: number of payments
-- @pmt: payment to be made on every period
-- @pv: present value. Default to 0.0
-- @type: 0 if the payment is made at the end of each period (default)
-- 1 if the payment is made at the beginning of each period
-----------------------------------------------------------

CREATE FUNCTION dbo.fn_FV

(@rate float, @nper int, @pmt money, @pv money = 0.0, @type bit = 0)
RETURNS money
AS
BEGIN
DECLARE @fv money

IF @rate = 0
SET @fv = @pv + @pmt * @nper
ELSE
SET @fv = @pv * POWER(1 + @rate, @nper) +
@pmt * (((POWER(1 + @rate, @nper + @type) - 1) / @rate) - @type)

RETURN (-@fv)
END
GO

-----------------------------------------------------------
-- Encrypt the string increasing the Unicode value of every
-- character by the number of characters in the string
-----------------------------------------------------------

CREATE Function dbo.SimpleEncrypt

(@string nvarchar(4000))
RETURNS nvarchar(4000)
AS
BEGIN
DECLARE @output nvarchar(4000)
DECLARE @i int, @l int, @c int

SET @i = 1
SET @l = len(@string)
SET @output = ''

WHILE @i <= @l
BEGIN

SET @c = UNICODE(SUBSTRING(@string, @i, 1))

SET @output = @output +
CASE
WHEN @c > 65535 - @l
THEN NCHAR(@c + @l - 65536)
ELSE NCHAR(@c + @l) END

375
Microsoft SQL Server 2000 Programming by Example

SET @i = @i + 1

END

RETURN @output

END
GO

-----------------------------------------------------------
-- Decrypt the string decreasing the Unicode value of every
-- character by the number of characters in the string
-----------------------------------------------------------

CREATE Function dbo.SimpleDecrypt

(@string nvarchar(4000))
RETURNS nvarchar(4000)
AS
BEGIN
DECLARE @output nvarchar(4000)
DECLARE @i int, @l int, @c int

SET @i = 1
SET @l = len(@string)
SET @output = ''

WHILE @i <= @l
BEGIN

SET @c = UNICODE(SUBSTRING(@string, @i, 1))

SET @output = @output +

CASE
WHEN @c - @l >= 0
THEN NCHAR(@c - @l)
ELSE NCHAR(@c + 65535 - @l) END

SET @i = @i + 1
END

RETURN @output

END
GO
Listing 10.8 shows a more complex example. In this case, we want to create functions to convert angles
from and to HMS format (hours, minutes, and seconds, or degrees, minutes, and seconds). We want to be
able to use angles either in sexagesimal (1 circle = 360°) or centesimal (1 circle = 400g) units.
Note

Some mathematical books and scientific calculators refer to centesimal degrees as grades.

376
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

First, create the ANG_HMS user-defined function, which converts an angle into HMS format. To know the
angle's units, define the @d_g parameter to specify the character that identifies the degrees unit. The @d_g
parameter has only two possible values: 'g' (centesimal) or '°' (sexagesimal).

You can see in Listing 10.8 comments along the code of the ANG_HMS function, but in summary, it extracts
from the angle in sequence: degrees, minutes, and seconds. The @precision parameter serves us to
specify the number of digits to show for the fraction of the second.

Already having the generic ANG_HMS function, you can create the functions DEG_HMS and GRAD_HMS now to
use directly in the ANG_HMS function.

Following a similar process, you can create the HMS_ANG, HMS_DEG, and HMS_GRAD functions.

Listing 10.8 Functions to Convert Angles to and from HMS Format

USE Northwind
GO

-----------------------------------------------------------
-- Generic function to convert an angle
-- into HMS format
-- @d_g indicates if the angle is measured in
-- sexagesimal degrees (CHAR(176) as the degrees symbol) or
-- centesimal degrees ('g', for grades)
-----------------------------------------------------------

CREATE FUNCTION dbo.ANG_HMS

(@angle float, @precision int = 0, @d_g CHAR(1))
RETURNS varchar(21)
AS
BEGIN

-- Declare variables

DECLARE @sign float

DECLARE @ncircles int

DECLARE @degrees int

DECLARE @minutes int
DECLARE @seconds float
DECLARE @secInt int
DECLARE @secDec int

DECLARE @frac float

DECLARE @hms varchar(20)

377
Microsoft SQL Server 2000 Programming by Example

-- Save the sign of the angle

SET @sign = SIGN(@angle)

-- Take out the sign of the angle

-- to avoid calculation problems

SET @angle = ABS(@angle)

-- Extract the integer part as degrees

SET @degrees = CAST(FLOOR(@angle) AS int)

-- Count how many complete circles the angle has

SET @ncircles = @degrees /

CASE @d_g WHEN 'g'THEN 400
ELSE 360 END

-- Convert the angle into an angle from the first circle

SET @degrees = @degrees %

CASE @d_g WHEN 'g'THEN 400
ELSE 360 END

-- Extract the decimal part from the angle

SET @frac = @angle - @degrees - (@ncircles *

CASE @d_g WHEN 'g'THEN 400
ELSE 360 END
)

-- Extract minutes from the decimal part

SET @minutes = FLOOR(@frac *

CASE @d_g WHEN 'g'THEN 100
ELSE 60 END
)

-- Extract the number of seconds

SET @seconds = (@frac *

CASE @d_g WHEN 'g'THEN 100
ELSE 60 END
- @minutes) *
CASE @d_g WHEN 'g'THEN 100
ELSE 60 END

-- Calculate the number of complete seconds

SET @secInt = FLOOR(@seconds)

-- Set a limit for the fraction of a second to 9 digits

IF @precision > 9
SET @precision = 9

-- Extract the fraction of a second in the given precision

SET @secDec = (@seconds - @secInt) * POWER(10, @precision)

378
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

-- Start creating the resulting string with the sign

-- only for negative numbers

SET @hms = CASE @sign

WHEN -1 THEN '- '
ELSE ''END

-- Create the HMS format

SET @hms = @hms

+ CAST(@degrees AS varchar(3))
+ CHAR(176) + ''
+ CAST(@minutes AS varchar(2))
+ 'm '
+ CAST(@secInt AS varchar(2))
+ 's'
+ CAST(@secDec AS varchar(10))

RETURN @hms

END
GO

----------------------------------
-- Function to convert angles
-- measured in centesimal degrees
-- into HMS format
----------------------------------

CREATE FUNCTION dbo.GRAD_HMS

(@angle float, @precision int = 0)
RETURNS varchar(21)
AS
BEGIN

-- Call the ANG_HMS function with the 'g'format

RETURN dbo.ANG_HMS(@angle, @precision, 'g')

END
GO

-----------------------------------
-- Function to convert angles
-- measured in sexagesimal degrees
-- into HMS format
-----------------------------------

CREATE FUNCTION dbo.DEG_HMS

(@angle float, @precision int = 0)
RETURNS varchar(21)
AS
BEGIN

-- Call the ANG_HMS function with the CHAR(176) format

RETURN dbo.ANG_HMS(@angle, @precision, CHAR(176))

END
GO
-----------------------------------------------------------

379
Microsoft SQL Server 2000 Programming by Example

-- Generic function to convert an angle

-- from HMS format
-- @d_g indicates if the angle is measured in
-- sexagesimal degrees (CHAR(176) as the degrees symbol) or
-- centesimal degrees ('g', for grades)
-----------------------------------------------------------

CREATE FUNCTION dbo.HMS_ANG

(@hms varchar(22), @d_g char(1))
RETURNS float
AS
BEGIN

-- Declare variables

DECLARE @sign float

DECLARE @pos0 int

DECLARE @posg int
DECLARE @posm int
DECLARE @poss int

DECLARE @degrees float

DECLARE @angle float
DECLARE @minutes float
DECLARE @seconds float
DECLARE @secInt float
DECLARE @secDec float

-- Extract the sign

IF RIGHT(@hms, 1) = '-'
SELECT @sign = -1, @pos0 = 2
ELSE
SELECT @sign = 1, @pos0 = 1

-- Search for the position in the string

-- of the character dividing degrees, minutes, and seconds

SET @posg = CHARINDEX(@d_g, @hms)

SET @posm = CHARINDEX('m', @hms)
SET @poss = CHARINDEX('s', @hms)

-- Extract the value of the degrees

SET @degrees = SUBSTRING(@hms, @pos0, @posg - @pos0)

-- Extract the value of the minutes

SET @minutes = SUBSTRING(@hms, @posg + 1, @posm - @posg - 1)

-- Extract the value of the seconds

-- as integer and decimal part

SET @secInt = SUBSTRING(@hms, @posm + 1, @poss - @posm - 1)

SET @secDec = SUBSTRING(@hms, @poss + 1, len(@hms) - @poss)
/ POWER(10.0, len(@hms) - @poss)

-- Calculate the angle

SET @angle = @sign * (

@degrees +

380
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

@minutes /
CASE @d_g WHEN 'g'THEN 100.0
ELSE 60.0 END +
(@secInt + @secdec) /
CASE @d_g WHEN 'g'THEN 10000.0
ELSE 3600.0 END)

-- Return the value

RETURN @angle

END
GO

----------------------------------
-- Function to convert angles
-- measured in sexagesimal degrees
-- from HMS format
----------------------------------

CREATE FUNCTION dbo.HMS_DEG

(@hms varchar(22))
RETURNS float
AS
BEGIN

-- Call the HMS_ANG function with the CHAR(176) format

RETURN dbo.HMS_ANG(@hms, CHAR(176))

END
GO

----------------------------------
-- Function to convert angles
-- measured in centesimal degrees
-- from HMS format
----------------------------------

CREATE FUNCTION dbo.HMS_GRAD

(@hms varchar(22))
RETURNS float
AS
BEGIN

-- Call the ANG_HMS function with the 'g'format

RETURN (dbo.HMS_ANG(@hms, 'g'))

END

GO

Invoking Scalar User-Defined Functions

You can use scalar user-defined functions anywhere in any Transact-SQL statement in which an expression
is allowed.
If you want to invoke a user-defined function, you must qualify the function name with its owner, usually dbo.
Listing 10.9 shows some statements using the MaxProductID in different valid ways.

Listing 10.9 How to Invoke a Scalar User-Defined Function

381
Microsoft SQL Server 2000 Programming by Example

SELECT dbo.MaxProductID()
GO

SELECT ProductID, dbo.MaxProductID() AS 'MaxID'

FROM Products
GO

UPDATE [Order Details]

SET ProductID = dbo.MaxProductID()
WHERE ProductID = 25
GO

SELECT ProductID, MaxID

FROM Products
CROSS JOIN (SELECT dbo.MaxProductID() AS 'MaxID') AS MI
GO

SELECT P.ProductID, Quantity

FROM Products AS P
JOIN [Order Details] AS OD
ON P.ProductID = OD.ProductID
AND P.ProductID = dbo.MaxProductID()
GO

SELECT P.ProductID, Quantity

FROM Products AS P
JOIN [Order Details] AS OD
ON P.ProductID = OD.ProductID
WHERE P.ProductID = dbo.MaxProductID()
GO

SELECT P.ProductID, SUM(Quantity)

FROM Products AS P
JOIN [Order Details] AS OD
ON P.ProductID = OD.ProductID
GROUP BY P.ProductID
HAVING P.ProductID = dbo.MaxProductID()
GO

SELECT ProductID, ProductName,

CASE ProductID
WHEN dbo.MaxProductID() THEN 'Last Product'
ELSE ''END AS Note
FROM Products
GO

DECLARE @ID int

SET @ID = dbo.MaxProductID()

Caution

382
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

User-defined functions are local to the database where they are created. Although it is possible to
create global user-defined functions, Microsoft does not support this functionality.

If a user-defined function uses parameters, you must specify a value for each parameter on every call, even if
they have default values. You can use the keyword DEFAULT to provide the default value for a parameter that
has a default value. You cannot omit a parameter when calling a user-defined function— because it has a
default value already— otherwise, you will receive a syntax error message, as in Listing 10.10, where you
can use the TotalPrice function without providing a value for discount, hoping that the function will use the
default value.

Listing 10.10 Failing to Provide a Parameter Produces a Syntax Error

USE Northwind
GO

-- This is an illegal call, because it does not provide

-- a value for the @discount parameter

SELECT dbo.TotalPrice (12, 25.4)

GO

-- This is a valid call, because it does provide

-- a value for every parameter

SELECT dbo.TotalPrice (12, 25.4, 0.0)

GO

Server: Msg 313, Level 16, State 2, Line 1

An insufficient number of arguments were supplied for the procedure or function
dbo.TotalPrice.

---------------------
304.8000
Listing 10.11 shows how to invoke some of the functions defined in Listings 10.4, 10.5, 10.6, 10.7, and
10.8. These examples show how to use these functions in different kinds of statements.

Listing 10.11 How to Invoke Scalar User-Defined Functions

383
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

-- dbo.MaxProductID

SELECT dbo.MaxProductID()
AS MaxProductID
GO

-- dbo.WhoWhere

SELECT dbo.WhoWhere() AS [Who from Where]

GO

-- Create a table using WHoWhere

-- As a DEFAULT constraint
IF OBJECT_ID('TestWhoWhere') IS NOT NULL
DROP TABLE TestWhoWhere
GO

CREATE TABLE TestWhoWhere(

ID int IDENTITY (1,1)
PRIMARY KEY,
Name nvarchar(40),
WhoWhere nvarchar(256) DEFAULT dbo.WhoWhere())

INSERT TestWhoWhere (Name)

VALUES ('New record')

SELECT *
FROM TestWhoWhere
GO

-- Create a trigger to Update automatically

-- Who did it and from where the change was done

CREATE TRIGGER trWhoWhere

ON TestWhoWhere
AFTER INSERT, UPDATE
AS
UPDATE TestWhoWhere
SET WhoWhere = dbo.WhoWhere()
FROM TestWhoWhere T
JOIN Inserted I
ON T.ID = I.ID
GO

INSERT TestWhoWhere
(Name, WhoWhere)
VALUES
('More records', 'nobody from nowhere')

SELECT *

384
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

FROM TestWhoWhere

-- dbo.Today

SELECT dbo.Today()
AS Today

-- Inserting an Order in the present month

INSERT Orders (CustomerID, OrderDate, ShippedDate)

VALUES ('WELLI', '2000-11-23', '2000-12-12')

-- Searching for Orders in the last 30 days

SELECT OrderID, CustomerID, OrderDate, ShippedDate

FROM Orders
WHERE DATEDIFF(day, ShippedDate, dbo.Today()) < 30
GO

-- dbo.PRand

SELECT dbo.PRand() AS PRand

GO

-- dbo.Get3Rand

SELECT dbo.Get3Rand() as Get3Rand1

GO

SELECT dbo.Get3Rand() as Get3Rand2

GO

-- dbo.NewPRand

SELECT dbo.NewPRand() as NewPRand

GO

-- dbo.MediumProductUnitPrice

SELECT dbo.MediumProductUnitPrice()
GO

-- Get the Medium 10 products

-- by UnitPrice

PRINT CHAR(10) + 'Medium 10 Products'+ CHAR(10)

SELECT *
FROM (SELECT TOP 6
UnitPrice, ProductID, ProductName
FROM Products
WHERE UnitPrice >= dbo.MediumProductUnitPrice()
ORDER BY UnitPrice ASC) AS A

UNION

SELECT *
FROM (SELECT TOP 5
UnitPrice, ProductID, ProductName
FROM Products
WHERE UnitPrice <= dbo.MediumProductUnitPrice()

385
Microsoft SQL Server 2000 Programming by Example

ORDER BY UnitPrice DESC) AS B

ORDER BY UnitPrice ASC

-- dbo.fn_FV

SELECT 0.07 AS Rate,

36 AS NPer,
1000 AS Pmt,
10000 AS Pv,
0 AS Type,
dbo.fn_FV(0.07, 36, 1000, 10000, 0) AS FV

-- dbo.SimpleEncrypt

SELECT dbo.SimpleEncrypt('Hello World')

AS [Encrypted Message]

-- dbo.SimpleDecrypt

SELECT dbo.SimpleDecrypt('Hello World')

AS [Decrypted version of a non-encrypted message]

SELECT dbo.SimpleDecrypt(dbo.SimpleEncrypt('Hello World'))

AS [Decrypted version of an encrypted message]

-- dbo.ANG_HMS

SELECT dbo.ANG_HMS(12.3456, 3, 'g') AS 'GRAD to HMS Using ANG_HMS'

SELECT dbo.ANG_HMS(12.3456, 3, CHAR(176)) AS 'DEG to HMS Using ANG_HMS'

-- dbo.GRAD_HMS

SELECT dbo.GRAD_HMS(12.3456, DEFAULT) AS 'GRAD to HMS Using GRAD_HMS'

-- dbo.DEG_HMS

SELECT dbo.DEG_HMS(12.3456, 2) AS 'DEG to HMS Using DEG_HMS'

-- dbo.HMS_ANG
SELECT dbo.HMS_ANG('12g 34m 56s789', 'g') AS 'HMS to GRAD Using HMS_ANG'

SELECT dbo.HMS_ANG('12 34m 56s789', ' ') AS 'HMS to DEG Using HMS_ANG'

-- dbo.HMS_DEG

SELECT dbo.HMS_DEG('12 34m 56s789') AS 'HMS to DEG Using HMS_DEG'

-- dbo.HMS_GRAD

SELECT dbo.HMS_GRAD('12g 34m 56s789') AS 'HMS to GRAD Using HMS_GRAD'

SELECT dbo.GRAD_HMS(dbo.HMS_GRAD('12g 34m 56s789'), 3)

AS 'HMS to GRAD and to HMS again using HMS_GRAD and GRAD_HMS'

GO

386
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

MaxProductID
------------
77

Who from Where

-------------------------------------------
SQLBYEXAMPLE\SQLAdmin FROM SQL Query Analyzer

ID Name WhoWhere
-------- --------------- ---------------------------------------------
1 New record SQLBYEXAMPLE\SQLAdmin FROM SQL Query Analyzer

ID Name WhoWhere
-------- --------------- ---------------------------------------------
1 New record SQLBYEXAMPLE\SQLAdmin FROM SQL Query Analyzer
2 More records SQLBYEXAMPLE\SQLAdmin FROM SQL Query Analyzer

Today
------------------------------------------------------
2000-12-26 00:00:00

OrderID CustomerID OrderDate ShippedDate

----------- ---------- ------------------------ -----------------------
11109 WELLI 2000-11-23 00:00:00.000 2000-12-12 00:00:00.000
11110 WELLI 2000-11-23 00:00:00.000 2000-12-12 00:00:00.000
11111 WELLI 2000-11-23 00:00:00.000 2000-12-12 00:00:00.000

PRand
-----------------------------------------------------
0.40721851426491701
Get3Rand1
-----------
475

Get3Rand2
-----------
575

NewPRand
-----------------------------------------------------
5.2412061195157997E-2

---------------------
19.4500

Medium 10 Products

UnitPrice ProductID ProductName

--------------------- ----------- ----------------------------------------

387
Microsoft SQL Server 2000 Programming by Example

18.0000 1 Chai
18.4000 40 Boston Crab Meat
19.0000 2 Chang
19.0000 36 Inlagd Sill
19.4500 44 Gula Malacca
19.5000 57 Ravioli Angelo
20.0000 49 Maxilaku
21.0000 11 Queso Cabrales
21.0000 22 Gustaf's Knäckebröd
21.0500 65 Louisiana Fiery Hot Pepper Sauce

Rate NPer Pmt Pv Type FV

---- ----------- ----------- ----------- ----------- ---------------------
.07 36 1000 10000 0 -263152.8817

Encrypted Message
---------------------------
Spwwz+bz} wo

Decrypted version of a non-encrypted message

---------------------------------------------
=Zaad_LdgaY

Decrypted version of an encrypted message

---------------------------------------------
Hello World

GRAD to HMS Using ANG_HMS

-------------------------
12 34m 55s999

DEG to HMS Using ANG_HMS

------------------------
12 20m 44s159

GRAD to HMS Using GRAD_HMS

--------------------------
12 34m 55s0

DEG to HMS Using DEG_HMS

------------`------------
12 20m 44s15

HMS to GRAD Using HMS_ANG

-----------------------------------------------------
12.345678899999999

HMS to DEG Using HMS_ANG

-----------------------------------------------------
12.582441388888888

HMS to DEG Using HMS_DEG

-----------------------------------------------------
12.582441388888888

HMS to GRAD Using HMS_GRAD

-----------------------------------------------------
12.345678899999999

HMS to GRAD and to HMS again using HMS_GRAD and GRAD_HMS

--------------------------------------------------------

388
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

12 34m 56s788
You can use fields as parameters when calling a user-defined function in any statement, such as in Listing
10.12.

Listing 10.12 You Can Apply Scalar User-Defined Functions to Table Fields

USE Northwind
GO

-- Use the TotalPrice function to retrieve

-- information from the Order Details table

SELECT OrderID, ProductID,

dbo.TotalPrice(Quantity, UnitPrice, Discount) AS TotalPrice
FROM [Order Details]
WHERE ProductID = 12

-- Use the SimpleEncrypt function to encrypt product names

SELECT ProductID,
dbo.SimpleEncrypt(ProductName) AS EncryptedName
FROM Products
WHERE CategoryID = 3

-- Use the SimpleDecrypt function to decrypt

-- a field encrypted with the SimpleEncrypt function

SELECT ProductID,
dbo.SimpleDecrypt(EncryptedName) AS ProductName
FROM (
SELECT ProductID, CategoryID,
dbo.SimpleEncrypt(ProductName) AS EncryptedName
FROM Products
) AS P
WHERE CategoryID = 3

OrderID ProductID TotalPrice

----------- ----------- ---------------------
10266 12 346.5600
10439 12 456.0000
10536 12 427.5000
10543 12 969.0000
10633 12 1162.8000

389
Microsoft SQL Server 2000 Programming by Example

10678 12 3800.0000
10695 12 152.0000
10718 12 1368.0000
10968 12 1140.0000
10979 12 760.0000
11018 12 760.0000
11046 12 722.0000
11049 12 121.6000
11077 12 72.2000

ProductID ProductName
----------- -----------------------------
11 Queso Cabrales
12 Queso Manchego La Pastora
31 Gorgonzola Telino
32 Mascarpone Fabioli
33 Geitost
59 Raclette Courdavault
60 Camembert Pierrot
69 Gudbrandsdalsost
71 Flotemysost
72 Mozzarella di Giovanni
A scalar user-defined function is a special form of stored procedure, and you can execute scalar user-defined
functions in the same way you execute stored procedures. In this case, you can omit parameters that have a
default value and alter the order of the parameters, providing the parameter name in the user-defined function
call in the same way you do for stored procedures. Listing 10.13 shows some examples of how to call a
scalar user-defined function using the EXECUTE statement.

Listing 10.13 You Can Use EXECUTE to Invoke Scalar User-Defined Functions

USE Northwind
GO

--Declare a variable to receive the result of the UDF

DECLARE @Total money

-- Use EXECUTE and provide values for every parameter

EXECUTE @Total = dbo.TotalPrice 12, 25.4, 0.0

SELECT @Total

-- Use EXECUTE and omit the @Discount parameter

-- because it has a default value

EXECUTE @Total = dbo.TotalPrice 12, 25.4

SELECT @Total

-- Use EXECUTE and omit the UDF owner, because it defaults to dbo

390
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

EXECUTE @Total = TotalPrice 12, 25.4

SELECT @Total

-- Use EXECUTE and provide values for every parameter

-- specifying parameter names

EXECUTE @Total = TotalPrice

@Quantity = 12,
@UnitPrice = 25.4,
@Discount = 0.2

SELECT @Total

-- Use EXECUTE and provide values for every parameter

-- specifying parameter by order and by name on any order

EXECUTE @Total = TotalPrice 12,

@Discount = 0.2,
@UnitPrice = 25.4

SELECT @Total
---------------------
304.8000

---------------------
304.8000

---------------------
304.8000

---------------------
243.8400

---------------------
243.8400
Tip

If you invoke a user-defined function with the EXECUTE statement, you do not have to qualify the
function with the owner name. However, it is always more efficient to qualify the objects you use,
because in this way SQL Server does not have to search first for an object owned by you, before
searching for the same object owned by dbo.

391
Microsoft SQL Server 2000 Programming by Example

Inline Table-Valued User-Defined Functions

You can use views in any DML statement as if they were tables. You can use a WHERE clause inside the view
definition to limit the view results to specific rows, but this restriction is fixed, because it is part of the view
definition. In the query where you use the view, you can use another WHERE clause to limit the search. In this
case, SQL Server combines both WHERE clauses in the final query plan, after the view definition is merged
with the outer query.
SQL Server 2000 gives you a new feature to create something similar to parameterized views: inline user-
defined functions.
An inline user-defined function contains a single SELECT statement but, unlike views, it can use several
parameters to restrict the query, providing an easier call interface than views.
Tip

You can identify inline user-defined functions because they return a table and their definition has
only a single SELECT statement, without a BEGIN...END block.

You can use an inline user-defined function wherever a table or view is accepted:

• In the SELECT clause of a SELECT statement, as part of a subquery that returns a single value (a
single row and single column result set).
• In the SET clause of an UPDATE statement, as part of a subquery that provides a single value for a
field in the table to be updated.
• In the FROM clause of any DML statement.
• In the WHERE or HAVING clauses of any DML statement, as part of a subquery that returns a single
value to be compared to any field or variable.
• In the WHERE clause of any DML statement, as part of a subquery introduced by EXISTS or NOT
EXISTS.
• In the WHERE or HAVING clause of any DML statement, as part of a subquery used with the IN or NOT
IN operators, as long as the subquery returns a single column.

Creating Inline Table-Valued User-Defined Functions

To create an inline user-defined function, you must use the CREATE FUNCTION statement, as shown in
Listing 10.14.
The first example creates the GetCustomersFromCountry function, which selects customers based in a
specific country. This function returns a result set with the same structure as the original Customers table.
Based on the GetCustomersFromCountry function, you can create the GetCustomersFromUSA function,
which retrieves customers based in the USA only.
The third example returns orders from a specific day using the GetOrdersFromDay function.
The fourth example creates the GetOrdersFromToday function to retrieve the list of today's orders. Note
that you cannot base this function in GetOrdersFromDay function, because you cannot invoke an inline
user-defined function using a scalar function as a parameter value.
Note

Inside the GetOrdersFromToday function you use the Today() scalar user-defined function,
because Getdate() is not valid inside the definition of a user-defined function.

Later in this chapter, in the "Deterministic and Nondeterministic Functions" section, you will
see the lists of functions that are not valid inside a user-defined function.

392
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

The next function is the OrdersWithValue function, which uses the TotalPrice scalar function to
calculate the total value for every order. This inline user-defined function retrieves a result set with the same
structure as the Orders table, plus the TotalValue field.

Based on the OrdersWithValue function, you can create the next two functions. OrdersByValue retrieves
the orders where TotalValue is greater than a specific target total value. TopTenOrders returns the top 10
orders by TotalValue.

Listing 10.14 Inline User-Defined Functions Examples Using the CREATE FUNCTION Statement

USE Northwind
GO

-- Returns customers from a specific country

CREATE FUNCTION dbo.GetCustomersFromCountry

(@country nvarchar(15))
RETURNS TABLE
AS
RETURN (
SELECT *
FROM Customers
WHERE Country = @country
)
GO

-- Returns USA-based customers

CREATE FUNCTION dbo.GetCustomersFromUSA

()
RETURNS TABLE
AS
RETURN (
SELECT *
FROM dbo.GetCustomersFromCountry(N'USA')
)
GO

-- Returns the orders from a specific day

CREATE FUNCTION dbo.GetOrdersFromDay

(@date as smalldatetime)
RETURNS TABLE
AS
RETURN (
SELECT *

393
Microsoft SQL Server 2000 Programming by Example

FROM Orders
WHERE DATEDIFF(day, OrderDate, @date) = 0
)
GO

-- Returns orders from today

CREATE FUNCTION dbo.GetOrdersFromToday

()
RETURNS TABLE
AS
RETURN (
SELECT *
FROM Orders
WHERE DATEDIFF(day, OrderDate, dbo.Today()) = 0
)
GO

-- Returns Orders with the total order value

CREATE FUNCTION dbo.OrdersWithValue

()
RETURNS TABLE
AS
RETURN (
SELECT O.*, TotalValue
FROM Orders O
JOIN (
SELECT OrderID, SUM(dbo.TotalPrice(Quantity, UnitPrice, Discount))
AS TotalValue
FROM [Order Details]
GROUP BY OrderID) AS OD
ON O.OrderID = OD.OrderID
)
GO

-- Returns orders with a value greater than

-- a specific target value

CREATE FUNCTION dbo.OrdersByValue

(@total money)
RETURNS TABLE
AS
RETURN (
SELECT *
FROM dbo.OrdersWithValue()
WHERE TotalValue > @total
)
GO

-- Returns the top 10 orders by total value

CREATE FUNCTION dbo.TopTenOrders

()
RETURNS TABLE
AS
RETURN (
SELECT TOP 10 WITH TIES *
FROM dbo.OrdersWithValue()
ORDER BY TotalValue DESC
)

394
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

GO
You can identify several parts of the CREATE FUNCTION syntax for inline user-defined function:

• CREATE FUNCTION ownername.functionname, where you can specify the owner of the function
and the name of the function.
• (parameter_lists), where every parameter is identified by its name and the data type. The
parameter list can be empty.
• RETURNS TABLE, because inline user-defined functions always return a tablelike result set.
• AS RETURN () marks the function definition body, which has to be enclosed inside the parenthesis
block.
• The SELECT statement, which defines the result set to return.

In essence, the creation of an inline user-define function is not very different from the creation of a view,
except for the parameters list, which gives inline user-defined functions extra functionality not available in
views.
Tip

Inline user-defined functions have the same restrictions that apply to views; look at Chapter 3,
"Working with Tables and Views," for more information about them. One of these restrictions is
that you cannot use ORDER BY, but you can use TOP 100 PERCENT...ORDER BY to produce
the same result. Listing 10.15 shows how to implement this trick.

Listing 10.15 Use TOP 100 PERCENT to Create an Inline User-Defined Function That Produces Sorted
Data

USE Northwind
GO

-- Returns Products ordered by ProductName

CREATE FUNCTION dbo.OrderedProducts()

RETURNS TABLE
AS
RETURN (SELECT TOP 100 PERCENT *
FROM Products
ORDER BY ProductName ASC)
GO

-- Test the function

SELECT TOP 10 ProductID, ProductName

FROM dbo.OrderedProducts()

395
Microsoft SQL Server 2000 Programming by Example

ProductID ProductName
----------- ----------------------------------------
17 Alice Mutton
3 Aniseed Syrup
40 Boston Crab Meat
60 Camembert Pierrot
18 Carnarvon Tigers
1 Chai
2 Chang
39 Chartreuse verte
4 Chef Anton's Cajun Seasoning
5 Chef Anton's Gumbo Mix

Invoking Data from Inline User-Defined Functions

You can invoke an inline user-defined function in the same way you invoke a table or a view in a DML
statement, with the only exception that you must use parentheses after the function name, even if there are
not any parameters to use.
SQL Server merges the definition of the Inline function with the definition of the query where the function is
invoked, to create a unique query plan. This is the same way that SQL Server uses views to execute
Transact-SQL statements.
Listing 10.16 shows how to invoke the inline user-defined functions defined in Listing 10.14 in different
ways.

Listing 10.16 Invoking Inline User-Defined Functions

USE Northwind
GO

PRINT CHAR(10) + 'Use GetCustomersFromCountry(''Mexico'')'+ CHAR(10)

SELECT CustomerID, CompanyName, City

FROM dbo.GetCustomersFromCountry('Mexico')

PRINT CHAR(10) + 'Use GetCustomersFromUSA()'+ CHAR(10)

Select CustomerID, CompanyName, City

FROM dbo.GetCustomersFromUSA()

PRINT CHAR(10)
+ 'Use GetCustomersFromCountry(''Mexico'') with the IN operator'
+ CHAR(10)

396
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

SELECT OrderID,
CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM Orders
WHERE CustomerID IN
(SELECT CustomerID
FROM dbo.GetCustomersFromCountry('Mexico'))
PRINT CHAR(10)
+ 'Joins OrdersByValue to Customers'
+ CHAR(10)

SELECT CompanyName, OrderID, TotalValue,

CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM dbo.OrdersByValue(10000) AS OBV
JOIN Customers C
ON OBV.CustomerID = C.CustomerID

PRINT CHAR(10)
+ 'Joins TopTenOrders to Customers'
+ CHAR(10)

SELECT CompanyName, OrderID, TotalValue,

CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM dbo.TopTenOrders() AS OBV
JOIN Customers C
ON OBV.CustomerID = C.CustomerID
Use GetCustomersFromCountry('Mexico')

CustomerID CompanyName City

---------- ---------------------------------------- ---------------
ANATR Ana Trujillo Emparedados y helados México D.F.
ANTON Antonio Moreno Taquería México D.F.
CENTC Centro comercial Moctezuma México D.F.
PERIC Pericles Comidas clásicas México D.F.
TORTU Tortuga Restaurante México D.F.

Use GetCustomersFromUSA()

CustomerID CompanyName City

---------- ---------------------------------------- ---------------
GREAL Great Lakes Food Market Eugene
HUNGC Hungry Coyote Import Store Elgin
LAZYK Lazy K Kountry Store Walla Walla
LETSS Let's Stop N Shop San Francisco
LONEP Lonesome Pine Restaurant Portland
OLDWO Old World Delicatessen Anchorage
RATTC Rattlesnake Canyon Grocery Albuquerque
SAVEA Save-a-lot Markets Boise
SPLIR Split Rail Beer & Ale Lander
THEBI The Big Cheese Portland

397
Microsoft SQL Server 2000 Programming by Example

THECR The Cracker Box Butte

TRAIH Trail's Head Gourmet Provisioners Kirkland
WHITC White Clover Markets Seattle

Use GetCustomersFromCountry('Mexico') with the IN operator

OrderID OrderDate
----------- ----------
10259 1996-07-18
10276 1996-08-08
10293 1996-08-29
10304 1996-09-12
10308 1996-09-18
10319 1996-10-02
10322 1996-10-04
10354 1996-11-14
10365 1996-11-27
10474 1997-03-13
10502 1997-04-10
10507 1997-04-15
10518 1997-04-25
10535 1997-05-13
10573 1997-06-19
10576 1997-06-23
10625 1997-08-08
10676 1997-09-22
10677 1997-09-22
10682 1997-09-25
10759 1997-11-28
10842 1998-01-20
10856 1998-01-28
10915 1998-02-27
10926 1998-03-04
10995 1998-04-02
11069 1998-05-04
11073 1998-05-05

Joins OrdersByValue to Customers

CompanyName OrderID TotalValue OrderDate

---------------------------------- ----------- ------------- ----------
Simons bistro 10417 11188.4000 1997-01-16
Rattlesnake Canyon Grocery 10479 10495.6000 1997-03-19
QUICK-Stop 10540 10191.7000 1997-05-19
QUICK-Stop 10691 10164.8000 1997-10-03
Königlich Essen 10817 10952.8450 1998-01-06
QUICK-Stop 10865 16387.5000 1998-02-02
Rattlesnake Canyon Grocery 10889 11380.0000 1998-02-16
Hungry Owl All-Night Grocers 10897 10835.2400 1998-02-19
Hanari Carnes 10981 15810.0000 1998-03-27
Save-a-lot Markets 11030 12615.0500 1998-04-17

Joins TopTenOrders to Customers

CompanyName OrderID TotalValue OrderDate

---------------------------------- ----------- ------------- ----------
QUICK-Stop 10865 16387.5000 1998-02-02
Hanari Carnes 10981 15810.0000 1998-03-27

398
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

Save-a-lot Markets 11030 12615.0500 1998-04-17

Rattlesnake Canyon Grocery 10889 11380.0000 1998-02-16
Simons bistro 10417 11188.4000 1997-01-16
Königlich Essen 10817 10952.8450 1998-01-06
Hungry Owl All-Night Grocers 10897 10835.2400 1998-02-19
Rattlesnake Canyon Grocery 10479 10495.6000 1997-03-19
QUICK-Stop 10540 10191.7000 1997-05-19
QUICK-Stop 10691 10164.8000 1997-10-03
You can update, insert, or delete rows in tables through inline user-defined functions with the same limits as
updating, inserting, or deleting rows in tables through views. Listing 10.17 shows an example of how to
insert, update, and delete rows in the Orders table through the GetOrdersWithValue inline user-defined
function.

Listing 10.17 Modify Data Through Inline User-Defined Functions

USE Northwind
GO

PRINT CHAR(10) + 'Before the Insert'+ CHAR(10)

SELECT CustomerID, CompanyName, Country

FROM dbo.GetCustomersFromUsa()
WHERE CustomerID > 'W'

INSERT dbo.GetCustomersFromUSA() (CustomerID, CompanyName, Country)

VALUES ('ZZZZZ', 'Dummy Customer', 'USA')

PRINT CHAR(10) + 'After the Insert'+ CHAR(10)

SELECT CustomerID, CompanyName, Country

FROM dbo.GetCustomersFromUsa()
WHERE CustomerID > 'W'
UPDATE dbo.GetCustomersFromUSA()
SET CompanyName = 'New Customer'
WHERE CustomerID = 'ZZZZZ'

PRINT CHAR(10) + 'After the Update'+ CHAR(10)

SELECT CustomerID, CompanyName, Country

FROM dbo.GetCustomersFromUsa()
WHERE CustomerID > 'W'

DELETE dbo.GetCustomersFromUSA()
WHERE CustomerID = 'ZZZZZ'

PRINT CHAR(10) + 'After the Delete'+ CHAR(10)

SELECT CustomerID, CompanyName, Country

FROM dbo.GetCustomersFromUsa()
WHERE CustomerID > 'W'
GO

399
Microsoft SQL Server 2000 Programming by Example

DECLARE @ID int

PRINT CHAR(10) + 'Before the Insert'+ CHAR(10)

SELECT CustomerID, OrderID, TotalValue, OrderDate

FROM dbo.OrdersWithValue()
WHERE CustomerID = 'VINET'

INSERT dbo.OrdersWithValue() (CustomerID, OrderDate)

SELECT 'VINET', dbo.Today()

-- Retrieve the latest Identity value in this session

SET @ID = SCOPE_IDENTITY()

INSERT [Order Details]

(OrderID, ProductID, Quantity, UnitPrice, Discount)
SELECT @ID, 28, 10, UnitPrice, 0.1
FROM Products
WHERE ProductID = 28

PRINT CHAR(10) + 'After the Insert'+ CHAR(10)

SELECT CustomerID, OrderID, TotalValue, OrderDate

FROM dbo.OrdersWithValue()
WHERE CustomerID = 'VINET'
DELETE [Order Details]
WHERE OrderID = @ID

DELETE Orders
WHERE OrderID = @ID

PRINT CHAR(10) + 'After the Delete'+ CHAR(10)

SELECT CustomerID, OrderID, TotalValue, OrderDate

FROM dbo.OrdersWithValue()
WHERE CustomerID = 'VINET'

Before the Insert

CustomerID CompanyName Country

---------- ---------------------------------------- ---------------
WHITC White Clover Markets USA

After the Insert

CustomerID CompanyName Country

---------- ---------------------------------------- ---------------
WHITC White Clover Markets USA
ZZZZZ Dummy Customer USA

400
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

After the Update

CustomerID CompanyName Country

---------- ---------------------------------------- ---------------
WHITC White Clover Markets USA
ZZZZZ New Customer USA

After the Delete

CustomerID CompanyName Country

---------- ---------------------------------------- ---------------
WHITC White Clover Markets USA

Before the Insert

CustomerID OrderID TotalValue OrderDate

---------- ----------- --------------------- -----------------------
VINET 10248 440.0000 1996-07-04 00:00:00.000
VINET 10274 538.6000 1996-08-06 00:00:00.000
VINET 10295 121.6000 1996-09-02 00:00:00.000
VINET 10737 139.8000 1997-11-11 00:00:00.000
VINET 10739 240.0000 1997-11-12 00:00:00.000

After the Insert

CustomerID OrderID TotalValue OrderDate

---------- ----------- --------------------- -----------------------
VINET 10248 440.0000 1996-07-04 00:00:00.000
VINET 10274 538.6000 1996-08-06 00:00:00.000
VINET 10295 121.6000 1996-09-02 00:00:00.000
VINET 10737 139.8000 1997-11-11 00:00:00.000
VINET 10739 240.0000 1997-11-12 00:00:00.000
VINET 11118 410.4000 2000-12-26 00:00:00.000

After the Delete

CustomerID OrderID TotalValue OrderDate

---------- ----------- --------------------- -----------------------
VINET 10248 440.0000 1996-07-04 00:00:00.000
VINET 10274 538.6000 1996-08-06 00:00:00.000
VINET 10295 121.6000 1996-09-02 00:00:00.000
VINET 10737 139.8000 1997-11-11 00:00:00.000
VINET 10739 240.0000 1997-11-12 00:00:00.000
Note

You cannot delete data from the OrdersWithValue function, because it joins multiple tables.

Table-Valued Functions

401
Microsoft SQL Server 2000 Programming by Example

Multistatement table-valued user-defined functions (or table-valued functions) are similar to inline user-defined
functions. Actually, you can use them in the same scenarios, but you are not restricted to defining them as a
single SELECT statement.
A table-valued function returns a single result set— the contents of a table variable with a predefined format
specified in the function declaration.
Caution

The result of a table-valued function is always read-only. The result of an inline user-defined
function can be read-only or not, depending on the SELECT statement used to define the function.

As mentioned in the preceding section, inline user-defined functions are similar to views in both structure and
use. Table-valued functions are similar to stored procedures, but they return a single result set.

You can define parameters in a table-valued function, as you did with scalar user-defined functions and stored
procedures. However, you cannot create an OUTPUT parameter in a user-defined function.

Creating Table-Valued Functions

To create a table-valued function, you must use the CREATE FUNCTION statement. Listing 10.18 shows
functions similar to those in Listing 10.14, but in table-valued versions.

Listing 10.18 Table-Valued Version of the Inline User-Defined Functions Examples from Listing
10.14

USE Northwind
GO

CREATE FUNCTION dbo.tv_GetCustomersFromCountry

(@country nvarchar(15))
RETURNS @List TABLE
(CustomerID nchar(5),
CompanyName nvarchar(40),
Country nvarchar(15))
AS
BEGIN

INSERT @List
SELECT CustomerID, CompanyName, Country
FROM Customers
WHERE Country = @country

RETURN
END
GO

402
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

-- Returns USA-based customers

CREATE FUNCTION dbo.tv_GetCustomersFromUSA

()
RETURNS @List TABLE
(CustomerID nchar(5),
CompanyName nvarchar(40),
Country nvarchar(15))
AS
BEGIN
INSERT @List
SELECT CustomerID, CompanyName, Country
FROM dbo.GetCustomersFromCountry(N'USA')

RETURN
END
GO

-- Returns the orders from a specific day

CREATE FUNCTION dbo.tv_GetOrdersFromDay

(@date as smalldatetime)
RETURNS @list TABLE
(OrderID int,
OrderDate datetime)
AS
BEGIN

INSERT @List
SELECT OrderID, OrderDate
FROM Orders
WHERE DATEDIFF(day, OrderDate, @date) = 0

RETURN
END
GO

-- Returns orders from today

CREATE FUNCTION dbo.tv_GetOrdersFromToday

()
RETURNS @list TABLE
(OrderID int,
OrderDate datetime)
AS
BEGIN

INSERT @List
SELECT OrderID, OrderDate
FROM Orders
WHERE DATEDIFF(day, OrderDate, dbo.Today()) = 0

RETURN
END
GO
-- Returns Orders with the total order value

CREATE FUNCTION dbo.tv_OrdersWithValue

()
RETURNS @list TABLE

403
Microsoft SQL Server 2000 Programming by Example

(OrderID int,
CustomerID nchar(5),
OrderDate datetime,
TotalValue money)
AS
BEGIN

INSERT @List
SELECT O.OrderID, CustomerID, OrderDate, TotalValue
FROM Orders O
JOIN (
SELECT OrderID, SUM(dbo.TotalPrice(Quantity, UnitPrice, Discount))
AS TotalValue
FROM [Order Details]
GROUP BY OrderID) AS OD
ON O.OrderID = OD.OrderID

RETURN
END
GO

-- Returns orders with a value greater than

-- a specific target value

CREATE FUNCTION dbo.tv_OrdersByValue

(@total money)
RETURNS @list TABLE
(OrderID int,
CustomerID nchar(5),
OrderDate datetime,
TotalValue money)
AS
BEGIN

INSERT @List
SELECT OrderID, CustomerID, OrderDate, TotalValue
FROM dbo.OrdersWithValue()
WHERE TotalValue > @total

RETURN
END
GO
-- Returns the top 10 orders by total value

CREATE FUNCTION dbo.tv_TopTenOrders

()
RETURNS @list TABLE
(OrderID int,
CustomerID nchar(5),
OrderDate datetime,
TotalValue money)
AS
BEGIN

INSERT @List
SELECT TOP 10 WITH TIES
OrderID, CustomerID, OrderDate, TotalValue
FROM dbo.OrdersWithValue()
ORDER BY TotalValue DESC

RETURN

404
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

END
GO
You can identify several parts in the CREATE FUNCTION syntax for table- valued functions:

• CREATE FUNCTION ownername.functionname, where you can specify the owner of the function
and the name of the function.
• (parameter_lists), where every parameter is identified by its name and the data type. The
parameter list can be empty. Parameters can have default values.
• RETURNS @tablename TABLE, because table-valued functions always return a tablelike result set,
which is the contents of the table variable defined in this line as @tablename.
• (Field definition, ...), definition of every field on the @tablename table variable, following
the same rules as declaring fields in a standard table object.
• AS BEGIN...END marks the function definition body, which must be written inside the BEGIN END
block.
• The RETURN statement, standalone, which sends the table variable back to the invoking process as a
result set.

The contents of a table-valued function are similar to a stored procedure that uses a temporary table to store
intermediate results and select from that temporary table at the end.
Tip

Inside scalar or table-valued user-defined functions you cannot create temporary tables; instead,
you can use table variables. However, it is a good practice to replace temporary tables with table
variables in stored procedures and scripts whenever this might be possible.

Looking at the examples in Listing 10.18, you could say that table-valued functions are similar to inline user-
defined functions, except that you must declare the fields of the returning table explicitly. However, table-
valued functions give you far more flexibility than inline user-defined functions, in a similar way as stored
procedures give you more flexibility than views.

Listing 10.19 shows two more complex examples in which the creation of a table-valued function solves the
problem.

The first function is called MakeList, and it converts a string with several items into a single column result
set where every item is stored in a single row. You can specify any separator to divide values, or use the
default '|'. This function can be useful in a WHERE clause introduced with the IN operator.

The second function is a bit more complex. It produces a subset of a fact table from the Order Details
table, adding some extra information as the CategoryName, ProductName, CompanyName, and
OrderDate from the Categories, Products, Customers, and Orders tables, and the TotalValue
using the TotalPrice scalar user-defined function.

To call this function, you can specify whether you want a full list (@Key IS NULL), a list for a specific order
(@Key = 'ORD' and @ID = OrderID), a specific Customer (@Key = 'CUS' and @ID = CustomerID), a
product (@Key = 'PRO' and ID = ProductID), or a category (@Key = 'CAT' and @ID = CategoryID).

405
Microsoft SQL Server 2000 Programming by Example

At the beginning of the function, you can find a sequence of IF structures to apply the most efficient method
to every case to prepare the initial list of rows to return.

At the end of the OrderDetailsComplete function, fill the missing information with values coming from
other tables.

Listing 10.19 Using Table-Valued Functions to Execute Complex Result Sets

USE Northwind
GO

-- Converts a string containing a list of items

-- into a single column table where every item
-- is in a separate row
-- using any character as a separator

CREATE FUNCTION dbo.MakeList

(@ParamArray as nvarchar(4000), @Separator as char(1) = '|')
RETURNS @List TABLE
(Item sql_variant)
AS
BEGIN

DECLARE @pos int, @pos0 int

SET @pos0 = 0

WHILE 1=1
BEGIN

SET @pos = CHARINDEX(@Separator, @ParamArray, @pos0 + 1)

INSERT @List
SELECT CASE @pos
WHEN 0 THEN SUBSTRING(@ParamArray, @pos0+1,
LEN(@ParamArray) - @pos -1)
ELSE SUBSTRING(@ParamArray, @pos0+1, @pos - @pos0-1)
END

IF @pos = 0 BREAK

SET @pos0 = @pos

END

RETURN

END
GO

406
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

-- Produces a list of orders

-- with full descriptive information
-- ProductName, CategoryName, CompanyName
-- OrderDate and TotalValue
-- with every primary key to link to other tables
-- The list can be produced for every
-- Order (@Key = 'ORD'),
-- Product (@Key = 'PRO'),
-- Customer (@Key = 'CUS'),
-- Category (@Key = 'CAT')
-- Full List (@Key NOT IN ('ORD', 'PRO', 'CUS', 'CAT'))
CREATE FUNCTION dbo.OrderDetailsComplete
(@ID sql_variant = NULL,
@Key char(3) = NULL)
RETURNS @Details TABLE
(OrderID int,
ProductID int,
CustomerID nchar(5) NULL,
CategoryID int NULL,
OrderDate smalldatetime NULL,
Value money NULL,
Category nvarchar(15) NULL,
Product nvarchar(40) NULL,
Company nvarchar(40) NULL)
AS
BEGIN

IF @Key = 'ORD'
BEGIN

INSERT @Details
(OrderID, ProductID, Value)
SELECT OrderID, ProductID,
dbo.TotalPrice(Quantity, UnitPrice, Discount)
FROM [Order Details]
WHERE OrderID = @ID

END
ELSE IF @Key = 'PRO'
BEGIN

INSERT @Details (OrderID, ProductID, Value)

SELECT OrderID, ProductID,
dbo.TotalPrice(Quantity, UnitPrice, Discount)
FROM [Order Details]
WHERE ProductID = @ID

END
ELSE IF @Key = 'CUS'
BEGIN

INSERT @Details (OrderID, ProductID, CustomerID, Value)

SELECT O.OrderID, ProductID, CustomerID,
dbo.TotalPrice(Quantity, UnitPrice, Discount)
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
WHERE CustomerID = @ID
END
ELSE IF @Key = 'CAT'
BEGIN

407
Microsoft SQL Server 2000 Programming by Example

INSERT @Details (OrderID, ProductID, CategoryID, Value)

SELECT OD.OrderID, P.ProductID, CategoryID,
dbo.TotalPrice(Quantity, OD.UnitPrice, Discount)
FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
WHERE CategoryID = @ID

END
ELSE
BEGIN

INSERT @Details
(OrderID, ProductID, Value)
SELECT OrderID, ProductID,
dbo.TotalPrice(Quantity, UnitPrice, Discount)
FROM [Order Details]

END

UPDATE D
SET D.CustomerID = O.CustomerID,
D.OrderDate = O.OrderDate
FROM @Details D
JOIN Orders O
ON O.OrderID = D.OrderID
WHERE D.CustomerID IS NULL

UPDATE D
SET D.CategoryID = P.CategoryID,
D.Product = P.ProductName
FROM @Details D
JOIN Products P
ON P.ProductID = D.ProductID
WHERE D.CategoryID IS NULL

UPDATE D
SET D.Category = C.CategoryName
FROM @Details D
JOIN Categories C
ON C.CategoryID = D.CategoryID

UPDATE D
SET D.Company = C.CompanyName
FROM @Details D
JOIN Customers C
ON C.CustomerID = D.CustomerID

RETURN

END
GO

Invoking Table-Valued User-Defined Functions

You can invoke table-value functions the same way you do for inline user-defined functions. Listing 10.20
shows examples of how to call some functions defined in Listing 10.18.
Note

408
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

Listing 10. 20 is a revised version of Listing 10.16, calling table-value functions instead of inline
user-defined functions. The output is the same as in Listing 10.16; therefore, it is not necessary
to show it again.

