CS301 Data Structures

___________________________________________________________________

Data Structure

Page 1 of 519

CS301 Data Structures

___________________________________________________________________
Data Structures............................................................................................................1
Lecture No. 01 ............................................................................................................3
Lecture No. 02 ..........................................................................................................12
Lecture No. 03 ..........................................................................................................21
Lecture No. 04 ..........................................................................................................33
Lecture No. 05 ..........................................................................................................47
Lecture No. 06 ..........................................................................................................57
Lecture No. 07 ..........................................................................................................64
Lecture No. 08 ..........................................................................................................72
Lecture No. 09 ..........................................................................................................84
Lecture No. 10 ..........................................................................................................97
Lecture No. 11 ........................................................................................................110
Lecture No. 12 ........................................................................................................128
Lecture No. 13 ........................................................................................................141
Lecture No. 14 ........................................................................................................149
Lecture No. 15 ........................................................................................................164
Lecture No. 16 ........................................................................................................177
Lecture No. 17 ........................................................................................................196
Lecture No. 18 ........................................................................................................210
Lecture No. 19 ........................................................................................................218
Lecture No. 20 ........................................................................................................227
Lecture No. 21 ........................................................................................................238
Lecture No. 22 ........................................................................................................247
Lecture No. 23 ........................................................................................................264
Lecture No. 24 ........................................................................................................280
Lecture No. 25 ........................................................................................................291
Lecture No. 26 ........................................................................................................306
Lecture No. 27 ........................................................................................................318
Lecture No. 28 ........................................................................................................332
Lecture No. 29 ........................................................................................................344
Lecture No. 30 ........................................................................................................359
Lecture No. 31 ........................................................................................................371
Lecture No. 32 ........................................................................................................381
Lecture No. 33 ........................................................................................................390
Lecture No. 34 ........................................................................................................398
Lecture No. 35 ........................................................................................................405
Lecture No. 36 ........................................................................................................418
Lecture No. 37 ........................................................................................................430
Lecture No. 38 ........................................................................................................438
Lecture No. 39 ........................................................................................................444
Lecture No. 40 ........................................................................................................454
Lecture No. 41 ........................................................................................................462
Lecture No. 42 ........................................................................................................473
Lecture No. 43 ........................................................................................................482
Lecture No. 44 ........................................................................................................488
Lecture No. 45 ........................................................................................................499

Page 2 of 519

CS301 Data Structures

___________________________________________________________________

Lecture No. 01
Reading Material
Data Structures and algorithm analysis in C++

Chapter. 3
3.1, 3.2, 3.2.1

Summary

Introduction to Data Structures

Selecting a Data Structure
Data Structure Philosophy
Goals of this Course
Array
List data structure

Welcome to the course of data structure. This is very important subject as the topics
covered in it will be encountered by you again and again in the future courses. Due to
its great applicability, this is usually called as the foundation course. You have
already studied Introduction to programming using C and C++ and used some data
structures. The focus of that course was on how to carry out programming with the
use of C and C++ languages besides the resolution of different problems. In this
course, we will continue problem solving and see that the organization of data in
some cases is of immense importance. Therefore, the data will be stored in a special
way so that the required result should be calculated as fast as possible.
Following are the goals of this course:

Prepare the students for (and is a pre-requisite for) the more advanced material
students will encounter in later courses.
Cover well-known data structures such as dynamic arrays, linked lists, stacks,
queues, trees and graphs.
Implement data structures in C++

You have already studied the dynamic arrays in the previous course. We will now
discuss linked lists, stacks, queues, trees and graphs and try to resolve the problems
with the help of these data structures. These structures will be implemented in C++
language. We will also do programming assignments to see the usage and importance
of these structures.

Information about Data Structure subject is available at: http://www.vu.edu.pk/ds.

Introduction to Data Structures

Lets discuss why we need data structures and what sort of problems can be solved
Page 3 of 519

CS301 Data Structures

___________________________________________________________________
with their use. Data structures help us to organize the data in the computer, resulting
in more efficient programs. An efficient program executes faster and helps minimize
the usage of resources like memory, disk. Computers are getting more powerful with
the passage of time with the increase in CPU speed in GHz, availability of faster
network and the maximization of disk space. Therefore people have started solving
more and more complex problems. As computer applications are becoming complex,
so there is need for more resources. This does not mean that we should buy a new
computer to make the application execute faster. Our effort should be to ensue that the
solution is achieved with the help of programming, data structures and algorithm.
What does organizing the data mean? It means that the data should be arranged in a
way that it is easily accessible. The data is inside the computer and we want to see it.
We may also perform some calculations on it. Suppose the data contains some
numbers and the programmer wants to calculate the average, standard deviation etc.
May be we have a list of names and want to search a particular name in it. To solve
such problems, data structures and algorithm are used. Sometimes you may realize
that the application is too slow and taking more time. There are chances that it may be
due to the data structure used, not due to the CPU speed and memory. We will see
such examples. In the assignments, you will also check whether the data structure in
the program is beneficial or not. You may have two data structures and try to decide
which one is more suitable for the resolution of the problem.
As discussed earlier, a solution is said to be efficient if it solves the problem within its
resource constraints. What does it mean? In the computer, we have hard disk, memory
and other hardware. Secondly we have time. Suppose you have some program that
solves the problem but takes two months. It will be of no use. Usually, you dont have
this much time and cannot wait for two months. Suppose the data is too huge to be
stored in disk. Here we have also the problem of resources. This means that we have
to write programs considering the resources to achieve some solution as soon as
possible. There is always cost associated with these resources. We may need a faster
and better CPU which can be purchased. Sometimes, we may need to buy memory.
As long as data structures and programs are concerned, you have to invest your own
time for this. While working in a company, you will be paid for this. All these
requirements including computer, your time and computer time will decide that the
solution you have provided is suitable or not. If its advantages are not obtained, then
either program or computer is not good.
So the purchase of a faster computer, while studying this course, does not necessarily
help us in the resolution of the problem. In the course of Computer Architecture
you will see how the more efficient solutions can be prepared with the hardware. In
this course, we will use the software i.e. data structures, algorithms and the recipes
through which the computer problems may be resolved with a faster solution.

Selecting a Data Structure

How can we select the data structure needed to solve a problem? You have already
studied where to use array and the size of array and when and where to use the
pointers etc. First of all, we have to analyze the problem to determine the resource
constraints that a solution must meet. Suppose, the data is so huge i.e. in Gega bytes

Page 4 of 519

CS301 Data Structures

___________________________________________________________________
(GBs) while the disc space available with us is just 200 Mega bytes. This problem can
not be solved with programming. Rather, we will have to buy a new disk.
Secondly, it is necessary to determine the basic operations that must be supported.
Quantify the resource constraints for each operation. What does it mean? Suppose you
have to insert the data in the computer or database and have to search some data item.
Lets take the example of telephone directory. Suppose there are eight million names
in the directory. Now someone asks you about the name of some particular person.
You want that this query should be answered as soon as possible. You may add or
delete some data. It will be advisable to consider all these operations when you select
some data structure.
Finally select the data structure that meets these requirements the maximum. Without,
sufficient experience, it will be difficult to determine which one is the best data
structure. We can get the help from internet, books or from someone whom you know
for already getting the problems solved. We may find a similar example and try to use
it. After this course, you will be familiar with the data structures and algorithms that
are used to solve the computer problems.
Now you have selected the data structure. Suppose a programmer has inserted some
data and wants to insert more data. This data will be inserted in the beginning of the
existing data, or in the middle or in the end of the data. Lets talk about the arrays and
suppose you have an array of size hundred. Data may be lying in the first fifty
locations of this array. Now you have to insert data in the start of this array. What will
you do? You have to move the existing data (fifty locations) to the right so that we get
space to insert new data. Other way round, there is no space in the start. Suppose you
have to insert the data at 25th location. For this purpose, it is better to move the data
from 26th to 50th locations; otherwise we will not have space to insert this new data at
25th location.
Now we have to see whether the data can be deleted or not. Suppose you are asked to
delete the data at 27th position. How can we do that? What will we do with the space
created at 27th position?
Thirdly, is all the data processed in some well-defined order or random access
allowed? Again take the example of arrays. We can get the data from 0th position and
traverse the array till its 50th position. Suppose we want to get the data, at first from
50th location and then from 13th. It means that there is no order or sequence. We want
to access the data randomly. Random access means that we cant say what will be the
next position to get the data or insert the data.

Data Structure Philosophy

Lets talk about the philosophy of data structure. Each data structure has costs and
benefits. Any data structure used in your program will have some benefits. For this,
you have to pay price. That can be computer resources or the time. Also keep in mind
that you are solving this problem for some client. If the program is not efficient, the
client will not buy it.
Page 5 of 519

CS301 Data Structures

___________________________________________________________________
In rare cases, a data structure may be better than another one in all situations. It means
that you may think that the array is good enough for all the problems. Yet this is not
necessary. In different situations, different data structures will be suitable. Sometimes
you will realize that two different data structures are suitable for the problem. In such
a case, you have to choose the one that is more appropriate. An important skill this
course is going to lend to the students is use the data structure according to the
situation. You will learn the programming in a way that it will be possible to replace
the one data structure with the other one if it does not prove suitable. We will replace
the data structure so that the rest of the program is not affected. You will also have to
attain this skill as a good programmer.
There are three basic things associated with data structures. A data structure requires:
space for each data item it stores
time to perform each basic operation
programming effort

Goals of this Course

Reinforce the concept that costs and benefits exist for every data structure. We will
learn this with practice.
Learn the commonly used data structures. These form a programmer's basic data
structure toolkit. In the previous course, you have learned how to form a loop,
functions, use of arrays, classes and how to write programs for different problems. In
this course, you will make use of data structures and have a feeling that there is bag
full of different data structures. In case of some problem, you will get a data structure
from the toolkit and use some suitable data structure.
Understand how to measure the cost of a data structure or program. These techniques
also allow you to judge the merits of new data structures that you or others might
develop. At times, you may have two suitable data structures for some problem. These
can be tried one by one to adjudge which one is better one. How can you decide
which data structure is better than other. Firstly, a programmer can do it by writing
two programs using different data structure while solving the same problem. Now
execute both data structures. One gives the result before the other. The data structure
that gives results first is better than the other one. But sometimes, the data grows too
large in the problem. Suppose we want to solve some problem having names and the
data of names grows to10 lakhs (one million). Now when you run both programs, the
second program runs faster. What does it mean? Is the data structure used in program
one not correct? This is not true. The size of the data, being manipulated in the
program can grow or shrink. You will also see that some data structures are good for
small data while the others may suit to huge data. But the problem is how can we
determine that the data in future will increase or decrease. We should have some way
to take decision in this regard. In this course we will do some mathematical analysis
and see which data structure is better one.

Page 6 of 519

CS301 Data Structures

___________________________________________________________________

Arrays
You have already studied about arrays and are well-versed with the techniques to
utilize these data structures. Here we will discuss how arrays can be used to solve
computer problems. Consider the following program:
main( int argc, char** argv )
{
int x[6];
int j;
for(j = 0; j < 6; j++)
x[j] = 2 * j;
}
We have declared an int array of six elements and initialized it in the loop.
Lets revise some of the array concepts. The declaration of array is as int x[6]; or
float x[6]; or double x[6]; You have already done these in your programming
assignments. An array is collection of cells of the same type. In the above program,
we have array x of type int of six elements. We can only store integers in this array.
We cannot put int in first location, float in second location and double in third
location. What is x? x is a name of collection of items. Its individual items are
numbered from zero to one less than array size. To access a cell, use the array name
and an index as under:
x[0], x[1], x[2], x[3], x[4], x[5]
To manipulate the first element, we will use the index zero as x[0] and so on. The
arrays look like in the memory as follows:

X[0]
Array cells are
contiguous in
computer memory

X[1]
X[2]
X[3]
X[4]
X[5]

Array occupies contiguous memory area in the computer. In case of the above
example, if some location is assigned to x[0], the next location can not contain data
other than x[1]. The computer memory can be thought of as an array. It is a very big
array. Suppose a computer has memory of 2MB, you can think it as an array of size 2
million and the size of each item is 32 bits. You will study in detail about it in the
computer organization, and Assembly language courses. In this array, we will put our

Page 7 of 519

CS301 Data Structures

___________________________________________________________________
programs, data and other things.
In the above program, we have declared an array named x. x is an arrays name but
there is no variable x. x is not an lvalue. If some variable can be written on the lefthand side of an assignment statement, this is lvalue variable. It means it has some
memory associated with it and some value can be assigned to it. For example, if we
have the code
int a, b; it can be written as b = 2; it means that put 2 in the
memory location named b. We can also write as a = b; it means whatever b has assign
it to a, that is a copy operation. If we write as a = 5; it means put the number 5 in the
memory location which is named as a. But we cannot write 2 = a; that is to put at
number 2 what ever the value of a is. Why cant we do that? Number 2 is a constant.
If we allow assignment to constants what will happen? Suppose a has the value
number 3. Now we assigned number 2 the number 3 i.e. all the number 2 will become
number 3 and the result of 2 + 2 will become 6. Therefore it is not allowed.
x is a name of array and not an lvalue. So it cannot be used on the left hand side in
an assignment statement. Consider the following statements
int x[6];
int n;
x[0] = 5; x[1] = 2;
x = 3;
//not allowed
x = a + b;
// not allowed
x = &n;
// not allowed
In the above code snippet, we have declared an array x of int. Now we can assign
values to the elements of x as x[0] = 5 or x[1] = 2 and so on. The last three
statements are not allowed. What does the statement x = 3; mean? As x is a name of
array and this statement is not clear, what we are trying to do here? Are we trying to
assign 3 to each element of the array? This statement is not clear. Resultantly, it can
not be allowed. The statement x = a + b is also not allowed. There is nothing wrong
with a + b. But we cannot assign the sum of values of a and b to x. In the statement x
= &n, we are trying to assign the memory address of n to x which is not allowed. The
reason is the name x is not lvalue and we cannot assign any value to it. For
understanding purposes, consider x as a constant. Its name or memory location can
not be changed. This is a collective name for six locations. We can access these
locations as x[0], x[1] up to x[5]. This is the way arrays are manipulated.
Sometimes, you would like to use an array data structure but may lack the information
about the size of the array at compile time. Take the example of telephone directory.
You have to store one lakh (100,000) names in an array. But you never know that the
number of entries may get double or decline in future. Similarly, you can not say that
the total population of the country is one crore (10 million) and declare an array of
one crore names. You can use one lakh locations now and remaining will be used as
the need arrives. But this is not a good way of using the computer resources. You
have declared a very big array while using a very small chunk of it. Thus the
remaining space goes waste which can, otherwise, be used by some other programs.
We will see what can be the possible solution of this problem?

Page 8 of 519

CS301 Data Structures

___________________________________________________________________
Suppose you need an integer array of size n after the execution of the program. We
have studied that if it is known at the execution of the program that an array of size 20
or 30 is needed, it is allocated dynamically. The programming statement is as follows:
int* y = new int[20];
It means we are requesting computer to find twenty memory locations. On finding it,
the computer will give the address of first location to the programmer which will be
stored in y. Arrays locations are contiguous i.e. these are adjacent. These twenty
locations will be contiguous, meaning that they will be neighbors to each other. Now
y has become an array and we can say y[0] =1 or y[5] = 15. Here y is an lvalue.
Being a pointer, it is a variable where we can store the address of some variable.
When we said int* y = new int[20]; the new returns the memory address of first of
the twenty locations and we store that address into y. As y is a pointer variable so it
can be used on the left-hand side. We can write it as:
y = &x[0];
In the above statement, we get the address of the fist location of the array x and store
it in y. As y is lvalue, so it can be used on left hand side. This means that the above
statement is correct.
y = x;
Similarly, the statement y = x is also correct. x is an array of six elements that holds
the address of the first element. But we cannot change this address. However we can
get that address and store it in some other variable. As y is a pointer variable and
lvalue so the above operation is legal. We have dynamically allocated the memory for
the array. This memory, after the use, can be released so that other programs can use
it. We can use the delete keyword to release the memory. The syntax is:
delete[ ] y;
We are releasing the memory, making it available for use by other programs. We will
not do it in case of x array, as new was not used for its creation. So it is not our
responsibility to delete x.

List data structure

This is a new data structure for you. The List data structure is among the most generic
of data structures. In daily life, we use shopping list, groceries list, list of people to
invite to a dinner, list of presents to give etc. In this course, we will see how we use
lists in programming.
A list is the collection of items of the same type (grocery items, integers, names). The
data in arrays are also of same type. When we say int x[6]; it means that only the
integers can be stored in it. The same is true for list. The data which we store in list
should be of same nature. The items, or elements of the list, are stored in some
particular order. What does this mean? Suppose in the list, you have the fruit first
which are also in some order. You may have names in some alphabetical order i.e. the

Page 9 of 519

CS301 Data Structures

___________________________________________________________________
names which starts with A should come first followed by the name starting with B and
so on. The order will be reserved when you enter data in the list.
It is possible to insert new elements at various positions in the list and remove any
element of the list. You have done the same thing while dealing with arrays. You
enter the data in the array, delete data from the array. Sometimes the array size grows
and at times, it is reduced. We will do this with the lists too.
List is a set of elements in a linear order. Suppose we have four names a1, a2, a3, a4
and their order is as (a3, a1, a2, a4) i.e. a3, is the first element, a1 is the second
element, and so on. We want to maintain that order in the list when data is stored in
the list. We dont want to disturb this order. The order is important here; this is not
just a random collection of elements but an ordered one. Sometimes, this order is due
to sorting i.e. the things that start with A come first. At occasions, the order may be
due to the importance of the data items. We will discuss this in detail while dealing
with the examples.
Now we will see what kind of operations a programmer performs with a list data
structure. Following long list of operations may help you understand the things in a
comprehensive manner.
Operation Name
createList()
copy()
clear();
insert(X, ?)
remove(?)
get(?)
update(X, ?)
find(X)
length()

Description
Create a new list (presumably empty)
Set one list to be a copy of another
Clear a list (remove all elements)
Insert element X at a particular position in the list
Remove element at some position in the list
Get element at a given position
Replace the element at a given position with X
Determine if the element X is in the list
Returns the length of the list.

createList() is a function which creates a new list. For example to create an array, we
use int x[6] or int* y = new int[20]; we need similar functionality in lists too. The
copy() function will create a copy of a list. The function clear() will remove all the
elements from a list. We want to insert a new element in the list, we also have to tell
where to put it in the list. For this purpose insert(X, position) function is used.
Similarly the function remove(position) will remove the element at position. To get an
element from the list get(position) function is used which will return the element at
position. To replace an element in the list at some position the function update(X,
position) is used. The function find(X) will search X in the list. The function length()
tells us about the number of elements in the list.
We need to know what is meant by particular position we have used ? for this in
the above table. There are two possibilities:

Use the actual index of element: i.e. insert it after element 3, get element
number 6. This approach is used with arrays
Use a current marker or pointer to refer to a particular position in the list.

Page 10 of 519

CS301 Data Structures

___________________________________________________________________
The first option is used in the data structures like arrays. When we have to manipulate
the arrays, we use index like x[3], x[6]. In the second option we do not use first,
second etc for position but say wherever is the current pointer. Just think of a pointer
in the list that we can move forward or backward. When we say get, insert or update
while using the current pointer, it means that wherever is the current pointer, get data
from that position, insert data after that position or update the data at that position. In
this case, we need not to use numbers. But it is our responsibility that current pointer
is used in a proper way.
If we use the current marker, the following four methods would be useful:
Functions Description
start()
Moves the current pointer to the very first element
tail()
Moves the current pointer to the very last element
next()
Move the current position forward one element
back()
Move the current position backward one element
In the next lecture, we will discuss the implementation of the list data structure and
write the functions discussed today, in C++ language.

Data Structures
Lecture No. 02
Reading Material

Page 11 of 519

CS301 Data Structures

___________________________________________________________________
Data Structures and Algorithm Analysis in C++

Chapter. 3
3.1, 3.2, 3.2.1, 3.2.2

Summary
1)

2)
3)
4)

List Implementation
add Method
next Method
remove Method
find Method
Other Methods
Analysis Of Array List
List Using Linked Memory
Linked List

Today, we will discuss the concept of list operations. You may have a fair idea of
start operation that sets the current pointer to the first element of the list while the
tail operation moves the current pointer to the last element of the list. In the previous
lecture, we discussed the operation next that moves the current pointer one element
forward. Similarly, there is the back operation which moves the current pointer one
element backward.

List Implementation
Now we will see what the implementation of the list is and how one can create a list
in C++. After designing the interface for the list, it is advisable to know how to
implement that interface. Suppose we want to create a list of integers. For this
purpose, the methods of the list can be implemented with the use of an array inside.
For example, the list of integers (2, 6, 8, 7, 1) can be represented in the following
manner where the current position is 3.
A

current

size

In this case, we start the index of the array from 1 just for simplification against the
usual practice in which the index of an array starts from zero in C++. It is not
necessary to always start the indexing from zero. Sometimes, it is required to start the
indexing from 1. For this, we leave the zeroth position and start using the array from
index 1 that is actually the second position. Suppose we have to store the numbers
from 1 to 6 in the array. We take an array of 7 elements and put the numbers from the
index 1. Thus there is a correspondence between index and the numbers stored in it.
This is not very useful. So, it does not justify the non-use of zeroth position of the
array out-rightly. However for simplification purposes, it is good to use the index
from 1.
add Method
Now we will talk about adding an element to the list. Suppose there is a call to add an
Page 12 of 519

CS301 Data Structures

___________________________________________________________________
element in the list i.e. add(9). As we said earlier that the current position is 3, so by
adding the element 9 to the list, the new list will be (2, 6, 8, 9, 7, 1).
To add the new element (9) to the list at the current position, at first, we have to make
space for this element. For this purpose, we shift every element on the right of 8 (the
current position) to one place on the right. Thus after creating the space for new
element at position 4, the array can be represented as
A

7
4

current

size

Now in the second step, we put the element 9 at the empty space i.e. position 4. Thus
the array will attain the following shape. The figure shows the elements in the array in
the same order as stored in the list.
A

current

size

We have moved the current position to 4 while increasing the size to 6. The size
shows that the elements in the list. Where as the size of the array is different that we
have defined already to a fixed length, which may be 100, 200 or even greater.
next Method
Now lets see another method, called next. We have talked that the next method
moves the current position one position forward. In this method, we do not add a new
element to the list but simply move the pointer one element ahead. This method is
required while employing the list in our program and manipulating it according to the
requirement. There is also an array to store the list in it. We also have two variablescurrent and size to store the position of current pointer and the number of elements in
the list. By looking on the values of these variables, we can find the state of the list
i.e. how many elements are in the list and at what position the current pointer is.
The method next is used to know about the boundary conditions of the list i.e. the
array being used by us to implement the list. To understand the boundary conditions,
we can take the example of an array of size 100 to implement the list. Here, 100
elements are added to the array. Lets see what happens when we want to add 101st
element to the array? We used to move the current position by next method and
reached the 100th position. Now, in case of moving the pointer to the next position
(i.e. 101st), there will be an error as the size of the array is 100, having no position
after this point. Similarly if we move the pointer backward and reach at the first
position regardless that the index is 0 or 1. But what will happen if we want to move
backward from the first position? These situations are known as boundary conditions
and need attention during the process of writing programs when we write the code to
use the list. We will take care of these things while implementing the list in C++
programs.
remove Method
We have seen that the add method adds an element in the list. Now we are going to
discuss the remove method. The remove method removes the element residing at the
current position. The removal of the element will be carried out as follows. Suppose
Page 13 of 519

CS301 Data Structures

___________________________________________________________________
there are 6 elements (2, 6, 8, 9, 7, 1) in the list. The current pointer is pointing to the
position 5 that has the value 7. We remove the element, making the current position
empty. The size of the list will become 5. This is represented in the following figure.

current

size

6
5

We fill in the blank position left by the removal of 7 by shifting the values on the
right of position 5 to the left by one space. This means that we shift the remaining
elements on the right hand side of the current position one place to the left so that the
element next to the removed element (i.e. 1) takes its place (the fifth position) and
becomes the current position element. We do not change the current pointer that is
still pointing to the position 5. Thus the current pointer remains pointing to the
position 5 despite the fact that there is now element 1 at this place instead of 7. Thus
in the remove method, when we remove an element, the element next to it on the right
hand side comes at its place and the remaining are also shifted one place to the right.
This step is represented by the following figure.

current

size

find Method
Now lets talk about a function, used to find a specific element in the array. The find
(x) function is used to find a specific element in the array. We pass the element, which
is to be found, as an argument to the find function. This function then traverses the
array until the specific element is found. If the element is found, this function sets the
current position to it and returns 1 i.e. true. On the other hand, if the element is not
found, the function returns 0 i.e. false. This indicates that the element was not found.
Following is the code of this find(x) function in C++.
int find (int x)
{
int j ;
for (j = 1; j < size + 1; j++ )
if (A[j] == x )
break ;
if ( j < size + 1)
// x is found
{
current = j ;
//current points to the position where x found
return 1 ;
// return true
}
return 0 ;
//return false, x is not found
}
In the above code, we execute a for loop to traverse the array. The number of
Page 14 of 519

CS301 Data Structures

___________________________________________________________________
execution of this loop is equal to the size of the list. This for loop gets terminated
when the value of loop variable (j) increases from the size of the list. However we
terminate the loop with the break statement if the element is found at a position.
When the control comes out from the loop, we check the value of j. If the value of j is
less than the size of the array, it means that the loop was terminated by the break
statement. We use the break statement when we find the required element (x) in the
list. The execution of break statement shows that the required element was found at
the position equal to the value of j. So the program sets the current position to j and
comes out the function by returning 1 (i.e. true). If the value of j is greater than the
size of the array, it means that the whole array has traversed and the required element
is not found. So we simply return 0 (i.e. false) and come out of the function.
Other Methods
There are some other methods to implement the list using an array. These methods are
very simple, which perform their task just in one step (i.e. in one statement). There is
a get() method , used to get the element from the current position in the array. The
syntax of this function is of one line and is as under
return A[current] ;
This statement returns the element to which the current is pointing to (i.e. the current
position) in the list A.
Another function is update(x). This method is used to change (set) the value at the
current position. A value is passed to this method as an argument. It puts that value at
the current position. The following statement in this method carries out this process.
A [current] = x ;
Then there is a method length( ).This method returns the size of the list. The syntax of
this method is
return size ;
You may notice here that we are returning the size of the list and not the size of the
array being used internally to implement the list. This size is the number of the
elements of the list, stored in the array.
The back() method decreases the value of variable current by 1. In other words, it
moves the current position one element backward. This is done by writing the
statement.
current -- ;
The -- is a decrement operator in C++ that decreases the value of the operand by one.
The above statement can also be written as
current = current -1 ;
The start() method sets the current position to the first element of the list. We know
that the index of the array starts from 0 but we use the index 1 for the starting

Page 15 of 519

CS301 Data Structures

___________________________________________________________________
position. We do not use the index zero. So we set the current position to the first
element by writing
current = 1 ;
Similarly, the end() method sets the current position to the last element of the list i.e.
size. So we write
current = size ;

Analysis of Array List

Now we analyze the implementation of the list while using an array internally. We
analyze different methods used for the implementation of the list. We try to see the
level upto which these are efficient in terms of CPUs time consumption. Time is the
major factor to see the efficiency of a program.
Add
First of all, we have talked about the add method. When we add an element to the list,
every element is moved to the right of the current position to make space for the new
element. So, if the current position is the start of the list and we want to add an
element in the beginning, we have to shift all the elements of the list to the right one
place. This is the worst case of adding an element to the list. Suppose if the size of the
list is 10000 or 20000, we have to do the shift operation for all of these 10000 or
20000 elements. Normally, it is done by shifting of elements with the use of a for
loop. This operation takes much time of the CPU and thus it is not a good practice to
add an element at the beginning of a list. On the other hand, if we add an element at
the end of the list, it can be done by carrying out no shift operation. It is the best
case of adding an element to the list. However, normally we may have to move half of
the elements. The usage of add method is the matter warranting special care at the
time of implementation of the list in our program. To provide the interface of the list,
we just define these methods.
Remove
When we remove an element at the current position in the list, its space gets empty.
The current pointer remains at the same position. To fill this space, we shift the
elements on the right of this empty space one place to the left. If we remove an
element from the beginning of the list, then we have to shift the entire remaining
elements one place to the left. Suppose there is a large number of elements, say 10000
or 20000, in the list. We remove the first element from the list. Now to fill this space,
the remaining elements are shifted (that is a large number). Shifting such a large
number of elements is time consuming process. The CPU takes time to execute the for
loop that performs this shift operation. Thus to remove an element at the beginning of
the list is the worst case of remove method. However it is very easy to remove an
element at the end of the list. In average cases of the remove method we expect to
shift half of the elements. This average does not mean that in most of the cases, you
will have to shift half the elements. It is just the average. We may have to shift all the
elements in one operation (if we remove at the beginning) and in the second
operation, we have to shift no element (if we remove at the end). Similarly, in certain
operations, we have to shift just 10, 15 elements.

Page 16 of 519

CS301 Data Structures

___________________________________________________________________
Find
We have discussed that the find method takes an element and traverses the list to find
that element. The worst case of the find method is that it has to search the entire list
from beginning to end. So, it finds the element at the end of the array or the element is
not found. On average the find method searches at most half the list.
The other methods get (), length () etc are one-step methods. They carry out their
operation in one instruction. There is no need of any loop or other programming
structures to perform the task. The get() method gets the value from the specified
position just in one step. Similarly the update() method sets a value at the specific
position just in one-step. The length () method returns the value of the size of the list.
The other methods back(), start() and end() also perform their tasks only in one step.

List using Linked Memory

We have seen the implementation of the list with the use of an array. Now we will
discuss the implementation of the list while using linked memory. In an array, the
memory cells of the array are linked with each other. It means that the memory of the
array is contiguous. In an array, it is impossible that one element of the array is
located at a memory location while the other element is located somewhere far from it
in the memory. It is located in very next location in the memory. It is a property of the
array that its elements are placed together with one another in the memory. Moreover,
when we have declared the size of the array, it is not possible to increase or decrease
it during the execution of the program. If we need more elements to store in the array,
there is need of changing its size in the declaration. We have to compile the program
again before executing it. Now array will be of the new size. But what happens if we
again need to store more elements? We will change the code of our program to
change the declaration of the array while recompiling it.
Suppose we have used the dynamic memory allocation and created an array of 100
elements with the use of new operator. In case of need of 200 elements, we will
release this array and allocate a new array of 200 elements. Before releasing the
previous array, it will wise to copy its elements to the new array so that it does not
lose any information. Now this new array is in ready for use position. Thus the
procedure of creating a new array is not an easy task.
To avoid such problems, usually faced by the programmers while using an array,
there is need of using linked memory in which the various cells of memory, are not
located continuously. In this process, each cell of the memory not only contains the
value of the element but also the information where the next element of the list is
residing in the memory. It is not necessary that the next element is at the next location
in the memory. It may be anywhere in the memory. We have to keep a track of it.
Thus, in this way, the first element must explicitly have the information about the
location of the second element. Similarly, the second element must know where the
third element is located and the third should know the position of the fourth element
and so on. Thus, each cell (space) of the list will provide the value of the element
along with the information about where the next element is in the memory. This
information of the next element is accomplished by holding the memory address of
the next element. The memory address can be understood as the index of the array. As
in case of an array, we can access an element in the array by its index. Similarly, we

Page 17 of 519

CS301 Data Structures

___________________________________________________________________
can access a memory location by using its address, normally called memory address.

Linked List
For the utilization of the concept of linked memory, we usually define a structure,
called linked list. To form a linked list, at first, we define a node. A node comprises
two fields. i.e. the object field that holds the actual list element and the next that holds
the starting location of the next node.

object

next

A chain of these nodes forms a linked list. Now lets consider our previous list, used
with an array i.e. 2, 6, 8, 7, 1. Following is the figure which represents the list stored
as a linked list.
Head
2

size = 5

current
This diagram just represents the linked list. In the memory, different nodes may occur
at different locations but the next part of each node contains the address of the next
node. Thus it forms a chain of nodes which we call a linked list.
While using an array we knew that the array started from index 1that means the first
element of the list is at index 1. Similarly in the linked list we need to know the
starting point of the list. For this purpose, we have a pointer head that points to the
first node of the list. If we dont use head, it will not be possible to know the starting
position of the list. We also have a pointer current to point to the current node of the
list. We need this pointer to add or remove current node from the list. Here in the
linked list, the current is a pointer and not an index as we used while using an array.
The next field of the last node points to nothing .It is the end of the list. We place the
memory address NULL in the last node. NULL is an invalid address and is
inaccessible.
Now again consider the list 2, 6, 8, 7, 1. The previous figure represents this list as a
linked list. In this linked list, the head points to 2, 2 points to 6, 6 points to 8, 8 points
to 7 and 7 points to 1. Moreover we have the current position at element 8.
This linked list is stored in the memory. The following diagram depicts the process
through which this linked list is stored in the memory.

Page 18 of 519

CS301 Data Structures

___________________________________________________________________

current

1051

1052

1063

1053
1054

1055

1051

1056
1057
1058

7
1060

1059

head

1060

1061
1062

0
1054

1063
1064

8
1057

1065
We can see in the figure that each memory location has an address. Normally in
programming, we access the memory locations by some variable names. These
variable names are alias for these locations and are like labels that are put to these
memory locations. We use head and current variable names instead of using the
memory address in numbers for starting and the current nodes. In the figure, we see
that head is the name of the memory location 1062 and the name current is used for
the memory address 1053. The head holds the address 1054 and the element 2, the
first one in the list, is stored in the location 1054. Similarly current holds the address
1063 where the element 8 is stored which is our current position in the list. In the
diagram, two memory locations comprise a node. So we see that the location 1054
holds the element 2 while the next location 1055 holds the address of the memory
location (1051) where the next element of the list (i.e. 6) is stored. Similarly the next
part of the node that has value 6 holds the memory address of the location occupied
by the next element (i.e. 8) of the list. The other nodes are structured in a similar
fashion. Thus, by knowing the address of the next element we can traverse the whole
list.

Page 19 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 03
Reading Material
Data Structures and algorithm analysis in C++
Chapter. 3
3.2.2, 3.2.3, 3.2.5

Summary

Linked List inside Computer Memory

Linked List Operations
Linked List Using C++
Example Program

In the previous lectures, we used an array to construct a list data structure and
observed the limitation that array being of fixed size can only store a fixed number of
elements. Therefore, no more elements can be stored after the size of the array is
reached.
In order to resolve this, we adopted a new data structure called linked list. We started
discussing, how linked lists are stored in computer memory and how memory chains
are formed.

Page 20 of 519

CS301 Data Structures

___________________________________________________________________

Linked List inside Computer Memory

currenten

1052

1063

1053

1063

1054

1055

1051

1056

head
2

1051

1057

1058

1060

1059

headent

1060

1061

1062

1054

1063

1064

1057

1065
Fig 1. Linked list in memory
There are two parts of this figure. On the left is the linked list chain that is actually the
conceptual view of the linked list and on the right is the linked list inside the
computer memory. Right part is a snapshot of the computer memory with memory
addresses from 1051 to 1065. The head pointer is pointing to the first element in the
linked list. Note that head itself is present in the memory at address 1062. It is
actually a pointer containing the memory address 1054. Each node in the above linked
list has two parts i.e. the data part of the node and the pointer to the next node. The
first node of the linked list pointed by the head pointer is stored at memory address
1054. We can see the data element 2 present at that address. The second part of the
first node contains the memory address 1051. So the second linked lists node starts at
memory address 1051. We can use its pointer part to reach the third node of the list
and in this way, we can traverse the whole list. The last node contains 1 in its data
part and 0 in its pointer part. 0 indicates that it is not pointing to any node and it is the
last node of the linked list.

Linked List Operations

The linked list data structure provides operations to work on the nodes inside the list.
The first operation we are going to discuss here is to create a new node in the
memory. The Add(9) is used to create a new node in the memory at the current
position to hold 9. You must remember while working with arrays, to add an
Page 21 of 519

CS301 Data Structures

___________________________________________________________________
element at the current position that all the elements after the current position were
shifted to the right and then the element was added to the empty slot.
Here, we are talking about the internal representation of the list using linked list. Its
interface will remain the same as in case of arrays.
We can create a new node in the following manner in the add() operation of the linked
list with code in C++:
Node * newNode = new Node(9);
The first part of the statement that is on the left of the assignment is declaring a
variable pointer of type Node. It may also be written as Node * newNode. On the
right of this statement, the new operator is used to create a new Node object as new
Node(9). This is one way in C++ to create objects of classes. The name of the class is
provided with the new operator that causes the constructor of the class to be called.
The constructor of a class has the same name as the class and as this a function,
parameters can also be passed to it. In this case, the constructor of the Node class is
called and 9 is passed to it as an int parameter.
Hence, the whole statement means:
Call the constructor of the Node class and pass it 9 as a parameter. After
constructing the object in memory, give me the starting memory address of the object.
That address will be stored in the pointer variable newNode.
To create an object of int type in the same manner, we can write as:
int * i = new int ;
Previously, we used the same technique to allocate memory for an array of ints as:
int * i = new int [10] ;
Now after the node has been created, how the node is fit into the chain of the linked
list.
Fig 2. Insertion of new Node into the linked list
2

current

3
9
newNode
In the above figure, there is a linked list that contains five nodes with data elements as
2, 6, 8, 7 and 1. The current pointer is pointing to the node with element as 8. We
want to insert a new node with data element 9. This new node will be inserted at the
current position (the position where the current pointer is pointing to). This insertion
operation is performed in a step by step fashion.
- The first step is to point next pointer of the new node (with data element as 9) to
Page 22 of 519

CS301 Data Structures

___________________________________________________________________
the node with data element as 7.
The second step is to point the next pointer of the node with data element 8 to the
node the new node with data element 9.
- The third step is to change the current pointer to point to the new node.
-

Now, the updated linked list has nodes with data elements as 2, 6, 8, 9, 7 and 1. The
list size has become 6.

Linked List Using C++

/* The Node class */
class Node
{
public:
int get() { return object; };
void set(int object) { this->object = object; };
Node * getNext() { return nextNode; };
void setNext(Node * nextNode) { this->nextNode = nextNode; };
private:
int object;
Node * nextNode;
};
Whenever, we write a class, it begins with the word class followed by the class-name
and the body of the class enclosed within curly braces. In the body, we write its public
variables, methods and then private variables and methods, this is normally the
sequence.
If there is no code to write inside the constructor function of a class, we need not to
declare it ourselves as the compiler automatically creates a default constructor for us.
Similarly, if there is nothing to be done by the destructor of the class, it will be better
not to write it explicitly. Rather, the compiler writes it automatically. Remember, the
default constructor and destructor do nothing as these are the function without any
code statements inside.
Lets start with the data members first. These are given at the bottom of the class body
with the scope mentioned as private. These data members are actually two parts of a
linked lists node. First variable is object of type int, present there to store the data
part of the node. The second variable is nextNode, which is a pointer to an object of
type Node. It has the address of the next element of the linked list.
The very first public function given at the top is get(). We have written its code within
the class Node. It returns back the value of the variable object i.e. of the type of int.
When we write class in C++, normally, we make two files (.h and .cpp) for a class.
The .h file contains the declarations of public and private members of that class. The
public methods are essentially the interface of the class to be employed by the users of
this class. The .cpp file contains the implementation for the class methods that has the
actual code. This is usually the way that we write two files for one class. But this is

Page 23 of 519

CS301 Data Structures

___________________________________________________________________
not mandatory. In the code given above, we have only one file .cpp, instead of
separating into two files. As the class methods are very small, so their code is written
within the body of the class. This facilitates us in carrying on discussion. Thus instead
of talking about two files, we will only refer to one file. On the other hand, compiler
takes these functions differently that are called inline functions. The compiler replaces
the code of these inline functions wherever the call to them is made.
The second method in the above-mentioned class is set() that accepts a parameter of
type int while returning back nothing. The accepted parameter is assigned to the
internal data member object. Notice the use of this pointer while assigning the value
to the internal data member. It is used whenever an object wants to talk to its own
members.
The next method is getNext() which returns a pointer to an object of type Node lying
somewhere in the memory. It returns nextNode i.e. a pointer to an object of type
Node. As discussed above, nextNode contains the address of next node in the linked
list.
The last method of the class is setNext() that accepts a pointer of type Node, further
assigned to nextNode data member of the object. This method is used to connect the
next node of the linked list with the current object. It is passed an address of the next
node in the linked list.
Lets discuss a little bit about classes. A very good analogy of a class is a factory.
Think about a car factory. On the placement of order, it provides us with the number
of vehicles we ordered for. Similarly, you can see number of other factories in your
daily-life that manufacture the specific products.
Lets take this analogy in C++ language. Suppose, we want to make a factory in C++.
By the way, what is our Node class? It is actually a factory that creates nodes. When
we want to make a new node, a new operator is used. By using new operator with the
Node class, actually, we send an order to Node factory, to make as many as nodes for
us.
So we have a good analogy, to think about a class as a factory. The products that are
made by the factory have their own characteristics. For example, a car made by an
automobile factory has an engine, wheels, steering and seats etc. These variables
inside a class are called state variables. Now the kinds of operations this car can do
are the methods of its class. A car can be driven, engine can be started, gears can be
shifted and an accelerator can be pressed to run it faster.
Similarly, the Node class creates nodes, where every node has two-state variables i.e.
object and nextNode. We have already seen its operations in the above code. We use
new to create new object or an array of new objects, stored in memory.
Lets see the code below.
/* List class */
#include <stdlib.h>
#include "Node.cpp"
class List
{
public:

Page 24 of 519

CS301 Data Structures

___________________________________________________________________
// Constructor
List() {
headNode = new Node();
headNode->setNext(NULL);
currentNode = NULL;
size = 0;
}
We are creating a list factory here employed to create list objects. Remember the list
operations; add, remove, next, back and start etc. Lets see the above class declaration
code in detail.
There are two include statements at the start. The first line is to include a standard
library stdlib.h while the second line includes the Node class file Node.cpp. This Node
class is used to create nodes that form a List object. So this List factory will order
Node class to create new nodes. The List class itself carries out the chain management
of these Node objects.
We have written our own constructor of List class as the default constructor is not
sufficient enough to serve the purpose. The List constructor is parameterless. The very
first step it is doing internally is that it is asking Node class to create a new node and
assigning the starting address of the new Nodes object to the headNode data member.
In the second statement, we are calling setNext() method of the Node class for the
object pointed to by the headNode pointer. This call is to set the nextNode data
member to NULL, i.e. Nodes object pointed to by the headNode pointer is not
pointing to any further Node. The next statement is to set the currentNode pointer to
NULL. So at the moment, we have initialized the currentNode pointer to NULL that is
not pointing to any Node object. The next statement is to initialize the size data
member to 0 indicating that there is no node present in the list. All this processing is
done inside the constructor of List class, as we want all this done when a list object is
created. Considering the analogy of car factory, the constructor function can perform
certain tasks: The oil is poured into the engine, the tyres are filled-in with air etc.
Lets see the add method of the List class:
/* add() class method */
void add (int addObject)
{
1. Node * newNode = new Node();
2. newNode->set(addObject);
3. if( currentNode != NULL )
4. {
5.
newNode->setNext(currentNode->getNext());
6.
currentNode->setNext( newNode );
7.
lastCurrentNode = currentNode;
8.
currentNode = newNode;
9. }
10. else
11. {
12.
newNode->setNext(NULL);

Page 25 of 519

CS301 Data Structures

___________________________________________________________________
13.
headNode->setNext(newNode);
14.
lastCurrentNode = headNode;
15.
currentNode = newNode;
16. }
17. size ++;
}
The interface or signatures of add() method is similar to the one discussed in case of
an array. This method takes the object to be added as a parameter. The
implementation of this add() method is a bit longer as the method is being
implemented for linked list. In the first statement, a new Node object is created with
its address stored in the newNode pointer variable. The second statement is to call
set() method of the Node object pointed to by the newNode pointer. You can note the
way the method is called. A pointer variable is at the left most side then an arrow sign
(->), then the name of the method with appropriate arguments within parenthesis. It is
followed by the if-statement that checks the currentNode is not NULL to perform
certain operations inside the if-code block. Inside the if-statement, at line 5, the
nextNode pointer of the new node is being set to the nextNode of the object pointed to
by the currentNode pointer. In order to understand the statements given in this code
properly, consider the fig 2 above, where we added a node in the linked list. We have
done step 1 at line5. At line 6, we are performing the second step by setting the
newNode in the nextNode pointer of the object pointed to by the currentNode. At line
7, we are saving the current position (address) of the currentNode pointer in the
pointer variable lastCurrentNode, which might be useful for backward traversing.
Although, the fig 1 (left part) indicates movement in one direction from left to right
but the lastCurrentNode pointer node can be used by the back() member function to
traverse one position back from right to left. At line 8, the currentNode pointer is
assigned the address of the object pointed to by newNode. This way, a new node is
added in already existent linked list.
Line 10 is start of the else part of if-statement. This is executed if the currentNode is
NULL. It means that there is no node present in the list previously and first node is
going to be added. At line 12, we are setting the nextNode pointer of the object
pointed to by newNode pointer. The nextNode is being set to NULL by calling the
setNext() method. Then at line 13, we point the head pointer (headNode) to this new
node pointed to by newNode pointer. Note that headNode is pointing to a node that is
there despite the fact that the size of the linked list is 0. Actually, we have allocated a
Node object for headNode pointer. Although, we dont need a Node object here, yet it
will be helpful when we perform other operations like remove() and find().
At line 14, the headNode address is being assigned to lastCurrentNode. At line 15,
currentNode pointer is assigned the address of newNode. At the end i.e. at line 17, the
size of the list is incremented by 1.

Page 26 of 519

CS301 Data Structures

___________________________________________________________________
headNod

List list;

size = 0
currentNod

headNod

list.add(2);

size = 1

lastCurrentNod
currentNod
list.add(6);

headNode

size = 2

lastCurrentNod
Fig 3. Add operation of linked list
Following is the crux of this add() operation :
Firstly, it will make a new node by calling Node class constructor. Insert the value
e.g. 2. of the node into the node by calling the set method. Now if the list already
exists (has some elements inside or its size is non-zero), it will insert the node after
the current position. If the list does not already exist, this node is added as the first
element inside the list.
Lets try to add few more elements into the above linked list in the figure. The
following are the lines of code to be executed to add nodes with values 8, 7 and 1 into
the linked list.
list.add(8); list.add(7);

list.add(1);
currentNode

headNode

size = 5

lastCurrentNode

Fig 4. More Nodes added into linked list

Now we will see the remaining methods of the linked list. The get() method of the

Page 27 of 519

CS301 Data Structures

___________________________________________________________________
List class is given below
/* get() class method */
int get()
{
if (currentNode != NULL)
return currentNode->get();
}
This method firstly confirms that the currentNode pointer is not NULL. If it is not
NULL, then it must be pointing to some Node object as inside the constructor of the
List class, we have initialized this pointer variable to NULL. That indicates that the
currentNode is NULL when there is no element inside the list. However, when a Node
object is added into it, it starts pointing to it. So, this get() returns the address of the
node pointed to by the currentNode pointer.
Further, we have another method given below:
/* next() class method */
bool next()
{
1. if (currentNode == NULL) return false;
2.
3. lastCurrentNode = currentNode;
4. currentNode = currentNode->getNext();
5. return true;
};
This is next() method, used to advance the currentNode pointer to the next node
inside the linked list. At line 1, the currentNode is being checked to confirm that there
are some elements present in the list to advance further. At line 1, the method is
returning false if there is no element present in the list. At line 3, it is storing the value
of the currentNode pointer into the lastCurrentNode. At line 4, currentNode is calling
the getNext() method to get the address of next node to be stored in the currentNode
pointer to advance the currentNode pointer to the next element. At line 5, it returns
true indicating the method is successful in moving to the next node.

Example Program
Given below is the full source code of the example program. You can copy, paste and
compile it right away. In order to understand the linked list concept fully, it is highly
desirable that you understand and practice with the below code.
#include <iostream.h>
#include <stdlib.h>
/* The Node class */
class Node
{
public:
int
get() { return object; };

Page 28 of 519

CS301 Data Structures

___________________________________________________________________
void set(int object) { this->object = object; };
Node * getNext() { return nextNode; };
void setNext(Node * nextNode) { this->nextNode = nextNode; };
private:
int object;
Node * nextNode;
};
/* The List class */
class List
{
public:
List();
void add (int addObject);
int get();
bool next();
friend void traverse(List list);
friend List addNodes();
private:
int
size;
Node * headNode;
Node * currentNode;
Node * lastCurrentNode;
};
/* Constructor */
List::List()
{
headNode = new Node();
headNode->setNext(NULL);
currentNode = NULL;
lastCurrentNode = NULL;
size = 0;
}
/* add() class method */
void List::add (int addObject)
{
Node * newNode = new Node();
newNode->set(addObject);
if( currentNode != NULL )
{
newNode->setNext(currentNode->getNext());
currentNode->setNext( newNode );
lastCurrentNode = currentNode;
currentNode = newNode;

Page 29 of 519

CS301 Data Structures

___________________________________________________________________
}
else
{
newNode->setNext(NULL);
headNode->setNext(newNode);
lastCurrentNode = headNode;
currentNode = newNode;
}
size ++;
}
/* get() class method */
int List::get()
{
if (currentNode != NULL)
return currentNode->get();
}
/* next() class method */
bool List::next()
{
if (currentNode == NULL) return false;
lastCurrentNode = currentNode;
currentNode = currentNode->getNext();
if (currentNode == NULL || size == 0)
return false;
else
return true;
}
/* Friend function to traverse linked list */
void traverse(List list)
{
Node* savedCurrentNode = list.currentNode;
list.currentNode = list.headNode;
for(int i = 1; list.next(); i++)
{
cout << "\n Element " << i << " " << list.get();
}
list.currentNode = savedCurrentNode;
}
/* Friend function to add Nodes into the list */
List addNodes()
{
List list;
list.add(2);

Page 30 of 519

CS301 Data Structures

___________________________________________________________________
list.add(6);
list.add(8);
list.add(7);
list.add(1);
cout << "\n List size = " << list.size <<'\n';
return list;
}
main()
{
List list = addNodes();
traverse(list);
}
The output of the example program is as follows:
List size = 5
Element
Element
Element
Element
Element

1
2
3
4
5

2
6
8
7
1

Page 31 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 04
Reading Material
Data Structures and algorithm analysis in C++
Chapter. 3
3.2.3, 3.2.4, 3.2.5

Summary

Methods of Linked List

Example of list usage
Analysis of Link List
Doubly-linked List
Circularly-linked lists
Josephus Problem

Methods of Linked List

In the previous lecture, we discussed the methods of linked list. These methods form
the interface of the link list. For further elucidation of these techniques, we will talk
about the start method that has the following code.
// position currentNode and lastCurrentNode at first element
void start() {
lastCurrentNode = headNode;
currentNode = headNode;
};
There are two statements in this method. We assign the value of headNode to both
lastCurrentNode and currentNode. These two pointers point at different nodes of the
list. Here we have pointed both of these pointers at the start of the list. On calling
some other method like next, these pointers will move forward. As we can move in
the singly-linked list in one direction, these pointers cannot go behind headNode.
We will now see how a node can be removed from the link list. We use the method
remove for this purpose.
void remove() {
if( currentNode != NULL && currentNode != headNode) {
(step 1) lastCurrentNode->setNext(currentNode->getNext());
(step 2) delete currentNode;
(step 3) currentNode = lastCurrentNode;

Page 32 of 519

CS301 Data Structures

___________________________________________________________________
(step 4) size--;
}
};
currentNode
headNode
6

Size = 5

lastCurrentNode
Suppose that the currentNode is pointing at the location that contains the value 6. A
request for the removal of the node is made. Resultantly, the node pointed by
currentNode should be removed. For this purpose, at first, the next pointer of the node
with value 2 (the node pointed by the lastCurrentNode pointer), that is before the
node with value 6, bypasses the node with value 6. It is, now pointing to the node
with value 8. The code of the first step is as:
lastCurrentNode->setNext(currentNode->getNext());
What does the statement currentNode->getNext() do? The currentNode is pointing to
the node with value 6 while the next of this node is pointing to the node with value 8.
That is the next pointer of node with value 6 contains the address of the node with
value 8. The statement lastCurrentNode->setNext(currentNode->getNext()) will set
the next pointer of the node pointed by the lastCurrentNode to the node with value 8.
So the next pointer of the node with value 2 is pointing to the node with value 8.
currentNode
headNode

Step1
2

Size = 5

lastCurrentNode

You see that the next pointer of the node having data element 2 contains the address
of the node having data element 8. The node with value 6 has been disconnected from
the chain while the node with value 2 is connected to the node with the value 8.
The code of the next step is:
delete currentNode;
You already know, in case of allocation of the memory with the help of the new

Page 33 of 519

CS301 Data Structures

___________________________________________________________________
keyword, the delete statement releases this memory which returns the memory to the
system. Pictorially it can be represented as:
currentNode
headNode

Step1
2

Size = 5

Step2
lastCurrentNode

In the next step, we have moved the currentNode to point the next node. The code is:
currentNode = lastCurrentNode;
In the fourth step, the size of the list has been reduced by 1 after the deletion of one
node i.e.
size--;
Step3
currentNode
headNode

Step1

Step4

Size = 4

Step2
lastCurrentNode

The next method is length() that simply returns the size of the list. The code is as
follows:
// returns the size of the list
int length()
{
return size;
};
The private data members of the list are:
private:
int size;
// contains the size of the list
Node *headNode;
// points to the first node of the list

Page 34 of 519

CS301 Data Structures

___________________________________________________________________
Node *currentNode, // current node
Node *lastCurrentNode; // last current node
The list class completed just now, can be termed as list factory. We have included all
the required methods in it. We may employ more methods if required. A programmer
can get the size of the list, add or remove nodes in it besides moving the pointers.

Example of list usage

Now lets see how we use the link list. Here is an example showing the use of list:
/* A simple example showing the use of link list */
#include <iostream>
#include <stdlib.h>
#include "List.cpp" // This contains the definition of List class
// main method
int main(int argc, char *argv[])
{
List list; // creating a list object
// adding values to the list
list.add(5);
list.add(13);
list.add(4);
list.add(8);
list.add(24);
list.add(48);
list.add(12);
// calling the start method of the list
list.start();
// printing all the elements of the list
while (list.next())
cout << "List Element: "<< list.get()<<endl;
}
The output of the program is:
List Element: 5
List Element: 13
List Element: 4
List Element: 8
List Element: 24
List Element: 48
List Element: 12

Page 35 of 519

CS301 Data Structures

___________________________________________________________________
Lets discuss the code of the above program. We have included the standard libraries
besides having the List.cpp file. Usually we do not include .cpp files. Rather, the .h
files are included. Whenever you write a class, two files will be created i.e. .h (header
file containing the interface of the class) and .cpp (implementation file). Here for the
sake of explanation, we have combined the two files into List.cpp file. At the start
of the main method, we have created a list object as:
List list;
Here the default constructor will be called. If you understand the concept of factory,
then it is not difficult to know that we have asked the List factory to create a List
object and named it as list. After creating the object, nodes have been added to it. We
have added the elements with data values 5, 13, 4, 8, 24, 48 and 12. Later, the start()
method of list is called that will position the currentNode and lastCurrentNode at the
start of the list. Now there is no need to worry about the implementation of the List.
Rather, we will use the interface of the List. So the start method will take us to the
start of the list and internally, it may be array or link list or some other
implementation. Then there is a while loop that calls the next() method of the List. It
moves the pointer ahead and returns a boolean value i.e. true or false. When we reach
at the end of the list, the next() method will return false. In the while loop we have a
cout statement that prints the value of the list elements, employing the get() method.
The loop will continue till the next() method returns true. When the pointers reach at
the end of the list the next() will return false. Here the loop will come to an end.
Please keep in mind that list is a very important data structure that will be used in the
entire programming courses.

Analysis of Link List

As stated earlier, we will be going to analyze each data structure. We will see whether
it is useful or not. We will see its cost and benefit with respect to time and memory.
Let us analyze the link list which we have created with the dynamic memory
allocation in a chain form.

add
For the addition purposes, we simply insert the new node after the current node.
So add is a one-step operation. We insert a new node after the current node in
the chain. For this, we have to change two or three pointers while changing the
values of some pointer variables. However, there is no need of traversing too
much in the list. In case of an array, if we have to add an element in the centre of
the array, the space for it is created at first. For this, all the elements that are after
the current pointer in the array, should be shifted one place to the right. Suppose if
we have to insert the element in the start of the array, all the elements to the right
one spot are shifted. However, for the link list, it is not something relevant. In link
lists, we can create a new node very easily where the current pointer is pointing.
We have to adjust two or three pointers. Its cost, in terms of CPU time or
computing time, is not much as compared to the one with the use of arrays.

remove
Remove is also a one-step operation. The node before and after the node to be

Page 36 of 519

CS301 Data Structures

___________________________________________________________________
removed is connected to each other. Update the current pointer. Then the node to
be removed is deleted. As a result, the node to be removed is deleted. Very little
work is needed in this case. If you compare it with arrays, for the deletion of an
element from the array, space is created. To fill this space, all the right elements
are shifted one spot left. If the array size is two thousand or three thousand, we
need to run a loop for all these elements to shift them to left.

find
The worst-case in find is that we may have to search the entire list. In find, we
have to search some particular element say x. If found, the currentNode pointer is
moved at that node. As there is no order in the list, we have to start search from
the beginning of the list. We have to check the value of each node and compare it
with x (value to be searched). If found, it returns true and points the currentNode
pointer at that node otherwise return false. Suppose that x is not in the list, in this
case, we have to search the list from start to end and return false. This is the worst
case scenario. Though time gets wasted, yet we find the answer that x is not in the
list. If we compare this with array, it will be the same. We dont know whether x
is in the array or not. So we have to search the complete array. In case of finding
it, we will remember that position and will return true. What is the average case? x
can be found at the first position , in the middle or at the end of the list. So on
average, we have to search half of the list.

back
In the back method, we move the current pointer one position back. Moving the
current pointer back, one requires traversing the list from the start until the node
whose next pointer points to current node. Our link list is singly linked list i.e. we
can move in one direction from start towards end. Suppose our currentNode
pointer and lastCurrentNode are somewhere in the middle of the list. Now we
want to move one node back. If we have the pointer of lastCurrentNode, it will be
easy. We will assign the value of lastCurrentNode to currentNode. But how can
we move the lastCurrentNode one step back. We dont have the pointer of
previous node. So the solution for this is to go at the start of the list and traverse
the list till the time you reach the node before the lastCurrentNode is pointing.
That will be the node whose next pointer contains the value lastCurrentNode. If
the currentNode and the lastCurrentNode are at the end of the list, we have to
traverse the whole list. Therefore back operation is not a one step operation. We
not only need a loop here but also require time.

Doubly-linked List
If you look at single link list, the chain is seen formed in a way that every node has a
field next that point to the next node. This continues till the last node where we set the
next to NULL i.e. the end of the list. There is a headNode pointer that points to the
start of the list. We have seen that moving forward is easy in single link list but going
back is difficult. For moving backward, we have to go at the start of the list and begin
from there. Do you need a list in which one has to move back or forward or at the
start or in the end very often? If so, we have to use double link list.
In doubly-link list, a programmer uses two pointers in the node, i.e. one to point to
next node and the other to point to the previous node. Now our node factory will

Page 37 of 519

CS301 Data Structures

___________________________________________________________________
create a node with three parts.

prev

element

next

First part is prev i.e. the pointer pointing to the previous node, second part is element,
containing the data to be inserted in the list. The third part is next pointer that points
to the next node of the list. The objective of prev is to store the address of the
previous node.
Lets discuss the code of the node of the doubly-link list. This node factory will create
nodes, each having two pointers. The interface methods are same as used in singly
link list. The additional methods are getPrev and setPrev. The method getPrev returns
the address of the previous node. Thus its return type is Node*. The setPrev method
sets the prev pointer. If we have to assign some address to prev pointer, we will call
this method. Following is the code of the doubly-linked list node.
/* this is the doubly-linked list class, it uses the next and prev pointers */
class Node {
public:
int get() { return object; }; // returns the value of the element
void set(int object) { this->object = object; }; // set the value of the element
Node* getNext() { return nextNode; }; // get the address of the next node
void setNext(Node* nextNode)
// set the address of the next node
{ this->nextNode = nextNode; };
Node* getPrev() { return prevNode; }; // get the address of the prev node
void setPrev(Node* prevNode)
// set the address of the prev node
{ this->prevNode = prevNode; };
private:
int object;
// it stores the actual value of the element
Node* nextNode; // this points to the next node
Node* prevNode; // this points to the previous node
};

Most of the methods are same as those in singly linked list. A new pointer prevNode
is added and the methods to get and set its value i.e. getPrev and setPrev. Now we
will use this node factory to create nodes.
You have to be very cautious while adding or removing a node in a doubly linked list.
The order in which pointers are reorganized is important. Lets have a pictorial view
of doubly-link list. The diagram can help us understand where the prevNode and
nextNode are pointing.

Page 38 of 519

CS301 Data Structures

___________________________________________________________________
head

size=5

current
This is a doubly link list. The arrows pointing towards right side are representing
nextNode while those pointing towards left side are representing prevNode. Suppose
we are at the last node i.e. the node with value 1. In case of going back, we usually
take the help of prevNode pointer. So we can go to the previous node i.e. the node
with value 7 and then to the node with value 8 and so on. In this way, we can traverse
the list from the end to start. We can move forward or backward in doubly-link list
very easily. We have developed this facility for the users to move in the list easily.
Lets discuss other methods of the doubly-linked list. Suppose we have created a new
node from the factory with value 9. We will request the node factory to create a new
object using new keyword. The newly created node contains three fields i.e. object,
prevNode and nextNode. We will store 9 into object and connect this new node in the
chain. Lets see how the pointers are manipulated to do that. Consider the above
diagram, the current is pointing at the node with value 6. The new node will be
inserted between the node with value 6 and the one with value 8.
In the first step, we assign the address of the node with value 8 to the nextNode of the
new node.
newNode->setNext( current->getNext() );
current

head

newNode

size=5

In the next step, a programmer points the prevNode of the newNode to the node with
value 6.
newNode->setprev( current );

Page 39 of 519

CS301 Data Structures

___________________________________________________________________
current

head

size=5

2
newNode

In the third step, we will set the previous node with value 8 to point to the newNode.
(current->getNext())->setPrev(newNode);
current

head

size=5

newNode

Now the prevNode of the node with value 8 is pointing to the node with value 9.
In the fourth step, the nextNode of the node with value 6 is pointing to the newNode
i.e. the node with value 9. Point the current to the newNode and add one to the size of
the list.
current->setNext( newNode );
current = newNode;
size++;

head

6
4

2
newNode

size=6

3
1

current

Now the newNode has been inserted between node with value 6 and node with value
8.

Page 40 of 519

CS301 Data Structures

___________________________________________________________________

Circularly-linked lists
Lets talk about circularly linked list. The next field in the last node in a singly-linked
list is set to NULL. The same is the case in the doubly-linked list. Moving along a
singly-linked list has to be done in a watchful manner. Doubly-linked lists have two
NULL pointers i.e. prev in the first node and next in the last node. A way around this
potential hazard is to link the last node with the first node in the list to create a
circularly-linked list.
The next method in the singly-linked list or doubly-linked list moves the current
pointer to the next node and every time it checks whether the next pointer is NULL or
not. Similarly the back method in the double-linked list has to be employed carefully
if the current is pointing the first node. In this case, the prev pointer is pointing to
NULL. If we do not take care of this, the current will be pointing to NULL. So if we
try to access the NULL pointer, it will result in an error. To avoid this, we can make a
circularly linked list.
We have a list with five elements. We have connected the last node with the first
node. It means that the next of the last node is pointing towards the first node.
current
head

size=5

The same list has been shown in a circular shape.

curren
6
8
hea

size=

2
7
1

You have noticed that there is no such node whose next field is NULL. What is the
benefit of this? If you use the next or back methods that move the current pointer, it
will never point to NULL. It may be the case that you keep on circulating in the list.
To avoid this, we get help from the head node. If we move the head node in the
circularly linked list, it will not be certain to say where it was pointing in the start. Its
advantages depend on its use. If we do not have to move too much in the list and have

Page 41 of 519

CS301 Data Structures

___________________________________________________________________
no problem checking the NULL, there is little need a circularly-linked list. But this
facility is available to us.
In this example, we made a circular linked list from a singly link list. In a singly link
list we move in one direction. We point the next pointer of the last node to the first
node. We can do the same with the doubly-linked list. The prev pointer of the first
node will point to the last node and the next pointer of the last node will point to the
first node. If you arrange all the nodes in a circle, one of the pointers (i.e. next
pointer) will move in clockwise direction while the prev pointers in anti-clockwise
direction. With the help of these pointers, you can move in the clockwise direction or
anti-clockwise direction. Head node pointer will remain at its position. You dont
need to change it. If there is a need to remove the node pointed by head node than you
have to move the head pointer to other node. Now we dont have any NULL pointer
in the doubly-linked list. We will not get any exception due to NULL pointers.

Josephus Problem
Now we will see an example where circular link list is very useful. This is Josephus
Problem. Consider there are 10 persons. They would like to choose a leader. The way
they decide is that all 10 sit in a circle. They start a count with person 1 and go in
clockwise direction and skip 3. Person 4 reached is eliminated. The count starts with
the fifth and the next person to go is the fourth in count. Eventually, a single person
remains.
You might ask why someone has to choose a leader in this way. There are some
historical stories attached to it. This problem is also studied in mathematics. Lets see
its pictorial view.
N=10, M=3
4

10
9

We have ten numbers representing the ten persons who are in a circle. The value of M
shows the count. As the value of M is three, the count will be three. N represents the
number of persons. Now we start counting clockwise. After counting up to three, we
have the number four. The number four is eliminated and put in the eliminated
column.

Page 42 of 519

CS301 Data Structures

___________________________________________________________________
Eliminated

N=10, M=3

10
9

After eliminating the number four, we will start our counting from number five.
Counting up to three, we have number eight which is eliminated and so on.
Eliminated
N=10, M=3
4
3
5

8
6

10
9

In the end, only number five will remain intact.

Eliminated
4

N=10, M=3
5

8
2
7
3

10
9
1
6

Page 43 of 519

CS301 Data Structures

___________________________________________________________________
If we have ten persons (N = 10) in a circle and eliminate after counting up to three (M
= 3). If we start our count from one, who will be the leader? We have studied this
earlier and know that the person who is sitting at the fifth position will become the
leader.

Suppose if the value of N is 300 or 400 and the value of M

is 5 or 10. Now who will be the leader? This is a
mathematical problem where we can change the values of
N and M. There is a formula where the values of N, M are
allotted. You can calculate who should become the leader.
Here we will not solve it mathematically. Rather, it will be
tackled as a computer problem. If you analyze the pictures
shown above, it gets clear that this can be solved with the
circular link list. We arrange these numbers in a circularlylinked list, point the head pointer at the starting number
and after calling the next method for three times, we will
reach the node which is to be removed. We will use the
remove method to remove the node. Then the next
method is called thrice from there and the node is
removed. We will continue this till we have only one node.
We are not concerned with the NULL pointers, internal to link list. However, if you
want to solve this problem and choose the best data structure, then circular link list is
the best option. We can also use the list to solve this.
Lets see the code of the program by which we can solve this problem. The code is as
under:
/*This program solves the Josephus Problem */
#include <iostream.h>
#include "CList.cpp" //contains the circularly-linked list definition
// The main method
void main(int argc, char *argv[])
{
CList list;
// creating an object of list
int i, N=10, M=3;
for(i=1; i <= N; i++ ) list.add(i); // initializing the list with values
list.start();

// pointing the pointers at the start of the list

// counting upto M times and removing the element

while( list.length() > 1 ) {
for(i=1; i <= M; i++ ) list.next();
Page 44 of 519

CS301 Data Structures

___________________________________________________________________
cout << "remove: " << list.get() << endl;
list.remove();
}
cout << "leader is: " << list.get() << endl;

//displaying the remaining node

}
We have included the CList.cpp. It means that we are using the circularly-linked
list. In the main method, CList factory is called to create a circular link list as CList
list; After this, we assign the values to N and M. We have used for loop to add the
nodes in the list. When this loop finishes, we have ten nodes in the list having values
from 1 to 10. But here a programmer may not pay attention to the internal details of
the list. We have created a list and stored ten numbers in it. Then we moved the
pointers of the list at the start of the list using the start method. It means that the
pointers are pointing at the position from where we want to start the counting of the
list.
There is a while loop that will continue executing until only one node is left in the list.
Inside this loop, we have a for loop. It will execute from 1 to M. It has only one
statement i.e. list.next(). This will move the pointer forward three times (as the value
of M is 3). Now the current pointer is at the 4th node. We called the remove method.
Before removing the node, we display its value on the screen using cout. Again we
come into the while loop, now the length of the list is 9. The for loop will be
executed. Now the list.next() is not starting from the start. It will start from the
position where the current pointer is pointing. The current pointer is pointing at the
next node to the node deleted. The count will start again. The list.next() will be called
for three times. The current pointer will point at the 8th node. Again the remove
method will be called and the current pointer moved to the next node and so on. The
nodes will be deleted one by one until the length of the list is greater than one. When
the length of the list is one, the while loop will be terminated. Now only one node is
left in the list i.e. the leader. We will display its value using the get method.
We can change the values of M and N. Similarly, these values can be read from the
file or can use the command line arguments to get values. There are many variations
of this problem. One variation is that the value of M keeps on changing. Sometimes, it
is 3, sometimes 4 or 5 and so on. Due to this, it will become difficult to think that who
will become leader. Make a picture in your mind that ten persons are sitting in a
circle. Every time the value of M is incremented by one. Now try to ascertain which
position you should sit to get chosen as a leader. You may like to write a program to
solve this or use the mathematical formula.

Page 45 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 05
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 3
3.1, 3.2.5, 3.3.1, 3.3.2
(array implementation)

Summary
5)
6)
7)
8)

Benefits of using circular list

Abstract Data Type
Stacks
Stack Implementation using arrays

In the previous lecture, we demonstrated the use of the circular list for the resolution
of the Josephus problem. After writing a program with the help of this data structure,
a leader among ten persons was selected. You must have noted many things while
trying to solve the problem. These things will help us to understand the usage of data
structures in C++, thus making the programming easy. The code of the program is
given below.
#include "CList.cpp"
void main(int argc, char *argv[])
{
CList list;
int i, N=10, M=3;
for(i=1; i <= N; i++ ) list.add(i);
list.start();
while( list.length() > 1 ) {
for(i=1; i <= M; i++ ) list.next();
cout << "remove: " << list.get() << endl;
list.remove();
}
cout << "leader is: " << list.get() << endl;
}
In the program, we include the file of the class CList and create its object i.e. list.
Then we solve the problem by using the add, start, length, next, remove and get
methods of the class CList.
In the program, we have included already-defined data structure CList. After defining
its different methods, we have an interface of Clist. There is no need to be worry

Page 46 of 519

CS301 Data Structures

___________________________________________________________________
about the nature of the list i.e. whether it is linked list, doubly linked list or an array.
For us, it is only a list to be manipulated according to our requirement. You will see
that a programmer may use different methods of the list object to solve the problem.
We add elements to the list by a simple call of add method and go to the first element
of the list by start method. Here, the length method is used in the condition of the
while loop. Then we remove elements from the list and use the next, get and remove
methods during this process. We get the current element by using the get method,
then remove it by calling the remove method and then go to the next element by the
method next. This way, all the elements are removed from the list except one element,
called the leader. This one element remains there as we execute the while loop one
less than the length of the list.
In singly linked list, the next returns false when it reaches to the last node due to the
fact that the next field of the last node is set to NULL. But in a circularly linked list
there is no NULL. It will be there only when there is no node in the list.
The whole process, which we carried out to solve the Josephus problem, can also be
carried out with functions in C++. While adopting this way (of writing functions), we
have to write these functions whenever we write another program that manipulates a
list. In this method, we define a class of the data structure list and its different
methods for the purpose of manipulation. This way, this class, obviously its methods
too, can be used in any program where the manipulation of a list is needed. Thus there
is re-usability of the code. In a class, we encapsulate the data and its methods. This
shows that we are no longer interested in the internal process of the class. Rather, we
simply use it wherever needed. The circular linked list, earlier used for the resolution
of the Josephus problem, can also be employed in other problems. We have a class
CList of this circular linked list through which any number of objects of data type of
circular linked list can be created. Thus we can assume the class CList as a factory,
creating as many objects of list as needed. This class and its objects in any program
can be used to solve the problems with the help of its interface. The interface of this
class consists of some methods like add, remove, next, back, get and some other
simple ones. While carrying out programming, we will see that these classes (objects)
help us very much to solve different problems.

Benefits of using circular list

While solving the Josephus problem, it was witnessed that the usage of circular linked
list helped us make the solution trivial. We had to just write a code of some lines that
solved the whole problem. In the program, we included the class CList (which is of
our data structure i.e. circular linked list) and used all of its methods according to the
requirements. There was no problem regarding the working of the methods. We just
called these methods and their definition in the class CList worked well.
Now we will see what happens if we solve the Josephus problem by using an array
instead of the class in our program. In this case, we have to define an array and write
code to move back and forth in the array and to remove different elements properly in
a particular order. A programmer needs to be very careful while doing this, to reach
the solution of the problem. Thus our code becomes very complex and difficult for
someone to understand and modify it. Moreover we cannot use this code in some
other problem. Note that here we are talking about the use of an array in the main
Page 47 of 519

CS301 Data Structures

___________________________________________________________________
program, not in the class that defines the CList data structure. There is no need to be
worried whether an array, singly linked list, doubly linked list is used or circular
linked list being employed internally in implementing the list in defining the class of
list data type. We only want that it should create objects of list. The usage of the class
of a data structure simplifies the code of the program. We can also use this class
wherever needed in other programs. This shows that the choice of appropriate data
structures can simplify an algorithm. It can make the algorithm much faster and
efficient. In this course, we will see that there are different data structures, which
makes the algorithms very easy to solve our problems. Later, we will see how some
elegant data structures lie at the heart of major algorithms. There is also a course
dedicated to study different algorithms and recipes that can be used to solve host of
complex problems. Moreover, we will study different data structures in detail and see
that with the use of a proper data structure, we can solve a problem efficiently. A
properly constructed data structure will always help in the solution of problems.

Abstract Data Type

A data type is a collection of values and a set of operations on those values. That
collection and these operations form a mathematical construct that may be
implemented with the use of a particular hardware or software data structure. The
term abstract data type (ADT) refers to the basic mathematical concept that defines
the data type. We have discussed four different implementations of the list data
structure. In case of implementation of the list with the use of an array, the size of the
array gives difficulty if increased. To avoid this, we allocate memory dynamically for
nodes before connecting these nodes with the help of pointers. For this purpose, we
made a singly linked list and connected it with the next pointer to make a chain.
Moving forward is easy but going back is a difficult task. To overcome this problem,
we made a doubly linked list using prev and next pointers. With the help of these
pointers, we can move forward and backward very easily. Now we face another
problem that the prev pointer of first node and the next pointer of the last node are
NULL. Therefore, we have to be careful in case of NULL pointers. To remove the
NULL pointers, we made the circular link list by connecting the first and last node.
The program employing the list data structure is not concerned with its
implementation. We do not care how the list is being implemented whether through
an array, singly linked list, doubly linked list or circular linked list. It has been
witnessed that in these four implementations of the list, the interface remained the
same i.e. it implements the same methods like add, get, next, start and remove etc.
This proves that with this encapsulation attained by making a class, we are not
concerned with its internal implementation. The implementation of these abstract data
types can be changed anytime. These abstract data types are implemented using
classes in C++. If the list is implemented using arrays while not fulfilling the
requirements, we can change the list implementation. It can be implemented with the
use of singly-link list or doubly link list. As long as the interface is same, a
programmer can change the internal implementation of the list and the program using
this list will not be affected at all. This is the abstract data type (ADT). What we care
about is the methods that are available for use, with the List ADT i.e. add, get, and
remove etc methods. We have not studied enough examples to understand all the
benefits of abstract data types. We will follow this theme while developing other
ADT. We will publish the interface and keep the freedom to change the
Page 48 of 519

CS301 Data Structures

___________________________________________________________________
implementation of ADT without effecting users of the ADT. The C++ classes provide
a programmer an ability to create such ADTs. What benefits can we get with the help
of these ADTs and classes? When we develop an ADT or a class or factory then the
users of this factory are independent of how this factory works internally. Suppose
that we have ordered the car factory (car class) to produce a new car and it replies
after a long time. If we ordered the remove method to remove one node and we are
waiting and it keeps on working and working. Then we might think that its
implementation is not correct. Although, we are not concerned with the internal
implementation of this ADT yet it is necessary to see whether this ADT is useful for
solving our problem or not. It should not become a bottleneck for us. If the method
we are using is too much time consuming or it has some problem in terms of
algorithm used. On one side, we only use the interfaces provided by these ADTs,
classes, or factories as long as they do what they promise. We are not concerned with
the internal details. On the other hand, we have to be careful that these factories or
methods should not take too much time so that these will not be useful for the
problem.
This distinction will always be there. Sometimes, the source code of classes is not
provided. We will be provided libraries, as standard libraries are available with the
compiler. These classes are in compiled form i.e. are in object form or in binary form.
On opening these files, you will not see the C++ code, rather binary code. When you
read the assembly language code, it will give some idea what this binary code is
about. You can view the interface methods in the .h file. As an application
programmer, you have to see that the ADTs being used are written in a better way.
The point to be remembered here is that you should not worry about the internal
implementation of these ADTs. If we want to change the internal implementation of
the ADTs, it can be done without affecting the users of these ADTs. While writing a
program, you should check its performance. If at some point, you feel that it is slow,
check the ADTs used at that point. If some ADT is not working properly, you can ask
the writer of the ADT to change the internal implementation of that ADT to ensure
that it works properly.

Stacks
Lets talk about another important data structure. You must have a fair idea of stacks.
Some examples of stacks in real life are stack of books, stack of plates etc. We can
add new items at the top of the stack or remove them from the top. We can only
access the elements of the stack at the top. Following is the definition of stacks.
Stack is a collection of elements arranged in a linear order
Lets see an example to understand this. Suppose we have some video cassettes. We
took one cassette and put it on the table. We get another cassette and put it on the top
of first cassette. Now there are two cassettes on the table- one at the top of other. Now
we take the third cassette and stack it on the two. Take the fourth cassette and stack it
on the three cassettes.
Now if we want to take the cassette, we can get the fourth cassette which is at the top
and remove it from the stack. Now we can remove the third cassette from the stack
and so on. Suppose that we have fifty cassettes stacked on each other and want to
Page 49 of 519

CS301 Data Structures

___________________________________________________________________
access the first cassette that is at the bottom of the stack. What will happen? All the
cassettes will fell down. It will not happen exactly the same in the computer. There
may be some problem. It does not mean that our data structure is incorrect. As we see
in the above example that the top most cassette will be removed first and the new
cassette will be stacked at the top. The same example can be repeated with the books.
In the daily life, we deal with the stacked goods very carefully.
Now we will discuss how to create a stack data structure or a factory, going to create
stack object for us. What will be the attributes of this object? During the discussion on
the list, we came to know that a programmer adds values in the list, removes values
from the list and moves forward and backward. In case of a stack too, we want to add
things and remove things. We will not move forward or backward in the stack. New
items can be added or removed at the top only. We can not suggest the removal of the
middle element of the stack.
Lets talk about the interface methods of the stacks. Some important methods are:
Method Name
push(x)
pop()
top()

Description
Insert x as the top element of the stack
Remove the top element of the stack and return it.
Return the top element without removing it from the stack.

The push(x) method will take an element and insert it at the top of the stack. This
element will become top element. The pop() method will remove the top element of
the stock and return it to the calling program. The top() method returns the top-most
stack element but does not remove it from the stack. The interface method names that
we choose has special objective. In case of list, we have used add, remove, get, set as
the suitable names. However, for stack, we are using push, pop and top. We can
depict the activity from the method name like push means that we are placing an
element on the top of the stack and pushing the other elements down.
The example of a hotels kitchen may help understand the concept of stacks in a
comprehensive manner. In the kitchen, the plates are stacked in a cylinder having a
spring on the bottom. When a waiter picks a plate, the spring moves up the other
plates. This is a stack of plates. You will feel that you are pushing the plates in the
cylinder and when you take a plate from the cylinder it pops the other plates. The top
method is used to get the top- most element without removing it.
When you create classes, interfaces and methods, choose such names which depicts
what these method are doing. These names should be suitable for that class or factory.
Lets discuss the working of stack with the help of a diagram.

Page 50 of 519

CS301 Data Structures

___________________________________________________________________

top
top
5

7
5

7
5

push(2)

push(5)

push(7)

push(1)

top
top

top
top

21

7
5

7
5

pop()

top

push(21)

7
5

top

2
21

2
pop()

top
pop()

2
5

pop()

At the start, the stack is empty. First of all, we push the value 2 in the stack. As a
result, the number 2 is placed in the stack. We have a top pointer that points at the top
element. Then we said push(5). Now see how 2 and 5 are stacked. The number 5 is
placed at the top of number 2 and the pointer top moves one step upward. Then we
pushed the number 7 which is placed on the top and the number 2 and 5 are below.
Similarly, we push number 1. The last figure in the first row shows the stacked values
of the numbers- 1, 7, 5 and 2.
Lets pop the elements from the stack. The first figure of second row shows the pop
operation. As a result, the number 1 is popped. Than again we push the number 21 on
the stack. The number 7, 5, and 2 are already in the stack and number 21 is pushed at
the top. If we pop now, the number 21 is popped. Now number 7 is at the top. If we
pop again, the number 7 is popped. Pop again and the number 5 is popped and number
2 remains in the stack. Here with the help of this diagram, we are proving that the
values are added at the top and removed at the top in a stack.
The last element to go into the stack is the first to come out. That is why, a stack is
known as LIFO (Last In First Out) structure. We know that the last element pushed in
the stack is at the top which is removed when we call pop. Lets see some other
scenarios. What happens if we call pop() while there is no element? One possible
way-out is that we have isEmpty() function that returns true if stack is empty and false
otherwise. This is a boolean function that returns true if there is no element in the
stack. Otherwise, it will return false. The second option is this that when we call pop
on an empty stack, it throws an exception. This is a concept of advanced C++.
Exception is also a way to convey that some unusual condition has arisen or
something has gone wrong. Suppose, if we have a division method and try to divide
some number with zero. This method will throw division by zero exception.
Page 51 of 519

CS301 Data Structures

___________________________________________________________________
Currently we will not throw an exception but use the isEmpty() method. The user who
is employing the stack is responsible to call the isEmpty() method before calling the
pop. Call the pop method if isEmpty() returns false . Otherwise, there will be a
problem.