Listing 10.20 Calling Table-Value Functions

USE Northwind
GO

PRINT CHAR(10) + 'Use tv_GetCustomersFromCountry(''Mexico'')'+ CHAR(10)

SELECT CustomerID, CompanyName, Country

FROM dbo.tv_GetCustomersFromCountry('Mexico')

PRINT CHAR(10) + 'Use tv_GetCustomersFromUSA()'+ CHAR(10)

Select CustomerID, CompanyName, Country

FROM dbo.tv_GetCustomersFromUSA()

PRINT CHAR(10)
+ 'Use tv_GetCustomersFromCountry(''Mexico'') with the IN operator'
+ CHAR(10)

SELECT OrderID,
CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM Orders
WHERE CustomerID IN
(SELECT CustomerID
FROM dbo.tv_GetCustomersFromCountry('Mexico'))
PRINT CHAR(10)
+ 'Joins tv_OrdersByValue to Customers'
+ CHAR(10)

SELECT CompanyName, OrderID, TotalValue,

CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM dbo.tv_OrdersByValue(10000) AS OBV
JOIN Customers C
ON OBV.CustomerID = C.CustomerID

PRINT CHAR(10)
+ 'Joins tv_TopTenOrders to Customers'
+ CHAR(10)

SELECT CompanyName, OrderID, TotalValue,

CONVERT(varchar(10), OrderDate, 120) AS OrderDate
FROM dbo.tv_TopTenOrders() AS OBV
JOIN Customers C
ON OBV.CustomerID = C.CustomerID

409
Microsoft SQL Server 2000 Programming by Example

However, the way SQL Server executes table-valued functions and inline user-defined functions is completely
different:

• The definition of inline user-defined functions is merged with the definition of the outer query,
producing a single query plan, in which you cannot find any traces of the user-defined function.
Figure 10.1 shows the query plan of this query, in which you can see an Index Scan on the
Customers table's clustered index. Every time you use an inline user-defined function its definition
must be merged again with the outer query, producing a new query plan.

Figure 10.1. Query plan of a query that uses an inline user-defined function.

• The first time you call a table-valued function, it is compiled and a query plan is placed in memory.
Every time you use this function, the calling query forces the execution of the saved query plan and
the function returns the table variable, which will be used in the outer query. Figure 10.2 shows the
query plan of this execution, simpler than the one from Figure 10.1, in which you can see a Table
Scan on the table variable returned from the function. In this case, most of the work is done in a
separate query plan, corresponding to the table- valued function.

Figure 10.2. Query plan of a query that uses a table-valued user-defined function.

Dropping User-Defined Functions

To drop a user-defined function, ,you must use the DROP FUNCTION statement. Listing 10.21 shows a
simple example of this statement.

Listing 10.21 Use DROP FUNCTION to Delete a User-Defined Function

410
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

USE Northwind
GO

DROP FUNCTION dbo.Today

Caution

Before dropping a user-defined function, as with any other database object, check its
dependencies.

You cannot drop a user-defined function if it is used in a constraint definition. If you drop a user-
defined function and it is used in other functions, views, triggers, or stored procedures, those
functions will produce an error on next execution.

Preventing the Alteration of Dependent Objects:The SCHEMABINDING

Option
You can prevent changes on the dependent objects of a user-defined function by using the SCHEMABINDING
option. Using this option, you cannot modify the definition of the dependent objects using any of the ALTER
statements, and you cannot drop dependent objects using any of the DROP statements. This link disappears
when the function is dropped or when you alter the function definition without using the SCHEMABINDING
option.
To use this option, you must ensure that the following conditions are met:

• Every function and view referenced in the function must be defined as SCHEMABINDING as well.
• Every object referenced in the function must be referenced using two-part names
(owner.objectname).
• Every object referenced in the function belongs to the same database as the function.
• The user who creates the function (not necessarily the owner) has REFERENCES permissions on
every object referenced inside the function. It is recommended that only members of the db_owner
role execute the CREATE FUNCTION statement.

Listing 10.22 shows how to use the SCHEMABINDING option and the effect when you try to modify a
dependent object. The process is as follows:

1. You create the NewCustomers table with data coming from the Customers table.
2. You create the GetCustomers table-valued function, reading the CustomerID and CompanyName
fields from the NewCustomers table.
3. You try to alter the NewCustomers table, dropping the CompanyName column, and it is successful
because the GetCustomers function was created without the SCHEMABINDING option.
4. Trying to use the GetCustomers function, you get error message 207 because the column
CompanyName does not exist.
5. You start all over, with the creation of the NewCustomers table.
6. Create the GetCustomers function with the SCHEMABINDING option, and use the NewCustomers
table without specifying the owner, and you get error 4512, because to use SCHEMABINDING you
must use two part names.
7. Create the GetCustomers function with the SCHEMABINDING option and use two-part names this
time. The operation succeeds.
8. Try to alter the NewCustomers table, dropping the CompanyName column. You get errors 5074 and
4922 because the function is created with the SCHEMABINDING option.

Listing 10.22 Effect of SCHEMABINDING on the Dependent Objects

411
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

IF OBJECT_ID('GetCustomers') IS NOT NULL

DROP FUNCTION GetCustomers
GO

IF OBJECT_ID('NewCustomers') IS NOT NULL

DROP TABLE NewCustomers
GO

SELECT *
INTO NewCustomers
FROM Customers
GO

CREATE FUNCTION dbo.GetCustomers()

RETURNS @List TABLE
(CustomerID nchar(5),
CompanyName nvarchar(40))
AS
BEGIN

INSERT @List
SELECT CustomerID, CompanyName
FROM NewCustomers

RETURN
END
GO

ALTER TABLE NewCustomers

DROP COLUMN CompanyName

PRINT CHAR(10)
+ 'ALTER TABLE statement successful without SCHEMABINDING'
+ CHAR(10)
GO

SELECT *
FROM GetCustomers()
GO

PRINT CHAR(10)
+ 'Execution of the GetCustomers table was unsuccessful'
+ CHAR(10)
+ 'because it references a non-existing field'
+ CHAR(10)
GO

IF OBJECT_ID('GetCustomers') IS NOT NULL

DROP FUNCTION GetCustomers
GO

412
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

IF OBJECT_ID('NewCustomers') IS NOT NULL

DROP TABLE NewCustomers
GO

SELECT *
INTO NewCustomers
FROM Customers
GO

CREATE FUNCTION dbo.GetCustomers()

RETURNS @List TABLE
(CustomerID nchar(5),
CompanyName nvarchar(40))
WITH SCHEMABINDING
AS
BEGIN

INSERT @List
SELECT CustomerID, CompanyName
FROM NewCustomers

RETURN
END
GO

PRINT CHAR(10)
+ 'CREATE FUNCTION failed with SCHEMABINDING'
+ CHAR(10)
+ 'because it did not use two part names'
+ CHAR(10)
GO

CREATE FUNCTION dbo.GetCustomers()

RETURNS @List TABLE
(CustomerID nchar(5),
CompanyName nvarchar(40))
WITH SCHEMABINDING
AS
BEGIN

INSERT @List
SELECT CustomerID, CompanyName
FROM dbo.NewCustomers
RETURN
END
GO

PRINT CHAR(10)
+ 'CREATE FUNCTION was successful with SCHEMABINDING'
+ CHAR(10)
+ 'because it did use two part names'
+ CHAR(10)
GO

ALTER TABLE NewCustomers

DROP COLUMN CompanyName
GO

PRINT CHAR(10)
+ 'ALTER TABLE statement failed with SCHEMABINDING'

413
Microsoft SQL Server 2000 Programming by Example

+ CHAR(10)
GO

(91 row(s) affected)

ALTER TABLE statement successful without SCHEMABINDING

Server: Msg 207, Level 16, State 3, Procedure GetCustomers, Line 12

Invalid column name 'CompanyName'.

Execution of the GetCustomers table was unsuccessful

because it references a non-existing field

(91 row(s) affected)

Server: Msg 4512, Level 16, State 3, Procedure GetCustomers, Line 14

Cannot schema bind function 'dbo.GetCustomers'because name NewCustomers'is
invalid for
schema binding. Names must be in two-part format and an object cannot reference
itself.

CREATE FUNCTION failed with SCHEMABINDING

because it did not use two part names

CREATE FUNCTION was successful with SCHEMABINDING

because it did use two part names
Server: Msg 5074, Level 16, State 3, Line 2
The object 'GetCustomers'is dependent on column 'CompanyName'.
Server: Msg 4922, Level 16, State 1, Line 2
ALTER TABLE DROP COLUMN CompanyName failed because one or more objects access
this column.

ALTER TABLE statement failed with SCHEMABINDING

Deterministic and Nondeterministic Functions

Some functions always return the same value when called with the same set of arguments. These functions
are called deterministic. This is important if you want to create a clustered index on a view or any index on a
computed column, because you can create these indexes only if they use deterministic functions.
Most of the built-in functions are deterministic, such as
ABS DATEDIFF PARSENAME
ACOS DAY POWER
ASIN DEGREES RADIANS
ATAN EXP ROUND
ATN2 FLOOR SIGN
CEILING ISNULL SIN
COALESCE ISNUMERIC SQUARE

414
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

COS LOG SQRT

COT LOG10 TAN
DATALENGTH MONTH YEAR
DATEADD NULLIF
Some built-in functions are deterministic or nondeterministic, depending on the way you use them:

• CAST is deterministic for every type of value except for conversion from datetime,
smalldatetime, and sql_variant containing a date value, because the final results depend on
regional settings.
• CONVERT is deterministic in the same cases as CAST and nondeterministic in the same cases as
CAST, except if you specify a style when converting datetime and smalldatetime data, the result
is always predictable and the function is deterministic in that case.
• CHECKSUM is deterministic if you specify the list of columns or an expression; it is nondeterministic if
you specify CHECKSUM(*).
• ISDATE is nondeterministic unless it is used with CONVERT and with a predictable style different from
0, 100, 9, or 109.
• RAND is deterministic if a seed value is specified; it is nondeterministic without a seed value.

Most of the other built-in functions are nondeterministic. For a full list, you can search for the "Deterministic
and Nondeterministic Functions" topic in Books Online.
A user-defined function is deterministic only if

• Every function— built-in or user-defined— referenced in the function is deterministic.

• The function is defined with the SCHEMABINDING option.
• The function does not references objects not defined inside the function itself, such as tables, views,
extended stored procedures.

Note

Creating a nondeterministic user-defined function is fine, as long as you are aware of their
limitations. Books Online incorrectly says that you cannot use built-in nondeterministic functions
inside a user-defined function. The only functions you cannot use inside a user-defined function are
contained in the list following this note.

Built-in functions that use the current time are not valid inside a user-defined function:

CURRENT_TIMESTAMP GETDATE
GetUTCDate IDENTITY
NEWID TEXTPTR
@@DBTS @@MAX_CONNECTIONS
Other functions that are not valid inside user-defined functions are the System Statistical functions:

@@CONNECTIONS @@PACK_RECEIVED
@@CPU_BUSY @@PACK_SENT
fn_virtualfilestats @@TIMETICKS
@@IDLE @@TOTAL_ERRORS
@@IO_BUSY @@TOTAL_READ

415
Microsoft SQL Server 2000 Programming by Example

@@PACKET_ERRORS @@TOTAL_WRITE

Altering User-Defined Functions Definition

To modify the definition of a user-defined function, you can use the ALTER FUNCTION statement in exactly
the same way you use the CREATE FUNCTION statement. In this case, the new definition replaces the old
definition of the user-defined function with the same name.
Listing 10.23 shows an example of how to use the ALTER FUNCTION statement to modify a preexisting
user-defined function and encrypt the definition with the WITH ENCRYPTION option.

Listing 10.23 Use ALTER FUNCTION to Modify a User-Defined Function

USE Northwind
GO

-- Returns the maximum ProductID from Products

ALTER FUNCTION dbo.MaxProductID

()
RETURNS int
WITH ENCRYPTION
AS
BEGIN

RETURN (
SELECT MAX(ProductID)
FROM dbo.Products
)

END
GOa user-defined function, you can use the ALTER FUNCTION statement
Caution

Not using the SCHEMABINDING option when you execute the ALTER FUNCTION statement
unbinds the dependent objects from the function.

Caution

Before encrypting a user-defined function definition, make sure you have a copy in a safe place,
because it will be impossible to decrypt it.

Security Implications of Using User-Defined Functions

416
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

You can grant or deny the permissions to use user-defined functions depending on the type of function:

• For scalar user-defined functions, you can grant or deny permissions on EXECUTE and REFERENCES.
• For inline user-defined functions, you can grant or deny permissions on SELECT, UPDATE, INSERT,
DELETE, or REFERENCES.
• For multistatement table-values user-defined functions, you can grant or deny permissions to SELECT
and REFERENCES.

As in stored procedures and views, if every object referenced in a user-defined function belongs to the same
owner as the user-defined function, and a user tries to use the function, permissions will be checked only on
the function, not on every object referenced in the function.

Applying User-Defined Functions

You convert commonly used formulas into scalar user-defined functions. In this case, the function's compiled
query plan remains in memory, as does any built-in function.
You can call user-defined functions from inside other user-defined functions, but only up to 32 levels, and this
limit applies to the total of stored procedures, triggers, and scalar or table-valued user-defined functions you
use.
Note

Use the @@NESTLEVEL system function to know how many nested levels you are using.

A good approach would be to create user-defined functions in a short number of layers, so the limit for nesting
levels will never be surpassed. This contributes to the clarity of your database design as well.

Be aware that modifying underlying objects could affect the result of a user-defining function, unless you
create the function with the SCHEMABINDING option.

This is still a new feature for Transact-SQL programmers, but client- application programmers will find user-
defined functions very close to their normal programming methods.

Converting Stored Procedures into User-Defined Functions

If the only reason for a stored procedure is to supply an output parameter, you can create a scalar user-
defined function instead. In this way, you can use this function in a more natural way than a stored procedure.
Listing 10.24 shows an example of converting the fn_FV function into the sp_FV stored procedure and how
to call them.

Listing 10.24 Comparing a Stored Procedure with a Scalar User-Defined Function

417
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

-- sp_fv with the same functionality

-- as the fn_fv function

CREATE PROCEDURE sp_fv

@rate float, @nper int, @pmt money,
@pv money = 0.0, @type bit = 0,
@FV money output
AS

IF @rate = 0
SET @fv = @pv + @pmt * @nper
ELSE
SET @fv = @pv * POWER(1 + @rate, @nper) +
@pmt * (((POWER(1 + @rate, @nper + @type) - 1) / @rate) - @type)

SET @fv = -@fv

GO

-- Call the sp_fv stored procedure

DECLARE @fv money

EXECUTE sp_fv 0.10, 24, 1000, 10000, 0, @fv OUTPUT

SELECT @fv 'From sp_fv'

GO

-- Call the sp_fv stored procedure

SELECT dbo.fn_fv(0.10, 24, 1000, 10000, 0) as 'From fn_fv'

GO

From sp_fv
---------------------
-186994.6535

From fn_fv
---------------------
-186994.6535
If a stored procedure returns a single read-only result set, you can convert it into a table-valued function with a
similar code, and you can use the function in the FROM clause of any DML statement, providing a better
programming flexibility. Listing 10.25 shows an example of a stored procedure with the same functionality as
the tv_TopTenOrders and how to call them.
If you have a stored procedure that provides read/write access to a table through a client library, you can
convert this procedure into an inline user-defined function.

Listing 10.25 Comparing Stored Procedures and Table-Valued User-Defined Functions

418
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

USE Northwind
GO

CREATE PROCEDURE sp_TopTenOrders

AS

DECLARE @list TABLE

(OrderID int,
CustomerID nchar(5),
OrderDate datetime,
TotalValue money)

INSERT @List
SELECT O.OrderID, CustomerID, OrderDate, TotalValue
FROM Orders O
JOIN (
SELECT OrderID, SUM(dbo.TotalPrice(Quantity, UnitPrice, Discount))
AS TotalValue
FROM [Order Details]
GROUP BY OrderID) AS OD
ON O.OrderID = OD.OrderID

SELECT TOP 10 WITH TIES

OrderID, CustomerID, OrderDate, TotalValue
FROM @List
ORDER BY TotalValue DESC
GO

EXECUTE sp_TopTenOrders
GO

SELECT *
FROM tv_TopTenOrders()
GO

OrderID CustomerID OrderDate TotalValue

----------- ---------- ----------------------------- ---------------------
10865 QUICK 1998-02-02 00:00:00.000 16387.5000
10981 HANAR 1998-03-27 00:00:00.000 15810.0000
11030 SAVEA 1998-04-17 00:00:00.000 12615.0500
10889 RATTC 1998-02-16 00:00:00.000 11380.0000
10417 SIMOB 1997-01-16 00:00:00.000 11188.4000
10817 KOENE 1998-01-06 00:00:00.000 10952.8450
10897 HUNGO 1998-02-19 00:00:00.000 10835.2400

419
Microsoft SQL Server 2000 Programming by Example

10479 RATTC 1997-03-19 00:00:00.000 10495.6000

10540 QUICK 1997-05-19 00:00:00.000 10191.7000
10691 QUICK 1997-10-03 00:00:00.000 10164.8000
OrderID CustomerID OrderDate TotalValue
----------- ---------- ----------------------------- ---------------------
10865 QUICK 1998-02-02 00:00:00.000 16387.5000
10981 HANAR 1998-03-27 00:00:00.000 15810.0000
11030 SAVEA 1998-04-17 00:00:00.000 12615.0500
10889 RATTC 1998-02-16 00:00:00.000 11380.0000
10417 SIMOB 1997-01-16 00:00:00.000 11188.4000
10817 KOENE 1998-01-06 00:00:00.000 10952.8450
10897 HUNGO 1998-02-19 00:00:00.000 10835.2400
10479 RATTC 1997-03-19 00:00:00.000 10495.6000
10540 QUICK 1997-05-19 00:00:00.000 10191.7000
10691 QUICK 1997-10-03 00:00:00.000 10164.8000

Converting Views into User-Defined Functions

You can convert views into inline user-defined functions very easily, but in this case, the only benefit you will
get is the possibility of having parameters. However, if you use a view to read data only, you will benefit from
converting this view into a table-valued function because it will be optimized and compiled on the first
execution, providing performance gains over a view.
Listing 10.26 shows the fv_TopTenOrders converted into a view, and how you call the view and the user-
defined function. The output is the same as the one for Listing 10.25.

Listing 10.26 Comparing Views and Table-Valued User-Defined Functions

USE Northwind
GO

CREATE VIEW vw_TopTenOrders

AS

SELECT TOP 10 WITH TIES

O.OrderID, CustomerID, OrderDate, TotalValue
FROM Orders O
JOIN (
SELECT OrderID, SUM(dbo.TotalPrice(Quantity, UnitPrice, Discount))
AS TotalValue
FROM [Order Details]
GROUP BY OrderID) AS OD
ON O.OrderID = OD.OrderID
ORDER BY TotalValue DESC

GO

SELECT *
FROM tv_TopTenOrders()
GO

SELECT *

420
 Chapter 10. Enhancing Business Logic: User-Defined Functions (UDF)

FROM vw_TopTenOrders
GO

Using User-Defined Functions in Constraints

You can use a scalar user-defined function anywhere an expression is allowed, and that includes

• DEFAULT constraints
• CHECK constraints
• DEFAULT objects
• RULE objects
• A PRIMARY KEY constraint defined in a computed column using a user-defined function, as long as
the returned values are unique
• A UNIQUE constraint defined in a computed column with a user-defined function, as long as the
returned values are unique

Therefore, it is possible to access values from other tables from inside a constraint, as long as the constraint
uses a user-defined function that searches for external data to produce its result.
The only place where you can use a table-valued user-defined function or an inline user-defined function is as
a subquery in a CHECK constraint but, unfortunately, CHECK constraints do not support subqueries.

What's Next?

This chapter covered the creation and use of user-defined functions— an exciting new feature that provides
extra programmability to the Transact-SQL language. The more you practice with user-defined functions, the
more you will wonder how you could have survived without them before SQL Server 2000 offered this feature.
Chapter 11 teaches you how to write complex queries, and in some cases, using user-defined functions that
could solve similar situations with less complexity.
In Chapter 12, you learn how to work with result sets row by row, using cursors. You can use cursors inside
user-defined functions to achieve complex operations that are impossible using rowset-oriented programming.

421
 Chapter 11. Using Complex Queries and Statements

Chapter 11. Using Complex Queries and Statements

In previous chapters, you learned how to execute queries to retrieve and modify data in SQL Server. You also
learned how to create and use database objects, such as tables, views, stored procedures, user-defined
functions, and triggers. Transact-SQL provides extended structures that can simplify the process of writing
queries to solve complex requests.
This chapter teaches you the following:

• How to create subqueries, which are queries inside other queries, to solve complex problems
• How to use the EXISTS keyword to test for existence of rows in a subquery
• How to use the IN operator to check for values returned from a subquery
• How to use derived tables, which are subqueries that can be used as virtual tables in the FROM clause,
to simplify complex queries
• How to use the CASE function to retrieve conditional values
• How to produce summary reports using the COMPUTE clause
• How to produce summary result sets using the CUBE and ROLLUP operators
• How to use optimizer hints to modify the way the query will be processed

Subqueries
A subquery is just a query contained inside another query. You can call the subquery an inner query
contained within an outer query, which in turn can be a standard query or another subquery.
If you think about standard queries, you can define three kinds of queries, according to the type of result they
provide:

• Scalar— Queries that produce a single value (one single row with only one column)
• List— Queries that produce a list of values (one or more rows with a single column only)
• Array— Queries that return a result set (one or more rows with one or more columns)

List queries can be considered single-column array queries. Scalar queries can be used as single-column,
single-row array queries as well.
Listing 11.1 shows different scalar queries that return a single value. This value can be a single constant,
the result of a system function, or the result of a standard query, as long as the query returns a single column
and a single row.

Listing 11.1 Scalar Queries Return a Single Value

USE Northwind
GO

SET NOCOUNT ON
GO

-- Select a single constant

SELECT 1

-- Select a scalar system niladic function

423
Microsoft SQL Server 2000 Programming by Example

SELECT SYSTEM_USER

-- Select a scalar system function

SELECT db_ID('Northwind')

-- Select the result of a User-Defined Function

-- Note this function does not exist

-- SELECT fn_getProductNameFromID(123)

-- Select the result of an aggregate function applied to a number of rows

SELECT COUNT(*) as NRows

FROM Northwind.dbo.Products

-- Select a single column from a single row in a table

SELECT ProductName
FROM Northwind.dbo.Products
WHERE ProductID = 5

-----------

1.00

----------------------------------------------------------
SQLBYEXAMPLE\GuestUser

------
6.00

NRows
-----------
77.00

ProductName
----------------------------------------
Chef Anton's Gumbo Mix
In Listing 11.2, you can see three examples of queries that provide a list of values. In the first example, you
select values from a single column. In the second example, you aggregate data, grouping the results by
another field. In the third example, you create a list query by combining several scalar queries using the
UNION operator.

Listing 11.2 List Queries Return a List of Values

424
 Chapter 11. Using Complex Queries and Statements

USE Northwind
GO

SET NOCOUNT ON
GO

-- Selecting a single column from a table

SELECT CategoryName
FROM Northwind.dbo.Categories

-- Selecting aggregate values from a single column from a table using GROUP BY

SELECT COUNT(*) AS "Products per Supplier"

FROM Northwind.dbo.products
GROUP BY SupplierID

-- Selecting different constant values using the UNION operator

SELECT 1 AS "Numbers"
UNION
SELECT 2
UNION
SELECT 3

CategoryName
---------------
Beverages
Condiments
Confections
Dairy Products
Grains/Cereals
Liquors
Liquors
Meat/Poultry
Produce
Seafood

Products per Supplier

---------------------
3.00
4.00
3.00
3.00

425
Microsoft SQL Server 2000 Programming by Example

2.00
3.00
5.00
4.00
2.00
1.00
3.00
5.00
1.00
3.00
3.00
3.00
3.00
2.00
2.00
3.00
2.00
2.00
3.00
3.00
2.00
2.00
1.00
2.00
2.00

Numbers
-----------
1.00
2.00
3.00
Listing 11.3 shows several array query examples that return result sets with multiple columns. The first
example selects two columns from a table. The second example selects several constants. The third example
selects the results of several scalar system functions. The last example combines the results of two array
queries to produce a single array query, using the UNION operator.

Listing 11.3 Array Queries Return a Complete Result Set

USE Northwind
GO

SET NOCOUNT ON
GO

-- Selecting multiple columns from a table

SELECT ProductName, UnitPrice

FROM Northwind.dbo.Products
WHERE CategoryID = 1

-- Selecting multiple constants

426
 Chapter 11. Using Complex Queries and Statements

SELECT 1 AS 'Lower',
2 AS 'Higher',
'Peter'AS 'Responsible'

-- Selecting values from system functions

SELECT CURRENT_TIMESTAMP AS 'Now',
CURRENT_USER AS 'Database User',
SYSTEM_USER AS 'System Login'

-- Selecting data from multiple tables using the UNION operator

SELECT CompanyName, ContactName

FROM Northwind.dbo.Customers
WHERE Country = 'Brazil'
UNION
SELECT CompanyName, ContactName
FROM Northwind.dbo.Suppliers
WHERE Country = 'Brazil'

ProductName UnitPrice
---------------------------------------- ---------------------
Chai $18.00
Chang $19.00
Guaraná Fantástica $4.50
Sasquatch Ale $14.00
Steeleye Stout $18.00
Côte de Blaye $263.50
Chartreuse verte $18.00
Ipoh Coffee $46.00
Laughing Lumberjack Lager $14.00
Outback Lager $15.00
Rhönbräu Klosterbier $7.75
Lakkalikööri $18.00

Lower Higher Responsible

----------- ----------- -----------
1.00 2.00 Peter

Now Database User System Login

------------------------- ---------------- ----------------------------
1/21/2001 4:38:42 PM dbo SQLBYEXAMPLE\AdminUser

CompanyName ContactName
---------------------------------------- ------------------------------
Comércio Mineiro Pedro Afonso
Familia Arquibaldo Aria Cruz
Gourmet Lanchonetes André Fonseca
Hanari Carnes Mario Pontes
Que Delícia Bernardo Batista
Queen Cozinha Lúcia Carvalho
Refrescos Americanas LTDA Carlos Diaz
Ricardo Adocicados Janete Limeira

427
Microsoft SQL Server 2000 Programming by Example

Tradição Hipermercados Anabela Domingues

Wellington Importadora Paula Parente
Most of the queries that use subqueries can be rewritten as simple queries without subqueries to produce the
same results. Actually, the Query Optimizer can decide to apply the same query plan regardless of the way
the query is written.
In the following sections, you will see the same solution with and without subqueries. In some cases, using a
subquery makes the query easier to read.

Scalar Subqueries

A scalar query can be used as a subquery anywhere in a Transact-SQL statement where an expression is
accepted:

• As part of any expression, because the result of the subquery is a scalar value.
• In the SELECT clause of a SELECT statement, as part of the output list.
• In the SET clause of an UPDATE statement, specifying the value to assign to a field.
• In the FROM clause of a SELECT statement, as a single row and single column derived table.
• In the WHERE clause, as a condition to test the value of a field, constant, variable, or the result of
another scalar subquery.
• In the HAVING clause, in the same cases as in the WHERE clause.

Listing 11.4 shows several examples of how to use scalar subqueries in the SELECT, SET, FROM, WHERE,
and HAVING clauses, inside other queries. The purpose of every query is documented throughout the code.
You can see in Listing 11.5 how to solve the same queries from Listing 11.4, without using any subquery.
Note that we do not show the output of Listing 11.5 because it is the same as for Listing 11.4.
Note

To use a query as a subquery inside another query, you must enclose the subquery in parentheses.

Listing 11.4 Use Scalar Subqueries Inside Other Queries

USE Northwind
GO

SET NOCOUNT ON
GO

-- In this case we combine the values returned by two subqueries

-- to get the medium unit price
SELECT (
(SELECT MIN(Unitprice)
FROM Products) +
(SELECT MAX(Unitprice)
FROM Products))/2 as NewPrice

428
 Chapter 11. Using Complex Queries and Statements

-- This query is not practically useful,

-- but it shows more choices on designing subqueries
SELECT 1, 2,
(SELECT 3)
GO

-- This query uses two subqueries to retrieve one single row

-- with the Maximum and Average UnitPrice

SELECT (
SELECT AVG(Unitprice)
FROM Products
) as AvgPrice
, (
SELECT MAX(Unitprice)
FROM Products
) as MaxPrice
GO

-- Compare the UnitPrice of every product

-- with the Average UnitPrice, produced by a subquery

SELECT ProductName, UnitPrice, (

SELECT AVG(Unitprice)
FROM Products
) as AvgPrice
FROM Products
WHERE CategoryID = 2
GO

-- Updates the UnitPrice of the product 11 to

-- 20% more than the maximum UnitPrice.

UPDATE Products
SET UnitPrice = (
SELECT MAX(Unitprice)
FROM Northwind..Products
) * 1.2
WHERE ProductID = 11
-- Show the product with maximum UnitPrice

SELECT ProductName, UnitPrice

FROM Products P
WHERE Unitprice = (
SELECT Max(UnitPrice) MPrice
FROM Products
)

-- You want to retrieve the Categories with average Unitprice

-- greater than the overall products average price

SELECT CategoryID, AVG(UnitPrice) AS 'Average Price'

FROM Products P
GROUP BY CategoryID
HAVING AVG(UnitPrice) > (
SELECT AVG(UnitPrice) MPrice
FROM Products
)

429
Microsoft SQL Server 2000 Programming by Example

NewPrice
---------------------
$190.97

----------- ----------- -----------

1.00 2.00 3.00

AvgPrice MaxPrice
--------------------- ---------------------
$33.88 $379.44

ProductName UnitPrice AvgPrice

---------------------------------------- --------------------- -----------
Aniseed Syrup $11.00 $33.88
Chef Anton's Cajun Seasoning $24.20 $33.88
Chef Anton's Gumbo Mix $23.49 $33.88
Grandma's Boysenberry Spread $27.50 $33.88
Northwoods Cranberry Sauce $44.00 $33.88
Genen Shouyu $17.05 $33.88
Gula Malacca $21.40 $33.88
Sirop d'érable $31.35 $33.88
Vegie-spread $48.29 $33.88
Louisiana Fiery Hot Pepper Sauce $23.16 $33.88
Louisiana Hot Spiced Okra $18.70 $33.88
Original Frankfurter grüne Soße$14.30 $33.88
ProductName UnitPrice
---------------------------------------- ---------------------
Queso Cabrales $455.33

CategoryID Average Price

----------- ---------------------
1.00 $37.98
4.00 $72.16
6.00 $54.01

Listing 11.5 Solving the Scalar Subqueries Examples from Listing 11.4 Without Subqueries

USE Northwind
GO

SET NOCOUNT ON

430
 Chapter 11. Using Complex Queries and Statements

GO

-- Get the medium unit price

SELECT (MIN(UnitPrice) +
MAX(UnitPrice))/2 AS NewPrice
FROM Products

-- Selects three constants

SELECT 1, 2, 3
GO

-- Retrieve one single row with the Maximum and Average UnitPrice

SELECT AVG(Unitprice) as AvgPrice,

MAX(Unitprice) as MaxPrice
FROM Products

GO

-- Compare the UnitPrice of every product

-- with the Average UnitPrice

DECLARE @UP money

SELECT @UP = AVG(Unitprice)

FROM Products

SELECT ProductName, UnitPrice, @UP AS AvgPrice

FROM Products
WHERE CategoryID = 2
GO

-- Updates the UnitPrice of the product 11 to

-- 20% more than the maximum UnitPrice.

DECLARE @UP money

SELECT @UP = MAX(Unitprice)

FROM Products

UPDATE Products
SET UnitPrice = @UP * 1.2
WHERE ProductID = 11

-- You want to show the product with maximum UnitPrice

DECLARE @UP money

SELECT @UP = MAX(Unitprice)

FROM Products

SELECT ProductName, UnitPrice

FROM Products P
WHERE Unitprice = @UP

-- You want to retrieve the Categories with average Unitprice

-- greater than the overall products average price

DECLARE @UP money

431
Microsoft SQL Server 2000 Programming by Example

SELECT @UP = AVG(Unitprice)

FROM Products

SELECT CategoryID, AVG(UnitPrice)

FROM Products P
GROUP BY CategoryID
HAVING AVG(UnitPrice) > @UP

List Subqueries

A List query can be used as a subquery inside a query in the following cases:

• In the WHERE clause of any query using the IN operator to specify the List query as a list of possible
values.
• In the WHERE clause when using any comparison operator with the SOME, ANY, or ALL operators.
• In the FROM clause of a SELECT statement, as a multirow and single- column derived table.
• In the WHERE clause, using the EXISTS or NOT EXISTS keywords to test for existence of values in
the List.

Listing 11.6 contains some examples of subqueries that produce lists of values. The first example uses a list
subquery in the WHERE clause introduced with the IN operator. The second example uses a list subquery in
the WHERE clause as well, with the ALL operator. The third example uses a list subquery as a derived table in
the FROM clause. The last example shows a subquery in the WHERE clause using the EXISTS operator.
Listing 11.7 contains the same examples, but without using list subqueries. The output is the same as in
Listing 11.6.

Listing 11.6 Using List Queries As Subqueries

USE Northwind
GO

SET NOCOUNT ON
GO

-- Orders placed by clients from London

-- and EmployeeID = 1

SELECT OrderID, CustomerID, EmployeeID, OrderDate

FROM Orders
WHERE CustomerID IN (
SELECT CustomerID
FROM Customers
WHERE City = 'London'
)
AND EmployeeID = 1

-- Select all the products with the UnitPrice

-- greater than all the products from Category 2

432
 Chapter 11. Using Complex Queries and Statements

SELECT ProductID, ProductName, UnitPrice

FROM Products
WHERE UnitPrice > ALL (
SELECT UnitPrice
FROM Products
WHERE CategoryID = 2
)
-- Select all the order details related to products from category 2
-- and OrderID between 10250 and 10300

SELECT OD.OrderID, OD.ProductID, OD.UnitPrice

FROM [Order Details] OD
JOIN (
SELECT ProductID
FROM Products
WHERE CategoryID = 2
) AS P
ON P.ProductID = OD.ProductID
WHERE OrderID BETWEEN 10250 AND 10300

-- List all the products only if there are any products never ordered

SELECT ProductID, ProductName

FROM Products
WHERE EXISTS (
SELECT Products.ProductID
FROM Products
LEFT OUTER JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE [Order Details].ProductID IS NULL
)

OrderID CustomerID EmployeeID OrderDate

----------- ---------- ----------- --------------------------
10,364.00 EASTC 1.00 11/26/1996 12:00:00 AM
10,377.00 SEVES 1.00 12/9/1996 12:00:00 AM
10,400.00 EASTC 1.00 1/1/1997 12:00:00 AM
10,453.00 AROUT 1.00 2/21/1997 12:00:00 AM
10,558.00 AROUT 1.00 6/4/1997 12:00:00 AM
10,743.00 AROUT 1.00 11/17/1997 12:00:00 AM
10,800.00 SEVES 1.00 12/26/1997 12:00:00 AM
11,023.00 BSBEV 1.00 4/14/1998 12:00:00 AM

ProductID ProductName UnitPrice

----------- ---------------------------------------- ---------------------
9.00 Mishi Kobe Niku $97.00
11.00 Queso Cabrales $455.33
18.00 Carnarvon Tigers $62.50
20.00 Sir Rodney's Marmalade $81.00
29.00 Thüringer Rostbratwurst $123.79
38.00 Côte de Blaye $263.50
51.00 Manjimup Dried Apples $53.00

433
Microsoft SQL Server 2000 Programming by Example

59.00 Raclette Courdavault $55.00

62.00 Tarte au sucre $49.30
OrderID ProductID UnitPrice
----------- ----------- ---------------------
10,250.00 65.00 $16.80
10,251.00 65.00 $16.80
10,256.00 77.00 $10.40
10,257.00 77.00 $10.40
10,258.00 5.00 $17.00
10,262.00 5.00 $17.00
10,278.00 44.00 $15.50
10,278.00 63.00 $35.10
10,283.00 15.00 $12.40
10,284.00 44.00 $15.50
10,289.00 3.00 $8.00
10,290.00 5.00 $17.00
10,290.00 77.00 $10.40
10,291.00 44.00 $15.50
10,293.00 63.00 $35.10
10,300.00 66.00 $13.60

ProductID ProductName
----------- ----------------------------------------

Listing 11.7 Solving the List Subqueries Examples from Listing 11.6 Without Subqueries

USE Northwind
GO

SET NOCOUNT ON
GO

-- Orders placed by clients from London

-- and EmployeeID = 1

SELECT O.OrderID, O.CustomerID, O.EmployeeID, O.OrderDate

FROM Orders O
JOIN Customers C
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND EmployeeID = 1

-- Select all the products with the UnitPrice

-- greater than all the products from Category 2

DECLARE @MP money

SELECT @MP = MAX(Unitprice)

FROM Products
WHERE CategoryID = 2
SELECT ProductID, ProductName, UnitPrice
FROM Products

434
 Chapter 11. Using Complex Queries and Statements

WHERE UnitPrice > @MP

-- Select all the order details related to products from category 2

-- and OrderID between 10250 and 10300

SELECT OD.OrderID, OD.ProductID, OD.UnitPrice

FROM [Order Details] OD
JOIN Products P
ON P.ProductID = OD.ProductID
WHERE CategoryID = 2
AND OrderID BETWEEN 10250 AND 10300

-- List all the products only if there are any products never ordered

DECLARE @n int

SELECT @n = COUNT(*)
FROM Products
LEFT OUTER JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE [Order Details].ProductID IS NULL

SELECT ProductID, ProductName

FROM Products
WHERE ISNULL(@n, 0) > 0

-- OR

IF EXISTS(
SELECT Products.ProductID
FROM Products
LEFT OUTER JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE [Order Details].ProductID IS NULL
)
SELECT ProductID, ProductName
FROM Products

Array Subqueries

An Array query, or standard query, can be used as a subquery inside a query in the following cases:

• In the FROM clause of a SELECT statement, as a multirow and multicolumn derived table.
• In the WHERE clause, using the EXISTS or NOT EXISTS keywords to test for existence of values in
the List. The EXISTS function does not return any rows, it evaluates to TRUE if the subquery returns
at least one row, and it evaluates to FALSE otherwise.

Listing 11.8 shows two examples of using array subqueries. The first example uses an array subquery in the
FROM clause as a derived table. The second example combines two result sets with the UNION operator,
introducing two array subqueries with the EXISTS operator.
Listing 11.9 solves the same problems from Listing 11.8 without using subqueries. Note that we do not
show the output for Listing 11.9 because it is exactly the same as for Listing 11.8.

Listing 11.8 Using Array Queries As Subqueries

435
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

SET NOCOUNT ON
GO

-- Show name of product and category for products of categories 1 to 3

SELECT CategoryName, ProductName

FROM Products P
JOIN (
SELECT CategoryID, CategoryName
FROM Categories
WHERE CategoryID
BETWEEN 1 AND 3
) AS C
ON C.CategoryID = P.CategoryID
WHERE ProductName LIKE 'L%'

-- Show if we have beverage sales

SELECT 'We do have beverage Sales'

AS [Beverage Sales]
WHERE EXISTS(
SELECT *
FROM [Order details] O
JOIN Products P
ON O.ProductID = P.ProductID
WHERE P.CategoryID = 1
)

UNION
SELECT 'We do not have any beverage Sales'
WHERE NOT EXISTS(
SELECT *
FROM [Order details] O
JOIN Products P
ON O.ProductID = P.ProductID
WHERE P.CategoryID = 1
)

CategoryName ProductName
--------------- ----------------------------------------
Beverages Lakkalikööri

436
 Chapter 11. Using Complex Queries and Statements

Beverages Laughing Lumberjack Lager

Condiments Louisiana Fiery Hot Pepper Sauce
Condiments Louisiana Hot Spiced Okra

Beverage Sales
---------------------------------
We do have beverage Sales

Listing 11.9 Solving the Array Subquery Examples from Listing 11.8 Without Subqueries

-- Show name of product and category for products of categories 1 to 3

SELECT CategoryName, ProductName

FROM Products P
JOIN Categories AS C
ON C.CategoryID = P.CategoryID
WHERE P.CategoryID
BETWEEN 1 AND 3
AND ProductName LIKE 'L%'

-- Show if we have beverage sales

-- Note: CASE will be explained later in this chapter

DECLARE @N int

SELECT @N = COUNT(*)
FROM [Order details] O
JOIN Products P
ON O.ProductID = P.ProductID
WHERE P.CategoryID = 1

SELECT CASE
WHEN ISNULL(@n, 0) > 0
THEN 'We do have beverage Sales'
ELSE 'We do not have any beverage Sales'
END
AS [Beverage Sales]
Caution

In some cases, converting a query that uses a subquery into another query without a subquery can
produce unexpected results, as in Listing 11.10. Make sure the query is defined properly to
retrieve the desired results.

Tip

437
Microsoft SQL Server 2000 Programming by Example

It is recommended to use IF EXISTS( SELECT * FROM ...) because Query Optimizer will
select the best available index to investigate the existence of rows.

Listing 11.10 Rewriting a Query to Avoid the Use of a Subquery Can Produce Unexpected Results If
Conditions Are Not Applied Properly

USE Northwind
GO

SET NOCOUNT ON
GO

-- Retrieve all the categories with products

-- This solution uses a subquery and retrieves 1 row

PRINT 'Using a Subquery'+ CHAR(10)

SELECT CategoryName
FROM Categories
WHERE CategoryID IN
(SELECT CategoryID
FROM Products)
AND Description LIKE '%pasta%'

-- This solution does not use a subquery but it retrieves 7 rows

PRINT 'Without Using a Subquery'+ CHAR(10)

SELECT CategoryName
FROM Categories
JOIN Products
ON Products.categoryID = Categories.CategoryID
WHERE Description LIKE '%pasta%'

-- This solution does not use a subquery but it retrieves 1 row again

PRINT 'Using DISTINCT Without a Subquery'+ CHAR(10)

SELECT DISTINCT CategoryName

FROM Categories
JOIN Products
ON Products.categoryID = Categories.CategoryID
WHERE Description LIKE '%pasta%'

Using a Subquery

438
 Chapter 11. Using Complex Queries and Statements

CategoryName
---------------
Grains/Cereals

Without Using a Subquery

CategoryName
---------------
Grains/Cereals
Grains/Cereals
Grains/Cereals
Grains/Cereals
Grains/Cereals
Grains/Cereals
Grains/Cereals

Using DISTINCT Without a Subquery

CategoryName
---------------
Grains/Cereals
From the previous examples, you can get the impression that every query that uses subqueries can be
defined as a standard query without any subquery. However, some problems can have an easier solution
using subqueries than not using subqueries at all. What if you wanted to know the best-selling product by
number of units and by total sale price? Listing 11.11 shows how to create subquery to solve this problem
and how to do it without a subquery.

Listing 11.11 In Some Cases the Easiest Solution Is to Use Subqueries

USE Northwind
GO

SET NOCOUNT ON
GO

-- Retrieve the best selling product by number of units and by revenue

-- This solution uses subqueries

SELECT 'By units'AS Criteria,
ProductName as 'Best Selling'
FROM Products

439
Microsoft SQL Server 2000 Programming by Example

WHERE ProductID = (
SELECT ProductID
FROM [Order Details]
GROUP BY ProductID
HAVING SUM(Quantity) = (
SELECT MAX(SQ)
FROM (
SELECT SUM(Quantity) as SQ
FROM [Order Details]
GROUP BY ProductID
) AS OD))

UNION

SELECT 'By revenue'AS Criteria,

ProductName as 'Best Selling'
FROM Products
WHERE ProductID = (
SELECT ProductID
FROM [Order Details]
GROUP BY ProductID
HAVING SUM(UnitPrice * Quantity * (1-Discount)) = (
SELECT MAX(SQ)
FROM (
SELECT SUM(UnitPrice * Quantity * (1-Discount)) as SQ
FROM [Order Details]
GROUP BY ProductID
) AS OD))

-- This solution uses subqueries as well

SELECT 'By units'AS Criteria,

ProductName as 'Best Selling'
FROM Products P
JOIN (
SELECT TOP 1 ProductID,
SUM(Quantity) AS SQ
FROM [Order Details]
GROUP BY productID
ORDER BY SQ DESC
) AS OD
ON OD.ProductID = P.ProductID

UNION
SELECT 'By revenue'AS Criteria,
ProductName as 'Best Selling'
FROM Products P
JOIN (
SELECT TOP 1 ProductID,
SUM(UnitPrice * Quantity * (1-Discount)) AS SR
FROM [Order Details]
GROUP BY ProductID
ORDER BY SR DESC
) AS OD
ON OD.ProductID = P.ProductID

-- This solution does not use subqueries.

-- However the execution is similar to the query that uses subqueries
SELECT ProductID,
SUM(Quantity) AS SQ,

440
 Chapter 11. Using Complex Queries and Statements

CAST(SUM(UnitPrice * Quantity * (1.0-Discount))AS money) as SR

INTO #BestSelling
FROM [Order Details]
WHERE ProductID IS NOT NULL
GROUP BY productID

DECLARE @MQ int, @MR money

DECLARE @PQ int, @PR int

SELECT @MQ = MAX(SQ),

@MR = MAX(SR)
FROM #BestSelling

SELECT @PQ = ProductID

FROM #BestSelling
WHERE SQ = @MQ

SELECT @PR = ProductID

FROM #BestSelling
WHERE SR = @MR

SELECT 'By units'AS Criteria,

ProductName as 'Best Selling'
FROM Products
WHERE ProductID = @PQ

UNION

SELECT 'By revenue'AS Criteria,

ProductName as 'Best Selling'
FROM Products
WHERE ProductID = @PR

-- drop temporary table

DROP TABLE #BestSelling

(Same output for every query)

Criteria Best Selling

---------- ----------------------------------------
By revenue Côte de Blaye
By units Camembert Pierrot

Correlated Subqueries
In the previous examples, the subqueries were normal queries that you could run isolated from the outer
query.
In some cases, you might need to write a subquery that depends on values from the outer query to retrieve its
results. In these cases, you call the subquery a correlated subquery. This usually forces the execution of the
subquery once per every row returned from the outer query.
Tip

441
Microsoft SQL Server 2000 Programming by Example

You can easily identify a correlated subquery because it is enclosed in parentheses and it cannot
be executed independently from the outer query.

Correlated subqueries can return a single scalar value, a list of values, or an array of values in the same way
as standard subqueries.

Suppose you wanted to know the suggested UnitPrice for every product, together with the average and
maximum selling price. You can solve this problem as in Listing 11.12.

Listing 11.12 Use Correlated Subqueries to Solve Complex Problems

-- Select the target UnitPrice, the minimum, average,

-- and maximum selling price for every product
-- show non-ordered products as well

-- With a Correlated Subquery

SELECT ProductID,
UnitPrice,
(SELECT AVG(UnitPrice)
FROM [Order Details]
WHERE [Order Details].ProductID =
Products.ProductID) AS AvgPrice,
(SELECT MIN(UnitPrice)
FROM [Order Details]
WHERE [Order Details].ProductID =
Products.ProductID) AS MaxPrice,
ProductName
FROM Products
WHERE CategoryID = 1

-- With a standard (non-correlated) subquery

SELECT P.ProductID,
P.UnitPrice,
MMP.AvgPrice,
MMP.MinPrice,
P.ProductName
FROM Products P
LEFT OUTER JOIN (
SELECT ProductID,
MIN(UnitPrice) AS MinPrice,
AVG(UnitPrice) AS AvgPrice
FROM [Order Details]
GROUP BY ProductID
) AS MMP

442
 Chapter 11. Using Complex Queries and Statements

ON MMP.ProductID =
P.ProductID
WHERE CategoryID = 1

-- Without subqueries

SELECT P.ProductID,
P.UnitPrice,
AVG(OD.UnitPrice) AS AvgPrice,
MIN(OD.UnitPrice) AS MinPrice,
P.ProductName
FROM Products P
LEFT OUTER JOIN [Order Details] OD
ON OD.ProductID =
P.ProductID
WHERE CategoryID = 1
GROUP BY P.productID, P.UnitPrice, P.ProductName
(Same output for every query)

ProductID UnitPrice AvgPrice MaxPrice ProductName

----------- ----------- ----------- ----------- --------------------------
1.00 $18.00 $17.15 $14.40 Chai
2.00 $19.00 $17.88 $15.20 Chang
24.00 $4.50 $4.24 $3.60 Guaraná Fantástica
34.00 $14.00 $12.97 $11.20 Sasquatch Ale
35.00 $18.00 $17.00 $14.40 Steeleye Stout
38.00 $263.50 $245.93 $210.80 Côte de Blaye
39.00 $18.00 $16.68 $14.40 Chartreuse verte
43.00 $46.00 $43.04 $36.80 Ipoh Coffee
67.00 $14.00 $13.72 $11.20 Laughing Lumberjack Lager
70.00 $15.00 $14.15 $12.00 Outback Lager
75.00 $7.75 $7.38 $6.20 Rhönbräu Klosterbier
76.00 $18.00 $16.98 $14.40 Lakkalikööri
Caution

It is important to qualify the column names with the table name inside a subquery to avoid
ambiguity. However, if you do not qualify column names, SQL Server will resolve them first from
the subquery, and then from the outer query.

Using subqueries produces similar results to OUTER JOIN queries. To implement functionality similar to
INNER JOIN queries, use EXISTS to test for existence in the inner table as in Listing 11.13.

Listing 11.13 Use EXISTS to Simulate Inner Queries When Using Subqueries

443
Microsoft SQL Server 2000 Programming by Example

-- As in the first example from Listing 11-12

-- but in this case we select only products with orders

SELECT ProductID,
UnitPrice,
(SELECT AVG(UnitPrice)
FROM [Order Details]
WHERE [Order Details].ProductID =
Products.ProductID) AS AvgPrice,
(SELECT MIN(UnitPrice)
FROM [Order Details]
WHERE [Order Details].ProductID =
Products.ProductID) AS MaxPrice,
ProductName
FROM Products
WHERE EXISTS (
SELECT *
FROM [Order Details]
WHERE [Order Details].ProductID = Products.productID
)
AND CategoryID = 1

(Same output for every query)

ProductID UnitPrice AvgPrice MaxPrice ProductName

----------- ----------- ----------- ----------- --------------------------
1.00 $18.00 $17.15 $14.40 Chai
2.00 $19.00 $17.88 $15.20 Chang
24.00 $4.50 $4.24 $3.60 Guaraná Fantástica
34.00 $14.00 $12.97 $11.20 Sasquatch Ale
35.00 $18.00 $17.00 $14.40 Steeleye Stout
38.00 $263.50 $245.93 $210.80 Côte de Blaye
39.00 $18.00 $16.68 $14.40 Chartreuse verte
43.00 $46.00 $43.04 $36.80 Ipoh Coffee
67.00 $14.00 $13.72 $11.20 Laughing Lumberjack Lager
70.00 $15.00 $14.15 $12.00 Outback Lager
75.00 $7.75 $7.38 $6.20 Rhönbräu Klosterbier
76.00 $18.00 $16.98 $14.40 Lakkalikööri
It is common in correlated subqueries to use the same object inside and outside the subquery. In this case,
you must provide an object alias to avoid ambiguity. In Listing 11.14, you want to know the list of customers
who are based in the same city of at least one more customer.

Listing 11.14 Use Table Aliases to Avoid Ambiguity When Using Correlated Subqueries

444
 Chapter 11. Using Complex Queries and Statements

USE Northwind
GO

SET NOCOUNT ON
GO

-- Retrieve the list of Cities where we have more than one customer,
-- ordered by City and CompanyName

-- With Correlated Subquery

SELECT City, CompanyName

FROM Customers C1
WHERE City IN (
SELECT City
FROM Customers C2
WHERE C2.CustomerID <> C1.CustomerID
)
AND Country = 'Argentina'
ORDER BY City, CompanyName

-- Without Subqueries

SELECT DISTINCT C1.City, C1.CompanyName

FROM Customers C1
JOIN Customers AS C2
ON C2.City = C1.City
WHERE C2.CustomerID <> C1.CustomerID
AND C1.Country = 'Argentina'
ORDER BY C1.City, C1.CompanyName

(Same output for every query)

City CompanyName
--------------- ----------------------------------------
Buenos Aires Cactus Comidas para llevar
Buenos Aires Océano Atlántico Ltda.
Buenos Aires Rancho grande

Derived Tables
Some of the examples from the previous section of this chapter can be solved using subqueries in the FROM
clause. Subqueries work as tables in the query, and they are called derived tables. SQL Server considers

445
Microsoft SQL Server 2000 Programming by Example

these subqueries in the same way as ad hoc views, because the definition of the derived tables is merged
with the outer query to produce a single query execution plan.
To use a derived table in a query, you must specify in the FROM clause a standard subquery with a table alias
for the derived table (see Listing 11.15).
Caution

Correlated queries cannot be used as derived tables.