Stack Implementation using array

Lets discuss the implementation of the stack. Suppose we implement the stack using
the arrays. The stack shown in the above diagram may be considered as an array. Here
the array is shown vertically. We can implement the stack using array. The interface
will remain as push and pop methods. The user of the stack does not need to know
that the stack is internally implemented with the help of array. The worst case for
insertion and deletion from an array may happen when we insert and delete from the
beginning of the array. We have to shift elements to the right for insertion and left for
removal of an element. We face the same problem while implementing the list with
the use of the array. If we push and pop the elements from the start of the array for
stack implementation, this problem will arise. In case of push, we have to shift the
stack elements to the right. However, in case of pop, after removing the element, we
have to shift the elements of stack that are in the array to the left. If we push the
element at the end of the array, there is no need to shift any element. Similarly as the
pop method removes the last element of the stack which is at the end of the array, no
element is shifted. To insert and remove elements at the end of the array we need not
to shift its elements. Best case for insert and delete is at the end of the array where
there is no need to shift any element. We should implement push() and pop() by
inserting and deleting at the end of an array.

top

7
5
2

top = 3

In the above diagram, on the left side we have a stack. There are four elements in the
stack i.e. 1, 7, 5 and 2. The element 1 is the extreme-most that means that it is inserted
in the end whereas 7, 5, and 2 have been added before. As this is a LIFO structure so
the element 1 should be popped first. On the right side we have an array with
positions 0, 1, 2, 3 and so on. We have inserted the numbers 2, 5, 7 and 1. We have
decided that the elements should be inserted at the end of the array. Therefore the
most recent element i.e. 1 is at position 3. The top is the index representing the
position of the most recent element. Now we will discuss the stack implementation in
detail using array.
We have to choose a maximum size for the array. It is possible that the array may
fill-up if we push enough elements. Now more elements cannot be pushed. Now
what should the user of the stack do? Internally, we have implemented the stack
using array which can be full. To avoid this, we write isFull() method that will return
Page 52 of 519

CS301 Data Structures

___________________________________________________________________
a boolean value. If this method returns true, it means that the stack (array) is full and
no more elements can be inserted. Therefore before calling the push(x), the user
should call isFull() method. If isFull() returns false, it will depict that stack is not full
and an element can be inserted. This method has become the part of the stack
interface. So we have two more methods in our interface i.e. isEmpty() and isFull().
Now we will discuss the actual C++ code of these operations. These methods are part
of stack class or stack factory. We have an array named A while current is its index.
The code of pop() method is as:
int pop()
{
return A[current--];
}
In this method, the recent element is returned to the caller, reducing the size of the
array by 1.
The code of push method is:
void push(int x)
{
A[++current] = x;
}
We know that ++current means that add one to the current and then use it. That also
shows that element x should be inserted at current plus one position. Here we are not
testing that this current index has increased from the array size or not. As discussed
earlier that before using the push method, the user must call isFull() method.
Similarly it is the responsibility of the user to call the isEmpty() method before calling
the pop method. Therefore there is no if statement in the push and pop method.
The code of the top() method is:
int top()
{
return A[current];
}
This method returns the element at the current position. We are not changing the
value of current here. We simply want to return the top element.
int isEmpty()
{
return ( current == -1 );
}
This method also tests the value of the current whether it is equal to -1 or not. Initially
when the stack is created, the value of current will be -1. If the user calls the

Page 53 of 519

CS301 Data Structures

___________________________________________________________________
isEmpty() method before pushing any element, it will return true.
int isFull()
{
return ( current == size-1);
}
This method checks that the stack is full or not. The variable size shows the size of the
array. If the current is equal to the size minus one, it means that the stack is full and
we cannot insert any element in it.
We have determined the cost and benefit of all the data structures. Now we will see
how much time these methods take. A quick examination shows that all the five
operations take constant time. In case of list, the find method takes too much time as it
has to traverse the list. Whereas the add and remove methods are relatively quick. The
methods of stack are very simple. There is no complexity involved. We insert element
at one side and also remove from that side not in the middle or some other place.
Therefore we need not to carry out a lot of work. During the usage of the array, the
stack methods push, pop, top, isFull and isEmpty all are constant time operations.
There is not much difference of time between them.
The complete code of the program is:
/* Stack implementation using array */
#include <iostream.h>
/* The Stack class */
class Stack
{
public:
Stack() { size = 10; current = -1;} //constructor
int pop(){ return A[current--];}
// The pop function
void push(int x){A[++current] = x;} // The push function
int top(){ return A[current];}
// The top function
int isEmpty(){return ( current == -1 );} // Will return true when stack is empty
int isFull(){ return ( current == size-1);} // Will return true when stack is full
private:
int object;
int current;
int size;
int A[10];

// The data element

// Index of the array
// max size of the array
// Array of 10 elements

// The main method

int main()
{
Stack stack;

// creating a stack object

};

Page 54 of 519

CS301 Data Structures

___________________________________________________________________
// pushing the 10 elements to the stack
for(int i = 0; i < 12; i++)
{
if(!stack.isFull())
// checking stack is full or not
stack.push(i);
// push the element at the top
else
cout <<"\n Stack is full, can't insert new element";
}
// pop the elements at the stack
for (int i = 0; i < 12; i++)
{
if(!stack.isEmpty())
// checking stack is empty or not
cout << "\n The popped element = " << stack.pop();
else
cout <<"\n Stack is empty, can't pop";
}
}
The output of the program is:
Stack is full, can't insert new element
Stack is full, can't insert new element
The popped element = 9
The popped element = 8
The popped element = 7
The popped element = 6
The popped element = 5
The popped element = 4
The popped element = 3
The popped element = 2
The popped element = 1
The popped element = 0
Stack is empty, can't pop
Stack is empty, can't pop
However, a programmer finds the size-related problems in case of an array. What
should we do when the array is full? We can avoid the size limitation of a stack
implemented with an array by using a linked list to hold the stack elements. Further
discussion on this issue will be made in the next lecture.

Page 55 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 06
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 3
3.3.2, 3.3.3
(Postfix expressions)

Summary

Stack From the Previous Lecture

Stack Using Linked List
Stack Implementation: Array or Linked List
Use of Stack
Precedence of Operators
Examples of Infix to Postfix

Stack From the Previous Lecture

We started discussing Stack data structure and its implementation in the previous
lecture. We also implemented stack structure using an array and wrote code for its
push(), pop() and top() operations. We realized that we have to specify the size of the
array before using it whether we declare it statically or dynamically. Arrays are of
fixed size and when they become full, no more elements can be added to them. In
order to get to know that the array has gone full, we wrote the isFull() method. It
became the responsibility of the user of the stack structure to call isFull() method
before trying to insert an element using the push() method otherwise the whole
program could crash.
isEmpty() method is implemented as a stack can be empty like a list or set structures.
It is important to understand that isFull() method is there in stack implementation
because of limitation of array but isEmpty() method is part of the stack characteristics
or functionality.
As previously in the implementation of list structure, we used linked list while
allocating nodes dynamicallyin order to avoid the fixed sized limitation of array. Now
in this case also, again to overcome the limitation of array, we are going to make use
of linked list in place of array to implement the stack data structure. Lets see, how we
can implement a stack structure using linked list and how the implementation code
will look like internally.

Stack Using Linked List

Page 56 of 519

CS301 Data Structures

___________________________________________________________________
We can avoid the size limitation of a stack implemented with an array, with the help
of a linked list to hold the stack elements.
As needed in case of array, we have to decide where to insert elements in the list and
where to delete them so that push and pop will run at the fastest.
Primarily, there are two operations of a stack; push() and pop(). A stack carries lifo
behavior i.e. last in, first out.
You know that while implementing stack with an array and to achieve lifo behavior,
we used push and pop elements at the end of the array. Instead of pushing and
popping elements at the beginning of the array that contains overhead of shifting
elements towards right to push an element at the start and shifting elements towards
left to pop an element from the start. To avoid this overhead of shifting left and right,
we decided to push and pop elements at the end of the array.
Now, if we use linked list to implement the stack, where will we push the element
inside the list and from where will we pop the element? There are few facts to
consider, before we make any decision:
- For a singly-linked list, insert at start or end takes constant time using the head
and current pointers respectively. As far as insertion is concerned, it is workable
and equally efficient at the start and end.
- Removing an element at the start is constant time but removal at the end requires
traversing the list to the node one before the last. So removing from the start is
better approach rather than from the end.
Therefore, it makes sense to place stack elements at the start of the list because
insertion and removal take constant time. As we dont need to move back and forth
within the list, therefore, there is no requirement of doubly or circular linked list.
Singly linked list can serve the purpose. Hence, the decision is to insert the element at
the start in the implementation of push operation and remove the element from the
start in the pop implementation.

head
top

1
7

5
2
Fig 1. Stack using array (on left side) and linked list (on right side)

There are two parts of above figure.On the left hand, there is the stack implemented
using an array. The elements present inside this stack are 1, 7, 5 and 2. The most
Page 57 of 519

CS301 Data Structures

___________________________________________________________________
recent element of the stack is 1. It may be removed if the pop() is called at this point
of time. On the right side, there is the stack implemented using a linked list. This
stack has four nodes inside it which are liked in such a fashion that the very first node
pointed by the head pointer contains the value 1. This first node with value 1 is
pointing to the node with value 7. The node with value 7 is pointing to the node with
value 5 while the node with value 5 is pointing to the last node with value 2. To make
a stack data strcuture using a linked list, we have inserted new nodes at the start of the
linked list.
Lets see the code below to implement pop() method of the stack.
int pop()
{
1. int x = head->get();
2. Node * p = head;
3. head = head->getNext();
4. delete p;
5. return x;
}
At line 1, we have declared x as an int and retrieved one element from the node of the
stack that is pointed by the head pointer. Remember, the Node class and its get()
method that returns the value inside the node.
At line 2, p is declared as a pointer of type Node and address inside the head pointer
is being saved inside this p pointer.
At line 3, the address of the next node is being retrieved with the help of the getNext()
method of the Node class and being assigned to head pointer. After this assignment,
the head pointer has moved forward and started pointing to the next element in the
stack.
At line 4, the node object pointed by the pointer p is being deallocated (deleted).
At line 5, the function is returning the value of the node retrieved in step 1.

head

top

5
2
Fig 2. A node removed from the stack after the pop() call

Lets see the code of the push() method of the stack:

void push(int x)

Page 58 of 519

CS301 Data Structures

___________________________________________________________________
{
1. Node * newNode = new Node();
2. newNode->set(x);
3. newNode->setNext(head);
4. head = newNode;
}
In line 1, a new node is created, using the new Node() statement and returned pointer
is assigned to a pointer newNode. So newNode starts pointing to the newly created
Node object.
In line 2, the value 2 is set into the newly created Node object.
In line 3, the next node of the newly created node is set to the node pointed to by the
head pointer using setNext(head).
In line 4, the head pointer is made to point to the newly created node.

head
top

9
7

5
2

newNode

Fig 3. A node added to the stack after the push(9) call

These are two primary methods of a stack. By using the push() method, we can keep
on pushing elements and using the pop() methods. Elements can be removed from the
stack till the time, it gets empty. As discussed earlier, isEmpty() is the stack
characteristic but isFull() was implemented because of the size limitation of the array.
We are no more using array to implement a stack. Rather, we have used linked list
here for stack implementation. Therefore, isFull() might not be required here. An
interesting question arises here. Can we add infinite elements to the stack now. We
should remember that this program of stack will run on computer that definitely has a
limited memory. Memory or Address space of a computer is the space (physical
memory and disk space) that can be addressed by the computer which is limited
inlcuding the limited physical memory. Disk space is used as the virtual memory ( we
will not discuss virtual memory in detail here). A computer with 32-bit addressing can
address upto 232-1 memory locations and similarly a computer with 64-bit addressing
can address upto 264-1 addresses. If this address space becomes full, the stack will
definitely be full. However, the stack implementation is not liable for this fullness of
address space and it is the limitation of a computer address space. Therefore, we dont
need to call isFull() before pushing the element. Rather, isEmpty() is called before
poping an element from the stack.

Page 59 of 519

CS301 Data Structures

___________________________________________________________________
Lets see the remaining methods of the stack while using linked list to implement it.
int top()
{
return head->get();
}
int isEmpty()
{
return ( head == NULL );
}
The above-mentioned methods i.e. top() and isEmpty() are very simple functions. One
statement inside the top() is retrieving the top element (pointed to by the head pointer)
from the stack and returning it back by value. It is important to note that top() is not
removing the element from the stack, but only retrieving it. The one statement inside
isEmpty() is a check to see if the head pointer is not pointing to any node and it is
NULL. If the head pointer is NULL that means the stack is empty, the method returns
true otherwise it returns false.
All four operations push(), pop(), top() and isEmpty() take constant time. These are
very simple methods and dont contain loops. They are also not CPU hungry
operation. Also note that we have not written isFull() while implementing stack with
the linked list.

Stack Implementation: Array or Linked List

Since both implementations support stack operations in constant time, we will see
what are the possible reasons to prefer one implementation to the other.
- Allocating and de-allocating memory for list nodes does take more time than preallocated array. Memory allocation and de-allocation has cost in terms of time,
especially, when your system is huge and handling a volume of requests. While
comparing the stack implementation, using an array versus a linked list, it
becomes important to consider this point carefully.
- List uses as much memory as required by the nodes. In contrast, array requires
allocation ahead of time. In the previous bullet, the point was the time required for
allocation and de-allocation of nodes at runtime as compared to one time
allocation of an array. In this bullet, we are of the view that with this runtime
allocation and de-allocation of nodes, we are also getting an advantage that list
consumes only as much memory as required by the nodes of list. Instead of
allocating a whole chunk of memory at one time as in case of array, we only
allocate memory that is actually required so that the memory is available for other
programs. For example, in case of implementing stack using array, you allocated
array for 1000 elements but the stack, on average, are using 50 locations. So, on
the average, 950 locations remain vacant. Therefore, in order to resolve this
problem, linked list is handy.
- List pointers (head, next) require extra memory. Consider the manipulation of
array elements. We can set and get the individual elements with the use of the
array index; we dont need to have additional elements or pointers to access them.
But in case of linked list, within each node of the list, we have one pointer
element called next, pointing to the next node of the list. Therefore, for 1000
Page 60 of 519

CS301 Data Structures

___________________________________________________________________
nodes stack implemented using list, there will be 1000 extra pointer variables.
Remember that stack is implemented using singly-linked list. Otherwise, for
doubly linked list, this overhead is also doubled as two pointer variables are stored
within each node in that case.
- Array has an upper limit whereas list is limited by dynamic memory allocation. In
other words, the linked list is only limited by the address space of the machine.
We have already discussed this point at reasonable length in this lecture.

Use of Stack
Examples of uses of stack include- traversing and evaluating prefix, infix and postfix
expressions.
Consider the expression A+B: we think of applying the operator + to the operands
A and B. We have been writing this kind of expressions right from our primary
classes. There are few important things to consider here:
Firstly, + operator requires two operators or in other words + is a binary operator.
Secondly, in the expression A+B, the one operand A is on left of the operator while
the other operand B is on the right side. This kind of expressions where the operator is
present between two operands called infix expressions. We take the meanings of this
expression as to add both operands A and B.
There are two other ways of writing expressions:
We could write +AB, the operator is written before the operands A and B. These
kinds of expressions are called Prefix Expressions.
We can also write it as AB+, the operator is written after the operands A and B.
This expression is called Postfix expression.
The prefixes pre and post refer to the position of the operator with respect to the two
operands.
Consider another expression in infix form: A + B * C. It consists of three operands A,
B, C and two operator +,* . We know that multiplication () is done before addition
(+), therefore, this expression is actually interpreted as: A + (B * C). The
interpretation is because of the precedence of multiplication (*) over addition (+). The
precedence can be changed in an expression by using the parenthesis. We will discuss
it a bit later.
Lets see, how can we convert the infix expression A + (B * C) into the postfix form.
Firstly, we will convert the multiplication to postfix form as: A + (B C *). Secondly,
we will convert addition to postfix as: A (B C *) + and finally it will lead to the
resultant postfix expression i.e. : A B C * +. Lets convert the expression (A + B) * C
to postfix. You might have noticed that to overcome the precedence of multiplication
operator (*) we have used parenthesis around A + B because we want to perform
addition operation first before multiplication.
(A + B) * C

infix form

(A B +) * C

convert addition

(A B +) C *

convert multiplication

AB+C*

postfix form
Page 61 of 519

CS301 Data Structures

___________________________________________________________________
These expressions may seem to be difficult to understand and evaluate at first. But
this is one way of writing and evaluating expressions. As we are normally used to
infix form, this postfix form might be little confusing. If a programmer knows the
algorithm, there is nothing complicated and even one can evaluate the expression
manually.

Precedence of Operators
There are five binary operators, called addition, subtraction, multiplication, division
and exponentiation. We are aware of some other binary operators. For example, all
relational operators are binary ones. There are some unary operators as well. These
require only one operand e.g. and +. There are rules or order of execution of
operators in Mathematics called precedence. Firstly, the exponentiation operation is
executed, followed by multiplication/division and at the end addition/subtraction is
done. The order of precedence is (highest to lowest):
Exponentiation

Multiplication/division *, /
Addition/subtraction
+, For operators of same precedence, the left-to-right rule applies:
A+B+C means (A+B)+C.
For exponentiation, the right-to-left rule applies:
A B C means A (B C)
We want to understand these precedence of operators and infix and postfix forms of
expressions. A programmer can solve a problem where the program will be aware of
the precedence rules and convert the expression from infix to postfix based on the
precedence rules.

Examples of Infix to Postfix

Lets consider few examples to elaborate the infix and postfix forms of expressions
based on their precedence order:
Infix
A+B
12 + 60 23
(A + B)*(C D )
A B * C D + E/F

Postfix
AB+
12 60 + 23
AB+CD*
A B C*D E F/+

In the next lecture we will see, how to convert infix to postfix and how to evaluate
postfix form besides the ways to use stack for these operations.

Page 62 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 07
Reading Material
Page 63 of 519

CS301 Data Structures

___________________________________________________________________
Data Structures and Algorithm Analysis in C++

Chapter. 3
3.3.3

Summary
9)
10)
11)

Evaluating postfix expressions

An example
Infix to postfix Conversion

Evaluating postfix expressions

In the previous lecture, we talked about infix and postfix expression and tried to
understand how to write the postfix notation of mathematical expressions. A
programmer can write the operators either after the operands i.e. postfix notation or
before the operands i.e. prefix notation. Some of the examples are as under:
Infix
A+B
12 + 60 23
(A + B)*(C D )
A B * C D + E/F

Postfix
AB+
12 60 + 23
AB+CD*
A B C*D E F/+

The last expression seems a bit confusing but may prove simple by following the
rules in letter and spirit. In the postfix form, parentheses are not used. Consider the
infix expressions as 4+3*5 and (4+3)*5. The parentheses are not needed in the
first but are necessary in the second expression. The postfix forms are:
4+3*5
(4+3)*5

435*+
43+5*

In case of not using the parenthesis in the infix form, you have to see the precedence
rule before evaluating the expression. In the above example, if we want to add first
then we have to use the parenthesis. In the postfix form, we do not need to use
parenthesis. The position of operators and operands in the expression makes it clear in
which order we have to do the multiplication and addition.
Now we will see how the infix expression can be evaluated. Suppose we have a
postfix expression. How can we evaluate it? Each operator in a postfix expression
refers to the previous two operands. As the operators are binary (we are not talking
about unary operators here), so two operands are needed for each operator. The nature
of these operators is not affected in the postfix form i.e. the plus operator (+) will
apply on two operands. Each time we read an operand, we will push it on the stack.
We are going to evaluate the postfix expression with the help of stack. After reaching
an operator, we pop the two operands from the top of the stack, apply the operator and
push the result back on the stack. Now we will see an example to comprehend the
working of stack for the evaluation of the postfix form. Here is the algorithm in
pseudo code form. After reading this code, you will understand the algorithm.

Page 64 of 519

CS301 Data Structures

___________________________________________________________________
Stack s;
// declare a stack
while( not end of input ) {
// not end of postfix expression
e = get next element of input
if( e is an operand )
s.push( e );
else {
op2 = s.pop();
op1 = s.pop();
value = result of applying operator e to op1 and op2;
s.push( value );
}
}
finalresult = s.pop();
We have declared a Stacks. There is a while loop along with not end of input
condition. Here the input is our postfix expression. You can get the expression from
the keyboard and use the enter key to finish the expression. In the next statement, we
get the next element and store it in e. This element can be operator or operand. The
operand needs not to be single digit. It may be of two digits or even more like 60 or
234 etc. The complete number is stored in the e. Then we have an if statement to
check whether e is an operand or not. If e is an operand than we wrote s.push(e)
i.e. we pushed the e onto the stack. If e is not the operand, it may be an operator.
Therefore we will pop the two elements and apply that operator. We pop the stack and
store the operand in op2. We pop the stack again and store the element in op1.
Then the operator in e is applied to op1 and op2 before storing the result in
value. In the end, we push the value on the stack. After exiting the loop, a
programmer may have only one element in the stack. We pop this element which is
the final result.
Consider the example of 4+3*2 having a postfix form of 432*+. Here 4, 3, and 2 are
operands whereas + and * are operators. We will push the numbers 4, 3 and 2 on the
stack before getting the operator *. Two operands will be popped from the stack and *
is being applied on these. As stack is a LIFO structure, so we get 2 first and then 3 as
a result of pop. So 2 is store in op1 and 3 in op2. Lets have a look on the program
again. On applying * on these, we will push the result (i.e. 6) on the stack. The while
loop will be executed again. In case of getting the next input as operand, we will
push it on the stack otherwise we will pop the two operands and apply the operator on
these. Here the next element is the operator +. So two operands will be popped from
the stack i.e. 6 and 4. We will apply the operator plus on these and push the result (i.e.
10) on the stack. The input is finished. Now we will pop the stack to get the final
result i.e. 10.

An Example
In the earlier example, we have used the stack to solve the postfix expression. Lets
see another comprehensive example. The postfix expression is:
623+-382/+*23+
Page 65 of 519

CS301 Data Structures

___________________________________________________________________
We want to evaluate this long expression using stack. Lets try to solve it on paper.
We have five columns here i.e. input, op1, op2, value and stack. We will run our
pseudo code program. In the start, we read the input as a result we get number 6. As 6
is operand, so it will be pushed on the stack. Then we have number 2 which will also
be pushed on the stack. Now 2 is the most recent element. The next element is the
number 3 that will also be pushed on the stack. Now, there are three elements on the
stack i.e. 3, 2 and 6. The number 3 is the most recent. On popping, we will get the
number 3 first of all. The next element is +, an operator. Now the else part of our
pseudo code is executed. We will pop two operands from the stack and apply the
operator (+) on these. The number 3 will be stored in variable op2 and number 2 in
op1. The operator (+) will be applied on these i.e. 2+3 and the result is stored in
value. Now we will push the value (i.e. 5) on the stack. Now we have two numbers on
the stack i.e. 5 and 6. The number 5 is the most recent element. The next element is . As it is also an operator, so we will pop the two elements from the stack i.e. 5 and 6.
Now we have 5 in op2 and 6 in op1. On applying the operator (-), we will get the
result as 1 (6-5). We cant say op2 - op1. The result (1) will be pushed on stack. Now
on the stack, we have only one element i.e. 1. Next three elements are operands so we
pushed 3, 8 and 2 on the stack. The most recent element is 2. The next input is an
operator in the expression i.e. /, we will pop two elements from the stack. The
number 2 will be stored in op2 while number 8 in op1. We apply the operator (/) on
the op1 and op2 i.e. (op1/op2), the result is 4 (i.e. 8/2). We push the result on the
stack. We have, now, three elements i.e. 4, 3, and 1 on the stack. The next element is
operator plus (+). We will pop the two elements i.e. 4 and 3 and will apply the
operator (+). The result (7) will be pushed on the stack. The next input element is
operator multiply (*). We will pop the two elements i.e. 7 and 1 and the result (7*1 =
7) is pushed on the stack. You have noted that whenever we have an operator in the
input expression, we have two or more elements on the stack. As the operators we are
using are binary and we need two operands for them. It will never be the case that you
want to pop two elements from the stack and there is only one or no element on the
stack. If this happens than it means there is an error in the program and you have
popped more values than required. The next input element is 2 that is pushed on the
stack. We have, now, the operator ( ) in the input. So we will pop the two elements,
op2 will hold 2 and op1 will have the number 7. The operator ( ) will be applied on
the operands i.e. (7 2) and the result (49) is pushed on the stack. We have, now, the
number 3 in the element being pushed on the stack. The last element is the operator
plus (+). So we pop the two elements i.e. 49 and 2 and apply the operator on these.
The result (49+3 = 52) is pushed on the stack. The input expression is finished,
resulting in the final result i.e. 52.
This the tabular form of the evaluation of the postfix expression.
Input
6
2
3

op1

op2

value

stack
6
2
6
3
2
6

Page 66 of 519

CS301 Data Structures

___________________________________________________________________
+

6
6

5
5

1
1

*
2

1
1

7
7

7
7

7
7

2
2

49
49

49

52

5
6
1
3
1
8
3
1
2
8
3
1
4
3
1
7
1
7
2
7
49
3
49
52

With the help of stack we can easily solve a very big postfix expression. Suppose you
want to make a calculator that is a part of some application e.g. some spreadsheet
program. This calculator will be used to evaluate expressions. You may want to
calculate the value of a cell after evaluating different cells. Evaluation of the infix
form programmatically is difficult but it can be done. We will see another data
structure which being used to solve the expressions in infix form. Currently, we have
to evaluate the values in different cells and put this value in another cell. How can we
do that? We will make the postfix form of the expression associated with that cell.
Then we can apply the above algorithm to solve the postfix expression and the final
result will be placed at that cell. This is one of the usages of the stack.

Infix to postfix Conversion

We have seen how to evaluate the postfix expressions while using the stack. How can
we convert the infix expression into postfix form? Consider the example of a
spreadsheet. We have to evaluate expressions. The users of this spreadsheet will
employ the infix form of expressions. Consider the infix expressions A+B*C and
(A+B)*C. The postfix versions are ABC*+ and AB+C* respectively. The order
of operands in postfix is the same as that in the infix. In both the infix expressions, we
have the order of operands as A, B and then C. In the postfix expressions too, the
order is the same i.e. A, B, followed by C. The order of operands is not changed in
postfix form. However, the order of operators may be changed. In the first expression
A+B*C, the postfix expression is ABC*+. In the postfix form multiplication
comes before the plus operator. In scanning from left to right, the operand A can be
inserted into postfix expression. First rule of algorithm is that if we find the operand
Page 67 of 519

CS301 Data Structures

___________________________________________________________________
in the infix form, put it in the postfix form. The rules for operators are different. The
+ cannot be inserted in the postfix expression until its second operand has been
scanned and inserted. Keep the expression A+B*C in your mind. What is the second
operand of the plus? The first operand is A and the second operand is the result of
B*C. The + has to wait until the * has not been performed. You do the same thing
while using the calculator. First you will multiply the B*C and then add A into the
result. The + has to be stored away until its proper position is found. When B is
seen, it is immediately inserted into the postfix expression. As B is the operand, we
will send the operand to the postfix form. Can the + be inserted now? In case of
A+B*C, we cannot insert + because * has precedence. To perform
multiplication, we need the second operand. The first operand of multiplication is B
while the second one is C. So at first, we will perform the multiplication before
adding result to A.
In case of (A+B)*C, the closing parenthesis indicates that + must be performed
first. After sending the A and B to postfix perform, we can perform the addition due
to the presence of the parenthesis. Then C will be sent to the postfix expression. It
will be followed by the multiplication of the C and the result of A + B. The postfix
form of this expression is AB+C*. Sometimes, we have two operators and need to
decide which to apply first like in this case + and *. In this case, we have to see
which operator has higher precedence. Assume that we have a function
prcd(op1,op2) where op1 and op2 are two operators. The function prcd(op1,op2)
will return TRUE if op1 has precedence over op2, FASLE otherwise. Suppose we call
this function with the arguments * and + i.e. prcd(*, +), it will return true. It will
also return true in case both op1 and op2 are + e.g. if we have A+B+C, then it does
not matter which + we perform first. The call prcd(+ , *) will return false as the
precedence of * is higher than the + operator. The + has to wait until * is performed.
Now we will try to form an algorithm to convert infix form into postfix form. For this
purpose, a pseudo code will be written. We will also write the loops and if conditions.
The pseudo code is independent of languages. We will be using a stack in this
algorithm. Here, the infix expression is in the form of a string. The algorithm is as
follows:
Stack s;
while( not end of input ) {
c = next input character;
if( c is an operand )
add c to postfix string;
else {
while( !s.empty() && prcd(s.top(),c) ){
op = s.pop();
add op to the postfix string;
}
s.push( c );
}
while( !s.empty() ) {
op = s.pop();
add op to postfix string;

Page 68 of 519

CS301 Data Structures

___________________________________________________________________
}
First we will declare a stack s. The while loop will continue till the end of input.
We read the input character and store it in the c. Here the input character does not
mean one character, but an operand or an operator. Then we have a conditional if
statement. If c is an operand, then we will have to add it to postfix string. Whenever
we get an operand in the infix form, it will be added to the postfix form. The order of
operands does not change in the conversion. However, in this case, the order of
operators may change. If c is the operator, then we will, at first, check that stack is
not empty besides identifying the precedence of the operators between the input
operator and the operator that is at the top of the stack. In case of the precedence of
the operator that is on the stack is higher, we will pop it from the stack and send to the
postfix string. For example if we have * on the stack and the new input operator is +.
As the precedence of the + operator is less than the * operator, the operands of the
multiplication has already been sent to the postfix expression. Now, we should send
the * operator to the postfix form. The plus operator (+) will wait. When the while
loop sends all such operators to the postfix string, it will push the new operator to the
stack that is in c. It has to wait till we get the second operand. Then we will again
get the input. On the completion of the input, the while loop will be finished. There
may be a case that input may be completed even at the time when there are still some
elements on the stack. These are operators. To check this, we have another while
loop. This loop checks if the stack is not empty, pops the operator and put it in the
postfix string. Lets take a look at a comprehensive example to understand it. In case
of the infix expression, A + B * C, we have three columns, one each for input symbol,
the postfix expression and the stack respectively. Now lets execute the pseudo code.
First of all, we get the A as input. It is an operand so we put it on the postfix string.
The next input is the plus operator (+) which will be pushed on the stack. As it is an
operator and we need two operands for it. On having a look at the expression, you
might have figure out that the second operand for the plus operator is B*C. The next
input is the operand B being sent to the postfix expression form. The next thing we
get is the input element as *. We know that the precedence of * is higher than that of
the +. Lets see how we can do that according to our pseudo code. The prcd(s.top(),
op) takes two operands. We will get the top element of the stack i.e. + will be used as
first argument. The second argument is the input operator i.e. *. So the function call
will be as prcd(+, *) while the function returns false because the precedence of the
plus operator is not higher than the multiplication operator. So far, we have only one
operand for multiplication i.e. B. As multiplication is also a binary operator, it will
also have to wait for the second operand. It has to wait and the waiting room is stack.
So we will push it on the stack. Now the top element of the stack is *. The next
symbol is C. Being an operand, C will be added to the postfix expression. At this
point, our input expression has been completed. Our first while loop executes till the
end of input. After the end of the input, the loop will be terminated. Now the control
goes to the second while loop which says if there is something on the stack, pop it and
add it the postfix expression. In this case, we have * and + on the stack. The * is at the
top of the stack. So when we pop, we get * which is at the top of the stack and it will
be added to the postfix expression. In the result of second pop, we get the plus
operator (+) which is also added to the postfix expression. The stack is empty now.
The while loop will be terminated and postfix expression is formed i.e. ABC*+.

Page 69 of 519

CS301 Data Structures

___________________________________________________________________
Symbol
A
+
B
*

postfix
A
A
AB
AB

ABC
ABC*
ABC*+

stack
+
+
*
+
*
+
+

If we have to convert the infix expression into the postfix form, the job is easily done
with the help of stack. The above algorithm can easily be written in C++ or C
language, specially, if you already have the stack class. Now you can convert very big
infix expressions into postfix expressions. Why we have done this? This can be
understood with the help of the example of spreadsheet programming where the value
of cell is the evaluation of some expression. The user of the spreadsheets will use the
infix expressions as they are used to it.
Sometimes we do need the parenthesis in the infix form. We have to evaluate the
lower precedence operator before the higher precedence operator. If we have the
expression (A+B) *C, this means that we have to evaluate + before the multiplication.
The objective of using parenthesis is to establish precedence. It forces to evaluate the
expression first of all. We also have to handle parenthesis while converting the infix
expression into postfix one. When an open parenthesis ( is read, it must be pushed
on the stack. This can be done by setting prcd(op,( ) to be FALSE. What is the
reason to put the parenthesis on the stack? It is due to the fact that as long as the
closing parenthesis is not found, the open parenthesis has to wait. It is not a unary or
binary operator. Actually, it is a way to show or write precedence. We can handle the
parenthesis by adding some extra functionality in our prcd function. When we call
prcd(op, (), it will return false for all the operators and be pushed on the stack. Also,
prcd( (,op ) is FALSE which ensures that an operator after ( is pushed on the stack.
When a ) is read. All operators up to the first ( must be popped and placed in the
postfix string. To achieve this our function prcd( op,) ) should return true for all the
operators. Both the ( and the) will not go to the postfix expression. In postfix
expression, we do not need parenthesis. The precedence of the operators is established
in such a way that there is no need of the parenthesis. To include the handling of
parenthesis, we have to change our algorithm. We have to change the line s.push(c)
to:
if( s.empty() || symb != ) )
s.push( c );
else
s.pop(); // discard the (
If the input symbol is not ) and the stack is not empty, we will push the operator on
the stack. Otherwise, it is advisable to pop the stack and discard the (. The following
functionality has to be added in the prcd function.

Page 70 of 519

CS301 Data Structures

___________________________________________________________________
prcd( (, op )
prcd( op, ) )
prcd( op, ) )
prcd( ), op )

=
=
=
=

FALSE
FALSE
TRUE
error

for any operator

for any operator other than )
for any operator other than (
for any operator.

In the next lecture we will see in detail an example regarding the use of parenthesis.

Page 71 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 08
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 3
3.3.3

Summary

Conversion from infix to postfix

C++ Templates
Implementation of Stack
Function Call Stack

Conversion from infix to postfix

In the previous lecture, we discussed the way to convert an infix notation into a
postfix notation. During the process of conversion, we saw that there may be need of
parenthesis in the infix especially at the times when we want to give a higher
precedence to an operator of lower precedence. For example, if there is a + operator
and * operator in an expression and a programmer wants the execution of addition
before the multiplication. To achieve this object, it is necessary to put parentheses
around the operands of + operator. Suppose, there is the expression A + B * C in
which we want to give the precedence to the + operator over * operator. This
expression will be written as (A + B) * C. Now we are going to discuss the conversion
of infix expression that includes parentheses to the postfix expression. We have
defined the return values for opening (and closing ) parentheses in the precedence
function. Lets try to understand this process with the help of an example of
converting the infix expression (A + B) * C into a postfix expression. We will see
how our algorithm, discussed earlier, converts this infix expression into a postfix
expression. To carry out the process of conversion we have three columns symbol,
postfix and stack. The column symbol has the input symbols from the expression. The
postfix column has the postfix string (expression) after each step and the stack is used
to put the operators on it. The whole process of converting the infix notation into a
postfix is given in the following table. This process of conversion is completed in
eight steps. Each of the rows of the table depicts one step.
Step No.

Symbol

Postfix

Stack

Page 72 of 519

CS301 Data Structures

___________________________________________________________________
1
2
3
4
5
6
7
8

(
A
+
B
)
*
C

A
A
AB
AB+
AB+
AB+C
AB+C*

(
(
(+
(+
*
*

First of all, there is the input symbol ((i.e. opening parenthesis). As this is not an
operand, it may be put on the stack. The next input symbol is A. Being an operand it
goes to the postfix string and the stack remains unchanged. Then there is + operator of
binary type. Moreover, there is one operand in the postfix string. We push this +
operator on the stack and it has to wait for its second operand. Now in the input
symbol, there is an operand B. We put his operand in the postfix string. Then after
this, there is the closing parenthesis ) in the input symbol. We know that the
presence of a closing parenthesis in the input means that an expression (within the
parentheses) has been completed. All of its operands and operators are present with in
the parentheses. As studied in the algorithm, we discard a closing parenthesis when it
comes in the input. Then the operators from the stack are popped up and put in the
postfix string. We also pop the opening parenthesis and discard it as we have no need
of opening as well as closing parenthesis in the postfix notation of an expression. This
process is carried out in the 5th row of the table. The + operator is put in the postfix
string. We also discard the opening parenthesis, as it is not needed in the postfix.
Now the next input symbol is *. We put this operator on the stack. There is one
operand for the * operator i.e. AB+. The * operator being a binary operator, has to
wait for the second operand. C is the Next input symbol that is an operand. We put it
in the postfix string. After this, the input string (expression) ends so we come out of
the loop. We check if there is any thing on the stack now? There is * operator in the
stack. We pop the operator and put it into the postfix string. This way, we get the
postfix form of the given infix expression that becomes AB+C*. In this postfix
expression, the + operator is before the * operator. So addition operation is done
before the multiplication. This is mainly due to the fact that in the infix expression,
we have put parentheses to give + operator the precedence higher than the * operator.
Note that there are no parentheses in the postfix form of the given infix expression.
Now we apply the evaluation algorithm on this postfix expression (i.e. AB+C*). The
two operands A and B, will go to the stack. Then operator + will pop these operands
from the stack, will add them and push the result back on the stack. This result
becomes an operand. Next C will go to the stack and after this * operator will pop
these two operands (result of addition and C). Their multiplication will lead to the
final result. The postfix notation is simple to evaluate as compared to the infix one. In
postfix, we need not to worry about what operation will be carried first. The operators
in this notation are in the order of evaluation. However, in the infix notation, we have
to force the precedence according to our requirement by putting parentheses in the
expression. With the help of a stack data structure, we can do the conversion and
evaluation of expressions easily.

Page 73 of 519

CS301 Data Structures

___________________________________________________________________

C++ Templates
We can use C++ templates for stack and other data structures. We have seen that
stack is used to store the operands while evaluating an expression. These operands
may be integers, floating points and even variables. We push and pop the operands to
and from the stack. In the conversion of an expression, a programmer uses the stack
for storing the operators like +, *, -, and / etc which are single characters. In both
cases, the functionality is the same. We push and pop things on and from the stack. At
times, we check if the stack is empty or not. Thus identical methods are employed
while using stack in evaluating and converting the expressions. However, there may
be a difference in the type of the elements (data) used in these cases. We may define a
stack, which can store different types of data but for the time being we are restricting
ourselves to the stack that can store elements of only one type. In C++ programming,
we will have to create two classes FloatStack and CharStack for operands and
operators respectively. These classes of stack have the same implementation. Thus,
we write the same code twice only with the difference of data type. Is there any
method to write the code for the stack once and then use it for different types of data?
This means is there any way that we can make a stack for storing integers, floating
points, characters or even objects with the same code written once. The language C++
provides us the facility of writing templates. A template can be understood with the
example of a factory that bakes biscuits. The factory may use flour, corn or starch as
ingredients of the product. But the process of baking biscuits is the same whatever
ingredients it uses. There is no difference in the machinery for producing biscuits with
different ingredients. So we call the factory as the template for the biscuits. Similarly
in C++ language, a template is a function or class that is written with a generic data
type. When a programmer uses this function or class, the generic data type is replaced
with the data type, needed to be used in the template function or in the template class.
We only give the data type of our choice while calling a template function or creating
an object of the template class. The compiler automatically creates a version of that
function or class with that specified data type. Thus if we write a template class for
stack, then later on we can use it for creating a stack for integers, floating points or
characters etc. So instead of writing code for different stacks of different data types,
we write one code as a template and reuse it for creating different stacks. We declare
the template class in a separate file in addition to the main program file. This file can
be used in our program by including it in that file. Following is the code of the
template class for stack. This is written in the file Stack.h.
template <class T>
class Stack
{
public:
Stack();
int empty(void);
// 1=true, 0=false
int push(T &);
// 1=successful,0=stack overflow
T pop(void);
T peek(void);
~Stack();
private:
int top;
T* nodes;
};

Page 74 of 519

CS301 Data Structures

___________________________________________________________________
In the above code the line
template <class T>
shows that we are going to write a template. Here T is a variable name for generic
data type. We can use any other name but generally T is used (T evolves from
template). A data type will replace this T whenever template is used in a program.
Then we declare member functions of the class. To begin with, there is a constructor
of the class with the same name as that of the class i.e. Stack. It is followed by the
empty () function and then function push that is declared as follows
int push(T &) ;
We have been using the data (element) type int or some other data type in the push()
function. Now there is T written as the data type in the push() function. This means
that the function takes an argument of type T, here T is a generic data type and we will
use a proper data type while calling this function in our program. This data type will
replace T. There are also pop and peek functions that take no arguments but return the
value which is of type T. The peek function is similar to the top function that returns
(shows) the element from the top but does not remove it from the stack.
In the private section of the class, there are two variables. The variable top is used to
point to the top of the stack. The pointer nodes is used to point to nodes of type T. We
allocate dynamic memory using this nodes pointer.
In the definition of this whole class, T is a substitution parameter. While using the
stack class in the program, we will replace this T with a proper data type.

Implementation
Now we will see how to implement this stack class in our program. Following is the
code of the program. We save this code in the file named Stack.cpp.
#include <iostream.h>
#include <stdlib.h>
#include "Stack.h"
#define MAXSTACKSIZE 50
template <class T>
Stack<T>::Stack()
{
top = -1;
nodes = new T[MAXSTACKSIZE];
}
template <class T>
Stack<T>::~Stack()
{
delete nodes;
}

Page 75 of 519

CS301 Data Structures

___________________________________________________________________
template <class T>
int Stack<T>::empty(void)
{
if( top < 0 ) return 1;
return 0;
}
template <class T>
int Stack<T>::push(T& x)
{
if( top < MAXSTACKSIZE )
{
nodes[++top] = x;
return 1;
}
cout << "stack overflow in push.\n";
return 0;
}
template <class T>
T Stack<T>::pop(void)
{
T x;
if( !empty() )
{
x = nodes[top--];
return x;
}
cout << "stack underflow in pop.\n";
return x;
}
In this code, we include different files in which one is Stack.h , written earlier to
declare the template class Stack. We have defined a constant size of the stack to 50 by
writing the line
#define MAXSTACKSIZE 50
Next is the definition of the constructor. Here before the signature of the constructor
function, we have written
template <class T>
Stack<T> :: Stack() ;
This means that we are writing a template function that uses T wherever a data type is
written. As we have declared the Stack class as a template class, <T> will be written
with the class name before the access specifier (i.e. ::) while defining a method of the
class. Then there is the implementation of the constructor in which we assign the
value 1 to top and allocate memory by using the new operator for stack of size
MAXSTACKSIZE of type T and put its starting address in the pointer nodes. It is
pertinent to note that we create an array of type T (i.e. a generic type).