Listing 11.15 You Can Use Derived Tables to Solve Common Queries Without Defining Views or
Temporary Tables

-- Using a Derived Table to compare UnitPrice

-- with the actual minimum and maximum prices
SELECT P.ProductID, P.UnitPrice
, (MinPrice + MaxPrice) / 2 AS MediumPrice
FROM Products P
JOIN (
SELECT ProductID,
MIN(UnitPrice) as MinPrice,
MAX(UnitPrice) as Maxprice
FROM [Order Details]
GROUP BY ProductID
) AS OD
ON OD.ProductID = P.ProductID

-- Using a Derived Table to produce a discounted price list

SELECT P.ProductID,
Discount * 100 AS Discount,
P.UnitPrice * (1 - Discount) as Price
FROM Products P
CROSS JOIN (
SELECT 0.0 AS Discount
UNION ALL
SELECT 0.10
UNION ALL
SELECT 0.20
UNION ALL
SELECT 0.30
) AS OD

446
 Chapter 11. Using Complex Queries and Statements

-- Partial result

ProductID UnitPrice MediumPrice

----------- --------------------- ---------------------
23 9.0000 8.1000
46 12.0000 10.8000
69 36.0000 32.4000
77 13.0000 11.7000
31 12.5000 11.2500
15 15.5000 13.9500
62 49.3000 44.3500
38 263.5000 237.1500

-- Partial result

ProductID Discount Price

----------- -------- --------------------------
1 .00 18.000000
1 10.00 16.200000
1 20.00 14.400000
1 30.00 12.600000
2 .00 19.000000
2 10.00 17.100000
2 20.00 15.200000
2 30.00 13.300000
Tip

Derived tables are good candidates to be defined as views. After converted into views, some of
them can be indexed, providing extra performance to your reporting queries. You can apply
permissions to views but not to derived tables.

The CASE Function

You can use the CASE function to provide conditional values to an expression.
Caution

You cannot use the CASE function to provide conditional execution of statements as in other
programming languages.

The result of a CASE function is always a scalar value. You can use CASE in two different ways:

447
Microsoft SQL Server 2000 Programming by Example

• Select a result depending on the possible values that a variable, column, or expression can have. This
is called Simple CASE. CASE returns the value correspondent to the first value that is equal to the
searched expression. Listing 11.16 contains three examples using the CASE function to provide a
verbose output.
• The CASE function evaluates in sequence several independent conditions. The CASE function returns
the result correspondent to the first condition that evaluates to TRUE. This is called Searched CASE.
Listing 11.17 shows three examples of Searched CASE.

Listing 11.16 Using Simple CASE Function to Provide an Easier-to-Understand Output

-- Use simple CASE to expand abbreviations

SELECT CASE Country

WHEN 'UK'THEN 'United Kingdom'
WHEN 'USA'THEN 'United States of America'
ELSE Country
END AS Country,
CompanyName
FROM Customers
ORDER BY 1, 2

-- Use simple CASE to define logical values

SELECT ProductName,
CASE Discontinued
WHEN 1 THEN 'Discontinued'
ELSE 'Available'
END AS Status
FROM Products

-- Use simple CASE to produce verbose results

SELECT CompanyName,
OrderID,
CASE YEAR(OrderDate)
WHEN YEAR(Getdate()) THEN 'This Year'
WHEN YEAR(GetDate()) - 1 THEN 'Last year'
ELSE CAST(DATEDIFF(year, OrderDate, Getdate()) AS varchar(5))
+ 'years ago'
END AS 'When'
FROM Orders
JOIN Customers
ON Orders.CustomerID = Customers.CustomerID

448
 Chapter 11. Using Complex Queries and Statements

(Partial output)

Country CompanyName
------------------------ ----------------------------------------
United Kingdom Around the Horn
United Kingdom B's Beverages
United Kingdom Seven Seas Imports
United States of America Great Lakes Food Market
United States of America Trail's Head Gourmet Provisioners
United States of America White Clover Markets
Venezuela GROSELLA-Restaurante
Venezuela LINO-Delicateses

(Partial output)

ProductName Status
---------------------------------------- ------------
Gustaf's Knäckebröd Available
Singaporean Hokkien Fried Mee Discontinued
Wimmers gute Semmelknödel Available

(Partial output)

CompanyName OrderID When

---------------------------------------- ----------- ---------------
Furia Bacalhau e Frutos do Mar 10963 2 years ago
Furia Bacalhau e Frutos do Mar 10664 3 years ago
Furia Bacalhau e Frutos do Mar 10328 4 years ago
Princesa Isabel Vinhos 11007 2 years ago
Princesa Isabel Vinhos 10433 3 years ago
Princesa Isabel Vinhos 10336 4 years ago

Listing 11.17 Using Searched CASE You Can Solve Complex Queries

-- Use searched CASE to define regions

SELECT CASE
WHEN Country IN ('Argentina', 'Brazil', 'Venezuela')
THEN 'South America'
WHEN Country IN ('Canada', 'USA')
THEN 'North America'
WHEN Country IN ('Mexico')
THEN 'Central America'
WHEN Country IN ('Austria', 'Belgium', 'Denmark',
'Finland', 'France', 'Germany', 'Ireland', 'Italy',
'Norway', 'Poland','Portugal','Spain',
'Sweden', 'Switzerland', 'UK')
THEN 'Europe'
ELSE 'Undefined'
END AS Continent,
CompanyName

449
Microsoft SQL Server 2000 Programming by Example

FROM Customers

-- Use searched CASE to define ranges of values

SELECT ProductName,
CASE
WHEN UnitPrice < 10 THEN 'Inexpensive'
WHEN UnitPrice < 50 THEN 'Fair'
WHEN UNitPrice < 100 THEN 'Expensive'
ELSE 'Very Expensive'
END AS Price
FROM products

-- Use searched CASE to get values that depend on data from other tables
-- in this example using a correlated subquery

SELECT CategoryName,
CASE
WHEN EXISTS (
SELECT *
FROM Products P
WHERE P.CategoryID = C.CategoryID)
THEN 'Yes'ELSE 'No'END AS 'Has products'
FROM Categories C

(Partial output)

Continent CompanyName
--------------- ----------------------------------------
Europe Around the Horn
Europe Seven Seas Imports
North America Great Lakes Food Market
North America White Clover Markets
South America GROSELLA-Restaurante
South America LINO-Delicateses

(Partial output)

ProductName Price
---------------------------------------- --------------
Gustaf's Knäckebröd Fair
Tunnbröd Inexpensive
Singaporean Hokkien Fried Mee Fair
Filo Mix Inexpensive
Gnocchi di nonna Alice Fair

(Partial output)

CategoryName Has products

--------------- ------------
Beverages Yes

450
 Chapter 11. Using Complex Queries and Statements

Grains/Cereals Yes
Meat/Poultry No
Produce Yes
Seafood Yes
Tip

It is not necessary to create mutually exclusive conditions in a CASE function because the search
will finish on the first condition that evaluates to true.

The COMPUTE Clause

In Chapter 4, you learned how to use aggregate functions to produce summary data. Queries using
aggregate functions show only summary information by grouping data.
The Transact-SQL COMPUTE clause provides the capability to create reports that show detailed and summary
information.
Caution

COMPUTE is not an ANSI standard feature.

To add summary information to any standard query, just add the COMPUTE clause at the end of the query,
specifying which aggregate function to compute, as in Listing 11.18.

Listing 11.18 Use COMPUTE to Show Summary Information at the End of a Standard Query

-- This is a normal query to retrieve Price and Discount from Order details

SELECT OrderID,
ProductID,
(UnitPrice * Quantity * (1-Discount)) AS Price,
Discount
FROM [Order Details]
-- This is the same query as before
-- but it shows the total of Price and the average discount

SELECT OrderID,
ProductID,
(UnitPrice * Quantity * (1-Discount)) AS Price,
Discount
FROM [Order Details]
COMPUTE SUM((UnitPrice * Quantity * (1-Discount))),
AVG(Discount)
With COMPUTE SUM and AVG (Partial results)

451
Microsoft SQL Server 2000 Programming by Example

OrderID ProductID Price Discount

----------- ----------- ------------------------ ------------------------
10250 41 77.0 0.0
10250 51 1261.4 0.15000001
10251 22 95.760002 5.0000001E-2
...
11077 52 14.0 0.0
11077 60 63.919998 5.9999999E-2
11077 73 29.700001 9.9999998E-3
11077 77 26.0 0.0
sum
===========================
1265793.0396184921
avg
============================
5.6167054202260661E-2
As you saw previously in Listing 11.18, the output is not a result set. It is a report that can be sent to a
printer, but it is difficult to use from a client application.
You can go one step further, because using the COMPUTE BY clause, you can show subtotals, as in Listing
11.19. In this case, you must use an ORDER BY clause in the query including, at least, the fields used in the
COMPUTE BY clause, and in the same order.

Listing 11.19 Use COMPUTE BY to Show Subtotals at Different Levels

-- Retrieve Price and discount from Order details

-- plus Category, product and customer
-- and show the Price subtotal per Category

SELECT CategoryName AS Category,

ProductName AS Product,
CompanyName AS Customer,
(OD.UnitPrice * Quantity * (1-Discount)) AS Price,
Discount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.categoryID = P.CategoryID
JOIN Customers CS
ON CS.CustomerID = O.CustomerID

452
 Chapter 11. Using Complex Queries and Statements

ORDER BY Category
COMPUTE SUM((OD.UnitPrice * Quantity * (1-Discount)))
BY Category

-- This is similar to the preceding query

-- but it shows the Average discount per Customer

SELECT CompanyName AS Customer,

(OD.UnitPrice * Quantity * (1-Discount)) AS Price,
Discount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Customers CS
ON CS.CustomerID = O.CustomerID
ORDER BY Customer
COMPUTE AVG(Discount)
BY Customer

-- This is a similar query as before

-- but it shows the Average discount per Category and Product
-- plus the overall average and grand total price

SELECT CategoryName AS Category,

ProductName AS Product,
(OD.UnitPrice * Quantity * (1-Discount)) AS Price,
Discount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.categoryID = P.CategoryID
ORDER BY Category, Product
COMPUTE AVG(Discount)
BY Category, Product
COMPUTE AVG(Discount)
BY Category
COMPUTE SUM((OD.UnitPrice * Quantity * (1-Discount)))
COMPUTE SUM(Price) BY Category (Partial results)

Category Product Customer Price Discount

---------- -------------------- ------------------------- ------- --------
Beverages Chartreuse verte Alfreds Futterkiste 283.5 0.25
Beverages Lakkalikööri Alfreds Futterkiste 270.0 0.0
Beverages Outback Lager Ana Trujillo Emparedados 60.0 0.0
...
Beverages Steeleye Stout Wolski Zajazd 54.0 0.0
Beverages Guaraná Fantástica Wolski Zajazd 54.0 0.0
Beverages Rhönbräu Klosterbier Wolski Zajazd 232.5 0.0

453
Microsoft SQL Server 2000 Programming by Example

sum
==================
267868.1805229187

Category Product Customer Price Discount

---------- -------------------- ------------------------- ------- --------
Condiments Vegie-spread Alfreds Futterkiste 878.0 0.0
Condiments Aniseed Syrup Alfreds Futterkiste 60.0 0.0
Condiments Original Frankfurter Alfreds Futterkiste 20.80 0.2
...
Seafood Inlagd Sill White Clover Markets 665.0 0.0
Seafood Boston Crab Meat Wilman Kala 220.8 0.0
Seafood Escargots de Bourgo Wolski Zajazd 159.0 0.0

sum
==================
131261.73742485046

Customer Price Discount

------------------------------ ---------- --------
Alfreds Futterkiste 513.0 0.25
Alfreds Futterkiste 283.5 0.25
Alfreds Futterkiste 18.0 0.25
...
Alfreds Futterkiste 503.5 5.0000001E-2
Alfreds Futterkiste 430.0 0.0
avg
=====================
0.08750000037252903

Customer Price Discount

------------------------------ ---------- --------
Ana Trujillo Emparedados y hel 28.799999 0.0
Ana Trujillo Emparedados y hel 60.0 0.0
...
Wolski Zajazd 232.5 0.0
Wolski Zajazd 591.59998 0.0

avg
=====================
0.0

COMPUTE AVG(Discount) BY Category, product

and SUM(price) Grand Total (Partial results)

Category Product Price Discount

----------- ---------------- ---------- -----------
Beverages Chai 288.0 0.0
Beverages Chai 144.0 0.0
Beverages Chai 576.0 0.2
Beverages Chai 72.0 0.2
...
Beverages Chai 162.0 0.1
Beverages Chai 337.5 0.25

avg
======================
7.7631580104169096E-2

Category Product Price Discount

----------- ---------------- ---------- -----------

454
 Chapter 11. Using Complex Queries and Statements

Beverages Chang 608.0 0.2

Beverages Chang 304.0 0.2
...
Beverages Steeleye Stout 360.0 0.0
Beverages Steeleye Stout 54.0 0.0
Beverages Steeleye Stout 720.0 0.0

avg
======================
4.7222223029368453E-2
avg
======================
6.1881189077797501E-2

Category Product Price Discount

----------- ---------------- ---------- -----------
Condiments Aniseed Syrup 400.0 0.0
Condiments Aniseed Syrup 140.0 0.0
Condiments Aniseed Syrup 60.0 0.0
...
Seafood Spegesild 216.0 0.0
Seafood Spegesild 252.0 0.0
Seafood Spegesild 36.0 0.0

avg
======================
6.3703704625368118E-2

avg
======================
6.0242425080275899E-2

sum
===================
1265793.0396184921

COMPUTE AVG(Discount) BY Customer (Partial results)

Caution

Avoid using COMPUTE and COMPUTE BY in client/server applications because they are difficult to
manage from the client side. Every partial output in the report is a different recordset to the client
application. COMPUTE is maintained only for backward compatibility.

Tip

To efficiently produce dynamic reports with summary information, take a look at SQL Server 2000
Analysis Services and Microsoft ActiveX Data Objects Multidimensional in Books Online.

The CUBE and ROLLUP Operators

To facilitate the use of summary information in client applications, SQL Server 2000 provides the ROLLUP and
CUBE operators.

455
Microsoft SQL Server 2000 Programming by Example

As you can see in Listing 11.20, using ROLLUP you can get aggregations at different levels in a single result
set. Remember that using COMPUTE, you obtain a reportlike result set (see Listing 11.19).

Listing 11.20 Use ROLLUP to Produce a Result Set with Aggregations at Different Levels

SELECT CategoryName AS Category,

ProductName AS Product,
SUM((OD.UnitPrice * Quantity * (1-Discount))) AS Price,
AVG(Discount) AS AvgDiscount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.categoryID = P.CategoryID
WHERE CategoryName LIKE 'C%'
GROUP BY CategoryName, ProductName
WITH ROLLUP
(Partial output)

Category Product Price AvgDiscount

------------ ------------------------------ --------------- ----------------
Condiments Aniseed Syrup 3044.0 1.66666669150E-2
Condiments Chef Anton's Cajun Seasoning 8567.89999389 7.50000009313E-2
...
Condiments Vegie-spread 16701.0950012 4.11764712018E-2
Condiments NULL 106047.084989 5.26388897733E-2
Confections Chocolade 1368.71252441 0.108333336810
Confections Gumbär Gummibärchen 19849.1445426 0.051562501117
...
Confections Valkoinen suklaa 3437.6875 0.025000000372
Confections Zaanse koeken 3958.07998657 7.38095243771E-2
Confections NULL 167357.224831 5.69461086474E-2
NULL NULL 273404.309821 5.52545463259E-2
Listing 11.20 shows a normal aggregate output, such as using GROUP BY, but you see three more rows
than usual:

456
 Chapter 11. Using Complex Queries and Statements

Category Product Price AvgDiscount

------------ ------------------------------ --------------- ----------------
Condiments NULL 106047.084989 5.26388897733E-2
Confections NULL 167357.224831 5.69461086474E-2
NULL NULL 273404.309821 5.52545463259E-2
The first two rows have a NULL in Product, which means they have the aggregates for categories
"Condiments" and "Confections". The last row has a NULL in both Category and Product because it
shows the Total SUM and AVG for this result set.
Looking at the output from Listing 11.20, you don't see any subtotals per product because the subtotals are
evaluated from left to right in the GROUP BY clause. If the GROUP BY clause has four fields, using ROLLUP
you will have totals at four levels, including the grand total.
Tip

ROLLUP is usually more efficient than COMPUTE BY.

Use the CUBE operator to obtain every combination of totals (see Listing 11.21).

Listing 11.21 Use CUBE to Produce a Result Set with Aggregations at Every Combination of Levels

SELECT CategoryName AS Category,

ProductName AS Product,
SUM((OD.UnitPrice * Quantity * (1-Discount))) AS Price,
AVG(Discount) AS AvgDiscount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C
ON C.categoryID = P.CategoryID
WHERE CategoryName LIKE 'C%'
GROUP BY CategoryName, ProductName
WITH CUBE
(Partial output)

457
Microsoft SQL Server 2000 Programming by Example

Category Product Price AvgDiscount

------------ ------------------------------ --------------- ----------------
Condiments Aniseed Syrup 3044.0 1.66666669150E-2
Condiments Chef Anton's Gumbo Mix 5347.20000457 7.50000007450E-2
...
Condiments Sirop d'érable 14352.6001586 7.08333346992E-2
Condiments Vegie-spread 16701.0950012 4.11764712018E-2
Condiments NULL 106047.084989 5.26388897733E-2
Confections Chocolade 1368.71252441 0.108333336810
Confections Gumbär Gummibärchen 19849.1445426 0.051562501117
...
Confections Teatime Chocolate Biscuits 5862.61998462 5.54054060296E-2
Confections Zaanse koeken 3958.07998657 7.38095243771E-2
Confections NULL 167357.224831 5.69461086474E-2
NULL NULL 273404.309821 5.52545463259E-2
NULL Aniseed Syrup 3044.0 1.66666669150E-2
...
NULL Vegie-spread 16701.0950012 4.11764712018E-2
NULL Zaanse koeken 3958.07998657 7.38095243771E-2
As you see in the preceding output, there are many rows where Category is NULL. Each one of these rows
represents subtotals per Product. The rest of the output is identical to the Listing 11.20 output.
If the GROUP BY clause has 4 fields, using ROLLUP you will have totals at 15 levels: 4 single-field levels, 6
double-field levels, 4 triple-field levels, and one grand total level.
Checking for NULL values in a client application to detect summary data can be inaccurate if the field
contained in the GROUP BY clause contain NULL values. SQL Server provides the function GROUPING to
detect aggregate rows. GROUPING(Column1) returns 1 if the row shows aggregations for Column1. The
example in Listing 11.22 adds two new columns, GrpCat and GrpProd, using the GROUPING function, to
specify which values are subtotals produced by the CUBE operator.

Listing 11.22 Use the GROUPING Function to Detect Aggregate Rows

SELECT CategoryName AS Category,

GROUPING(CategoryName) AS GrpCat,
ProductName AS Product,
GROUPING(ProductName) AS GrpProd,
AVG(Discount) AS AvgDiscount
FROM [Order Details] OD
JOIN Orders O
ON O.OrderID = OD.OrderID
JOIN Products P
ON P.ProductID = OD.ProductID
JOIN Categories C

458
 Chapter 11. Using Complex Queries and Statements

ON C.categoryID = P.CategoryID
WHERE CategoryName LIKE 'C%'
GROUP BY CategoryName, ProductName
WITH CUBE
(Partial output)

Category GrpCat Product GrpProd AvgDiscount

----------- ------ ------------------------------ ------- ----------------
Condiments 0 Aniseed Syrup 0 1.66666669150E-2
Condiments 0 Chef Anton's Cajun Seasoning 0 7.50000009313E-2
...
Condiments 0 Sirop d'érable 0 7.08333346992E-2
Condiments 0 Vegie-spread 0 4.11764712018E-2
Condiments 0 NULL 1 5.26388897733E-2
Confections 0 Chocolade 0 0.108333336810
Confections 0 Gumbär Gummibärchen 0 0.051562501117
...
Confections 0 Valkoinen suklaa 0 0.025000000372
Confections 0 Zaanse koeken 0 7.38095243771E-2
Confections 0 NULL 1 5.69461086474E-2
NULL 1 NULL 1 5.52545463259E-2
NULL 1 Aniseed Syrup 0 1.66666669150E-2
NULL 1 Chef Anton's Cajun Seasoning 0 7.50000009313E-2
...
NULL 1 Vegie-spread 0 4.11764712018E-2
NULL 1 Zaanse koeken 0 7.38095243771E-2
Tip

In a client application, you can hide the GROUPING columns because they make sense only to the
programmer, not to the user.

Using Hints
In some cases, you might think that there is a better way to execute your queries than the query plan selected
by the Query Optimizer. SQL Server 2000 provides several optimizer hints to tailor the execution to your
needs.
You can specify the type of join to execute by using the LOOP, HASH, MERGE, or REMOTE hints of the JOIN
clause. The query in Listing 11.23 forces a LOOP join to connect the Orders and Order Details tables, a
MERGE join between the Products and Order Details table, and a HASH join to connect the Categories
and Products tables. The purpose of this example is to show how to use the optimizer hints; the output is
the same you can have without using these optimizer hints.
Note

Joins were covered in Chapter 5. The REMOTE hint allows you to select which server will execute
the join when joining tables from different servers.

459
Microsoft SQL Server 2000 Programming by Example

Listing 11.23 Use JOIN Optimizer Hints to Force One Specific Join Type

-- Force LOOP JOIN for the first JOIN,

-- MERGE JOIN for the second JOIN,
-- and HASH JOIN for the third JOIN

SELECT CategoryName,
ProductName,
OrderDate,
Quantity
FROM [Order Details] OD
INNER LOOP JOIN Orders O
ON O.OrderID = OD.OrderID
INNER MERGE JOIN Products P
ON P.ProductID = OD.ProductID
INNER HASH JOIN Categories C
ON C.CategoryID = P.CategoryID
WHERE OrderDate
BETWEEN '1997-01-01'
AND '1997-03-31'
SQL Server selects automatically the order of joins, but you can specify a join order by using the FORCE
ORDER optimizer hint.
You can control how to group data by using the HASH and ORDER hints of the GROUP BY or COMPUTE clause.
To select how to execute the UNION operator, you can select the MERGE, HASH, or CONCAT hints.
To create a query plan that will execute efficiently with a bigger number of rows that will allow for rapid growth,
you can use the ROBUST PLAN optimizer hint.
The user usually perceives speed as how fast the server retrieves the first set of rows. SQL Server tries to
optimize the query as a complete set, and this can produce slow perceived speed. You can use the FAST n
to speed up the retrieval of the first n rows.
Caution

You should not use the FASTFIRSTROW optimizer hint because it is maintained only for backward
compatibility. Use FAST 1 instead.

Tip

Use the FAST n optimizer hint, as in Listing 11.24, for gridlike client applications, and FAST 1 for
single-record, form-based client applications.

Listing 11.24 Use the FAST Optimizer Hint to Improve the Response Time of Your Client Applications

460
 Chapter 11. Using Complex Queries and Statements

-- Speed up the retrieval of the first 10 rows

SELECT ProductName,
OrderDate,
Quantity
FROM [Order Details] OD
INNER JOIN Orders O
ON O.OrderID = OD.OrderID
INNER JOIN Products P
ON P.ProductID = OD.ProductID
WHERE OrderDate
BETWEEN '1997-01-01'
AND '1997-03-31'
OPTION (FAST 10)
Query Optimizer selects the best indexes to use to execute the query effi ciently, based on available statistics.
In some cases, you can find a better query plan than Query Optimizer. In these cases, you can force the use
of specific indexes by using the INDEX optimizer hint for every table (see Listing 11.25).

Listing 11.25 Use the INDEX Optimizer Hint to Force the Selection of Specific Indexes

-- This query uses a Clustered Index Scan (Table scan)

SELECT DISTINCT CustomerID,

OrderDate
FROM Orders
WHERE OrderDate
BETWEEN '1997-01-01'
AND '1997-03-31'

-- This query uses an Index Seek

-- On the OrderDate index

SELECT DISTINCT CustomerID,

OrderDate
FROM Orders (Index (OrderDate))
WHERE OrderDate
BETWEEN '1997-01-01'
AND '1997-03-31'
Caution

You can use the optimizer hint (INDEX (index1, index2, ...)) to force using a specific
index. The (INDEX = index1) syntax is maintained for backward compatibility.

461
Microsoft SQL Server 2000 Programming by Example

SQL Server selects automatically the correct locking strategy for every query. In some cases, you might have
specific locking needs. For these cases, you can specify lock-optimizer hints, such as HOLDLOCK,
SERIALIZABLE, PAGLOCK, UPDLOCK, and XLOCK. Locks will be covered in Chapter 13, "Maintaining
Data Consistency: Transactions and Locks."

Caution

Always document the reasons for using an optimizer hint and test these reasons from time to time
because, as volume of data increases, they might be invalid.

Note

We want to finish this chapter with a familiar note: Keep it simple.

Write your queries having performance and maintenance in your mind. Complex queries are
difficult to optimize and difficult to maintain.

Query Optimizer usually does an excellent job. Use optimizer hints only when necessary, and
document your reasons.

What's Next?

In this chapter, you learned the use of advanced query structures.

In Chapter 12, "Row-Oriented Processing: Us ing Cursors," you will learn how to use cursors to solve
complex programming needs.
Chapter 13 covers transactions and locks. Understanding them provides the basis to improve concurrency in
your database.
Working in a heterogeneous environment is a common scenario for a database programmer. Chapter 14,
"Transferring Data to and from SQL Server," shows how to transfer data from different databases, and
Chapter 15, "Working with Heterogeneous Environments: Setting Up Linked Servers," introduces
the concept of distributed data by using linked servers.
Delivered for Nenad Apostoloski Last updated on 1/18/2002
Swap Option Available: 2/24/2002 Microsoft SQL Server 2000 Programming by Example, © 2002 Que

462
 Chapter 12. Row-Oriented Processing: Using Cursors

Chapter 12. Row-Oriented Processing: Using Cursors

In previous chapters you learned how SQL Server processes complete result sets. SQL Server is optimized to
work with operations that affect complete result sets, and Query Optimizer decides in which order to process
the rows to get the job done in the most efficient way.
In some cases, you would be required to process individual rows from a result set in a specific order and, in
these cases, you can use cursors. SQL Server supports Transact-SQL cursors and application cursors.
This chapter teaches you

• The differences between set- and row-oriented processing

• The type of cursors and when to use them
• How to implement Transact-SQL cursors
• The scope of cursors
• How to use cursor variables
• How to use cursors inside triggers to apply row-by -row processing in a
• multirow operation
• How to use application cursors

Row-by-Row Versus Set-Oriented Processing

You can apply a specific business process to a group of rows, in two completely different ways:

• Navigate the result set in the way you prefer, and apply the business process to every row individually,
sending one or more Transact-SQL statements to SQL Server for every row.
• Send to SQL Server a single Transact-SQL statement that describes how to apply the business
process to the entire result set, and let SQL Server decide how to apply it, in the optimal way.

I'll explain the difference between these two ways with a practical example.
Consider it is the end of the year and you want to increase product prices by 5%. The manual process will
involve modifying each product's UnitPrice. The automatic process cannot be much different from this,
because the final result must be new UnitPrice values 5% more expensive than before.
If you think about how to do this operation from a client application, such as Visual Basic, you could write a
loop to retrieve one product at a time and modify its UnitPrice. Listing 12.1 shows how to use the
ActiveX Data Object (ADO) library to access SQL Server and apply these changes row by row.

Listing 12.1 Access to SQL Server Row by Row from Visual Basic and ADO

Dim conNorthwind As ADODB.Connection

Dim rstProducts As ADODB.Recordset

Set conNorthwind = New ADODB.Connection

conNorthwind.Provider = "SQLOLEDB"

conNorthwind.Open "Server=MSSQLFGG\s2k;UID=sa;PWD=;Database=Northwind;"

Set rstProducts = New ADODB.Recordset

463
Microsoft SQL Server 2000 Programming by Example

rstProducts.Open "select ProductID, UnitPrice FROM Products", _

conNorthwind, adOpenForwardOnly, adLockPessimistic

On Error GoTo CancelChanges

conNorthwind.BeginTrans

While Not rstProducts.EOF

rstProducts.Fields("UnitPrice").Value = _
rstProducts.Fields("UnitPrice").Value * 1.05
rstProducts.Update

rstProducts.MoveNext
Wend

conNorthwind.CommitTrans

GoTo CloseObjects

CancelChanges:
conNorthwind.RollbackTrans

CloseObjects:

rstProducts.Close
conNorthwind.Close

Set rstProducts = Nothing

Set conNorthwind = Nothing
However, SQL Server can achieve this operation with a single UPDATE statement, as in Listing 12.2.

Listing 12.2 Convert a Cursor Operation into a Single Transact-SQL Statement

-----------------------
-- From Query Analyzer
-----------------------

UPDATE Poducts
SET UnitPrice = UnitPrice * 1.05

----------------------
-- From Visual Basic
----------------------

Dim conNorthwind As ADODB.Connection

Set conNorthwind = New ADODB.Connection

conNorthwind.Provider = "SQLOLEDB"

conNorthwind.Open "Server=MSSQLFGG\s2k;UID=sa;PWD=;Database=Northwind;"

464
 Chapter 12. Row-Oriented Processing: Using Cursors

conNorthwind.Execute "UPDATE Products SET UnitPrice =UnitPrice / 1.05"

conNorthwind.Close

Set conNorthwind = Nothing

Using Profiler, you can inspect which statements are sent to SQL Server by the Visual Basic client application.
In Listing 12.1, you can see the output from Listing 12.3 (the output has been simplified to avoid repetitive
lines).

Listing 12.3 Execution Trace of the Visual Basic Example from Listing 12.1

TraceStart
SQL:StmtCompleted select ProductID, UnitPrice FROM Products 61
RPC:Completed declare @P1 int
set @P1=180150000
declare @P2 int
set @P2=2
declare @P3 int
set @P3=2
declare @P4 int
set @P4=-1
exec sp_cursoropen @P1 output, N'select ProductID, UnitPrice FROM Products', @P2
output,
@P3 output, @P4 output
select @P1, @P2, @P3, @P4 61
SQL:StmtCompleted set implicit_transactions on 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
SQL:StmtCompleted UPDATE [Products] SET [UnitPrice]=@Param000004 61
RPC:Completed exec sp_cursor 180150000, 33, 1, N'Products', @UnitPrice =
$19.8450 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
SQL:StmtCompleted UPDATE [Products] SET [UnitPrice]=@Param000004 61
RPC:Completed exec sp_cursor 180150000, 33, 1, N'Products', @UnitPrice =
$20.9475 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
SQL:StmtCompleted UPDATE [Products] SET [UnitPrice]=@Param000004 61
... (output truncated here)
RPC:Completed exec sp_cursor 180150000, 33, 1, N'Products', @UnitPrice =
$8.5444 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
SQL:StmtCompleted UPDATE [Products] SET [UnitPrice]=@Param000004 61
RPC:Completed exec sp_cursor 180150000, 33, 1, N'Products', @UnitPrice =
$19.8450 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
SQL:StmtCompleted UPDATE [Products] SET [UnitPrice]=@Param000004 61
RPC:Completed exec sp_cursor 180150000, 33, 1, N'Products', @UnitPrice =
$14.3325 61
RPC:Completed exec sp_cursorfetch 180150000, 32, 1, 1 61
RPC:Completed exec sp_cursorclose 180150000 61
SQL:StmtCompleted IF @@TRANCOUNT > 0 61

465
Microsoft SQL Server 2000 Programming by Example

SQL:StmtCompleted COMMIT TRAN 61

SQL:StmtCompleted 61
TraceStop
However, as you can see in Listing 12.4, the trace produced by the example of Listing 12.2 ismuch simpler.

Listing 12.4 Execution Trace of the Example from Listing 12.2

TraceStart
SQL:StmtCompleted UPDATE Products SET UnitPrice = UnitPrice / 1.05 61
TraceStop
The difference between the outputs in Listing 12.3 and Listing 12.4 has a tremendous importance in terms
of network traffic and server overhead. As you can imagine, sending one UPDATE statement plus two calls to
stored procedures for every product cannot be as efficient as sending a single UPDATE statement for the
complete list of products.
You can create a Transact-SQL script that works in a similar way as the Visual Basic application in Listing
12.1. Listing 12.5 shows an example of how to convert the Visual Basic program into a Transact-SQL script
using cursors.
Note

Later in this chapter, you will learn the meaning of the statements used in Listing 12.5. The
intention of this example is only to show the similarities between Transact-SQL cursors and
Application cursors.

Listing 12.5 Using Transact-SQL Cursors to Apply Row-by-Row Changes

-- Declare host variables

DECLARE @ID int, @UnitPrice money

-- Declare the cursor

DECLARE MyProducts CURSOR LOCAL

FORWARD_ONLY
FOR
SELECT ProductID, UnitPrice
FROM Products

466
 Chapter 12. Row-Oriented Processing: Using Cursors

-- Open the cursor

OPEN MyProducts

-- Fetch the first row in the cursor

FETCH NEXT FROM MyProducts

INTO @ID, @UnitPrice
WHILE @@FETCH_STATUS = 0

-- While the fetch is successful

BEGIN

-- Update the current product

UPDATE Products
SET UnitPrice = @UnitPrice * 1.05
WHERE CURRENT OF MyProducts

-- Fetch next product

FETCH NEXT FROM MyProducts

INTO @ID, @UnitPrice

END

-- Close the cursor

CLOSE MyProducts

-- Deallocate the cursor

DEALLOCATE MyProducts
Listing 12.6 shows the execution trace of the example in Listing 12.5. You can see that the trace in Listing
12.6 is structurally similar to the output from Listing 12.3. Note that the trace has been truncated to avoid
too many repetitions.

Listing 12.6 Execution Trace of Listing 12.5 (Simplified)

TraceStart
SQL:StmtCompleted SELECT ProductID, UnitPrice FROM Products 52
SQL:StmtCompleted OPEN MyProducts 52
SQL:StmtCompleted FETCH NEXT FROM MyProducts INTO @ID, @UnitPrice 52
SQL:StmtCompleted WHILE @@FETCH_STATUS = 0 52
SQL:StmtCompleted UPDATE Products SET UnitPrice = @UnitPrice * 1.05 WHERE
CURRENT OF
MyProducts 52
SQL:StmtCompleted FETCH NEXT FROM MyProducts INTO @ID, @UnitPrice 52
SQL:StmtCompleted 52

467
Microsoft SQL Server 2000 Programming by Example

SQL:StmtCompleted WHILE @@FETCH_STATUS = 0 52

SQL:StmtCompleted UPDATE Products SET UnitPrice = @UnitPrice * 1.05 WHERE
CURRENT OF
MyProducts 52
SQL:StmtCompleted FETCH NEXT FROM MyProducts INTO @ID, @UnitPrice 52
SQL:StmtCompleted 52
... (Trace truncated)
SQL:StmtCompleted WHILE @@FETCH_STATUS = 0 52
SQL:StmtCompleted UPDATE Products SET UnitPrice = @UnitPrice * 1.05 WHERE
CURRENT OF
MyProducts 52
SQL:StmtCompleted FETCH NEXT FROM MyProducts INTO @ID, @UnitPrice 52
SQL:StmtCompleted 52
SQL:StmtCompleted WHILE @@FETCH_STATUS = 0 52
SQL:StmtCompleted UPDATE Products SET UnitPrice = @UnitPrice * 1.05 WHERE
CURRENT OF
MyProducts 52
SQL:StmtCompleted FETCH NEXT FROM MyProducts INTO @ID, @UnitPrice 52
SQL:StmtCompleted 52
SQL:StmtCompleted WHILE @@FETCH_STATUS = 0 52
SQL:StmtCompleted CLOSE MyProducts 52
SQL:StmtCompleted DEALLOCATE MyProducts 52
TraceStop
It is obvious that SQL Server has to modify the UnitPrice column individually, regardless of the method
used to send the UPDATE. However, if you do not use cursors, Query Optimizer can find the best strategy to
modify this value without any constraints imposed by the way you navigate the cursor.
For example, if you want to retrieve information about orders placed by customers in the USA, and for every
order you want to retrieve the order date and the name of the customer, you can do it in the following way:

1. Open a cursor in the Customers table.

2. Navigate the Customer cursor row by row, searching for customers in the USA.
3. For every customer in the USA, open a cursor in the Orders table for orders from this specific
customer only.
4. Navigate the Orders cursor to show every order date.
5. After the last order of the current customer has been retrieved, you can go to the next customer and
start with step 2..

Perhaps this strategy looks similar to a loop join— actually, SQL Server follows a similar process to execute
loop joins. You can avoid using cursors to solve this problem by sending a simple query to SQL Server joining
the Customers and Orders tables. If you send a standard join, Query Optimizer will have the final decision
about which type of join is more appropriate to solve this particular query.
Listing 12.7 provides two exampleson how to solve this problem using Transact-SQL language. The first
example does not use cursors, and the second example does.

Listing 12.7 Solving a Business Problem With and Without Cursors

------------------
-- Without cursors
------------------

SELECT CompanyName, OrderDate

468
 Chapter 12. Row-Oriented Processing: Using Cursors

FROM Customers
JOIN Orders
ON Customers.CustomerID = Orders.CustomerID
WHERE Country = 'USA'
ORDER BY CompanyName, OrderDate

----------------
-- Using cursors
----------------

-- Declare host variables

DECLARE @ID nchar(5), @Name nvarchar(40),

@Country nvarchar(15), @OrderDate datetime

-- Declare products cursor

DECLARE MyCustomers CURSOR LOCAL

FOR
SELECT CustomerID, CompanyName, Country
FROM Customers

-- Open Customers Cursor

OPEN MyCustomers

-- Search for first customer

FETCH NEXT FROM MyCustomers

INTO @ID, @Name, @Country

WHILE @@FETCH_STATUS=0
BEGIN

IF @Country = 'USA'
BEGIN
-- Declare Orders cursor

DECLARE MyOrders CURSOR LOCAL

FOR
SELECT OrderDate
FROM Orders
WHERE CustomerID = @ID

-- Open Orders cursor

OPEN MyOrders

-- Search for first Order

FETCH NEXT FROM MyOrders

INTO @OrderDate

WHILE @@FETCH_STATUS=0
BEGIN

SELECT @Name AS 'Company Name',

@OrderDate AS 'Order Date'

-- Search for next Order

469
Microsoft SQL Server 2000 Programming by Example

FETCH NEXT FROM MyOrders

INTO @OrderDate

END

-- Close Orders cursor

CLOSE MyOrders

--Deallocate Orders cursor

DEALLOCATE MyOrders

END

-- Search for next Customer

FETCH NEXT FROM MyCustomers

INTO @ID, @Name, @Country

END
-- Close Customers cursor

CLOSE MyCustomers

-- Deallocate Customers cursor

DEALLOCATE MyCustomers
Tip

Use cursors as a last resort. First, consider whether you can achieve the same results without
using cursors.

To work with user-defined cursors, SQL Server must maintain specific memory structures and execute
complex operations not required in standard, cursorless operations. In other words, using cursors is
expensive for SQL Server. However, cursors are necessary to solve especially complex problems where
result set operations cannot provide an easy solution. Later in this chapter, you will see an example of how to
use cursors inside triggers to solve multirow operations, in which case, using cursors might be the only
available choice.

Types of Cursors

SQL Server 2000 supports four types of cursors:

• Forward-only cursors can retrieve data from the first to the last row, without any other navigation
capabilities.
• Static cursors provide a snapshot of the data to navigate without being affected by other connections.
• Dynamic cursors retrieve a dynamic result set that sees modifications to the data made from outside
the cursor.
• Keyset-driven cursors create a fixed set of rows to navigate.

470
 Chapter 12. Row-Oriented Processing: Using Cursors

Static cursors use more storage space than dynamic or keyset-driven, but SQL Server uses fewer resources
to navigate the data after their creation. Static cursors provide a fixed set of data that does not detect changes
made by other connections.
Dynamic cursors use little storage space, but SQL Server uses more resources to navigate the data after their
creation than static or keyset- driven cursors. Dynamic cursors provide a flexible set of data that reflects
changes made to the data by other connections.
Keyset-driven cursors balance storage space requirements with navigation resources. Keyset-driven cursors
provide a set of data with a fixed number of rows that reflects changes made by other connections.
Transact-SQL does not consider forward-only as a different kind of cursor, but as an optional property of the
other types of cursors.

Forward-Only

A forward-only cursor does not provide the option to scroll through the data. Using this type of cursor, you can
retrieve rows from the beginning of the result set to the end only.
Rows are not retrieved until they are requested by using the FETCH statement. For this reason, using this
cursor type, you can see changes made to the data by any connection to data that is not yet retrieved.
Changes to the data already retrieved cannot be seen because this cursor does not support scrolling
backward.
By default, static, dynamic, and keyset-driven cursors are scrollable, unless you specify the FORWARD_ONLY
keyword in the DECLARE CURSOR statement. Later in this chapter, you will learn about the syntax of the
DECLARE CURSOR statement.
Microsoft provides an improvement to the forward-only cursors with the fast forward-only cursor. This option
opens an optimized read-only forward-only cursor. To use this optimization, you must specify the
FAST_FORWARD in the DECLARE CURSOR statement.
In some cases, it is not valid to create a fast forward-only cursor, and SQL Server must convert the cursor to
the appropriate type. The following are some of these cases:

• The SELECT statement references BLOB columns, such as columns with text, ntext, or image data
types.
• The cursor is not open as read-only.
• The query references remote tables from linked servers.

Tip

If you have to use a cursor, try to use a fast forward-only cursor, because it uses fewer resources
in the server, providing a faster and more efficient operation.

Static

A static cursor provides a snapshot of the query. SQL Server builds the complete result set in TempDB when
the cursor is open, and remains there untouched until the cursor is closed. Because the information is
selected when the cursor is open, changes from other statements are not visible when you navigate the
cursor. The creation of the worktable takes not only storage space, but some time as well, providing a slower
start to the application that uses it.
Note

After the static cursor is populated, it contains a fixed number of rows with fixed column values until
the cursor is closed.

471
Microsoft SQL Server 2000 Programming by Example

When searching for rows, the cursor retrieves its information exclusively from the worktable created in
TempDB. This is why static cursors provide fast data navigation.

The cursor does not detect the insertion of new rows, the deletion of existing ones, or changes to column
values, unless you close the cursor and reopen it again.

If you create a static cursor on customers from London and retrieve 10 customers, you can navigate the rows
returned from that cursor, and you will always retrieve the same values for the same 10 customers—
regardless of any changes made to these customers from other statements in the same connection or
different connections. Listing 12.8 shows the effect of changes in data used in a static cursor. Where the
output does not show the change on ProductName, it still shows the row where ProductID = 75, after its
deletion by a DELETE statement, and it does not show the new inserted row.

Listing 12.8 You Have an Isolated Result Set Using a Static Cursor

USE Northwind
GO

-- Start a new transaction

-- so we can rollback the changes

BEGIN TRAN

DECLARE @ID int, @name nvarchar(40)

-- Declare the cursor

DECLARE MyProducts CURSOR STATIC

FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID
-- Open the cursor

OPEN MyProducts

-- Search for a new row from the cursor

FETCH NEXT FROM MyProducts

INTO @ID, @name

-- Update directly the Products table

UPDATE Products
SET ProductName = 'NewName'
WHERE ProductID > 70

472
 Chapter 12. Row-Oriented Processing: Using Cursors

-- Delete one row from the Products table

DELETE [Order details]

WHERE ProductID = 75

DELETE Products
WHERE ProductID = 75

-- Add a new row to the Products table

INSERT Products (ProductName)

VALUES ('New Product')

-- While the fetch is successful

WHILE @@FETCH_STATUS <> -1

BEGIN

IF @@FETCH_STATUS = -2
-- Row has been deleted
PRINT CHAR(10) + 'Missing Row'+ CHAR(10)
ELSE
-- Show the values from the cursor
SELECT @ID, @Name

-- Search for the new row

FETCH NEXT FROM MyProducts

INTO @ID, @Name
END
-- Close the cursor

CLOSE MyProducts

-- Destroy the cursor

DEALLOCATE MyProducts

-- Undo changes

ROLLBACK TRAN

----------- ----------------------------------------
71 Flotemysost

----------- ----------------------------------------
72 Mozzarella di Giovanni

----------- ----------------------------------------
73 Röd Kaviar

473
Microsoft SQL Server 2000 Programming by Example

----------- ----------------------------------------
74 Longlife Tofu

----------- ----------------------------------------
75 Rhönbräu Klosterbier

----------- ----------------------------------------
76 Lakkalikööri

----------- ----------------------------------------
77 Original Frankfurter grüne Soße
Note

Later in this chapter, you will see, step by step, how to use cursors, covering the statements
contained in Listing 12.8.

Note

Static cursors are read-only in SQL Server.

Note

In some documentation, you canfind the term snapshot or insensitive applied to the same concept
as static cursor.

Caution

Because the result of a static cursor must be stored in a worktable in TempDB, the total size of the
columns selected in the cursor cannot exceed the maximum number of bytes in a standard row.

If your application makes extensive use of static cursors, make sure you have enough free space
on TempDB.

Dynamic

474
 Chapter 12. Row-Oriented Processing: Using Cursors

A dynamic cursor reflects all the changes made by other connections. This type of cursor does not create a
table in TempDB. Because this cursor does not require creating any temporary data in TempDB, it opens
faster and uses less storage space than static cursors.
Note

After the dynamic cursor is opened, it contains an undefined number of rows with potentially
changing column values until the cursor is closed.

Every time you need to fetch a new row, SQL Server must execute the required query plan to select the new
row, and that operation takes some time, providing a slower navigation than static cursors.

Any changes to existing data, either by inserting new rows, deleting existing ones, or changing values in
columns will be reflected in this type of cursor. Listing 12.9 shows how the dynamic cursor is affected by
changes in the underlying data, where the output

• Shows the change on ProductName

• Hides the deleted row where ProductID = 75
• Shows the newly inserted row where ProductName = 'NewProduct'

Listing 12.9 Using Dynamic Cursors to See Changes from Other Connections— Same As Listing
12.8 Except

-- Declare the cursor

DECLARE MyProducts CURSOR DYNAMIC

FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID

----------- ----------------------------------------
71 Flotemysost

----------- ----------------------------------------

475
Microsoft SQL Server 2000 Programming by Example

72 NewName

----------- ----------------------------------------
73 NewName

----------- ----------------------------------------
74 NewName

----------- ----------------------------------------
76 NewName

----------- ----------------------------------------
77 NewName

----------- ----------------------------------------
88 New Product
Changes to the data made in the same connection are always visible to the dynamic cursor. However,
changes made by other connections are not visible until the transactions that modified the data are committed,
unless you specify a Read Uncommitted isolation level for the transaction holding the cursor.

Keyset-Driven Cursors

To solve some of theproblems of both static and dynamic cursors, you can create a keyset-driven cursor.This
type of cursor creates in TempDB a list of unique values, called the keyset, from the original query, where
every key uniquely identifies one single row in the result set. This keyset contains bookmarks to the actual
data.
Because the keyset is completely built when the cursor is open, the number of rows is fixed until the cursor is
closed. However, if some rows are deleted from the underlying tables, you will get a "row missing" message
trying to fetch them, because the bookmarks of these cursor rows will point to an invalid location.
Rows inserted in the data that could be considered part of the cursor result set are not visible unless you
close the cursor and reopen it again. Changes to column values made by any connections are visible inside
the cursor, as long as the modified columns are not part of the keyset.
Listing 12.10 shows the same script as in Listing 12.8 and 12.9, but in this case the cursor is declared as
KEYSET. As you can see in Listing 12.10, the output

• Shows the change on ProductName

• Produces an error message for the deleted row where ProductID = 75
• Does not show the newly inserted row where ProductName = 'NewProduct'

Listing 12.10 Using Keyset Cursors— Same As Listing 12.8 Except

-- Declare the cursor

DECLARE MyProducts CURSOR KEYSET
FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID

476
 Chapter 12. Row-Oriented Processing: Using Cursors

----------- ----------------------------------------
71 Flotemysost

----------- ----------------------------------------
72 NewName

----------- ----------------------------------------
73 NewName

----------- ----------------------------------------
74 NewName

Missing Row

----------- ----------------------------------------
76 NewName
----------- ----------------------------------------
77 NewName

Steps to Use Cursors

As shown in Listing 12.8, to use a cursor you must follow this sequence:

1. Use the DECLARE statement to declare the cursor. This step specifies the type of cursor and the
query that defines the data to retrieve. SQL Server creates the memory structures that support the
cursor. No data is retrieved yet.
2. Execute the OPEN statement to open the cursor. In this step, SQL Server executes the query specified
in the cursor definition and prepares the data for further navigation.
3. Execute the FETCH statement to search for rows. In this step, you move the cursor pointer to the
required row and, optionally, retrieve column values into variables. Repeat this step as many times as
necessary to complete the required task. Optionally, you can modify data through the cursor, unless
the cursor is read-only.
4. Execute the CLOSE statement to close the cursor when the data contained is no longer necessary.
The cursor is still created, but it does not contain any data. To retrieve the data again, you must
execute the OPEN statement to open the cursor again.
5. Execute the DEALLOCATEstatement to drop the cursor when you don't have intentions of reusing the
cursor any longer.

The following sections look at these steps in more detail.

Declaring Cursors

To declare a cursor, you must use the DECLARE CURSOR statement. A simplified syntax of the DECLARE
CURSOR statement could be

DECLARE CursorName CURSOR

[CursorScope]
[CursorBehavior]

477
Microsoft SQL Server 2000 Programming by Example

[CursorType]
[CursorLock]
[TYPE_WARNING]
FOR SelectStatement
[UpdateOfColumns]
Note

SQL Server 2000 accepts the SQL-92 DECLARE CURSOR syntax as well.

Every cursor must have a name, and this name is a Transact-SQL identifier that must follow the same
guidelines as for other object identifiers.

After the cursor name, you can specify whether the cursor is GLOBAL to the connection or LOCAL to the batch,
stored procedure, trigger, or user-defined function where the cursor is created. Cursor scope is covered in
more detail later in this chapter.

You can control cursor behavior by using two keywords: FORWARD_ONLY or SCROLL (see Listings 12.11
and 12.12). The default cursor behavior is FORWARD_ONLY, which means that the cursor can move only
forward, using the FETCH NEXT statement, row by row to the end of the result set.

Listing 12.11 Using the FORWARD_ONLY Keyword

-- This is a LOCAL FORWARD_ONLY cursor

DECLARE MyProducts CURSOR

LOCAL FORWARD_ONLY
FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID ASC
Declaring a cursor as SCROLL enables you to use any of the FETCH statements. Later in this chapter, you will
learn the different FETCH options.

Listing 12.12 Using the SCROLL Keyword

478
 Chapter 12. Row-Oriented Processing: Using Cursors

-- This is a GLOBAL SCROLL cursor

DECLARE MyProducts CURSOR

GLOBAL SCROLL
FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductName DESC
Caution

FORWARD_ONLY and FAST_FORWARD are mutually exclusive. Declaring a cursor using both
keywords produces a syntax error.

Cursor types have been explained earlier in this chapter. To define the cursor type, you can use the STATIC,
KEYSET, DYNAMIC, or FAST_FORWARD keywords, as seen in Listings 12.1 to 12.3.