Page 76 of 519

CS301 Data Structures

___________________________________________________________________
Similarly we define the destructor ~Stack, which frees the memory by deleting the
nodes. Then there are the different methods of the stack i.e. empty(), push(), and
pop(). We define all the methods in the same way as done in case of the constructor. It
means that while writing template <class T> at the start, we use T wherever a data
type can be used. The function empty() returns a Boolean parameter. It returns 1 that
means TRUE if the stack is empty i.e. if top is less than 0. Otherwise it returns zero
i.e. FALSE if the stack is not empty.
We define the push function as under
int Stack<T>::push(T& x)
{
if( top < MAXSTACKSIZE )
{
nodes[++top] = x;
return 1;
}
cout << "stack overflow in push.\n";
return 0;
}
This function takes an argument x by reference. It checks whether there is space in the
stack by checking the value of the top. If top is less than the MAXSTACKSIZE, it
means there is space available in the stack. Then it puts the element x on the stack and
returns 1, indicating that push operation has succeeded. Otherwise, it displays a
message of stack overflow and returns zero which indicates that the element was not
put on the stack.
Next comes the pop method. Here we see that the value returned by this method is of
type T. In the body of the function, we define a local variable of type T and check if
the stack is empty. If it is not empty, we pop the value from the top in variable x and
return it.
Now lets see the use of this template stack. Here we write the main program in a
separate file including the stack.cpp (we have written before shortly) to use the stack.
Following is the program written in the file main.cpp. This program demonstrates the
implementation of the stack.
#include "Stack.cpp"
int main(int argc, char *argv[])
{
Stack<int> intstack;
Stack<char> charstack;
int x=10, y=20;
char c='C', d='D';
intstack.push(x); intstack.push(y);
cout << "intstack: " << intstack.pop() << ", " << intstack.pop() << "\n";
charstack.push(c); charstack.push(d);
cout << "charstack: " << charstack.pop() << ", " << charstack.pop() << "\n";
}

Page 77 of 519

CS301 Data Structures

___________________________________________________________________
In the above code, consider the line
Stack <int> intstack ;
This line means that while creating an object intstack of Stack, the generic data type T
should be replaced by the type int, In other words, it will be a stack for integers. The
compiler will replace T with int wherever it exists in the code, providing a version of
code with data type int. The compiler does this automatically.
Similarly the next line
Stack <char> charstack ;
creates an object of Stack that has name charstack and replaces the type T with char.
It shows that it will be a stack of characters. Here T is replaced with char and a
version of code is provided by the compiler, used for the char data type. Thus we
create two objects of two types (i.e. int and char) of Stack by using the same code of
the template class.
To demonstrate the implementation of Stack, we declare two variables of type int and
two of type char. Then we push and pop these variables on and from the proper stack.
We push the int values on the intstack (which we have created for int data type). The
values of type other than int on the stack intstack can not be pushed as we have
created it to store the int data type. And then we poop these values from the stack and
show on the screen. Similarly we push the char values on the charstack (a stack to
store char values) before displaying these values on the screen with the help of the
pop method to get these values from the stack.
Now we have the three files of our code i.e. stack.h, stack.cpp and main.cpp. Having
these files in the same directory, we compile the main file (main.cpp) and execute it.
Following is the output of the above program.
intstack: 10, 20
charstack: C, D
In the above example, we create two objects of Stack to use for two different data
types. The compiler automatically provides us two versions of the template code, one
for int and the other for char using the same code, written as template. So it is only
due to the use of the template utility, provided by the C++ language only. No other
language including C provides this utility of templates. If we write this program in C,
the code of the functions of Stack has to be written repeatedly for each data type.
Similarly, we will have to write different versions for using different data types. But
in templates, a programmer writes the code once and the compiler automatically
produces the version of the code with the needed data type. We have implemented the
stack by using array. It can also be implemented with the linked list. The use of array
or linked list does not matter here. The implementation of the stack remains the same.
The templates are so important that C++ provides a library in which a large number of
common use functions are provided as templates. This library is a part of the official
standard of C++. It is called STL i.e. Standard Template Library. As a library, it is a
tested code base. We can use these templates and implement different concepts for
our own data types. STL is an important code, pre-developed for us. It is available as
a library. Different data structures like stack, queue etc is also there in STL. We can
write programs by using them. But here in this course, our goal is to know what the
data structures are, what is functioning and how can they be written? So we are

Page 78 of 519

CS301 Data Structures

___________________________________________________________________
writing and discussing the stack templates. You can use data structures from STL in
the programming courses or in the professional life. You need not to write the stack or
queue from the scratch you can simply use them in your programs from STL.

Function Call Stack

Lets talk about another example of the stack. We know the functionality of the
function calls. Whenever a programmer calls a function, he or she passes some
arguments or parameters to the function. The function does work on these arguments
and returns a value to the calling function or program. This value is known as the
return value of the function. We declare some variables inside the function which are
local variables of the function. These variables are demolished when the execution of
the function ends. If there are variables in the function that need to be preserved, we
have to take care of them. For this purpose, we use global variables or return a pointer
to that variable. Now lets see how a stack is used in function calls.
We are using devC++ compiler that actually uses GCC (glue compiler), a public
domain compiler. Whenever we call a function, the compiler makes a stack that it
uses to fulfill this function call. The compiler puts the entries on the stack in the way
that first of all i.e. on the top (i.e. first entry in the stack) is the return address of the
function where the control will go back after executing the function. After it, the next
entries on the stack are the arguments of the function. The compiler pushes the last
argument of the call list on the stack. Thus the last argument of the call list goes to the
bottom of the stack after the return address. This is followed by the second last
argument of the call list to be pushed on the stack. In this way, the first argument of
the call list becomes the first element on the stack. This is shown in the following
figure.
last argument

second argument
first argument
top ------> return address
In the calling function, after the execution of the function called, the program
continues its execution form the next line after the function call. The control comes
back here because when the execution of the function ends the compiler pops the
address from the stack which it has pushed when the function call was made. Thus the
control goes at that point in the program and the execution continues in the calling
function or program.
Consider the following code of a function that takes two integer arguments a, b and
returns the average of these numbers.
int i_avg (int a, int b)
{
return (a + b) / 2;
}
Page 79 of 519

CS301 Data Structures

___________________________________________________________________
To understand the use of stack, look at the assembly language code of the above
function that is written as under.
globl _i_avg
_i_avg:
movl 4(%esp), %eax
addl 8(%esp), %eax
sarl $1, %eax
ret

# Add the args

# Divide by 2
# Return value is in %eax

The first statement is globl_i_avg which shows that its a global function that can be
called by other functions or programs. After it, there is a label, written as _i_avg:
The next statement is movl 4(%esp), %eax. Here in this statement, there is the use of
stack. Here esp is a register in assembly language that is now a stack pointer for us
(i.e. top). The movl (move long) takes offset 4 from top (4 is number of bytes, we use
4 bytes as in C++ an integer is of 4 bytes.) that means after 4 bytes from the top in the
stack it gets the value and put it in the eax register. We know that the compiler pushes
the arguments of the function in reverse order on the stack. And pushes return address
at the end. Thus the order of stack will be that on the top will be the return address
and immediately after it will be the first argument. Here in the assembly code
generated by the compiler, the compiler pops first argument from offset 4 and puts it
in eax register. The next statement is
addl 8(%esp), %eax
The addl takes offset 8 from the stack pointer that is second argument and adds it to
eax. Thus in the previous two statements, the compiler got the first argument i.e. a
from the stack and the second argument b from the stack, added them before putting
the result in eax. The next statement
sarl $1, %eax
is the division statement of assembly language. This statement divides the value in
eax by 2 and thus eax has the resultant value. The last statement i.e. ret, returns the
value on the top of the stack to the caller function.
So we have seen the use of stack in the execution of a function and how the
arguments are passed to the function, how the functions return its return value and
finally how the control goes back to the caller function .All this process is executed
by using a stack. All the things about the functionality of the function calls are
necessary to understand as these will be needed while going to write our own
compilers. We will read this in the compilers course. The whole process we have
discussed about the use of stack in function calling is known as run time environment.
Different data structures are also used in run time environment of the computer. We
know that an executable program while in run, is loaded in the memory and becomes
a process. This process is given a block of memory which it uses during its execution.
Even the operating system, in which we are working, itself takes memory. Suppose
we are running many programs simultaneously, which for example include browser,
MS Word, Excel and dev-C++. We can also run programs written by us. Every

Page 80 of 519

CS301 Data Structures

___________________________________________________________________
program which we run takes a block of memory and becomes a process. The
following figure shows a part of memory in which different programs occupy a block
of memory for their execution.
Process 1
(browser)
Process 3
(Word)
Process 4
(Excel)
Process 2
(Dev-C++)
Windows Os

We can also see the details of all the programs running at a specific time. If we press
the key combination Ctrl-Alt-Del, there appears a window task manager on the
screen. Following is the figure of the task manager. In the figure, there are many
columns i.e. PID (process ID), CPU, CPU time, Memory usage, page faults, I/O
Reads, I/O Writes, I/O Read Bytes. These all things we will read in the course of
Operating Systems in detail.

Page 81 of 519

CS301 Data Structures

___________________________________________________________________

Here the thing of our interest is the first, second and fifth column. These columns are
Image Name, PID and Mem Usage (i.e. memory usage). Now look at the row where
explorer.exe is written in the first column. The process ID (PID) of it is 888 and
memory usage is 4696K. This means that the process size of explorer.exe in the
memory is 4696 Kilo Bytes (KB). All the processes in the first column of the task
manager are present in the memory of the computer at that time. The column Image
name has the names of the processes being executed. These have extension .exe but
there may be other executable programs that have extension other than .exe.

Page 82 of 519

CS301 Data Structures

___________________________________________________________________
The following figure shows the internal memory organization of a process.

Process 1
(browser)

Code

Process 3
(Word)

Static data

Process 4
(Excel)

Stack

Process 2
(Dev-C++)
Windows Os

Heap

This shows that the first part of the memory of the process is for the code. This is the
code generated by the compiler of C++, JAVA or VB etc with respect to the language
in which the actual code was written. Then the static data of the program occupies the
memory. This holds the global variables and different variables of objects. Then in the
memory of the process, there is stack. After it there is heap. The stack and heap are
used in function calls. We have discussed the use of stack in function calls. When we
allocate memory dynamically in our programs, it is allocated from the heap. The use
of heap is a topic related to some programming course or to the operating system
course.

Page 83 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 09
Reading Material
Data Structures and Algorithm Analysis in C++
Chapter. 3
3.3.3, 3.4.1, 3.4.2

Summary

Memory Organization
Stack Layout During a Function Call
Queues
Queue Operations
Implementing Queue
Queue using Array
Use of Queues

Memory Organization
By the end of last lecture, we discussed the uses of stack to develop a process from an
executable file and then in function calls. When you run an executable, the operating
system makes a process inside memory and constructs the followings for that purpose.
- A code section that contains the binary version of the actual code of the program
written in some language like C/C++
- A section for static data including global variables
- A stack and
- Finally, a heap

Page 84 of 519

CS301 Data Structures

___________________________________________________________________
Stack is used in function calling while heap area is utilized at the time of memory
allocation in dynamic manner.
Process 1
(Browser)

Code

Process 3
(Word)

Static Data

Process 4
(Excel)

Stack

Process 2
(Dev-C++)

Heap

Windows OS

Fig 1. Memory Organization

Stack Layout during a Function Call

Parameters (F)

Parameters (F)

Local variables (F)

Local variables (F)

Return address (F)

Return address (F)

Parameters (G)
sp
sp

Parameters (G)
Local variables (G)
Return address (G)

Parameters (F)
Local variables (F)
Return address (F)
sp

Fig 2: Stack Layout; When function F calls function G

The above diagrams depict the layout of the stack when a function F calls a function
G. Here sp stands for stack pointer. At the very left, you will find the layout of the
stack just before function F calls function G. The parameters passed to function F are

Page 85 of 519

CS301 Data Structures

___________________________________________________________________
firstly inserted inside the stack. These are followed by the local variables of the
function F and finally the memory address to return back after the function F finishes.
Just before function is made to the function G, the parameters being passed to the
function G, are inserted into the stack.
In the next diagram, there is layout of the stack on the right side after the call to the
function G. Clearly, the local variables of the function G are inserted into the stack
after its parameters and the return address. If there are no local variables for a
function, the return address is inserted (pushed) on to the stack.
The layout of the stack, when the function G finishes execution is shown on the right.
You can see that the local variables of function G are no more in the stack. They have
been removed permanently along with the parameters passed to the function G. Now,
it is clear that when a function call is made, all local variables of the called function
and the parameters passed to it, are pushed on to the stack and are destroyed, soon
after the the completion of the called functions execution.
In C/C++ language, the variables declared as static are not pushed on the stack.
Rather, these are stored in another separate section allocated for static data of a
program. This section for global or static data can be seen in the fig 1 of this lecture.
It is not destroyed till the end of the processs execution. If a variable, say x is
declared as static inside function G, x will be stored in the static data section in the
processs memory. Whereas, its value is preserved across G function calls. The
visibility of x is restricted to the function G only. But a static variable declared as a
class data is available to all member functions of the class and a static variable
declared at global scope (outside of any class or function body) is available to all
functions of the program.
Now, lets move on to another data structure called queue.

Queues
A queue is a linear data structure into which items can only be inserted at one end and
removed from the other. In contrast to the stack, which is a LIFO (Last In First Out)
structure, a queue is a FIFO (First In First Out) structure.
The usage of queue in daily life is pretty common. For example, we queue up while
depositing a utility bill or purchasing a ticket. The objective of that queue is to serve
persons in their arrival order; the first coming person is served first. The person, who
comes first, stands at the start followed by the person coming after him and so on. At
the serving side, the person who has joined the queue first is served first. If the
requirement is to serve the people in some sort of priority order, there is a separate
data structure that supports priorities. The normal queue data structure, presently
under discussion, only supports FIFO behavior.
Now, lets see what are the operations supported by the queue.

Queue Operations
The queue data structure supports the following operations:
Operation
enqueue(X)
dequeue()

Description
Place X at the rear of the queue.
Remove the front element and return it.

front()

Return front element without removing it.

Page 86 of 519

CS301 Data Structures

___________________________________________________________________
isEmpty()

Return TRUE if queue is empty, FALSE otherwise

Page 87 of 519

CS301 Data Structures

___________________________________________________________________

Implementing Queue
There are certain points related to the implementation of the queue. Suppose we are
implementing queue with the help of the linked -list structure. Following are the key
points associated with the linked list implementations:
- Insert works in constant time for either end of a linked list.
- Remove works in constant time only.
- Seems best that head of the linked list be the front of the queue so that all removes
will be from the front.
- Inserts will be at the end of the list.

The above figure shows queue elements on the left with two pointers front and rear.
This is an abstract view of the queue, independent of its implementation method of
array or linked list. On the right side is the same queue ,using linked list and pointers
of front and rear. When dequeue() function is called once, the front element 1 is
removed. The picture of the queue showing one element removal is also depicted
below. Note that front pointer has been moved to the next element 7 in the list afer
removing the front element 1.
After dequeue() is called once

Now at this stage of the queue, we will call enqueue (9) to insert an element 9 in it. .
The following figure shows that the new element is inserted at the rear end and rear
pointer starts pointing this new node with element 9.
At this point of time, the code of these functions of dequeue() and enqueue() should
not be an issue.

Page 88 of 519

CS301 Data Structures

___________________________________________________________________

Note that in this queue data structure, the new elements are inserted at rear end and
removed from the front. This is in contrast to stack structure where the elements are
inserted and removed from the same end.
Lets see the code for queue operations:
/* Remove element from the front */
1. int dequeue()
2. {
3. int x = front->get();
4.
Node* p = front;
5.
front = front->getNext();
6.
delete p;
7.
return x;
8. }
/* Insert an element in the rear */
9. void enqueue(int x)
10. {
11.
Node* newNode = new Node();
12.
newNode->set(x);
13.
newNode->setNext(NULL);
14.
rear->setNext(newNode);
15.
rear = newNode;
16. }
In dequeue() operation, at line 3, the front element is retrieved from the queue and
assigned to the int variable x.
In line 4, the front pointer is saved in Node pointer variable p.
In line 5, the front pointer is moved forward by retrieving the address of the next node
by using front->getNext() and assigning it to the front pointer.
In line 6, the node pointed to by the front pointer is deleted by using delete front
statement.
At the end of dequeue() implementation, the value of deleted node that was saved in
the int variable x, is returned back.
The enqueue(int ) is used to add an element in the queue. It inserts the element in the
rear of the queue. At line 11, a new Node object is created using the new Node()
statement and the returned starting address of the created object is assigned to the
newNode pointer variable.

Page 89 of 519

CS301 Data Structures

___________________________________________________________________
In line 12, the value of the passed in parameter x, is set in the newly created node
object using the set() method.
In line 13, the next pointer in the newly created node is set to NULL.
In line 14, the newly created node is set as the next node of the node currently pointed
by the rear pointer.
Ine line 15, the rear pointer is set to point to the newly created node.
The code of two smaller functions is as under:
/* To retrieve the front element */
int front()
{
return front->get();
}
/* To check if the queue is empty */
int isEmpty()
{
return ( front == NULL );
}
The front() method is used to retrieve the front element. This is the oldest element
inserted in the queue. It uses the get() method of the Node class.
The isEmpty() method is used to check whether the queue is empty or not. It checks
the address inside the front pointer, if it is NULL. It will return true indicating that the
queue is empty or vice versa.
While studying stack data structure, we implemented it by using both array and
linked list. For queue, until now we have been discussing about implementing queue
using linked list. Now, lets discuss implementing queue with the help of an array.

Queue using Array

A programmer keeps few important considerations into view account before
implementing a queue with the help of an array:
If we use an array to hold the queue elements, both insertions and removal at the front
(start) of the array are expensive. This is due to the fact that we may have to shift up
to n elements.
For the stack, we needed only one end but for a queue, both are required. To get
around this, we will not shift upon removal of an element.

Page 90 of 519

CS301 Data Structures

___________________________________________________________________
front

rea

front

rear

Fig 6. Queue implemented using an array

In the above figure, queue implementation using array is shown. As the array size is
8, therefore, the index of the array will be from 0 to 7. The number of elements inside
array are 1, 7, 5 and 2, placed at start of the array. The front and rear in this
implementation are not pointers but just indexes of arrays. front contains the starting
index i.e. 0 while rear comprises 3.
Lets see, how the enqueue() works:
enqueue(6)
rea

front

front

rear

As shown in the above diagram, an element i.e. 6 has been inserted in the queue.
Now, the rear index is containing 4 while the front has the same 0 index. Lets see
the figure of the array when another element 8 is inserted in the queue.

front

enqueue(8)
rea

1 7 5 2 6 8

front

rear

When an element is removed from the queue. It is removed from the front index.
dequeue( )
front
rea
7

7 5 2 6 8
0

5
2
front

2
3

6
4

8 Page 91 of 519
5

6
rear

CS301 Data Structures

___________________________________________________________________

After another call of dequeue() function:

dequeue( )
rea
front
5

5 2 6 8
0

front

rear

With the removal of element from the queue, we are not shifting the array elements.
The shifting of elements might be an expensive exercise to perform and the cost is
increased with the increase in number of elements in the array. Therefore, we will
leave them as it is.
enqueue(9)
enqueue(12)
front

rea
5

5 2 6 8 9 12
0

6
4

front

rear

12
7

After insertion of two elements in the queue, the array that was used to implement it,
has reached its limit as the last location of the array is in use now. We know that there
is some problem with the array after it attained the size limit. We observed the similar
problem while implementing a stack with the help of an array.
We can also see that two locations at the start of the array are vacant. Therefore, we
should can consider how to use those locations appropriately in to insert more

Page 92 of 519

CS301 Data Structures

___________________________________________________________________
elements in the array.
Although, we have insert and removal operations running in constantly, yet we
created a new problem that we cannot insert new elements even though there are two
places available at the start of the array. The solution to this problem lies in allowing
the queue to wrap around.
How can we wrap around? We can use circular array to implement the queue. We
know how to make a linked list circular using pointers. Now we will see how can we
make a circular array.
0
front
5

front

rea
2

12

12

2
8

Fig 12. Circular array to implement queue

rear
7

The number of locations in the above circular array are also eight, starting from index
0 to index 7. The index numbers are written outside the circle incremented in the
clock-wise direction. To insert an element 21 in the array , we insert this element in
the location, which is next to index 7.
enqueue(21)
front

0
rea

5 2 6 8 9 12 21
6
Fig 13. An element added in circular array

1
front

21
12

2
8
5

6
4

size

rear

noElements

Now, we will have to maintain four variables. front has the same index 2 while the,
size is 8. rear has moved to index 0 and noElements is 7. Now, we can see that rear
index has decreased instread of increasing. It has moved from index 7 to 0. front is
containing index 2 i.e. higher than the index in rear. Let see, how do we implement
the enqueue() method.
void enqueue( int x)
{
1. rear = (rear + 1) % size;
2. array[rear] = x;
3. noElements = noElements + 1;
}
In line 1 of the code, 1 is added in rear and the mod operator (that results in

Page 93 of 519

CS301 Data Structures

___________________________________________________________________
remainder of the two operands) is applied with size variable. This expression on the
right of assignment in line 1 can result from 0 to 7 as size is containing value 8. This
operator ensures that value of this expression will always be from 0 to 7 and increase
or decrease from this. This resultant is assigned to the rear variable.
In line 2, the x (the value passed to enqueue() method to insert in the queue) is
inserted in the array at the rear index position. Therefore, in the above case, the new
element 21 is inserted at index 0 in the array.
In line 3, noElements is added to accumulate another element in the queue.
Lets add another element in the queue.
enqueue(7)
rear

front

0
7

5 2 6 8 9 12 21 7
6

1
front

21
12

2
8

5
Fig 14. Another element added in circular array

size

rear

noElements

Now, the queue, rather the array has become full. It is important to understand, that
queue does not have such characteristic to become full. Only its implementation array
has become full. To resolve this problem, we can use linked list to implement a
queue. For the moment, while working with array, we will write the method isFull(),
to determine the fullness of the array.
int isFull()
{
return noElements == size;
}
int isEmpty()
{
return noElements == 0;
}
isFull() returns true if the number of elements (noElements) in the array is equal to the
size of the array. Otherwise, it returns false. It is the responsibility of the caller of the
queue structure to call isFull() function to confirm that there is some space left in the
queue to enqueue() more elements.
Similarly isEmpty() looks at the number of elements (noElements) in the queue. If
there is no element, it returns true or vice versa..
Lets see the dequeue() method.
0

dequeue()
front

rea

6 8 9 12 21 7
6

1
21

front

12
9
8
5

6
4

size
8

Page 94 of 519
rear noElements
1

CS301 Data Structures

___________________________________________________________________

int dequeue()
{
int x = array[front];
front = (front + 1) % size;
noElements = noElements - 1;
return x;
}
In the first line, we take out an element from the array at front index position and
store it in a variable x. In the second line, front is incremented by 1 but as the array is
circular, the index is looped from 0 to 7. That is why the mod (%) is being used. In the
third line, number of elements (noElements) is reduced by 1 and finally the saved
array element is returned.

Use of Queues
We saw the uses of stack structure in infix, prefix and postfix expressions. Lets see
the usage of queue now.
Out of the numerous uses of the queues, one of the most useful is simulation. A
simulation program attempts to model a real-world phenomenon. Many popular video
games are simulations, e.g., SimCity, Flight Simulator etc. Each object and action in
the simulation has a counterpart in the real world. Computer simulation is very
powerful tool and it is used in different high tech industries, especially in engineering
projects. For example, it is used in aero plane manufacturing. Actually Computer
Simulation is full-fledged subject of Computer Science and contains very complex
Mathematics, sometimes. For example, simulation of computer networks, traffic
networks etc.
If the simulation is accurate, the result of the program should mirror the results of the
real-world event. Thus it is possible to understand what occurs in the real-world
without actually observing its occurrence.
Let us look at an example. Suppose there is a bank with four tellers.
A customer enters the bank at a specific time (t1) desiring to conduct a transaction.
Any one of the four tellers can attend to the customer. The transaction (withdraws,
deposit) will take a certain period of time (t2). If a teller is free, the teller can process
the customers transaction immediately and the customer leaves the bank at t1+t2. It is
possible that none of the four tellers is free in which case there is a line of customers
at each teller. An arriving customer proceeds to the back of the shortest line and waits
for his turn. The customer leaves the bank at t2 time units after reaching the front of
the line.
The time spent at the bank is t2 plus time waiting in line.
So what we want to simulate is the working environment of the bank that there are

Page 95 of 519

CS301 Data Structures

___________________________________________________________________
specific number of queues of customers in the bank in front of the tellers. The tellers
are serving customers one by one. A customer has to wait for a certain period of time
before he gets served and by using simulation tool, we want to know the average
waiting time of a bank customer. We will talk about this simulation in the next lecture
and will do coding also in order to understand it well.

Page 96 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 10
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 3, 6
3.4.3, 6.1

Summary
12)
13)
14)
15)

Queues
Simulation Models
Priority Queue
Code of the Bank simulation

Queues
In the previous lecture, we discussed the queue data structure and demonstrated its
implementation by using array and link list. We also witnessed the usefulness of
queue as data structure in the simulation example. This concept can be further
elaborated with a daily life example relating to banking sector. Suppose, customers
want to deposit or withdraw money from a bank, having four cashiers or tellers. The
teller helps you in depositing the money or withdrawing the money from the same
window. This window is known as teller window. The customer needs some service
from the bank like depositing the money, bill etc. This transaction needs some time
that may be few minutes. A person enters the bank and goes to the teller who is free
and requests him to do the job. After the completion of the transaction, the person
goes out of the bank. Now we will discuss a scenario when there is a lot of rush of
customers. The tellers are just four. Now the tellers are busy and the new customers
will form a queue. In this example, we need a queue in front of each of the tellers. A
new customer enters the bank and analyzes the four queues and wants to join the
shortest queue. This person has to wait for the persons in front of him to be served.
Another person may come behind him. In this simulation, we will restrict the person
from changing the queue. A person comes into the bank at 10 O clock. His transaction
time is 5 minutes. He has to wait for another fifteen minutes in the queue. After this,
the teller serves him in 5 min. This person comes at 10 am and waits for fifteen
minutes. As the transaction time is 5 minutes, so he will leave the bank at 1020. Now
this is the situation of simulation and we have to write a program for this. We can go
to some bank and analyze this situation and calculate the time. At the end of the day,
we can calculate the average time for each of the customer. This time can be 30
minutes. Here we will simulate this situation with the help of a computer program.
This is the real life example. Lets see the picture of simulations to understand what is
happening in the bank.
In the picture below, we have four tellers and four queues, one for each of the tellers.
Each teller is serving a customer. When the transaction of a customer is completed, he

Page 97 of 519

CS301 Data Structures

___________________________________________________________________
will leave the bank.

teller
1

teller
2

teller
3

teller
4

A person enters the bank. He sees that all the four tellers are busy and in each queue
there are two persons waiting for their turn. This person chooses the queue no. 3.
Another person enters the bank. He analyzed all the queues. The queue no 3 is the
biggest and all other are having 2 persons in the queue. He chooses the queue no 1.

teller
1

teller
2

teller
3

teller
4

Now we have three persons waiting in queue no 1 and 3 and two persons waiting in
queue no 2 and 4. The person in queue no.1 completes his transaction and leaves the
bank. So the person in the front of the queue no. 1 goes to the teller and starts his
transaction. Similarly the person at queue No. 3 finishes his transaction and leaves the
premises. The person in front of queue number 3 goes to the teller.
Another person enters the bank and goes to the queue No. 1. This activity goes on.
Page 98 of 519

CS301 Data Structures

___________________________________________________________________
The queues become bigger and shorter. The persons coming in the bank have to wait.
If the queues are shorter, people have to wait for less time. However, if the queues are
longer, people have to wait for more time. The transactions can also take much more
time, keeping the people waiting. Suppose four persons come with big amount and
their transaction takes too much time. These are all parameters which we can
incorporate in our simulation making it real. For this, we have carry out more
programming. With the introduction of these parameters in the simulation, it will be
more close to the real life situation. Simulation, being a very powerful technique, can
yield the results, very close to some real life phenomenon.

Simulation Models
Lets discuss little bit about the simulation models. Two common models of
simulation are time-based simulation and event-based simulation. In time-based
simulation, we maintain a timeline or a clock. The clock ticks and things happen
when the time reaches the moment of an event.
Suppose we have a clock in the computer. The minute hand moves after every minute.
We know the time of the customers entry into the bank and are aware that his
transaction takes 5 minutes. The clock is ticking and after 5 minutes, we will ask the
customer to leave the bank. In the program, we will represent the person with some
object. As the clock continues ticking, we will treat all the customers in this way.
Note that when the customer goes to some teller, he will take 5 minutes for his
transaction. During this time, the clock keeps on ticking. The program will do nothing
during this time period. Although some other customer can enter the bank. In this
model, the clock will be ticking during the transaction time and no other activity will
take place during this time. If the program is in some loop, it will do nothing in that
loop until the completion of the transaction time.
Now consider the bank example. All tellers are free. Customer C1 comes in 2 minutes
after the opening of the bank. Suppose that bank opens at 9:00 am and the customer
arrives at 9:02 am. His transaction (withdraw money) will require 4 minutes.
Customer C2 arrives 4 minutes after the bank opens (9:04 am). He needs 6 minutes
for transaction. Customer C3 arrives 12 minutes after the bank opens and needs 10
minutes for his transaction.
We have a time line and marks for every min.
Time (minutes)
0

C1 in

10

11

12

13

14

15

C1 out
C2 in

C2 out
C3 in

C1 comes at 2 min, C2 enters at 4 min. As C1 needs 4 min for his transaction, so he

leaves at 6 min. C2 requires 6 min for the processing so C2 leaves at 10 min. Then C3
Page 99 of 519

CS301 Data Structures

___________________________________________________________________
enters at 12 min and so on. This way, the activity goes on. Therefore, we can write a
routine of the clock. We take a variable clock representing the clock. The clock will
run for 24 hrs. Banks are not open for 24 hrs but we will run our loop for 24 hrs. The
pseudo code is as under:
clock = 0;
while ( clock <= 24*60 ) { // one day
read new customer;
if customer.arrivaltime == clock
insert into shortest queue;
check the customer at head of all four queues.
if transaction is over
remove from queue.
clock = clock + 1;
}
The variable clock is initialized to zero. The while loop runs for 24 hrs. Then we read
a new customer. This information may be coming from some file. The if statement is
checking the arrival time of the customer. He comes 10 minutes after the opening of
the bank. So when this time is equal to the clock, we insert the customer in the
shortest queue. Then we will check the transaction time of all the four customers at
each teller. If the transaction time of any customer ends, we will remove it from the
queue and he will leave the bank. In the end, we increment the clock with one minute.
As seen in the above statements, some activity takes place when the clock reaches at
the event time (that is the time to enter the bank or leave the bank arrives). If the
customers arrival time has not come, the first if statement becomes false and we do
nothing. Similarly if the transaction of customer is not finished, the second if
statement becomes false. Then this while loop will simply add one min to the clock.
This is the clock- based (time- based) simulation.
Lets discuss the other type of simulation i.e. the event-based simulation. Dont wait
for the clock to tick until the next event. Compute the time of next event and maintain
a list of events in increasing order of time. Remove an event from the list in a loop
and process it. Lets see the time line again.

Page 100 of 519

CS301 Data Structures

___________________________________________________________________
Time (minutes)
0

C1 in

10

11

12

13

14

15

C1 out
C2 in

Event 1:
2 mins C1 in
Event 2: 4 mins C2 in
Event 3: 6 mins C1 out
Event 4: 10 mins C2 out
Event 5: 12 mins C3 in

C2 out
C3 in

The customer C1 comes at 2 min and leaves at 6 min. Customer C2 comes at 4 min
and leaves at 10 min and so on. We have written the events list in the above figure.
Do not see the clock but see the events on time. Event 1 occurs at 2 min that is the
customer C1 enters the bank 2 minutes after its opening. Event 2 is that C2 enters at 4
min. Event 3 is that the customer C1 leaves the bank at 6 min. Event 4 is that the C2
leaves the bank at 10 min and event 5 is that C3 enters the bank at 12 min. Here we
have a list of events. How can we process them? We will make a queue of these
events. Remove the event with the earliest time from the queue and process it. Insert
the newly created events in the queue. A queue where the de-queue operation depends
not on FIFO, is called a priority queue.

Priority Queue
As stated earlier, the queue is a FIFO (First in first out) structure. In daily life, you
have also seen that it is not true that a person, who comes first, leaves first from the
queue. Lets take the example of traffic. Traffic is stopped at the signal. The vehicles
are in a queue. When the signal turns green, vehicles starts moving. The vehicles
which are at the front of the queue will cross the crossing first. Suppose an ambulance
comes from behind. Here ambulance should be given priority. It will bypass the queue
and cross the intersection. Sometimes, we have queues that are not FIFO i.e. the
person who comes first may not leave first. We can develop such queues in which the
condition for leaving the queue is not to enter first. There may be some priority. Here
we will also see the events of future like the customer is coming at what time and
leaving at what time. We will arrange all these events and insert them in a priority
queue. We will develop the queue in such a way that we will get the event which is
going to happen first of all in the future. This data structure is known as priority
queue. In a sense, FIFO is a special case of priority queue in which priority is given to
the time of arrival. That means the person who comes first has the higher priority
while the one who comes later, has the low priority. You will see the priority queue
being used at many places especially in the operating systems. In operating systems,
we have queue of different processes. If some process comes with higher priority, it
will be processed first. Here we have seen a variation of queue. We will use the
priority queue in the simulation. The events will be inserted in the queue and the
event going to occur first in future, will be popped.

Page 101 of 519

CS301 Data Structures

___________________________________________________________________
What are the requirements to develop this simulation? We need the C++ code for the
simulation. There will be a need of the queue data structure and obviously, the
priority queue. Information about the arrival of the customers will be placed in an
input file. Each line of the file contains the items (arrival time, transaction duration).
Here are a few lines from the input file.
00 30 10 <- customer 1
00 35 05 <- customer 2
00 40 08
00 45 02
00 50 05
00 55 12
01 00 13
01 01 09
The first line shows the customer 1. 00 30 10 means Customer 1 arrives 30 minutes
after the opening of the bank. He will need 10 minutes for his transaction. The last
entry 01 01 09 means customer arrives one hour and one minute after the bank
opened and his transaction will take 9 minutes and so on. The file contains similar
information about the other customers. We will collect the events now. The first event
to occur is the arrival of the first customer. This event is placed in the priority queue.
Initially, the four teller queues are empty. The simulation proceeds as follows: when
an arrival event is removed from the priority queue, a node representing the customer
is placed on the shortest teller queue. Here we are trying to develop an algorithm
while maintaining the events queue.
After the opening of the bank, the arrival of the first customer is the first event. When
he enters the bank all the four tellers are free. Suppose he goes to the first teller and
starts his transaction. After the conclusion of his transaction, he leaves the bank. With
respect to events, we have only two events, one is at what time he enters the bank and
other is at what time he leaves the bank. When other customers arrive, we have to
maintain their events.
If the customer is the only one on a teller queue, an event for his departure is placed
on the priority queue. At the same time, the next input line is read and an arrival event
is placed in the priority queue. When a departure event is removed from the event
priority queue, the customer node is removed from the teller queue. Here we are
dealing with the events, not with the clock. When we come to know that a person is
coming at say 9:20am, we make an event object and place it in the priority queue.
Similarly if we know the time of leaving of the customer from the bank, we will make
an event and insert it into the priority queue. When the next customer in the queue is
served by the teller, a departure event is placed on the event priority queue. When the
other customer arrives, we make an event object and insert it into the priority queue.
Now the events are generated and inserted when the customer arrives. But the dequeue is not in the same fashion. When we de-queue, we will get the event which is
going to occur first.
When a customer leaves the bank, the total time is computed. The total time spent by

Page 102 of 519

CS301 Data Structures

___________________________________________________________________
the customer is the time spent in the queue waiting and the time taken for the
transaction. This time is added to the total time spent by all customers. At the end of
the simulation, this total time divided by the total customers served will be average
time consumed by customers. Suppose that 300 customers were served, then we will
divide the total time by 300 to get the average time. So with the help of simulation
technique, we will get the result that x customers came today and spent y time in the
bank and the average time spent by a customer is z.

Code of the Bank Simulation

Lets have a look on the C+ code of this simulation.
#include <iostream>
#include <string>
#include <strstream.h>
#include "Customer.cpp"
#include "Queue.h"
#include "PriorityQueue.cpp"
#include "Event.cpp"
Queue q[4];
// teller queues
PriorityQueue pq; //eventList;
int totalTime;
int count = 0;
int customerNo = 0;
main (int argc, char *argv[])
{
Customer* c;
Event* nextEvent;
// open customer arrival file
ifstream data("customer.dat", ios::in);
// initialize with the first arriving customer.
ReadNewCustomer(data);
While( pq.length() > 0 )
{
nextEvent = pq.remove();
c = nextEvent->getCustomer();
if( c->getStatus() == -1 ){ // arrival event
int arrTime = nextEvent->getEventTime();
int duration = c->getTransactionDuration();
int customerNo = c->getCustomerNumber();
processArrival(data, customerNo,
arrTime, duration , nextEvent);
}
else { // departure event
Page 103 of 519

CS301 Data Structures

___________________________________________________________________
int qindex = c->getStatus();
int departTime = nextEvent->getEventTime();
processDeparture(qindex, departTime, nextEvent);
}
}
We have included lot of files in the program. Other than the standard libraries, we
have Customer.cpp, Queue.h, PriorityQueue.cpp and Event.cpp. With the help of
these four files, we will create Customer object, Queue object, PriorityQueue object
and Event object. You may think that these are four factories, creating objects for us.
As there are four tellers, so we will create equal number of queues (Queue q[4]).
Then we create a priority queue object pq from the PriorityQueue factory. We declare
totalTime, count and customerNo as int. These are global variables.
In the main method, we declare some local variables of customer and event.
Afterwards, the customer.dat file for the input data is opened as:
ifstream data("customer.dat", ios::in);
We read the first customers data from this file as:
readNewCustomer(data);
Here data is the input file stream associated to customer.dat. We will read the arrival
time and time of transaction from the file of the first customer. After reading it, we
will process this information.
Now there is the while loop i.e. the main driver loop. It will run the simulation. First
thing to note is that it is not clock-based which is that the loop will execute for 24
hours. Here we have the condition of priority queues length. The variable pq
represents the event queue. If there are some events to be processed, the queue pq will
not be empty. Its length will not be zero. We get the next event from the priority
queue, not from the queue. The method pq.remove() (de-queue method) will give us
the event which is going to happen first in future. The priority of events is according
the time. In the event object we have the customerNo. In the if statement, we check
the status of the customer. If the status is 1, it will reflect that this is the new
customer arrival event.
We know that when a new customer enters the bank, he will look at the four tellers
and go to the teller where the queue is smallest. Therefore in the program, we will
check which is the smallest queue and insert the customer in that queue. If the event is
about the new customer, the if statement returns true. We will get its arrival time,
duration and customer number and assign it to the variables arrTime, duration and
customerNo respectively. We will call the method processArrival() and pass it the
above information.
If the status of the customer is not equal to 1, it means that the customer is in one of
the four queues. The control will go to else part. We will get the status of the

Page 104 of 519

CS301 Data Structures

___________________________________________________________________
customer which can be 0, 1, 2 and 3. Assign this value to qindex. Later on, we will see
how these values are assigned to the status. We will get the departure time of the
customer and call the processDeparture() method.
In the main driver loop, we will get the next event from the event queue. In this case,
events can be of two types i.e. arrival event and the departure event. When the person
enters the bank, it is the arrival event. If any queue is empty, he will go to the teller.
Otherwise, he will wait in the queue. After the completion of the transaction, the
customer will leave the bank. It is the departure event.
Lets discuss the function readNewCustomer(). This function is used to read the data
from the file.
void readNewCustomer(ifstream& data)
{
int hour,min,duration;
if (data >> hour >> min >> duration) {
customerNo++;
Customer* c = new Customer(customerNo,
hour*60+min, duration);
c->setStatus( -1 ); // new arrival
Event* e = new Event(c, hour*60+min );
pq.insert( e ); // insert the arrival event
}
else {
data.close(); // close customer file
}
}
Here, we have used the >> to read the hour, minute and duration from the file. Then
we create a customer object c from the customer factory with the new keyword. We
pass the customerNo, arrival time and transaction duration to the constructor of the
customer object. After the object creation, it is time to set its status to 1. This means
that it is an arriving customer. Then we create an event object e passing it the
customer c and the arrival time. We insert this event into the priority queue pq. If
there is no more data to read, we go into the else part and close the data file.
Lets see the function processArrival(). We have decided that when the customer
arrives and no teller is available, he will go to the shortest queue.
int processArrival(ifstream &data, int customerNo, int arrTime, int duration,
Event* event)
{
int i, small, j = 0;
// find smallest teller queue
small = q[0].length();
for(i=1; i < 4; i++ )
if( q[i].length() < small ){
small = q[i].length(); j = i;

Page 105 of 519

CS301 Data Structures

___________________________________________________________________
}
// put arriving customer in smallest queue
Customer* c = new Customer(customerNo, arrTime, duration );
c->setStatus(j); // remember which queue the customer goes in
q[j].enqueue(c);
// check if this is the only customer in the.
// queue. If so, the customer must be marked for
// departure by placing him on the event queue.
if( q[j].length() == 1 ) {
c->setDepartureTime( arrTime+duration);
Event* e = new Event(c, arrTime+duration );
pq.insert(e);
}
// get another customer from the input
readNewCustomer(data);
}
First of all, we will search for the smallest queue. For this purpose, there is a for loop
in the method. We will check the length of all the four queues and get the smallest
one. We store the index of the smallest queue in the variable j. Then we create a
customer object. We set its status to j, which is the queue no. Then we insert the
customer in the smallest queue of the four. The customer may be alone in the queue.
In this case, he does not need to wait and goes directly to the teller. This is the real
life scenario. When we go to bank, we also do the same. In the banks, there are
queues and everyone has to enter in the queue. If the queue is empty, the customers
go straight to the teller. Here we are trying to simulate the real life scenario. Therefore
if the length of the queue is one, it will mean that the customer is alone in the queue
and he can go to the teller. We calculate his departure time by adding the arrival time
and transaction time. At this time, the person can leave the bank. We create a
departure event and insert it into the priority queue. In the end, we read a new
customer. This is the way; a programmer handles the new customers. Whenever a
new person enters the bank, we create an event and insert it into the smallest queue. If
he is alone in the queue, we create a departure event and insert it into the priority
queue. In the main while loop, when we remove the event, in case of first future
event, it will be processed. After the completion of the transaction, the person leaves
the bank.
We may encounter another case. There may be a case that before leaving the bank,
more persons arrive and they have to wait in the queue for their turn. We handle this
scenario in the departure routine. The code is:
int processDeparture( int qindex, int departTime, Event* event)
{
Customer* cinq = q[qindex].dequeue();

Page 106 of 519

CS301 Data Structures

___________________________________________________________________
int waitTime = departTime - cinq->getArrivalTime();
totalTime = totalTime + waitTime;
count = count + 1;
// if there are any more customers on the queue, mark the
// next customer at the head of the queue for departure
// and place him on the eventList.
if( q[qindex].length() > 0 ) {
cinq = q[qindex].front();
int etime = departTime + cinq->getTransactionDuration();
Event* e = new Event( cinq, etime);
pq.insert( e );
}
}
In this method, we get the information about the qindex, departTime and event from
the main method. We get the customer by using the qindex. Then we calculate the
wait time of the customer. The wait time is the difference of departure time and the
arrival time. The total time holds the time of all the customers. We added the wait
time to the total time. We incremented the variable count by one. After the departure
of this customer, next customer is ready for his transaction. The if statement is doing
this. We check the length of the queue, in case of presence of any customer in the
queue, we will check the customer with the front() method. We set its departure time
(etime) by adding the depart time of the previous customer and his transaction time.
Then we create an event and insert it in the priority queue.
In the end, we calculate the average time in the main loop and print it on the screen.
Average time is calculated by dividing the total time to total customer.
// print the final average wait time.
double avgWait = (totalTime*1.0) / count;
cout << "Total time: " << totalTime << endl;
cout << Customer:
" << count << endl;
cout << "Average wait: " << avgWait << endl;
You may be thinking that the complete picture of simulation is not visible. How will
we run this simulation? Another important tool in the simulation is animation. You
have seen the animation of traffic. Cars are moving and stopping on the signals.
Signals are turning into red, green and yellow. You can easily understand from the
animation. If the animation is combined with the simulation, it is easily understood.
We have an animated tool here that shows the animation of the events. A programmer
can see the animation of the bank simulation. With the help of this animation, you can
better understand the simulation.
In this animation, you can see the Entrance of the customers, four tellers, priority
queue and the Exit. The customers enter the queue and as the tellers are free. They go

Page 107 of 519

CS301 Data Structures

___________________________________________________________________
to the teller straight. Customer C1<30, 10> enters the bank. The customer C1 enters
after 30 mins and he needs 10 mins for the transaction. He goes to the teller 1. Then
customer C2 enters the bank and goes to teller 2. When the transaction ends, the
customer leaves the bank. When tellers are not free, customer will wait in the queue.
In the event priority queue, we have different events. The entries in the priority queue
are like arr, 76 (arrival event at 76 min) or q1, 80 (event in q1 at 80 min) etc. Lets
see the statistics when a customer leaves the bank. At exit, you see the customer
leaving the bank as C15<68, 3><77, 3>, it means that the customer C15 enters the
bank at 68 mins and requires 3 mins for his transaction. He goes to the teller 4 but the
teller is not free, so the customer has to wait in the queue. He leaves the bank at 77
mins.
This course is not about the animation or simulation. We will solve the problems,
using different data structures. Although with the help of simulation and animation,
you can have a real sketch of the problem.

Page 108 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 11
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.1.1, 4.2.1

Summary

Implementation of Priority Queue

Tree
Binary Tree
Terminologies of a Binary Tree
Strictly Binary Tree
Level
Complete Binary Tree
Level of a Complete Binary Tree
Tips

In the previous lecture, we demonstrated the technique of data structure priority queue
with the help of the example from the banking system. Data structure priority queue is
a variation of queue data structure to store the events. We have witnessed this
simulation by animation. The simulation and animation help a programmer to
understand the functionality of a program. As these are not the part of the current
course, so we will study simulation and animation in detail in the coming courses.
A queue is a FIFO structure (i.e. first in first out). But there are its variations
including the priority queue, discussed in the previous lecture. In priority queue, we
were adding data to a queue but did not get the data from the queue by First In, First
Out (FIFO) rule. We put events in the queue and got these from the queue with
respect to the time of occurrence of the event. Thus we get the items from a priority
queue in a particular order with respect to a priority given to the items (events). The
priority queue is also an important data structure. Lets see its implementation.

Implementation of Priority Queue

In the priority queue, we put the elements in the queue to get them from the queue
with a priority of the elements. Following is the C++ code of the priority queue.
#include "Event.cpp"
#define PQMAX 30
class PriorityQueue
{

Page 109 of 519

CS301 Data Structures

___________________________________________________________________
public:
PriorityQueue()
{
size = 0; rear = -1;
};
~PriorityQueue() {};
int full(void)
{
return ( size == PQMAX ) ? 1 : 0;
};
Event* remove()
{
if( size > 0 )
{
Event* e = nodes[0];
for(int j=0; j < size-2; j++ )
nodes[j] = nodes[j+1];
size = size-1; rear=rear-1;
if( size == 0 ) rear = -1;
return e;
}
return (Event*)NULL;
cout << "remove - queue is empty." << endl;
};
int insert(Event* e)
{
if( !full() )
{
rear = rear+1;
nodes[rear] = e;
size = size + 1;
sortElements(); // in ascending order
return 1;
}
cout << "insert queue is full." << endl;
return 0;
};
int length() { return size; };
};
In this code, the file Events.cpp has been included. Here we use events to store in the
queue. To cater to the need of storing other data types too, we can write the
PriorityQueue class as a template class.
In the above code, we declare the class PriorityQueue. Then there is the public part of
the class. In the public part, at first a programmer encounters the constructor of the
class. In the constructor, we assign the value 0 to size and 1 to rear variables. A

Page 110 of 519

CS301 Data Structures

___________________________________________________________________
destructor, whose body is empty, follows this. Later, we employ the method full()
which checks the size equal to the PQMAX to see whether the queue is full. If the size
is equal to PQMAX, the function returns 1 i.e. TRUE. Otherwise, it returns 0 i.e.
FALSE. We are going to implement the priority queue with an array. We can also use
linked list to implement the priority queue. However, in the example of simulation
studied in the previous lecture, we implemented the queue by using an array. We have
seen in the simulation example that there may be a maximum of five events. These
events include one arrival event and four departure events. That means four queues
from where the customers go to the teller and one to go out of the bank after the
completion of the transaction. As we know that there is a need of only five queues, so
it was decided to use the array instead of dynamic memory allocation and
manipulating the pointers.
In the remove() method, there are some things which are the property of the priority
queue. We dont have these in the queue. In this method, first of all we check the size
of the priority queue to see whether there is something in the queue or not. If size is
greater than 0 i.e. there are items in the queue then we get the event pointer (pointer to
the object Event) e from the first position (i.e. at index 0) of the array, which we are
using internally to implement the queue. At the end of the method, we return this
event object e. This means that we are removing the first object from the internal
array. We already know that the removal of an item from the start of an array is very
time consuming as we have to shift all the remaining items one position to the left.
Thus the remove() method of the queue will execute slowly. We solved this problem
by removing the item from the position where the front pointer is pointing. As the
front and rear went ahead and we got empty spaces in the beginning, the circular
array was used. Here, the remove() method is not removing the element from the
front. Rather, it is removing element from the first position (i.e. index 0). Then we
execute a for loop. This for loop starts from 0 and executes to size-2. We can notice
that in this for loop, we are shifting the elements of the array one position left to fill
the space that has been created by removing the element from the first position. Thus
the element of index 1 becomes at index 0 and element of index 2 becomes at index 1
and so on. Afterwards, we decrease size and rear by 1. By decreasing the size 1 if it
becomes zero, we will set rear to 1. Now by the statement
return e ;
We return the element (object e), got from the array. The outer part of the if block
return (Event*)NULL;
cout << "remove - queue is empty." << endl;
is executed if there is nothing in the queue i.e. size is less than 0. Here we return
NULL pointer and display a message on the screen to show that the queue is empty.
Now lets look at the insert() method. In this method, first of all we check whether the
array (we are using internally) is full or not. In case, it is not full, we increase the
value of rear by 1. Then we insert the object e in the nodes array at the position rear.
Then the size is increased by 1 as we have inserted (added) one element to the queue.
Now we call a method sortElements() that sorts the elements of the array in an order.
We will read different algorithms of sorting later in this course.
We have said that when we remove an element from a priority queue, it is not

Page 111 of 519

CS301 Data Structures

___________________________________________________________________
according to the FIFO rule. We will remove elements by some other rule. In the
simulation, we had decided to remove the element from the priority queue with
respect to the time of occurrence of an event (arrival or departure). We will remove
the element (event) whose time of occurrence is before other events. This can be
understood from the example of the traffic over a bridge or crossing. We will give
higher priority to an ambulance as compared to a bus. The cars will have the lower
priority. Thus when a vehicle has gone across then after it we will see if there is any
ambulance in the queue. If it is there, we will remove it from the queue and let go
across the bridge. Afterwards, we will allow a bus to go and then the cars. In our
simulation example, we put a number i.e. time of occurrence, with the object when we
add it to the queue. Then after each insertion, we sort the queue with these numbers in
ascending order. Thus the objects in the nodes array get into an order with respect to
the time of their occurrence. After sorting, the first element in the array is the event,
going to be occurring earliest in the future. Now after sorting the queue we return 1
that shows the insert operation has been successful. If the queue is full, we display a
message to show that the queue is full and return 0, indicating that the insert operation
had failed.
Then there comes the length() method, having a single statement i.e.
return size ;
This method returns the size of the queue, reflecting the number of elements in the
queue. It is not the size of the array used internally to store the elements of the queue.
We have seen the implementation of the priority queue by using an array. We will use
the priority queue in some other algorithms later. Now, we will see another
implementation of the priority queue using some thing other than array, which is
much better than using an array. This will be more efficient. Its remove and insert
methods will be faster than the ones in array technique. Here in the simulation, we
were making only five queues for the events. Suppose, if these go to hundreds or
thousands, a lot of time will be spent to remove an element from the queue. Similarly,
when an element is added, after adding the element, to sort the whole array will be a
time consuming process. Thus the application, with the use of the priority queue, will
not be more efficient with respect to the time.

Tree
Now lets talk about a data structure called tree. This is an important data structure.
This data structure is used in many algorithms. We will use it in most of our
assignments. The data structures that we have discussed in previous lectures are linear
data structures. The linked list and stack are linear data structures. In these structures,
the elements are in a line. We put and get elements in and from a stack in linear order.
Queue is also a linear data structure as a line is developed in it. There are a number of
applications where linear data structures are not appropriate. In such cases, there is
need of some non-linear data structure. Some examples will show us that why nonlinear data structures are important. Tree is one of the non-linear data structures.
Look at the following figure. This figure (11.1) is showing a genealogy tree of a
family.

Page 112 of 519

CS301 Data Structures

___________________________________________________________________

T
.

Muhammad Aslam Khan

Sohail Aslam

Haaris

Saad

Javed Aslam

Qasim

Asim

Yasmeen Aslam

Fahd

Ahmed

Sara

Omer

In this genealogy tree, the node at the top of the tree is Muhammad Aslam Khan i.e.
the head of the family. There are three nodes under this one. These are Sohail Aslam,
Javed Aslam and Yasmeen Aslam. Then there are nodes under these three nodes i.e.
the sons of these three family members. You may have seen the tree like this of some
other family. You can make the tree of your family too. The thing to be noted in this
genealogical tree is that it is not a linear structure like linked list, in which we have to
tell that who is the father and who is the son. We make it like a tree. We develop the
tree top-down, in which the father of the family is on the top with their children
downward. We can see the similar tree structure of a company. In the tree of a
company, the CEO of the company is on the top, followed downwardly by the
managers and assistant managers. This process goes downward according to the
administrative hierarchy of the company. Our tree structure is different from the
actual tree in the way that the actual tree grows from down to up. It has root downside
and the leaves upside. The data structure tree is opposite to the real tree as it goes
upside down. This top-down structure in computer science makes it easy to
understand the tree data structure.
There may be situations where the data, in our programs or applications, is not in the
linear order. There is a relationship between the data that cannot be captured by a
linked list or other linear data structure. Here we need a data structure like tree.
In some applications, the searching in linear data structures is very tedious. Suppose
we want to search a name in a telephone directory having 100000 entries. If this
directory is in a linked list manner, we will have to traverse the list from the starting
position. We have to traverse on average half of the directory if we find a name. We
may not find the required name in the entire directory despite traversing the whole
list. Thus it would be better to use a data structure so that the search operation does
not take a lot of time. Taking into account such applications, we will now talk about a
special tree data structure, known as binary tree.

Binary Tree
The mathematical definition of a binary tree is
A binary tree is a finite set of elements that is either empty or is partitioned into three
disjoint subsets. The first subset contains a single element called the root of the tree.
The other two subsets are themselves binary trees called the left and right sub-trees.
Each element of a binary tree is called a node of the tree.
Following figure shows a binary tree.

Page 113 of 519

CS301 Data Structures

___________________________________________________________________
A

Fig 11.2: A Binary Tree

There are nine nodes in the above figure of the binary tree. The nine nodes are A, B,
C, D, E, F, G, H, and I. The node A is at the top of the tree. There are two lines from
the node A to left and right sides towards node B and C respectively. Lets have a
look at the left side of the node A. There are also two lines from node B to left and
right, leading to the nodes D and E. Then from node, E there is only one line that is to
the left of the node and leads to node G. Similarly there is a line from node C towards
the node F on right side of the node C. Then there are two lines from node F that leads
to the nodes H and I on left and right side respectively.
Now we analyze this tree according to the mathematical definition of the binary tree.
The node A is the root of the tree. And tree structure (i.e. the nodes B, D, E, G and
their relation) on the left side of the node A is a sub-tree, called the Left subtree.
Similarly the nodes (C, F, H and I) on the right side of the node A comprise the subtree termed as Right subtree. Thus we made three parts of the tree. One part contains
only one node i.e. A while the second part has all the nodes on left side of A, called as
Left subtree. Similarly the third part is the Right subtree, containing the nodes on right
side of A. The following figure depicts this scenario.

Page 114 of 519

CS301 Data Structures

___________________________________________________________________

Root

Left subtree

Right subtree
Fig 11.3: Analysis of a binary tree

The same process of sub classing the tree can be done at the second level. That means
the left and right subtrees can also be divided into three parts similarly. Now consider
the left subtree of node A. This left subtree is also a tree. The node B is the root of
this tree and its left subtree is the node D. There is only one node in this left subtree.
The right subtree of the node B consists of two nodes i.e. E and G. The following
figure shows the analysis of the tree with root B.

Page 115 of 519

CS301 Data Structures

___________________________________________________________________

A
Root
B

Left subtree

Right subtree
Fig 11.4: Analysis of a binary tree
On going one step down and considering right subtree of node B, we see that E is the
root and its left subtree is the single node G. There is no right subtree of node E or we
can say that its right subtree is empty. This is shown in the following figure.
A

Root

Leftsubtree

Fig 11.5: Analysis of a binary tree

Page 116 of 519

CS301 Data Structures

___________________________________________________________________
Now the left sub tree of E is the single node G. This node G can be considered as the
root node with empty left and right sub trees.
The definition of tree is of recursive nature. This is due to the fact that we have seen
that which definition has been applied to the tree having node A as root, is applied to
its subtrees one level downward. Similarly as long as we go down ward in the tree the
same definition is used at each level of the tree. And we see three parts i.e. root, left
subtree and right subtree at each level.
Now if we carry out the same process on the right subtree of root node A, the node C
will become the root of this right subtree of A. The left subtree of this root node C is
empty. The right subtree of C is made up of three nodes F, H and I. This is shown in
the figure below.
A

Root

Right subtree
Fig 11.6: Analysis of a binary tree
Now we apply the same recursive definition on the level below the node C. Now the
right subtree of the node C will be considered as a tree. The root of this tree is the
node F. The left and right subtrees of this root F are the nodes H and I respectively.
The following figure depicts this.

Page 117 of 519

CS301 Data Structures

___________________________________________________________________
A

C
Root

Left subtree

Right subtree

Fig 11.7: Analysis of a binary tree

We make (draw) a tree by joining different nodes with lines. But we cannot join any
nodes whichever we want to each other. Consider the following figure.
A

Fig 11.8: A non-tree structure

Page 118 of 519

CS301 Data Structures

___________________________________________________________________
It is the same tree, made earlier with a little modification. In this figure we join the
node G with D. Now this is not a tree. It is due to the reason that in a tree there is
always one path to go (from root) to a node. But in this figure, there are two paths
(tracks) to go to node G. One path is from node A to B, then B to D and D to G. That
means the path is A-B-D-G. We can also reach to node G by the other path that is the
path A-B-E-G. If we have such a figure, then it is not a tree. Rather, it may be a
graph. We will discuss about graphs at the end of this course.
Similarly if we put an extra link between the nodes A and B, as in the figure below,
then it is also no more a tree. This is a multiple graph as there are multiple (more than
1) links between node A and B.
A

Fig 11.9: A non-tree structure

Similarly if we put other links between different nodes, then the structure thus
developed will not be a tree. The following figure is also an example of a structure
that is not a tree as there are multiple links between the different nodes.

Page 119 of 519

CS301 Data Structures

___________________________________________________________________
A

Fig 11.10: A non-tree structure

Terminologies of a binary tree

Now lets discuss different terminologies of the binary tree. We will use these
terminologies in our different algorithms. Consider the following figure.
parent

A
Left descendant

Right descendant
B

Leaf nodes

Leaf nodes
Fig 11.11: Terminologies used in a binary tree

Page 120 of 519

CS301 Data Structures

___________________________________________________________________
We have discussed that the node on the top is said root of the tree. If we consider the
relationship of these nodes, A is the parent node with B and C as the left and right
descendants respectively. C is the right descendant of A. Afterwards, we look at the
node B where the node D is its left descendant and E is its right descendant. We can
use the words descendant and child interchangeably. Now look at the nodes D, G, H
and I. These nodes are said leaf nodes as there is no descendant of these nodes. We
are just introducing the terminology here. In the algorithms, we can use the words
root or parent, child or descendant. So the names never matter.

Strictly Binary Tree

There is a version of the binary tree, called strictly binary tree. A binary tree is said to
be a strictly binary tree if every non-leaf node in a binary tree has non-empty left and
right subtrees.
Now consider the previous figure 11.11. We know that a leaf node has no left or right
child. Thus nodes D, G, H and I are the leaf nodes. The non-leaf nodes in this tree are
A, B, C, E and F. Now according to the definition of strictly binary tree, these nonleaf nodes (i.e. A, B, C, E and F) must have left and right subtrees (Childs). The node
A has left child B and right child C. The node B also has its left and right children that
are D and E respectively. The non-leaf node C has right child F but not a left one.
Now we add a left child of C that is node J. Here C also has its left and right children.
The node F also has its left and right descendants, H and I respectively. Now the last
non-leaf node E has its left child, G but no right one. We add a node K as the right
child of the node E. Thus the tree becomes as shown below in the figure 11.12.
A

Fig 11.12: A Strictly binary tree

Now all the non-leaf nodes (A, B, C, E and F) have their left and right children so
according to the definition, it is a strictly binary tree.
Page 121 of 519

CS301 Data Structures

___________________________________________________________________

Level
The level of a node in a binary tree is defined as follows:
Root has level 0,
Level of any other node is one more than the level its parent (father).
The depth of a binary tree is the maximum level of any leaf in the tree.
To understand level of a node, consider the following figure 11.13. This figure shows
the same tree of figure 11.2, discussed so far.
A

---------------------------------------- Level 0

D 2

1 ------------------- Level 1

2 ------- Level 2

-Level 3

Fig 11.13: Level of nodes of a tree

In the above figure, we have mentioned the level of each node. The figure also shows
that the root node has level 0. We have talked in the definition of the level that the
level of a node, other than root, is one more than the level of its parent. Now in the
tree, the parent of nodes B and C is node A with level 0. Now by definition, the level
of B and C will be 1 (i.e. level of A + 1). Now if go downward in the tree and see the
nodes D, E and F, the parent node of D and E is B the level of which is 1. So the level
of D and E is 2 (i.e. 1 + 1). Similarly the level of F is 2 as the level of its parent (i..e.
C) is 1. In the same way, we can easily understand that the level of G, H and I is 3 as
the parent nodes of these (E and F) have level 2. Thus we see that tree has multi levels
in the downward direction. The levels of the tree increase, as the tree grows
downward more. The number of nodes also increases.
Now lets talk why we call it a binary tree. We have seen that each node has a
maximum of two subtrees. There might be two, one or no subtree of a node. But it
cannot have more than two. So, we call such a tree a binary one due to the fact that a
node of it can have a maximum of two subtrees. There are trees whose nodes can have
more than two subtrees. These are not binary trees. We will talk about these trees
later.
By seeing the level of a tree, we can tell the depth of the tree. If we put level with
each node of the binary tree, the depth of a binary tree is the maximum level. In the
Page 122 of 519

CS301 Data Structures

___________________________________________________________________
figure 11.13, the deepest node is at level 3 i.e. the maximum level is 3. So the depth
of this binary tree is 3.

Complete Binary Tree

Now, if there are left and right subtrees for each node in the binary tree as shown
below in figure 11.14, it becomes a complete binary tree.

H
3

J
3

K
3

L
3

M
3

N
3

O
3

Fig 11.14: A Complete Binary Tree

The definition of the complete binary tree is
A complete binary tree of depth d is the strictly binary tree all of whose leaves are at
level d.
Now look at the tree, the leaf nodes of the tree are at level 3 and are H, I, J, K, L, M,
N and O. There is no such a leaf node that is at some level other than the depth level d
i.e. 3. All the leaf nodes of this tree are at level 3 (which is the depth of the tree i.e. d).
So this is a complete binary tree. In the figure, all the nodes at level 3 are highlighted.
We know that the property of the binary tree is that its node can have a maximum of
two subtrees, called as left and right subtrees. The nodes increase at each level, as the
tree grows downward. In a complete binary tree, the nodes increase in a particular
order. If we reconsider the previous tree (figure 11.14), we note that the number of
node at level 0 is 1. We can say it 20, as 20 is equal to 1. Down to this, the node is the
level 1 where the number of nodes is 2, which we can say 21, equal to 2. Now at the
next level (i.e. level 2), the number of nodes is 4 that mean 22. Finally the number of
nodes at level 3 is 8, which can be called as 23. This process is shown pictorially in
the following figure 11.15.

Page 123 of 519

CS301 Data Structures

___________________________________________________________________

---------- Level 1: 21 nodes ----------

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

------------- Level 0: 20 nodes

Level 2: 22 nodes

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

O
-

------------------------------------ Level 3: 23 nodes -------------------------------------------Fig 11.15: Number of nodes at each level in a complete binary tree
By observing the number of nodes at a particular level, we come to the conclusion
that the number of nodes at a level is equal to the level number raising to the power of
two. Thus we can say that in a complete binary tree at a particular level k, the number
of nodes is equal to 2k. Note that this formula for the number of nodes is for a
complete binary tree only. It is not necessary that every binary tree fulfill this
criterion. Applying this formula to each level while going to the depth of the tree (i.e.
d), we can calculate the total number of nodes in a complete binary tree of depth d by
adding the number of nodes at each level. This can be written as the following
summation.
20+ 21+ 22 + + 2d = d j=0 2j = 2d+1 1
Thus according to this summation, the total number of nodes in a complete binary tree
of depth d will be 2d+1 1. Thus if there is a complete binary tree of depth 4, the
total number of nodes in it will be calculated by putting the value of d equal to 4. It
will be calculated as under.
24+1 - 1 = 25 1 = 32 1 = 31
Thus the total number of nodes in the complete binary tree of depth 4 is 31.
We know that the total number of nodes (leaf and non-leaf) of a complete binary tree
of depth d is equal to 2d+1 1. In a complete binary tree, all the leaf nodes are at the
Page 124 of 519

CS301 Data Structures

___________________________________________________________________
depth level d. So the number of nodes at level d will be 2d . These are the leaf nodes.
Thus the difference of total number of nodes and number of leaf nodes will give us
the number of non-leaf (inner) nodes. It will be (2d+1 1) 2d i.e. 2d 1. Thus we
conclude that in a complete binary tree, there are 2d leaf nodes and 2d 1 non-leaf
(inner) nodes.

Level of a Complete Binary Tree

We can conclude some more results from the above mathematical expression. We can
find the depth of a complete binary tree if we know the total number of nodes. If we
have a complete binary tree with n total nodes, then by the equation of the total
number of nodes we can write
Total number of nodes = 2d+1 1 = n
To find the value of d, we solve the above equation as under
2d+1 1 = n
2d+1 = n + 1
d + 1 = log2 (n + 1)
d = log2 (n + 1) 1
After having n total nodes, we can find the depth d of the tree by the above equation.
Suppose we have 100,000 nodes. It means that the value of n is 100,000, reflecting a
depth i.e. d of the tree will be log2 (100000 + 1) 1, which evaluates to 20. So the
depth of the tree will be 20. In other words, the tree will be 20 levels deep. The
significance of all this discussion about the properties of complete binary tree will
become evident later.

Operations on Binary Tree

We can define different operations on binary trees.
If p is pointing to a node in an existing tree, then
left(p) returns pointer to the left subtree
right(p) returns pointer to right subtree
parent(p) returns the father of p
brother(p) returns brother of p.
info(p) returns content of the node.
We will discus these methods in detail in next lecture.

Tips

A priority queue is a variation of queue that does not follow FIFO rule.
Tree is a non-linear data structure.
The maximum level of any leaf in a binary tree is called the depth of the tree.
Other than the root node, the level of each node is one more than the level of

Page 125 of 519

CS301 Data Structures

___________________________________________________________________

its parent node.

A complete binary tree is necessarily a strictly binary tree but not vice versa.
At any level k, there are 2k nodes at that level in a complete binary tree.
The total number of nodes in a complete binary tree of depth d is 2d+1 1.
In a complete binary tree there are 2d leaf nodes and 2d 1 non-leaf nodes.

Page 126 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 12
Reading Material
Data Structures And Algorithm analysis in C++
Chapter 4
4.2, 4.3(4.3.2, 4.3.4)

Summary

Operations on Binary Tree

Applications of Binary Tree
Searching for Duplicates
C++ Implementation of Binary Tree
Trace of insert

Operations on Binary Tree

In the last lecture, we talked about the uses of binary tree, which is an abstract data
type. We discussed about the functions to find the information inside a node, the
parent, the siblings(brothers), the left and right children of a node. In this lecture, we
will talk about the algorithms and implementation of those functions.
When we discuss about an abstract data type, firstly we focus what it can do for us
and dont bother about the how part. The how part or implementation is thought out
later.
While implementing these abstract data types, the implementation is hidden in a class
so it is abstract to the user. The user only needs to be aware of the interface. But there
can be situations when the users may like to know about the implementation detail,
for example, when a data type is performing quite slower than promised.
For now, we start our discussion from the methods of tree data type. Consider a tree
has been formed already, following methods will be used to perform different
operations on a node of this tree:
Operation
left(p)
right(p)
parent(p)
brother(p)
info(p)

Description
Returns a pointer to the left sub-tree
Returns a pointer to the right sub-tree
Returns the father node of p
Returns the brother node of p
Returns the contents of node p

These methods have already been discussed at the end of the previous lecture,
however, few more methods are required to construct a binary tree:

Page 127 of 519

CS301 Data Structures

___________________________________________________________________
Operation
setLeft(p, x)
setRight(p, x)

Description
Creates the left child node of p and set
the value x into it.
Creates the right child node of p, the
child node contains the info x.

All these methods are required to build and to retrieve values from a tree.

Applications of Binary Tree

Lets take few examples to understand how the tree data type is used and what are its
benefits. We will also develop some algorithms that may by useful in future while
working with this data type.
Binary tree is useful structure when two-way decisions are made at each point.
Suppose we want to find all duplicates in a list of the following numbers:
14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5
This list may comprise numbers of any nature. For example, roll numbers, telephone
numbers or voters list. In addition to the presence of duplicate number, we may also
require the frequency of numbers in the list. As it is a small list, so only a cursory
view may reveal that there are some duplicate numbers present in this list. Practically,
this list can be of very huge size ranging to thousands or millions.
Searching for Duplicates

One way of finding duplicates is to compare each number with all those that precede
it. Lets see it in detail.
14 15 4, 9, 7, 4,
18 3, 5, 16 4, 20 17 9, 14 5

14 15 4, 9, 7, 4,
18 3, 5, 16 4, 20 17 9, 14 5
Fig 12.1: Search for Duplicates
Suppose, we are looking for duplicates for the number 4, we will start scanning from
the first number 14. While scanning, whenever we find the number 4 inside, we
remember the position and increment its frequency counter by 1. This comparison
will go on till the end of the list to get the duplicates or fequence of the number 4.
You might have understood already that we will have to perform this whole scanning
of list every time for each number to find duplicates. This is a long and time
consuming process.
So this procedure involves a large number of comparisons, if the list of numbers is
large and is growing.
A linked list can handle the growth. But it can be used where a programmer has no
idea about the size of the data before hand. The number of comparisons may still be
large. The comparisons are not reduced after using linked list as it is a linear data

Page 128 of 519

CS301 Data Structures

___________________________________________________________________
structure. To search a number in a linked list, we have to begin from the start of the
list to the end in the linear fashion, traversing each node in it. For optimizing search
operation in a list, there is no real benefit of using linked list. On the contrary, the
search operation becomes slower even while searching in an array because the linked
list is not contiguous like an array and traversing is done in the linked list with the
help of pointers.
So, the solution lies in reducing the number of comparisons. The number of
comparisons can be drastically reduced with the help of a binary tree. The benefits of
linked list are there, also the tree grows dynamically like the linked list.
The binary tree is built in a special way. The first number in the list is placed in a
node, designated as the root of the binary tree. Initially, both left and right sub-trees
of the root are empty. We take the next number and compare it with the number
placed in the root. If it is the same, this means the presence of a duplicate. Otherwise,
we create a new tree node and put the new number in it. The new node is turned into
the left child of the root node if the second number is less than the one in the root.
The new node is turned into the right child if the number is greater than the one in the
root.

14
14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

Fig 12.2: First number in the list became the root

In the above figure, the first number in the list 14 is placed in a node , making it the
root of the binary tree. You can see that it is not pointing to any further node.
Therefore, its left and right pointers are NULL at this point of tree construction time.
Now, lets see, how do we insert the next element into it.
15

14

14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

Fig 12.3: A new node is created to insert it in the binary tree

As a first step to add a new node in the tree, we take the next number 15 in the list and
compare it with 14, the number in the root node. As they are different, so a new node
is created and the number 15 is set into it.
14
15
Page 129 of 519

CS301 Data Structures

___________________________________________________________________

14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

The next step is to add this node in the tree. We compare 15, the number in the new
node with 14, the number in the root node. As number 15 is greater than number 14,
therefore, it is placed as right child of the root node.
The next number in the list i.e. 4, is compared with 14, the number in the root node.
As the number 4 is less than number 14, so we see if there is a left child of the root
node to compare the number 4 with that. At this point of time in the tree, there is no
further left child of the root node. Therefore, a new node is created and the number 4
is put into it.
The below figure shows the newly created node.

14
15

14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

Fig 12.5: A new node is created and number 4 put into it

Next, the newly created node is added as the left child of the root node. It is shown in
the figure below.

Page 130 of 519

CS301 Data Structures

___________________________________________________________________

14
4

15

14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

Fig 12.6: The node is added as the left child of the root node
The next number in the list is 9. To add this number in the tree, we will follow the
already defined and experimented procedure. We compare this number first with the
number in the root node of the tree. This number is found to be smaller than the
number in the root node. Therefore, left sub-tree of the root node is sought. The left
child of the root node is the one with number 4. On comparison, number 9 is found
greater than the number 4. Therefore, we go to the right child of the node with
number 4. At the moment, there is no further node to the right of it, necessitating the
need of creating a new node. The number 9 is put into it and the new node is added as
the right child of the node with number 4. The same is shown in the figure given
below.
14
4

15
9
14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5
Fig 12.7: A new node is added in the tree

We keep on adding new nodes in the tree in line with the above practiced rule and
eventually, we have the tree shown in the below figure.

Page 131 of 519

CS301 Data Structures

___________________________________________________________________

14
4
3

15
9

7
5

18
16

20
17

14, 15, 4, 9, 7, 18, 3, 5, 16, 4, 20, 17, 9, 14, 5

Fig 12.8: Binary tree of whole list of numbers
It is pertinent to note that this is a binary tree with two sub-nodes or children of each
node. We have not seen the advantage of binary tree, the one we were earlier talking
about i.e. it will reduce the number of comparisons. Previously, we found that search
operation becomes troublesome and slower as the size of list grows. We will see the
benefit of using binary tree over linked list later. Firstly, we will see how the tree is
implemented.

C++ Implementation of Binary Tree

See the code below for the file treenode.cpp.
/* This file contains the TreeNode class declaration. TreeNode contains the
functionality for a binary tree node */
1. #include <stdlib.h>
2.
3. template <class Object>
4.
5. class TreeNode
6. {
7.
public:
8. // constructors
9. TreeNode()
10. {
11.
this->object = NULL;
12.
this->left = this->right = NULL;
13. };
14.
15. TreeNode( Object * object )
16. {

Page 132 of 519

CS301 Data Structures

___________________________________________________________________
17.
this->object = object;
18.
this->left = this->right = NULL;
19. };
20.
21. Object * getInfo()
22. {
23.
return this->object;
24. };
25.
26. void setInfo(Object * object)
27. {
28.
this->object = object;
29. };
30.
31. TreeNode * getLeft()
32. {
33.
return left;
34. };
35.
36. void setLeft(TreeNode * left)
37. {
38.
this->left = left;
39. };
40.
41 TreeNode * getRight()
42. {
43.
return right;
44. };
45.
46. void setRight(TreeNode * right)
47. {
48.
this->right = right;
49. };
50.
51. int isLeaf( )
52. {
53.
if( this->left == NULL && this->right == NULL )
54.
return 1;
55.
return 0;
56. };
57.
58. private:
59.
Object * object;
60.
TreeNode * left;
61.
TreeNode * right;
62. }; // end class TreeNode
For implementation, we normally write a class that becomes a factory for objects of

Page 133 of 519

CS301 Data Structures

___________________________________________________________________
that type. In this case too, we have created a class TreeNode to be used to create
nodes of the binary tree. As we want to use this class for different data types,
therefore, we will make it a template class, the line 2 is doing the same thing. Inside
the class body, we start with the private data members given at the bottom of the class
declaration. At line 59, the object is a private data element of Object *, used to store
the tree element (value) inside the node of the tree. left is a private data member of
type TreeNode*, it is used to store a pointer to the left sub-tree. right is a private data
member of type TreeNode*, employed to store a pointer to the right sub-tree.
Now, we go to the top of the class declaration to see the public functions. At line 9, a
public parameter-less constructor is declared. The data members have been initialized
in the constructor. At line 11, the object data member is initialized to NULL. Similarly
left and right data members are initialized to NULL at line 12.
There is another constructor declared at line 15- that takes object value as a parameter
to construct a TreeNode object with that object value. While the pointers for right and
left sub-trees are pointing to NULL.
At line 21, there is method getInfo(), which returns the object i.e. the element of the
TreeNode object.
At line 26, the method setInfo(Object * ) sets the value of the object data member to
the value passed to it as the argument.
The method getLeft() returns the pointer to the left sub-tree. Similarly, the getRight()
returns the right sub-tree. Note that both of these methods return a pointer to the
object of type TreeNode.
The setLeft(TreeNode *) method is used to set the pointer left to left sub-tree.
Similarly, setRight(TreeNode *) is used to set the pointer right to right sub-tree. Both
of these methods accept a pointer of type TreeNode.
The isLeaf() method at line 51, is to see whether the current node is a leaf node or not.
The method returns 1 if it is leaf node. Otherwise, it returns 0.
Using the above TreeNode, nodes for the binary tree can be created and linked up
together to form a binary tree. We can write a separate method or a class to carry out
the node creation and insertion into tree.
Lets use this class by writing couple of functions. Below is the code of main program
file containing the main() and insert() functions.
1. #include <iostream>
2. #include <stdlib.h>
3. #include "TreeNode.cpp"
4.
5. int main(int argc, char * argv[])
6. {
7. int x[] = {14,15,4,9,7,18,3,5,16,4,20,17,9,14,5,-1};
8. TreeNode <int> * root = new TreeNode<int>();
9. root->setInfo( &x[0] );
10. for(int i = 1; x[i] > 0; i++ )
11. {
12.
insert( root, &x[i] );
13. }
14. }
15.

Page 134 of 519

CS301 Data Structures

___________________________________________________________________
16. void insert (TreeNode <int> * root, int * info)
17. {
18. TreeNode <int> * node = new TreeNode <int> (info);
19. TreeNode <int> * p, * q;
20. p = q = root;
21. while( *info != *(p->getInfo()) && q != NULL )
22. {
23.
p = q;
24.
if( *info < *(p->getInfo()) )
25.
q = p->getLeft();
26.
else
27.
q = p->getRight();
28. }
29.
30. if( *info == *( p->getInfo() ) )
31. {
32.
cout << "attempt to insert duplicate: " << *info << endl;
33.
delete node;
34. }
35. else if( *info < *(p->getInfo()) )
36.
p->setLeft( node );
37. else
38. p->setRight( node );
39. } // end of insert
We have used the same list of numbers for discussion in this lecture. It is given at line
7 in the code. It is the same, only the last number is 1. This is used as delimiter or
marker to indicate that the list has finished.
At line 8, we are creating a new TreeNode object i.e. a root node as the name implies.
This node will contain an int type element as evident from the syntax.
At line 9, the first number in the list is set into the root node. At line 10, the for loop
is started, which is inserting all the elements of the list one by one in the tree with the
use of the insert() function. Most of the time, this loop will be reading the input
numbers from the users interactively or from a file. It it is Windows application then a
form can by used to take input from the user. In this implementation, our objective is
to insert numbers in the tree and see the duplicates, so hard coding the numbers within
our program will serve the purpose.
The insert() method starts from line 16. It is accepting two parameters. The first
parameter is pointer to a TreeNode object, containing a value of int type while second
is the info that is an int *.
In the first line of the function, line 18, a new node has been by calling the
parameterized constructor of the TreeNode class.
Then at line 19, two pointer variables p and q are declared.
In line 20, we are initializing variables p and q to the root node.
In line 21, the while loop is being started, inside the while loop we are checking the
equality of the number inside the node (being pointed to by pointer p) with the
number being passed. The control is entered into the loop, if both the numbers are not
equal and q is not pointing to NULL. This while loop will be terminated if the
numbers in the two nodes are equal or end of a sub-tree is reached.

Page 135 of 519

CS301 Data Structures

___________________________________________________________________
At line 23 inside the loop, p is assigned the value of q.
At line 24 inside the loop, if the number inside the node is smaller than the number in
the node pointed to by the pointer p. Then at line 25, we are getting the left sub-tree
address (left) and assigning it to the pointer q.
Otherwise, if this not the case, means the number in the node to be inserted is not
smaller than the number in the node pointed to by the pointer p. Then at line 27, the
right sub-tree address is retrieved and assigned to pointer q.
At line 30, the comparison is made to see if the both the values are equal (number in
the node to be inserted and the number inside the node pointed to by the pointer p). In
case they are equal, it displays a message with the number informing that a duplicate
number has been found. This also means that the upper while loop was terminated
because of equality of the two numbers. In the line 33, the newly created node is
deleted afterwards.
At line 35, we are checking if the number in the newly constructed node is less than
the number in the node at the end of a sub-tree. If this is so then the newly constructed
node is inserted to the left of the tree node. The insertion code is at line 36.
If the number in the newly constructed node is greater than the number in the tree
node, the newly constructed node will be inserted to the right of the tree node as
shown on line 38. To make it clearer, lets see the same thing, the insert() method
pictorially.

Trace of insert
We will take the tree and the figure, we constructed above. At this time, we want to
insert some new numbers in the tree as given in the figure below:
p
14
q
17
4

15
9

3
7
5

18
16

20

17, 9, 14, 5

Fig 12.9: Start of insertion of new node in the tree

Initially the pointers p and q are pointing to the start (root) of the tree. We want to
insert the number 17 in the tree. We compare the number in the root node (14) with
number 17. Because number 17 is greater than the number 14, so as we did in the
while loop in the function above, we will move toward the right sub-tree. In the next
picture below, we can see that the pointer q has moved to the right sub-tree of the
root.

After moving the pointer q forward, we make the pointer p point to the same node.
We do this operation as the first step inside the while loop. It can be seen at line 23
above. So following will be the latest position of pointers in the tree.
14

17
4
3

15

p
q

9
7

18
16

20

5
17, 9, 14, 5

Fig 12.11: Insertion of a new node in progress

Now, the number 15 in the tree node is compared with new number 17. Therefore, the
pointer q is again moved forward to the right child node i.e. the node with number 18.
In the next step, we move forward the p pointer also. The following figure depicts the
current picture of the tree:
14

17
4
3

15
9

18

p
q

16

20

Page 137 of 519

Fig 12.12: Insertion of a new node in progress

CS301 Data Structures

___________________________________________________________________
17, 9, 14, 5

The previous process is repeated again (while loop is the repeating construct) that the
number 17 is compared with the number 18. This time the left child node is traversed
and q pointer starts pointing the node with number 16 inside. In the next step, p is
moved forward to point to the same node as q.
The same comparison process starts again, number 17 is found to be greater than the
number 16, therefore, the right child is seek. But we see that the current tree node
does not have right child node, so it returns NULL. Following figure depicts it.
14

17
4

15

18

9
7

20

16

p
5

q
17, 9, 14, 5
Fig 12.13: Insertion of a new node in progress

Above shown (q is pointing to NULL) is the condition that causes the while loop in
the code above to terminate. Later we insert the new node as the right child of the
current node.
14
4
3

15
18

9
7

20

16

p
5
17, 9, 14, 5

node

17

p->setRight( node ) ;
Fig 12.14: A new node is inserted in the tree
Page 138 of 519

CS301 Data Structures

___________________________________________________________________

It is recommended to execute this algorithm manually to insert the remaining numbers

in the tree. This will make the understanding more clear. This tree algorithm is going
to be used rigorously in the future and complete understanding of it will make the
complex future implementations easier to comprehend. The specific advantages of
using binary tree over linked list will be discussed in more detail in the next lecture.

Page 139 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 13
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.3, 4.6

Summary

Cost of Search
Binary Search Tree (BST)
Traversing a Binary Tree
C++ code
Example
Exercise

In the previous lecture, we had written and demonstrated the use of C++ code of
insert routine for a tree through an example. We also saw how a new node is inserted
into a binary tree. If the to-be-inserted number (node) is already in the tree i.e. it
matches a number already present in the tree, we display a message that the number is
already in the tree. In the last lecture, the advantages of the tree data structure vis-vis linked list data structure were also discussed. In a linked list, a programmer has to
search the whole list to find out a duplicate of a number to be inserted. It is very
tedious job as the number of stored items in a linked list is very large. But in case of
tree data structure, we get a dynamic structure in which any number of items as long
as memory is available, can be stored. By using tree data structure, the search
operation can be carried out very fast. Now we will see how the use of binary tree can
help in searching the duplicate number in a very fast manner.

Cost of Search
Consider the previous example where we inserted the number 17 in the tree. We
executed a while loop in the insert method and carried out a comparison in while loop.
If the comparison is true, it will reflect that in this case, the number in the node where
the pointer p is pointing is not equal to 17 and also q is not NULL. Then we move p
actually q to the left or right side. This means that if the condition of the while loop is
true then we go one level down in the tree. Thus we can understand it easily that if
there is a tree of 6 levels, the while loop will execute maximum 6 times. We conclude
from it that in a given binary tree of depth d, the maximum number of executions of
the while loop will be equal to d. The code after the while loop will do the process
depending upon the result of the while loop. It will insert the new number or display a
message if the number was already there in the tree.