Caution

SCROLL and FAST_FORWARD are mutually exclusive. Declaring a cursor using both keywords
produces a syntax error.

You can control how to lock the cursor data using the READ_ONLY, SCROLL_LOCKS, and OPTIMISTIC
keywords in the DECLARE CURSOR statement.

Declaring a cursor as READ_ONLY prevents updates to the cursor using the UPDATE or DELETE statements
with the WHERE CURRENT OF clause (see Listing 12.13). Using the SCROLL_LOCKS option forces the data
to be locked when the cursor reads it, to guarantee potential updates. This locking behavior is often called
pessimistic locking.

Using the OPTIMISTIC option frees the lock on the data after the data is loaded into the cursor, and will lock
the data only to update or delete, if required. In this case, SQL Server must check whether the row has been
modified by other connections between the reading and writing operations.

479
Microsoft SQL Server 2000 Programming by Example

SQL Server can check for changes on the data by inspecting the actual value of a timestamp (rowversion)
column and comparing this value to the value obtained during the read operation. If the data does not contain
any timestamp column, SQL Server can use a checksum of the existing column values.

Tip

You can increase the concurrency of your application by selecting OPTIMISTIC concurrency.

STATIC and FAST_FORWARD cursors default to READ_ONLY locking. However, KEYSET and DYNAMIC cursors
default to OPTIMISTIC locking.

Caution

SQL Server 2000 does not support OPTIMISTIC concurrency in a FAST_FORWARD cursor.

Listing 12.13 Use the READ_ONLY Option to Protect the Cursor from Updates

BEGIN TRAN

-- Declare the cursor

DECLARE MyProducts CURSOR

FORWARD_ONLY READ_ONLY
FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID

-- Open the cursor

OPEN MyProducts

-- Fetch the first row

FETCH NEXT FROM MyProducts

-- Try to update the data

-- in the current row

480
 Chapter 12. Row-Oriented Processing: Using Cursors

-- gives an error
-- on a READ_ONLY cursor

update Products
set productname = 'Modified name'
where current of MyProducts

-- Close the cursor

CLOSE MyProducts

-- Deallocate the cursor

DEALLOCATE MyProducts

ROLLBACK TRAN

ProductID ProductName
----------- ----------------------------------------
71 Flotemysost

(1 row(s) affected)

Server: Msg 16929, Level 16, State 1, Line 26

The cursor is READ ONLY.
The statement has been terminated.
In some cases, the cursor type must be changed because of restrictions in the definition of the cursor, as
mentioned before in this chapter. In this case, you can get a notification of this change by using the
TYPE_WARNING option (see Listing 12.14).

Listing 12.14 Using the TYPE_WARNING Option

-- Declare the cursor

-- as FAST_FORWARD
-- but it is converted into KEYSET
-- because it uses ntext fields
-- and ORDER BY

DECLARE MyCategories CURSOR

FAST_FORWARD READ_ONLY
TYPE_WARNING
FOR
SELECT CategoryID, CategoryName, Description

481
Microsoft SQL Server 2000 Programming by Example

FROM Categories
ORDER BY CategoryName ASC

-- Open the cursor

OPEN MyCategories

-- Fetch the first row

FETCH NEXT FROM MyCategories

-- Close the cursor

CLOSE MyCategories

-- Deallocate the cursor

DEALLOCATE MyCategories
Cursor created was not of the requested type.

CategoryID CategoryName Description

----------- --------------- -------------------------------------------
1 Beverages Soft drinks, coffees, teas, beers, and ales
The cursor must be defined for a SELECT statement. This is a normal SELECT statement with a few
exceptions. You cannot use COMPUTE, COMPUTE BY, FOR BROWSE, or INTO in a SELECT statement that
defines a cursor.
Caution

If the SELECT statement produces a result set that is not updatable, the cursor will be READ_ONLY.
This can happen because of the use of aggregate functions, insufficient permissions, or retrieving
read-only data.

You can restrict the columns to update inside the cursor using the FOR UPDATE clause, as shown in Listing
12.15. This clause can be used in two ways:

• FOR UPDATE OF Column1, ..., ColumnN—Use this option to define columns Column1 to
ColumnN as updatable through the cursor.
• FOR UPDATE—This is the default option, and it declares all the cursor columns as updatable.

Listing 12.15 Using the FOR UPDATE Clause

482
 Chapter 12. Row-Oriented Processing: Using Cursors

DECLARE MyCategories CURSOR

KEYSET
FOR
SELECT CategoryID, CategoryName, Description
FROM Categories
ORDER BY CategoryName ASC
FOR UPDATE OF CategoryName, Description
Note

When you declare a cursor, SQL Server creates some memory structures to use the cursor, but the
data is not retrieved until you open the cursor.

Opening Cursors

To use a cursor, you must open it. You can open a cursor using the OPEN statement. If the cursor was
declared as STATIC or KEYSET, SQL Server must create a worktable in TempDB to store either the full result
set, in a STATIC cursor, or the keyset only in a keyset-driven cursor. In these cases, if the worktable cannot
be created for any reason, the OPEN statement will fail.
SQL Server can optimize the opening of big cursors by populating the cursor asynchronously. In this case,
SQL Server creates a new thread to populate the worktable in parallel, returning the control to the application
as soon as possible.
You can use the @@CURSOR_ROWS system function to control how many rows are contained in the cursor. If
the cursor is using asynchronous population, the value returned by @@CURSOR_ROWS will be negative and
represents the approximate number of rows returned since the opening of the cursor.
For dynamic cursors, @@CURSOR_ROWS returns -1, because it is not possible to know whether the full result
set has been returned already, because of potential insertions by other operations affecting the same data.
Caution

The @@CURSOR_ROWS function returns the number of rows of the last cursor opened in the current
connection. If you use cursors inside triggers, the result of this function from the main execution
level could be misleading. Listing 12.16 shows an example of this problem.

To specify when SQL Server will decide to populate a cursor asynchronously, you can use the
sp_configure system-stored procedure to change the server setting "cursor threshold", specifying
the maximum number of rows that will be executed directly without asynchronous population.

Caution

483
Microsoft SQL Server 2000 Programming by Example

Do not fix the "cursor threshold" value too low, because small result sets are more efficiently
opened synchronously.

Listing 12.16 Using the @@CURSOR_ROWS System Function

-- Create a procedure to open

-- a cursor on Categories

CREATE PROCEDURE GetCategories

AS
DECLARE MyCategories CURSOR STATIC
FOR
SELECT CategoryID, CategoryName
FROM Categories

OPEN MyCategories

-- Shows the number of rows in the cursor

SELECT @@CURSOR_ROWS 'Categories cursor rows after open'

CLOSE MyCategories

DEALLOCATE MyCategories
GO

-- Create a cursor on Products

DECLARE MyProducts CURSOR STATIC

FOR
SELECT ProductID, ProductName
FROM Products

OPEN MyProducts

-- Shows the number of rows in the last opened cursor,

which is MyProducts

SELECT @@CURSOR_ROWS 'Products cursor rows'

EXEC GetCategories
-- Shows the number of rows in the last opened cursor
-- in the current connection, which is MyCategories

SELECT @@CURSOR_ROWS 'Categories cursor rows after close and deallocated'

CLOSE MyProducts

484
 Chapter 12. Row-Oriented Processing: Using Cursors

DEALLOCATE MyProducts

Products cursor rows

--------------------
77

Categories cursor rows after open

---------------------------------
8

Categories cursor rows after close and deallocated

--------------------------------------------------
0

Fetching Rows

You can use the FETCH statement to navigate an open cursor, as shown in Listing 12.17. Every time you
execute the FETCH statement, the cursor moves to a different row.
FETCH FROM CursorName retrieves the next row in the cursor. This is a synonym of FETCH NEXT FROM
CursorName. If the FETCH statement is executed right after the OPEN statement, the cursor is positioned in
the first row. If the current row is the last one in the result set, executing FETCH NEXT again will send the
cursor beyond the end of the result set and will return an empty row, but no error message will be produced.
Caution

After opening a cursor with the OPEN statement, the cursor does not point to any specific row, so
you must execute a FETCH statement to position the cursor in a valid row.

FETCH PRIOR moves the cursor to the preceding row. If the cursor was positioned already at the beginning of
the result set, using FETCH PRIOR will move the pointer before the starting of the result set, retrieving an
empty row, but no error message will be produced.

FETCH FIRST moves the cursor pointer to the beginning of the result set, returning the first row.

FETCH LAST moves the cursor pointer to the end of the result set, returning the last row.

FETCH ABSOLUTE n moves the cursor pointer to the n row in the result set. If n is negative, the cursor
pointer is moved n rows before the end of the result set. If the new row position does not exist, an empty row
will be returned and no error will be produced. If n is 0, no rows are returned and the cursor pointer goes out
of scope.

485
Microsoft SQL Server 2000 Programming by Example

FETCH RELATIVE n moves the cursor pointer n rows forward from the current position of the cursor. If n is
negative, the cursor pointer is moved backward n rows from the current position. If the new row position does
not exist, an empty row will be returned and no error will be produced. If n is 0, the current row is returned.

You can use the @@FETCH_STATUS system function to test whether the cursor points to a valid row after the
last FETCH statement. @@FETCH_SATUS can have the following values:

• 0 if the FETCH statement was successful and the cursor points to a valid row.
• -1 if the FETCH statement was not successful or the cursor points beyond the limits of the result set.
This can be produced using FETCH NEXT from the last row or FETCH PRIOR from the first row.
• -2 the cursor is pointing to a nonexistent row. This can be produced by a keyset-driven cursor when
one of the rows has been deleted from outside the control of the cursor.

Caution

@@FETCH_STATUS is global to the connection, so it reflects the status of the latest FETCH
statement executed in the connection. That is why it is important to test it right after the FETCH
statement.

Listing 12.17 Use FETCH to Navigate the Cursor

DECLARE MyProducts CURSOR STATIC

FOR
SELECT ProductID, ProductName
FROM Products
ORDER BY ProductID ASC

OPEN MyProducts

SELECT @@CURSOR_ROWS 'Products cursor rows'

SELECT @@FETCH_STATUS 'Fetch Status After OPEN'

FETCH FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After first FETCH'

FETCH NEXT FROM MyProducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH NEXT'

FETCH PRIOR FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH PRIOR'

486
 Chapter 12. Row-Oriented Processing: Using Cursors

FETCH PRIOR FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH PRIOR the first row'

FETCH LAST FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH LAST'

FETCH NEXT FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH NEXT the last row'

FETCH ABSOLUTE 10 FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH ABSOLUTE 10'

FETCH ABSOLUTE -5 FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH ABSOLUTE -5'

FETCH RELATIVE -20 FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH RELATIVE -20'

FETCH RELATIVE 10 FROM Myproducts

SELECT @@FETCH_STATUS 'Fetch Status After FETCH RELATIVE 10'

CLOSE MyProducts

SELECT @@FETCH_STATUS 'Fetch Status After CLOSE'

DEALLOCATE MyProducts

Products cursor rows

--------------------
77

Fetch Status After OPEN

-----------------------
0

ProductID ProductName
----------- ----------------------------------------
1 Chai

Fetch Status After first FETCH

------------------------------
0

ProductID ProductName

487
Microsoft SQL Server 2000 Programming by Example

----------- ----------------------------------------
2 Chang

Fetch Status After FETCH NEXT

-----------------------------
0

ProductID ProductName
----------- ----------------------------------------
1 Chai

Fetch Status After FETCH PRIOR

------------------------------
0

ProductID ProductName
----------- ----------------------------------------

Fetch Status After FETCH PRIOR the first row

--------------------------------------------
-1

ProductID ProductName
----------- ----------------------------------------
77 Original Frankfurter grüne Soße

Fetch Status After FETCH LAST

-----------------------------
0

ProductID ProductName
----------- ----------------------------------------

Fetch Status After FETCH NEXT the last row

------------------------------------------
-1

ProductID ProductName
----------- ----------------------------------------
10 Ikura

Fetch Status After FETCH ABSOLUTE 10

------------------------------------
0

ProductID ProductName
----------- ----------------------------------------
73 Röd Kaviar

Fetch Status After FETCH ABSOLUTE -5

------------------------------------
0

ProductID ProductName
----------- ----------------------------------------
53 Perth Pasties

Fetch Status After FETCH RELATIVE -20

-------------------------------------
0

488
 Chapter 12. Row-Oriented Processing: Using Cursors

ProductID ProductName
----------- ----------------------------------------
63 Vegie-spread

Fetch Status After FETCH RELATIVE 10

------------------------------------
0

Fetch Status After CLOSE

------------------------
0
At the same time you are moving the cursor with the FETCH statement, you can use the INTO clause to
retrieve the cursor fields directly into user-defined variables (see Listing 12.18). In this way, you later can
use the values stored in these variables in further Transact-SQL statements.

Listing 12.18 Use FETCH INTO to Get the Values of the Cursor Columns into Variables

SET NOCOUNT ON
GO

DECLARE @ProductID int,

@ProductName nvarchar(40),
@CategoryID int

DECLARE MyProducts CURSOR STATIC

FOR
SELECT ProductID, ProductName, CategoryID
FROM Products
WHERE CategoryID BETWEEN 6 AND 8
ORDER BY ProductID ASC

OPEN MyProducts

FETCH FROM Myproducts

INTO @ProductID, @ProductName, @CategoryID

WHILE @@FETCH_STATUS = 0
BEGIN

SELECT @ProductName as 'Product',

CategoryName AS 'Category'
FROM Categories
WHERE CategoryID = @CategoryID

FETCH FROM Myproducts

INTO @ProductID, @ProductName, @CategoryID

END

CLOSE MyProducts

DEALLOCATE MyProducts

489
Microsoft SQL Server 2000 Programming by Example

Product Category
---------------------------------------- ---------------
Uncle Bob's Organic Dried Pears Produce
Product Category
---------------------------------------- ---------------
Mishi Kobe Niku Meat/Poultry

Product Category
---------------------------------------- ---------------
Ikura Seafood

Product Category
---------------------------------------- ---------------
Konbu Seafood

Product Category
---------------------------------------- ---------------
Tofu Produce

Product Category
---------------------------------------- ---------------
Alice Mutton Meat/Poultry

Product Category
---------------------------------------- ---------------
Carnarvon Tigers Seafood

Product Category
---------------------------------------- ---------------
Rössle Sauerkraut Produce

Product Category
---------------------------------------- ---------------
Thüringer Rostbratwurst Meat/Poultry

Product Category
---------------------------------------- ---------------
Nord-Ost Matjeshering Seafood

Product Category
---------------------------------------- ---------------
Inlagd Sill Seafood

Product Category
---------------------------------------- ---------------
Gravad lax Seafood
Product Category
---------------------------------------- ---------------
Boston Crab Meat Seafood

Product Category

490
 Chapter 12. Row-Oriented Processing: Using Cursors

---------------------------------------- ---------------
Jack's New England Clam Chowder Seafood

Product Category
---------------------------------------- ---------------
Rogede sild Seafood

Product Category
---------------------------------------- ---------------
Spegesild Seafood

Product Category
---------------------------------------- ---------------
Manjimup Dried Apples Produce

Product Category
---------------------------------------- ---------------
Perth Pasties Meat/Poultry

Product Category
---------------------------------------- ---------------
Tourti\'e8re Meat/Poultry

Product Category
---------------------------------------- ---------------
Pâté chinois Meat/Poultry

Product Category
---------------------------------------- ---------------
Escargots de Bourgogne Seafood

Product Category
---------------------------------------- ---------------
Röd Kaviar Seafood

Product Category
---------------------------------------- ---------------
Longlife Tofu Produce

If the cursor is updatable, you can modify values in the underlying tables sending standard UPDATE or
DELETE statements and specifying WHERE CURRENT OF CursorName as a restricting condition (see
Listing 12.19).

Listing 12.19 Using WHERE CURRENT OFto Apply Modifications to the Current Cursor Row

BEGIN TRAN

-- Declare the cursor

DECLARE MyProducts CURSOR

FORWARD_ONLY
FOR

491
Microsoft SQL Server 2000 Programming by Example

SELECT ProductID, ProductName

FROM Products
WHERE ProductID > 70
ORDER BY ProductID

-- Open the cursor

OPEN MyProducts

-- Fetch the first row

FETCH NEXT FROM MyProducts

-- UPdate the name of the product

-- and the UnitPrice in the current cursor position

update Products
set ProductName = ProductName + '(to be dicontinued)',
UnitPrice = UnitPrice * (1.0 + CategoryID / 100.0)
where current of MyProducts

SELECT *
from Products

-- Close the cursor

CLOSE MyProducts

-- Deallocate the cursor

DEALLOCATE MyProducts

ROLLBACK TRAN
Note

You can update through cursor columns that are not part of the cursor definition, as long as the
columns are updatable

Closing Cursors

Use the CLOSE statement to close a cursor, freeing any locks used by it. The cursor structure is not destroyed,
but it is not possible to retrieve any data from the cursor after the cursor is closed.
Tip

It is a good practice to close cursors as soon as they are not necessary. This simple practice can
provide better concurrency to your application.

Most of the listings in this chapter use the CLOSE statement.

492
 Chapter 12. Row-Oriented Processing: Using Cursors

Deallocating Cursors

To destroy the cursor completely, you can use the DEALLOCATE statement. After this statement is executed, it
is not possible to reopen the cursor without redefining it again.
After DEALLOCATE you can reuse the cursor name to declare any other cursor, with identical or different
definition.
Tip

To reuse the same cursor in different occasions in a long batch or a complex stored procedure, you
should declare the cursor as soon as you need it and deallocate it when it is no longer necessary.
Between the DECLARE and DEALLOCATE statements, use OPEN and CLOSE to access data as
many times as necessary to avoid long-standing locks. However, consider that each time you open
the cursor the query has to be exec uted. This could produce some overhead.

Scope of Cursors
In the DECLARE CURSOR statement, you can specify the scope of the cursor after its name. The default scope
is GLOBAL, but you can change the default scope, changing the database option default to local cursor.
Caution

You should not rely on the default cursor scope of SQL Server. It is recommended that you declare
the cursor explicitly as either LOCAL or GLOBAL, because the default cursor scope might change in
future versions of SQL Server.

You can use a global cursor anywhere in the same connection in which the cursor was created, whereas local
cursors are valid only within the scope of the batch, procedure, user-defined function, or trigger where the
cursor is created. The cursor is automatically deallocated when it goes out of scope (see Listing 12.20).

Listing 12.20 Using Global Cursors

-- Declare the cursor as GLOBAL

DECLARE MyProducts CURSOR GLOBAL

FOR
SELECT ProductID, ProductName
FROM Products
WHERE ProductID > 70
ORDER BY ProductID
However, you can assign the cursor to an OUTPUT parameter in a stored procedure. In this case, the cursor
will be deallocated when the last cursor variable that references the cursor goes out of scope.
Note

493
Microsoft SQL Server 2000 Programming by Example

Cursor variables are covered later in this chapter.

Global and local cursors have two different name spaces, so it is possible to have a global cursor with the
same name as a local cursor, and they can have completely different definitions. To avoid potential problems,
SQL Server use local cursors.

Local Cursors

Local cursors are a safety feature that provides the creation of local cursors inside independent objects, such
as stored procedures, triggers, and user-defined functions. Local cursors are easier to manage than global
cursors because you do not have to consider potential changes to the cursor in other procedures or triggers
used by your application.

Global Cursors

Global cursors are useful in scenarios where different procedures must manage a common result set, and
they must dynamically interact with it. It is recommended you use local cursors whenever possible. If you
require sharing a cursor between two procedures, consider using a cursor variable instead, as is covered in
the next section.

Using Cursor Variables

It is possible to declare variables using the cursor data type, which is very useful if you need to send a
reference of your cursor to another procedure or user-defined function. Using cursor variables is similar to
using standard cursors (see Listing 12.21).

Listing 12.21 Using Cursor Variables

-- Declare the cursor variable

DECLARE @Products AS CURSOR

-- Assign the cursor variable a cursor definition

SET @Products = CURSOR STATIC

FOR
SELECT ProductID, ProductName
FROM Products

-- Open the cursor

OPEN @Products

-- Fetch the first cursor row

494
 Chapter 12. Row-Oriented Processing: Using Cursors

FETCH NEXT FROM @Products

-- Close the cursor

CLOSE @Products

-- Deallocate the cursor

DEALLOCATE @Products
SQL Server provides system stored procedures to retrieve information about cursors. These procedures use
cursor variables to communicate its data:

• sp_cursor_list produces a list of available cursors in the current connection.

• sp_describe_cursor retrieves the attributes of an open cursor. The out put is the same as the
output produced with sp_cursor_list, but sp_describe_cursor refers to a single cursor.
• sp_describe_cursor_columns describes the columns retrieved by the cursor.
• sp_describe_cursor_tables gets information about the tables used in the cursor.

These stored procedures use cursor variables to retrieve results. In this way, calling procedures and batches
can use the result one row at a time.
Listing 12.22 shows how to execute these system stored procedures to get information about cursors and
cursors variables.

Listing 12.22 Retrieving Information About Cursors with System Stored Procedures

USE Northwind
GO

-- Declare some cursors

DECLARE CCategories CURSOR LOCAL

DYNAMIC
FOR
SELECT CategoryName
FROM Categories

DECLARE CCustomers CURSOR LOCAL

FAST_FORWARD
FOR
SELECT CompanyName
FROM Customers

DECLARE COrdersComplete CURSOR GLOBAL

KEYSET
FOR
SELECT O.OrderID, OrderDate,
C.CustomerID, CompanyName,
P.ProductID, ProductName,
Quantity, OD.UnitPrice, Discount
FROM Orders O
JOIN [Order Details] OD

495
Microsoft SQL Server 2000 Programming by Example

ON OD.OrderID = O.OrderID
JOIN Customers C
ON C.CustomerID = O.CustomerID
JOIN Products P
ON P.ProductID = OD.ProductID

-- Declare a cursor variable to hold

-- results from the stored procedures
DECLARE @OutputCursor AS CURSOR

-- Get information about declared local cursors

EXEC sp_cursor_list @OutputCursor OUTPUT, 1

-- deallocate the cursor, so we can reuse the cursor variable

DEALLOCATE @OutputCursor

-- Or get information about declared global cursors

EXEC sp_cursor_list @OutputCursor OUTPUT, 2

-- deallocate the cursor, so we can reuse the cursor variable

DEALLOCATE @OutputCursor

-- Or get information about declared global and local cursors

-- note that status = -1 means cursor closed

PRINT CHAR(10) + 'sp_cursor_list cursor OUTPUT'+ CHAR(10)

EXEC sp_cursor_list @OutputCursor OUTPUT, 3

FETCH NEXT FROM @OutputCursor

WHILE @@FETCH_STATUS = 0
FETCH NEXT FROM @OutputCursor

-- deallocate the cursor, so we can reuse the cursor variable

DEALLOCATE @OutputCursor

-- Open the CCategories cursor

OPEN CCategories

-- Get information about a cursor

-- note that status = 1 means cursor open

EXEC sp_describe_cursor @OutputCursor OUTPUT,

N'local', N'CCategories'

PRINT CHAR(10) + 'sp_describe_cursor cursor OUTPUT'+ CHAR(10)

FETCH NEXT FROM @OutputCursor

WHILE @@FETCH_STATUS = 0
FETCH NEXT FROM @OutputCursor

-- deallocate the cursor, so we can reuse the cursor

variable

496
 Chapter 12. Row-Oriented Processing: Using Cursors

DEALLOCATE @OutputCursor

CLOSE CCategories

-- Open the CCustomers cursor

OPEN CCustomers

-- Get information about a cursor

-- note that status = 1 means cursor open

EXEC sp_describe_cursor_columns @OutputCursor OUTPUT,

N'local', N'CCustomers'

PRINT CHAR(10) + 'sp_describe_cursor_columns cursor OUTPUT'+ CHAR(10)

FETCH NEXT FROM @OutputCursor

WHILE @@FETCH_STATUS = 0
FETCH NEXT FROM @OutputCursor

-- deallocate the cursor, so we can reuse the cursor variable

DEALLOCATE @OutputCursor

CLOSE CCustomers

-- Open the CCategories cursor

OPEN COrdersComplete

-- Get information about a cursor

-- note that status = 1 means cursor open

EXEC sp_describe_cursor_tables @OutputCursor OUTPUT,

N'global', N'COrdersComplete'

PRINT CHAR(10) + 'sp_describe_cursor_tables cursor OUTPUT'+ CHAR(10)

FETCH NEXT FROM @OutputCursor

WHILE @@FETCH_STATUS = 0
FETCH NEXT FROM @OutputCursor

DEALLOCATE @OutputCursor

CLOSE COrdersComplete

DEALLOCATE CCategories

DEALLOCATE CCustomers

DEALLOCATE COrdersComplete
Note

Books Online contains a full description of the sp_cursor_list, sp_describe_cursor,

sp_describe_cursor_columns, and sp_describe_cursor_tables system stored
procedures.

Use this information to interpret the output from Listing 12.22.

497
Microsoft SQL Server 2000 Programming by Example

Using Cursors to Solve Multirow Actions in Triggers

In many cases, dealing with multirow operations inside triggers is not an easy task. If the single-row solution is
solved, you can use cursors to convert multirow operations into single-row operations inside the trigger, to
apply to them the same proved logic of the single-row cases.
Consider the following example: You want to assign a credit limit to every customer following an automated
process applied by the AssignCreditLimit stored procedure. To automate the process, you can create a
trigger AFTER INSERT to calculate the credit limit for every new customer.
The AssignCreditLimit stored procedure can work with only one customer at a time. However, an
INSERT operation can insert multiple rows at the same time, using INSERT SELECT.
You can create the trigger with two parts; one will deal with single row and the other with multiple rows, and
you will check which part to apply using the result of the @@ROWCOUNT function as described in Listing 12.23.

Listing 12.23 Using Cursors to Convert Multirow Operations into Single-Row Operations Inside
Triggers

USE Northwind
GO

ALTER TABLE Customers

ADD CreditLimit money
GO

CREATE PROCEDURE AssignCreditLimit

@ID nvarchar(5)
AS

-- Write here your own CreditLimit function

UPDATE Customers
SET CreditLimit = 1000
WHERE CustomerID = @ID
GO

CREATE TRIGGER isr_Customers

ON Customers
FOR INSERT AS

SET NOCOUNT ON

DECLARE @ID nvarchar(5)

IF @@ROWCOUNT > 1
-- Multirow operation
BEGIN

-- Open a cursor on the Inserted table

DECLARE NewCustomers CURSOR

FOR SELECT CustomerID
FROM Inserted

498
 Chapter 12. Row-Oriented Processing: Using Cursors

ORDER BY CustomerID

OPEN NewCustomers

FETCH NEXT FROM NewCustomers

INTO @ID

WHILE @@FETCH_STATUS = 0
BEGIN
-- Assign new Credit Limit to every new customer

EXEC AssignCreditLimit @ID

FETCH NEXT FROM NewCustomers

INTO @ID
END

-- close the cursor

CLOSE NewCustomers
DEALLOCATE NewCustomers
END

ELSE
-- Single row operation
BEGIN
SELECT @ID = CustomerID
FROM Inserted

IF @ID IS NOT NULL

-- Assign new Credit Limit to the new customer

EXEC AssignCreditLimit @ID

END

GO

-- Test it

INSERT customers (CustomerID, CompanyName)

VALUES ('ZZZZZ', 'New Company')

SELECT CreditLimit
FROM Customers
WHERE CustomerID = 'ZZZZZ'

Application Cursors
When a client application requests information from SQL Server using the default settings in ADO, OLE DB,
ODBC, or DB-Library, SQL Server must follow this process:

1. The client application sends a request to SQL Server in a network package. This request can be any
Transact-SQL statement or a batch containing multiple statements.
2. SQL Server interprets the request and creates a query plan to solve the request. The query plan is
compiled and executed.
3. SQL Server packages the results in the minimum number of network packets and sends them to the
user.

499
Microsoft SQL Server 2000 Programming by Example

4. The clients start receiving network packets, and these packets are waiting in the network buffer for the
application to request them.
5. The client application receives the information contained in the network packages row by row.

The client application cannot send any other statement through this connection until the complete result set is
retrieved or cancelled.
This is the most efficient way to retrieve information from SQL Server, and it is called a default result set. It is
equivalent to a FORWARD_ONLY READ_ONLY cursor with a row set size set to one row.
Note

Some articles and books refer to the default result set as a "Firehose" cursor, which is considered
an obsolete term.

SQL Server supports three types of cursors:

• Transact-SQL cursors— These are the cursors you studied in the previous sections of this chapter.
• Application Programming Interface (API) server cursors— These are cursors created in SQL Server,
following requests from the database library, such as ADO, OLE DB, ODBC, or DB-Library. Listings
12.1 and 12.3 contain examples of this type of cursor.
• Client cursors— These cursors are implemented in the client side by the database library. The client
cache contains the complete set of rows returned by the cursor, and it is unnecessary to have any
communication to the server to navigate the cursor.

Caution

Do not mix API cursors with Transact-SQL cursors from a client application, or SQL Server will try
to map an API cursor over Transact-SQL cursors, with unexpected results.

Tip

Use Transact-SQL cursors in stored procedures and triggers and as local cursors in Transact-SQL
batches, to implement cursors that do not require user interaction.

Use API cursors from client applications where the cursor navigation requires user interaction.

Using a default result set is more efficient than using a server cursor, as commented in previous sections in
this chapter.

Caution

500
 Chapter 12. Row-Oriented Processing: Using Cursors

You cannot open a server cursor in a stored procedure or batch if it contains anything other than a
single SELECT statement with some specific Transact-SQL statements. In these cases, use a client
cursor instead.

Using server cursors is more efficient than using client cursors because client cursors must cache the
complete result set in the client side, whereas server cursors send to the client the fetched rows only. To open
a client cursor using ADO, you can set the CursorLocation property to adUseClient in the Connection
or Recordset objects. The default value is adUseServer for server API cursor.

What's Next?

In this chapter, you learned how to use Transact-SQL cursors.

In Chapter 13, you will learn about transactions and locks, which are both important aspects of using cursors.
The concurrency of a database application depends directly on how the application manages transactions and
locks.

501
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

Chapter 13. Maintaining Data Consistency: Transactions

and Locks
SQL Server 2000 is designed to serve multiuser environments. If multiple users try to access the same data,
SQL Server must protect the data to avoid conflicting requests from different processes. SQL Server uses
transactions and locks to prevent concurrency problems, such as avoiding simultaneous modifications to the
same data from different users.
This chapter teaches you the following:

• Basic concepts about transactions

• How to use Transact-SQL statements to manage transactions
• How to understand the common concurrency problems and avoid them when they arise
• How to apply the right transaction isolation level
• Lock types available in SQL Server
• How to detect and avoid deadlocks

Characteristics of Transactions (ACID)

A transaction is a sequence of operations executed as a single logical operation, which must expose the ACID
(Atomicity, Consistency, Isolation, and Durability) properties. These are as follows:

• Atomicity— The transaction must be executed as an atomic unit of work, which means that it either
completes all of its data modifications or none at all.
• Consistency— The data is consistent before the transaction begins, and the data is consistent after
the transaction finishes. To maintain consistency, all integrity checks, constraints, rules, and triggers
must be applied to the data during the transaction. A transaction can affect some internal SQL Server
data structures, such as allocation maps and indexes, and SQL Server must guarantee that these
internal modifications are applied consistently. If the transaction is cancelled, the data should go back
to the same consistent state it was in at the beginning of the transaction.
• Isolation— The transaction must be isolated from changes made to the data by other transactions, to
prevent using provisional data that is not committed. This implies that the transaction must either see
the data in its previous state or the transaction must wait until the changes from other transactions are
committed.
• Durability— After the transaction completes, its changes to the data are permanent, regardless of the
event of a system failure. In other words, when a client application receives notification that a
transaction has completed its work successfully, it is guaranteed that the data is changed permanently.

Every RDBMS uses different ways to enforce these properties. SQL Server 2000 uses Transact-SQL
statements to control the boundaries of transactions to guarantee which operations must be considered as an
atomic unit of work.
Constraints and other integrity mechanisms are used to enforce logical consistency of every transaction. SQL
Server internal engines are designed to provide physical internal consistency to every operation that modifies
data, maintaining allocation structures, indexes, and metadata.
The programmer must enforce correct transaction and error management to enforce an appropriate atomicity
and consistency. Later in this chapter, in the "Transactions and Runtime Errors" section, you will learn
about transaction and error management.
Programmers can select the right level of isolation by specifying Transaction Isolation Level or using locking
hints. Later in this chapter, in the "Isolation Levels" section, you will learn how to apply transaction isolation
levels. The section "Types of Locks" gives you details on how to use locking hints.
SQL Server guarantees durability by using the Transaction log to track all the changes to the database and
uses the recovery process when necessary to enforce data consistency in case of system failure or
unexpected shutdown.

Using Transactions

503
Microsoft SQL Server 2000 Programming by Example

To consider several operations as members of the same transaction, it is necessary to establish the
transaction boundaries by selecting the transaction starting and ending points.
You can consider three different types of transactions:

• Auto commit transactions— SQL Server always starts a transaction whenever any statement needs to
modify data. SQL Server automatically commits the transaction if the statement finishes its work
successfully. However, if the statement produces any error, SQL Server will automatically roll back all
changes produced by this incomplete statement. In this way, SQL Server automatically maintains data
consistency for every statement that modifies data.
• Explicit transactions— The programmer specifically declares the transaction starting point and decides
either to commit or rollback changes depending on programming conditions.
• Implicit transactions— SQL Server starts a transaction automatically whenever any statement needs to
modify data, but it is the programmer's responsibility to specify the transaction ending point and
confirm or reject applied changes.

Note

It is impossible to instruct SQL Server to disable the creation of Auto commit transactions. This is
why inside a trigger you are always inside a transaction.

A Transact-SQL batch is not a transaction unless stated specifically. In Listing 13.1, Operations 1 through 3
are independent; there is no link between them, so they don't form a single transaction. If there is an error in
one of the operations, the others can still be committed automatically. However, operations 4 through 6 are
part of the same transaction, and either all of them or none of them will be applied permanently.

Using the @@IDENTITY function can be wrong in this case, because this system function returns the latest
Identity value generated in this connection. If a trigger inserts data in a table where you have an Identity field,
the @@IDENTITY function will return the value generated inside the trigger, not the one generated by the
original action that fired the trigger.

Tip

Use the SCOPE_IDENTITY() function to retrieve the latest Identity value inserted in the current
scope.

Listing 13.1 Setting the Transaction Boundaries

USE Northwind

504
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

GO

-----------------------
-- Without Transactions
-----------------------

DECLARE @CatID int,

@ProdID int

-- Operation 1
-- Create a new Category

INSERT Categories
(CategoryName)
VALUES ('Cars')

-- Retrieves the latest IDENTITY value inserted

SET @CatID = SCOPE_IDENTITY()

-- Operation 2
-- Create a new product
-- in the new Category

INSERT Products
(ProductName, CategoryID)
VALUES ('BigCars', @CatID)

-- Retrieves the latest IDENTITY value inserted

SET @ProdID = SCOPE_IDENTITY()

-- Operation 3
-- Change UnitsInStock
-- for the new product

UPDATE Products
SET UnitsInStock = 20
WHERE ProductID = @ProdID

-----------------------
-- With Transactions
-----------------------

-- Start a new transaction

BEGIN TRAN

-- Operation 4
-- Create a new Category

INSERT Categories
(CategoryName)
VALUES ('HiFi')

IF @@ERROR <> 0 GOTO AbortTransaction

SELECT @CatID = CategoryID

FROM Categories

505
Microsoft SQL Server 2000 Programming by Example

WHERE CategoryName = 'HiFi'

-- Operation 2
-- Create a new product
-- in the new Category

INSERT Products
(ProductName, CategoryID)
VALUES ('GreatSound', @CatID)

IF @@ERROR <> 0 GOTO AbortTransaction

SELECT @ProdID = ProductID

FROM Products
WHERE ProductName = 'GreatSound'

-- Operation 3
-- Change UnitsInStock
-- for the new product
UPDATE Products
SET UnitsInStock = 50
WHERE ProductID = @ProdID

IF @@ERROR <> 0 GOTO AbortTransaction

COMMIT TRAN
PRINT 'Transaction committed'

GOTO EndTransaction

AbortTransaction:

ROLLBACK TRAN
PRINT 'Transaction rolled back'

EndTransaction:

PRINT 'Transaction finished'

BEGIN TRAN

To start a new local transaction, you can use the BEGIN TRANSACTION (or BEGIN TRAN) statement. This
statement starts a new transaction, if there aren't any transactions already started, or creates a new level of
nested transactions if the execution was already inside another transaction.
As mentioned before, any time you execute a statement that modifies data, SQL Server automatically starts a
new transaction. If you were already inside a transaction when the statement started to run and this operation
fired a trigger inside the trigger, you will be in the second level of a nested transaction.
The same situation happens if you define a stored procedure to apply some data changes, and you need to
apply these data changes as a single transaction. In this case, you start a new transaction inside the stored
procedure and decide at the end of it whether you want to commit or roll back. This stored procedure will
execute its statements in a transaction state regardless of the existence of a transaction in the calling
procedure or batch.
It is possible to have any number of nested transactions in SQL Server 2000. The @@TRANCOUNT system
function gives you the number of open transactions you have at any given time. Any time you execute BEGIN
TRAN, the result of the function @@TRANCOUNT is increased by one. Listing 13.2 shows an example of how
the @@TRANCOUNT function works.

Listing 13.2 Values of the @@TRANCOUNT Function After Using BEGIN TRAN

506
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

BEGIN TRAN

SELECT @@TRANCOUNT 'First Transaction'

BEGIN TRAN

SELECT @@TRANCOUNT 'Second Transaction'

BEGIN TRAN

SELECT @@TRANCOUNT 'Third Transaction'

BEGIN TRAN

SELECT @@TRANCOUNT 'Fourth Transaction'

ROLLBACK TRAN

First Transaction
-----------------
1

Second Transaction
------------------
2

Third Transaction
-----------------
3

Fourth Transaction
------------------
4
Note

Using nested transactions is not considered a good practice. SQL Server considers nested
transactions as one single transaction, starting on the first BEGIN TRAN and finishing on the last
COMMIT TRAN or the first ROLLBACK TRAN. Having multiple transaction levels in the same batch
or stored procedure makes the code harder to understand and maintain.

The reason for having nested transactions is to be able to start a new transaction inside a stored
procedure, or trigger, regardless of the existence of a previous transaction in the process.

507
Microsoft SQL Server 2000 Programming by Example

You can assign a name to a transaction to easily identify it in code. In this case, this name only helps you to
identify possible errors in code, but you cannot commit or roll back an individual transaction by providing its
name, unless you save the transaction. You learn how to do this later in this chapter in the "ROLLBACK
TRAN" section. Listing 13.3 uses transactions with names and shows how they are written to the transaction
log.

Listing 13.3 Transactions with Names Can Be Identified in the Transaction Log

USE Northwind
GO

BEGIN TRAN ChangeNameAllCustomers

UPDATE Customers
SET CompanyName = CompanyName
WHERE COUNTRY = 'USA'

SELECT [Current LSN], Operation, [Transaction Name]

FROM ::fn_dblog(NULL, NULL)

COMMIT TRAN

(Partial output)

Current LSN Operation Transaction Name

---------------------- ------------------------- --------------------
00000046:000000ca:0002 LOP_BEGIN_CKPT NULL
00000046:000000cb:0001 LOP_END_CKPT NULL
00000046:000000cc:0001 LOP_BEGIN_XACT ChangeNameAllCustome
00000046:000000cc:0002 LOP_DELETE_ROWS NULL
00000046:000000cc:0003 LOP_MODIFY_HEADER NULL
00000046:000000cc:0004 LOP_SET_BITS NULL
00000046:000000cc:0005 LOP_DELETE_ROWS NULL
00000046:000000cc:0006 LOP_INSERT_ROWS NULL
00000046:000000cc:0007 LOP_INSERT_ROWS NULL
00000046:000000cc:0008 LOP_DELETE_ROWS NULL
00000046:000000cc:0009 LOP_DELETE_ROWS NULL
00000046:000000cc:000a LOP_INSERT_ROWS NULL
00000046:000000cc:000b LOP_INSERT_ROWS NULL

508
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

00000046:000000cc:000c LOP_DELETE_ROWS NULL

00000046:000000cc:000d LOP_DELETE_ROWS NULL
00000046:000000cc:000e LOP_INSERT_ROWS NULL
00000046:000000cc:000f LOP_INSERT_ROWS NULL
. . . [deleted rows from output]
Caution

Microsoft does not support the fn_dblog function. It is used here only to show how transactions
are written to the Transaction log.

Note

Before using the fn_dblog function, you should change the Northwind database to Full Logging
mode and perform a full database backup.

Whenever you start a new transaction, it is marked in the transaction log, as you saw in Listing 13.3. You
can restore a transaction log, specifying to stop the restore process either before or after a specific marked
transaction. To achieve this, you must use the WITH MARK option in the BEGIN TRAN statement, as you can
see in Listing 13.4.

Listing 13.4 Starting Transactions Using the WITH MARK Option

USE Northwind
GO

BEGIN TRAN ChangeUnitsInStock_1_10_20

WITH MARK

UPDATE Products
SET UnitsInStock = UnitsInStock * 1.1
WHERE ProductID in (1, 10, 20)

COMMIT TRAN
Caution

Mark your long complex administrative transactions with a name and use WITH MARK, so you can
restore the database in the same state it was before the execution of this complex, and potentially
dangerous, operation.

509
Microsoft SQL Server 2000 Programming by Example

Note

Using named nested transactions, only the name of the outermost transaction is recorded in the
transaction log.

COMMIT TRAN

To confirm the changes made inside a transaction, you must execute the COMMIT TRANSACTION (or COMMIT
TRAN) statement.
Caution

Explicit transactions must be committed using COMMIT TRAN; otherwise, they will be rolled back
when the connection is closed, or during the recovery process in case of a system shutdown or
failure.

If the transaction was at the first level of transactions, executing COMMIT TRAN forces SQL Server to consider
the changes made to the database as permanent. If the transaction was inside another transaction or
transactions, changes to the database must wait until the outermost transaction is committed. In this case, the
value of @@TRANCOUNT decreases by 1, but no changes to the data are confirmed yet. Only when the value
of @@TRANCOUNT changes from 1 to 0 because of a COMMIT TRAN statement are changes on the data
considered permanent.

While the transaction is not finally committed, the modified data is locked to other transactions. SQL Server
will free these locks as soon as the transaction finally terminates, and this happens only when the outermost
transaction terminates.

Listing 13. 5 shows the effect of COMMIT TRAN in the value of @@TRANCOUNT.

Listing 13.5 Every Time You Execute COMMIT TRAN, @@TRANCOUNT Is Decreased by One

USE Northwind
GO

BEGIN TRAN Customers

510
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

UPDATE Customers
SET ContactTitle = 'President'
WHERE CustomerID = 'AROUT'

SELECT @@TRANCOUNT 'Start Customers Transaction'

BEGIN TRAN Products

UPDATE Products
SET UnitPrice = UnitPrice * 1.1
WHERE CategoryID = 3

SELECT @@TRANCOUNT 'Start Products Transaction'

BEGIN TRAN Regions

INSERT Region
VALUES (5, 'Europe')
SELECT @@TRANCOUNT 'Start Regions Transaction'

COMMIT TRAN Regions

SELECT @@TRANCOUNT 'Commit Regions Transaction'

BEGIN TRAN Orders

UPDATE Orders
SET ShippedDate = CONVERT(VARCHAR(10), Getdate(), 120)
WHERE OrderID = 10500

SELECT @@TRANCOUNT 'Start Orders Transaction'

COMMIT TRAN Orders

SELECT @@TRANCOUNT 'Commit Orders Transaction'

COMMIT TRAN Products

SELECT @@TRANCOUNT 'Commit Products Transaction'

COMMIT TRAN Customers

SELECT @@TRANCOUNT 'Commit Customers Transaction'

Start Customers Transaction

---------------------------
1

Start Products Transaction

--------------------------
2

511
Microsoft SQL Server 2000 Programming by Example

Start Regions Transaction

-------------------------
3

Commit Regions Transaction

--------------------------
2

Start Orders Transaction

------------------------
3
Commit Orders Transaction
-------------------------
2

Commit Products Transaction

---------------------------
1

Commit Customers Transaction

----------------------------

0
Caution

Although you can provide a transaction name to the COMMIT TRAN statement, this name is ignored
and the latest open transaction is committed instead. However, it is a good practice to provide
names to transactions in long scripts and stored procedures to provide extra help to detect code
errors.

ROLLBACK TRAN

To cancel the changes applied during a transaction, you can use the ROLLBACK TRANSACTION (or
ROLLBACK TRAN) statement. Calling ROLLBACK TRAN inside a nested transaction undoes all the changes
applied from the starting point of the outermost transaction. Because ROLLBACK TRAN cancels the active
transaction, all the resources locked by the transaction are freed after the transaction terminates. After the
execution of the ROLLBACK TRAN statement, the TRANCOUNT function returns 0.
SQL Server 2000 supports the ANSI standard statement ROLLBACK WORK, which is equivalent to ROLLBACK
TRAN, but in this case you cannot specify a transaction name.
Listing 13.6 shows the effect of ROLLBACK TRAN in the value of @@TRANCOUNT.

Listing 13.6 When You Execute ROLLBACK TRAN, @@TRANCOUNT Is Decremented to Zero

BEGIN TRAN Customers

512
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

UPDATE Customers
SET ContactTitle = 'President'
WHERE CustomerID = 'AROUT'

SELECT @@TRANCOUNT 'Start Customers Transaction'

BEGIN TRAN Products

UPDATE Products
SET UnitPrice = UnitPrice * 1.1
WHERE CategoryID = 3

SELECT @@TRANCOUNT 'Start Products Transaction'

BEGIN TRAN Orders

UPDATE Orders
SET ShippedDate = CONVERT(VARCHAR(10), Getdate(), 120)
WHERE OrderID = 10500

SELECT @@TRANCOUNT 'Start Orders Transaction'

COMMIT TRAN Orders

SELECT @@TRANCOUNT 'Commit Orders Transaction'

-- Note: the following statement produces an error,

-- because the specified transaction name is invalid

ROLLBACK TRAN Products

SELECT @@TRANCOUNT 'Rollback Products Transaction'

ROLLBACK TRAN Customers

SELECT @@TRANCOUNT 'Rollback Customers Transaction'

Start Customers Transaction

---------------------------
1

Start Products Transaction

--------------------------
2

Start Orders Transaction

------------------------
3

Commit Orders Transaction

-------------------------

513
Microsoft SQL Server 2000 Programming by Example

Server: Msg 6401, Level 16, State 1, Line 29

Cannot roll back Products. No transaction or savepoint of that name was found.

Rollback Products Transaction

-----------------------------
2

Rollback Customers Transaction

------------------------------

0
The way that ROLLBACK TRAN works depends on the point from which you execute it:

• When executed inside a batch, it cancels the active transaction, but the execution continues with the
remaining statements of the batch. To prevent this situation, check the value of the @@TRANCOUNT
function.
• Using ROLLBACK TRAN inside a stored procedure cancels the active transaction, even if the
outermost transaction was declared outside the stored procedure. However, the execution continues
with the remaining statements of the stored procedure. In this case, the process that called this
procedure receives a warning because @@TRANCOUNT changed its value inside the procedure.
• If you execute ROLLBACK TRAN inside a trigger, the transaction is completely cancelled but the
execution of the trigger continues. Any changes made inside the trigger after ROLLBACK TRAN are
made permanent, because these modifications are not inside a transaction anymore. However, when
the execution of the trigger terminates, the batch is cancelled and no more instructions will be
executed.

Tip

You can cancel the operation that fires the trigger, without using ROLLBACK TRAN, using the
information contained in the Inserted and Deleted tables to execute an action that
compensates the action just made. For example, you can cancel a DELETE operation reinserting in
the target table the content of the Deleted table.

In some cases, it could be interesting to consider part of a transaction as provisional, being able to roll back
this portion without affecting the outer transaction status. In this case, you can create a savepoint and roll
back only to the savepoint.

As an example, consider a transaction has been created to insert a new order and to insert some rows in
Order Details. As a part of the same trans action, you want to try a 10% discount to products ordered in more
than 5 units in this transaction, but only if the order costs more than $500 after the discount. To solve this
problem, you can declare a savepoint before applying the extra discount. After the extra discount is applied,
you can test whether the total price of this order is lower than $500, in which case this extra discount, and only
this extra discount, must be rolled back. Listing 13.7 shows this example.

Listing 13.7 Use Savepoints to Roll Back a Portion of a Transaction

514
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

BEGIN TRAN NewOrder

-- Insert a new Order

DECLARE @ID int

INSERT Orders (CustomerID, OrderDate)

VALUES ('BOTTM', '2000-11-23')

-- Obtain the newly inserted OrderID

SET @ID = @@IDENTITY

-- Insert [Order details] data

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @ID, 23, 9, 12, 0.10

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @ID, 18, 62.5, 5, 0.05

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @ID, 32, 32, 5, 0.05

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @ID, 9, 97, 4, 0.10
-- try the discount

-- Create a Savepoint

SAVE TRAN Discount

-- Increase the discount to

-- products where Quantity >= 5

UPDATE [Order Details]

SET Discount = Discount + 0.1
WHERE OrderID = @ID
AND QUANTITY >= 5

-- Check the total price, after the extra discount, to see if

-- this order qualifies for this discount.

IF (SELECT SUM(Quantity * UnitPrice * (1-Discount))

FROM [Order Details]

515
Microsoft SQL Server 2000 Programming by Example

WHERE OrderID = @ID) < 500

-- Does not qualify, roll back the discount

ROLLBACK TRAN Discount

-- Commit the transaction, inserting the order permanently

COMMIT TRAN NewOrder

Caution

Transaction names are case sensitive in SQL Server 2000.

Caution

In a ROLLBACK TRAN statement, the only names allowed are the name of a saved transaction or
the name of the outermost transaction.

Using Implicit Transactions

As commented earlier in this chapter, SQL Server starts a transaction automatically every time you modify
data. However, these transactions are automatically committed when the operation terminates. In this way,
each statement by itself is a transaction in SQL Server.
You can set a connection in Implicit Transactions mode. In this mode, the first time you modify data, SQL
Server starts a transaction and keeps the transaction open until you decide explicitly to commit or roll back the
transaction.
To set the Implicit Transactions mode in a connection, you must execute the SET
IMPLICIT_TRANSACTIONS ON statement. Listing 13.8 shows an example of implicit transactions.
Caution

SQL Server 2000 connections start with Implicit Transactions mode off, so any change you make
to a database is permanent, unless it is executed inside a user-defined transaction.

Listing 13.8 Using the Implicit Transactions Mode

516
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

SET NOCOUNT ON
SET IMPLICIT_TRANSACTIONS ON
GO

SELECT @@TRANCOUNT
AS 'Transactions levels before UPDATE'

UPDATE Customers
SET ContactName = 'Peter Rodriguez'
WHERE CustomerID = 'ANATR'

SELECT @@TRANCOUNT
AS 'Transactions levels after UPDATE'

ROLLBACK TRAN

SELECT @@TRANCOUNT
AS 'Transactions levels after ROLLBACK TRAN'

Transactions levels before UPDATE

---------------------------------
0

Transactions levels after UPDATE

--------------------------------
1

Transactions levels after ROLLBACK TRAN

---------------------------------------
0

Tip

Use Implicit Transactions mode in testing environments. In this mode, you can always roll back
your actions.

Transactions and Runtime Errors

It is a common misconception that errors inside a transaction force the transaction to roll back. However, this
is not always true, and you should provide the appropriate error control to decide when to roll back the
changes after an error.