Page 140 of 519

CS301 Data Structures

___________________________________________________________________
Now suppose we have another method find. This method does not insert a new
number in the tree. Rather, it traverses a tree to find if a given number is already
present in the tree or not. The tree which the find method traverses was made in such
an order that all the numbers less than the number at the root are in the left sub-tree of
the root while the right sub-tree of the root contains the numbers greater than the
number at the root. Now the find method takes a number x and searches out its
duplicate in the given tree. The find method will return true if x is present in the tree.
Otherwise, it will return false. This find method does the same process that a part of
the insert method performs. The difference is that the insert method checks for a
duplicate number before putting a number in the tree whereas the find method only
finds a number in the tree. Here in the find method, the while loop is also executed at
maximum equal to the number of levels of the tree. We do a comparison at each level
of the tree until either x is found or q becomes NULL. The loop terminates in case, the
number is found or it executes to its maximum number, i.e. equal to the number of
levels of the tree.
In the discussion on binary tree, we talked about the level and number of nodes of a
binary tree. It was witnessed that if we have a complete binary tree with n numbers of
nodes, the depth d of the tree can be found by the following equation:
d = log2 (n + 1) 1
Suppose we have a complete binary tree in which there are 100,000 nodes, then its
depth d will be calculated in the following fashion.
d = log2 (100000 + 1) 1 = log2 (100001) 1= 20
The statement shows that if there are 100,000 unique numbers (nodes) stored in a
complete binary tree, the tree will have 20 levels. Now if we want to find a number x
in this tree (in other words, the number is not in the tree), we have to make maximum
20 comparisons i.e. one comparison at each level. Now we can understand the benefit
of tree as compared to the linked list. If we have a linked list of 100,000 numbers,
then there may be 100,000 comparisons (supposing the number is not there) to find a
number x in the list.
Thus in a tree, the search is very fast as compared to the linked list. If the tree is
complete binary or near-to-complete, searching through 100,000 numbers will require
a maximum of 20 comparisons or in general, approximately log2(n). Whereas in a
linked list, the comparisons required could be a maximum of n.
Tree with the linked structure, is not a difficult data structure. We have used it to
allocate memory, link and set pointers to it. It is not much difficult process. In a tree,
we link the nodes in such a way that it does not remain a linear structure. If instead of
100,000, we have 1 million or 10 million or say, 1 billion numbers and want to build a
complete binary tree of these numbers, the depth (number of levels) of the tree will be
log2 of these numbers. The log2 of these numbers will be a small number, suppose 25,
30 or 40. Thus we see that the number of level does not increase in such a ratio as the
number of nodes increase. So we can search a number x in a complete binary tree of 1

Page 141 of 519

CS301 Data Structures

___________________________________________________________________
billion nodes only in 30-40 comparisons. As the linked list of such a large number
grows large, the search of a number in such a case will also get time consuming
process. The usage of memory space does not cause any effect in the linked list and
tree data structures. We use the memory dynamically in both structures. However,
time is a major factor. Suppose one comparison takes one micro second, then one
billion seconds are required to find a number from a linked list (we are supposing the
worst case of search where traversing of the whole list may be needed). This time will
be in hours. On the other hand, in case of building a complete binary tree of these one
billion numbers, we have to make 30-40 comparisons (as the levels of the tree will be
30-40), taking only 30-40 microseconds. We can clearly see the difference between
hours and microseconds. Thus it is better to prefer the process of building a tree of the
data to storing it in a linked list to make the search process faster.

Binary Search Tree

While discussing the search procedure, the tree for search was built in a specific
order. The order was such that on the addition of a number in the tree, we compare it
with a node. If it is less than this, it can be added to the left sub-tree of the node.
Otherwise, it will be added on the right sub-tree. This way, the tree built by us has
numbers less than the root in the left sub-tree and the numbers greater than the root in
the right sub-tree. A binary tree with such a property that items in the left sub-tree are
smaller than the root and items in the right sub-tree are larger than the root is called a
binary search tree (BST). The searching and sorting operations are very common in
computer science. We will be discussing them many times during this course. In most
of the cases, we sort the data before a search operation. The building process of a
binary search tree is actually a process of storing the data in a sorted form. The BST
has many variations, which will be discussed later. The BST and its variations play an
important role in searching algorithms. As data in a BST is in an order, it may also be
termed as ordered tree.

Traversing a Binary Tree

Now lets discuss the ways to print the numbers present in a BST. In a linked list, the
printing of stored values is easy. It is due to the fact that we know wherefrom, a
programmer needs to start and where the next element is. Equally is true about
printing of the elements in an array. We execute a for loop starting from the first
element (i.e. index 0) to the last element of the array. Now lets see how we can
traverse a tree to print (display) the numbers (or any data items) of the tree.
We can explain this process with the help of the following example in which we
traverse a binary search tree. Suppose there are three nodes tree with three numbers
stored in it as shown below.

Page 142 of 519

CS301 Data Structures

___________________________________________________________________
14

15
Fig 13.1: A three node binary tree

Here we see that the root node has the number 14. The left sub-tree has only one node
i.e. number 4. Similarly the right sub-tree consists of a single node with the number
15. If we apply the permutations combinations on these three nodes to print them,
there may be the following six possibilities.
1: (4, 14, 15)
2: (14, 4, 15)
3: (15, 4, 14)
4: (4, 15, 14)
5: (14, 15, 4)
6: (15, 14, 4)
Look at these six combinations of printing the nodes of the tree. In the first
combination, the order of printing the nodes is 4-14-15. It means that left subtreeroot-right subtree. In the second combination the order is root-left subtree-right
subtree. In the third combination, the order of printing the nodes is right subtree-rootleft subtree. The fourth combination has left subtree-right subtree-root. The fifth
combination has the order root-rigth subtree- left subtree. Finally the sixth
combination has the order of printing the nodes right subtree-root-left subtree. These
six possibilities are for the three nodes only. If we have a tree having a large number
of nodes, then there may increase number of permutations for printing the nodes.
Lets see the general procedure of traversing a binary tree. We know by definition that
a binary tree consists of three sets i.e. root, left subtree and right subtree. The
following figure depicts a general binary tree.
N

node

R
left
subtre

right
subtree

Fig 13.2: A generic binary tree

Page 143 of 519

CS301 Data Structures

___________________________________________________________________
In this figure, we label the root node with N. The left subtree in the figure is in a
triangle labeled as L. This left subtree may consist of any number of nodes. Similarly
the right subtree is enclosed in a triangle having the label R. This triangle of right
subtree may contain any number of nodes. Following are the six permutations, which
we have made for the three nodes previously. To generalize these permutations, we
use N, L and R as abbreviations for root, left subtree and right subtree respectively.
1: (L, N, R)
2: (N, L, R)
3: (R, L, N)
4: (L, R, N)
5: (N, R, L)
6: (R, N, L)
In these permutations, the left and right subtree are not single nodes. These may
consist of several nodes. Thus where we see L in the permutations, it means traversing
the left subtree. Similarly R means traversing the right subtree. In the previous tree of
three nodes, these left and right subtrees are of single nodes. However, they can
consist of any number of nodes. We select the following three permutations from the
above six. The first of these three is (N, L, R), also called as preorder traversal. The
second permutation is (L, N, R) which we called inorder traversal. Finally the third
permutation, also termed as postorder traversal is (L, R, N). Now we will discuss
these preorder, inorder and postorder traversal in detail besides having a look on their
working. We will also see the order in which the numbers in the tree are displayed by
these traversing methods.

C++ code
Lets write the C++ code for it. Following is the code of the preorder method.
void preorder(TreeNode<int>* treeNode)
{
if( treeNode != NULL )
{
cout << *(treeNode->getInfo())<<" ";
preorder(treeNode->getLeft());
preorder(treeNode->getRight());
}
}
In the arguments, there is a pointer to a TreeNode. We may start from any node and
the pointer of the node will be provided as argument to the preorder method. In this
method, first of all we check whether the pointer provided is NULL or not. If it is not
NULL, we print the information stored in that node with the help of the getInfo()
method. Then we call the getLeft() method that returns a pointer of left node, which
may be a complete subtree. With the help of this method, we get the root of that
subtree. We call the preorder method again passing that pointer. When we return

Page 144 of 519

CS301 Data Structures

___________________________________________________________________
from that, the preorder method is called for the right node. Lets see what is
happening in this method. We are calling the preorder method within the preorder
method. This is actually a recursive call. Recursion is supported in C++ and other
languages. Recursion means that a function can call itself. We may want to know why
we are doing this recursive call. We will see some more examples in this regard and
understand the benefits of recursive calls. For the time being, just think that we are
provided with a tree with a root pointer. In the preorder, we have to print the value of
the root node. Dont think that you are in the preorder method. Rather keep in mind
that you have a preorder function. Suppose you want to print out the left subtree in the
preorder way. For this purpose, we will call the preorder function. When we come
back, the right subtree will be printed. In the right subtree, we will again call the
preorder function that will print the information. Then call the preorder function for
the left subtree and after that its right subtree. It will be difficult if you try to do this
incursion in the mind. Write the code and execute it. You must be knowing that the
definition of the binary tree is recursive. We have a node and left and right subtrees.
What is left subtree? It is also a node with a left subtree and right subtree. We have
shown you how the left subtree and right subtree are combined to become a tree. The
definition of tree is itself recursive. You have already seen the recursive functions.
You have seen the factorial example. What is factorial of N? It is N multiplied by N-1
factorial. What is N-1 factorial? N-1 factorial is N-1 multiplied by N-2 factorial. This
is the recursive definition. For having an answer, it is good to calculate the factorial of
one less number till the time you reach at the number 2. You will see these recursions
or recursive relations here and also in mathematic courses. In the course of discrete
mathematics, recursion method is used. Here we are talking about the recursive calls.
We will now see an example to understand how this recursive call works and how can
we traverse the tree using the recursion. One of the benefits of recursion that it prints
out the information of all the nodes without caring for the size of the tree. If the tree
has one lakh nodes, this simple four lines routine will print all the nodes. When
compared with the array printing, we have a simple loop there. In the link list also, we
have a small loop that executes till the time we have a next pointer as NULL. For tree,
we use recursive calls to print it.
Here is the code of the inorder function.
void inorder(TreeNode<int>* treeNode)
{
if( treeNode != NULL )
{
inorder(treeNode->getLeft());
cout << *(treeNode->getInfo())<<" ";
inorder(treeNode->getRight());
}
}
The argument is the same as in the preorder i.e. a pointer to the TreeNode. If this node
is not NULL, we call getLeft() to get the left node and call the inorder function. We
did not print the node first here. In the inorder method, there is no need to print the

Page 145 of 519

CS301 Data Structures

___________________________________________________________________
root tree first of all. We start with the left tree. After completely traversing the
complete left tree, we print the value of the node. In the end, we traverse the right
subtree in recursion.
Hopefully, you have now a fair knowledge about the postorder mechanism too. Here
is the code of the postorder method.
void postorder(TreeNode<int>* treeNode)
{
if( treeNode != NULL )
{
postorder(treeNode->getLeft());
postorder(treeNode->getRight());
cout << *(treeNode->getInfo())<<" ";
}
}
In the postorder, the input argument is a pointer to the TreeNode. If the node is not
NULL, we traverse the left tree first. After that we will traverse the right tree and
print the node value from where we started.
As all of these above routines are function so we will call them as:
cout << "inorder: ";
preorder( root);
cout << "inorder: ";
inorder( root );
cout << "postorder: ";
postorder( root );
Here the root represents the root node of the tree. The size of the tree does not matter
as the complete tree will be printed in preorder, inorder and postorder. Lets discuss
an example to see the working of these routines.

Example
Lets have a look on the following tree.

Page 146 of 519

CS301 Data Structures

___________________________________________________________________
14

15

9
7

18
16

20
17

Fig 13.3:
20

Preorder: 14 4 3 9 7 5 15 18 16 17

This is the same tree we have been using previously. Here we want to traverse the
tree. In the bottom of the figure, the numbers are printed with the help of preorder
method. These numbers are as 14 4 3 9 7 5 15 18 16 17 20. Now take these
numbers and traverse the tree. In the preorder method, we print the root, followed by
traversing of the left subtree and the right subtree respectively. As the value of the
root node is 14, so it will be printed first of all. After printing the value of the root
node, we call the preorder for the left node which is 4. Forget the node 14 as the root
is 4 now. So the value 4 is printed and we call the preorder again with the left sub tree
i.e. the node with value 3. Now the root is the node with value 3. We will print its
value before going to its left. The left side of node with value 3 is NULL. Preorder
will be called if condition is false. In this case, no action will be taken. Now the
preorder of the left subtree is finished. As the right subtree of this node is also NULL,
so there is no need of any action. Now the left subtree of the node with value 4 is
complete. The method preorder will be called for the right subtree of the node with
value 4. So we call the preorder with the right node of the node with value 4. Here,
the root is the node with value 9 that is printed. We will call its left subtree where the
node value is 7. It will be followed by its left subtree i.e. node 5 which will be printed.
In the preorder method, we take the root i.e. 14 in this case. Its value is printed,
followed by its left subtree and so on. This activity takes us to the extreme left node.
Then we back track and print the right subtrees.
Lets try to understand the inorder method from the following statement.

Page 147 of 519

CS301 Data Structures

___________________________________________________________________
14

15

9
7

18
16

20
17

Fig 13.4:

Inorder: 3 4 5 7 9 14 15 16 17 18 20

When we call the inorder function, the numbers will be printed as 3 4 5 7 9 14 15 16

17 18 20. You might have noted that these numbers are in sorted order. When we
build the tree, there was no need of sorting the numbers. But here we have the sorted
numbers in the inorder method. While inserting a new node, we compare it with the
existing ones and place the new node on the left or right side. So during the process of
running the inorder on the binary tree, a programmer gets the numbers sorted.
Therefore we have a sorting mechanism. If we are given some number, it will not be
difficult to get them sorted. This is a new sorting algorithm for you. It is very simple.
Just make BST with these numbers and run the inorder traversal. The numbers
obtained from the inorder traversal are sorted.
In the inorder method, we do not print the value of node first. We go to traverse its
left subtree. When we come back from the left subtree and print the value of this
node. Afterwards, we go to the right subtree. Now consider the above tree of figure
13.4. If we start from number 14, it will not be printed. Then we will go to the left
subtree. The left subtree is itself a tree. The number 4 is its root. Now being at the
node 4, we again look if there is any left subtree of this node. If it has a left subtree,
we will not print 4 and call the inorder method on its left subtree. As there is a left
subtree of 4 that consists a single node i.e. 3, we go to that node. Now we call the
inorder of node 3 to see if there is a subtree of it. As there is no left subtree of 3, the if
statement that checks if the node is not NULL will become false. Here the recursive
calls will not be executed. We will come back from the call and print the number 3.
Thus the first number that is printed in the inorder traversal is 3. After printing 3, we
go to its right subtree that is also NULL. So we come back to node 3. Now as we have
traversed the left and right subtrees of 3 which itself is a left subtree of 4, thus we
have traversed the left subtree of 4. Now we print the value 4 and go to the right
subtree of 4. We come at node 9. Before printing 9 we go to its left subtree that leads
us to node 7. In this subtree with root 7, we will not print 7. Now we will go to its left
subtree which is node 5. We look for the left subtree of 5 which is NULL. Now we
print the value 5. Thus, we have so far printed the numbers 3, 4 and 5. Now we come

Page 148 of 519

CS301 Data Structures

___________________________________________________________________
back to 7. After traversing its left subtree, we print the number 7. The right subtree of
7 is NULL. Thus finally, we have traversed the whole tree whose root is 7. It is a left
subtree of 9. Now, we will come back to 9 and print it (as its left subtree has been
traversed). The right subtree of 9 is NULL. So there is no need to traverse it. Thus the
whole tree having 9 as its root has been traversed. So we come back to the root 4, the
right subtree of which is the tree with root 9. Resultantly, the left and right subtree of
4 have been traversed now. Thus we have printed the numbers 3, 4, 5, 7 and 9. Now
from here (i.e. node 4) we go to the node 14 whose left subtree has been traversed.
Now we print 14 and go to its right subtree. The right subtree of 14 has 15 as the root.
From here, we repeat the same process earlier carried out in case of the left subtree of
14. As a result of this process, the numbers 15, 16, 17, 18 and 20 are printed. Thus we
get the numbers stored in the tree in a sorted order. And we get the numbers printed as
3, 4, 5, 7, 9, 14, 15, 16, 17, 18, 20. This sorted order is only in the inorder traversal.
When we apply the post order traversal on the same tree, the numbers we get are not
in sorted order. The numbers got through postorder traversal are in the following
order:
3 5 7 9 4 17 16 20 18 15 14
We see that these are not in a sorted order.
We know that every function has a signature. There is return type, function name and
its parameter list in the signature of a function. Then there comes the body of the
function. In the body of a function, there may be local variables and calls to some
other functions. Now if a statement in the body of a function is calling the same
function in which it exists, it will be termed as a recursive call. For example if a
statement in function A is calling the function A, it is a recursive call and the process
is called recursion.
Now lets see how recursion is carried out. We know that the function calls are
implemented through stacks at run time. Lets see what happens when there is a
recursive call. In a recursive call that means calling a function itself makes no
difference as far as the call stack is concerned. It is the same as a function F calls a
function G, only with the difference that now function F is calling to function F
instead of G. The following figure shows the stack layout when function F calls
function F recursively.

Page 149 of 519

CS301 Data Structures

___________________________________________________________________
Parameters(F)

Parameters(F)

Local variables(F)

Local variables(F)

Local variables(F)
Return address(F)

Return address(F)
sp

Parameters(F)

Parameters(F)

Parameters(F)

Return address(F)
sp

Local variables(F)
sp
At point of

Return address(F)
During execution of

After

Fig 13.5
When F is called first time, the parameters, local variables and its return address are
put in a stack, as some function has called it. Now when F calls F, the stack will
increase the parameters, local variables and return address of F comes in the stack
again. It is the second call to F. After coming back from this second call, we go to the
state of first call of F. The stack becomes as it was at first call. In the next lecture, we
will see the functionality of recursive calls by an example.

Exercise
Please find out the preorder, inorder and postorder traversal of the tree given below:

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 14
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.3, 4.6

Summary

Recursive Calls
Preorder Recursion
Inorder Recursion
Non-Recursive Traversal
Traversal Trace

We discussed the methods of traversal of the binary tree in the previous lecture. These
methods are- preorder, inorder and postorder. It was witnessed that in the C++ code
that the coding of these methods becomes very short with the use of recursive calls.
The whole tree can be traversed with a few lines of code. We also demonstrated the
benefits of the methods with the help of an example in which a tree was traversed by
preorder, inorder and postorder methods. Implementation of the recursion also came
under discussion in the previous lecture.

Recursive Calls
We know that function calls are made with the help of stacks. When a function calls
some other function, the parameters and return address of the function is put in a
stack. The local variables of the function are also located at the stack and the control
is passed to the called function. When the called function starts execution, it performs
its job by using these parameters and local variables. When there in the function there
comes a return statement or when the function ends then the return address, already
stored in the stack, is used and control goes back to the next statement after the
calling function. Similarly when a function is calling to itself, there is no problem
regarding the stack. We know, in recursion a function calls to itself instead of calling
some other function. Thus the recursion is implemented in the way as other function
calls are implemented.

Preorder Recursion
Now lets see the preorder traversal with respect to the recursion. We will see what
changes happen when preorder function calls itself recursively. Consider the
following binary search tree used in our previous example.

Page 151 of 519

CS301 Data Structures

___________________________________________________________________

14
15

9
7
5

18
20

16
17

Fig 14.1 (a) : preorder recursion

preorder(14)
14
..preorder(4)
4
....preorder(3)
3
......preorder(NULL)
......preorder(NULL)
....preorder(9)
9
......preorder(7)
7
........preorder(5)
5
..........preorder(NULL)
..........preorder(NULL)
........preorder(NULL)
......preorder(NULL)

For the preorder traversal, we call the preorder function and pass it the root node of
the tree (i.e. 14) as a parameter. We have discussed in the preorder traversing that
preorder method first prints the value of the node that is passed to it. So number 14 is
printed. After it, the preorder method calls itself recursively. Thus we will call the
preorder and give it the left subtree of 14 (actually the root node of left subtree) as an
argument. So this call will be preorder (4). As it is preorder traversing, the value of
the root node i.e. 4 will be printed. Now the preorder method looks if there is a left
subtree of the node (i.e.4). If it has a left subtree, the preorder function will call itself
again. The root node of this left subtree will be passed as an argument to the new
function call. Here it is 3, so the number 3 is printed. Thus by the previous three calls
(i.e. preorder(14), preorder(4) and preorder(3) ) we have printed the values 14, 4 and
3. Now the left subtree of 3 is NULL. We know that in preorder method, first of all
we check whether the node is NULL or not. If it is not NULL, the recursion process
continues. It means that we print the value of the node and call the preorder function
again. If the node is NULL, the process ends.
There is always a terminating condition in recursive call or recursive procedure. In the
previous lecture, while trying to end the factorial recursion, we defined the condition
that the factorial of zero is 1. When the factorial of n starts, the factorial is called
again and again by decreasing the value of n by 1. So when we reach at 0, there is no
further recursive call for the factorial as we have set the condition for it i.e. the
factorial of 0 is equal to 1. Thus the recursion ends. Equally is true about the preorder
process. When a node is NULL, then if statement becomes false and we exit the
function (see code of preorder function in previous lecture).
In the above figure, the preorder process that started from node 14, came to an end at
node 3 as the left subtree of 3 is NULL. Now this preorder process will make no
further recursive call. We will come back from here by using the return address of the
call preorder (NULL) from the stack. If we see the stack, this fourth call will come
back to the third call i.e. preorder (3). The left subtree of 3 has ended. Now we have
Page 152 of 519

CS301 Data Structures

___________________________________________________________________
to traverse the right subtree of 3. Here the right subtree of 3 is also NULL. So the
traversing of tree with node 3 as root has been completed. Now we return from the
call preorder (3) and reach one level backward at the node 4. The left subtree of node
4 has been traversed. Now we go to traverse the right subtree of 4. The preorder call
for the right subtree has started. This started from the call preorder (9).Note in the
figure that this call has made at the same level as was for node 3. It is because we are
traversing the right subtree of the same tree that has node 4 as its root and 3 was its
left subtree. After printing the number 9 we go to call the preorder process for the left
subtree of 9. That is preorder (7). When we go to 7, we print it and call the preorder
process for its left subtree i.e. preorder (5). Now after printing the number 5, the
preorder is applied to its left subtree i.e. NULL. So the preorder traversal of left
subtree of 5 ends this way. Now the preorder is applied to the right subtree of 5 which
is also NULL. Thus the next two preorder calls (one for left subtree of 5 and one for
right subtree of 5) after the call preorder (5) are preorder (NULL). Here the returning
address takes the control back to the call preorder (7) through the use of stack. The
left subtree of 7 has been traversed and now we go to the right subtree of it i.e. NULL.
Thus there is a call preorder (NULL). So we come back to node 9 completing the
traversal of its left subtree. Now we traverse the right subtree of 9 which again
happens to be NULL. So there is one more call of preorder with NULL. The number
of dots with the preorder calls in the figure describes the state of the stack. When we
go ahead to traverse a node the dots increase in figure as the call made is put on the
stack. When we come back to a node, the returning address is popped from the stack
and we mention it by decreasing a dot in the figure.
After these four calls of preorder with NULL as argument, we return to the node 4.
We have completed the tree (left and right subtree ) with node 4 as root. As seen in
the figure, this tree (having node 4 as root) is a left subtree of node 14. The numbers
that we got printed upto now are 14, 4, 3, 9, 7 and 5. So after traversing the left
subtree of 14, we come back to node 14 and start to traverse its right subtree. We
traverse the right subtree by applying the recursive calls of preorder in the way as
done while traversing the left subtree of 14. The recursive calls involved in it are
shown in the following figure.

Page 153 of 519

CS301 Data Structures

___________________________________________________________________
14
15

9
7
5

18
20

16
17

..preorder(15)
15
....preorder(NULL)
....preorder(18)
18
......preorder(16)
16
........preorder(NULL)
........preorder(17)
17
..........preorder(NULL)
..........pre3
der(NULL)
......preord er(20)
20
........preorder(NULL)
preorder(NULL)

Fig 14.1 (b): preorder recursion

Now we are at node 14 and know that in preorder, the value of the node is printed first
before going to its left subtree. And after traversing the left subtree, a programmer
comes back and goes to traverse the right subtree. As we have traversed the left
subtree of 14 and have come back to 14, now we will traverse the right subtree of 14.
In the right subtree, we reach at node 15. We print the value 15 and look at its left
subtree. The left subtree is NULL so the call to the preorder method is preorder
(NULL). This call makes the condition of if statement false and the recursive calls
ends on this side. Afterwards, we go to the right subtree of 15. The call preorder (18)
is made, followed by printing of the value 18. Then we go to the left subtree of it that
is node 16. After calling preorder(16) and printing it, we go to the left subtree of 16
which is NULL. Now we will move to right subtree of 16 and the call preorder (17)
prints the number 17. After it, the two next calls are preorder (NULL) as the left and
right subtree of 17 are NULL. Thus after traversing the left and right subtree of 16
(which itself is a left subtree of 18), we return to node 18. Then we go to the right
subtree of 18 and reach at node 20. So preorder (20) prints the value 20. There are
again two preorder (NULL) calls as the left and right subtree of 20 are NULL. The
preorder recursive call ends here. Now we go back and reach at node 14 whose left
and right subtrees have been traversed. That means the whole tree having 14 as the
root has traversed.

Inorder Recursion
Now lets see the inorder recursive calls for traversing a tree. It is not different from
the preorder. The pattern of recursive calls will be changed as in the inorder we
traverse the left subtree first before printing the root. Afterwards, the right subtree is
traversed. The following figures (Fig 14.2(a) and 14.2(b)) explains this phenomenon
of inorder recursion by showing the recursive calls.

Page 154 of 519

CS301 Data Structures

___________________________________________________________________
14
15

9
7
5

18
20

16
17

inorder(14)
..inorder(4)
....inorder(3)
......inorder(null)
3
......inorder(null)
4
....inorder(9)
......inorder(7)
........inorder(5)
..........inorder(null)
5
..........inorder(null)
7
........inorder(null)
9
......inorder(null)
14

Fig 14.2 (a): inorder recursion

We start the inorder with the call inorder (14) as the root of the tree is 14. Now in the
inorder traversing, a programmer has to traverse the left subtree before printing the
value of the root node and then going to the right subtree. So the call to the left
subtree of 14 i.e. inorder (4) will lead to the node 4. At the node 4, the same process
of calling its left subtree, before printing its value, will be applied and we reach at
node 3 that is the left subtree of 4. From the node 3, we go to its left subtree. This left
subtree is NULL. Here in the inorder method, we have the same terminating condition
as that seen in the preorder i.e. we will not make further recursive calls if there
becomes a NULL node in the method call. We end the recursion at that node and
come back. Now when we come back from the NULL left subtree of 3 this shows that
we have visited the left subtree of 3. So the value of the node i.e. 3 is printed. Now we
go to the right subtree of 3. Here the inorder call is inorder (NULL). This call will
make the if statement, in the inorder method, false, leading to the end of the recursion.
Thus we have visited the whole tree whose root is node 3. So we come back to node
4. As the left subtree of node 4 has been visited, we print the value of node i.e. 4.
Thus we have printed the numbers 3 and 4. Now we go to the right subtree of 4. The
right subtree of 4 is the node 9. Now we send 9 as an argument to inorder method. So
there is a call inorder (9). As it is inorder traversing, we go to traverse its left subtree
before printing the number 9. We reach the node 7 and make a call inorder (7). Later,
we go to the left subtree of 7 i.e. the node 5. Now we visit the left subtree of 5 which
is NULL. So here the inorder recursion ends and we come back to node 5, print it and
go to the right subtree of 5. The right subtree of 5 is also NULL. So ending recursion
here, we have finally traversed the tree with 5 as its root. After traversing it, we come
back to the node 7. After visiting the left subtree of 7, we print 7 and go to the right
subtree of 7 which is NULL.
Page 155 of 519

CS301 Data Structures

___________________________________________________________________
After the call inorder (NULL) we have traversed the tree having 7 as the root and it is
a left subtree of 9. Thus we have visited the left subtree of 9. We print the number 9
before going to traverse its right subtree. The right subtree of 9 is NULL so the
recursion ends here and we come back to the node 4. The tree having node 4 as root
has been traversed now. This tree is a left subtree of node 14. Thus after traversing it
we come back to node 14 and print it (as we have traversed its left subtree i.e. tree
with 4 as root). It is evident here that before going to traverse the right subtree of 14,
we have printed the numbers 3, 4, 5, 7, 9 and 14.
Now after printing the number 14, we go to traverse the right subtree of 14. The
traversal of this right subtree with 15 as a root, is being carried out in the way, we
traversed the left subtree. In the right subtree, we reach at node 15 and look at its left
subtree before printing it. This left subtree is NULL. The call inorder (NULL) makes
the condition of if statement false and we come back to the root node i.e. 15. We print
this number and go to the right subtree of 15. The right subtree of 15 is also a tree the
root node of which is 18. From this root node i.e. 18, we go to its left subtree and
reach at node 16. NULL being the left subtree of 16, makes us to return to 16 and
print it (as its left subtree has been traversed, which is NULL). After printing 16, we
go to the right subtree of 16. This leads us to node 17. As we are traversing the tree by
inorder process, so before printing the value 17, it will be necessary to go to traverse
its left subtree. Its left subtree is NULL. So after the call inorder (NULL) we return to
17 and print it. Now we look at the right subtree of 17 that is also NULL. So again
there is a call inorder (NULL) which ends the recursive calls for this tree. Thus we
have traversed the whole tree having node 17 as root and it is a right subtree of 16.
We have already traversed the left subtree of 16.

14
15

9
7
5

18
20

16
17

..inorder(15)
....inorder(null)
15
....inorder(18)
......inorder(16)
........inorder(null)
16
........inorder(17)
..........inorder(null)
17
..........inorder(null)
18
......inorder(20)
........inorder(null)
20
........inorder(null)

Fig 14.2 (b): inorder recursion

Page 156 of 519

CS301 Data Structures

___________________________________________________________________
So we have traversed the tree with node 16 as its root. This tree is a left subtree of
node 18. After traversing the left subtree of 18, we print the number 18 and go to its
right subtree and reach at node 20. Now we go to the left subtree of 20 i.e. NULL and
come back to 20 and print it. Then go to its right subtree, which is also NULL that
ends the recursion. Thus we have traversed the tree having node 20 as its root. This
tree (having 20 as root) is a right subtree of 18 which itself is a right subtree of 15.
We have already traversed the left subtree of 15 and gone through the whole tree
having 15 as root. This tree is a right subtree of the main tree with 14 as the root. We
have already traversed its left subtree. So we have traversed the whole tree by the
inorder process. From the figure 14.3 we see that by the inorder traversal we have
printed the numbers in the order as below
3, 4, 5, 7, 9, 14, 15, 16, 17, 18, 20
We have successfully demonstrated the preorder and inorder traversal of a tree by
using recursion method. The postorder traversal has been left for you as an exercise.
You can do it with the same above tree or build a tree on your own and traverse it.
The tree data structure by nature is a recursive data structure. In the coming lecture,
we will see that most of the methods we write for tree operations are recursive. From
the programming point of view, the recursion is implemented internally by using call
stack. This stack is maintained by the run time environment. The recursion is there in
the machine code generated by the compiler. It is the programmers responsibility to
provide a terminating condition to stop the recursion. Otherwise, it will become an
infinite recursion. If we do not put a terminating condition, the recursive calls will be
continued and the call stack will go on increasing. We know that when a program
executes, it becomes a process and some memory is allocated to it by the system for
its call stack. This memory has a limit. Thus when the recursive calls do not end, there
is no memory left to be used for increasing call stack after a certain time. This will
crash the program or the program will halt. So we should be very careful while using
the recursive calls and ensure the provision of a terminating condition in the recursive
calls.

Non Recursive Traversal

We can also carry out these traversing with the help of non-recursive calls. Think for
a few moments that how can we keep track of going left or going right or coming
back to the node with out using the recursion.. In the preorder, inorder and postorder
traversal, we visit nodes at different level and then return. We know the way to
comeback or go further deep. Now if we are not using the recursion, how these things
can be managed? We have to manage all this in our methods. Lets see how we can
write the non-recursive tree traversal methods. We can implement non-recursive
versions of the preorder, inorder and postorder traversal by using an explicit stack.
We cannot avoid stack. The stack will be used to store the tree nodes in the
appropriate order. Lets try to write methods without using recursion. For this
purpose, we have to create stack.
Here, for example, is the routine for inorder traversal that uses a stack.
void inorder(TreeNode<int>* root)

Page 157 of 519

CS301 Data Structures

___________________________________________________________________
{
Stack<TreeNode<int>* > stack;
TreeNode<int>* p;
p = root;
do
{
while( p != NULL )
{
stack.push( p );
p = p->getLeft();
}
// at this point, left tree is empty
if( !stack.empty() )
{
p = stack.pop();
cout << *(p->getInfo()) << " ";
// go back & traverse right subtree
p = p->getRight();
}
} while ( !stack.empty() || p != NULL );
}
The signature is same. The name of the function that is inorder, its return type is void
and the argument is pointer to TreeNode. Then there is stack abstract data type. We
can store int, float or other data type in the stack. Here we want to store the TreeNode
in the stack and then in the nodes we want to store integers. The statement is:
Stack<TreeNode<int>* > stack;
We send a message to the Stack factory that creates a stack to store TreeNode. The
integers will be stored in the TreeNode. We have almost two levels of template. Then
we declare a TreeNode pointer as p to traverse the tree up and down. We point this
pointer p to the root node. Then we have a do-while loop. The loop checks the value
of p is not NULL or stack is not empty. Inside the outer loop, we have an inner while
loop which checks that p is not NULL. If p is not NULL, we push p on the stack i.e.
we push the root on the stack. Then we replace the p with p->getLeft() . This shows
that p is now pointing to the left subtree node. In case of having a left subtree of p, we
come back in the while loop as p is not NULL. Again we push p on the stack. Now p
is pointing to the left subtree. You have presumed that this while loop will take us to
the left subtree of the root. On the exit from the while loop, we have pushed the node
in the stack whose left subtree is NULL. Here, p is pointing to NULL. Now we check
that stack is empty or not. If we have pushed some nodes in the stack, it will not be
empty. We pop a value from the stack and assign it to p before printing the info stored
in that node and assign p the right subtree of that node. We comeback to the outer dowhile loop. It will continue executing till stack is empty or p is NULL.
To understand this code, we will trace the code on the same tree, earlier used in the
recursive inorder function.

Page 158 of 519

CS301 Data Structures

___________________________________________________________________

14
15

9
7

18
20

16

17

Fig 14.3: Non-Recursive inorder

push(14)
..push(4)
....push(3)
3
4
..push(9)
....push(7)
......push(5)
5
7
9
14
push(15)
15
push(18)
..push(16)
16
..push(17)
17
18
push(20)
20

This is the same tree earlier discussed with reference to node as 14. We create a stack
and assign the root to the pointer p and have an inner while loop. In the while loop,
we pushed the root node in the stack i.e. push(14). After this, we assign the pointer p
to the left subtree of the root and return to while loop again. Now p is pointing to the
node with value 4, so we push 4 and then 3 on the stack. As the left subtree if 3 is null
so we exit from the inner loop. Afterwards, we pop the stack in the inner loop and
print it. As we know that stack is a LIFO structure (last in, first out). Finally, we have
pushed the node 3. So this node will be popped and the number 3 will be printed.
Now we will assign the right subtree of the node 3 to the p. As this is NULL, the inner
loop will not be executed and again come to the if statement. We will pop the stack in
the if statement and as a result, the node 4 will be popped and printed. Then we will
assign p to the right subtree of node 4 i.e. node 9. The control comes to the inner loop
in which we will push the node 9 and all of its left nodes one by one i.e. push(9),
push(7) and push(5). As the left subtree of the node 5 is NULL, the inner while loop
will be terminated. We will pop the stack and print 5. As the right subtree of node 5 is
NULL, the node 7 will be popped and printed. Here the right subtree of node 7 is
NULL. The stack is again popped and the number 9 will be printed. Here, the right
subtree of node 9 is also NULL. The stack will be popped resulting in popping and
printing of the node 14. Then we go the right subtree of the node 14 which is the node
15. The same rules of left subtree are applied here. You can understand it from the
above table.

Traversal Trace
Page 159 of 519

CS301 Data Structures

___________________________________________________________________
Lets compare the recursive inorder and non-recursive inorder and see whether there
is any difference between them. We find no difference between them. Function call
takes place with the help of stack. Explicit stack is also a stack and its behavior is also
the same. Lets compare these. In the left column, we have recursive inorder and on
the right column there is non-recursive inorder.
recursive inorder
nonrecursive inorder
inorder(14)
push(14)
..inorder(4)
..push(4)
....inorder(3)
....push(3)
3
3
4
4
..inorder(9)
..push(9)
....inorder(7)
....push(7)
......inorder(5)
......push(5)
5
5
7
7
9
9
14
14
inorder(15)
push(15)
15
15
inorder(18)
push(18)
..inorder(16)
..push(16)
16
16
..inorder(17)
..push(17)
17
17
18
18
inorder(20)
push(20)
20
20
In recursive inorder ,the function name is inorder(). We are printing the values. In
case of non-recursive, push() method is employed. You can see the order in which
tree nodes are being pushed explicitly on the stack. Then we print the values and use
the pop() method in a special order to see the values of the nodes.
In the recursive inorder we have inorder(14) while in the non-recursive inorder, there
is push(14). In the recursive inorder, we have inorder(4), in the non-recursive inorder
we have push(4). In the recursive inorder we have inorder(3. Whereas in the nonrecursive inorder, you may have push(3). It is pertinent to note that the only
difference in these two cases is that inorder() is in the left column while push() will
be found in the right column. You should not be surprised over this as stack is being
used in both recursive calls and non-recursive calls.
We can use recursive calls in which stack is used internally. However, in case of nonrecursive calls, the stack is used explicitly. Now the question arises which one of
these methods is better. Lets see some other aspects of recursion. When should we
use the recursion with respect to efficiency and implementation? How many
statements were there in the inorder recursive function. The first statement was an if
statement which is the terminating condition. After this, we have only three
statements, two calls and one cout statement. The order in the inorder is inorder(left),
cout and then inorder(right). Thats the complete function. Now see the non recursive
function. There are a lot of statements in this function so that there should be more

Page 160 of 519

CS301 Data Structures

___________________________________________________________________
code in the non-recursive function. The second thing is the readability. The
readability and understanding is better in the recursive function. One can easily
understand what is happening. To understand the non-recursive code, one has to read
the code with care. He has to go through every statement to see what you are doing. It
will also help ascertain if you are doing it right or not.
This is also an important aspect of programming. Program readability is also an issue.
Suppose you have written some program. Will you understand if after some months
that why you have written this and how? The first thing may be that what you are
doing in the program and how you do it? After going through the program, you will
remember it and recall that you have used this data structure for some purpose.
Always comment the code. Comment the data structure, logic and algorithm of the
program. Recursive procedure is an elegant procedure. Both the data structure and
procedure are recursive. We have traversed a tree with only three four statements. No
matter whatever is the size of the tree.
When the recursion happens with the help of function calls and stack.. There are some
other values also included. It has return address, local variables and parameters. When
a function calls another function irrespective of recursive or non recursive like
function F is calling function G. It will take time to put the values in the stack. If you
create your own stack, it takes time for creation and then push and pop will also
consume time. Now think that which one of these will take more time. The situation is
this that function calls always takes place using stack irrespective of the language.
The implementation of using stack is implemented very efficiently in the Assembly
language. In the computer architecture or Assembly language program, you will study
that the manipulation of stack calls that is push and pop and other methods are very
efficiently coded. Now you may think that there is a lot of work needed for the
recursive calls and non- recursive calls are faster. If you think that the non-recursive
function will work fast, it is wrong. The experience shows that if recursive calls are
applied on recursive data structures, it will be more efficient in comparison with the
non-recursive calls written by you. We will see more data structures which are
inherently recursive. If the data structures are recursive, then try to write recursive
methods. We will further discuss binary tree, binary search tree and see more
examples. Whenever we write a method dealing with these data structures and use
recursion. With the use of recursion, our program will be small, efficient and less
error prone. While doing programming, we should avoid errors. We dont want to see
there are errors while writing a program and executing it.
Try to do these traversals by hands on trees. These traversals are very important in
trees. We will write more methods while dealing with tree, used internally in the
inorder or postorder traversal.
There is yet another way of traversing a binary tree that is not related to recursive
traversal procedures discussed previously. In level-order traversal, we visit the nodes
at each level before proceeding to the next level. At each level, we visit the nodes in a
left-to-right order. The traversal methods discussed earlier are inorder, preorder and
post order. We print the node or to the left subtree or to the right subtree. In the
preorder, we traverse the complete left subtree before coming to the right subtree.
These procedures are in one way , depth-oriented. Now we may like to traverse the

Page 161 of 519

CS301 Data Structures

___________________________________________________________________
tree in a way that we go at one level and print all the nodes at that level. There are
some algorithm in which we need level order traversal.
In the following figure, the level order traversal is represented using the arrows.

14
4

15

9
7
5

Level-order:

18
16

20
17

14 4 15 3 9 18 7 16 20 5 17
Fig 14.4: level order traversal

We started from the root with 14 as a root node. Later, we printed its value. Then we
move to the next level. This is the binary tree so that we may have two elements at
next level i.e. 4 and 15. So we printed 4 and 15 while moving from left to right. Now
we are not moving to left or right subtree of any node. Rather, we move to the next
level. At next level, a programmer will have three nodes that are from left to right as
3, 9 and 18. We printed these values. If we take root node as zero level, it will be at
level 2. Now we move to the level 3. Here at level 3, we have 7, 16 and 20. At next
level, we have 5 and 17. So the numbers we are having are- 14 4 15 3 9 18 7 16
20 5 17. How can we programmatically traverse the tree in level order? We will
discuss it in the next lecture.

Page 162 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 15
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.3, 4.3.5, 4.6

Summary

Level-order Traversal of a Binary Tree

Storing Other Types of Data in Binary Tree
Binary Search Tree (BST) with Strings
Deleting a Node From BST

Level-order Traversal of a Binary Tree

In the last lecture, we implemented the tree traversal in preorder, postorder and
inorder using recursive and non-recursive methods. We left one thing to explore
further that how can we print a binary tree level by level by employing recursive or
non-recursive method.
See the figure below:

14
4

15

9
7

18
16

20
17

Fig 15.1: Level-order Traversal

In the above figure, levels have been shown using arrows:
At the first level, there is only one number 14.
Page 163 of 519

CS301 Data Structures

___________________________________________________________________
At second level, the numbers are 4 and 15.
At third level, 3, 9 and 18 numbers are present.
At fourth level the numbers are 7, 16 and 20.
While on fifth and final level of this tree, we have numbers 5 and 17.
This will also be the order of the elements in the output if the level-order traversal is
done for this tree. Surprisingly, its implementation is simple using non-recursive
method and by employing queue instead of stack. A queue is a FIFO structure, which
can make the level-order traversal easier.
The code for the level-order traversal is given below:
1. void levelorder( TreeNode <int> * treeNode )
2. {
3.
Queue <TreeNode<int> *> q;
4.
5.
if( treeNode == NULL ) return;
6.
q.enqueue(treeNode);
7.
while( !q.empty() )
8.
{
9.
treeNode = q.dequeue();
10.
cout << *(treeNode->getInfo()) << " ";
11.
if(treeNode->getLeft() != NULL )
12.
q.enqueue( treeNode->getLeft());
13.
if(treeNode->getRight() != NULL )
14.
q.enqueue( treeNode->getRight());
15.
}
16.
cout << endl;
17. }
The name of the method is levelorder and it is accepting a pointer of type TreeNode
<int>. This method will start traversing tree from the node being pointed to by this
pointer. The first line ( line 3) in this method is creating a queue q by using the Queue
class factory interface. This queue will be containing the TreeNode<int> * type of
objects. Which means the queue will be containing nodes of the tree and within each
node the element is of type int.
The line 5 is checking to see, if the treeNode pointer passed to the function is pointing
to NULL. In case it is NULL, there is no node to traverse and the method will return
immediately.
Otherwise at line 6, the very first node (the node pointed to by the treeNode pointer)
is added in the queue q.
Next is the while loop (at line 7), which runs until the queue q does not become
empty. As we have recently added one element (at line 6), so this loop is entered.
At line 9, dequeue() method is called to remove one node from the queue, the element
at front is taken out. The return value will be a pointer to TreeNode. In the next line
(line 10), the int value inside this node is taken out and printed. At line 11, we check
to see if the left subtree of the tree node (weve taken out in at line 9) is present. In
case the left subtree exists, it is inserted into the queue in the next statement at line 12.
Next, we see if the right subtree of the node is there, it is inserted into the queue also.

Page 164 of 519

CS301 Data Structures

___________________________________________________________________
Next statement at line 15 closes the while loop. The control goes back to the line 7,
where it checks to see if there is some element left in the queue. If it is not empty, the
loop is entered again until it becomes empty.
Lets execute this method levelorder( TreeNode <int> * ) by hand on a tree shown in
the fig 15.1 to understand the level order traversal better. We can see that the root
node of the tree is containing element 14. This node is inserted in the queue first, it is
shown in the below figure in gray shade.

14
4

15

18

16

20
17

Queue: 14
Output:
i

d i i

di

After insertion of node containing number 14, we reach the while loop statement,
where this is taken out from the queue and element 14 is printed. Next the left and
right subtrees are inserted in the queue. The following figure represents the current
stage of the tree.

Page 165 of 519

CS301 Data Structures

___________________________________________________________________

14
4

15

18

16

20
17

Queue: 4 15
Output: 14
i
i i d d
l
dd d i h
The element that has been printed on the output is shown with dark gray shade while
the elements that have been inserted are shown in the gray shade. After this the
control is transferred again to the start of the while loop. This time, the node
containing number 14 is taken out (because it was inserted first and then the right
node) and printed. Further, the left and right nodes (3 and 9) are inserted in the queue.
The following figure depicts the latest picture of the queue.

14
4

15

9
7
5

18
16

20
17

Queue: 15 3 9
Output: 14 4
i
In the above Queue, numbers 15, 3 and 9 have been inserted in the queue until now.
We came back to the while loop starting point again and dequeued a node from the
queue. This time a node containing number 15 came out. Firstly, the number 15 is
printed and then we checked for its left and right child. There is no left child of the

CS301 Data Structures

___________________________________________________________________
node (with number 15) but the right child is there. Right child is a node containing
number 18, so this number is inserted in the queue.

This time the node that is taken out is containing number 3. It is printed on the output.
Now, node containing number 3 does not have any left and right children, therefore,
no new nodes is inserted in the queue in this iteration.

14
4

15

9
7
5

18
16

20
17

Queue: 9 18
Output: 14 4 15 3
i we take out the node containing number 9. Print it on the output.
In the next iteration,
Look for its left subtree, it is there, hence inserted in the queue. There is no right
subtree of 9.

CS301 Data Structures

___________________________________________________________________

Similarly, if we keep on executing this function. The nodes are inserted in the queue
and printed. At the end, we have the following picture of the tree:

14
4

15

18

16

20
17

Queue:
Output: 14 4 15 3 9 18 7 16 20 5 17
i
As shown in the figure above, the Queue has gone empty and in the output we can see
that we have printed the tree node elements in the level order.
In the above algorithm, we have used queue data structure. We selected to use queue
data structure after we analyzed the problem and sought most suitable data structure.
The selection of appropriate data structure is done before we think about the
programming constructs of if and while loops. The selection of appropriate data
structure is very critical for a successful solution. For example, without using the

Page 168 of 519

CS301 Data Structures

___________________________________________________________________
queue data structure our problem of levelorder tree traversal would not have been so
easier.
What is the reason of choosing queue data structure for this problem. We see that for
this levelorder (level by level) tree traversal, the levels come turn by turn. So this turn
by turn guides us of thinking of using queue data structure.
Always remember that we dont start writing code after collecting requirements,
before that we select the appropriate data structures to be used as part of the design
phase. It is very important that as we are getting along on our course of data
structures, you should take care why, how and when these structures are employed.

Storing Other Types of Data in Binary Tree

Until now, we have used to place int numbers in the tree nodes. We were using int
numbers because they were easier to understand for sorting or comparison problems.
We can put any data type in tree nodes depending upon the problem where tree is
employed. For example, if we want to enter the names of the people in the telephone
directory in the binary tree, we build a binary tree of strings.

Binary Search Tree (BST) with Strings

Lets write C++ code to insert non-integer data in a binary search tree.
void wordTree()
{
TreeNode<char> * root = new TreeNode<char>();
static char * word[] = "babble", "fable", "jacket",
"backup", "eagle","daily","gain","bandit","abandon",
"abash","accuse","economy","adhere","advise","cease",
"debunk","feeder","genius","fetch","chain", NULL};
root->setInfo( word[0] );
for(i = 1; word[i]; i++ )
insert(root, word[i] );
inorder( root );
cout << endl;
}
The function name is wordTree(). In the first line, a root tree node is constructed, the
node will be containing char type of data. After that is a static character array word,
which is containing words like babble, fable etc. The last word or string of the array is
NULL. Next, we are putting the first word (babble) of the word array in the root node.
Further is a for loop, which keeps on executing and inserting words in the tree using
insert (TreeNode<char> *, char *) method until there is a word in the word array
(word is not NULL). You might have noticed that we worked in the same manner
when we were inserting ints in the tree. Although, in the int array, we used 1 as the
ending number of the array and here for words we are using NULL. After inserting the
whole array, we use the inorder( ) method to print the tree node elements.
Now, we see the code for the insert method.

Page 169 of 519

CS301 Data Structures

___________________________________________________________________
void insert(TreeNode<char> * root, char * info)
{
TreeNode<char> * node = new TreeNode<char>(info);
TreeNode<char> *p, *q;
p = q = root;
while( strcmp(info, p->getInfo()) != 0 && q != NULL )
{
p = q;
if( strcmp(info, p->getInfo()) < 0 )
q = p->getLeft();
else
q = p->getRight();
}
if( strcmp(info, p->getInfo()) == 0 )
{
cout << "attempt to insert duplicate: " << * info << endl;
delete node;
}
else if( strcmp(info, p->getInfo()) < 0 )
p->setLeft( node );
else
p->setRight( node );
}
The insert(TreeNode<char> * root, char * info) is accepting two parameters. First
parameter root is a pointer to a TreeNode, where the node is containing element char
type. The second parameter info is pointer to char.
In the first line of the insert method, we are creating a new TreeNode containing the
char value passed in the info parameter. In the next statement, two pointers p and q of
type TreeNode<char> are declared. Further, both pointers p and q start pointing to
the tree node passed in the first parameter root.
You must remember while constructing binary tree of numbers, we were incrementing
its count instead of inserting the new node again if the same number is present in the
node. One the other hand, if the same number was not there but the new number was
less than the number in the node, we used to traverse to the left subtree of the node. In
case, the new number was greater than the number in the node, we used to seek its
right subtree.
Similarly, in case of strings (words) here, we will increment the counter if it is already
present, and will seek the left and right subtrees accordingly, if required.
In case of ints we could easily compare them and see which one is greater but what
will happen in case of strings. You must remember, every character of a string has an
associated ASCII value. We can make a lexicographic order of characters based on
their ASCII values. For example, the ASCII value of B (66) is greater than A (65),
therefore, character B is greater than character A. Similarly if a word is starting with
the letter A, it is smaller than the words starting from B or any other character up to Z.
This is called lexicographic order.
C++ provides us overloading facility to define the same operators (<, >, <=, >=, ==
etc) that were used for ints to be used for other data types for example strings. We can
also write functions instead of these operators if we desire. strcmp is similar kind of

Page 170 of 519

CS301 Data Structures

___________________________________________________________________
function, part of the standard C library, which compares two strings.
In the code above inside the while loop, the strcmp function is used. It is comparing
the parameter info with the value inside the node pointed to by the pointer p. As info
is the first parameter of strcmp, it will return a negative number if info is smaller, 0 if
both are equal and a positive number if info is greater. The while loop will be
terminated if the same numbers are found. There is another condition, which can
cause the termination of loop that pointer q is pointing to NULL.
First statement inside the loop is the assignment of pointer q to p. In the second
insider statement, the same comparison is done again by using strcmp. If the new
word pointed to by info is smaller, we seek the left subtree otherwise we go to the
right subtree of the node.
Next, we check inside the if-statement, if the reason of termination of loop is
duplication. If it is, a message is displayed on the output and the newly constructed
node (that was to be inserted) is deleted (deallocated).
If the reason of termination is not duplication, which means we have reached to the
node where insertion of the new node is made. We check if the info is smaller than the
word in the current tree node. If this is the case, the newly constructed node is
inserted to the left of the current tree node. Otherwise, it is inserted to the right.
This insert() method was called from inside the for loop in the wordTree() method.
That loop is teminated when the NULL is reached at the end of the array word. At the
end, we printed the inserted elements of the tree using the inorder() method.
Following is the output of inorder():
Output:
abandon
abash
accuse
adhere
advise
babble
backup
bandit
cease
chain
daily
debunk
eagle
economy
fable
feeder
fetch
gain
genius
jacket
Notice that the words have been printed in the sorted order. Sorting is in increasing
order when the tree is traversed in inorder manner. This should not come as a surprise
if you consider how we built the binary search tree. For a given node, values less than
the info in the node were all in the left subtree and values greater or equal were in the

Page 171 of 519

CS301 Data Structures

___________________________________________________________________
right. Inorder prints the left subtree first, then the node itself and at the end the right
subtree.
Building a binary search tree and doing an inorder traversal leads to a sorting
algorithm.
We have found one way of sorting data. We build a binary tree of the data, traverse
the tree in inorder fashion and have the output sorted in increasing order. Although,
this is one way of sorting, it may not be the efficient one. When we will study sorting
algorithms, will prove Mathematically that which method is the fastest.

Deleting a Node From BST

Until now, we have been discussing about adding data elements in a binary tree but
we may also require to delete some data (nodes) from a binary tree. Consider the case
where we used binary tree to implement the telephone directory, when a person leaves
a city, its telephone number from the directory is deleted.
It is common with many data structures that the hardest operation is deletion. Once
we have found the node to be deleted, we need to consider several possibilities.
For case 1, If the node is a leaf, it can be deleted quite easily.
See the tree figure below.

6
2

4
3

Fig 15.9: BST

Suppose we want to delete the node containing number 3, as it is a leaf node, it is
pretty straight forward. We delete the leaf node containing value 3 and point the right
subtree pointer to NULL.
For case 2, if the node has one child, the node can be deleted after its parent adjusts a
pointer to bypass the node and connect to inorder successor.

6
2

4
3

Fig 15.10: Deleting a Node From BST

Page 172 of 519

CS301 Data Structures

___________________________________________________________________

If we want to delete the node containing number 4 then we have to adjust the right
subtree pointer in the node containing value 2 to the inorder successor of 4. The
important point is that the inorder traversal order has to be maintained after the
delete.

6
2

6
8

6
8

2
1

Fig 15.11: Deletion in steps

The case 3 is bit complicated, when the node to be deleted has both left and right
subtrees.
The strategy is to replace the data of this node containing the smallest data of the right
subtree and recursively delete that node.
Lets see this strategy in action. See the tree below:

6
2

5
3

Inorder successor

Page 173 of 519

CS301 Data Structures

___________________________________________________________________

In this tree, we want to delete the node containing number 2. Lets do inorder
traversal of this tree first. The inorder traversal give us the numbers: 1, 2, 3, 4, 5, 6
and 8.
In order to delete the node containing number 2, firstly we will see its right subtree
and find the left most node of it.
The left most node in the right subtree is the node containing number 3. Pay attention
to the nodes of the tree where these numbers are present. You will notice that node
containing number 3 is not right child node of the node containing number 2 instead it
is left child node of the right child node of number 2. Also the left child pointer of
node containing number 3 is NULL.
After we have found the left most node in the right subtree of the node containing
number 2, we copy the contents of this left most node i.e. 3 to the node to be deleted
with number 2.
6

4
4
4
Inorder
successor
Next step isFi
to delete
most node containing
15 13 the
d lleft
t (2)
th i d value 3. Now being the left most
node, there will be no left subtree of it, it might have a right subtree. Therefore, the
deletion of this node is the case 2 deletion and the delete operation can be called
recursively to delete the node.
6
6
3

8
3

5
1

3
4

Now if we traverse the tree in inorder, we get the numbers as: 1, 3, 4, 5, 6 and 8.

Page 174 of 519

CS301 Data Structures

___________________________________________________________________
Notice that these numbers are still sorted. In the next lecture, we will also see the C++
implementation code for this deletion operation.

Page 175 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 16
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.3 (all sub sections)

Summary

Deleting a node in BST

C++ code for remove
Binary Search Tree Class (BST)

Deleting a node in BST

In the previous lecture, we talked about deleting a node. Now we will discuss the
ways to delete nodes of a BST (Binary Search Tree). Suppose you want to delete one
of the nodes of BST. There are three cases for deleting a node from the BST. Case I:
The node to be deleted is the leaf node i.e. it has no right or left child. It is very
simple to delete such node. We make the pointer in the parent node pointing to this
node as NULL. If the memory for the node has been dynamically allocated, we will
release it. Case II: The node to be deleted has either left child (subtree) or right child
(subtree). Case III: The node to be deleted has both the left and right children
(subtree). This is the most difficult case. Lets recap this case. The following figure
shows the tree and the node to be deleted.

Page 176 of 519

CS301 Data Structures

___________________________________________________________________
6

Inorder successor will be the leftmost node in the right subtree of 2.

The inorder successor will not have a

left child because if it did, that child
would be the left-most node.

Inorder
successor

In the above example, the node to be deleted is 2. It has both right and left subtrees.
As this is the BST, so while traversing the tree in inorder, we will have sorted data.
The strategy is to find the inorder successor of the node to be deleted. To search the
inorder successor, we have to go to its right subtree and find the smallest element.
Afterwards we will copy this value into the node to be deleted. In the above example,
the number 3 is the smallest in the right subtree of the node 2. A view of data of
inorder traversal shows that number 3 is the inorder successor of number 2. We copy
the number 3 in the node to be deleted. Now in this transition tree, the number 3 is at
two places. Now we have to delete the inorder successor node i.e. node 3. The node 3
has only right child. So this is the Case II of deletion.
6

Page 177 of 519

CS301 Data Structures

___________________________________________________________________
6

We will get the inorder successor of node 3 i.e. number 4. Therefore we will connect
the node 5 with node 4 that is the left pointer of 5 which was earlier pointing to node
3. But now it is pointing to node 4. We delete the node 3. Remember that number 3
has been copied in the node 2. This way, we will delete the node having right and left
both subtrees.

C++ code for remove

Now we will see the C++ code to carry out this deletion. In the above figures, we
have named the method as delete, but delete is a keyword of C++. So we cannot write
a method named delete. We have to name it something else, say remove. Keep in
mind all the three cases of delete and the figures.
Here is the code of the remove method.
/* This method is used to remove a node from the BST */
TreeNode<int>* remove(TreeNode<int>* tree, int info)
{
TreeNode<int>* t;
int cmp = info - *(tree->getInfo());
if( cmp < 0 ){
// node to delete is in left subtree
t = remove(tree->getLeft(), info);
tree->setLeft( t );
}
else if( cmp > 0 ){
// node to delete is in right subtree
t = remove(tree->getRight(), info);
tree->setRight( t );
}
Page 178 of 519

CS301 Data Structures

___________________________________________________________________
//two children, replace with inorder successor
else if(tree->getLeft() != NULL && tree->getRight() != NULL ){
TreeNode<int>* minNode;
MinNode = findMin(tree->getRight());
tree->setInfo( minNode->getInfo() );
t = remove(tree->getRight(), *(minNode->getInfo()));
tree->setRight( t );
}
else { // case 1
TreeNode<int>* nodeToDelete = tree;
if( tree->getLeft() == NULL ) //will handle 0 children
tree = tree->getRight();
else if( tree->getRight() == NULL )
tree = tree->getLeft();
else tree = NULL;
delete nodeToDelete; // release the memory
}
return tree;
}
The return type of the remove method is a pointer to TreeNode. The argument list
contains two arguments, the pointer to TreeNode pointing the root of the tree and an
integer info containing the value of the node to be deleted. A thorough analysis of the
delete method reveals that you need these two things.
Here we are dealing with the templates and the nodes containing the int values. Inside
the function, we create a temporary variable t of type TreeNode pointer. Then we
have int type variable cmp. The value of the TreeNode provided in the argument is
subtracted from the info (this is the value of the node to be deleted) and assigned to
the cmp. Then we have conditional if statement which checks if the value of cmp is
less than 0, we call remove method recursively. As the cmp is less than zero, it means
that the info is not the node provided in the argument. The value of cmp is less than
zero, depicting that the value of the current node is greater than info. Therefore if the
node with value info exists, it should be in the left subtree. So we call the remove
method on the left subtree. When we return from this call, the left subtree pointer of
tree is assigned to our temporary variable t.
If the first if statement evaluates to true, it means that our required node is in the left
subtree of the current node. We call the remove method recursively and remove that
node. On coming back, we assign t to the left subtree pointer of the tree. Suppose that
the node to be deleted is the immediate left child of the node. It will be deleted by the
remove method. When the node is deleted, some other node will come to its place,
necessitating the readjustment of the left pointer of the parent node. We can do it by
pointing the left subtree pointer to t.
If the value of cmp is greater than zero, the node with info value should be in the right
subtree of the current node. Therefore we call the remove method recursively,
providing it the pointer to the right subtree and the info. When we return from that

Page 179 of 519

CS301 Data Structures

___________________________________________________________________
call, the right subtree is readjusted by assigning it the t.
In the third case, we are at the node, which is to be deleted. Here we have to deal with
the three delete cases- 1) the node to be deleted is the leaf node 2) node to be deleted
has either right or left child and 3) the node to be deleted has both left and right
subtree. We will begin with the third case. We will check in the else-if statement that
the left and right subtrees are not NULL. It means that the node has both left and right
subtrees. We create a temporary variable minNode of type TreeNode pointer. This
variable minNode is local to this if-else block. We assign it value by calling the
findMin method. We call the findMin method as:
findMin(tree->getRight())
In this delete case, we need to find the inorder successor of the node to be deleted
which is the minimum value in the right subtree. Then we will put the value of this
node to the node to be deleted. This node is saved in the variable minNode. The node
to be deleted is now pointed by tree. Then we get the value of the inorder successor,
stored in the minNode and assign it to the node to be deleted. Now we have to delete
the minNode. For this purpose, we call the remove method recursively providing it the
right subtree and the value of the minNode. We have also changed the value of the
current node and can also send this value to the remove method. For clarity purposes,
we provide it the minNode value. The node returned by the remove method, is stored
in the temporary variable t. Then we assign t to the right subtree of tree. We do it due
to the fact that the tree has to re-adjusted after the deletion of a node in the right
subtree.
Now we will talk about other two cases i.e. the node to be deleted is the leaf node or
has left or right child. We will deal with these two cases together. The else statement
is being employed in these cases. We create a temporary variable nodeToDelete and
assign it to the tree. Now we have to check whether this is the leaf node or has either
left or right child. First we check that the left subtree of the tree is NULL before
assigning the right subtree to the tree. Then we check that the right subtree of the tree
is NULL. If it is NULL, the tree will be replaced by the left subtree. In the end, we
make the tree NULL and delete the nodeToDelete. Here the memory is released.
While dealing with this case of delete, we find the inorder successor from the right
subtree i.e. the minimum value in the right subtree. That will also be the left most
node of the right subtree. We use the findMin method for this purpose. Lets have a
look on the code of this method.
/* This method is used to find the minimum node in a tree
*/
TreeNode<int>* findMin(TreeNode<int>* tree)
{
if( tree == NULL )
return NULL;
if( tree->getLeft() == NULL )
return tree; // this is it.

Page 180 of 519

CS301 Data Structures

___________________________________________________________________
return findMin( tree->getLeft() );
}
The return type is a pointer to TreeNode which will be the minimum node. The input
argument is also a pointer to TreeNode. This input argument is a root of some subtree
in which we have to find the minimum node. Inside the function, we have a couple of
if statements. If the tree is NULL, it will result in the return of the NULL. If the left
child of the tree is NULL, the tree will be returned. In the end, we are calling findMin
recursively providing it the left subtree as argument. When we were talking about the
inorder successor or the left most node, it was stated that this node will not contain a
left child because if it has a left child then it will not be a left most child. The leftmost node is the node which either does not have a left child or its left pointer is
NULL. The above method works the same way. It looks for the left node having left
pointer NULL. This is the node with the minimum value. You must have noted that in
this function, we are making a recursive call in the end of the function. If a function
has recursive call as the last statement, it is known as tail recursion. We can replace
the tail recursion with the loop. How can we do that? This is an exercise for you. In
case of tail recursion, you have either to use the tail recursion or loop. While opting
for loop, you will get rid of the recursion stack.
You can trace this function on the sample tree we have been using in the previous
lecture. It can also be hand-traced. Provide the root of the BST to the method and try
to find the minNode while employing the above method. You will see that the value
returned by this function is the minimum node. Also try to write the FindMin in a
non-recursive way i.e. use of the loop. You will get the same result.

Binary Search Tree Class (BST)

Lets see the code of the binary search tree (BST). We put the interface in the .h file.
We also have the state variable in addition to the public and private methods of the
class in it. In the public methods, the user of the class calls with their objects. Lets
create a .h file for binary search tree and the objects of binary search tree obtained
from the factory will use the attributes defined in the .h file. Here is the .h file.
/* binarysearchtree.h file contains the interface of binary search tree (BST) */
#ifndef _BINARY_SEARCH_TREE_H_
#define _BINARY_SEARCH_TREE_H_
#include <iostream.h>

// For NULL

// Binary node and forward declaration

template <class EType>
class BinarySearchTree;
template <class EType>
class BinaryNode
{
EType element;
BinaryNode *left;
Page 181 of 519

CS301 Data Structures

___________________________________________________________________
BinaryNode *right;
public:
// constructor
BinaryNode( const EType & theElement, BinaryNode *lt, BinaryNode *rt )
: element( theElement ), left( lt ), right( rt ) { }
friend class BinarySearchTree<EType>;
};
// continued
At the start of the file, we have conditional statement using the ifndef that checks a
constant _BINARY_SEARCH_TREE_H_. If this constant is not defined, we define it
with the help of define keyword. This is a programming trick. Sometimes, we include
the same .h file more than once. If we do not use this trick, the compiler will give
error on the inclusion of the file for the second time. Here, we have a statement:
class BinarySearchTree;
This is the forward declaration of the class BinarySearchTree. We have not defined
this class so far. First we define the class BinaryNode and then by combining these
nodes, a binary tree is created. This class is a template class. The nodes can contain
data other than integers. Then we define the BinaryNode class and define three state
variables. The data of the node is stored in the element which is of type EType. Being
a binary node, it may have a left and right child. We have declared two pointers for
this node. Then we have its constructor that takes three arguments, used to initialize
the element, left and right variables. We wrote the colon(:) and the initialization list.
There are three state variables in the BinaryNode. These are element, left and right.
When we create an object of BinaryNode, all of these variables are created. The
EType can also be some class so its object will be created and a call to its constructor
is made. Whenever an object is created, its constructor will be called. We have to
provide the arguments. If we do not call the constructor explicitly, the default
constructor will be called. In the constructor of BinaryNode, after the colon, we have
written element(theElement). (Here theElement is the argument in the constructor of
BinaryNode). It seems to be a function call but actually it is a call to the constructor
for element. If element was of type String, constructor of String will be called and
theElement is sent as argument. Through this way of writing the variables after the
colon, we are actually calling the explicit constructor for the state variables. While
calling the constructor, we are not using the name of the class but the variable names.
There should be no ambiguity. Then we have left( lt ) and right( rt ), as left and right
are of type BinaryNode so the constructor for BinaryNode will be called. The body of
the constructor of the BinaryNode is empty.
You might not be familiar with this type of coding but it is commonly used in C++
coding. You can get further information on this from the C++ book or from the
internet. In some cases, a programmer is forced to use this syntax for having no other
option. The last statement of the class definition has a friend class:
friend class BinarySearchTree<EType>;

Page 182 of 519

CS301 Data Structures

___________________________________________________________________
You are familiar with the friend keyword. The class BinaryNode has defined the
BinarySearchTree class as its friend class. It means that the BinarySearchTree class
can access the state variables of BinaryNode class. We can also declare friend method
of some class which can access the state variable of that class. We have not defined
the BinarySearchTree class yet. We have just declared the class with the name
BinarySearchTree as the forward declaration. This forward declaration is for
compilers. When the compiler reads the file from top to bottom and encounters the
line defining the BinarySearchTree as friend, it gives an error. The compiler does not
know about the BinarySearchTree. To overcome this error we use forward
declaration. With this forward declaration, the compiler is communicated that we
want to use the BinarySearchTree as a class. Therefore, when the compiler reads the
line, which defines the BinarySearchTree as a friend of BinaryNode, it knows that it
is a class, which used EType template and will be defined later. Now there is no
ambiguity for the compiler. We must have to define this class later on.
Lets see the code of BinarySearchTree. We are giving it EType template parameter.
We want to make our BinarySearchTree generic and it can be used with integers,
strings, characters or with some other data type. Therefore, we have defined it as a
template class. The user of this class will create a BinarySearchTree of its specific
data type. The code is as follows:
/* binarysearchtree.h file also contains the definition of the
BinarySearchTree */
template <class EType>
class BinarySearchTree
{
public:
BinarySearchTree( const EType& notFound );
BinarySearchTree( const BinarySearchTree& rhs );
~BinarySearchTree( );
const EType& findMin( ) const;
const EType& findMax( ) const;
const EType& find( const EType & x ) const;
bool isEmpty( ) const;
void printInorder( ) const;
void insert( const EType& x );
void remove( const EType& x );
const BinarySearchTree & operator = ( const BinarySearchTree & rhs
);
private:
BinaryNode<EType>* root;
// ITEM_NOT_FOUND object used to signal failed finds
const EType ITEM_NOT_FOUND;

Page 183 of 519

CS301 Data Structures

___________________________________________________________________
const EType& elementAt( BinaryNode<EType>* t );
void insert(const EType& x, BinaryNode<EType>* & t);
void remove(const EType& x, BinaryNode<EType>* & t);
BinaryNode<EType>* findMin(BinaryNode<EType>* t);
BinaryNode<EType>* findMax(BinaryNode<EType>* t);
BinaryNode<EType>* find(const EType& x, BinaryNode<EType>* t
);
void makeEmpty(BinaryNode<EType>* & t);
void printInorder(BinaryNode<EType>* t);
};
#endif
We start the definition of our class with the public interface. The user of this class is
interested in the public interface of the class. These are the methods, which the user
can employ. We have two constructors for this class. One constructor takes EType
reference parameter while other takes BinarySearchTree as a parameter. The types of
constructors depend on the usage. We can have more constructors if needed. Then we
have a destructor of BinarySearchTree. Besides, we have interface methods for
BinarySearchTree. There are also findMin and findMax methods, which will return
the minimum and maximum value of EType in the tree respectively. We have used the
const keyword with these functions. The reference variables in functions are also
being used. Next thing we are going to discuss is the find method. Its signature is as
under:
const EType& find( const EType & x ) const;
This method will takes an argument x of EType. It will search the tree whether x exists
in the tree or not. Then we have isEmpty method that will ascertain if the tree is empty
or not. There is also the printInorder method, which will print the tree in inorder
traversal. If the tree has integers, we have sorted integers as a result of inorder
traversal. Next thing we have is the insert (const EType& x) method, which inserts the
x as a new node in the tree. After this, there is the remove method i.e.delete method.
We have renamed delete as remove because delete is a keyword of C++. We can also
name it as deleteNode or some other name which you think is suitable. The interface
of the class is almost complete. Now think as a user of BinarySearchTree and decide
if you need more methods. With the help of methods defined above, we can do a lot
of work on BinarySearchTree. If you feel to add more methods due to the usage need,
these can be added later.
We also need some private variables and methods for this class. At First, we have
defined a pointer to BinaryNode as root before defining an EType variable as
ITEM_NOT_FOUND. Then there are elementAt, insert and remove methods. But the
method signatures of these methods are different than public methods. The signature
of insert method is:
void insert(const EType& x, BinaryNode<EType>* & t);
In the public insert method, we have only one argument of type EType. Similarly
there are findMin, findMax and find methods with different signatures. Then we have

Page 184 of 519

CS301 Data Structures

___________________________________________________________________
makeEmpty and printInorder methods. Why we are talking about these methods
especially when these are the private ones. In C++, we have to do like this. In .h file
we define the class and in .cpp file, the implementation of the class is given. The user
of this class is advised not to look at the private part of the class. He should be
interested only in the public part of the class. We are not the users of this class, but
only the developers. In the BinarySeachTree.h file, we have defined two classes i.e.
BinaryNode and BinarySearchTree. Are these classes private or public? Can the user
of BinarySearchTree benefit from the BinaryNode class? Is it possible or not? This is
an exercise for you. Study about the reference data types before coming to next
lecture.

Sample Program
Here is the code of the program. BinarySearchTree.h file.
/* This file contains the declaration of binary node and the binary search tree */
#ifndef BINARY_SEARCH_TREE_H_
#define BINARY_SEARCH_TREE_H_
#include <iostream.h>

// For NULL

// Binary node and forward declaration

template <class EType>
class BinarySearchTree;
template <class EType>
class BinaryNode
{
EType element;
BinaryNode *left;
BinaryNode *right;
BinaryNode( const EType & theElement, BinaryNode *lt, BinaryNode *rt )
: element( theElement ), left( lt ), right( rt ) { }
friend class BinarySearchTree<EType>;
};
// BinarySearchTree class
//
// CONSTRUCTION: with ITEM_NOT_FOUND object used to signal failed
finds
//
// ******************PUBLIC OPERATIONS*********************
// void insert( x )
--> Insert x
// void remove( x )
--> Remove x
// EType find( x ) --> Return item that matches x
// EType findMin( ) --> Return smallest item
// EType findMax( ) --> Return largest item
Page 185 of 519

CS301 Data Structures

___________________________________________________________________
// boolean isEmpty( ) --> Return true if empty; else false
// void makeEmpty( )
--> Remove all items
// void printTree( )
--> Print tree in sorted order
template <class EType>
class BinarySearchTree
{
public:
BinarySearchTree( const EType & notFound );
BinarySearchTree( const BinarySearchTree & rhs );
~BinarySearchTree( );
const EType & findMin( ) const;
const EType & findMax( ) const;
const EType & find( const EType & x ) const;
bool isEmpty( ) const;
void printTree( ) const;
void makeEmpty( );
void insert( const EType & x );
void remove( const EType & x );
const BinarySearchTree & operator=( const BinarySearchTree & rhs );
private:
BinaryNode<EType> *root;
const EType ITEM_NOT_FOUND;
const EType & elementAt( BinaryNode<EType> *t ) const;
void insert( const EType & x, BinaryNode<EType> * & t ) const;
void remove( const EType & x, BinaryNode<EType> * & t ) const;
BinaryNode<EType> * findMin( BinaryNode<EType> *t ) const;
BinaryNode<EType> * findMax( BinaryNode<EType> *t ) const;
BinaryNode<EType> * find( const EType & x, BinaryNode<EType> *t )
const;
void makeEmpty( BinaryNode<EType> * & t ) const;
void printTree( BinaryNode<EType> *t ) const;
BinaryNode<EType> * clone( BinaryNode<EType> *t ) const;
};
#include "BinarySearchTree.cpp"
#endif
BinarySearchTree.cpp file.
/* This file contains the implementation of the binary search tree */
#include <iostream.h>

Page 186 of 519

CS301 Data Structures

___________________________________________________________________
#include "BinarySearchTree.h"
/**
* Construct the tree.
*/
template <class EType>
BinarySearchTree<EType>::BinarySearchTree( const EType & notFound ) :
ITEM_NOT_FOUND( notFound ), root( NULL )
{
}
/**
* Copy constructor.
*/
template <class EType>
BinarySearchTree<EType>::
BinarySearchTree( const BinarySearchTree<EType> & rhs ) :
root( NULL ), ITEM_NOT_FOUND( rhs.ITEM_NOT_FOUND )
{
*this = rhs;
}
/**
* Destructor for the tree.
*/
template <class EType>
BinarySearchTree<EType>::~BinarySearchTree( )
{
makeEmpty( );
}
/**
* Insert x into the tree; duplicates are ignored.
*/
template <class EType>
void BinarySearchTree<EType>::insert( const EType & x )
{
insert( x, root );
}
/**
* Remove x from the tree. Nothing is done if x is not found.
*/
template <class EType>
void BinarySearchTree<EType>::remove( const EType & x )
{
remove( x, root );
}

Page 187 of 519

CS301 Data Structures

___________________________________________________________________

/**
* Find the smallest item in the tree.
* Return smallest item or ITEM_NOT_FOUND if empty.
*/
template <class EType>
const EType & BinarySearchTree<EType>::findMin( ) const
{
return elementAt( findMin( root ) );
}
/**
* Find the largest item in the tree.
* Return the largest item of ITEM_NOT_FOUND if empty.
*/
template <class EType>
const EType & BinarySearchTree<EType>::findMax( ) const
{
return elementAt( findMax( root ) );
}
/**
* Find item x in the tree.
* Return the matching item or ITEM_NOT_FOUND if not found.
*/
template <class EType>
const EType & BinarySearchTree<EType>::
find( const EType & x ) const
{
return elementAt( find( x, root ) );
}
/**
* Make the tree logically empty.
*/
template <class EType>
void BinarySearchTree<EType>::makeEmpty( )
{
makeEmpty( root );
}
/**
* Test if the tree is logically empty.
* Return true if empty, false otherwise.
*/
template <class EType>
bool BinarySearchTree<EType>::isEmpty( ) const
{

Page 188 of 519

CS301 Data Structures

___________________________________________________________________
return root == NULL;
}
/**
* Print the tree contents in sorted order.
*/
template <class EType>
void BinarySearchTree<EType>::printTree( ) const
{
if( isEmpty( ) )
cout << "Empty tree" << endl;
else
printTree( root );
}
/**
* Deep copy.
*/
template <class EType>
const BinarySearchTree<EType> &
BinarySearchTree<EType>::
operator=( const BinarySearchTree<EType> & rhs )
{
if( this != &rhs )
{
makeEmpty( );
root = clone( rhs.root );
}
return *this;
}
/**
* Internal method to get element field in node t.
* Return the element field or ITEM_NOT_FOUND if t is NULL.
*/
template <class EType>
const EType & BinarySearchTree<EType>::
elementAt( BinaryNode<EType> *t ) const
{
if( t == NULL )
return ITEM_NOT_FOUND;
else
return t->element;
}
/**
* Internal method to insert into a subtree.
* x is the item to insert.
* t is the node that roots the tree.

Page 189 of 519

CS301 Data Structures

___________________________________________________________________
* Set the new root.
*/
template <class EType>
void BinarySearchTree<EType>::
insert( const EType & x, BinaryNode<EType> * & t ) const
{
if( t == NULL )
t = new BinaryNode<EType>( x, NULL, NULL );
else if( x < t->element )
insert( x, t->left );
else if( t->element < x )
insert( x, t->right );
else
; // Duplicate; do nothing
}
/**
* Internal method to remove from a subtree.
* x is the item to remove.
* t is the node that roots the tree.
* Set the new root.
*/
template <class EType>
void BinarySearchTree<EType>::
remove( const EType & x, BinaryNode<EType> * & t ) const
{
if( t == NULL )
return; // Item not found; do nothing
if( x < t->element )
remove( x, t->left );
else if( t->element < x )
remove( x, t->right );
else if( t->left != NULL && t->right != NULL ) // Two children
{
t->element = findMin( t->right )->element;
remove( t->element, t->right );
}
else
{
BinaryNode<EType> *nodeToDelete = t;
t = ( t->left != NULL ) ? t->left : t->right;
delete nodeToDelete;
}
}
/**
* Internal method to find the smallest item in a subtree t.
* Return node containing the smallest item.
*/

Page 190 of 519

CS301 Data Structures

___________________________________________________________________
template <class EType>
BinaryNode<EType> *
BinarySearchTree<EType>::findMin( BinaryNode<EType> *t ) const
{
if( t == NULL )
return NULL;
if( t->left == NULL )
return t;
return findMin( t->left );
}
/**
* Internal method to find the largest item in a subtree t.
* Return node containing the largest item.
*/
template <class EType>
BinaryNode<EType> *
BinarySearchTree<EType>::findMax( BinaryNode<EType> *t ) const
{
if( t != NULL )
while( t->right != NULL )
t = t->right;
return t;
}
/**
* Internal method to find an item in a subtree.
* x is item to search for.
* t is the node that roots the tree.
* Return node containing the matched item.
*/
template <class EType>
BinaryNode<EType> *
BinarySearchTree<EType>::
find( const EType & x, BinaryNode<EType> *t ) const
{
if( t == NULL )
return NULL;
else if( x < t->element )
return find( x, t->left );
else if( t->element < x )
return find( x, t->right );
else
return t; // Match
}
/****** NONRECURSIVE VERSION*************************
template <class EType>
BinaryNode<EType> *
BinarySearchTree<EType>::

Page 191 of 519

CS301 Data Structures

___________________________________________________________________
find( const EType & x, BinaryNode<EType> *t ) const
{
while( t != NULL )
if( x < t->element )
t = t->left;
else if( t->element < x )
t = t->right;
else
return t; // Match
return NULL; // No match
}
*****************************************************/
/**
* Internal method to make subtree empty.
*/
template <class EType>
void BinarySearchTree<EType>::
makeEmpty( BinaryNode<EType> * & t ) const
{
if( t != NULL )
{
makeEmpty( t->left );
makeEmpty( t->right );
delete t;
}
t = NULL;
}
/**
* Internal method to print a subtree rooted at t in sorted order.
*/
template <class EType>
void BinarySearchTree<EType>::printTree( BinaryNode<EType> *t ) const
{
if( t != NULL )
{
printTree( t->left );
cout << t->element << endl;
printTree( t->right );
}
}
/**
* Internal method to clone subtree.
*/
template <class EType>
BinaryNode<EType> *

Page 192 of 519

CS301 Data Structures

___________________________________________________________________
BinarySearchTree<EType>::clone( BinaryNode<EType> * t ) const
{
if( t == NULL )
return NULL;
else
return new BinaryNode<EType>( t->element, clone( t->left ), clone( t>right ) );
}
TestBinarySearchTree.cpp file. This file contains the main program.
/* This file contains the test program for the binary search tree */
#include <iostream.h>
#include "BinarySearchTree.h"
// Test program
int main( )
{
const int ITEM_NOT_FOUND = -9999;
BinarySearchTree<int> t( ITEM_NOT_FOUND );
int NUMS = 30;
int i;
cout << "Inserting elements (1 to 30) in the tree .......)" << endl;
for( i = 0; i <= NUMS; i++ )
t.insert( i );
cout << "Printing the values of the nodes in tree .......)" << endl;
t.printTree();
cout << "Removing the even number elements in the tree .......)" << endl;
for( i = 0; i <= NUMS; i+= 2 )
t.remove( i );
cout << "Printing the values of the nodes in tree .......)" << endl;
t.printTree();
int abc;
cin >> i;
return 0;
}
When we run the TestBinarySearch program the following output is obtained.
Inserting elements (1 to 30) in the tree .......)
Printing the values of the nodes in tree .......)
0
1
2

Page 193 of 519

CS301 Data Structures

___________________________________________________________________
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Removing the even number elements in the tree .......)
Printing the values of the nodes in tree .......)
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29

Page 194 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 17
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4

Summary

Reference Variables
Sample Program

After demonstrating the use of const and reference variables in the interface class for
a binary search tree in the previous lectures, we will now see what these reference
variables are and how does internal process go on during the use of reference
variables.

Reference Variables
Before proceeding ahead, there is need to know why the reference variables are used;
and what benefits a programmer can get while employing them. We will discuss these
all things in detail with the help of examples.
The symbol &, used for reference variable has a few different purposes with respect
to its occurrence in the code. In C++ programming, we have seen that when the
ampersand sign i.e. & appears in front of a variable name, it is the address operator. It
returns the address of the variable in front of which it is written. Thus for example, if
x is a variable name, then &x ; will return the address of the variable x. In general we
can say that
Page 195 of 519

CS301 Data Structures

___________________________________________________________________
&variablename ;
will return the address of the variable. We know that an address can be stored in a
pointer. To further understand this concept, lets suppose that there are following lines
in our code.
int x ;
int* ptr = &x;
The first line declares a variable x of type int while the second one declares a pointer
to int and assigns it the address of x. This address of variable x is returned by the &
sign written in front of the variable x. Thus a pointer variable is initialized by
assigning it the address of a variable. This address of the variable is gotten by using
the & sign with the variable name.
The other place where & sign can appear is the signature of the function where it
appears after the type of the parameter. Consider the insert and remove methods from
BinarySearchTree class that were declared as the following.
void insert( const EType& x );
void remove( const EType& x );
Notice that the & sign is after the type. Here we define the class and functions as
templates. Whenever, we use these methods, the type EType will be replaced with a
proper data type.
Suppose that we have designed the BinarySearchTree class to hold the integers only.
This means that there are no templates and the class can be defined only for integers.
So the insert and remove methods of this class will be as follows.
void insert( const int& x );
void remove( const int& x );
Here, in the function signature, & sign after the type of the parameter indicates that
the parameter is a reference variable. We can use this & sign with any data type i.e.
built- in or user defined, of an argument to show that this is a reference variable. This
is the syntax for the reference variable. Now lets see why and how we use it and
what is its advantage? Lets look at an example that will explain the reference
variable. Suppose we have the following three different functions.
The name of the first function is intMinus1, written as under.
//Function 1
int intMinus1( int oldVal)
{
oldVal = oldVal 1;
return oldVal;
}
This function takes an integer as an argument. The local name of this argument is

Page 196 of 519

CS301 Data Structures

___________________________________________________________________
oldVal. The first statement in the body of the function decreases the value of oldVal
by 1 then the next statement returns this value to the calling function.
The second function is intMinus2. This function takes a pointer to integer as an
argument. In the statement
*oldVal = *oldVal 2 ;
*oldVal means the value at the location where oldVal is pointing to. To get the value
this way is called dereferencing. After experiencing a decrease by 2, this value
remains at the same location i.e. where oldVal is pointing to. Then the pointer to
integer is returned. This function is defined as follows.
// Function 2
int intMinus2( int* oldVal)
{
*oldVal = *oldVal 2;
return *oldVal;
}
The third function intMinus3 takes a reference variable as an argument. Its definition
is as under.
// Function 3
int intMinus3( int& oldVal)
{
oldVal = oldVal 3;
return oldVal;
}
The & sign after the type in the signature (declaration line) indicates that this
argument is a reference variable. Notice that & sign is used only in the function
declaration, leaving no need for its use in the body of the function. The statements in
the body of the function decrease the vale of oldVal by 3 and return it to the calling
function.
We can see that these three functions have the same return type and their body
performs the task of decreasing the passed value. However it is clear from the
declaration of these functions that the way of passing argument to these functions is
different. We will now see what changes occur while passing the argument in
different ways. We will write calling functions those call the above three functions
one by one. We will see how these calls are made and what differences happen during
the process of passing same value (in different ways) to these functions. The use of
call stack will help in ascertaining what happens to these local variables and
parameters.
Following is the function that calls the first function i.e. intMinus1.