517
Microsoft SQL Server 2000 Programming by Example

You can use the @@ERROR system function to detect the error caused by the latest statement sent to SQL
Server in your connection. If the statement was successful, @@ERROR will return 0.
In some cases, you can consider an error as something that is perfectly valid for SQL Server. For example,
you can execute an INSERT statement, and because of the restricting conditions, the statement does not
insert any row. For SQL Server, the statement was completed successfully and @@ERROR returned 0.
However, if you called the @@ROWCOUNT function, you can see that it returns 0 as well, because 0 rows were
affected by the latest statement.
Another potential problem might be trying to commit a nonexistent transaction, because of a previous rollback
in the same connection. A rollback does not cancel the batch, and the execution continues, potentially arriving
at a COMMIT TRAN or ROLLBACK TRAN again, producing a runtime error.
Listing 13.9 shows the following sequence:

1. You start a transaction.

2. You try to create a new product in category 10 and get an error message because category 10 does
not exist.
3. You create a new order.
4. Unaware of the previous product error, the execution continues and you try to insert the new product
in an existing order, thinking that this product exists. You do not get an error message, but nothing is
actually inserted.
5. You commit the transaction thinking that the order contains a product, but actually it doesn't.
6. You start a new transaction.
7. You try to create a new product in category 10 and you get an error message because category 10
does not exist.
8. The batch detects the error and decides to roll back the transaction.

Listing 13.9 The Same Transaction With and Without Error Control

USE Northwind
GO

------------------------
-- Without Error Control
------------------------

DECLARE @PID int, @OID int

PRINT CHAR(10) + 'Start a transaction without error control'+ CHAR(10)

BEGIN TRAN

INSERT Products (ProductName, CategoryID, UnitPrice)

VALUES ('New Cars Magazine Year Subscription', 10, 35.0)

SET @PID = SCOPE_IDENTITY()

INSERT Orders (CustomerID, OrderDate)

VALUES ('COMMI', '2000-11-23')

SET @OID = SCOPE_IDENTITY()

INSERT [Order Details]

518
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

(OrderID, ProductID, UnitPrice, Quantity, Discount)

SELECT @OID, ProductID, UnitPrice, 1, 0.3
FROM Products
WHERE ProductID = @PID

COMMIT TRAN

PRINT CHAR(10)
+ 'The transaction has been partially applied'
+ CHAR(10)

SELECT ProductName
FROM Products
WHERE ProductID = @PID

SELECT CustomerID, OrderDate

FROM Orders
WHERE OrderID = @OID

SELECT UnitPrice, Quantity

FROM [Order Details]
WHERE ProductID = @PID
GO

---------------------
-- With Error Control
---------------------

DECLARE @PID int, @OID int

PRINT CHAR(10)
+ 'Start a transaction with error control'
+ CHAR(10)

BEGIN TRAN

INSERT Products (ProductName, CategoryID, UnitPrice)

VALUES ('New Cars Magazine Year Subscription', 10, 35.0)

IF @@ERROR <> 0
GOTO CancelOrder

SET @PID = SCOPE_IDENTITY()

INSERT Orders (CustomerID, OrderDate)

VALUES ('COMMI', '2000-11-23')

IF @@ERROR <> 0
GOTO CancelOrder

SET @OID = SCOPE_IDENTITY()

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @OID, ProductID, UnitPrice, 1, 0.3
FROM Products
WHERE ProductID = @PID

IF @@ERROR <> 0 OR @@ROWCOUNT=0

GOTO CancelOrder

519
Microsoft SQL Server 2000 Programming by Example

COMMIT TRAN
GOTO CheckOrder

CancelOrder:

ROLLBACK TRAN
CheckOrder:

PRINT CHAR(10)
+ 'The transaction has been completely rolled back'
+ CHAR(10)

SELECT ProductName
FROM Products
WHERE ProductID = @PID

SELECT CustomerID, OrderDate

FROM Orders
WHERE OrderID = @OID

SELECT UnitPrice, Quantity

FROM [Order Details]
WHERE ProductID = @PID
Start a transaction without error control

Server: Msg 547, Level 16, State 1, Line 1

<$I~listings;transactions;error control>
INSERT statement conflicted with COLUMN FOREIGN KEY constraint
'FK_Products_Categories'.
The conflict occurred in database 'Northwind', table 'Categories', column
'CategoryID'.
The statement has been terminated.

(1 row(s) affected)

(0 row(s) affected)

The transaction has been partially applied

ProductName
----------------------------------------

(0 row(s) affected)

CustomerID OrderDate
---------- ------------------------------------------------------
COMMI 2000-11-23 00:00:00.000

(1 row(s) affected)

520
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

UnitPrice Quantity
--------------------- --------
(0 row(s) affected)

Start a transaction with error control

Server: Msg 547, Level 16, State 1, Line 1

INSERT statement conflicted with COLUMN FOREIGN KEY constraint
'FK_Products_Categories'.
The conflict occurred in database 'Northwind', table 'Categories', column
'CategoryID'.
The statement has been terminated.

The transaction has been completely rolled back

ProductName
----------------------------------------

(0 row(s) affected)

CustomerID OrderDate
---------- ------------------------------------------------------

(0 row(s) affected)

UnitPrice Quantity
--------------------- --------

(0 row(s) affected)

Concurrency Problems
In a multiuser environment, as several users try to access the same data at the same time, trying to perform
different actions, you can encounter various concurrency problems.
In the following sections you learn every potential problem and, later in this chapter, you will see how to use
the right isolation level to solve the concurrency problem and how SQL Server uses locks to support these
isolation levels.
Every concurrency problem is illustrated with a figure. Every figure shows two connections, called Connection
A and Connection B. These two connections can be made from the same computer or from different
computers, and by the same user or by different users. SQL Server has to solve concurrency problems
between connections, regardless of which user or client com puter is used to establish these connections.

Lost Updates

You can experience a lost update situation whenever two connections modify the same data in sequence,
because SQL Server will maintain only the last successful update. Consider the example of Figure 13.1:

Figure 13.1. Updating data from two connections can produce lost updates.

521
Microsoft SQL Server 2000 Programming by Example

1. Connection A retrieves the UnitPrice of Product 25 in the variable @UP for later use.
2. Connection B retrieves the UnitPrice of Product 25 in the variable @UP for later use.
3. Connection A updates the price of Product 25, increasing the price 20% over the price saved in the
variable @UP. The change is permanent.
4. Connection B updates the price of Product 25, increasing the price 20% over the price saved in the
variable @UP. Connection B is unaware that Connection A changed the product after step 2. The new
price is changed permanently— overwriting the price modified by Connection A in step 3.
5. Connection A checks the new price and compares it with the calculated price and finds that they are
different. The update from Connection A is lost.
6. Connection B checks the new price and it is the same as the intended computed value.

This problem can be prevented by writing atomic UPDATE statements in both Connection A and Connection B.
An atomic UPDATE contains the reading operation and the writing operation in a single statement, as in
Listing 13.10.

Listing 13.10 Write Atomic UPDATE Statements to Prevent Lost Updates

522
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

UPDATE Products
SET UnitPrice = UnitPrice * 1.2
WHERE ProductID = 25

Uncommitted Dependency (Dirty Read)

Reading data without using locks can produce unexpected results. Other connections can modify data
temporarily, and you can use that new data in your calculations, driving you to incorrect results.
In this case, you could say that you are reading dirty data, because the data has not been committed yet. In
other words, your calculations depend on uncommitted data.
Figure 13.2 shows an example of this situation:

Figure 13.2. Reading data without using locks can produce dirty reads.

523
Microsoft SQL Server 2000 Programming by Example

1. Connection A starts a new transaction to update product prices. Inside the transaction, Connection A
increases product prices by 20%. The transaction is not committed yet, so these changes are
provisional.
2. Connection B decides to work reading uncommitted data. Connection B retrieves the price for product
25. This price is dirty, because Connection A has changed this value temporarily to be 20% higher
than before.
3. Connection A changes its mind and decides to roll back the changes on product prices. In this case,
product 25 recovers its original price.
4. Connection B, unaware of the changes made by Connection A in step 3, creates a new Order and a
new Order Detail for product 25, using the price incorrectly retrieved in step 2. These insertions in the
database are permanent.

SQL Server 2000 prevents this problem automatically by using READ COMMITTED as the default isolation
level. In this way, other connections cannot see the changes before they are considered permanent. Listing
13.11 shows how to declare the same operation in Connection B as READ COMMITTED, using an atomic
INSERT instead of a SELECT followed by INSERT. In this case, Connection B must wait for Connection A to
liberate its locks before proceeding with the UPDATE operation.

Listing 13.11 Do Not Use the READ UNCOMMITTED Isolation Level If You Want to Prevent Dirty Reads

524
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

SET TRANSACTION ISOLATION LEVEL READ COMMITTED

DECLARE @ID int

INSERT Orders
(CustomerID, OrderDate)
VALUES ('BSBEV', '2000-11-25')

SET @ID = @@IDENTITY

INSERT [Order Details]

(OrderID, ProductID, UnitPrice, Quantity, Discount)
SELECT @ID, ProductID, UnitPrice, 10, 0.1
FROM Products
WHERE ProductID = 25
Note

Later in this chapter, you will learn the READ UNCOMMITTED and READ COMMITTED isolation
levels.

Inconsistent Analysis (Nonrepeatable Read)

In a multiuser environment, other users dynamically modify the data stored in a database.
Trying to execute a long-running process, such as a monthly report, can produce some inconsistencies
because of changes produced in the database from the beginning of the report to the end. This can be
considered an inconsistent analysis, because every time you read data, the data is different. This situation is
called nonrepeatable reads.
To produce a long-running report, you must retrieve the number of orders from the Order Details table
because you are interested only in orders with Order Details. Later in the report, you might want to calculate
the average of total price per order, using the value retrieved previously.
Figure 13.3 shows an example of this situation:

Figure 13.3. This is an example of inconsistent analysis.

525
Microsoft SQL Server 2000 Programming by Example

1. Connection A retrieves the total number of orders with at least one row in the Order Details table,
and stores this value in the @Count variable.
2. Connection B updates the Order Details table changing the quantity of the Product 20 in the
Order 10272.
3. The total number of rows has not been changed from Connection B, and now you can calculate the
total price of the complete Order Details table and store this value in the @Total variable. This
total includes the change made by Connection B in step 2.
4. Connection B updates the Order Details table again, changing the discount to Product 20.
5. Connection A is not aware of the changes on discount that Connection B applied to Product 20. So, it
does not know that the content of the variable @Total is invalid. Connection A calculates the average
price per order in two ways and every way provides a different result.

This problem can be prevented by minimizing the repeated reads and trying to convert them into atomic
operations, or using the REPEATABLE READ isolation level. When using the REPEATABLEREAD isolation level,
the retrieved data is locked until the transaction terminates, preventing changes to the data from other
connections.
Listing 13.12 shows that the average price calculated directly or indirectly produces the same results in an
atomic SELECT statement.

526
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

Listing 13.12 Use Atomic SELECT Statements to Prevent Repeatable Reads

USE Northwind
GO

SELECT AVG(TotalPrice) AS 'Actual AVG',

SUM(TotalPrice) / COUNT(OrderID) AS 'Calc AVG'
FROM (
SELECT OrderID, SUM(UnitPrice * Quantity * (1 - Discount))
AS TotalPrice
FROM [Order Details]
GROUP BY OrderID) AS TotOrders
Note

Later in this chapter, you will learn the REPEATABLE READ isolation levels.

Phantom Reads

In the preceding section, you covered the concurrency problem due to updates in data previously read. If
other connections are inserting data in the range of data you are analyzing, you can find that those new rows
appear in your result sets with no apparent reason, from your connection point of view. These new rows are
called phantoms.
This problem produces inconsistent analysis as well, because your previous totals are no longer valid after the
insertion of new rows.
Figure 13.4 shows a very simple example of this problem:

Figure 13.4. Insertions from other connections can appear in your results as phantom rows.

527
Microsoft SQL Server 2000 Programming by Example

1. Connection A retrieves the list of orders that includes the Product 25, and it produces a result set with
6 rows, including orders 10259, 10337, 10408, 10523, 10847, and 10966.
2. Connection B inserts a new order detail in the Order 10615 with the product 25.
3. Connection A resends the same statement as in step 1, but it retrieves a new row, correspondent to
the Order 10615. Connection A was unaware of this insertion, so it considers this row to be a
phantom.

Caution

Having phantoms is nothing to avoid per se, unless they produce inconsistent analysis as well,
which is the usual case.

528
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

Preventing this problem is more difficult than in the previous cases, because SQL Server needs to lock rows
that do not exist yet, locking non-existing data.

You can use the SERIALIZABLE isolation level, or the SERIALIZABLE optimizer hint, to prevent phantom
reads. Listing 13.13 shows how to use the SERIALIZABLE optimizer hint.

Listing 13.13 Using SERIALIZABLE Optimizer Hint Inside a Transaction to Prevent Phantom Reads

USE Northwind
GO

SELECT OrderID,
ProductID, UnitPrice
FROM [Order Details] (SERIALIZABLE)
WHERE ProductID = 37
Note

Later in this chapter, you will learn how SQL Server implements the SERIALIZABLE isolation level.

Isolation Levels

In previous sections of this chapter, you learned about the concurrency problems experienced in a multiuser
environment.
It is important to decide how changes from other connections can affect your results, or when to allow other
connections to apply changes to the data you use in your connection.
You can select the isolation level of your transactions in SQL Server 2000 either per transaction or per table.
You can use the SET ISOLATION LEVEL statement to specify the isolation level at the transaction level. To
select the isolation level per table, you can use locking hints.
SQL Server 2000 supports four isolation levels:

• READ COMMITTED
• READ UNCOMMITTED
• REPEATABLE READ
• SERIALIZABLE

Isolation levels are defined to solve specific concurrency problems as shown in Figure 13.5.

Figure 13.5. Every isolation level solves specific concurrency problems.

529
Microsoft SQL Server 2000 Programming by Example

Note

Note that SQL Server avoids lost updates automatically by using exclusive locks inside
transactions. However, after the transaction terminates, other connections can modify the same
data again, and SQL Server will always keep the latest modified value.

READ COMMITTED

SQL Server 2000 uses by default the READ COMMITTED isolation level. Using this setting, transactions
cannot see changes from other connections while they are not committed or rolled back. SQL Server
implements this isolation level, locking the modified data exclusively; therefore, other transactions cannot read
the modified data.
Note

ANSI-92 specifies that if a connection tries to read data that is being modified by another
connection, it can either wait for data to be unlocked or it must see the previous state of the data.

SQL Server implements exclusive locks to solve this situation, so other connections must wait.
Other RDBMS products implement the other way, so users can see the previous state of the data
until changes are made permanent.

530
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

Every implementation has advantages and disadvantages.

Using this isolation level you can prevent dirty reads, because you cannot see data modifications from other
connections that have not been committed; however, other connections can change some of the data you
have read already during your transaction, producing nonrepeatable reads, or add new rows to the data set
you are reading, producing phantoms.

Tip

Try to use the READ COMMITTED isolation table as a standard isolation level because it is less
intrusive than REPEATABLE READ and SERIALIZABLE, and it provides a better concurrency to
your database application.

Setting the READ COMMITTED locking hint for a table, in the FROM clause of a query, overrides the transaction
isolation level. Listing 13.14 shows how to use the READ COMMITTED isolation level and READCOMMITTED
locking hint.

Listing 13.14 Use READ COMMITTED to Prevent Dirty Reads

USE Northwind
GO

-- Use the READCOMMITTED optimizer hint for a table

SELECT *
FROM Products (READCOMMITTED)
GO

-- Use the READ COMMITTED ISOLATION LEVEL for a transaction

SET ISOLATION LEVEL READ COMMITTED

GO

SELECT *
FROM Products

READ UNCOMMITTED

531
Microsoft SQL Server 2000 Programming by Example

In some cases, it might be interesting to read data without being affected by any potential lock held by other
connections. A typical example is to search for some general information about the data contained in a
table— such as the total number of rows— while the table is being modified by other connections. In these
cases, you don't care about dirty reads, and reading provisional data is acceptable.
Using the READ UNCOMMITTED isolation level, SQL Server does not check for exclusive locks affecting the
data, and does not lock the data being read with shared locks. You can use this isolation level by using the
SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED statement.
You can use the READUNCOMMITTED locking hint to specify the READ UNCOMMITTED isolation level for a
specific table in a query, regardless of the current transaction isolation level. Listing 13.15 shows an
example of this locking hint.
Note

The READUNCOMMITTED and NOLOCK locking hints are equivalent.

Listing 13.15 Using the READ UNCOMMITTED Isolation Level

USE Northwind
GO

-- Use the READUNCOMMITTED or the NOLOCK locking hint

-- to avoid lockings on specific tables

SELECT COUNT(*)
FROM [Order Details] (READUNCOMMITTED) –- or (NOLOCK)
GO

-- Use the READ UNCOMMITTED Isolation level

-- to specify this setting for the current connection

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED

GO

SELECT COUNT(*)
FROM [Order Details]

SELECT MAX(OrderID)
FROM Orders
Caution

Using the READ UNCOMMITTED isolation level makes sense only when you are reading for data.
Applying this isolation level to a table that is being modified by the same transaction does not
prevent the production of exclusive locks on this table.

532
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

REPEATABLE READ

Use the REPEATABLE READ isolation level to guarantee that the retrieved data will not change during the
transaction. To achieve this isolation level, SQL Server must lock the data retrieved, preventing updates from
other connections. This is important on long-running processes where the results must be consistent all during
the process, such as during long-running reports.
This isolation level provides consistent analysis, but it does not prevent the appearance of phantom reads.
You can override the default transaction isolation level for one specific table in a query using the
REPEATABLEREAD locking hint in the FROM clause of the query, as in Listing 13.16.

Listing 13.16 Use the REPEATABLE READ Isolation Level to Prevent Inconsistent Analysis

USE Northwind
GO

-- Use the REPEATABLEREAD to prevent inconsistent analysis

-- when reading from a table
-- that is being modified by other connections

SELECT COUNT(*)
FROM [Order Details] (REPEATABLEREAD)
GO

-- Set REPEATABLE READ isolation level at connection setting

-- to affect every table used in the connection

SET TRANSACTION ISOLATION LEVEL REPEATABLE READ

GO

SELECT COUNT(*) AS NRows

FROM [Order Details]

SELECT SUM(Quantity * UnitPrice * (1-Discount)) AS Total

FROM [Order Details]
Caution

Use the REPEATABLE READ isolation level with caution, because it locks connections that try to
modify the data affected by this transaction.

SERIALIZABLE

Use the SERIALIZABLE isolation level to prevent all the concurrency problems, including the appearance of
phantom reads.

533
Microsoft SQL Server 2000 Programming by Example

SQL Server must lock enough resources to prevent the possibility that other connections could insert or
modify data covered by the range specified in your query. The way SQL Server enforces this isolation level
depends on the existence of a suitable index, adequate to the WHERE clause of the query or queries to protect:

• If the table does not have a suitable index, SQL Server must lock the entire table.
• If there is an appropriate index, SQL Server locks the keys in the index corresponding to the range of
rows in your query, plus the next key to the range.

Tip

Create an index according to the fields used in the WHERE clause of a query that references a table
with a SERIALIZABLE locking hint, to prevent locks at table level.

Caution

Applying a SERIALIZABLE isolation level to a table used in a query without a WHERE clause will
lock the entire table.

You can use the SERIALIZABLE locking hint to apply this isolation level to a specific table in a query,
overriding the actual transaction isolation level.

Note

The SERIALIZABLE locking hint is equivalent to the HOLDLOCK locking hint.

This isolation level is important to prevent insertion that could invalidate your analysis. Listing 13.17 shows
an example of this isolation level.

Listing 13.17 Use the SERIALIZABLE Locking Hint to Prevent an Inconsistent Analysis Because of
Phantom Reads

USE Northwind
GO

534
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

-- Use the SERIALIZABLE or HOLDLOCK locking hints to prevent

-- inconsistent analysis due to Phantom Reads

DECLARE @ID int

BEGIN TRANSACTION

SELECT @ID = MAX(RegionID)

FROM Region (SERIALIZABLE)

INSERT Region
(RegionID, RegionDescription)
SELECT @ID + 1, 'Europe Union'

COMMIT TRANSACTION
GO

-- Use the SERIALIZABLE Isolation level to prevent

-- inconsistent analysis due to Phantom Reads in a long running process

SET TRANSACTION ISOLATION LEVEL SERIALIZABLE

GO

BEGIN TRANSACTION

-- Your queries

COMMIT TRANSACTION
Caution

SERIALIZABLE is the isolation level that allows the lowest concurrency. Use it only when
necessary and with the shortest possible transactions.

Types of Locks

SQL Server uses locks to provide concurrency while maintaining data consistency and integrity. Whenever
you read data, SQL Server locks the data while reading, to prevent other connections from modifying provide
the data at the same time.
If you modify data, SQL Server locks that data as long as the transaction lasts to avoid dirty reads in other
connections. These locks are called exclusive locks, and they are incompatible with any other locks.
SQL Server uses other types of locks to prevent specific concurrency problems and to guarantee the
consistency of internal operations.
Locks are maintained as long as they are necessary, depending on the isolation level selected. It is the
programmer's responsibility to design short transactions that execute quickly and do not lock the data for a
long time.
If a connection is trying to get a lock in a resource that is locked by another connection, the connection has to
wait until the resource is free or is locked in a compatible mode to continue the execution. SQL Server does
not produce any error message when a connection is being blocked by another transaction. Usually, this
blocking disappears shortly and the transaction continues.
If a connection is blocked for a longer time than the query timeout specified in the client database access
library, the application can determine whether the transaction should continue or should be aborted, perhaps
with the user's consent. In this case, it is the responsibility of the client application to search for the reasons
for this timeout, because SQL Server does not report specifically the existence of any blocking situation.
If necessary, SQL Server can lock any of the following resources:

• A row in a table, or RID.

535
Microsoft SQL Server 2000 Programming by Example

• A key in an index page.

• A page in a table.
• A page in an index.
• An extent in a database. SQL Server locks extents during the process of allocation and deallocation of
new extents and pages.
• An entire table.
• An index.
• A database. SQL Server always grants a shared lock on a database to the process that connects to a
database. This is helpful to detect whether users are connected to a database before trying intrusive
actions, such as dropping or restoring the database.

In some cases, SQL Server decides to escalate locks to an upper level to keep locking costs at an appropriat e
level. By using locking hints (PAGLOCK, ROWLOCK, TABLOCK, TABLOCKX), you can suggest that Query
Optimizer use locks at a certain level.
You can use the sp_lock system stored procedure to list the locks currently existing in the server. Specifying
the SPID (system process ID) of an existing process when calling sp_lock, you will retrieve the list of locks
that belong to a specific process. Listing 13.18 shows how to use sp_lock to get information about locks.

Listing 13.18 Use sp_lock to Get Information About Lockings

USE Northwind
GO

BEGIN TRAN

DECLARE @UP money

-- Read and lock the data to prevent updates

SELECT @UP = AVG(UnitPrice)

FROM [Order Details] (REPEATABLEREAD)
WHERE ProductID = 37

-- Update the Product 37

UPDATE Products
SET UnitPrice = @UP
WHERE ProductID = 37

-- List locks server wide

EXECUTE sp_lock

-- List locks for this connection only

EXECUTE sp_lock @@SPID

-- Cancel changes

ROLLBACK TRAN

536
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
52 8 0 0 DB S GRANT
53 6 0 0 DB S GRANT
54 6 0 0 DB S GRANT
56 6 0 0 DB S GRANT
56 6 325576198 4 KEY (84006690ff8d) S GRANT
56 6 117575457 0 TAB IX GRANT
56 6 325576198 1 KEY (cd0058f4cb88) S GRANT
56 6 325576198 1 KEY (fb00d3668ea6) S GRANT
56 6 117575457 1 KEY (2500ef7f5749) X GRANT
56 6 325576198 1 PAG 1:148 IS GRANT
56 6 325576198 1 KEY (4000a82ec576) S GRANT
56 6 325576198 1 PAG 1:182 IS GRANT
56 6 325576198 1 PAG 1:181 IS GRANT
56 6 325576198 4 PAG 1:198 IS GRANT
56 6 325576198 1 PAG 1:200 IS GRANT
56 6 325576198 1 PAG 1:208 IS GRANT
56 6 325576198 4 KEY (38008da3b95f) S GRANT
56 6 117575457 1 PAG 1:276 IX GRANT
56 6 325576198 4 KEY (86009acb8f9e) S GRANT
56 6 325576198 1 KEY (840022f8faea) S GRANT
56 6 325576198 0 TAB IS GRANT
56 6 325576198 4 KEY (cd00166aca95) S GRANT
56 1 85575343 0 TAB IS GRANT
56 6 325576198 4 KEY (fb00d769ae1d) S GRANT
56 6 325576198 1 KEY (860044a822ed) S GRANT
56 6 325576198 1 KEY (3800a93e5703) S GRANT
56 6 325576198 4 KEY (400055e1cf9b) S GRANT
57 6 0 0 DB S GRANT
58 8 0 0 DB S GRANT

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
56 6 0 0 DB S GRANT
56 6 325576198 4 KEY (84006690ff8d) S GRANT
56 6 117575457 0 TAB IX GRANT
56 6 325576198 1 KEY (cd0058f4cb88) S GRANT
56 6 325576198 1 KEY (fb00d3668ea6) S GRANT
56 6 117575457 1 KEY (2500ef7f5749) X GRANT
56 6 325576198 1 PAG 1:148 IS GRANT
56 6 325576198 1 KEY (4000a82ec576) S GRANT
56 6 325576198 1 PAG 1:182 IS GRANT
56 6 325576198 1 PAG 1:181 IS GRANT
56 6 325576198 4 PAG 1:198 IS GRANT
56 6 325576198 1 PAG 1:200 IS GRANT
56 6 325576198 1 PAG 1:208 IS GRANT
56 6 325576198 4 KEY (38008da3b95f) S GRANT
56 6 117575457 1 PAG 1:276 IX GRANT
56 6 325576198 4 KEY (86009acb8f9e) S GRANT
56 6 325576198 1 KEY (840022f8faea) S GRANT

537
Microsoft SQL Server 2000 Programming by Example

56 6 325576198 0 TAB IS GRANT

56 6 325576198 4 KEY (cd00166aca95) S GRANT
56 1 85575343 0 TAB IS GRANT
56 6 325576198 4 KEY (fb00d769ae1d) S GRANT
56 6 325576198 1 KEY (860044a822ed) S GRANT
56 6 325576198 1 KEY (3800a93e5703) S GRANT
56 6 325576198 4 KEY (400055e1cf9b) S GRANT
Using the SET LOCK_TIMEOUT statement, you can specify the number of milliseconds that a session will wait
for any lock to be released before reporting a timeout error in the current session, independently of the query
timeout defined in the client application. Use the @@TIMEOUT system function to get the current lock timeout
value, which can be

• @@TIMEOUT = 0—SQL Server cancels the query as soon as it detects a blocking situation, sending
the error 1222 "Lock request time out period exceeded."
• @@TIMEOUT = -1—SQL Server waits until the block disappears.
• @@TIMEOUT > 0—Indicates the number of milliseconds that SQL Server will wait before sending the
error 1222.

When the lock timeout occurs, the batch is cancelled and the execution goes to the next batch. Listing 13.19
shows an example of LOCK_TIMEOUT for two connections.

Listing 13.19 Use LOCK_TIMEOUT to Control How Much Time a Connection Will Wait for Locks to Be
Released

-- Connection A

USE Northwind
GO

BEGIN TRANSACTION

SELECT *
FROM Products (HOLDLOCK)

-- Note that we do not terminate the transaction.

---------------
-- Connection B
---------------

USE Northwind
GO

-- Specify 2000 milliseconds (2 seconds) as lock timeout

SET LOCK_TIMEOUT 2000

GO

UPDATE Products
SET UnitPrice = UnitPrice * 0.9
GO

538
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

IF @@ERROR = 1222
PRINT CHAR(10) + 'Lock Timeout produced'+ CHAR(10)

SELECT GetDate() as Now

Server: Msg 1222, Level 16, State 1, Line 1

Lock request time out period exceeded.

Lock Timeout produced

Now
------------------------------------------------------
2000-12-11 00:55:48.210
SQL Server uses the following lock types:

• Shared locks
• Exclusive locks
• Update locks
• Intent locks
• Schema locks
• Bulk Update locks

Note

Bulk Update locks are used to prevent other connections from using a table while importing data in
the table using the TABLOCK hint or setting the Table lock on Bulk Load option for the target table
with the sp_tableoption system stored procedure.

SQL Server can grant a lock in a resource depending on other existing locks. Figure 13.6 shows the
compatibility between locks.

Figure 13.6. This table shows the compatibility between locks.

539
Microsoft SQL Server 2000 Programming by Example

Caution

The locks produced in a process do not affect any statement in the same process, so it is
impossible to get blocked by any action performed in the same process.

A transaction is not affected by its own locks.

Shared Locks

During the process of reading data, SQL Server uses shared locks (S) to prevent changes to the data from
other processes. However, shared locks do not prevent reading operations from other connections. SQL
Server releases shared locks as soon as possible, unless the reading operation is inside a user-defined
transaction with an isolation level of REPEATABLE READ or SERIALIZABLE.
If a query reads many rows from a table, in a transaction with the default isolation level, only portions of that
table are locked at any given time. In this case, as soon as the statement terminates, the shared locks are
released without waiting to finish the transaction. Use the HOLDLOCK lock hint to maintain these shared locks
until the end of the transaction.
Note

Any DML statement (SELECT, UPDATE, INSERT, or DELETE) can produce shared locks, as long
as these statements are reading data from one or more tables.

540
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

Listing 13.20 shows how, after executing a SELECT statement, locks are released; however, after the
INSERT statement, you can see shared locks in the Products and Orders tables, because you use them
for the HOLDLOCK and REPEATABLEREAD lock hints.

Listing 13.20 Examples of Shared Locks

USE Northwind
GO

-- Start a transaction

BEGIN TRAN

-- Get OBJECT_IDs to interpret the sp_lock output

SELECT OBJECT_ID('Order details') as 'Order details'

SELECT OBJECT_ID('Orders') as 'Orders'
SELECT OBJECT_ID('Products') as 'Products'

-- Get current locks to use as a baseline to further executions of sp_lock

PRINT 'Initial lock status'+ CHAR(10)

EXEC sp_lock

-- Execute a standard SELECT statement

SELECT MAX(OrderID)
FROM Orders
WHERE CustomerID = 'ALFKI'

-- Get locks status after the SELECT statement

PRINT CHAR(10) + 'Lock status after SELECT'+ CHAR(10)

EXEC sp_lock

-- Perform an INSERT statement in Order details,

-- reading from Orders and Products

INSERT [Order Details]

(OrderID, ProductID, Unitprice, Quantity, Discount)
SELECT
(SELECT MAX(OrderID)
FROM Orders (HOLDLOCK)
WHERE CustomerID = 'ALFKI'),
ProductID, UnitPrice, 10, 0.1
FROM Products (REPEATABLEREAD)
WHERE ProductName = 'Tofu'
-- Get locks status after the INSERT statement

541
Microsoft SQL Server 2000 Programming by Example

PRINT CHAR(10) + 'Lock status after INSERT'+ CHAR(10)

EXEC sp_lock

ROLLBACK TRAN

-- Get locks status after the ROLLBACK statement

PRINT CHAR(10) + 'Lock status after ROLLBACK'+ CHAR(10)

EXEC sp_lock

Order details
-------------
325576198

(1 row(s) affected)

Orders
-----------
21575115

(1 row(s) affected)

Products
-----------
117575457

(1 row(s) affected)

Initial lock status

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 1 85575343 0 TAB IS GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

-----------
11011

(1 row(s) affected)

Lock status after SELECT

spid dbid ObjId IndId Type Resource Mode Status

542
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

------ ------ ----------- ------ ---- ---------------- -------- ------

51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 1 85575343 0 TAB IS GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

(1 row(s) affected)

Lock status after INSERT

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 117575457 0 TAB IS GRANT
55 6 0 0 DB S GRANT
55 6 325576198 5 PAG 1:189 IX GRANT
55 6 325576198 4 PAG 1:187 IX GRANT
55 6 325576198 3 PAG 1:202 IX GRANT
55 6 325576198 2 PAG 1:201 IX GRANT
55 6 325576198 1 PAG 1:211 IX GRANT
55 6 21575115 3 PAG 1:248 IS GRANT
55 6 21575115 3 KEY (d300d99e322e) RangeS-S GRANT
55 6 117575457 1 PAG 1:276 IS GRANT
55 6 117575457 4 PAG 1:281 IS GRANT
55 6 21575115 0 TAB IS GRANT
55 6 325576198 1 KEY (1100203d3739) X GRANT
55 6 325576198 2 KEY (1100203d3739) X GRANT
55 6 325576198 3 KEY (1100203d3739) X GRANT
55 6 325576198 0 TAB IX GRANT
55 6 21575115 3 KEY (18013d14dae8) RangeS-S GRANT
55 6 117575457 4 KEY (c80063450f36) S GRANT
55 1 85575343 0 TAB IS GRANT
55 6 325576198 5 KEY (11004b34faa9) X GRANT
55 6 325576198 4 KEY (11004b34faa9) X GRANT
55 6 117575457 1 KEY (0e00d057643e) S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

Lock status after ROLLBACK

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 1 85575343 0 TAB IS GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

543
Microsoft SQL Server 2000 Programming by Example

Exclusive Locks

SQL Server uses exclusive locks (X) to protect the modified data from other processes until the transaction
terminates. This lock mode is incompatible with any other lock. In the preceding section, you can see in
Listing 13.20 how an INSERT statement produces exclusive locks.
SQL Server creates exclusive locks when necessary, regardless of the selected isolation level. Inserted,
deleted, or updated resources are always locked exclusively until the transaction terminates.
You can still see these resources from other processes using the READ UNCOMMITTED isolation level or the
NOLOCK locking hint.
Tip

You can use mock updates to produce an exclusive lock in a row without modifying it. To SQL
Server, you have modified the data, even if the new values are the same as the old ones, as in
Listing 13.21.

Listing 13.21 Modifying Data Produces Exclusive Locks

-- Start a transaction

BEGIN TRAN

-- Get OBJECT_ID to interpret the sp_lock output

SELECT OBJECT_ID('Products') as 'Products'

-- Get current locks to use as a baseline to further executions of sp_lock

PRINT 'Initial lock status'+ CHAR(10)

EXEC sp_lock

-- Execute a mock UPDATE statement

UPDATE Products
SET UnitPrice = UnitPrice
WHERE ProductID = 28

-- Get locks status after the mock UPDATE statement

PRINT CHAR(10) + 'Lock status after the mock UPDATE'+ CHAR(10)

EXEC sp_lock

-- Execute a real UPDATE statement

UPDATE Products

544
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

SET UnitPrice = UnitPrice * 1.5

WHERE ProductID = 35

-- Get locks status after the real UPDATE statement

PRINT CHAR(10) + 'Lock status after the real UPDATE'+ CHAR(10)

EXEC sp_lock

ROLLBACK TRAN

Products
-----------
117575457

(1 row(s) affected)
Initial lock status

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 1 85575343 0 TAB IS GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

(1 row(s) affected)

Lock status after the mock UPDATE

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 117575457 0 TAB IX GRANT
55 6 0 0 DB S GRANT
55 6 117575457 1 PAG 1:276 IX GRANT
55 1 85575343 0 TAB IS GRANT
55 6 117575457 1 KEY (1c00c4c874c4) X GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT

(1 row(s) affected)

545
Microsoft SQL Server 2000 Programming by Example

Lock status after the real UPDATE

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 117575457 0 TAB IX GRANT
55 6 0 0 DB S GRANT
55 6 117575457 1 KEY (230033203c6c) X GRANT
55 6 117575457 1 PAG 1:276 IX GRANT
55 1 85575343 0 TAB IS GRANT
55 6 117575457 1 KEY (1c00c4c874c4) X GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
Note

Modifying a row in a table with no clustered index produces a RID (Row ID) lock. If the table has a
clustered index, the index leaf level is the collection of data pages; this is why the row lock is
considered a key lock in the sp_lock output.

Update Locks

Update locks are an intermediate state between a shared lock and an exclusive lock. You use an update lock
when you are reading some data with intentions of modifying the same data. To prevent modifications from
other connections, you could lock the data exclusively, but in that case, you are preventing readings from
other connections. Using update locks prevents undesired modifications, while allowing read access to the
data from other connections.
If you used only a shared lock in the resource lock, and other connection gets another shared lock in the
same resource, you can get a deadlock situation if both connections tried to convert their shared lock into
exclusive locks.
In Listing 13.22, you can see an example of update locks. To execute this example, you must establish two
connections to SQL Server:

1. Execute the first part of the script in Connection A, which reads a row from the Products table, using
the UPDLOCK locking hint.
2. Attempt to execute an UPDATE statement of the same row from Connection B. The statement waits
for locks to be released, because the existing update lock is incompatible with the exclusive lock
required to modify data.
3. Execute an UPDATE statement from Connection A. Because this is the connection that holds the
update lock, it is possible to update the data directly. In this case, the update lock is converted into an
exclusive lock.
4. Connection B is still waiting because the existing exclusive lock is incompatible with the requested
exclusive lock.

Listing 13.22 An Example of How to Use UPDLOCK

546
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

-- Start a transaction

BEGIN TRAN

-- Get OBJECT_ID to interpret the sp_lock output

SELECT OBJECT_ID('Products') as 'Products'

-- Get current locks to use as a baseline to further executions of sp_lock

PRINT 'Initial lock status'+ CHAR(10)

EXEC sp_lock

SELECT ProductID, ProductName, UnitPrice

FROM Products (UPDLOCK)
WHERE ProductID = 28

PRINT CHAR(10) + 'lock status after SELECT'+ CHAR(10)

EXEC sp_lock

-- Change to connection B and execute:

/*

UPDATE Products
SET UnitPrice = UnitPrice
WHERE ProductID = 28

*/

-- Change to connection A and execute:

PRINT CHAR(10) + 'lock status after UPDATE from Connection B'+ CHAR(10)

EXEC sp_lock

-- Execute a mock UPDATE statement

UPDATE Products
SET UnitPrice = UnitPrice
WHERE ProductID = 28

-- Get locks status after the mock UPDATE statement

PRINT CHAR(10) + 'Lock status after the mock UPDATE'+ CHAR(10)

EXEC sp_lock

ROLLBACK TRAN

547
Microsoft SQL Server 2000 Programming by Example

Products
-----------
117575457

(1 row(s) affected)

Initial lock status

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
55 1 85575343 0 TAB IS GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT

ProductID ProductName UnitPrice

----------- ---------------------------------------- ---------------------
28 Rössle Sauerkraut 45.6000

lock status after SELECT

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
55 6 117575457 0 TAB IX GRANT
55 6 117575457 1 PAG 1:276 IU GRANT
55 1 85575343 0 TAB IS GRANT
55 6 117575457 1 KEY (1c00c4c874c4) U GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT
lock status after UPDATE from Connection B

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
55 6 117575457 0 TAB IX GRANT
55 6 117575457 1 PAG 1:276 IU GRANT

548
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

55 1 85575343 0 TAB IS GRANT

55 6 117575457 1 KEY (1c00c4c874c4) U GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT
58 6 117575457 1 KEY (1c00c4c874c4) X WAIT
58 6 117575457 1 PAG 1:276 IX GRANT
58 6 117575457 0 TAB IX GRANT

Lock status after the mock UPDATE

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
55 6 117575457 0 TAB IX GRANT
55 6 117575457 1 PAG 1:276 IX GRANT
55 1 85575343 0 TAB IS GRANT
55 6 117575457 1 KEY (1c00c4c874c4) X GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT
58 6 117575457 1 KEY (1c00c4c874c4) X WAIT
58 6 117575457 1 PAG 1:276 IX GRANT
58 6 117575457 0 TAB IX GRANT

Intent Locks

SQL Server can lock resources at different hierarchical levels: rows, pages, extents, tables, and databases. If
one process holds a lock in a row, a different process could get a lock on the page where this row is stored,
compromising the lock in the other connection. Another process could get a lock on the table, adding more
complexity to this locking situation.
To prevent this situation, SQL Server uses intent locks at a higher level in the hierarchy, preventing
incompatible levels from other connections.
If SQL Server grants shared locks at row level to a process, it grants an intent share (IS) lock at page and
table level also, protecting the requested shared row lock. Figure 13.7 shows this case.

Figure 13.7. SQL Server uses intent locks to protect granted locks at lower hierarchical levels.

549
Microsoft SQL Server 2000 Programming by Example

To protect an exclusive (X) row lock, SQL Server grants an intent exclusive (IX) lock at page and table level.
In some cases, a SQL Server transaction requests shared locks on some rows and exclusive locks on other
rows. In these cases, SQL Server will produce intent shared (IS) locks to protect the data pages that have
only rows with shared locks. However, if a data page has at least one row with an exclusive lock, SQL Server
will protect this page with an intent exclusive (IX) lock. Because some pages have intent shared locks and
other pages have exclusive shared locks, at the table level SQL Server will use the shared with intent
exclusive (SIX) lock.

Schema Locks

When SQL Server receives Data Definition Language (DDL) statements, it locks the object being created or
modified with a schema modification lock (Sch-M), as in the example of Listing 13.23.

Listing 13.23 SQL Server Uses Schema Modification Locks to Protect Objects During Creation or
Modification

550
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

USE Northwind
GO

BEGIN TRAN

CREATE TABLE TestSchema

(ID int)

-- Show names of the objects 1, 2 and 3 of Northwind,

-- because they are used in sp_lock

SELECT object_name(1) AS ID1

SELECT object_name(2) AS ID2

SELECT object_name(3) AS ID3

-- Show the ID of TestSchema table to identify locks on this table

SELECT Object_ID('TestSchema') AS TestSchemaID

EXEC sp_lock

ROLLBACK TRAN

ID1
-------------
sysobjects

ID2
-------------
sysindexes

ID3
-------------
syscolumns

TestSchemaID
------------
1813581499

(1 row(s) affected)

551
Microsoft SQL Server 2000 Programming by Example

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
57 6 1 0 TAB IX GRANT
57 6 3 0 TAB IX GRANT
57 6 2 0 TAB IX GRANT
57 6 1 3 KEY (bb006c403485) X GRANT
57 6 2 1 KEY (bb00da29dced) X GRANT
57 6 1813581499 0 TAB Sch-M GRANT
57 6 1 2 KEY (f00169258f34) X GRANT
57 6 3 2 KEY (04018aa462b1) X GRANT
57 1 85575343 0 TAB IS GRANT
57 6 1 1 KEY (bb00194052c1) X GRANT
57 6 3 1 KEY (bc00403fd981) X GRANT
58 6 0 0 DB S GRANT
As you saw in the output of Listing 13.23, there is a schema modification (Sch-M) lock on the table
TestSchema and exclusive locks on some rows of the system tables sysobjects, syscolumns, and
sysindexes.
During the process of compiling a query plan, SQL Server uses a schema stability (Sch-S) lock on the objects
used in the query plan to prevent modifications on their definition that could potentially invalidate the query
plan being compiled.
Changes to objects referenced in an existing compiled query plan force the query plan to be invalid, and SQL
Server marks the query plan to be recompiled next time it is executed.

Key-Range Locks

SQL Server uses key-range locks to avoid phantom reads. As commented in the "Concurrency Problems"
section earlier in this chapter, this is a special case of inconsistent analysis because of the insertion of rows
from other processes in the range of rows affected by the process.
Key-range locks help serialize the following operations:

• Queries restricted by a range of values, to prevent other rows from being inserted in this range by
other processes.
• Fetching a nonexistent row, with intentions of inserting the row at a later point in the transaction.
• Insert operations.

SQL Server implements range locks by locking keys in an index when you select the SERIALIZABLE isolation
level, as you can see in Listing 13.24.

Listing 13.24 Use the SERIALIZABLE Locking Hint to Force Range Locks

USE Northwind
GO

552
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

-- Start a transaction

BEGIN TRAN

-- Get OBJECT_ID to interpret the sp_lock output

SELECT OBJECT_ID('Products') as 'Products'

-- Get current locks to use as a baseline to further executions of sp_lock

PRINT 'Initial lock status'+ CHAR(10)

EXEC sp_lock

SELECT ProductID, ProductName, UnitPrice

FROM Products (SERIALIZABLE)
WHERE CategoryID = 3
PRINT CHAR(10) + 'lock status after SELECT'+ CHAR(10)

EXEC sp_lock

ROLLBACK TRAN
Products
-----------
117575457

(1 row(s) affected)

Initial lock status

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT
61 6 0 0 DB S GRANT
61 1 85575343 0 TAB IS GRANT
62 6 0 0 DB S GRANT
63 6 0 0 DB S GRANT
68 6 0 0 DB S GRANT
69 6 0 0 DB S GRANT

ProductID ProductName UnitPrice

----------- ---------------------------------------- ---------------------
16 Pavlova 17.4500
19 Teatime Chocolate Biscuits 9.2000

553
Microsoft SQL Server 2000 Programming by Example

20 Sir Rodney's Marmalade 81.0000

21 Sir Rodney's Scones 10.0000
25 NuNuCa Nuß-Nougat-Creme 14.0000
26 Gumbär Gummibärchen 31.2300
27 Schoggi Schokolade 43.9000
47 Zaanse koeken 9.5000
48 Chocolade 12.7500
49 Maxilaku 20.0000
50 Valkoinen suklaa 16.2500
62 Tarte au sucre 49.3000
68 Scottish Longbreads 12.5000

(13 row(s) affected)

lock status after SELECT

spid dbid ObjId IndId Type Resource Mode Status

------ ------ ----------- ------ ---- ---------------- -------- ------
51 4 0 0 DB S GRANT
52 4 0 0 DB S GRANT
53 10 0 0 DB S GRANT
54 7 0 0 DB S GRANT
55 6 0 0 DB S GRANT
56 10 0 0 DB S GRANT
57 6 0 0 DB S GRANT
58 6 0 0 DB S GRANT
61 6 0 0 DB S GRANT
61 6 117575457 0 TAB IS GRANT
61 6 117575457 1 KEY (310027bf2c96) S GRANT
61 6 117575457 1 KEY (3200c9109984) S GRANT
61 6 117575457 1 KEY (3e0071af4fce) S GRANT
61 6 117575457 1 KEY (2f008b9fea26) S GRANT
61 6 117575457 2 KEY (1700bde729cb) RangeS-S GRANT
61 6 117575457 2 KEY (1e00ebf74a93) RangeS-S GRANT
61 6 117575457 1 KEY (1b007df0a359) S GRANT
61 6 117575457 1 KEY (14002be0c001) S GRANT
61 6 117575457 1 PAG 1:276 IS GRANT
61 6 117575457 2 PAG 1:277 IS GRANT
61 6 117575457 2 KEY (32001d9803ec) RangeS-S GRANT
61 6 117575457 2 KEY (4100e7a8a604) RangeS-S GRANT
61 6 117575457 2 KEY (3400b1b8c55c) RangeS-S GRANT
61 6 117575457 2 KEY (35005f17704e) RangeS-S GRANT
61 6 117575457 1 PAG 1:360 IS GRANT
61 1 85575343 0 TAB IS GRANT
61 6 117575457 1 KEY (440089efcdca) S GRANT
61 6 117575457 1 KEY (300042d8902e) S GRANT
61 6 117575457 2 KEY (1c00603f4339) RangeS-S GRANT
61 6 117575457 2 KEY (1d008e90f62b) RangeS-S GRANT
61 6 117575457 2 KEY (160004dffe56) RangeS-S GRANT
61 6 117575457 2 KEY (1300ea704b44) RangeS-S GRANT
61 6 117575457 2 KEY (1800d8809573) RangeS-S GRANT
61 6 117575457 1 KEY (15004e877cb9) S GRANT
61 6 117575457 1 KEY (130092d8179c) S GRANT
61 6 117575457 1 KEY (10007c77a28e) S GRANT
61 6 117575457 1 KEY (1900f638aaf3) S GRANT
61 6 117575457 1 KEY (1a0018971fe1) S GRANT
61 6 117575457 2 KEY (3300d4df79e4) RangeS-S GRANT
61 6 117575457 2 KEY (0f006da996c9) RangeS-S GRANT
61 6 117575457 2 KEY (47001fe82400) RangeS-S GRANT
62 6 0 0 DB S GRANT

554
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

63 6 0 0 DB S GRANT
68 6 0 0 DB S GRANT
69 6 0 0 DB S GRANT

A Serious Problem to Avoid: Deadlocks

Imagine that your database application has two users: Paul and Mary.
Paul starts a transaction and modifies some attributes of the Acme Ltd. customer. Later, inside the same
transaction, Paul tries to modify this customer's payments. However, Paul cannot modify these payments
because Mary holds an exclusive lock on these payment records. Paul must wait for these records to be
unlocked before completing the transaction.
Mary is modifying customers' payments, and that's why this information is locked. Inside the same transaction,
Mary tries to modify some data about the Acme Ltd. customer. At this moment, Paul, who modified this record
just a few minutes ago, locks this information.
Mary cannot update this information because Paul is holding an exclusive lock on it, so Mary must wait for this
resource to be unlocked before proceeding with her transaction. However, Paul cannot continue with his
transaction because he's waiting for Mary to unlock the information he needs to update.
This situation of mutual blockings is called deadlock. If SQL Server detects this situation, it decides which
process has a bigger execution cost, and selects this process as a winner. After the winner is selected, SQL
Server notifies the other processes waiting in this deadlock situation with error 1205, telling them that they
have been selected as victims in a deadlock situation.
If the processes involved in a deadlock situation are blocking one another in a circular reference, SQL Server
selects which process can be selected to break the deadlock with the least overall cost, and notifies this
process with error 1205.
Note

You can propose your specific session as a potential deadlock victim by using the statement SET
DEADLOCK_PRIORITY LOW.

Two processes can create a deadlock situation when they access resources in opposite orders and try to
convert a shared lock into an exclusive lock at the same time. Figure 13.8 illustrates this scenario:

Figure 13.8. A typical deadlock situation.

555
Microsoft SQL Server 2000 Programming by Example

1. Connection A starts a transaction and reads the UnitPrice column from the Product 37. This
connection uses the HOLDLOCK locking hint to maintain the shared lock on the row corresponding to
Product 37.
2. Connection B starts a transaction and reads the average UnitPrice from the Order Details
table for Product 37. This connection uses the HOLDLOCK locking hint to maintain the shared lock on
the Order Details rows from Product 37.
3. Connection A tries to update the Order Details table to reset the unit price of Product 37 to the
value stored in the Products table. To execute this statement, Connection A needs an exclusive lock
on the affected rows, but this exclusive lock must wait because Connection B holds a shared lock on
the same rows.
4. Connection B tries to update Product 37 in the Products table with the average unit price retrieved
from the Order Details table. Connection B requests an exclusive lock on Product 37, but this lock
must wait because Connection A holds a shared lock on it.
5. SQL Server detects this deadlock situation, selects Connection B as victim of this situation, and sends
message 1205 to Connection B. Resources locked by Connection B are unlocked.

556
 Chapter 13. Maintaining Data Consistency: Transactions and Locks

6. After Connection B has been selected as a victim and its locks have been released, Connection A can
continue its operation.

Another typical case is when two transactions want to convert an existing shared lock on a common locked
resource into an exclusive lock. To prevent this situation, you should use the UPDLOCK locking hint in
transactions in which you read data with intentions of updating it later in the same transaction.
Caution

When a transaction is selected as a victim in a deadlock situation, the process is cancelled and
changes applied are rolled back. However, the calling application could usually resend the
transaction and, hopefully, the previous locks have disappeared.

Avoiding deadlock is not always possible; however, you can help to reduce deadlocks by following these
guidelines:

• Keep transactions as short as possible.

• Avoid user interaction inside transactions. In other words, start a transaction only when required and
release it as soon as possible.
• Always access resources in the same order and check for potential circular references.
• Use the READ COMMITTED isolation level if possible, because it produces fewer locks than higher
isolation levels. Try to avoid SERIALIZABLE as much as possible.
• If an application uses several connections, bind them to share the same locking space. You can
execute the stored procedure sp_bindsession to keep more than one session in the same
transaction.

What's Next?

Transactions and locks are key aspects to provide the adequate concurrency to your database application in a
multiuser environment. However, they are restricted, as covered in this chapter, to a single-server operation.
The following two chapters focus on the multiserver environment from two different perspectives:

• Chapter 14 shows how to transfer data to and from SQL Server databases stored in the same or
different servers. Data Transformation Services (DTS) is a feature-rich application which, integrated in
SQL Server or as a standalone subsystem, transfers data between hetero-geneous systems,
including all the required transformations.
• Chapter 15 discusses the multiserver environment and the implications of the distributed
transactions. In Chapter 15, you learn how to use linked servers to maintain data in multiple servers,
as an alternative to DTS and Replication.

557
 Chapter 14. Transferring Data to and from SQL Server

Chapter 14. Transferring Data to and from SQL Server

In a standard business environment, it is quite common to have different system platforms, different operating
systems, heterogeneous networks, and different database systems. Linking existing data from different
sources is a convenient way to work with heterogeneous data to gain data consistency through the company
without creating any data redundancy. However, in some cases, you might need to transfer data from one
system to another.
Importing and exporting data is a common task for a database administrator, and it is not our intention to
cover this subject in detail. However, as a database programmer, you should know the basics of importing
and exporting data, and this chapter will teach you how to solve this problem.
This chapter teaches you the following:

• Why you need to transfer and transform data

• SQL Server 2000 tools for transferring data
• How to use the BULK INSERT statement
• How to use the bcp command-line utility
• How to use the Copy Database Wizard

The Need for Transferring Data

If your company has a single database, in a single server, and you never need to receive data from other
systems or send data to other servers, you could skip this chapter.
Many systems receive their data through direct user input. However, there are some cases where transferring
data is important:

• You want to migrate to a new system and you want to populate the new database with data coming
from your old system.
• Your accounting system works in a mainframe and you do not want to change this system. However,
it would be useful to have some accounting information in the SQL Server Sales database. In this
case, you must periodically refresh this information from the mainframe.
• The post office changes the national postal code information and they distribute this new information
as a CSV file. You need to import this file into your system to update the Customer Management
application.
• The Inland Revenue changes their requirements and now the annual accounts must be sent in a
different format. You must create the process of exporting data in exactly the way they require.
• You create a testing server in your network and you want to have the same databases as in your
production server to test a new indexing strategy.
• Your sales managers visit customers, and they want to have a copy of the Sales System database in
their laptops so they can look at sales figures when they are at the customer site.
• Your corporation has many different companies in different countries, and you want to receive
periodic financial information from them. Every one of these companies uses a different system, and
the only way to receive data is by text files, so you can import them easily.
• You have a Documents database and you receive many documents from different sources. You want
to import them into the Documents database efficiently.
• You are running a Geographical Information System and your field teams send you files every week
with their field measurements. You need to integrate this new data with your existing GIS database.
• You just finished a new Agricultural Census in your county and you want to compare this new data
with the latest census's data. The old data is in a different system and you want to import the old data
to consolidate both databases.
• Your remote offices need to produce reports about their local sales figures. They complain because
they need to access your central mainframe to produce these reports, but the mainframe connection
is not always available. You decide that a good solution is to have a local database with local data to
produce reports locally. You need to refresh these local databases periodically to have their data
synchronized with the central database.
• Your network administrators are concerned about a potential bottleneck on your central database
system. A feasible solution is to install departmental servers with replicated data. In this way, users

559
Microsoft SQL Server 2000 Programming by Example

can receive data from a local server, in the same network segment, without traversing the entire
network to arrive to the data center.

SQL Server 2000 provides different tools to transfer data from any source to any destination. Depending on
your specific requirements, one tool can be more appropriate than another. You will learn about the SQL
Server 2000 tools used to transfer data in the next section of this chapter.
In other cases, the problem is not only transferring data, but also modifying data from the source database to
meet the requirements of the destination database system. Some examples are as follows:

• You have a relational database and you need to create a data warehouse database with a different
database schema; in this case, it could be a star schema.
• Your legacy system in the USA stores dates in a different format (mmddyyyy) from the legacy system
you have in France (ddmmyyyy). You want to make sure you can import dates correctly to your
central server in Indonesia, which uses the ISO/ODBC standard format (yyyy-mm-dd).
• After a company merge, you need to consolidate data from two different systems. In one system, the
codes used in lookup tables are different from the codes used in the other system. In the Spanish
system, end users can be S (solteros), C (casados), D (divorciados o separados), V (viudos). In the
British system, end users can be S (single), M (married), D (divorced), W (widow or widower). You need
to agree about new codes and transform the old ones.
• You just bought a bank in Morocco, and you see that their database system identifies customer
accounts by their full name, including title. You want to provide a new account identification number
and store title, family name, and first name in separate fields.
• You work in an international project and you need to integrate data in different currencies. Your
system selects Euro as the standard internal currency and you must transform all quantities into
Euros and store the exchange rate applied to every amount in a different field.
• You created a weather database to help on global weather forecasts. This system receives
continuous information from weather systems around the world, each one using different units for
temperature, rainfall, pressure, and so on. You must convert the data to uniform units to be able to
produce consistent results.

Data Transformation Services 2000 can help you create complex packages that transfer and transform the
data to meet the requirements of the destination database.

Tools for Transferring Data Using SQL Server 2000

SQL Server 2000 offers many different choices to transfer data. Every tool has advantages and
disadvantages. You can use the following examples as guidelines to select the data distribution tool to use:

• Using distributed queries, you can directly access data from different servers. Chapter 15,
"Working with Heterogeneous Environments: Setting Up Linked Servers," covers distributed
queries in detail.
• You can use replication to copy data from one server to another, on demand or at regular intervals. If
you need to distribute data to mobile users, and they need to modify data locally, merge replication is
an excellent solution. Transactional replication is a very efficient mechanism to distribute data
changes to remote servers, if the latency inherent to replication is acceptable in your case. Replication
is not covered in this book. Books Online contains a full section about replication, with comprehensive
information about how replication works.
• You can back up a database in a SQL Server 7.0 or 2000 server and restore it in another SQL Server
2000 server. If you restore a SQL Server 7.0 database into SQL Server 2000, the restore process
modifies the database internal physical structure to adapt it to the new SQL Server 2000 physical
structure. Restoring SQL Server 2000 databases into SQL Server 7.0 is not supported. Backup is not
covered in this book because it is an administrative task. Books Online contains the "Backing Up and
Restoring Databases" section, where you can find more information about this topic.

Note

560
 Chapter 14. Transferring Data to and from SQL Server

Contrary to what happened with SQL Server 7.0, in SQL Server 2000 you can restore databases
from servers with different collations, because every database has its own collation, independent
from the server default collation.

• You can detach a database from a server running SQL Server 7.0 or 2000, copy the database files to
another server, and attach them to the destination server. This procedure is more efficient than using
backup and restore. After you attach a SQL Server 7.0 database into SQL Server 2000, it is converted
to the new database structure. Attaching SQL Server 2000 databases into SQL Server 7.0 is not
supported. Look in Books Online for information on how to use the stored procedures sp_detach_db
and sp_attach_db.

Note

In SQL Server 2000, you can attach databases that have been detached from servers with different
collations, because every database has its own collation, independent from the server default
collation.

• You can convert SQL Server 6.5 databases into SQL Server 2000 running the SQL Server Upgrade
Wizard. Look in Books Online for information on "Upgrading Databases from SQL Server 6.5
(Upgrade Wizard)."
• Data Transformation Services (DTS) is a flexible and powerful tool that you can use to import and
export data, and transform the data as well. You will learn how to import and export data using DTS
later in this chapter, in the "Using Data Transformation Services" section.
• Use the bcp command-line utility to import and export data to and from SQL Server 2000. You learn
how to use bcp in the next section of this chapter.
• Use the new BULK INSERT statement in a batch or stored procedure to import data from a file into a
SQL Server 2000 table. The next section of this chapter covers this tool in detail.
• You can use the ODBC bulk copy application programming interface (API), as the bcp utility does,
using any programming language to create your own transferring application. To get more information
about this interesting programming solution, search in Books Online for the section "How to Bulk Copy
with the SQL Server ODBC Driver (ODBC)."
• You can write an application using the SQL-DMO library, and use the Transfer and Transfer2
objects'properties and methods to transfer data and schema between SQL Server 2000 or SQL
Server 7.0 servers. Search in Books Online for the "Transfer Object" topic.

The BULK INSERT Statement and bcp

You can use the bcp command-line utility to export a table or the result of a query to an external data file. You
can copy this file over the network or the Internet, or use any media to send it to its destination. It also can be
used to import the data file into a single table.
You can use the bcp native mode to export data to and from SQL Server databases. If you export data from
SQL Server in native mode, you cannot import that data into any other database system than SQL Server.
However, using character-based files provides better flexibility, because the data can be exported to any
database system that supports importing from text files.
Tip

561
Microsoft SQL Server 2000 Programming by Example

Using bcp in native mode, between SQL Server databases, is more efficient than using character
mode.

To use bcp, you must open a command prompt window and execute this utility from there.

If you want to import data from a data file into SQL Server, using Transact-SQL language, you can use the
new BULK INSERT statement. This method of importing data is highly efficient, and you should use it to
perform simple import operations of big data files.

Using bcp and BULK INSERT is faster than inserting the same information, row by row, either manually or
from a client application.

By default, constraints and triggers are ignored when importing data using bcp or BULK INSERT, providing a
faster inserting operation. However, you should check the data to guarantee that it complies with the existing
constraints.

Tip

If you define triggers in your tables to maintain denormalized data in other tables, you should
create a stored procedure with similar functionality to apply to the imported data after the bulk
operation terminates. In this case, it is better if the stored procedure executes both operations in
sequence: the import process and the post-import maintenance operations.

In the next section, you will see how to enable or disable constraint checking and trigger execution during the
bulk copy operations.

If your destination database uses a full recovery model, the import operation must be fully logged, and you will
be potentially running out of space in the transaction log.

The fastest way to import data into SQL Server is by executing a minimally logged bulk-copy operation, which
can be performed if all these conditions are met:

• The database recovery model is set to simple or bulk-logged.

• The destination table is not replicated.
• The destination table does not have any triggers.
• The destination table is empty or does not have any indexes.
• You run the bulk copy operation, specifying the TABLOCK hint.

If the destination of the bulk copy operation does not meet any of these conditions, the operation will be
logged.
Tip

562
 Chapter 14. Transferring Data to and from SQL Server

If the destination table has indexes, it is recommended to drop the indexes before importing the
data and re-creating them after the data is imported. In this case, the sequence should be as
follows:

1. Drop nonclustered indexes.

2. Drop the clustered index, if it exists.
3. Import the data.
4. Create the clustered index.
5. Create the nonclustered indexes.

However, for extremely big tables, when the data to import does not represent an appreciable
percentage of the existing volume of data, this technique is not recommended, because the internal
index maintenance during the importing process will be more efficient than the full rebuild of
existing indexes.

Tip

The first time you import a new type of file using a bulk-copy operation, you should import the data
to a provisional table first, check the data you just imported to see whether the importing process is
done correctly, and when you are certain that the operation works as expected, you can consider
the process as valid, and perform the bulk-copy operation on the destination table.

Using the bcp Command-Line Utility

The bcp command-line utility copies data from SQL Server to an external data file and imports data from an
external data file into SQL Server.
Note

The bcp utility uses the ODBC Bulk Copy Application Programming Interface (API). It is compatible
with any version of SQL Server.

To test the bcp utility, open a command prompt window and execute bcp /?, as in Listing 14.1.

Listing 14.1 Get Syntax Help About How to Execute bcp

C:\TEMP>bcp /?

563
Microsoft SQL Server 2000 Programming by Example

usage: D:\Program Files\Microsoft SQL Server\80\Tools\BINN\bcp.exe {dbtable |

query} {in | out | queryout | format} datafile
[-m maxerrors] [-f formatfile] [-e errfile]
[-F firstrow] [-L lastrow] [-b batchsize]
[-n native type] [-c character type] [-w wide character type]
[-N keep non-text native] [-V file format version] [-q quoted identifier]
[-C code page specifier] [-t field terminator] [-r row terminator]
[-i inputfile] [-o outfile] [-a packetsize]
[-S server name] [-U username] [-P password]
[-T trusted connection] [-v version] [-R regional enable]
[-k keep null values] [-E keep identity values]
[-h "load hints"]
In this section, we will take a look at some of these options, step by step.
Now, in the same command prompt window, you can write the instruction from Listing 14.2 to export the
Northwind.dbo.Region table to the external file region.txt in character format, and use your NT or
Windows 2000 credentials to connect to SQL Server.

Listing 14.2 Export the Region Table to the region.txt External File Using bcp

C:\TEMP>bcp northwind.dbo.region out region.txt -S YourServer\YourInstance -T -c

Starting copy...

4 rows copied.
Network packet size (bytes): 4096
Clock Time (ms.): total 20
Looking at the instruction you just typed, see the following options:

• bcp is the program to execute.

• northwind.dbo.region is the fully qualified name of the table to export. You can specify the name
of a view, an inline user-defined function, or a table-valued function, as shown in Listing 14.3.
• out specifies that you want to export data.
• region.txt is the name of the file to fill with the exported data.
• -S YourServer\YourInstance specifies the server and instance to connect to. If you want to
export from the default instance, use –S YourServer instead.
• -T instructs bcp to use your NT or Windows 2000 credentials to connect to SQL Server, using
integrated authentication.
• -c means the data is exported using text mode.

Listing 14.3 Export the Result of the dbo.TopTenOrders Inline User-Defined Function to the
topten.txt External File Using bcp

564
 Chapter 14. Transferring Data to and from SQL Server

C:\TEMP>bcp northwind.dbo.toptenorders() out topten.txt -S

YourServer\YourInstance -T -c
Starting copy...

10 rows copied.
Network packet size (bytes): 4096
Clock Time (ms.): total 541
Note

You created the TopTenOrders inline user-defined function in Listing 10.14 from Chapter 10,
"Enhancing Business Logic : User-Defined Functions (UDF)."

To look at the file region.txt, you can use the type command, as seen in Listing 14.4.

Listing 14.4 Inspect the Contents of the Exported File region.txt

C:\TEMP>type region.txt
1 Eastern
2 Western
3 Northern
4 Southern
Now, you can try to import the file into a new table, using bcp again. In the same command-prompt window,
write the instruction contained in Listing 14.5.

Listing 14.5 Import the region.txt File into a New Table Called NewRegions Using bcp

C:\TEMP>bcp northwind.dbo.NewRegions in region.txt -S YourServer\YourInstance -T

-c

565
Microsoft SQL Server 2000 Programming by Example

SQLState = S0002, NativeError = 208

Error = [Microsoft][ODBC SQL Server Driver][SQL Server]Invalid object name
'northwind.dbo.NewRegions'.
You got an error message because bcp can import data only into existing tables, either directly, through
appropriate views, or inline user-defined functions.
To solve this problem, you must create the destination table first. You can do it easily from Query Analyzer
with the CREATE TABLE statement, or graphically in Enterprise Manager. However, you can do it as well from
the command prompt, using the osql utility to connect to SQL Server and execute the CREATE TABLE
statement. Listing 14.6 shows the execution of both osql and bcp.

Listing 14.6 Create the NewRegions Table and Import the region.txt File

C:\TEMP>osql -S YourServer\YourInstance -E -d Northwind -Q "CREATE TABLE

NewRegions (ID
int, Name nchar(50))"

C:\TEMP>bcp northwind.dbo.NewRegions in region.txt - S YourServer\YourInstance -T

–c
Starting copy...

4 rows copied.
Network packet size (bytes): 4096
Clock Time (ms.): total 311
Now, you can use osql again, as in Listing 14.7, to look at the new table NewRegions and test whether the
import operation succeeded.

Listing 14.7 Use osql to Read Data from the NewRegions Table

C:\TEMP>osql -S YourServer\YourInstance -E -d Northwind -Q "SELECT * FROM

NewRegions"
ID Name
---------- ----------------------------------------------------
1 Eastern
2 Western
3 Northern
4 Southern

(4 rows affected)
You can use bcp to export the result from any query into a file, using the queryout option, as in Listing
14.8.

566
 Chapter 14. Transferring Data to and from SQL Server

Listing 14.8 Export the Result of a Query to the query.txt External File Using bcp and the queryout
Option

C:\TEMP>bcp "SELECT CategoryID, CategoryName FROM Northwind.dbo.Categories"

queryout
query.txt -S YourServer\YourInstance -T -c
Starting copy...

10 rows copied.
Network packet size (bytes): 4096
Clock Time (ms.): total 1
You can limit the number of errors to accept during the bulk copy operation by using the -m option. The
default value is 10. Every row that produces an error is disregarded by bcp, and the execution continues
until the number of errors is greater than 10 or the number specified within the –m option, in which case the
operation is cancelled.
Using the –e err_file option, bcp sends rows with transfer errors to the err_file file. You can later
review this file, correct any error, and retry the import operation only with these rows.
If you want to import specific rows only from the data file, use -F first_row and –L last_row to specify
the first and last rows to import. If you do not use the –F option, the transfer process starts from the first row. If
you do not use the –L option, the transfer continues to the end of the file.
The default field terminator is the tab character (\t or CHAR(9)), but you can specify your own field
terminator with the –t option. The default row terminator is the newline character (\n or CHAR(10)), but you
can specify your own row terminator with the –r option.
In the examples from Listings 14.1 through 14.8, we always used character format. However, bcp accepts
more formats:

• -n uses native SQL Server mode; therefore, every field is exported using its native storage format.
This mode is very efficient if you need to transfer data between SQL Server databases. Use the –N
option to send character data as UNICODE, and any other data type in its native format.
• -c uses the character data type. This option uses the tab character (\t) as field separator and the
newline character (\n) as row terminator. Use this format to transfer data to non-SQL Server
databases. Use the –w option if you want to output data in UNICODE (double byte) format.
• -V60, -V65, -V70 uses data types from old versions of SQL Server.

If the query to execute is too long to be written inline with the bcp command, you can create a text file and
use it as an input file with the –i input_file option. For similar reasons, if you expect too many messages
to fit in the command-prompt window, you can specify an output file with the –o output_file option.
In the preceding examples, we used integrated authentication (with the –T option) to connect bcp to SQL
Server, but you can use SQL Server authentication using the –U login_id and -P password options.
By default, bcp does not fire any AFTER INSERT or INSTEAD OF INSERT triggers on the destination table,
but you can force the execution of triggers using the –h "FIRE_TRIGGERS" hint. This option is valid only if
the in option is specified. The triggers are fired only once per batch during the bulk copy operation, and the
inserted and deleted tables contain the complete set of imported rows on that batch.
As with triggers, constraints are not checked during data import operations using bcp. If you want to enforce
constraints for every imported row, you can use the –h "CHECK_CONSTRAINTS" hint.
If you want to perform a minimum logged bulk copy operation, you must use the –h "TABLOCK" hint as well,
as mentioned earlier in this chapter.
If you want to use more than one hint, you can specify them using a single –h option with every hint separated
by commas, such as –h "FIRE_TRIGGERS, CHECK_CONSTRAINTS, TABLOCK".

567
Microsoft SQL Server 2000 Programming by Example

You can use the format option, instead of the in, out, or queryout options, to produce a format file. By
editing the format file, you can perform complex import operations, such as selecting which columns to import
from the file, change the order of the columns to import, or specify different delimiters for every column. Later
in this chapter, you will see how to use the format file to import WAV files into SQL Server. You can search in
Books Online for the "Using Format Files" topic to get information about the different options you have when
using the format file.

Using the BULK INSERT Statement

The BULK INSERT statement imports a data file into a table either directly or through a view. This way is
similar to the bcp utility, but you use BULK INSERT from Transact-SQL, not from the command prompt.
Listing 14.9 shows a simple example to import data from the region.txt file created in Listing 14.2. To
execute this example, you can open a session in SQL Server using Query Analyzer.

Listing 14.9 the BULK INSERT Statement to Import a Data File into a Table

USE Northwind
GO

TRUNCATE TABLE NewRegions

GO

SELECT *
FROM NewRegions
GO

BULK INSERT NewRegions FROM 'C:\Temp\region.txt'

GO

SELECT *
FROM NewRegions

ID Name
----------- --------------------------------------------------

(0 row(s) affected)

(4 row(s) affected)

ID Name
----------- --------------------------------------------------
1 Eastern

568
 Chapter 14. Transferring Data to and from SQL Server

2 Western
3 Northern
4 Southern

(4 row(s) affected)
You can use the FIRSTROW and LASTROW options in the same way you used the –F and –L options in bcp.
Listing 14.10 show an example of importing rows 5 to 8 from the topten.txt file produced in Listing 14.3.

Listing 14.10 Use the FIRSTROW and LASTROW Options to Specify Which Rows to Import

USE Northwind
GO

-- Create the destination table

-- with no rows and the same structure as
-- the result set from TopTenOrders function

SELECT *
INTO TopTen
FROM dbo.TopTenOrders()
where OrderID < 1000
GO

-- Import rows 5 to 8 from the file

BULK INSERT TopTen FROM 'C:\Temp\topten.txt'

WITH
(
FIRSTROW = 5,
LASTROW = 8
)
GO

-- Test the rows imported

SELECT OrderID, CustomerID
FROM TopTen

ID Name
----------- --------------------------------------------------

(0 row(s) affected)

(4 row(s) affected)

569
Microsoft SQL Server 2000 Programming by Example

ID Name
----------- --------------------------------------------------
1 Eastern
2 Western
3 Northern
4 Southern

(4 row(s) affected)
BULK INSERT has a similar functionality as bcp for importing operations. Table 14.1 maps every option in
the bcp utility to the corresponding option in the BULK INSERT statement.
Table 14.1. Options Equivalence Between BULK INSERT and bcp
BULK INSERT bcp
FROM 'data_file' in data_file
BATCHSIZE = batch_size -b batch_size
CHECK_CONSTRAINTS -h "CHECK_CONSTRAINTS"
CODEPAGE = 'ACP' -C ACP
CODEPAGE = 'OEM' -C OEM
CODEPAGE = 'RAW' -C RAW
CODEPAGE = 'code_page' -C code_page
DATAFILETYPE = 'char' -c
DATAFILETYPE = 'native' -n
DATAFILETYPE = 'widechar' -w
DATAFILETYPE = 'widenative' -N
FIELDTERMINATOR = 'field_terminator' -t field_term
FIRSTROW = first_row -F first_row
FIRE_TRIGGERS -h "FIRE_TRIGGERS"
FORMATFILE = 'format_file' -f format_file
KEEPIDENTITY -E
KEEPNULLS -k
KILOBYTES_PER_BATCH = kb_per_batch (Not available)
(Not available) -a packet_size
BULK INSERT bcp
LASTROW = last_row -L last_row
MAXERRORS = max_errors -m max_errors
ORDER (column [ASC|DESC],... n) -h "ORDER (column [ASC|DESC],... n)"
ROWS_PER_BATCH = rows_per_batch -h "ROWS_PER_BATCH = bb"
ROWTERMINATOR = 'row_terminator' -r row_term
(Not available) out
(Not available) queryout
(Not available) format
(Not available) -e err_file
(Not available) format
(Not available) -V 60
(Not available) -V 65
(Not available) -V 70
(Not available) -6
(Not available) -q
(Not available) -o output file
(Not available) -i input file

570
 Chapter 14. Transferring Data to and from SQL Server

(Not available) -S server_name\instance

(Not available) -U login_id
(Not available) -P password
(Not available) -T
(Not available) -v
(Not available) -R
Note

For descriptions of individual options not described in this chapter, look at the "BULK INSERT"
topic in Books Online.

Caution

Only members of the sysadmin role can execute the BULK INSERT statement. SQL Server uses
the SQL Server service account to read the file. Therefore, you should make sure that the service
account has permissions to read the file.

It is not required to be a member of the sysadmin role to execute the bcp command-line utility,
but the user needs to have appropriate permissions on the source and destination tables, as well
as the files and directories used by bcp.

BULK INSERT imports data into a table, but you do not have a BULK EXPORT statement to export data from
a table to an external file. You can execute bcp from the command prompt to export data from SQL Server to
a file. Can you execute bcp from Transact-SQL?

You can use the xp_cmdshell system stored procedure to execute any OS command, and that includes
bcp. Listing 14.11 shows an example of how to export a table to an external file, using bcp with
xp_cmdshell, create a new destination table, and import the file into the new table using BULK INSERT.

Listing 14.11 Use bcp with xp_cmdshell to Export Data from Transact-SQL

USE Northwind
GO

PRINT CHAR(10)

571
Microsoft SQL Server 2000 Programming by Example

+ 'Exporting the Products Table in widenative mode'

+ CHAR(10)

EXECUTE master.dbo.xp_cmdshell 'bcp northwind.dbo.products out

c:\temp\products.txt -S
MSSQLFGG\S2K -T -N'
GO

PRINT CHAR(10)
+ 'Creating the NewProducts table '
+ 'with the same structure as '
+ CHAR(10)
+ 'the Products table but empty'
+ CHAR(10)

SELECT *
INTO NewProducts
FROM Products
WHERE ProductID = -1
GO

PRINT CHAR(10)
+ 'Checking the NewProducts table'
+ CHAR(10)

SELECT COUNT(*)
FROM NewProducts
GO

PRINT CHAR(10)
+ 'Importing the Products.txt file into the NewProducts Table'
+ CHAR(10)

BULK INSERT NewProducts FROM 'c:\temp\Products.txt'

WITH
(
DATAFILETYPE = 'widenative'
)
GO
PRINT CHAR(10)
+ 'Checking the NewProducts table'
+ CHAR(10)

SELECT COUNT(*) AS NRows

FROM NewProducts
GO
Exporting the Products Table in widenative mode

output
-----------------------------------
NULL
Starting copy...

572
 Chapter 14. Transferring Data to and from SQL Server

NULL
77 rows copied.
Network packet size (bytes): 4096
Clock Time (ms.): total 411
NULL

(7 row(s) affected)

Creating the NewProducts table with the same structure as

the Products table but empty

(0 row(s) affected)

Checking the NewProducts table

-----------
0

(1 row(s) affected)

Importing the Products.txt file into the NewProducts Table

(77 row(s) affected)

Checking the NewProducts table

NRows
-----------
77

(1 row(s) affected)
Another common problem is inserting images, or any document, from individual files into a table. This case is
more difficult than reading data from a single file because in this case, the individual files are not part of any
data file exported from a database application.
To solve this problem, you must create a format file to import every file, one by one. As an example, you can
create the WAVFiles table, as in Listing 14.12, to store WAV files, and you want to save the WAV files
included in the WINNT\MEDIA directory in this table. Using one of these files (START.WAV), you must first
know how big it is, to write a format file for it. When you look at the directory, you will find that the START.WAV
file is exactly 1,192 bytes in size. The format file to create it is included in Listing 14.12. Create a file called
wav.fmt in the WINNT\MEDIA directory with the contents of Listing 14.13.

Listing 14.12 Create the WAVFiles Table

USE Northwind
GO

CREATE TABLE WAVFiles (

ID int NOT NULL
IDENTITY(1,1)
PRIMARY KEY,

573
Microsoft SQL Server 2000 Programming by Example

FullFileName varchar(1024) NULL,

WAV image NULL)
GO

Listing 14.13 WAV.FMT File to Import the START.WAV File Using BULK INSERT

8.0
1
1 SQLIMAGE 0 1192 "" 3 wav ""
The WAV.FMT file created on Listing 14.13 contains the following sections:

• First line (8.0)— This is the version number of the bcp.exe application, corresponding to SQL Server
2000.
• Second line (1)— This is the number of fields the source file contains. In this case, the file contains a
single field: the wav field.
• Third line (1)— Field number in the file. There is only one field in this case:

SQLIMAGE Data file in the destination database. Because this is nontext BLOB information, the data type
should be SQLIMAGE.
0 Prefix length. In this case, you want to read from the beginning of the file.
1192 Length of the field. In this case, it is the length of the file: 1192 bytes.
"" Field terminator. In this case, it must be empty, because there is only one field in the file.
3 Import this information in the third field of the table.
wav Target field name.
"" Target field collation. It must be empty for an image field.
Now, you execute the BULK INSERT statement to import this file into the table, as in Listing 14.14. After
importing the file, the script updates the record with the original filename and tests the length of the
information just imported.

Listing 14.14 Import the WAV File into the WAVFile Table

USE Northwind
GO

DECLARE @ID int

BULK INSERT WAVFiles FROM 'd:\winnt\media\start.wav'

WITH (
FORMATFILE = 'd:\winnt\media\wav.fmt'

574
 Chapter 14. Transferring Data to and from SQL Server

SET @ID = IDENT_CURRENT('WAVFiles')

UPDATE WAVFiles
SET FullFileName = 'D:\WINNT\MEDIA\start.wav'
WHERE ID = @ID

SELECT ID,
DATALENGTH(wav) AS WAVELength,
FullFileName
FROM WAVFiles
WHERE ID = @ID

ID WAVELength FullFileName
----------- ----------- ------------------------------------
1 1192 D:\WINNT\MEDIA\start.wav
To automate the process, you can create the stored procedure ImportWavFiles, as defined in Listing
14.15. The ImportWavFiles stored procedure uses the CreaWavFmt stored procedure, defined in Listing
14.15 as well, to automatically create the WAV.FMT file for every WAV file in the required directory.
Tip

The CreaWavFmt stored procedure uses the DOS ECHO command to write text to a file. You can
use xp_cmdshell to execute ECHO commands and write information to a short file from Transact-
SQL, as in this example.

Listing 14.15 Stored Procedures to Import WAV Files from Any Directory into the WAVFile Table

USE Northwind
GO

---------------------------
-- CreaWavFmt
---------------------------
CREATE PROCEDURE CreaWavFmt
@dir varchar(255), -- directory ended with '\'
@length int -- file length
AS

575
Microsoft SQL Server 2000 Programming by Example

/*
** This is the required step to import
** image files with BULK INSERT
**
** We should do it manually, but we
** have xp_cmdshell for?
*/

DECLARE @cmd varchar(8000)

-- Remove wav.fmt file if exists

SET @cmd = 'del '

+ @dir + 'wav.fmt'

EXEC master.dbo.xp_cmdshell @cmd, no_output

-- Create the first line of the format file

SET @cmd = 'echo 8.0 >>'

+ @dir + 'wav.fmt'

EXEC master.dbo.xp_cmdshell @cmd, no_output

-- Write the second line to the file

SET @cmd = 'echo 1 >>'

+ @dir + 'wav.fmt'

EXEC master.dbo.xp_cmdshell @cmd, no_output

/*
** Add the third line to the file, specifying:
** 1 (the first field = entire file)
** SQLIMAGE as datatype
** 0 as field prefix length
** length of the field (file in this case)
** no field separator
** third field on the table
** Wav field
** Empty collation
*/

SET @cmd = 'echo 1 SQLIMAGE 0 '

+ CONVERT(varchar(10), @length)
+ '"" 3 wav "" >>'
+ @dir + 'wav.fmt'

EXEC master.dbo.xp_cmdshell @cmd, no_output

-- wav.fmt is created already for this file

GO

---------------------------
-- InsertWavFiles
---------------------------

CREATE PROCEDURE InsertWavFiles

@dir varchar(255)
AS

576
 Chapter 14. Transferring Data to and from SQL Server

DECLARE @sdir varchar(256)

/*
** Create temporary table to hold
** directory contents
*/

CREATE TABLE #tdir(

FileDir varchar(200) NULL,
length int NULL)

SET @sdir = 'dir '

+ @dir + '*.WAV'

INSERT #tdir (FileDir)

EXEC master.dbo.xp_cmdshell @sdir

-- Filter undesired rows

-- you can add your own conditions

DELETE #tdir
WHERE FileDir NOT LIKE '%.WAV'
OR FileDir IS NULL

-- Obtain file length and

-- filename cleansing
--
-- You could check with
-- EXEC master.dbo.xp_cmdshell 'dir c:\*.*'
-- that lengths are correct

UPDATE #tdir
SET length = CONVERT(int,
CONVERT(money,
LTRIM(RTRIM(SUBSTRING(FileDir, 20, 20))), 1)),
FileDir = LTRIM(RTRIM(SUBSTRING(FileDir, 40, 40)))

DECLARE @file varchar(256)

DECLARE @length int
DECLARE @sql varchar(8000)

DECLARE c_files CURSOR

FOR SELECT FileDir, length
FROM #tdir

OPEN c_files

FETCH NEXT FROM c_files INTO @file, @length

WHILE @@FETCH_STATUS = 0
BEGIN

-- Create bcp.fmt file to import the file

EXEC CreaWavFmt @dir, @length

-- Import the file

SET @sql ='BULK INSERT WAVFiles FROM '''

+ @dir
+ @file

577
Microsoft SQL Server 2000 Programming by Example

+ '''WITH (FORMATFILE = '''

+ @dir
+ 'wav.fmt'')'

EXECUTE (@sql)

-- Update the imported record

UPDATE WAVFiles
SET FUllFileName = @dir + @file
WHERE ID = IDENT_CURRENT('WAVFiles')

FETCH NEXT FROM c_files INTO @file, @length

END

CLOSE c_files

DEALLOCATE c_files

DROP TABLE #tdir

GO

-------------------------------------
-- Test the InsertWavFiles procedure
-------------------------------------

EXEC InsertWavFiles 'd:\winnt\media\'

SELECT ID,
DATALENGTH(wav) AS WAVELength,
FullFileName
FROM WAVFiles

ID WAVELength FullFileName
----------- ----------- --------------------------------------
1 55776 d:\winnt\media\chimes.wav
2 97016 d:\winnt\media\chord.wav
3 80856 d:\winnt\media\ding.wav
4 15906 d:\winnt\media\ir_begin.wav
5 42728 d:\winnt\media\ir_end.wav
6 75508 d:\winnt\media\ir_inter.wav
7 119384 d:\winnt\media\notify.wav
8 25434 d:\winnt\media\recycle.wav
9 10026 d:\winnt\media\ringin.wav
10 5212 d:\winnt\media\ringout.wav
11 1192 d:\winnt\media\start.wav
12 171100 d:\winnt\media\tada.wav
13 135876 d:\winnt\media\The Microsoft Sound.wav
14 95708 d:\winnt\media\Utopia Asterisk.WAV
15 4616 d:\winnt\media\Utopia Close.WAV
16 5824 d:\winnt\media\Utopia Critical Stop.WAV

578
 Chapter 14. Transferring Data to and from SQL Server

17 9946 d:\winnt\media\Utopia Default.WAV

18 24596 d:\winnt\media\Utopia Error.WAV
19 13026 d:\winnt\media\Utopia Exclamation.WAV
20 14922 d:\winnt\media\Utopia Maximize.WAV
21 3462 d:\winnt\media\Utopia Menu Command.WAV
22 2692 d:\winnt\media\Utopia Menu Popup.WAV
23 14990 d:\winnt\media\Utopia Minimize.WAV
24 10760 d:\winnt\media\Utopia Open.WAV
25 13084 d:\winnt\media\Utopia Question.WAV
26 98330 d:\winnt\media\Utopia Recycle.WAV
27 5120 d:\winnt\media\Utopia Restore Down.WAV
28 15372 d:\winnt\media\Utopia Restore Up.WAV
29 86798 d:\winnt\media\Utopia Windows Exit.WAV
30 156760 d:\winnt\media\Utopia Windows Start.WAV
31 344108 d:\winnt\media\Windows Logoff Sound.wav
32 486188 d:\winnt\media\Windows Logon Sound.wav
Tip

You can use a similar strategy to import any kind of files or documents into SQL Server.

Using Data Transformation Services

Data Transformation Services is a powerful tool introduced with SQL Server 7.0. It is a versatile tool that
enables developers to design packages that transfer and transform the data efficiently between two data
sources.
Using Data Transformation Services, you can

• Select any data source, not necessarily SQL Server, if you have an ODBC driver or OLE DB provider
to access its data.
• Select any data destination (it doesn't have to be SQL Server) if you have the ODBC driver or OLE
DB provider to connect to it.
• Define the transformation process to convert the source data into the structure and format required on
destination.
• Define complex tasks using Transact-SQL or any scripting language.
• Transfer database objects between two SQL Server databases in the same or different servers.
• Define a package with a complete sequence of DTS tasks with rich flow control, to specify the order of
execution.
• Save the DTS package in SQL Server 2000's msdb database, SQL Server 2000 Meta Data Services,
a COM structured file, or as a Visual Basic file.

It is not the purpose of this book to cover in detail this important tool. How ever, we want to show you how to
perform two common tasks, step by step:

• How to transfer database objects between two Microsoft SQL Server 2000 databases.
• How to export tables and views from Microsoft SQL Server 2000 to Microsoft Access 2000.

Note

To execute the examples from the next three sections, you must have two instances of SQL Server
2000 installed, or access to two different severs in your network with SQL Server 2000 installed.

579
Microsoft SQL Server 2000 Programming by Example

Transfer Objects Between Two SQL Server 2000 Databases

In this section, you learn how to transfer database objects from the Northwind database in a SQL Server 2000
instance to a new database in a different SQL Server 2000 instance in the same server. Note that this
example works as well between two different servers.
To perform this task, you will use the DTS Import/Export Wizard.
To start the wizard, you can run Enterprise Manager, open the Tools menu, select Wizards, Data
Transformation Services, and DTS Export Wizard.
A different way to start the wizard is by choosing All Tasks, Export Data from the context menu of the
Databases folder, as shown in Figure 14.1.

Figure 14.1. Start the DTS Export Wizard using the context menu of the Databases folder.

You see the DTS Import/Export Wizard Welcome screen. Here you can click Next.
Figure 14.2 shows the next step, which is to choose a data source. If you started the wizard from a specific
database context menu, you will find the selected server and database here. If you started the wizard from the
Databases folder, you will see the selected server and the default database for your connection.

Figure 14.2. You can select the data source from where to read the data.

580
 Chapter 14. Transferring Data to and from SQL Server

In this step, you can select any data source and specify any required settings to connect to the data source. In
this case, we accept the following default settings:

• Microsoft SQL OLE DB provider for SQL Server.

• Server SQLBE\Inst3, which is the named instance Inst3 in the SQLBE server.
• DTS uses Windows Authentication Mode to connect to the SQLBE\Inst3 server. You could select SQL
Server authentication instead, and, in that case, you must supply a valid username and password.
• Use the Northwind database as data source.

Click Next to arrive at the next step, which is to choose a destination, as you can see in Figure 14.3. In this
step, you can select the following settings:

Figure 14.3. You can select the destination where the data will be sent.

581
Microsoft SQL Server 2000 Programming by Example

• Microsoft SQL OLE DB Provider for SQL Server.

• Server SQLBE\Inst2, which is the named instance Inst2 in the SQLBE server.
• DTS will use Windows Authentication mode to connect to the SQLBE\Inst3 server. You could select
SQL Server Authentication instead and, in that case, you must supply a valid username and password.
• Use a new database as destination.

When you select New database, the DTS Import/Export Wizard will show you the Create Database form, as
shown in Figure 14.4. In this form, you can specify the name of the new database, NewNorthwind, and the
initial size of the data and log files— in this case, 2MB for each file.

Figure 14.4. You can create a new destination database, if required.

Tip

582
 Chapter 14. Transferring Data to and from SQL Server

We recommend that you create the destination database using the Transact-SQL CREATE
DATABASE statement before starting the wizard, because the CREATE DATA BASE statement
gives you greater flexibility in how and where to create the new database.

When you accept the creation of the new database, you return to the wizard and you can see the new
database selected, as in Figure 14.5.

Figure 14.5. You can select the newly created database as the destination database.

Click Next and you arrive at the Specify Table Copy or Query step, as shown in Figure 14.6. This step is
different, depending on which data source and destination you selected in the previous steps. In this case,
from SQL Server to SQL Server, you have three choices:

Figure 14.6. You can select different ways to select which data to copy.

583
Microsoft SQL Server 2000 Programming by Example

• Copy Table(s) and View(s) from the Source Database— Selecting this option, you will be presented
with a list of available tables and views to select, as in Figure 14.16, in the next section of this
chapter.

Figure 14.16. Select which tables and views to select as data source.

584
 Chapter 14. Transferring Data to and from SQL Server

• Use a Query to Specify the Data to Transfer— This is a very flexible way of defining the data source,
because you can write your own SELECT statement to retrieve the required data.
• Copy Objects and Data Between SQL Server Databases— This option is available only when you
select SQL Server as a source and destination. Using this option, you can transfer objects with or
without data.

In this case, select Copy Objects and Data Between SQL Server Databases, and click Next.
The next step is to Select Objects to Copy, as shown in Figure 14.7. In this step, you can select

Figure 14.7. You can select which database objects to copy.

• Whether or not to create destination objects— In this case, you can specify to drop the object first,
include all dependent objects, and include extended properties.

Caution

If you do not select Include All Dependent Objects, you can find errors when scripting objects that
depend on objects that will be created later in the same package. If you try to export a view, and its
base tables are not transferred, SQL Server throws an error during the transfer process.

• Transfer the data— You can uncheck this option to transfer only the schema: the definition of the
database objects. If you selected to transfer the data, you can select to overwrite the existing data or
to append the new data to the existing data.
• Specify to translate collations— When transferring data between SQL Server 2000 databases, this
setting affects only data added to an existing table. If your DTS package creates the destination object,
the columns will have the same collation as in the source database.

585
Microsoft SQL Server 2000 Programming by Example

Tip

Using UNICODE data when working with servers or databases with different code pages saves
translation problems. If you work only with SQL Server 2000, specify the collation at database or
column level, if necessary, and the columns will be transferred with their collation definition.

• Select to copy all objects or only some of them.

• Use default security and table options— Uncheck this option and you can select to transfer logins,
permissions, indexes, triggers, and constraints.

Accept the default options in the Select Objects to Copy step and click Next.
You are now in the Save, Schedule, and Replicate Package step, as you can see in Figure 14.8. In this step,
you can set several options that affect how and when the package will be executed:

Figure 14.8. You can select where to save the DTS package and when to execute it.

• Run Immediately— To execute the package right after the wizard completes.
• Use Replication to Publish Destination Data— This option starts the Create Publication Wizard after
the DTS Import/Export Wizard completes.
• Schedule DTS Package for Later Execution— This option causes a SQL Server Agent job to be
executed automatically, according to the required schedule.
• Save DTS Package— Use this option to store the package so you can modify it later. You can store
the package in the MSDB database in SQL Server, in the SQL Server Meta Data Services, in a COM
structured storage file, or as a Visual Basic File.

586
 Chapter 14. Transferring Data to and from SQL Server

Select SQL Server as the storage location for the package, and click Next. You arrive at the Save DTS
Package step, and you will see something similar to Figure 14.9. In this step, you can specify several
options:

Figure 14.9. To save the DTS package, you must specify a name and provide a description, as well as
owner and user passwords.

• Name and description of the package.

• Owner Password— This password will be required for users trying to modify the package.
• User Password— Users must specify this password to run the package.

Note

DTS packages can be stored outside SQL Server. Therefore, the authentication mode selected in
SQL Server will not necessarily protect the DTS packages from being modi fied or executed.

• Server— You can select in which server to store the package and specify the authentication mode to
connect to the server.

After setting the required options, click Next and you will arrive at the last step in the DTS Import/Export
Wizard, the Completing the DTS Import/ Export Wizard. In this stage, you can still cancel the execution of the
package by clicking Cancel.
Click Finish, and the wizard creates the package according to the settings you selected. Because you
selected to run the package immediately, the package will run right after the wizard completes, as you see in
Figure 14.10.

Figure 14.10. You can see the execution progress of the package.

587
Microsoft SQL Server 2000 Programming by Example

When the package finishes its execution, you will receive a confirmation message.
You saved this package in SQL Server, so you can open it again and modify its definition if you need to.
You can open the Data Transformation Services folder in Enterprise Manager and inside Local Packages you
could find the CopyObjects package you just created. Open the package and you will see, as in Figure
14.11, a single Copy SQL Server Objects task. You are now in the DTS Designer environment, where you
can modify the package adding more tasks, modifying the sequence of execution, and so on.

Figure 14.11. You can open an existing DTS package in the DTS package designer.

Double-click on the task and you can see its properties. Figure 14.12 shows the Data Source properties, and
it shows the same settings you selected previously in the Wizard in the Choose a Data Source step, as shown
in Figure 14.12.

588
 Chapter 14. Transferring Data to and from SQL Server

Figure 14.12. You can see the data source properties in the Source tab of the Object Transfer Task
Properties form.

Click the Copy tab, and what you will see, as shown in Figure 14.13, is very similar to the Select Objects to
Copy step shown in Figure 14.7.

Figure 14.13. Use the Copy tab in the Object Transfer Task Properties form to see which database
objects are selected to be copied.

589
Microsoft SQL Server 2000 Programming by Example

Now that the package has been created and saved, you can execute it any time you need it. To run the
package at any time, just right-click it and select Execute Package.

Export a SQL Server Table to an Access Database

In this case, you are going to copy the tables and views from the Northwind database in SQL Server to a
database in Access 2000.
The first task you must do is to create an empty Access database, named DTS.MDB, in a well-known location
in your hard disk. For this example, we selected the location D:\SQL.
You must select the DTS Import/Export Wizard, as in the previous section, and you will arrive at the welcome
screen.
Click Next and you will arrive at the Choose a Data Source step as described earlier in this chapter. Refer to
Figure 14.3, which shows the default options for this example.
Click Next to arrive at the Choose a Destination step. In this case, you must select the Microsoft Access
destination, and the form, as shown in Figure 14.14, will be different from the form displayed in Figure 14.3.

Figure 14.14. Specify the Access file to export to, as well as a valid username and password for
Access.

590
 Chapter 14. Transferring Data to and from SQL Server

As you select the Access destination, you must provide

• File Name— In this case, D:\SQL\DTS.MDB, or the full path of the file you created at the beginning of
this section, if it is different.
• Username and Password— If you have a secured Access environment, you must provide a valid
username and password. Otherwise, you can provide Admin as username and a blank password.

Click the Advanced button if you want to specify different extra settings for the Jet OLE DB Provider.
Click Next to go to the Specify Table Copy or Query step, as shown in Figure 14.15. This figure is similar to
the one in Figure 14.6, but you cannot copy objects from SQL Server to Access; that is why this option is not
available.

Figure 14.15. You can select whether to read data from tables or views, or create your own queries.

591
Microsoft SQL Server 2000 Programming by Example

Select Copy Table(s) and View(s) from the Source Database, and click Next to arrive at the Select Source
Tables and Views step, as shown in Figure 14.16. In this step, you can

• Select which tables and views to copy. For this example, click Select All to select all tables and views.
• Select the name for every destination table.

Note

Both tables and views are transferred as tables to the Access database.

• Specify how to transform the data between source and destination, clicking the "…" button under the
Transform column.

Now you have to follow the same steps as those in Figures 14.8 and 14.9 to save and schedule the DTS
package.
After this process, the package will be saved and executed, as shown in Figure 14.17. Note that this figure is
different from Figure 14.10. In this case, there are several tasks running in parallel. You will see later in this
section that this DTS package is made out of individual subpackages for every table or view to copy.

Figure 14.17. You can see how SQL Server executes different DTS tasks in parallel.

592
 Chapter 14. Transferring Data to and from SQL Server

When the package finishes its execution, you will see a confirmation message with information about how
many objects have been transferred from Microsoft SQL Server to Microsoft Access.
Now you can edit the package, using the DTS Designer. You can use Enterprise Manager to open the Local
Packages folder in the Data Transformation Services section, and you will see the ExportAccess package
(see Figure 14.18).

Figure 14.18. You can see the list of local DTS packages from Enterprise Manager.

Right -click the package and select Design Package, and the DTS Designer environment will show a similar
screen as shown in Figure 14.19.

Figure 14.19. The DTS Designer Environment shows the ExportAccess package.

593
Microsoft SQL Server 2000 Programming by Example

Figure 14.19 shows the package in design mode and you can see many little tasks there. Use the Zoom
icon in the toolbar to show the package design at 100%, and your screen will show something similar to
Figure 14.20.

Figure 14.20. Zoom 100% to see every task in the DTS Designer window.

You can see in Figure 14.20 that for every table or view, the DTS package contains three elements:

• An Execute SQL Task to create the table in Access

• A connection to SQL Server
• A connection to Access

594
 Chapter 14. Transferring Data to and from SQL Server

Note that DTS opens two connections to SQL Server— Connection 1 and Connection 3— and two connections
to Access— Connection 2 and Connection 4. In this way, the package can run tasks in parallel.
There is an arrow between the Execute SQL Task and the SQL Server Connection. If you right-click on this
arrow, you will be prompted with a context menu.
Selecting Properties in the context menu will bring you to the Workflow Properties form (see Figure 14.21).
There you can specify the source and destination steps, as well as the precedence criteria:

Figure 14.21. You can select precedence properties based on completion, failure, or success of the
preceding task.

• Select None for independent tasks, with no precedence declared between them.
• Select Completion for unconditional precedence, where the destination task will be executed after the
source task completes, regardless of the execution success or failure.
• Select Failure to execute the destination task only in the event of a source task execution failure.
• Select Success to execute the destination task only if the execution of the source tasks is successful.

If you open the destination Access database, you will see the table list, where every table and view in the
source database has been converted into an Access table (see Figure 14.22).

Figure 14.22. You can see the tables list in the Destination Access database.

595
Microsoft SQL Server 2000 Programming by Example

You can create a similar package to import data from Access into SQL Server or export from an Access
database into another Access database, if this is what you need to do.
Note

Remember that you can create a DTS package to transfer data from ANY data source to ANY
destination, as long as you have an ODBC driver or an OLEDB provider for the data source and
the destination.

The Copy Database Wizard

SQL Server includes a new tool to copy databases between two servers: the Copy Database Wizard. The
Copy Database Wizard is implemented as a custom DTS package with the following custom tasks:

• Database Move/Copy Task— Using this task, you can select source and destination servers, as well
as a database or databases to move or copy. Using this custom task, you specify the location for the
database files in the destination server.
• Logins Copy Tasks— Used to copy existing logins from the source to the destination server.
• Master Stored Procedures Copy Task— Used to copy user-defined stored procedures from the master
database in the source server to the master database in the destination server.
• Jobs Copy Task— Used to copy jobs from the msdb database, in the source server, to the msdb
database, in the destination server.
• Error Messages Copy Task— Used to copy the messages contained in the
master.dbo.sysmessages system table between the source and destination servers.

Tip

The Copy Database Wizard Custom tasks can be very useful in your own DTS packages.

To illustrate how the Copy Database Wizard works, copy a database from one instance of SQL Server to
another instance in the same server. If you have two servers in your network with SQL Server 2000 installed,
you can follow this example to copy a database from one server to another.

596
 Chapter 14. Transferring Data to and from SQL Server

Before starting with the wizard, you must create a new database. In this example, we created a new database
called TimDB.

To start the Copy Database Wizard from Enterprise Manager, you can display the context menu for the
Databases folder, as in Figure 14.23.

Figure 14.23. Start the Copy Database Wizard from the Databases context menu.

When you start the Copy Database Wizard, you get the Welcome to the Copy Database Wizard form, where
you can find a summary of the actions that this wizard will execute.
Click Next, and you will see the Select a Source Server step (see Figure 14.24). In this form, you must
select the SQL Server source server and specify which authentication mode the package will use to connect
to the source server.

Figure 14.24. Select a source server for the Copy Database Wizard.

597
Microsoft SQL Server 2000 Programming by Example

Note

It is not necessary to register a server in Enterprise Manager to be able to select the server as
source or destination server.

Click Next and the wizard will take you to the Select a Destination Server step (see Figure 14.25). Here you
will select another instance of SQL Server 2000 in the same or a different server from the source server.

Figure 14.25. Select a destination server for the Copy Database Wizard.

598
 Chapter 14. Transferring Data to and from SQL Server

Note

You cannot select the same server as both source and destination.

The next step is Select the Databases to Move or Copy (see Figure 14.26). In this form, you will see that
only databases that do not exist in the destination server are available for moving or copying. This excludes
system databases, because they exist in both servers. In this example, you must select to copy the database
you just created at the beginning of this section.

Figure 14.26. Select valid databases to move or copy.

599
Microsoft SQL Server 2000 Programming by Example

Click Next and you will arrive at the Database File Location step (see Figure 14.27). In this form, you will see
every file used by every database selected in the previous step.

Figure 14.27. Select the location of the database files.

The wizard selects as the destination the default data directory of the destination server, which is the directory
where the master database primary data file is stored. You can change the directory of these files to any valid
folder in the destination server.

600
 Chapter 14. Transferring Data to and from SQL Server

Click Modify to go to the Database Files form, where you can change the name and location of every file
individually (see Figure 14.28). If the file already exists on the destination directory, you will see a conflict
mark. Note that you cannot change the file size in this wizard. Database files retain their size after the transfer
to the destination server.

Figure 14.28. The Database Files form enables you to change database filenames and locations.

Note

Filenames are unique per database, not serverwide. You can have conflicts only if the destination's
physical file already exists.

In SQL Server 2000, every database is as self-contained as possible. How ever, your database might need
some objects that are defined in other databases. This is why the Copy Database Wizard contains the Select
Related Objects step (see Figure 14.29). In this step, you can

Figure 14.29. Select related objects, such as logins, jobs, and messages.

601
Microsoft SQL Server 2000 Programming by Example

• Select to copy logins from the source server to the destination. This is a recommended practice to
avoid orphan database users in the destination database, because of nonexisting logins in the
destination server. You can specify which logins to copy.
• If you created stored procedures in the master database to be shared by every database, you can
transfer them, too. Perhaps the source database contains stored procedures and triggers that access
the shared stored procedures from a master.
• You can have jobs defined in msdb that are referenced from your source database. You can copy
them to the destination server, also.
• Perhaps you defined custom error messages in the source server and your stored procedures,
triggers, and user-defined functions use these custom messages; then, you will need to transfer them
to the destination server.

Caution

Have clear numbering criteria for your custom messages, so you can avoid overlapped messages
from one database system to another. Before copying messages to the destination server, check
that these messages do not exist in the destination server.

The next step is Schedule the DTS Package, because the Copy Database Wizard creates a DTS package
and stores it in the destination server. Figure 14.30 shows this form, where you can provide a name to the
DTS package and specify a schedule in the usual way.

Figure 14.30. You can schedule when to execute the Copy Database Wizard DTS package.

602
 Chapter 14. Transferring Data to and from SQL Server

The Completing the Copy Database Wizard form will show a summary of the tasks to execute, and you can
click Finish to complete the wizard's work.
The package execution starts, as you see in Figure 14.31, and you can click More Info to get a full
description of the tasks to execute (see Figure 14.32).

Figure 14.31. Copy Database Wizard. Log details, without extra details.

Figure 14.32. Copy Database Wizard. Log details, with extra details.

603
Microsoft SQL Server 2000 Programming by Example

As you can see in Figure 14.33, the Copy Database Wizard executed the fol lowing tasks successfully:

Figure 14.33. The Copy Database Wizard execution completed successfully.

1. Copied selected logins to the destination server.

2. Copied selected stored procedures to the destination server.
3. Copied selected jobs from the source to the destination server.
4. Copied selected messages from the source to the destination server.
5. Checked that the source database does not have active connections.
6. Put the source database in single-user mode.
7. Detached the source database.
8. Copied the database files to the destination server.
9. Attached the database to the destination server.

What's Next?

604
 Chapter 14. Transferring Data to and from SQL Server

Transferring and transforming data is a common administrative task in multiserver environments.

Chapter 15, "Working with Heterogeneous Environments: Setting Up Linked Servers," discusses
the multiserver environment and the implications of distributed transactions. In Chapter 15, you learn how to
use linked servers to maintain data in multiple servers, as an alternative to DTS and replication.

605
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Chapter 15. Working with Heterogeneous Environments:

Setting Up Linked Servers
Chapter 14, "Transferring Data to and from SQL Server," introduced the idea of transferring data
between different databases, servers, or platforms. In this chapter, you will practice with SQL Server 2000
features that enable you to execute queries that span multiple servers.
In some cases, you might want to access data that is not stored in SQL Server, that perhaps is not even
relational. As long as you have an ODBC driver or an OLE DB provider to connect to this particular data, you
can use linked servers, or rowset functions, to access this data from any Transact-SQL batch, stored
procedure, trigger, or user-defined function.
This chapter teaches you the following:

• What distributed queries are and why you might need them
• How to use Transact-SQL rowset functions to execute queries in remote servers
• How to define linked servers
• How to execute queries using linked servers
• How to run queries that are executed remotely in a remote server
• How to design and use partitioned views

Distributed Queries
Using any programming language, you can use ADO and OLE DB to connect to any data source. SQL Server
2000 uses OLE DB to give you access to any available data source.
You can write ad hoc queries in SQL Server to access external data using a rowset function in the FROM
clause, as if it were a result set from a local table. All you need to use this functionality is an OLE DB provider
and the properties required to establish a connection to the data.
For common queries, it is more efficient to define a linked server, declaring the connection properties
permanently, so any user connected to the local server will have access to the remote server without
specifying any connection property manually.
Note

For this chapter, you should install SQL Server 2000 three times:

• Default instance: SQLBE

• First named instance: SQLBE\Inst2
• Second named instance: SQLBE\Inst3

Read Appendix A, "Using SQL Server Instances," to learn about how to install and work with
SQL Server 2000 instances.

Tip

If your server has a different name than SQLBE, you can use the SQL Server 2000 Client Network
Utility to create an alias to your first available SQL Server 2000 server or instance and call it
SQLBE. Select two more SQL Server 2000 servers and create aliases for them called SQLBE\ Inst2
and SQLBE\Inst3.

In this way, you can execute the examples of this chapter with minimal or no changes.

607
Microsoft SQL Server 2000 Programming by Example

Ad Hoc Queries

SQL Server 2000 provides, two rowset functions to access heterogeneous data from any query:

• OPENDATASOURCE—To open any relational data source that exposes the data organized in catalogs,
schemas, and data objects. SQL Server is a typical example, because it exposes data as
DatabaseName.ObjectOwner.ObjectName, where the object can be a table, view, stored
procedure, or user-defined function.
• OPENROWSET—To open any data source, relational or nonrelational, as long as you can connect to the
data source through OLE DB.

Both rowset functions provide a similar functionality. The main difference is the way you invoke them,
although the result is the same in both cases: a result set.

Using OPENDATASOURCE

Any database object in SQL Server can be identified by its fully qualified name:
ServerName.CatalogName.SchemaName.ObjectName. In SQL Server, the CatalogName is the name
of the database in which the object is stored. The SchemaName is usually the name of the owner of the object.
Note

In SQL Server 2000, although not very common, you can create schemas using the CREATE
SCHEMA statement. If you are interested in this topic, you can take a look at the topic " CREATE
SCHEMA " in Books Online.

Working with ADO, from any programming language, you need to establish a connection to a server before
trying to access its data. In Transact-SQL, use the OPENDATASOURCE function to connect to a server and
retrieve data from there (see Listing 15.1).

Listing 15.1 Use OPENDATASOURCE to Connect to a Server and Retrieve Data

SELECT ProductID, ProductName

FROM OPENDATASOURCE(
'SQLOLEDB',
'Data Source=SQLBE\Inst3;User ID=sa;Password=;').Northwind.dbo.Products
WHERE UnitPrice > 50.0

608
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

ProductID ProductName
----------- ----------------------------------------
9 Mishi Kobe Niku
18 Carnarvon Tigers
20 Sir Rodney's Marmalade
29 Thüringer Rostbratwurst
38 Côte de Blaye
51 Manjimup Dried Apples
59 Raclette Courdavault
As you saw in Listing 15.1, you use OPENDATASOURCE as a server name to fully qualify a table in a remote
server. The OPENDATASOURCE function has two parameters:

• The OLE DB provider to use. In this example, you use the SQLOLEDB OLE DB provider to access a
SQL Server 2000 server. You can use this provider to access any version of SQL Server.
• The connection string, required by the OLE DB provider, to connect to the data source.

Every OLE DB provider requires a different connection string. In this case, the connection string contains the
following data:

• Data Source=SQLBE\Inst3;—In this case, it is the name of the SQL Server 2000 server you want
to connect to, SQLBE, and the instance name Inst3. You can use the Server keyword instead of
Data Source.
• User ID=sa;—This is the SQL Server login used to connect to the remote SQL Server. In this
example, you connect to SQL Server using the sa account. You can use the UID keyword instead of
User ID.
• Password=;—In this case, you provide a blank password. You can substitute the Password keyword
with PWD.

Caution

Try to avoid using the sa account to connect to SQL Server; use integrated security instead.
However, if you must use the sa account, provide a hard-to-guess password to the sa account as
soon as possible, and restrict the number of users who know this password.

To run the examples in this chapter, you should connect to SQL Server using integrated security,
as in Listing 15.2.

The example in Listing 15.2 uses integrated security to connect to SQL Server through OPENDATASOURCE.
As you can see in that example, the only difference is the inclusion of the Integrated Security=SSPI;
string, or Trusted_Connection=yes;, instead of User ID=sa;Password=;.

609
Microsoft SQL Server 2000 Programming by Example

Listing 15.2 You Can Use Integrated Security When Using OPENDATASOURCE to Connect to SQL Server

SELECT CustomerID, CompanyName

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE;Integrated Security=SSPI;').Northwind.dbo.Customers
WHERE City = 'London'

CustomerID CompanyName
---------- ----------------------------------------
AROUT Around the Horn
BSBEV B's Beverages
CONSH Consolidated Holdings
EASTC Eastern Connection
NORTS North/South
SEVES Seven Seas Imports
In the examples from Listings 15.1 and 15.2, you might think that SQL Server connects to the remote
servers, retrieves the data from the specified table, and locally applies the filter declared in the WHERE clause.
However, the Query Processor is intelligent enough to detect that the WHERE clause applies exclusively to
remote data and send the query to the remote server to be filtered remotely. In this way, the overall
performance is improved, because the data is filtered where it belongs. Figure 15.1 shows the query plan of
the query in Listing 15.2. In this query plan, you can see only one step, a remote query, defined as

Figure 15.1. The Query Processor sends the WHERE clause to the remote server to be processed
remotely.

SELECT Tbl1001."CustomerID" Col1003

,Tbl1001."CompanyName" Col1004
FROM "Northwind"."dbo"."Customers" Tbl1001
WHERE Tbl1001."City"=N'London'

You can use the OPENDATASOURCE function to join it to a local table (see Listing 15.3). In this case, the
query plan is shown in Figure 15.2, where you can still see the remote query, with a merge join to connect
the remote result set with the local result set.

610
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Figure 15.2. You can join a Remote Query to a local table.

Listing 15.3 You Can Join the Result of OPENDATASOURCE to a Local Table

SELECT OrderID, OrderDate, O.CustomerID, CompanyName

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE;Integrated Security=SSPI;').Northwind.dbo.Customers
AS C
JOIN Northwind.dbo.Orders
AS O
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND OrderDate BETWEEN '1996-12-01'
AND '1996-12-15'
OrderID OrderDate CustomerID CompanyName
-------- ------------------------ ---------- --------------------

10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports

If you modify the example from Listing 15.3 to retrieve the Customer information from the SQLBE server and
the Orders information from the SQLBE\Inst2 server, the query should be as in Listing 15.4, and the query
plan is shown in Figure 15.3.

Figure 15.3. You can join a remote query to another remote query.

611
Microsoft SQL Server 2000 Programming by Example

Listing 15.4 Use OPENDATASOURCE to Connect to a Server and Retrieve Data

SELECT OrderID, OrderDate, O.CustomerID, CompanyName

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE;Integrated Security=SSPI;').Northwind.dbo.Customers
AS C
JOIN OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE\Inst2;Integrated Security=SSPI;').Northwind.dbo.Orders
AS O
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND OrderDate BETWEEN '1996-12-01'
AND '1996-12-15'

OrderID OrderDate CustomerID CompanyName

-------- ------------------------ ---------- --------------------
10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports
Caution

You can use OPENDATASOURCE only to retrieve data from tables and views. Stored procedures
and user-defined functions are not allowed in OPENDATASOURCE

You can use OPENDATASOURCE to retrieve data from an Access database, as the DTS.MDB database
defined in Chapter 14, in the "Export a SQL Server Table to an Access Database" section. Listing

612
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

15.5 shows how to use the Microsoft Jet OLE DB provider to connect to an Access database, and Figure
15.4 shows its query plan.

Figure 15.4. You can join a remote query to SQL Server to a remote query to Access.

Listing 15.5 Use OPENDATASOURCE to Connect to a Remote SQL Server and an Access Database

SELECT OrderID, OrderDate--, O.CustomerID, CompanyName

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE;Integrated Security=SSPI;').Northwind.dbo.Customers
AS C
JOIN OPENDATASOURCE(
'Microsoft.Jet.OLEDB.4.0',
'Data Source="D:\SQL\DTS.MDB";User ID=Admin;Password=;')...Orders
AS O
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND OrderDate BETWEEN '1996-12-01'
AND '1996-12-15'

OrderID OrderDate CustomerID CompanyName

-------- ------------------------ ---------- --------------------
10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports
If you compare the query plan from Figure 15.4 to the one from Figure 15.3, you can see two extra steps
after retrieving the Orders data from Access:

• A filter to select the data range.

• A sort operation to be able to execute the merge join.

These extra steps are required because SQL Server cannot ask the Jet OLE DB provider to execute these
tasks remotely.

613
Microsoft SQL Server 2000 Programming by Example

You can use OPENDATASOURCE to modify a table in an UPDATE, DELETE, or INSERT statement, as in Listing
15.6.

Listing 15.6 Use OPENDATASOURCE to Specify a Source for UPDATE, INSERT, or DELETE Statements

PRINT 'Before moving BSBEV from London'+ CHAR(10)

SELECT CustomerID, CompanyName, City

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE\Inst3;Integrated Security=SSPI;').Northwind.dbo.Customers
WHERE CustomerID = 'BSBEV'

PRINT CHAR(10) + 'Moving BSBEV from London to Southampton'+ CHAR(10)

UPDATE OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE\Inst3;Integrated Security=SSPI;').Northwind.dbo.Customers
SET City = 'Southampton'
WHERE CustomerID = 'BSBEV'

PRINT CHAR(10) + 'After BSBEV moved from London to Southampton'+ CHAR(10)

SELECT CustomerID, CompanyName, City

FROM OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE\Inst3;Integrated Security=SSPI;').Northwind.dbo.Customers
WHERE CustomerID = 'BSBEV'

PRINT CHAR(10) + 'Moving BSBEV back to London'+ CHAR(10)

UPDATE OPENDATASOURCE(
'SQLOLEDB',
'Server=SQLBE\Inst3;Integrated Security=SSPI;').Northwind.dbo.Customers
SET City = 'London'
WHERE CustomerID = 'BSBEV'
Before moving BSBEV from London

CustomerID CompanyName City

---------- ---------------------------------------- ---------------
BSBEV B's Beverages London

(1 row(s) affected)

614
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Moving BSBEV from London to Southampton

(1 row(s) affected)

After BSBEV moved from London to Southampton

CustomerID CompanyName City

---------- ---------------------------------------- ---------------
BSBEV B's Beverages Southampton

(1 row(s) affected)

Moving BSBEV back to London

(1 row(s) affected)

Using OPENROWSET

You can use OPENROWSET to retrieve result sets from any data source, in a way similar to OPENDATASOURCE.
The main difference is that you can send the query to the data source, using the syntax that the OLE DB
provider accepts, and the OLE DB provider will return the requested result set.
Listing 15.7 shows how to use OPENROWSET to retrieve data from an instance of SQL Server, in a way
similar to that seen earlier in Listing 15.1. How ever, you can send the entire query, including the WHERE
clause, directly to the OPENROWSET function, as shown in the second example of Listing 15.7.

Listing 15.7 Use OPENROWSET to Connect to a Server and Retrieve Data

SELECT ProductID, ProductName

FROM OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst3;UID=sa;PWD=;', Northwind.dbo.Products)
WHERE UnitPrice > 50.0

SELECT *
FROM OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst3;UID=sa;PWD=;',
'SELECT ProductID, ProductName
FROM Northwind.dbo.Products
WHERE UnitPrice > 50.0')

615
Microsoft SQL Server 2000 Programming by Example

ProductID ProductName
----------- ----------------------------------------
9 Mishi Kobe Niku
18 Carnarvon Tigers
20 Sir Rodney's Marmalade
29 Thüringer Rostbratwurst
38 Côte de Blaye
51 Manjimup Dried Apples
59 Raclette Courdavault
The query plans of the two queries shown in Listing 15.7 are not the same. In Figure 15.5, you can see a
remote query step for the first query and a remote scan for the second query; however, the results are the
same, and in both cases the query is sent to the remote server to be executed there.

Figure 15.5. OPENROWSET produces either a remote query or a remote scan when reading remote data.

You can join the result set returned by OPENROWSET to other result sets. Listing 15.8 shows an example
similar to Listing 15.4, but in this case you retrieve the same result in three different ways:

• The first query joins two OPENROWSET functions against two servers (SQLBE\Inst2 and SQLBE\Inst3).
• The second query joins two OPENROWSET functions against the same server (SQLBE\Inst2).
• The third query uses OPENROWSET just once to retrieve the entire result set from a single server
(SQLBE\Inst2).

Listing 15.8 Use OPENROWSET to Retrieve a Result Set to JOIN to Other Result Sets

PRINT 'Using OPENROWSET twice againt two servers'+ CHAR(10)

SELECT OrderID, OrderDate, O.CustomerID, CompanyName

FROM OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst2;Trusted_Connection=yes;',
Northwind.dbo.Customers)
AS C
JOIN OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst3;Trusted_Connection=yes;',
Northwind.dbo.Orders)
AS O
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND OrderDate BETWEEN '1996-12-01'
AND '1996-12-15'

PRINT 'Using OPENROWSET twice againt one server'+ CHAR(10)

616
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

SELECT OrderID, OrderDate, O.CustomerID, CompanyName

FROM OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst2;Trusted_Connection=yes;',
Northwind.dbo.Customers)
AS C
JOIN OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst2;Trusted_Connection=yes;',
Northwind.dbo.Orders)
AS O
ON O.CustomerID = C.CustomerID
WHERE City = 'London'
AND OrderDate BETWEEN '1996-12-01'
AND '1996-12-15'

PRINT 'Using OPENROWSET once'+ CHAR(10)

SELECT *
FROM OPENROWSET(
'SQLOLEDB',
'Server=SQLBE\Inst2;Trusted_Connection=yes;',
'SELECT OrderID, OrderDate, O.CustomerID, CompanyName
FROM Northwind.dbo.Customers AS C
JOIN
Northwind.dbo.Orders AS O
ON O.CustomerID = C.CustomerID
WHERE City = ''London''
AND OrderDate BETWEEN ''1996-12-01''
AND ''1996-12-15''')
Using OPENROWSET twice againt two servers

OrderID OrderDate CustomerID CompanyName

-------- ------------------------ ---------- --------------------
10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports

Using OPENROWSET twice againt one server

OrderID OrderDate CustomerID CompanyName

-------- ------------------------ ---------- --------------------
10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports

Using OPENROWSET once

OrderID OrderDate CustomerID CompanyName

-------- ------------------------ ---------- --------------------
10377 1996-12-09 00:00:00.000 SEVES Seven Seas Imports
Figure 15.6 shows the query plan produced to execute the first query from Listing 15.8. You can see that
this query plan is similar to the query plan shown in Figure 15.3.

Figure 15.6. You can join a remote query to another remote query using OPENROWSET.

617
Microsoft SQL Server 2000 Programming by Example

Figure 15.7 shows the query plan produced when the second query from Listing 15.8 is executed.
Because both OPENROWSET functions connect to the same server, only one remote query step is required to
retrieve the final result set. This is more efficient than returning two complete result sets and joining them
locally. The Query Processor makes this decision automatically, and through this way you can save network
bandwidth because only the requested data is transferred from the remote server.

Figure 15.7. The Query Processor optimizes access to remote servers when you connect twice, using
OPENROWSET.

Finally, you can use OPENROWSET to send a complex query to a remote SQL Server, as seen in the third
example in Listing 15.8. Figure 15.8 shows the query plan produced to execute this query. This query
contains only one OPENROWSET function, whereas the query plan shows a single remote scan step.

Figure 15.8. You can use OPENROWSET to send complex queries to a remote server.

In the examples in this section, you used the same kind of connection string as in the OPENDATASOURCE
function. You can use a different syntax in OPENROWSET to specify the server name, the user ID, and the
password, as illustrated in Listing 15.9.

Listing 15.9 You Can Use a Simplified Syntax in OPENROWSET to Connect to a Data Source

PRINT 'From SQL Server'+ CHAR(10)

SELECT CategoryID, CategoryName

FROM OPENROWSET(
'SQLOLEDB',
'SQLBE\Inst3';'sa';'', Northwind.dbo.Categories)

618
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

PRINT 'From Access'+ CHAR(10)

SELECT CategoryID, CategoryName

FROM OPENROWSET(
'Microsoft.Jet.OLEDB.4.0',
'D:\SQL\DTS.MDB';'Admin';'',
'SELECT * FROM Categories
ORDER BY CategoryName DESC')
From SQL Server

CategoryID CategoryName
----------- ---------------
1 Beverages
2 Condiments
3 Confections
4 Dairy Products
5 Grains/Cereals
6 Meat/Poultry
7 Produce
8 Seafood

From Access

CategoryID CategoryName
----------- ---------------
8 Seafood
7 Produce
6 Meat/Poultry
5 Grains/Cereals
4 Dairy Products
3 Confections
2 Condiments
1 Beverages

Tip

Avoid the syntax from Listing 15.9, because with the syntax of Listings 15.7 and 15.8 you can
reuse the same connection string used in any application that uses OLE DB to connect to the data
source.

You can use OPENROWSET, with the OLE DB provider for ODBC, to connect to any Data Source Name (DSN)
defined in the ODBC Manager. Listing 15.10 uses the LocalServer system DSN, which points to the default
SQL Server instance. You can use this provider to establish DSN-less connections to any data source also,
specifying the ODBC driver to use as seen in the second example in Listing 15.10 .

Caution

619
Microsoft SQL Server 2000 Programming by Example

Test the LocalServer DSN with the ODBC Manager before trying the example from Listing 15.10.

Listing 15.10 You Can Use the OLE DB Provider for ODBC with OPENROWSET to Access Any ODBC
Data Source Name

SELECT *
FROM OPENROWSET('MSDASQL',
'DSN=LocalServer',
'SELECT * FROM Northwind.dbo.Region')

RegionID RegionDescription
----------- --------------------------------------------------
1 Eastern
2 Western
3 Northern
4 Southern

Tip

You can use OPENROWSET to invoke system functions or administrative stored procedures from the
local server as result sets to be filtered in a SELECT statement, as in Listing 15.11, where you
use OPENROWSET to execute sp_who remotely.

Listing 15.11 Use OPENROWSET to Work with Transact-SQL Administrative Statements As Result Sets

620
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

SELECT spid, cmd, dbname

FROM OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
'EXEC sp_who')
WHERE dbname = 'Northwind'

spid cmd dbname

------ ---------------- -----------------
62 SELECT Northwind
63 AWAITING COMMAND Northwind
65 AWAITING COMMAND Northwind
You can retrieve results from remote stored procedures or user-defined functions by using OPENROWSET, as
illustrated in Listing 15.12.

Listing 15.12 Use OPENROWSET to Open Remote User-Defined Functions

SELECT OrderID, CustomerID, TotalValue, ShipCountry, OrderDate

FROM OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
'Select * FROM Northwind.dbo.TopTenOrders()')

OrderID CustomerID TotalValue ShipCountry OrderDate

--------- ---------- --------------- --------------- ------------------------
10865 QUICK 16387.5000 Germany 1998-02-02 00:00:00.000
10981 HANAR 15810.0000 Brazil 1998-03-27 00:00:00.000
11030 SAVEA 12615.0500 USA 1998-04-17 00:00:00.000
10889 RATTC 11380.0000 USA 1998-02-16 00:00:00.000
10417 SIMOB 11188.4000 Denmark 1997-01-16 00:00:00.000
10817 KOENE 10952.8450 Germany 1998-01-06 00:00:00.000
10897 HUNGO 10835.2400 Ireland 1998-02-19 00:00:00.000
10479 RATTC 10495.6000 USA 1997-03-19 00:00:00.000
10540 QUICK 10191.7000 Germany 1997-05-19 00:00:00.000
10691 QUICK 10164.8000 Germany 1997-10-03 00:00:00.000

621
Microsoft SQL Server 2000 Programming by Example

If you want to update data through the OPENROWSET function, you must use the OPENROWSET function in the
UPDATE clause as if it were a table. Listing 15.13 shows an example of UPDATE, INSERT, and DELETE
statements using OPENROWSET .

Listing 15.13 You Can UPDATE, INSERT, and DELETE a Result Set Returned by OPENROWSET

DECLARE @ID int

PRINT 'Insert a new Category'+ CHAR(10)

INSERT OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
Northwind.dbo.Categories)
(CategoryName)
VALUES ('New Category')

PRINT 'Retrieve the CategoryID'+ CHAR(10)

SELECT @ID = CategoryID

FROM OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
'SELECT CategoryID
FROM Northwind.dbo.Categories
WHERE CategoryName = ''New Category''')

PRINT 'Update the name of the new Category'+ CHAR(10)

UPDATE OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
Northwind.dbo.Categories)
SET CategoryName = 'Other'
WHERE CategoryID = @ID

PRINT 'Delete the new Category'+ CHAR(10)

DELETE OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
Northwind.dbo.Categories)
WHERE CategoryID = @ID
Insert a new Category

(1 row(s) affected)

Retrieve the CategoryID

Update the name of the new Category

(1 row(s) affected)

Delete the new Category

(1 row(s) affected)

622
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Note

You cannot use the SCOPE_IDENTITY function in the example from Listing 15.13 because the
insertion takes place in a remote server.

You can use OPENROWSET to create a more accurate version of the scalar user-defined function
dbo.Today(), created in Listing 10.4, because using OPENROWSET you can actually reconnect to SQL
Server. Almost any instruction is available in OPENROWSET, including not valid built-in functions in user-
defined functions, such as GetDate. Listing 15.14 shows how to implement the Today() function using
OPENROWSET.

Listing 15.14 You Can Use OPENROWSET Inside User-Defined Functions

USE Northwind
GO

IF OBJECT_ID('Today', 'FN') IS NOT NULL

DROP FUNCTION dbo.Today
GO
-- Returns the actual system date
-- obtained by using OPENROWSET as a callback function

CREATE FUNCTION dbo.Today

()
RETURNS smalldatetime
AS
BEGIN
DECLARE @sdt smalldatetime

SELECT @SDT = CONVERT(varchar(10), TodaysDate, 120)

FROM OPENROWSET('SQLOLEDB',
'Server=SQLBE;Trusted_Connection=yes;',
'SELECT Getdate() AS TodaysDate')
RETURN @SDT

END
GO

SELECT dbo.Today()
AS Today

623
Microsoft SQL Server 2000 Programming by Example

Today
------------------------------------------------------
2001-01-02 00:00:00

Linked Servers

Any client application can establish connections to more than one server at a time, but it is not possible to join
directly result sets from different connections.
Using the rowset functions from the last section, you can execute queries that relate information coming from
different data sources. However, SQL Server must establish a connection on every call to OPENDATASOURCE
or OPENROWSET, using the connection string or connection parameters sent along with the function call.
If you are a Microsoft Access user, you will be familiar with the concept of a linked table. This is a permanent
definition of a logical connection to an external data source.
SQL Server 2000 implements links to any OLE DB data source, as linked servers, to any SQL Server instance.
Any user connected to an instance of SQL Server can access any linked server defined in that instance
without knowing the parameters to connect to this particular data source. In this way, you have the flexibility of
the OPENROWSET and OPENDATASOURCE functions without exposing to the users the complexity inherent to
any OLE DB connection.
Caution

Having a SQL Server registered in Enterprise Manager does not mean that you have declared that
server as a linked server. This is only a setting in a client application, Enterprise Manager, stored in
a specific client computer, perhaps the server itself, and it does not have to be visible to any other
client connecting to the server.

Users can access objects on linked servers, using fully qualified four-part names and using any data access
statement. In this way, you can use any kind of information exposed by the OLE DB provider as if it were a
table on a database, and join that information to other tables in the local server.

In the following sections, you will learn how to set up and use linked servers.

Setting Up and Querying Linked Servers

The fi rst thing you need to set up a linked server, which connects to an external data source, is an appropriate
OLE DB provider. Microsoft has tested the following OLE DB providers to use in a linked server:

• Microsoft OLE DB Provider for SQL Server (SQLOLEDB)— Use this provider to connect to SQL
Server 6.5, 7.0, and 2000.
• Microsoft OLE DB Provider for ODBC (MSDASQL)— Use this provider to connect to any data source,
as long as you have a valid ODBC driver for this particular data source.

624
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

• Microsoft OLE DB Provider for Jet (Microsoft.Jet.OLEDB.4.0)— This provider connects you to
Microsoft Access databases, Microsoft Excel spreadsheets, and text files.
• Microsoft OLE DB Provider for DTS Packages (DTSPackageDSO)— This provider gives you access
to the result set of a transformation step from a DTS package.
• Microsoft OLE DB Provider for Oracle (MSDAORA).
• Microsoft OLE DB Provider for Microsoft Directory Services (ADSDSOObject)— Use this provider to
get information from the Active Directory information on Microsoft Windows 2000 or Microsoft
Exchange 2000.
• Microsoft OLE DB Provider for Microsoft Indexing Service (MSIDXS)— This provider gives you access
to local files indexed by the Microsoft Indexing Service.
• Microsoft OLE DB Provider for DB2 (DB2OLEDB)— This provider is part of the Microsoft Host
Integration Server, and gives you connectivity to IBM DB2 databases.

To set up a linked server, you can use the sp_addlinkedserver system stored procedure. Listing 15.15
shows how to create a linked server in the SQLBE server to connect to the SQLBE\Inst2 instance of SQL
Server .

Listing 15.15 Setting Up a Linked Server Using the sp_addlinkedserver System Stored Procedure

-- Use sp_addlinkedserver with

-- SQL Server as a product name

EXEC sp_addlinkedserver
@server = N'SQLBE\Inst3',
@srvproduct = N'SQL Server'
GO

-- Use sp_addlinkedserver with

-- SQLOLEDB as a provider name

EXEC sp_addlinkedserver
@server = N'SQLBEInst2',
@srvproduct = N'',
@provider = N'SQLOLEDB',
@datasrc = N'SQLBE\Inst2'
GO

-- Use sp_addlinkedserver with

-- SQLOLEDB as a provider name
-- and with an initial catalog

EXEC sp_addlinkedserver
@server = N'NewSQLBEInst2',
@srvproduct = N'',
@provider = N'SQLOLEDB',
@datasrc = N'SQLBE\Inst2',
@catalog = N'Northwind'
GO
To execute the sp_addlinkedserver system stored procedure to create a linked server to a SQL Server
instance, you must supply

625
Microsoft SQL Server 2000 Programming by Example

• The actual name of the SQL Server default instance or named instance (@server)
• N'SQL Server' as product name (@srvproduct)

or

• The logical name you want to provide to the linked server (@server).
• N'' as product name (@srvproduct).
• The name of the OLE DB provider used to connect to the data source— in this case, N'SQLOLEDB'
(@provider).
• The actual name of the SQL Server default instance or named instance to connect (@datasrc).
• Optionally, you can specify the catalog or database to which to connect (@catalog). However, this
parameter is used only to specify an initial database to connect. After the connection is made, you can
access any database on that server, providing you have permissions to use it and the @catalog
parameter is disregarded.

To query a linked server, you must use a fully qualified name, using four-part names, as seen in Listing
15.16. In this way, tables from linked servers can be used as any local table on any DML operation, such as
SELECT, INSERT, UPDATE, or DELETE. The last example on Listing 15.16 shows how to link remote
tables to other remote and local tables, as if all these tables were local tables .

Listing 15.16 You Can Use LinkedServers to Access Remote Tables Using Fully Qualified Names

PRINT 'Selecting data from a linked server'

+ CHAR(10)

SELECT CategoryID, CategoryName

FROM [SQLBE\Inst3].northwind.dbo.categories
WHERE CategoryID BETWEEN 1 AND 3

PRINT 'Inserting a row into a linked server'

+ CHAR(10)

INSERT SQLBEInst2.Northwind.dbo.Categories
(CategoryName)
VALUES('More products')

PRINT 'Updating a row from a linked server'

+ CHAR(10)

UPDATE SQLBEInst2.Northwind.dbo.Categories
SET CategoryName = 'Extra Products'
WHERE CategoryName = 'More products'

PRINT 'Deleting a row from a linked server'

+ CHAR(10)

DELETE NewSQLBEInst2.Northwind.dbo.Categories
WHERE CategoryName = 'Extra Products'

626
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

PRINT 'Join data coming from linked servers'

+ CHAR(10)

SELECT OrderDate, Quantity, OD.UnitPrice,

CategoryName, ProductName
FROM [SQLBE\Inst3].northwind.dbo.categories C
JOIN SQLBEInst2.Northwind.dbo.Products P
ON P.CategoryID = C.CategoryID
JOIN NewSQLBEInst2.Northwind.dbo.[Order Details] OD
ON OD.ProductID = P.ProductID
JOIN Northwind.dbo.Orders O
ON O.OrderID = OD.OrderID
WHERE P.CategoryID = 1
AND Year(OrderDate ) = 1998
AND CustomerID = 'BSBEV'
Selecting data from a linked server

CategoryID CategoryName
----------- ---------------
1 Beverages
2 Condiments
3 Confections

(3 row(s) affected)

Inserting a row into a linked server

(1 row(s) affected)

Updating a row from a linked server

(1 row(s) affected)

Deleting a row from a linked server

(1 row(s) affected)

Join data coming from linked servers

OrderDate Quantity UnitPrice CategoryName ProductName

----------------------- -------- ----------- --------------- -----------
1998-04-14 00:00:00.000 30 46.0000 Beverages Ipoh Coffee

(1 row(s) affected)

Caution

You cannot omit any of the four parts of the fully qualified name when referencing a remote table
from a linked server.

627
Microsoft SQL Server 2000 Programming by Example

If you want to execute a stored procedure from a linked server, you must first enable RPC (Remote Procedure
Calls) on the linked server. Listing 15.17 shows an example of how to enable RPC in a linked server by
using the sp_serveroption system stored procedure, and how to call a stored procedure remotely.

Listing 15.17 You Can Execute Remote Stored Procedures in a Linked Server

-- Set RPC OUT true

-- to accept remote procedure calls

EXECUTE sp_serveroption N'SQLBEinst2', 'RPC OUT', 'true'

-- Execute a remote procedure

EXECUTE sqlbeinst2.northwind.dbo.CustOrderHist 'BSBEV'

ProductName Total
---------------------------------------- -----------
Aniseed Syrup 30
Boston Crab Meat 10
Geitost 15
Gnocchi di nonna Alice 20
Gustaf's Knäckebröd 21
Ipoh Coffee 30
Konbu 23
Manjimup Dried Apples 3
Maxilaku 6
Mozzarella di Giovanni 1
Outback Lager 7
Raclette Courdavault 4
Ravioli Angelo 6
Sir Rodney's Scones 29
Spegesild 15
Steeleye Stout 20
Tarte au sucre 10
Uncle Bob's Organic Dried Pears 34
Wimmers gute Semmelknödel 9

(19 row(s) affected)

628
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

You can define a linked server to connect to an Access database. Listing 15.18 shows how to create a
linked server to connect to the DTS.MDB Access database created in Chapter 14. In this example, you must
write in @datasrc the location of the MDB file .
Note

Linking an Access database, the value of @srvproduct is only informative. And the location of the
database must be sent using the @datasrc parameter, not the @location parameter.

Listing 15.18 Setting Up a Linked Server to an Access Database

EXEC sp_addlinkedserver
@server = N'DTSMDB',
@srvproduct = N'Access 2000',
@provider = N'Microsoft.Jet.OLEDB.4.0',
@datasrc = N'd:\sql\dts.mdb'
GO

-- Map every login in SQL Server

-- to the Admin login in Access

EXEC sp_addlinkedsrvlogin
@rmtsrvname = 'DTSMDB',
@useself= false,
@locallogin = NULL,
@rmtuser = 'Admin',
@rmtpassword = NULL
GO

-- Read data from Access

-- through the linked server

SELECT ProductName, UnitPrice

FROM DTSMDB...Products
WHERE UnitPrice > 100
GO

ProductName UnitPrice
---------------------------------------- ---------------------

629
Microsoft SQL Server 2000 Programming by Example

Thüringer Rostbratwurst 123.7900

Côte de Blaye 263.5000

As you saw in Listing 15.18, it is not enough to create the linked server to access its data. In some cases, it
is necessary to map the local logins to remote logins to be able to connect to the linked server. To map logins,
use the sp_addlinkedsrvlogin system stored procedure. This procedure accepts the following
parameters:

• @rmtsrvname—The name of the linked server. In this case, it is 'DTSMDB'.

• @useself— True to map every local account to the same account in the linked server, so the
@locallogin, @rmtuser, and @rmtpassword parameters will be ignored. In this case, you don't
want to automatically map every local user to a remote user, because your Access database is not
secured in this case, so you give a value of @rmtuser = false.
• @locallogin—Name of the local login to map, only if @useself = false. You specify NULL in
this case because you want to map all local logins to the same remote login.
• @rmtuser—Name of the remote login to map the @locallogin. If you use an unsecured Access
database, @rmtuser = 'Admin'.
• @rmtpassword—Password to use in the remote server for the remote user specified in @rmtuser.
In this case, it must be a blank password, @rmtpassword = NULL.

You can create a linked server to read text files from a directory. To test it, you can create a little text file, like
the one in Listing 15.19, in the D:\SQL directory.

Listing 15.19 Ages.txt File

ID Age Name
1 55 Joseph
2 32 John
3 34 Frederick
4 70 Antony
5 65 Francis
6 75 Jon
7 43 Manfred
8 21 Dick
9 18 Louis
Now, you can create a linked server to read files in any directory, as in the example in Listing 15.20, using
the OLE DB provider for Jet.

Listing 15.20 Setting Up a Linked Server to a Disk Directory

-- Create a Linked Server

630
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

-- To read text files from

-- the D:\SQL directory

EXEC sp_addlinkedserver
@server = N'TextFiles',
@srvproduct = N'Text files',
@provider = N'Microsoft.Jet.OLEDB.4.0',
@datasrc = N'D:\SQL',
@provstr='Text'

GO

-- Map every login in SQL Server

-- to the Admin login in Jet

EXEC sp_addlinkedsrvlogin
@rmtsrvname = 'TextFiles',
@useself= false,
@locallogin = NULL,
@rmtuser = 'Admin',
@rmtpassword = NULL
GO

-- Read data from the Ages.txt file

SELECT *
FROM TextFiles...ages#txt
GO

ID_Age_Name
-----------------------------------
1 55 Joseph
2 32 John
3 34 Frederick
4 70 Antony
5 65 Francis
6 75 Jon
7 43 Manfred
8 21 Dick
9 18 Louis

Note

Note that, as in Listing 15.20, you must convert the character "." ("ages.txt") from the
filename into the character '#' ("ages#txt"), because the character "." is not valid inside a
table name in SQL Server.

631
Microsoft SQL Server 2000 Programming by Example

Pass-Through Queries

When working with linked servers, SQL Server 2000 always tries to send the queries to the linked servers to
be processed remotely. This decreases the network traffic. In particular, the query execution is more efficient
because it is performed in the same server in which the affected data is stored. In this case, the query is
"passed through" the linked server for remote execution.
You can force the execution of pass-through queries remotely by using the OPENQUERY function. OPENQUERY
is similar to the OPENDATASOURCE and OPENROWSET functions, because it connects to a remote data source
and returns a result set. However, OPENQUERY uses a linked server definition to connect to the remote server.
In this way, you have a persistent definition of the connection properties, providing easier maintenance of your
database application.
As you can see in the examples from Listing 15.21, the syntax of OPENQUERY is very simple: You provide
the linked server name to send the query and the query to be executed.

Listing 15.21 Using OPENQUERY to Send Pass-Through Queries to a Linked Server

-- Gets the date and time in the linked server

SELECT *
FROM OPENQUERY(SQLBEinst2,
'SELECT Getdate() AS Now')

-- Reads some data from the linked server

SELECT DISTINCT ProductName, UnitPrice

FROM OPENQUERY(SQLBEinst2,
'SELECT DISTINCT P.ProductID, ProductName,
OD.UnitPrice
FROM Northwind.dbo.Products P
JOIN Northwind.dbo.[Order details] OD
ON OD.ProductID = P.ProductID
WHERE OD.UnitPrice > 100')

-- Updating data through OPENQUERY

UPDATE OPENQUERY(SQLBEinst2,
'SELECT *
FROM Northwind.dbo.Categories')
SET CategoryName = 'Obsolete'
WHERE CategoryID = 3

-- Testing changes

SELECT categoryname
FROM SQLBEInst2.Northwind.dbo.categories
WHERE CategoryID = 3
GO

632
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Now
------------------------------------------------------
2001-02-20 17:33:16.370

ProductID ProductName UnitPrice

----------- ---------------------------------------- ---------------------
29 Thüringer Rostbratwurst 123.7900
38 Côte de Blaye 210.8000
38 Côte de Blaye 263.5000

categoryname
---------------
Obsolete
The first query in Listing 15.21 retrieves the system data and time from the linked server.
The second query remotely executes a query that joins two tables. When the combined result set is returned,
the local server performs the DISTINCT operation.
The third query updates data remotely using the OPENQUERY function.
Caution

OPENROWSET, OPENDATASOURCE, and OPENQUERY accept only string constants as values for
their parameters. String variables are not accepted.

Partitioned Views

Consider you have a very big table, SalesInfo, with your worldwide sales information. You have different
regions and you want to be able to execute queries to any subset of the complete sales table, regardless of
the region.
This table is too big and the maintenance is starting to be difficult. You decide to divide the table among four
servers, North, West, South, and East, each one storing data from only one region.
To ensure that you store on every server only data related to that specific server, create a check constraint
that enforces the value for the particular regions this server manages.
Now you want to access any data from anywhere, so, on every server you create a view that combines the
data from every server with the data from the other servers by a UNION ALL. Use UNION ALL because you
do not want to remove duplicates in the final result set. This view is called a partitioned view.
You can test a simple version of this technique using the example from Listing 15.22. This script can be run
in a single server and single instance, and still it uses the partitioned view technique. This is the only
simplification used in this example. You can change this script to create every table in a different instance or
server and modify the view to retrieve every table from the appropriate server, as shown in Listing 15.22.

Listing 15.22 Create a Partitioned View Based on Four Tables

633
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

-- Create the partitioned table

-- RegionID is the partitioning column
-- it is part of the PRIMARY KEY
-- and it has a check constraint
-- to delimit ranges per table

-- RegionID = 3 North

CREATE TABLE SalesInfoNorth (

OrderID int NOT NULL,
RegionID int NOT NULL
CHECK (RegionID = 3),
SaleDate datetime,
Amount money,
EmployeeID int,
CONSTRAINT PK_SI_North
PRIMARY KEY (OrderID, RegionID))

-- RegionID = 4 South

CREATE TABLE SalesInfoSouth (

OrderID int NOT NULL,
RegionID int NOT NULL
CHECK (RegionID = 4),
SaleDate datetime,
Amount money,
EmployeeID int,
CONSTRAINT PK_SI_South
PRIMARY KEY (OrderID, RegionID))

-- RegionID = 1 East

CREATE TABLE SalesInfoEast (

OrderID int NOT NULL,
RegionID int NOT NULL
CHECK (RegionID = 1),
SaleDate datetime,
Amount money,
EmployeeID int,
CONSTRAINT PK_SI_East
PRIMARY KEY (OrderID, RegionID))

-- RegionID = 2 West

CREATE TABLE SalesInfoWest (

OrderID int NOT NULL,
RegionID int NOT NULL
CHECK (RegionID = 2),
SaleDate datetime,

634
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

Amount money,
EmployeeID int,
CONSTRAINT PK_SI_West
PRIMARY KEY (OrderID, RegionID))
GO

-- Create a View that gets the entire

-- SalesInfo informations
-- Note the use of UNION ALL
-- This is the Partitioned View

CREATE VIEW SalesInfo

AS
SELECT *
FROM SalesInfoNorth
UNION ALL
SELECT *
FROM SalesInfoSouth
UNION ALL
SELECT *
FROM SalesInfoEast
UNION ALL
SELECT *
FROM SalesInfoWest
GO

-- Populate the partitioned tables

-- using the SalesInfo view directly
-- The partitioned view mechanism
-- will send every row to the appropriate
-- destination table automatically.

INSERT SalesInfo
SELECT o.OrderID,
T.RegionID,
O.OrderDate,
sum(UnitPrice * Quantity * (1-Discount)),
O.EmployeeID
FROM Orders O
JOIN [Order Details] OD
ON O.OrderID = OD.OrderID
JOIN EmployeeTerritories ET
ON ET.EmployeeID = O.EmployeeID
JOIN Territories T
ON T.TerritoryID = ET.TerritoryID
GROUP BY O.OrderID, T.RegionID, O.OrderDate, O.EmployeeID
GO
-- Checking number of rows in every table and total

SELECT COUNT(*) AS CountNorth

FROM SalesInfoNorth

SELECT COUNT(*) AS CountSouth

FROM SalesInfoSouth

SELECT COUNT(*) AS CountEast

FROM SalesInfoEast

SELECT COUNT(*) AS CountWest

FROM SalesInfoWest

635
Microsoft SQL Server 2000 Programming by Example

SELECT COUNT(*) AS CountTotal

FROM SalesInfo
GO

CountNorth
-----------
147

CountSouth
-----------
127

CountEast
-----------

417
CountWest
-----------

139
CountTotal
-----------
830

As you have seen in Listing 15.22, it is not necessary to insert data in the individual tables, because SQL
Server detects that you are using a partitioned view and sends every row to the appropriate table (even if the
table is stored in a different server).
Caution

The term partitioned view, although it is the official term that Microsoft gives to this technique, can
be misleading: The view is not partitioned; actually, it is the data that is divided, or partitioned,
across different tables. Using this technique, the view integrates the entire data set from the
partitioned table.

If every individual table is stored in a different server, or instance, the view is called a distributed partitioned
view. This technique provides great improvements on performance. Microsoft used this technique to execute
the SQL Server 2000 performance tests sent to the Transaction Processing Council (http://www.tpc.org).

Note

You can use partitioned views to speed up data retrieval in SQL Server 7.0. However, only SQL
Server 2000 supports updatable partitioned views. If you update a field that is part of the partitioned
key, SQL Server 2000 moves the affected rows to the appropriate table, according to the defined
partition schema.

636
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

When selecting data from the partitioned view, SQL Server decides automatically which table and server must
serve the request, and then divides the execution across the relevant servers .

Distributed Transactions

As you learned on Chapter 13, "Maintaining Data Consistency: Transactions and Locks," you can
consider a group of Transact-SQL statements as part of the same transaction. If the data affected by a
transaction is spread across different servers, you need to create a distributed transaction.
To create a distributed transaction, you must start the Microsoft Distributed Transaction Coordinator service
(MS-DTC), and the connection must use the SET_XACT_ABORT ON setting.
MS-DTC implements a two-phase commit mechanism to guarantee transaction consistency across different
servers. This process can be described as follows:

1. You connect to a SQL Server instance and start a distributed transaction, using the SET
XACT_ABORT ON and BEGIN DISTRIBUTED TRANSACTION statements.
2. You send a DML statement to another instance or server. MS-DTC running on your server must
contact the MS-DTC running on the other server to start a distributed transaction and to send the DML
statement to be executed remotely.
3. You can send other commands to other instances or servers, including the server you are connected
to. In every case, MS-DTC will check whether this connection already has a distributed transaction
with the target server.
4. When your operations have terminated and you want to commit all changes, send the COMMIT TRAN
statement.
5. MS-DTC takes control of the commit process and asks every participant server whether they are
ready to commit.
6. Every server answers the commit request, sending an affirmative or negative vote.
7. MS-DTC counts the votes received. If there is one negative vote, it informs every participant server
that they must roll back the operation. If all votes are affirmative, MS-DTC informs them that they can
finally commit.

Listing 15.23 shows an example of a distributed transaction, where you can update from two different
instances as part of the same transaction.

Listing 15.23 Use Distributed Transactions to Maintain Transaction Consistency Across Multiple
Servers

-- This setting is required

-- to start Distributed Transactions

SET XACT_ABORT ON
GO

-- Start a Distributed Transaction

BEGIN DISTRIBUTED TRANSACTION

637
Microsoft SQL Server 2000 Programming by Example

-- Modify data locally

UPDATE Northwind.dbo.Products
SET UnitPrice = UnitPrice * 1.1
WHERE CategoryID = 2

-- Modify the same data remotely

UPDATE SQLBEInst2.Northwind.dbo.Products
SET UnitPrice = UnitPrice * 1.1
WHERE CategoryID = 2

-- Confirm changes

COMMIT TRANSACTION
GO
(12 row(s) affected)

(12 row(s) affected)

You can create the DistSalesInfo distributed partitioned view, as shown in Listing 15.24. This view is
created based on the tables SalesInfoNorth and SalesInfoSouth that you created in the local server
with Listing 15.22. Now you must create the SalesInfoEast and SalesInfoWest tables in the
SQLBE\Inst2 instance, using the same script from Listing 15.22.
You can modify a record from the distributed partitioned view, as in Listing 15.24, to change the RegionID
field from North (3) to East (1). You can see how the record has been moved automatically from the
SalesInfoNorth table in the local server to the SalesInfoEast table in the linked server. The output
shows one row less in the local SalesInfoNorth table, and one more row in the remote SalesInfoEast
table.
In this case, it is not necessary to start a distributed transaction, because SQL Server does it automatically for
you, executing this statement in autocommitted mode.

Listing 15.24 Use Distributed Transactions to Maintain Transaction Consistency Across Multiple
Servers

----------------------------------------------------
-- NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE
--
-- Execute this script in the SQLBE\Inst2 instance
--
-- NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE
----------------------------------------------------

USE Northwind
GO

-- RegionID = 1 East

CREATE TABLE SalesInfoEast (

OrderID int NOT NULL,
RegionID int NOT NULL

638
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

CHECK (RegionID = 1),

SaleDate datetime,
Amount money,
EmployeeID int,
CONSTRAINT PK_SI_East
PRIMARY KEY (OrderID, RegionID))
GO
-- RegionID = 2 West

CREATE TABLE SalesInfoWest (

OrderID int NOT NULL,
RegionID int NOT NULL
CHECK (RegionID = 2),
SaleDate datetime,
Amount money,
EmployeeID int,
CONSTRAINT PK_SI_West
PRIMARY KEY (OrderID, RegionID))
GO

--------------------------------------------
-- NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE
--
-- Execute from here in the SQLBE instance
-- and make sure that MS-DTC is running
--
-- NOTE NOTE NOTE NOTE NOTE NOTE NOTE NOTE
--------------------------------------------
USE Northwind
GO

SET XACT_ABORT ON
GO

-- Populate the new table with information

-- from the same table in the SQLBE instance

INSERT SQLBEInst2.Northwind.dbo.SalesInfoEast
SELECT *
FROM SalesInfoEast

INSERT SQLBEInst2.Northwind.dbo.SalesInfoWest
SELECT *
FROM SalesInfoWest
GO

-- Create a View that gets the entire

-- SalesInfo informations
-- Note the use of UNION ALL

CREATE VIEW DistSalesInfo

AS
SELECT *
FROM Northwind.dbo.SalesInfoNorth
UNION ALL
SELECT *
FROM Northwind.dbo.SalesInfoSouth
UNION ALL
SELECT *
FROM SQLBEInst2.Northwind.dbo.SalesInfoEast
UNION ALL

639
Microsoft SQL Server 2000 Programming by Example

SELECT *
FROM SQLBEInst2.Northwind.dbo.SalesInfoWest
GO

SELECT COUNT(*) AS NorthBefore

FROM SalesInfoNorth

SELECT COUNT(*) AS EastBefore

FROM SQLBEInst2.Northwind.dbo.SalesInfoEast

UPDATE DistSalesInfo
SET RegionID = 1
WHERE OrderID = 10602

SELECT COUNT(*) AS NorthAfter

FROM SalesInfoNorth

SELECT COUNT(*) AS EastAfter

FROM SQLBEInst2.Northwind.dbo.SalesInfoEast
GO

NorthBefore
-----------
147

EastBefore
-----------
417

NorthAfter
-----------
146

EastAfter
-----------
418

What's Next?

In this chapter, you learned how to work with data from different instances, different servers, or even different
environments.
To execute the exercises in this chapter, review Appendix A, "Using SQL Server Instances," where you
can learn how to set up multiple SQL Server 2000 instances in the same server.
In this book, we tried to show you how to use SQL Server 2000 from a database developer's point of view.
Now you can try to apply these techniques to your own database environment.
You can obtain support and updated code from this book on
http://www.sqlserverbyexample.com

640
 Chapter 15. Working with Heterogeneous Environments: Setting Up Linked Servers

You can find extra SQL Server support in the Microsoft SQL Server public newsgroups, where Microsoft SQL
Server engineers, SQL Server Most Valuable Professionals (MVP), and many SQL Server professionals try
every day to learn a bit more about SQL Server and share their knowledge with their colleagues:

news://msnews.microsoft.com/microsoft.public.sqlserver.ce
news://msnews.microsoft.com/microsoft.public.sqlserver.clients
news://msnews.microsoft.com/microsoft.public.sqlserver.clustering
news://msnews.microsoft.com/microsoft.public.sqlserver.connect
news://msnews.microsoft.com/microsoft.public.sqlserver.datamining
news://msnews.microsoft.com/microsoft.public.sqlserver.datawarehouse
news://msnews.microsoft.com/microsoft.public.sqlserver.dts
news://msnews.microsoft.com/microsoft.public.sqlserver.fulltext
news://msnews.microsoft.com/microsoft.public.sqlserver.mseq
news://msnews.microsoft.com/microsoft.public.sqlserver.odbc
news://msnews.microsoft.com/microsoft.public.sqlserver.olap
news://msnews.microsoft.com/microsoft.public.sqlserver.programming
news://msnews.microsoft.com/microsoft.public.sqlserver.replication
news://msnews.microsoft.com/microsoft.public.sqlserver.security
news://msnews.microsoft.com/microsoft.public.sqlserver.server
news://msnews.microsoft.com/microsoft.public.sqlserver.setup
news://msnews.microsoft.com/microsoft.public.sqlserver.tools
news://msnews.microsoft.com/microsoft.public.sqlserver.xml
news://msnews.microsoft.com/microsoft.public.ae.arabic.sqlserver
news://msnews.microsoft.com/microsoft.public.de.sqlserver
news://msnews.microsoft.com/microsoft.public.es.sqlserver
news://msnews.microsoft.com/microsoft.public.espanol.sqlserver.administracion
news://msnews.microsoft.com/microsoft.public.espanol.sqlserver.olap
news://msnews.microsoft.com/microsoft.public.fr.sqlserver
news://msnews.microsoft.com/microsoft.public.il.hebrew.sqlserver
news://msnews.microsoft.com/microsoft.public.jp.sqlserver.server

641
 Appendix A. Using SQL Server Instances

Appendix A. Using SQL Server Instances

In previous versions of SQL Server, it was possible to install more than one instance of the SQL Server
engine in the same machine using some Registry tricks. Although this method did the trick, it was very tedious
and, more importantly, it was not supported by Microsoft. One of the new features Microsoft introduced in SQL
Server 2000 is the capability to run more than one copy or instance of SQL Server in the same computer. This
new feature is called multi-instance support and basically allows you to maintain multiple installations of SQL
Server running independently in just one server.
In some cases, it might be beneficial to maintain separate installations of SQL Server in one computer for
various reasons. For example, suppose two different customers of an application-hosting company need
administrative access to SQL Server, and the hosting company doesn't want each SQL administrator to
interfere with the activities of the other customer's administrator. To deal with this issue, the hosting company
can manage two different installations of SQL Server, one for each customer, keeping each one of them from
interfering with the other's installation. Previously, the way to overcome this limitation was to use one server
for each installation. How ever, using multi-instance support in SQL Server 2000, all installations can run
simultaneously and independently in just one server. In general, multi-instance is a cost-effective solution,
because you need to buy and manage just one server, instead of maintaining as many servers as installations
you have to support.
This appendix teaches you the following:

• How to install a new SQL Server instance

• How to connect to different instances of SQL Server in the same machine
• System functions used in multi-instance installations
• Current limitations of SQL Server instances

Installing SQL Server Instances

A SQL Server instance is a completely independent installation of the SQL Server engine and its related
services, such as SQL Agent. There are two types of SQL Server instances: default and named. A default
instance is identified by the name of the server where it runs, whereas named instances are identified by the
server name and the instance name (servername\instancename). There can be just one default instance
running in a server, and it works exactly as previous versions of SQL Server. Regarding named instances,
you can install as many named instances as you want in a specific server, even if there isn't a default instance
installed. However, Microsoft supports only a maximum of 16 named instances per machine.
A named instance can only be SQL Server 2000, whereas the default instance can be either SQL Server 6.5,
7.0, or 2000. Furthermore, the default instance can use version switching between SQL Server 6.5 and 7.0, or
between SQL Server 6.5 and 2000. Version switching enables you to keep two versions installed as default,
but only one of them is active at a time. Therefore, you can maintain three versions of SQL Server (6.5, 7.0,
and 2000) in the same machine, using the following configuration:

• Version switching between 6.5 and 7.0 as a default instance

• One or more named instances of SQL Server 2000

This type of environment is great for developers who must develop and test applications in multiple versions
of SQL Server.
Using named instances, different versions of SQL Server 2000 can be installed on the same machine. In
particular, you might have an instance running the standard edition of SQL Server, and another one running
the Enterprise Edition in the same machine. Regarding licensing, an additional license must be purchased for
each new instance installed in a server. As you can see, multi-instance can be useful for testing purposes,
because you can test the features of the Standard Edition and the Enterprise Edition using just one server,
instead of two. Moreover, because every instance runs independently, you can have the same version of SQL
Server 2000 in different instances with different service packs.
Caution

Be aware that all instances in one server share the system memory of the server. This is the
reason it is usually not recommended to install multiple instances in production systems.

643
Microsoft SQL Server 2000 Programming by Example

In regard to related services, each SQL Server instance has its own instance of SQL Server Agent.
Nevertheless, the Microsoft Distributed Transaction Coordinator (MSDTC) service and the Full text search
service have only one instance, which is shared among all SQL Server instances. Similarly, client tools, such
as Enterprise Manager, Query Analyzer, Profiler, Server Network Utility, Client Network Utility, isql, and osq,
are shared among instances.

In this appendix, you will install a new named instance of SQL Server 2000 Enterprise Edition in a server
called SQLBYEXAMPLE. Be aware that Internet Explorer 5.0 is a prerequisite of the installation of SQL
Server 2000, because the MMC (SQL Server snap-in) and the Books online are HTML based. Maybe you
already have this prerequisite if there's already a default instance or another named instance of SQL Server
2000 in the machine where the new named instance will be installed.

Also, make sure that any service related to SQL Server is stopped before proceeding with the installation
process of the new SQL Server named instance. For example, if you have a Web server or a transaction
server connecting to SQL Server, stop these services first, and proceed to install the ne0w instance.

To begin the installation process, insert the SQL Server Enterprise Edition CD and this will automatically begin
the installation process.

The first step is to choose whether it will be a local or a remote install. If you choose the remote install option,
SQL Server creates an unattended installation file called setup.iss, copies the installation files to the
remote server, and then performs the actual installation. In our case, you have to choose Local Computer
because you will be performing a local installation, as shown in Figure A.1.

Figure A.1. Performing a local or remote installation.

644
 Appendix A. Using SQL Server Instances

In the next window, you must choose the type of task to perform in the installation. By choosing the first
choice (the one you will choose), a new instance of SQL Server is installed. The second choice allows you to
modify an existing installation of SQL Server, and the last option is used to manage cluster installations,
rebuild the Registry, or create an unattended installation file (an .iss file). Figure A.2 shows this window.

Figure A.2. Choosing the action to take in the installation process.

Next, you are required to enter your name and company name, and then you have to accept the license
agreement. In the next window, you must choose the components that you want to install. In this case, Server
and Client Tools is selected because you want to install the SQL Server engine and also the client tools.
Using the third option, you can install just the Microsoft Data Access Components (MDAC 2.6).
This window is shown in Figure A.3. Be aware that if you select the first or the second option, the installation
process overwrites any client tools that you have previously installed on the server, because there can be only
one copy of the client tools in a server, regardless of the number of instances running.

Figure A.3. Choosing the components to be installed.

645
Microsoft SQL Server 2000 Programming by Example

In the next window, you must specify the type of instance to install, either default or named. If there's already
a default instance installed on the server, the first choice (install a default instance) is grayed out because
there can be only one default instance, and you will only be allowed to enter an instance name, as Figure
A.4 shows. The name of the instance you'll be installing is APPENDIXA.

Figure A.4. Choosing the type of instance to install.

Then, choose the type of installation of SQL Server you are performing (Typical, Minimum, or Custom), and
also the path where files will be installed. Specifically, the elements of the SQL Server installation are the SQL
Server engine, replication tools, full-text search, client tools, client connectivity, Books Online, upgrade tools,

646
 Appendix A. Using SQL Server Instances

development tools, and code samples. The typical installation includes all elements except the code samples,
and the minimum installation includes only the engine, replication, full-text search, and client connectivity.
These options appear in Figure A.5.

Figure A.5. Type of installation and path of files.

If you chose Custom installation, you will see the screen shown in Figure A.6, in which you must choose the
elements to install. Otherwise, if you choose either Typical or Minimum, the installation process takes you
directly to the screen shown in Figure A.7.

Figure A.6. Choosing each individual component to install.

647
Microsoft SQL Server 2000 Programming by Example

In the next window, you configure the accounts that will be used by the SQL Server service and the SQL
Agent service of the new instance, as shown in Figure A.7. Also, these services can be set to start
automatically when Windows starts up. You can configure both services to use the same account or to use
different ones.

Figure A.7. Configuring the service accounts.

The account(s) used by SQL Server and SQL Agent can be either localsystem, which is an account with
administrative privileges on the local server (similar to a local administrator), or a domain account. If SQL
Server won't be performing any activity that involves any other server in the domain, localsystem may be the
solution. However, if SQL Server performs activities that involve other servers— for example, taking backups
and storing them in another server— SQL Server must use a domain account that has ap propriate
permissions on the server where the backup files will be stored.
In the next window, Figure A.8, the authentication mode used by SQL Server is set up. Windows
authentication mode is the recommended one and the default one in the installation process (the installation is
secure out of the box). Using this type of authentication, SQL Server doesn't store passwords; it just stores
the SID, which is an identifier of Windows login, and users don't have to specify a password when connecting
to SQL Server (because they were already validated by Windows). Therefore, any user who has a Windows
NT 4.0 or Windows 2000 account can benefit from using the Windows authentication mode or trusted
connections.

Figure A.8. Configuring the Authentication mode.

648
 Appendix A. Using SQL Server Instances

Mixed mode allows users to connect to SQL Server using a SQL Server login and password. This mode is
useful when you have non-Windows clients connecting to SQL Server. If you choose SQL Server and
Windows authentication (mixed mode), you have the choice to leave the sa password blank, which is not
recommended because this leaves the system unprotected (any user can log in as sa with a blank password).
In this case, it's highly recommended that you change the sa password immediately after the installation
process.
You must choose the default collation (sort order plus character set) used by this installation of SQL Server.
The collation defines the way data is stored and compared in SQL Server. In some cases, it might be
beneficial to change the default collation if you want to store data in a different language.
In SQL Server 2000, you can have different collations in the same instance, even at column level; thus, a
single table can have different collations. This window is shown in Figure A.9.

Figure A.9. Choosing the default collation.

649
Microsoft SQL Server 2000 Programming by Example

In the next window, the network libraries used by this instance of SQL Server are configured. The only ones
you can choose when installing a named instance are named pipes, TCP/IP, and NWLink because you
cannot connect to a SQL Server named instance using any of the other network libraries.
Basically, a network library enables different clients to communicate with SQL Server. To be able to
communicate with SQL Server, clients must connect to the server using a network library used by the server.
Also, SQL Server might be using more than one network library simultaneously. For example, if the enabled
network libraries in SQL Server are named pipes and TCP/IP, some clients can connect using named pipes,
and others can connect using TCP/IP.
If you leave the port number as the default (0) in the TCP/IP network library, the TCP port used by the SQL
Server instance is automatically chosen by SQL Server every time it is started, unless you specify a port
different from zero at installation time. To check the current TCP port used by a SQL Server instance, use the
Server Network utility, and check the properties of the TCP/IP network library.
After the installation process, network libraries can be reconfigured using the Server Network Utility. Figure
A.10 shows the window used to configure network libraries.
Afterward, a window states that to complete the setup process it only needs the licensing information, which is
configured in the next window of the installation process.
Two licensing modes are available: per seat and per processor. Use per seat mode when you know
beforehand how many clients will be connecting to SQL Server. In this mode, you need a server license for
each server or instance, and a client access license (CAL) for each device that will be connecting to SQL
Server.

Figure A.10. Configuring the server network libraries.

650
 Appendix A. Using SQL Server Instances

Use per processor mode if you want to allow unlimited connections to SQL Server, directly or indirectly. For
example, if you run a Web site, you can use per processor licensing to allow unlimited connections to SQL
Server from the Web server (indirect connections). In this mode, a license for each processor of the server is
required. For example, if SQL Server has two processors, you need two per processor licenses (even if SQL
Server is configured to use just one processor).
The licensing window is shown in Figure A.11.
Caution

The Internet connector license was used in previous versions of SQL Server to allow unlimited
connections to SQL Server. This type of license was discontinued in SQL Server 2000. Now, use
the per processor license instead.

After the installation process gathers all necessary information, it begins the actual installation of the
components in the following order:

1. Microsoft Data Access Components (MDAC)

2. Distributed Transaction Coordinator (MSDTC)
3. SQL Server engine
4. Run configuration scripts
5. Register ActiveX components

Figure A.11. Choosing the licensing mode.

651
Microsoft SQL Server 2000 Programming by Example

Each SQL Server instance has its own directory with its data files. However, the common tools (client tools)
for all instances are stored in a directory called 80 that is located inside the Microsoft SQL Server folder.
Every SQL Server named instance has its own SQL Server and SQL Agent services. The names of these
services are mssql$instancename (SQL Server service) and sqlagent$instancename (SQL Agent). These
services are listed in the Services window (located in the Control Panel in Windows NT 4.0, or in the
Administrative tools folder in Windows 2000), which is shown in Figure A.12.

Figure A.12. The Services window.

652
 Appendix A. Using SQL Server Instances

These services can be started, stopped, and paused from the Services window, the command prompt,
Enterprise Manager, or the SQL Server Service Manager. In particular, to start, stop, or pause the SQL Server
service of an instance from the command prompt, use the following commands, respectively:

net start mssql$instancename

net stop mssql$instancename
net pause mssql$instancename
To uninstall a SQL Server named instance, use one of these two approaches:

• Use the Add/Remove Programs window located in the Control Panel. In this window, shown in Figure
A.13, you can see a list of all instances installed in the local server and the option to change or
completely remove the instance from the server.

Figure A.13. Uninstalling a SQL Server instance.

• Rerun the SQL Server installation program and choose the Remove Components option.

Connecting to Instances

After an instance is completely installed, you can verify that the installation was successful connecting to this
new instance. In particular, to connect to a named instance, use the following syntax:
servername\instancename. Notice that when there are many instances in one server, from the client's
point of view, each instance works as a completely different server.
There are many ways to test the connection to SQL Server. For example, you can use osql, which is a
command-line utility that connects to SQL Server using ODBC. Figure A.14 shows how this command is
used from the DOS prompt.

Figure A.14. Using osql to test connectivity to SQL Server.

653
Microsoft SQL Server 2000 Programming by Example

Clients that connect to a named instance of SQL Server 2000 must have installed at least the Microsoft Data
Access Components (MDAC) 2.6, which are the ones installed by SQL Server 2000. Therefore, if you install
SQL Server 2000's client utilities in a client machine, you will be able to connect to named instances. However,
you can install MDAC 2.6 separately in a client machine without installing the client tools of SQL Server 2000.
MDAC 2.6 can be downloaded directly from Microsoft's Web site (http://www.microsoft.com/data).
Another way to connect to a named instance of SQL Server 2000 is by creating an alias in the client machine
using the Client Network utility, which is one of the SQL Server client tools. When creating the alias, you must
specify the name of the alias (this is the one you will use to connect to the named instance), the network
library, and the name of the instance— for example, SQLBYEXAMPLE\APPENDIXA. Figure A.15 shows the
creation of an alias called TESTINSTANCE using the Client Network utility.

Figure A.15. Using the Client Network utility to create aliases.

After the alias is created, you can connect to the named instance using the alias name (TESTINSTANCE),
instead of servername\instancename. Figure A.16 shows how to connect to a named instance using an
alias in the Query Analyzer.

Figure A.16. Connecting to a SQL Server instance using an alias.

654
 Appendix A. Using SQL Server Instances

If you create an alias using the TCP/IP network library, it is recommended you set the Dynamically Determine
Port option, because a SQL Server named instance, by default, chooses an available TCP port every time it is
started. Therefore, you don't know the TCP port used by the named instance to accept incoming connections
beforehand, unless you specify a port for the named instance in the Server Network utility. Figure A.17
shows how to configure a SQL Server 2000 named instance to use a specific TCP port (8888 in this case).

Figure A.17. Specifying the TCP port used by an instance to accept incoming connections.

Caution

If you change the TCP port used by a named instance, the SQL Server service of this instance
must be restarted for this change to take effect.

Then, you can create an alias to connect to the named instance using the port specified in the Server Network
utility. This is shown in Figure A.18.

Usually, applications connect to SQL Server using either ODBC or OLE DB. Specifically, in OLE DB
connection strings, the name of the server is specified using the following syntax: Data
Source=computername\instancename. The network library can also be specified in the connection
string. For example, Listing A.1 shows a connection string that connects to the Northwind database located
in SQLBYEXAMPLE\APPENDIXA, using integrated security (Integrated Security=SSPI) and the
TCP/IP network library (Network Library=dbmssocn).

655
Microsoft SQL Server 2000 Programming by Example

Figure A.18. Specifying the TCP port when creating an alias.

Listing A.1 An OLE DB Connection String Used to Connect to a Named Instance

"Provider=SQLOLEDB.1;Data Source=dev08\instance2;Integrated Security=SSPI;

Initial Catalog=Northwind;Network Library=dbmssocn"
You can issue distributed queries to a named instance through a linked server. Specifically, when querying
tables in the linked server use the following syntax:

[servername\instancename].database.username.objectname
Listing A.2:shows how a linked server is configured when connecting to a named instance, and then it issues
a query against the linked server.

Listing A.2 Creating a Linked Server Using Instances

sp_addlinkedserver @server = 'dev08\instance1'

GO

SET ANSI_NULLS ON
SET ANSI_WARNINGS ON
GO

SELECT * FROM [dev08\instance1].Northwind.dbo.Shippers

GO
(1 row(s) affected)

656
 Appendix A. Using SQL Server Instances

(1 row(s) affected)
ShipperID CompanyName Phone
----------- ---------------------------------------- ------------------------
1 Speedy Express (503) 555-9831
2 United Package (503) 555-3199
3 Federal Shipping (503) 555-9931
(3 row(s) affected)

System Functions Used in Multi-Instance Installations

Two system functions return information related to instances:

• SERVERPROPERTY—This function takes one parameter and returns information about the server
where the instance is installed. If the parameter is machinename, this function returns the name of
the server. If the parameter used is servername, it returns the name of the server, along with the
name of the instance. If instancename is used as the parameter, the function returns the name of
the current instance, and if it's the default instance, it returns NULL. Listing A.3 shows how to use
this function with the parameters described.

Listing A.3 Using the SERVERPROPERTY Function

USE Master

SELECT SERVERPROPERTY('machinename')
SELECT SERVERPROPERTY('servername')
SELECT SERVERPROPERTY('instancename')
GO

-------------------------

SQLBYEXAMPLE

(1 row(s) affected)

657
Microsoft SQL Server 2000 Programming by Example

-------------------------
SQLBYEXAMPLE\APPENDIXA

(1 row(s) affected)

-------------------------
APPENDIXA

(1 row(s) affected)

• @@SERVERNAME—This function returns the name of the server and current instance of SQL Server.
This is equivalent to SERVERPROPERTY using the servername parameter. @@SERVERNAME is shown
in Listing A.4.

Listing A.4 Using the @@SERVERNAME Function

USE Master

SELECT @@SERVERNAME
GO
-------------------------

SQLBYEXAMPLE\APPENDIXA

(1 row(s) affected)

Current Limitations
As any other new feature, SQL Server's multi-instances has a lot of great and useful characteristics, but also
has some limitations.

• First of all, not all network libraries work with instances. Specifically, only named pipes, TCP/IP, and
NWLink can be configured in a server with multiple instance installations. The remaining ones that
don't work with SQL Server instances are Multiprotocol, AppleTalk, and Bany an VINES, which
Microsoft didn't enhance in SQL Server 2000. This is the reason they cannot be used to communicate
with named instances.
• In a server with multiple instances, there's only one version of the client tools. For example, if there's a
default instance running SQL Server 7.0, and then a SQL Server 2000 named instance is installed,
the client tools of SQL Server 2000 overwrite SQL Server 7.0's tools. Therefore, you cannot have

658
 Appendix A. Using SQL Server Instances

different versions of the client utilities installed in one machine. The only exception is Books Online,
which is kept in the system even if a newer version of the client tools is installed. For example, you
can browse the Books Online of SQL Server 7.0 and 2000 in the same computer.
• A limitation of the installation process is that the client tools are always installed in the system drive
(usually C). Sometimes, this can be a serious problem if no space is available in the system drive.

Finally, any suggestion that you might have regarding instances or any other aspect of SQL Server can be
sent by email to sqlwish@microsoft.com. The SQL Server development team constantly monitors this
email address; therefore, there's a good chance that your comment will reach the right person, and eventually,
it might become a feature of an upcoming version of SQL Server.

659
 Appendix B. Using SQL Query Analyzer

Appendix B. Using SQL Query Analyzer

As a relational database management system, SQL Server stores data, enforces security, maintains data
integrity, and executes user queries to retrieve any requested data.
In a client/server environment, the client application sends queries to SQL Server to be executed at the server
side and is the client application that manages the query results and presents them to the user. SQL Server
provides a client application to execute queries and retrieve results; this application is SQL Query Analyzer.
Although SQL Query Analyzer does not pretend to substitute the role of a standard application, you cannot
expect users to work with SQL Query Analyzer as their standard front end for database applications. However,
SQL Query Analyzer is very useful for specific activities such as the following:

• Creating and testing the database objects during the development process
• Tuning and optimizing the database while in production
• Executing common administrative tasks

In this book, you use SQL Query Analyzer version 8.00.194, as included in the original release (RTM) of
Microsoft SQL Server 2000.
This appendix teaches you how to

• Connect to SQL Server

• Set user options
• Write and execute queries
• Search for database objects
• Use the Object Browser to navigate your database structure
• Analyze queries
• Use the Transact-SQL Debugger

Installing SQL Query Analyzer

You install SQL Query Analyzer as part of the SQL Server 2000 installation process. The Typical and
Minimum setups both install SQL Query Analyzer by default, but the Custom setup gives you the option of
installing this tool, as you can see in Figure B.1.

Figure B.1. Using the Custom setup, you can select which SQL Server components to install.

661
Microsoft SQL Server 2000 Programming by Example

It is possible to install SQL Query Analyzer on a client computer even if the server components are not
installed on the same computer. This allows for remote access to SQL Server.
A special case is the remote access to a computer through the Internet. In this case, the simplest way is to
create a server alias using the Client Network utility, specifying the IP address as server name and a server
alias name to be used when connecting by SQL Query Analyzer and other client tools. Figure B.2 shows a
typical example of this alias creation form.

Figure B.2. Using the Client Network utility to define a server alias simplifies the access to remote
servers.

The Query Analyzer Workplace

662
 Appendix B. Using SQL Query Analyzer

When you run SQL Query Analyzer, you will get a connection form, as you can see in Figure B.3, where you
can select which server or server instance to connect to, and which authentication mode to use. If you
selected SQL Server Authentication mode, you must provide a valid login and password to connect.

Figure B.3. In the Logon screen, you can select the server you want to connect to and which
authentication mode to use.

Caution

If you try to connect to SQL Server using the SQL Server aut hentication mode and you fail to
supply a valid login name, SQL Query Analyzer will not try to connect using your Windows NT 4.0
or Windows 2000 credentials, as it did in previous versions, so the connection will be rejected. It is
recommended to use integrated security if possible.

Initially, the Server drop-down list will not contain any entries. The server name will be added to the list only
after successful connection to a server.

Tip

SQL Query Analyzer will try to connect to the local default instance if you leave the server name
blank or write a dot '.'or '(local)'as the server name.

Caution

If you installed SQL Server 2000 with SQL Server 7.0 or 6.5 on the same computer, the default
instance will be either SQL Server 7.0 or SQL Server 6.5, which means that many of the examples
in this book will not work because of the different functionality between SQL Server 2000 and
earlier versions. So, make sure you connect to the right instance of SQL Server 2000 before trying
any of the examples.

663
Microsoft SQL Server 2000 Programming by Example

When connected, SQL Query Analyzer will show a workplace as shown in Figure B.4.

As in other Windows applications, you can see some common elements such as the menu bar, the toolbar,
and the status bar.

Figure B.4. The SQL Query Analyzer workplace is similar to all Windows applications in look and feel.

SQL Query Analyzer is a Multi Document Interface (MDI) application where every connection to SQL Server
uses its own child window, inside the parent window. You can select the connection window by using the
application window menu as usual, and you can arrange the windows in the common ways: Cascade, Tile
Horizontally, and Tile Vertically.
Tip

If you have established more than one connection to SQL Server from SQL Query Analyzer, it is
usually better to maximize the connection window to have extra visible space to work with the
Editor and Results panes. As soon as you maximize one of the connection windows, all the other
connection windows will be maximized as well.

The Editor Pane

The Editor pane of a connection window is a text editor you can use to edit your queries and SQL statements.
These statements can come from different sources as follows:

• Scripts written directly in the Editor pane

• Saved SQL script files you can retrieve in the Editor pane for further editing

664
 Appendix B. Using SQL Query Analyzer

• Scripts produced from the Object Browser and Object Search, as you will see later
• Predesigned templates you can reuse to speed up and facilitate the process of writing code

This is a code-oriented text editor with many features to help you write SQL language code. You can find
information about these features in Books Online, but it is interesting to pay attention to the following points:

• The text is color coded to differentiate between keywords, system objects and system stored
procedures, operators, comments, and so on.
• You can use shortcuts to speed the editing process. You can find the full shortcut list in the topic "SQL
Query Analyzer Keyboard Shortcuts" in Books Online, but the more frequently used are in Table B.1.
• It is possible to select the font, tab length, and other options, as you can see in Figures B.13 and
B.15 later in this appendix.
• You can select which part of the code to execute by selecting the block either with the mouse or with
the keyboard combining the Shift key with the navigation keys.

Table B.1. SQL Query Analyzer's More Useful Shortcuts

Shortcut Action
F5, Ctrl+E, or Executes the selected query
Alt+X
Ctrl+F5 Checks the syntax of the selected query
Alt+Break Cancels the execution of the query that is being executed
Shift+F1 Searches in Books Online for the selected word, SQL statement, keyword, or SQL Server
object in Books Online
Ctrl+Shift+Del Clears the current Editor pane window to provide an empty window to start editing again
(see the tip following this table)
Ctrl+C or Copies the marked block to the Clipboard
Ctrl+Insert
Ctrl+X or Shift+Del Cuts the marked block and send it to the Clipboard
Ctrl+V or Pastes the contents of the Clipboard
Shift+Insert
Ctrl+Shift+C Marks the current block of text as comments (adds -- at the beginning of every line)
Ctrl+Shift+R Uncomment the current block of text (removes the -- from the beginning of every line)
Ctrl+Shift+L Converts the selected text to lowercase
Ctrl+Shift+U Converts the selected text to uppercase; this is useful to highlight keywords in the code
Ctrl+Z Undoes the latest editing action
Tip

As you can select which block of your code to execute, it is not advisable to clear the window to
write new code, unless, of course, you wanted to start completely from scratch. Clearing the Editor
Pane window prevents you from being able to copy previous statements to save coding time.

The Object Browser

Writing queries requires precise knowledge of database objects definition and complete syntax of statements,
functions, and stored procedures. Developers can use the SQL Query Analyzer Object Browser to get
information about database objects. Moreover, Object Browser is a dynamic tool with embedded scripting
capabilities that can greatly improve your coding productivity.

665
Microsoft SQL Server 2000 Programming by Example

As you can see in Figure B.5, Object Browser has two main sections: the server structure, with information
about databases and database object of the server that the user is connected to, and the common objects,
with information about system functions and data types.

Figure B.5. In the Object Browser structure, you can see two main sections: server and common
objects.

Caution

It is a common mistake to consider the system stored procedures as system functions, because
system stored procedures, as described in Chapter 8, are no more than standard stored
procedures created during the installation process in the Master database. That's why Object
Browser shows them in their original place, which is under the Server section, Master database,
Stored Procedures.

As you can see in Figure B.6, you can right-click any object to see a context menu from where you can
create a script of that object either to the Clipboard, a new window, or a file. Depending on the object type, it is
possible to script different statements. In the example shown in Figure B.6, a user table, the available
statements are CREATE, DROP, SELECT, INSERT, UPDATE, and DELETE. For stored procedures, you can
choose to script the CREATE, ALTER, DROP, and EXECUTE statements.

Tip

666
 Appendix B. Using SQL Query Analyzer

During the coding process, scripting to the Clipboard is usually more flexible than scripting to a new
window because after it is in the Clipboard, you can paste the script anywhere in your code, even
several times.

Figure B.6. Right-click any object to show the Object Browser context menu.

In Figure B.7, you can see how the Object Browser shows object dependencies, as well as definitions of
columns and parameters. The Object Browser enables the user to drag and drop any object's name to the
Editor Pane, reducing the likelihood of misspelling an object name.
Caution

Always check dependencies before altering or dropping any object. Failing to do so could produce
errors when other objects must reference the modified or missing object.

An interesting feature in SQL Server 2000 is the possibility of defining extended properties for databases and
database objects. From the Object Browser it is possible to edit them using the context menu, as you can see
in Figure B.8.

Figure B.7. Object Browser shows complete information about tables such as Columns, Indexes,
Constraints, Dependencies, and Triggers.

667
Microsoft SQL Server 2000 Programming by Example

Figure B.8. Extended properties for a table, identifying the default form position and size for a
fictitious Visual Basic client application.

668
 Appendix B. Using SQL Query Analyzer

The Object Search

Working with several databases, and hundreds or thousands of objects, makes it difficult to find specific
objects in Object Browser. Using the Object Search you can efficiently search for any object on any database.
Figure B.9 shows the Object Search window. You can use wildcards in the Object name field, as seen in
Figure B.9. Searching for ___help* means any three characters followed by the word help followed by any
string. Other combinations, such as *sys* sys* or authors, are valid as well.
You can search for objects with specific extended properties values.
Tip

Object Search is especially useful when you want to know which table contains the information you
need to select in your query. In this case, provide an approximate column name, using wildcards,
and specify Column as object type.

Figure B.9. Searching for an object is very efficient using the Object Search window.

The Results Pane

The lower section of every connection's window, below the Editor pane, is the Results pane, in which SQL
Query Analyzer shows the results of the queries and any output message from SQL Server.
SQL Query Analyzer can show results in Text or Grid, in a spreadsheetlike format. It is possible to send
results directly to a file, in which case the Results pane shows only SQL Server messages and confirmation of
the file write operation as shown in Listing B.1.
To send results to a file, select the menu Query— Results to File. With this setting, whenever you execute a
query, SQL Query Analyzer will prompt you to provide a filename, and the results will be sent to the selected
file. To test how SQL Query Analyzer sends results to a file, make sure the menu Query— Results to File is
checked and execute the example of Listing B.1.

Listing B.1 The Results of a Query Can Be Sent to a File

669
Microsoft SQL Server 2000 Programming by Example

USE Northwind
GO

SELECT *
FROM Products
(77 row(s) affected)
The following file has been saved successfully:
C:\WINNT\Profiles\Administrator\My Documents\fn.rpt 14242 bytes
In Figure B.10, you can see the same results in text, and in Figure B.11 in grid.

Figure B.10. You can show query results in text mode.

Note

Results in grid are read-only. If you want to edit table values directly in a grid from SQL Query
Analyzer, go to Object Browser, right-click in the object, and select Open. This way, SQL Query
Analyzer will open a new window with a fully editable grid.

670
 Appendix B. Using SQL Query Analyzer

Caution

Executing more than a single query in the same batch when Grid mode is selected produces
several grids in the Result pane. This can be a bit confusing because of the number of different
slide bars in the Result pane. The main slider belongs to the Result pane, and every individual grid
has its own slider, too.

Figure B.11. Dealing with query results with multiple columns is easier in grid mode.

Managing User and Connection Options

SQL Query Analyzer is a flexible environment that can be customized to specific user needs, as you can do in
Microsoft Office applications. Some of these settings affect the way SQL Query Analyzer looks, the way it
shows information. and how it interacts with the user, whereas other settings affect the way it connects to SQL
Server.

Customizing SQL Query Analyzer

You can change SQL Query Analyzer settings using two different menus under the Tools menu title: Options
and Customize. In Figure B.12, you can see how to change some general settings about SQL Query
Analyzer.
Tip

Check the Select SQL Query Analyzer as the Default Editor for Query File Extensions option,
because it can help you open files using Windows Explorer or any standard folder window.

671
Microsoft SQL Server 2000 Programming by Example

Caution

You are not forced to use the recommended extensions. However, if every computer uses different
extensions for scripts, results, or templates, sharing files will be more complex, and association of
filename extensions to Query Analyzer can be a more complex task.

Figure B.12. You can select SQL Query Analyzer as the default editor for scripts file with the general
options.

In Figure B.13, SQL Query Analyzer enables you to change some settings that influence how it displays the
Editor pane.

Figure B.13. You can change tab size and tabbed mode using the Editor Options.

672
 Appendix B. Using SQL Query Analyzer

Adjust the tab size to a value big enough to facilitate the programming logic readability, but not too high or you
will be forced to use the horizontal sliders to read the code, and it will not print properly. Normal values for tab
size are 3, 4, or 5, but feel free to select your own value if it suits you better.
The lower part of the form offers you the choice of showing the connection window as a tabbed window, with
the Edit and Results panes under different tabs.
In Figure B.14, you can see how to change the way SQL Query Analyzer produces results. You can change
the default results target in a way similar to the toolbar.

Figure B.14. It is possible to control the results output.

673
Microsoft SQL Server 2000 Programming by Example

Tip

If you intend to export the results to a spreadsheet application, you should consider setting the
Results Output Format option to Comma Delimited (CSV).

Caution

The Maximum Characters per Column setting is useful to limit the output in the Results pane, but if
you selected results in a file, be aware that columns results will be truncated to the maximum
characters per column value you selected.

SQL Query Analyzer uses different font settings to help the developers identify statements, operators, system
objects, and other specific system-supplied keywords during the code-writing process. You can change these
settings as shown in Figure B.15. For every category, you can select which font SQL Query Analyzer will use
for text, but for the Editor category, you can specifically select which font to use to highlight every keyword
type.

It is possible to set how SQL Query Analyzer produces scripts using the Script tab in the Options form.

674
 Appendix B. Using SQL Query Analyzer

It is useful to select the Prefix the Script with a Check for Existence option because this will avoid
unnecessary error messages when running the script, if the objects to be created already exist in the
destination database.

Select Script Object-Level Permissions to guarantee the destination objects will have the same access level
as the original ones.

Figure B.15. Selecting the right fonts for every part of the code makes it more readable.

Unless you have the same user-defined data types in the destination database, you should select the Convert
Specification of User-Defined Data Types to the Appropriate SQL Server Base Data Type option so that only
native data types will be scripted. If the destination server is SQL Server 7.0 or 6.5, check that the data types
are supported in the target version.
In Figure B.16, you can see some of the scripting options that SQL Query Analyzer offers you.

Figure B.16. SQL Query Analyzer offers many options to create scripts according to the programmer's
needs.

675
Microsoft SQL Server 2000 Programming by Example

Caution

If you had to use the script to re-create objects on any previous version, including SQL Server 7.0,
make sure you check the script completely to test that it includes only available features in the
target version. Selecting the Only Script 7.0 Compatible Features option does not always produce
the expected results.

It is possible to create user-defined shortcuts to speed up writing some repetitive sentences. SQL Query
Analyzer provides the following predefined shortcuts:

Shortcut Action
Alt+F1 sp_help
Ctrl+1 sp_who
Ctrl+2 sp_lock
The previous shortcuts are fixed and cannot be changed, but programmers can define actions for shortcuts
Ctrl+F1, Ctrl+0, and Ctrl+3 to Ctrl+9.

When you use these shortcuts, nothing is written to the Editor pane, the sentence is sent to SQL Server, and
the Results pane shows the results of the statement.

Figure B.17 shows the Customize form, where you can define new shortcuts or modify or remove existing
ones.

676
 Appendix B. Using SQL Query Analyzer

Figure B.17. It is possible to create shortcuts using the Customize menu.

Connection Settings

Every connection to SQL Server uses different settings, which affect the way SQL Server produces results
and, in some cases, the way SQL Server performs some operations.
SQL Query Analyzer offers three ways to change connection settings:

• Use the Tools, Options menu to change the default connection settings for new connections.
• Use the Query, Current Connection Properties menu, or the correspondent icon in the toolbar, to
modify specific settings for the current connection.
• Write the required SET commands, in the Editor Pane, to change specific settings for the current
connection.

The settings of the current connection are changed any time the default connection settings change, but not
vice versa.
Tip

If you want to start a new connection with the same settings as the current connection, you can
click the New Query icon in the toolbar, press Ctrl+N, or choose File, New, Blank Query Window.
Otherwise, a new connection with the default settings will be created.

Figure B.18 shows some connection options. You can access this configuration form from the Tools, Options
menu.

677
Microsoft SQL Server 2000 Programming by Example

Figure B.18. SQL Query Analyzer provides full connection settings control using the Options menu.

The value for Query Time-Out and Lock Time-Out are related to each other. By default, SQL Query Analyzer
assumes unlimited Lock Time-Out and Query Time-Out, but it is possible to change this setting to abort
queries that have waited for longer than a specified time for locks to be released by adjusting the Lock Time-
Out to a shorter time than the Query Time-Out.
Caution

Take care to remember that Query Time-Out is specified in seconds and Lock Time-Out in
milliseconds.

The default value for Query Time-Out is 0 seconds, which means indefinite waiting time. It is impossible to set
Query Time-Out to no waiting time because SQL Server would be unable to execute any query.

You can select 0 milliseconds to specify "no waiting time" for Lock Time-Out, so as soon as the query detects
any locks on required resources, the execution of the query will be cancelled.

If your client and your server use different collation settings, you should check the Perform Translation for
Character Data option. This instructs the driver to translate the character data between collations. This option
is useful when, for example, the client has a Western European collation and the server uses an Eastern
European collation.

The last setting, Use regional settings when displaying currency, number, dates, and times, formats the query
results according to the regional settings specified in Control Panel.

678
 Appendix B. Using SQL Query Analyzer

Caution

Setting Use regional settings when displaying currency, number, dates, and times affects only the
way results are formatted, but not the format of input values. To change the format for input date
values, you should execute the SET DATEFORMAT statement first.

The connection properties manage the way SQL Server executes the query. Figure B.19 shows the
connection properties for the default connection.

Figure B.19. It is possible to manage specific connection properties for the current connection or
default properties for new connections.

Selecting Set parseonly instructs SQL Server to check the syntax correctness of the query, without executing
it. Object names are not resolved with this option set, so it is possible to have some nonexistent objects
referenced in your queries.
Selecting Set noexec instructs SQL Server to check syntax, resolve names, and compile the query, saving the
compiled query plan in the procedure cache but without executing the query.
To avoid unnecessary network traffic it is best to select the Set nocount option.
Later in this appendix, you'll learn how to use the execution options that provide execution statistic information.
To select the default options, you can click the Reset All button.

Defining and Using Templates for Query Analyzer

To help in the writing of repetitive queries, SQL Query Analyzer provides templates. These are scripts, saved
in the directory Templates/SQL Query Analyzer, with special directives to SQL Query Analyzer to accept
template parameters.

679
Microsoft SQL Server 2000 Programming by Example

To define a new template, write your script as usual and save it with the .tql extension in the SQL Query
Analyzer templates directory (it is C:\Program Files\Microsoft SQL
Server\80\Tools\Templates\SQL Query Analyzer by default).
To use a template, you can use the templates pane of the Object Browser or the Edit, Insert template menu.
Whichever one you use, the template will be inserted in the Editor pane. Figure B.20 shows the Templates
pane of the Object Browser.
If the template uses parameters, you can either edit the template to change values manually, use the Edit,
Replace Template Parameters menu, or Ctrl+Shift+M to edit in the Replace Template Parameters form as
shown in Figure B.21.
Tip

It is more efficient to save frequent scripts as templates, instead as normal scripts, because they
provide parameters functionality, and they are listed in the Templates pane, which makes it easy to
reuse these scripts.

Figure B.20. By using templates, you can save time reusing frequent scripts.

Figure B.21. Specifying parameter values and data types is easy using the Replace Template
Parameters form.

680
 Appendix B. Using SQL Query Analyzer

Analyzing Queries
SQL Query Analyzer is not only a client tool to edit SQL queries and to show the correspondent results. SQL
Query Analyzer provides analyzing capabilities that can help tune our databases and the queries they have to
answer.
Analyzing queries is an advanced task, but it is important to at least show the very basic analyzing tasks every
programmer should know.

Obtaining Information About Query Execution

The first question about query execution is usually: How long will it take to execute?
To answer this question, you can execute the SET STATISTICS TIME ON statement, or you can use the
connection properties page and check the Set Statistics Time check box. Programmers can use these results
to compare different strategies to retrieve the same results. Listing B.2 shows a typical output of Statistics
time.

Listing B.2 Execution of a Query with Information About Time Statistics

USE Northwind
GO

SET STATISTICS TIME ON

SELECT Products.ProductID,
Products.ProductName,

681
Microsoft SQL Server 2000 Programming by Example

[Order Details].OrderID,
[Order Details].Quantity,
[Order Details].UnitPrice,
[Order Details].Discount
FROM Products
JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE Products.ProductID = 50

SET STATISTICS TIME OFF

SQL Server Execution Times:

CPU time = 0 ms, elapsed time = 0 ms.
SQL Server parse and compile time:
CPU time = 0 ms, elapsed time = 0 ms.

SQL Server Execution Times:

CPU time = 0 ms, elapsed time = 0 ms.
ProductID ProductName OrderID Quantity UnitPrice Discount
---------- ------------------ ----------- --------- ---------- ------------
50.00 Valkoinen suklaa 10,350.00 15.00 £13.00 0.10
50.00 Valkoinen suklaa 10,383.00 15.00 £13.00 0.00
50.00 Valkoinen suklaa 10,429.00 40.00 £13.00 0.00
50.00 Valkoinen suklaa 10,465.00 25.00 £13.00 0.00
50.00 Valkoinen suklaa 10,637.00 25.00 £16.25 5.0000001E-2
50.00 Valkoinen suklaa 10,729.00 40.00 £16.25 0.00
50.00 Valkoinen suklaa 10,751.00 20.00 £16.25 0.10
50.00 Valkoinen suklaa 10,920.00 24.00 £16.25 0.00
50.00 Valkoinen suklaa 10,948.00 9.00 £16.25 0.00
50.00 Valkoinen suklaa 11,072.00 22.00 £16.25 0.00

(10 row(s) affected)

SQL Server Execution Times:

CPU time = 10 ms, elapsed time = 151 ms.
Execution time depends mainly on the amount of data to access. It is possible to obtain this information by
executing the SET STATISTICS IO ON statement, or selecting the Set Statistics IO check box in the
Connection Properties form. Listing B.3 shows an example using the same query from Listing B.2 as a
basis.

Listing B.3 Execution of a Query with Information About Data Access Statistics

USE Northwind

682
 Appendix B. Using SQL Query Analyzer

GO

SET STATISTICS IO ON

SELECT Products.ProductID,
Products.ProductName,
[Order Details].OrderID,
[Order Details].Quantity,
[Order Details].UnitPrice,
[Order Details].Discount
FROM Products
JOIN [Order Details]
ON Products.ProductID = [Order Details].ProductID
WHERE Products.ProductID = 50

SET STATISTICS IO OFF

ProductID ProductName OrderID Quantity UnitPrice Discount

---------- ------------------ ----------- --------- ---------- ------------
50.00 Valkoinen suklaa 10,350.00 15.00 £13.00 0.10
50.00 Valkoinen suklaa 10,383.00 15.00 £13.00 0.00
50.00 Valkoinen suklaa 10,429.00 40.00 £13.00 0.00
50.00 Valkoinen suklaa 10,465.00 25.00 £13.00 0.00
50.00 Valkoinen suklaa 10,637.00 25.00 £16.25 5.0000001E-2
50.00 Valkoinen suklaa 10,729.00 40.00 £16.25 0.00
50.00 Valkoinen suklaa 10,751.00 20.00 £16.25 0.10
50.00 Valkoinen suklaa 10,920.00 24.00 £16.25 0.00
50.00 Valkoinen suklaa 10,948.00 9.00 £16.25 0.00
50.00 Valkoinen suklaa 11,072.00 22.00 £16.25 0.00

(10 row(s) affected)

Table 'Order Details'. Scan count 1, logical reads 22, physical reads 10,
read-ahead reads 0.
Table 'Products'. Scan count 1, logical reads 2, physical reads 2,
read-ahead reads 0.
As you can see in the results of Listing B.3, to execute the query, SQL Server had to

• Scan the table [Order Details] once, reading 10 pages from disk and reading data pages 22 times in
total, so it had to read some of the pages several times.
• Scan the Products table once, reading 2 pages from disk and reading data from data pages 2 times in
total.

Analyzing Query Execution Plan

According to the comments of the previous section, there is information on how long it takes to execute a
query and how much data SQL Server must access to achieve the expected result. However, how does SQL
Server actually execute this query?
SQL Query Analyzer provides a graphical tool to show the query plan selected by SQL Server to execute the
query. To see this query plan on query execution, you can select the Show Execution Time menu, or press

683
Microsoft SQL Server 2000 Programming by Example

Ctrl+K. To show the estimated query plan without actually executing the query, click the Display Estimated
Execution Plan icon in the toolbar, or press Ctrl+L.
Figure B.22 shows the graphical execution plan for the query from Listing B.2.
Note

Query optimization is both an art and a science. Showing in detail how to master the analytical
capabilities of Query Analyzer to optimize queries is not the purpose of this book.

Extra information is available in Books Online in the topic "Graphically Displaying the Execution
Plan Using SQL Query Analyzer."

Figure B.22. SQL Server uses the graphical execution plan to show how the query is executed.

Managing Indexes from Query Analyzer

Indexes have an important role to play in the execution of queries. SQL Query Analyzer provides index
management from the same environment. The menu Tools, Manage Indexes gives you access to the Manage
Indexes form. Figure B.23 shows this form.

Figure B.23. Managing indexes from SQL Query Analyzer facilitates query tuning and optimization.

684
 Appendix B. Using SQL Query Analyzer

From the graphical query plan it is possible to manage indexes for a given table by right -clicking the table
object and selecting Manage Indexes from the table's context menu. This is just another way to arrive at the
Manage Indexes form.
Indexes are discussed in detail in Chapter 6, "Optimizing Access to Data: Indexes."

Working with the Transact-SQL Debugger

Transact-SQL is a rich and powerful programming language designed to be used on databases. Visual Basic
programmers appreciate the debugging features of this popular language, and for SQL Server–stored
procedures they benefit from a Transact-SQL debugger add-in. However, some SQL Server programmers
don't use Visual Studio and they would appreciate having a full-featured Transact-SQL debugger to test the
execution of stored procedures, triggers, and user-defined functions. This new version of SQL Query Analyzer
integrates a Transact-SQL debugger.
To test how Transact-SQL Debugger works, create a user-defined function and a stored procedure as in
Listing B.4.

Listing B.4 The User-Defined Function and the Stored Procedure You Can Debug Using the Transact-
SQL Debugger

CREATE FUNCTION testdebug(@a int, @b int)

RETURNS int

685
Microsoft SQL Server 2000 Programming by Example

AS
BEGIN
SET @a = @a * 2
SET @b = @b * 4

DECLARE @c int

SET @c = @a * @b

SET @c = @c + @a

SET @c = @c + @b

RETURN @c
END
GO

CREATE PROCEDURE spTestdebug

AS
DECLARE @a0 int, @b0 int
DECLARE @c0 int
SET @a0 = 1
SET @b0 = 10

SET @c0 = dbo.testdebug(@a0, @b0)

SELECT @c0
GO
You will now use Object Search to search for the procedure you just created. Figure B.24 shows the Object
Search form and the results, as well as the context menu for the selected stored procedure.

Figure B.24. Searching for objects is easy using the SQL Query Analyzer Object Search.

Select Debug from the context menu to open the Transact-SQL Debugger window, where you can trace the
execution of this stored procedure step by step by pressing F11. When the stored procedure execution arrives
at the step where the user-defined function must be evaluated, the debugger goes to the user-defined function

686
 Appendix B. Using SQL Query Analyzer

code and opens the code to start the debug process there as well. Figure B.25 shows the Debugger window
while debugging the user-defined function TestDebug. The following are the components of the window:

• The Debugger toolbar controls the procedure's execution.

• The Source Code window uses a yellow arrow to indicate the next step to execute and a red bullet to
indicate the breakpoints.
• The Local Variables window shows local variables, including parameters, available in the current
scope. In this window, you can change the values of the variables.
• The Globals window shows the value of several system functions. By default in this window, you can
see the system functions @@connections and @@trancount, but it would be useful to add @@error
as well as any other useful system function.
• The Callstack window shows the logical sequence of procedure calls up to this execution point. As
you see in Figure B.25, this case shows two procedures: the Northwind.dbo.TestDebug user-defined
function and the Northwind.dbo.spTestDebug stored procedure. Double-click any of these procedures
and the debugger shows the selected procedure's code and a yellow arrow points to the local
execution point.
• The Results Text window shows any results your procedure might produce.

Figure B.25. Use the Transact-SQL Debugger to trace the execution of stored procedures, triggers,
and user-defined functions.

Tip

To debug the execution of triggers, create a stored procedure to modify the table, forcing the
trigger to be executed. When the debugger arrives at the DML statement, which fires the trigger, it
will jump to the trigger code to continue debugging.

Summary

In this chapter, you learned how to use SQL Query Analyzer. This is an important tool that helps you to design,
test, and tune a database in SQL Server.

687
Microsoft SQL Server 2000 Programming by Example

The Object Browser helps you identify the definition of any object, without using metadata stored procedures
or external tools. Using drag-and-drop helps you avoid misspellings.
You learned how to use the scripting facilities of Object Browser and Object Search. Now, you know how to
produce results in text, grid, or to a file, according to your needs.
As any client tool, you can manage connection settings that help you adjust the connection to the user
requirements.
Query Analyzer is a powerful analysis tool that helps you understand how SQL Server executes queries.
Graphical Query Plan and the Transact-SQL debugger are great tools that will help you during the
development and maintenance of a database system.
SQL Query Analyzer is the main tool that you use to test the examples provided in this book, and the main
client tool you use in the development environment as a database developer.
We tried to cover in this appendix the main uses of SQL Query Analyzer. However, if you still need more
information, Books Online contains a comprehensive description of the functionality of this powerful and useful
tool.
Delivered for Nenad Apostoloski Last updated on 1/18/2002
Swap Option Available: 2/24/2002 Microsoft SQL Server 2000 Programming by Example, © 2002 Que

688