Page 197 of 519

CS301 Data Structures

___________________________________________________________________
void caller()
{
int myInt = 31;
int retVal;
retVal = intMinus1( myInt );
cout << myInt << retVal;
}
In this function (caller), we declare an integer variable myInt and assign it a value 31.
We also declare a variable retVal of type int. In the statement
retVal = intMinus1( myInt );
We call the function intMinus1 by passing it the variable myInt (the value of which is
31) and assign the return value of this function call to retVal. In the next statement,
we print the value of these two variables.
While talking about runtime environment, we noticed that an executable program
while in run, is loaded in the memory. It later becomes a process. This process is
given a block of memory, which it uses during its execution. Suppose we are running
many programs simultaneously i.e. browser, MS Word, Excel and dev-C++. We can
also run programs written by us. The figure 17.1 shows that process4 is a program
written by us. Every program, which we run takes a block of memory and becomes a
process. The memory occupied by a process is further partitioned into different parts.
The first part of the memory is for the code of the program. This code is in binary
form i.e. it is the compiled code. Afterwards, there is some area of memory for static
data that holds static and global variables. Then in the memory, there becomes the
stack. This stack is used in function calls. At the end, there is some area, called heap.
When we allocate memory dynamically by using new operator, this memory is
allocated from the heap. The following figure shows the internal memory
organization of a program (ourtest.exe).

Process 1
(browser)

Code

Process 3
(Word)

Static data

Process 4
(ourtest.exe)

Stack

Process 2
(Dev-C++)
Windows Os

Heap

Page 198 of 519

CS301 Data Structures

___________________________________________________________________
Fig 17.1: Internal memory organization of a process (ourtest.exe)
We have discussed in detail the call stack that is used in function calls. We know that
when a function call is made, its all the arguments go on the stack. The return
address of the function also goes to the stack. The local variables are placed on the
stack. When a function calls some other function, an activation record is made. It has
much information including all these things (parameters, return address etc). The
detail of activation record relates to the compiler construction course. Here we are
concerned with only the call stack. Following figure shows the call stack layout.
Parameters (caller)
Local variables(caller)
Return address (caller)

sp

Parameters (intMinus1)
Local
variables(intMinus1)
R
dd
stack grows downwards
Fig 17.2: call stack layout

In the stack layout figure, there are the entries of the caller in the upper portion. This
is due to the fact that the caller function itself was called (executed) by some one else.
The lower portion of the stack describes the stack status when the function intMinus1
is called. The stack contains parameters, local variables and return address of the
function intMinus1. There is a stack pointer sp that points to the top of the stack (the
figure shows the stack downward; it should not be confused with the top). As the
caller has called intMinus1, this function is on the top of the stack.
The following figure shows the contents of the stack when the function intMinus1 is
called.

1072
1068

31
?

myInt
retVal

calling function caller

callers other
stuff
1060

31

1056

oldVal
Called function

1052

sp
stack grows downward

Page 199 of 519

Fig 17.3: call stack layout when intMinus1 is called

CS301 Data Structures

___________________________________________________________________

The figure shows the stack in two parts. The first part that is in the upper curly
bracket, shows the contents of the calling function i.e. caller. The lower part shows
the contents of the called function i.e. intMinus1. Recall that we declared two
variables in the caller function. These variables were myInt and retVal. We also
assigned a value 31 to myInt. This value (31) is in the memory location of myInt in the
stack. When we declared the variable retVal, there was no value assigned to it. So its
memory location contains nothing yet. After it, there is the other stuff of the caller
function. We have also shown the memory addresses of the locations on the left hand
side. We can see that the memory address of myInt is 1072 while that of retVal is
1068. Similarly, we see the addresses 1060, 1056 and 1052 further down. We note
that the addresses are with a difference of four. This is because of the use of the
integer variable, as an integer requires four bytes in the memory for storing. Similarly
the pointer variables are also of four bytes. The memory addresses are in 32 bits that
is four bytes (eight bits make a byte). Generally we start addressing the memory from
down. The lowest byte is addressed 0, followed by 1, 2, 3 and so on to upward
direction. If we have larger variables, for example, objects created by our defined
classes, these will require more memory. In that case, the memory chunks will not be
of four bytes. Rather, it will be equal to the size of the object. In C++, we have an
operator size or sizeof by which the size of a type or variable, acquired in the memory,
can be found. There are some machines that do not use byte addressing. These
machines use word addressing. Every word is of 48 bits. All these things (memory
addressing etc) relate to computer architecture course.
Now in the stack, there are the entries of the called function. There is the parameter
oldVal of the called function in the stack. It has the value 31. This is the same value as
that of myInt in the caller function. As studied earlier, in call by value phenomenon,
when arguments (whatever type they have) are sent to a function, a copy of these
arguments is made before sending. That copy is sent to the called function. The called
function uses this copy of the argument. Thus a copy of myInt is passed to the called
function intMinus1. Thus its parameter oldVal has value 31. Now intMinus1 function
uses this value (i.e. oldVal) to do its functionality. It makes alternations to this copy.
The original value that is myInt does not change. This phenomenon is shown in the
following figure. Here the function intMinus1 decreases the value of oldVal by 1 and
becomes 30. The original value (myInt) remains unchanged.
1072
1068

31
?

myInt
retVal

calling function caller

callers other
stuff
1060

31 30

1056

oldVal
Called function

1052

sp

Page 200 of 519

stack grows downward

Fig 17.4: call stack layout after subtraction in intMinus1

CS301 Data Structures

___________________________________________________________________

When the return call of intMinus1 is executed, the control comes back to the calling
function. The stuff of the called function intMinus1 is popped from the stack and the
pointer sp moves up accordingly. In other words, the activation record of the function
intMinus1 has been removed from the stack. The oldval (value of which is 30 now) is
returned to the caller function. We have written in the code that this value is assigned
to retVal variable. So in the call stack layout, the memory location of retVal has the
value 30. The figure 17.5 shows the stack layout after return from the called function
i.e. intMinus1.
1072
1068

31
30

myInt
retVal

calling function caller

callers other
stuff
sp

stack grows downward

Fig 17.5 call stack layout after return from intMinus1

In the call by value phenomenon of function calling, we can pass a literal or even an
expression. Thus in the previous function caller, we could have called the function
intMinus1 as follows.
void caller()
{
int retVal;
retVal = intMinus1( 31 ); // a literal is passed
cout << retVal;
}
Here we did not put the value 31 in an integer variable. Rather, it was directly passed
as an argument in the function call i.e. intMinus1 (31). We are not passing a variable
as it was in the previous code of the function caller. The calling statement also be
written as under:
retVal = intMinus1( 30 + 1 );
In this statement, first the expression 30 + 1 is evaluated and then the copy of the
result goes to the call stack as a value of oldVal of called function.

Page 201 of 519

CS301 Data Structures

___________________________________________________________________
In the previous calling function (i.e. caller), it was witnessed that the value of the
passed variable (myInt) remained same when we passed the variable by value. Now
there may be situations where we want to actually change the value of a variable of
the caller function from within the called function. In such a situation, we send the
pointer to the variable (i.e. the address of the variable) as an argument. Look at the
following code of the function caller. In this code, we call the function intMinus2 by
passing it the address of the variable myInt.
void caller()
{
int retVal;
int myInt = 31;
retVal = intMinus2( &myInt );
cout << myInt << retVal;
}
Following is the function calling statement
retVal = intMinus2( &myInt );
The & sign before the name of the variable myInt means that the address of the
variable is being passed. The called function will take it as a pointer. Remember that
we have declared the argument of the function intMinus2 as int* oldVal that means
this function takes a pointer as an argument.
Now, keeping in mind that a pointer is being sent to the function, lets see what is
happening in the call stack. The stack portion which has the stuff of the calling
function i.e. caller is the same as seen in the call by value process. The variable myInt
has value 31 and value of retVal will be written after returning from the called
function. Its stack layout is shown in the figure 17.6 below.

1072
1068

31
?

myInt
retVal

calling function caller

callers other
stuff
1060

1072

1056

oldVal
Called function

1052

sp
stack grows downward

Fig 17.6 call stack layout when intMinus2 is called

Page 202 of 519

CS301 Data Structures

___________________________________________________________________
In the stuff of the called function i.e. intMinus2, the memory location of the oldVal
holds the value 1072. We can see on the left side of the stack in the figure above that
this 1072 is the address of the variable myInt. Look at the statement
*oldVal = *oldVal 2 ;
of intMinus2. Here *oldVal can be expressed as the value at the memory address that
is in oldVal. In the stack layout, we see that the memory address in oldVal is 1072.
The value stored at the address 1072 is 31. The following statement
*oldVal = *oldVal 2 ;
decreases this value 2 and thus brings the value at the memory address 1072 down to
29. As 1072 is the address of myInt, thus actually the value of myInt is changed. The
following figure of stack layout depicts this process.
1072
1068

31 29
?

myInt
retVal

calling function caller

callers other
stuff
1060

1072

oldVal

1056

Called function

1052

sp
stack grows downward

Fig 17.7: call stack layout after *oldVal = *oldVal 2;

Now when the execution of the called function ends after returning *oldVal, the
activation record of this function is removed from the stack. The pointer of the stack
comes up. The value 29 is put in the variable retVal of the calling function. Following
is the stack layout after return from the function intMinus2.

1072
1068

31 29
29

myInt
retVal

calling function caller

callers other
stuff
sp

stack grows downward

Page 203 of 519

Fig 17.8: call stack after return from intMinus2

CS301 Data Structures

___________________________________________________________________

Notice that the value of myInt of the caller function has also been changed.
We have seen that in call by value a copy of the argument is passed and used by the
called function. In this method, the original value of the argument in the calling
function remains unchanged. In the other method, we saw that when we pass the
address of the variable, the pointer is used to manipulate the variable data. In this
case, the original value of the variable in the calling function is changed. Suppose that
we want a function to change an object (variable/argument) but dont want to send the
copy of the object to the function. The reason of not sending a copy is that the object
may be large (as we have seen the object Customer in bank simulation example) and
making a copy of it costs time and memory space. Say, the object Customer is of 500
bytes. Now when a calling function will call a function using call by value method, a
copy of this object will be made on the call stack same as in our previous example a
copy of myInt was made. The copy constructor will make a copy of this whole object
that costs time. Moreover, the copy of this object will take 500 bytes on the stack that
is a lot of memory consuming. There will be as many copies as the number of calls. In
case of a large number of calls, there may be inefficient memory as call stack has a
limited memory. So we do not use call by value methodology. Furthermore, we want
to avoid the massive syntax of pointers. For this purpose, we do not use pointers in
function calling. Now the question arises is there any way through which we can
fulfill our requirement. (i.e. we dont want to make a copy and want to change the
objective without using the pointers ). The use of reference variables may be a
suitable answer to this very ticklish situation. The phenomenon of function calls using
reference variables is termed as call by reference. Following is the code of the caller
function that involves call by reference.

Void caller()
{
int retVal;
int myInt = 31;
retVal = intMinus3( myInt );
cout << myInt << retVal;
}
Note that this is almost the same function, we wrote at first to call intMinus1. The
only difference is that here we are calling the function intMinus3 instead of
intMinus1. We did not use & sign with myInt as used in the call for intMinus2. This is
only due to the fact that we are not sending an address. It is pertinent to note that in
the definition of the function intMinus3, we have used & sign with the argument
variable. We have written it as
int intMinus3( int& oldVal)

Page 204 of 519

CS301 Data Structures

___________________________________________________________________
This statement means that the function intMinus3 will take a reference as an
argument. In other words, a reference of the argument variable will be passed to this
function. Thus the idea of reference variable is that an object will be used exactly as it
exists in the caller function. The called function simply accesses it through a different
name i.e. other than the name in the caller function. Here in our example, the called
function intMinus3 accesses (and uses) the passed argument i.e. myInt with the name
oldVal. The function intMinus3 cannot use the name myInt as it is in the callers
scope. Now the both variable names (myInt and oldVal) refer to the same object. It
means that the same memory cell where the object lies. We have read the scope of
variables that the local variables can be used only by the function in which they are
declared. The other functions cannot use the local variables of a function. But in
reference variables, we can access the memory location of a local variable in another
function by using another name. Thus in our example, we access the local variable
myInt (actually the memory location) of the function caller in the other function
intMinus3 with the name oldVal. The following figure explains this phenomenon of
reference variables with the help of call stack layout.
1072

myInt

31
?

1068

retVal

calling function caller

callers other
stuff
oldVal

1060
1056

called function intMinus3

1052

sp
stack grows downward

Fig 17.9: call stack when intMinus3 is called

The caller function part of the stack contains myInt, retVal and other stuff of the caller
function. In the stack part intMinus3, the name oldVal shows that it has nothing in this
portion but it is the other name of the memory location 1072, named as myInt in the
caller function. The dotted arrow shows this.
In the following figure 17.10, we show the oldVal along with the myInt at same
memory location. Here we want to show that the oldVal of called function and myInt
of caller function are the two names of the same memory location.

OldVal

1072
1068

31
?

myInt
retVal

calling function caller

callers other
stuff
Page 205 of 519
1060
1056

called function intMinus3

CS301 Data Structures

___________________________________________________________________

Now when the body of the intMinus3 executes the statement i.e.
oldVal = oldVal 3 ;

CS301 Data Structures

___________________________________________________________________

This phenomenon of call by reference is actually implemented by the compiler by

using pointers. The obtaining of address and de-referencing will be done behind the
scene. We have no concern to do this. For a programmer, it is simply a renaming
abstraction, in which we rename the argument variable of the caller function and use
it in the called function.

Sample Program
Following is the program, which demonstrate the above three function calls that we
discussed in the previous example. We define the three functions intMinus1,
intMinus2 and intMinus3. These functions accept the argument as value, pointer and
reference variable respectively. The endl puts a new line and is used in the program
for the output clarity. Here is the code of the program followed by the output of the
program.
/*This program demonstrate tha how the value in a caller function is effected when it
is passed to a function by using call by value, by using pointers and by using call by
reference methods.
*/
#include <iostream.h>
//Function 1, call by value
int intMinus1( int oldVal)
{
oldVal = oldVal 1;
return oldVal;
}
// Function 2, call by using pointers
int intMinus2( int* oldVal)
{
*oldVal = *oldVal 2;
return *oldVal;
}
// Function 3, call by reference

Page 207 of 519

CS301 Data Structures

___________________________________________________________________
int intMinus3( int& oldVal)
{
oldVal = oldVal 3;
return oldVal;
}
void main ()
{
int myInt = 31;
int retVal;
retVal = intMinus1( myInt ); //call by value
cout << After returning from the called function intMinus1 << endl ;
cout << The value returned by the called function (retVal) is : << retVal ;
cout << endl ;
cout << The value of the calling functions variable (myInt) is : << myInt ;
cout << endl << endl;
// now pass the argument by using pointer, also initialize the value of myInt
myInt = 31 ;
retVal = intMinus2( &myInt ); //call by passing a pointer
cout << After returning from the called function intMinus2 << endl;
cout << The value returned by the called function (retVal) is : << retVal ;
cout << endl;
cout << The value of the calling functions variable (myInt) is : << myInt ;
cout << endl << endl;
// now pass the argument by as reference, also initialize the value of myInt
myInt = 31 ;
retVal = intMinus3( myInt ); //call by passing a reference
cout << After returning from the called function intMinus3 << endl;
cout << The value returned by the called function (retVal) is : << retVal ;
cout << endl;
cout << The value of the calling functions variable (myInt) is : << myInt ;
}

Following is the output of the program.

After returning from the called function intMinus1
The value returned by the called function (retVal) is : 30
The value of the calling function's variable (myInt) is : 31
After returning from the called function intMinus2
The value returned by the called function (retVal) is : 29

Page 208 of 519

CS301 Data Structures

___________________________________________________________________
The value of the calling function's variable (myInt) is : 29
After returning from the called function intMinus3
The value returned by the called function (retVal) is : 28
The value of the calling function's variable (myInt) is : 28
We can see from the output of the program how the passed variable of the caller
function is affected by these different ways of function calling. Note that the values of
the variables used are the same as we have discussed with the help of call stack
layout.

Data Structures
Lecture No. 18
Reading Material
Data Structures and Algorithm Analysis in C++
4.3.3

Chapter. 4

Summary

Reference Variables
const keyword
Tips

Reference Variables
In the last lecture we were discussing about reference variables, we saw three
examples; call by value, call by reference and call by pointer. We saw the use of stack
when a function is called by value, by reference or by pointer. The arguments passed
to the function and local variables are pushed on to the stack.
There is one important point to note that in this course, we are using C/C++ but the
usage of stack is similar in most of the computer languages like FORTRAN and Java .
The syntax we are using here is C++ specific, like we are sending a parameter by
pointer using & sign. In Java, the native data types like int, float are passed by value
and the objects are passed by reference. In FORTRAN, every parameter is passed by
Page 209 of 519

CS301 Data Structures

___________________________________________________________________
reference. In PASCAL, you can pass a parameter by value or by reference like C++.
You might have heard of ALGOL, this language had provided another way of passing
parameter called call by name. These kinds of topics are covered in subjects like
Study of Computer Languages or Compilers Theory.
It is recommended while you are doing your degree, you study other computer
languages and compare them from different aspects. Java is quite popular now a day,
quite similar in syntax to C++. May be as a next language, you can study that and
compare its different aspects with C/C++. The concepts like how a program is loaded
into memory to become a process, how the functions are called and the role of stack
etc are similar in all major languages.
we have discussed when the variables are passed by reference then behind the scene
what goes on inside the stack. There are few important things to take care of while
using reference variables:
One should be careful about transient objects that are stored by reference in data
structures.
We know that the local variables of a function are created on call stack. Those
variables are created inside the function, remains in memory until the control is inside
the function and destroyed when the function exits. Activation record comprise of
function call parameters, return address and local variables. The activation record
remains inside stack until the function is executing and it is destroyed once the control
is returned from the function.
Lets see the following code that stores and retrieves objects in a queue:
void loadCustomer( Queue & q)
{
Customer c1(irfan);
Customer c2(sohail);
q.enqueue( c1 );
q.enqueue( c2 );
}
Above given is a small function loadCustomer( Queue &), which accepts a parameter
of type Queue by reference. Inside the function body, firstly, we are creating c1 and
c2 Customer objects. c1 and c2 both are initialized to string values irfan and sohail
respectively. Then we queue up these objects c1 and c2 in the queue q using the
enqueue() method and finally the function returns.
Now, the objects created inside the above function are c1 and c2. As local variables
are created on stack, therefore, objects are also created on stack, no matter how big is
the size of the data members of the object. In the Bank example, in previous lecture,
we saw that for each customer we have the name (32 characters maximum), arrival
time ( int type, 4 bytes), transaction time ( int type)and departure time (int type) of
the customer. So the size of the Customer object is 44 bytes. Our c1 and c2 objects are
created on stack and have 44 bytes occupied. It is important to mention here that we
are referring each 44 bytes of allocation with the name of the object. The allocated 44
bytes are bound with the name of the object c1 or c2. Another significant point here is
that the function enqueue() accepts the object Customer by reference. See the code
below of serviceCustomer() method, which is executed after the loadCustomer().
void serviceCustomer( Queue & q)

Page 210 of 519

CS301 Data Structures

___________________________________________________________________
{
Customer c = q.dequeue();
cout << c.getName() << endl;
}
The serviceCustomer(Queue &) also accepts one parameter of type Queue by
reference. In the first statement, it is taking out one element from the queue and
assigning to newly created object c. Before assignment of address of c1 object (c1
because it was inserted first), the object c is constructed by calling the default
(parameter less) constructor. In the next statement, c.getName() function call is to get
the name of the customer and then to print it. What do you think about it? Will this
name be printed or not? Do you see any problem in its execution? In short, this
statement will not work.
To see the problem in this statement, we have to understand the mechanism; where
the object was created, what was pushed on the stack, when the function
loadCustomer() returned and what had happened to the objects pushed on to the stack.
The objects c1 and c2, which were created locally in loadCustomer() function,
therefore, they were created on stack. After creating the objects, we had added their
addresses, not the objects themselves in the queue q. When the function
loadCustomer() returned, the local objects c1 and c2 were destroyed but their
addresses were there in the queue q. After some time, the serviceCustomer() is called.
The address of the object is retrieved from the queue and assigned to another newly
created local object c but when we wanted to call a method getName() of c1 object
using its retrieved address, we encountered the problem.
This shows that this is true that use of reference alleviate the burden of copying of
object but storing of references of transient objects can create problems because the
transient object (object created on stack) is destroyed when the function execution
finishes.
The question arises, what can we do, if we do not want the objects created in a
function to be destroyed. The answer to this is dynamic memory allocation. All the
variables or objects created in a function that we want to access later are created on
memory heap (sometimes called free store) using the dynamic memory allocation
functions or operators like new. Heap is an area in computer memory that is allocated
dynamically. You should remember that all the objects created using new operator
have to be explicitly destroyed using the delete operator.
Lets see the modified code of loadCustomer() function, where the objects created in
a function are not transient, means they are created on heap to be used later in the
program outside the body of the function loadCustomer().
void loadCustomer( Queue & q)
{
Customer * c1 = new Customer(irfan);
Customer * c2 = new Customer(sohail);
q.enqueue( c1 ); // enqueue takes pointers
q.enqueue( c2 );
}

Page 211 of 519

CS301 Data Structures

___________________________________________________________________
This time, we are creating the same two objects using the new operator and assigning
the starting addresses of those objects to c1 and c2 pointers. Nameless objects (objects
accessed by pointers) are called anonymous objects.
Here c1 and c2 are pointers to the objects not the actual objects themselves, as it was
previously. These starting addresses c1 and c2 of the objects are then queued using
the enqueue() method. As the objects lie on the heap, so there will not be any problem
and the objects will be accessible after the function loadCustomer() returns.
There is a bit tricky point to understand here. Although, the objects are created on
heap but the pointer variables c1 and c2 are created on stack and they will be
definitely destroyed after the loadCustomer() activation record is destroyed.
Importantly, you should understand the difference between the pointer variables and
the actual objects created on heap. The pointer variables c1 and c2 were just used to
store the starting addresses of the objects inside the function loadCustomer(), once the
function is returned the pointer variables will not be there. But as the starting
addresses of the objects are put in the queue, they will be available to use later after
retrieving them from queue using the dequeue() operation. These dynamic objects will
live in memory (one heap) unless explicitly deleted.
By the way, there is another heap, heap data structure that we are going to cover later
in this course.
At the moment, see the layout of computer memory and heap as we previously saw in
this course. Heap is an area in memory given to a process from operating system
when the process does dynamic memory allocation.

Process 1
(Browser)

Code

Process 3
(Word)

Static Data

Process 4
(Excel)

Stack

Process 2
(Dev-C++)
Windows OS

Heap

Fig 18.1: Memory Organization

One the left of the picture, we can see different processes in the computer memory.
When we zoomed into the one of the processes, we saw the picture on the right. That
firstly, there is a section for code, then for static data and for stack. Stack grows in the
downward section. You can see the heap section given at the end, which grows
upward. An interesting question arises here is that why the stack grows downward
and heap in the upward direction. Think about an endless recursive call of a function
to itself. For every invocation, there will be an activation record on stack. So the
stack keeps on growing and growing even it overwrites the heap section. One the

Page 212 of 519

CS301 Data Structures

___________________________________________________________________
other hand, if your program is performing dynamic memory allocation endlessly, the
heap grows in the upward direction such that it overwrites the stack section and
destroys it.
You might have already understood the idea that if a process has some destructive
code then it will not harm any other process, only its own destruction is caused. By
the way, lot of viruses exploit the stack overflow to change the memory contents and
cause further destruction to the system.
Consider that we allocate an array of 100 elements of Customer objects dynamically.
As each object is 44 bytes, therefore, the size of memory allocated on heap will be
4400 bytes (44 * 100). To explain the allocation mechanism on stack and heap, lets
see the figure below where the objects are created dynamically.

heap grows upwards

c

688

sohail

Customer(sohail) -> c2

644

irfan

Customer(irfan) -> c1

600

Fig 18.2: Heap layout during call to loadCustomer

Page 213 of 519

CS301 Data Structures

___________________________________________________________________
The objects are shown in this figure by using the names of the customers inside them.
Actually, there are three more int type variables inside each object. You can see that
the object with string irfan is from memory address 600 to 643 and object with name
customer name as sohail is from address 644 to 687. Now when these objects are
inserted in the queue, only their starting addresses are inserted as shown in the below
figure.

1072
1068

600
644

loadCustome
c

.
.
.
.
1060

(644

(elt

1056
1052

enqueue

sp
stack grows downwards
Fig 18.3: Stack layout when q.enqueue(2) called from loadCustomer
The loadCustomer() is being executed. It is containing two pointers c1 and c2
containing the addresses 600 and 643 respectively. enqueue(elt) method is called and
the parameter values (which actually are addresses) 600 and 643 are inserted in the
queue.

Page 214 of 519

CS301 Data Structures

___________________________________________________________________
Because the objects have been allocated on heap, therefore, there will no issue with
them. The pointer variables c1 and c2, which we used to store addresses, are
destroyed. But the queue q, which is passed by reference to loadCustomer will be
there and it is containing the starting addresses of the Customer objects. Those are
valid addresses of valid objects, so they can used in the program later to access the
customer objects. See the function below:
void serviceCustomer( Queue & q)
{
Customer* c = q.dequeue();
cout << c->getName() << endl;
delete c; // the object in heap dies
}
You can see that we are taking one pointer out of the queue and in the second line
calling the method of the Customer object getName() with c->. We are using ->
operator because we are taking out pointer from the queue.
Now, we should be sure that this method will be executed successfully because the
object was created dynamically inside the loadCustomer() method. The last statement
inside the method is delete, which has been used to deallocate the object.
So now, we understand that we cannot pass references to transient objects. If we want
to use the objects later we create them on heap and keep the address. There is another
point to mention here that in case, the object has already been deallocated and we are
accessing it (calling any of its member), it may the cause the program to crash. The
pointer of the object (when object has already been deallocated or released) is called
dangling pointer.

The const Keyword

The const keyword is used for something to be constant. The actual meanings depend
on where it occurs but it generally means something is to held constant. There can be
constant functions, constant variables or parameters etc.
The references are pointers internally, actually they are constant pointers. You cannot
perform any kind of arithmetic manipulation with references that you normally do
with pointers. You must be remembering when we wrote header file for binary tree
class, we had used const keyword many times. The const keyword is often used in
function signatures. The function signature is also called the function prototype where
we mention the function name, its parameters and return type etc.
Here are some common uses of const keyword.
1. The const keyword appears before a function parameter. E.g., in a chess program:
int movePiece(const

Piece & currentPiece)

The function movePiece() above is passed one parameter, which is passed by

reference. By writing const, we are saying that parameter must remain constant for the
life of the function. If we try to change value, for example, the parameter appears on
the left side of the assignment, the compiler will generate an error. This also means

Page 215 of 519

CS301 Data Structures

___________________________________________________________________
that if the parameter is passed to another function, that function must not change it
either.
Use of const with reference parameters is very common. This is puzzling; why are we
passing something by reference and then make it constant, i.e., dont change it?
Doesnt passing by reference mean we want to change it?
Think about it, consult your C++ book and from the internet. We will discuss about
the answer in the next lecture.

Page 216 of 519

CS301 Data Structures

___________________________________________________________________

Tips
The arithmetic operations we perform on pointers, cannot be performed on
references
Reference variables must be declared and initialized in one statement.
To avoid dangling reference, dont return the reference of a local variable
(transient) from a function.
In functions that return reference, return global, static or dynamically allocated
variables.
The reference data types are used as ordinary variables without any dereference
operator. We normally use arrow operator (->) with pointers.
const objects cannot be assigned any other value.
If an object is declared as const in a function then any further functions called
from this function cannot change the value of the const object.

Page 217 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 19
Reading Material
Data Structures and Algorithm Analysis in C++
4.4

Chapter. 4

Summary

Usage of const keyword

Degenerate Binary Search Tree
AVL tree

Usage of const keyword

In the previous lecture, we dealt with a puzzle of constant keyword. We send a
parameter to a function by using call by reference and put const with it. With the help
of the reference variable, a function can change the value of the variable. But at the
same time, we have used the const keyword so that it does not effect this change.
With the reference parameter, we need not to make the copy of the object to send it to
the calling function. In case of call by value, a copy of object is made and placed at
the time of function calling in the activation record. Here the copy constructor is used
to make a copy of the object. If we dont want the function to change the parameter
without going for the use of time, memory creating and storing an entire copy of, it is
advisable to use the reference parameter as const. By using the references, we are not
making the copy. Moreover, with the const keyword, the function cannot change the
object. The calling function has read only access to this object. It can use this object
in the computation but can not change it. As we have marked it as constant, the
function cannot alter it, even by mistake. The language is supportive in averting the
mistakes.
There is another use of keyword const. The const keyword appears at the end of class
members function signature as:
EType& findMin( ) const;
This method is used to find the minimum data value in the binary tree. As you have
noted in the method signature, we had written const at the end. Such a function cannot
change or write to member variables of that class. Member variables are those which
appear in the public or private part of the class. For example in the BinaryTree, we
have root as a member variable. Also the item variable in the node class is the
member variable. These are also called state variables of the class. When we create an
object from the factory, it has these member variables and the methods of this class
which manipulate the member variables. You will also use set and get methods,

Page 218 of 519

CS301 Data Structures

___________________________________________________________________
generally employed to set and get the values of the member variables. The member
function can access and change the public and private member variables of a class.
Suppose, we want that a member function can access the member variable but cannot
change it. It means that we want to make the variables read only for that member
function. To impose that constraint on the member function, a programmer can put
the keyword const in the end of the function. This is the way in the C++ language. In
other languages, there may be alternative methods to carry out it. These features are
also available in other object oriented languages. This type of usage often appears in
functions that are supposed to read and return member variables. In the Customer
example, we have used a method getName that returns the name of the customer. This
member function just returns the value of member variable name which is a private
data member. This function does not need to change the value of the variable. Now
we have written a class and its functions. Why we are imposing such restrictions on
it? This is the question of discipline. As a programmer when we write programs,
sometimes there are unintentional mistakes. On viewing the code, it seems
unbelievable that we have written like this. If these codes contain mistakes, the user
will get errors. At that time, it was thought that we have imposed restrictions on the
function and can avoid such mistakes at compile time or runtime. The discipline in
programming is a must practice in the software engineering. We should not think that
our programs are error-free. Therefore, the programming languages help in averting
the common errors. One of the examples of such support is the use of const keyword.
There is another use of const. The const keyword appears at the beginning of the
return type in function signature:
const EType& findMin( ) const;
The return type of the findMin() function is ETyper& that means a reference is
returned. At the start of the return type, we have const keyword. How is this
implemented internally? There are two ways to achieve this. Firstly, the function puts
the value in a register that is taken by the caller. Secondly, the function puts the value
in the stack that is a part of activation record and the caller functions gets the value at
that point from the stack and use it. In the above example, we have return value as a
reference as EType&. Can a function return a reference of its local variable? When the
function ends, the local variables are destroyed. So returning the reference of local
variable is a programming mistake. Therefore, a function returns the reference of
some member variable of the class. By not writing the & with the return type, we are
actually returning the value of the variable. In this case, a copy of the returning
variable is made and returned. The copy constructor is also used here to create the
copy of the object. When we are returning by value, a copy is created to ascertain
whether it is a local variable or member variable. To avoid this, we use return by
reference. Now we want that the variable being returned, does not get changed by the
calling function especially if it is the member variable.
When we create an object from the factory, the member variable has some values. We
do not want that the user of this object has direct access to these member variables. So
get and set methods are used to obtain and change the value of these member
variables. This is a programming practice that the values of the object should be
changed while using these methods. This way, we have a clean interface. These

Page 219 of 519

CS301 Data Structures

___________________________________________________________________
methods are in a way sending messages to the object like give me the name of the
customer or change the name of the customer. The presence of a queue object can
help us send a message to it that gets an object and returns it. In these function-calling
mechanisms, there are chances that we start copying the objects that is a time
consuming process. If you want that the function returns the reference of the member
variable without changing the value of the member variable using this reference, a
construct is put at the start of the function. It makes the reference as a const reference.
Now the value of this member variable cannot be changed while using this reference.
The compiler will give error or at the runtime, you will get the error. When we return
an object from some function, a copy is created and returned. If the object is very big,
it will take time. To avoid this, we return this through the reference. At this point, a
programmer has to be very careful. If you do not use the const with the reference,
your object is not safe and the caller can change the values in it.
These are the common usage of const. It is mostly used with the member function. It
is just due to the fact that we avoid creating copy of the object and secondly we get
our programming disciplined. When we send a reference to some function or get a
reference from some function, in both cases while using the const, we guard our
objects. Now these objects cannot be changed. If the user of these objects needs to
change the object, he should use the set methods of the object.
We have used such methods in the BinarySearchTree.h file. However, the
implementation of this class has not been discussed so far. We advise you to try to
write its code yourself and experiment with it.

Degenerate Binary Search Tree

Consider the tree as shown below:
BST for 14, 15, 4, 9, 7, 18, 3, 5, 16, 20, 17
14

15

18

16

20

17

Page 220 of 519

CS301 Data Structures

___________________________________________________________________
The above tree contains nodes with values as 14, 15, 4, 9, 7, 18, 3, 5, 16, 20, 17
respectively. The root node is 14. The right subtree contains the numbers greater than
14 and the left subtree contains the numbers smaller than 14. This is the property of
the binary search tree that at any node, the left subtree contains the numbers smaller
than this node and the right subtree contains the numbers greater than this node.
Now suppose that we are given data as 3, 4, 5, 7, 9, 14, 15, 16, 17, 18, 20 to create a
tree containing these numbers. Now if our insert method takes the data in the order as
given above, what will be the shape of our tree? Try to draw a sketch of the tree with
some initial numbers in your mind. The tree will look like:
BST for 3, 4, 5, 7, 9, 14, 15, 16, 17, 18, 20
3
4
5
7
9
14
15
16
17
18
20

It does not seem to be a binary tree. Rather, it gives a look of a linked list, as there is a
link from 3 to 4, a link from 4 to 5, a link from 5 to 7. Similarly while traversing the
right link of the nodes, we reached at the node 20. There is no left child of any node.
Thats why, it looks like a link list. What is the characteristic of the link list? In link
list, every node has a pointer that points to the next node. While following this
pointer, we can go to the next node. The root of this tree is 3. Now we have to find the
node with value 20 in this tree. Remember that it is a tree, not a link list. We will use
find method to search the number 20 in this tree. Now we will start from the root. As
20 is greater than 3, the recursive call to the method find will be made and we come to
the next node i.e. 4. As 20 is greater than 4, so again a recursive call is generated.
Similarly we will come to 5, then 7, 9 and so on. In the end, we will reach at 20 and
the recursion will stop here.

Page 221 of 519

CS301 Data Structures

___________________________________________________________________
Now if we search the above tree through the method in which we started from 3, then
4, 5 and so on, this will be the same technique as adopted in the link list. How much
time it will take to find the number? We have seen in the link list that if the number to
be searched is at the last node, a programmer will have to traverse all the nodes. This
means that in case of nodes having strength of n, the loop will execute n times.
Similarly as shown in the above tree, our find method will be called recursively equal
to number of nodes in the tree. We have designed binary search tree in such a fashion
that the search process is very short. You must be remembering the example of
previous lecture that if we have one lakh numbers, it is possible to find the desired
number in 20 iterations. If we have link list for one lakh elements, the required results
can be obtained only after executing the loop for one lakh times if the element to be
searched is the last element. However, in case of BST, there are only 20 steps. The
BST technique, as witnessed earlier, is quite different as compared to this tree. They
have both left and right subtrees. What happened with this tree? The benefit we have
due to BST is not applicable here. It seems that it is a link list. This is only due to the
fact that the data of the tree was given in the sorted order.
If you want to create a tree out of a sorted data with the insert method, it will look like
the above tree. It means that you do not want to have sorted data. But it is not easy, as
you might not have control over this process. Consider the example of polling. It is
not possible that all the voters come to the polling station in some specific order. But
in another example, if you are given a list of sorted data and asked to create a BST
with this data. If you create a BST with data that is in an ascending order, it will look
like a link list. In the link list, the search takes a lot of time. You have created a BST
but the operations on it are working as it is a singly link list. How can we avoid that?
We know that the BST is very beneficial. One way to avoid this is that some how we
get the sorted data unsorted. How this can be done. It is not possible, as data is not
always provided as a complete set. Data is provided in chunks most of the times. Now
what should we do? We will apply a technique here so that we can get the benefits of
the BST. We should keep the tree balanced. In the above tree, nodes have left child
and no right child. So this tree is not balanced. One way to achieve it is that both the
left and right subtrees have the same height. While talking about the binary search
tree, we discussed the height, depth and level of BST. Every node has some level. As
we go down to the tree from the root, the levels of the tree increased and also the
number of nodes, if all the left and right subtrees are present. You have earlier seen
different examples of tree. The complete binary tree is such a tree that has all the left
and right subtrees and all the leaf nodes in the end. In the complete binary tree, we
can say that the number of nodes in the left subtree and right subtree are equal. If we
weigh that tree on the balance, from the root, both of its sides will be equal as the
number of nodes in the right subtree and left subtree are equal. If you have such a
balanced binary search tree with one lakh nodes, there will need of only 20
comparisons to find a number. The levels of this tree are 20. We have also used the
formula log2 (100,000). The property of such a tree is that the search comparison can
be computed with the help of log because subtrees are switched at every comparison.
Now lets see the above tree which is like a singly link list. We will try to convert it
into a balanced tree. Have a look on the following figure.

Page 222 of 519

CS301 Data Structures

___________________________________________________________________
14
9
7
5

15
16
17

18
20

This tree seems to be a balanced tree. We have made 14 as the root. The nodes at the
left side occur at the left of all the nodes i.e. left subtree of 14 is 9, the left subtree of 9
is 7, the left subtree of 7 is 5 and so on. Similarly the right subtree contains the nodes
15, 16, 17, 18, 20. This tree seems to be a balanced tree. Lets see its level. The node
14 i.e. the root is at level zero. Then at level one, we have 9 and 15. At level two,
there are 7 and 16. Then 5 and 17, followed by 4 and 18. In the end, we have 3 and
20. It seems that we have twisted the tree in the middle, taking 14 as a root node. If
we take other nodes like 9 or 7, these have only left subtree. Similarly if we take 15 or
16, these have right subtrees only. These nodes do not have both right and left
subtree. In the earlier example, we have seen that the nodes have right and left
subtrees. In that example, the data was not sorted. Here the tree is not shallow. Still
we can not get the required BST. What should we do? With the sorted data, the tree
can not become complete binary search tree and the search is not optimized. We want
the data in unsorted form that may not be available.
We want to make a balanced tree, keeping in mind that it should not be shallow one.
We could insist that every node must have left and right subtrees of same height. But
this requires that the tree be a complete binary tree. To achieve it, there must be (2d+1
1) data items, where d is the depth of the tree. Here we are not pleading to have
unsorted data. Rather, we need as much data which could help make a balanced
binary tree. If we have a tree of depth d, there will be need of (2d+1 1) data items i.e.
we will have left and right subtrees of every node with the same height. Now think
yourself that is it possible that whenever you build a tree or someone uses your BST
class can fulfill this condition. This is not possible that whenever we are going to
create a tree, there will be (2d+1 1) data items for a tree of depth d. The reason is
that most of the time you do not have control over the data. Therefore this is too rigid
condition. So this is also not a practical solution.

AVL Tree
AVL tree has been named after two persons Adelson-Velskii and Landis. These two
had devised a technique to make the tree balanced. According to them, an AVL tree is
identical to a BST, barring the following possible differences:

Height of the left and right subtrees may differ by at most 1.

Height of an empty tree is defined to be (1).

Page 223 of 519

CS301 Data Structures

___________________________________________________________________
We can calculate the height of a subtree by counting its levels from the bottom. At
some node, we calculate the height of its left subtree and right subtree and get the
difference between them. Lets understand this with the help of following fig.

An AVL Tree

Level
0

2
3

This is an AVL tree. The root of the tree is 5. At next level, we have 2 and 8, followed
by 1, 4 and 7 at next level where 1, 4 are left and right subtrees of node 2 and 7 is the
left subtree of node 8. At the level three, we have 3. We have shown the levels in the
figure at the right side. The root is at level 0, followed by the levels 1, 2 and 3. Now
see the height of the left subtree of 5. It is 3. Similarly the height of the right subtree
is 2. Now we have to calculate the difference of the height of left subtree and right
subtree of 5. The height of left subtree of 5 is 3 and height of right subtree of 5 is 2.
So the difference is 1. Similarly, we can have a tree in which right subtree is deeper
than left subtree. The condition in the AVL tree is that at any node the height of left
subtree can be one more or one less than the height of right subtree. These heights, of
course, can be equal. The difference of heights can not be more than 1. This
difference can be -1 if we subtract the height of left subtree from right subtree where
the height of left subtree is one less than the height of right subtree. Remember that
this condition is not at the root. It should satisfy at any level at any node. Lets
analyze the height of left subtree and right subtree of node 2. This should be -1, 0 or
1. The height of left subtree of node 2 is 1 while that of right subtree of the node 2 is
2. Therefore the absolute difference between them is 1. Similarly at node 8, the height
of left subtree is 1 and right subtree does not exist so its height is zero. Therefore the
difference is 1. At leaves, the height is zero, as there is no left or right subtree. In the
above figure, the balanced condition is satisfactory at every level and node. Such trees
have a special structure.
Lets see another example. Here is the diagram of the tree.

Page 224 of 519

CS301 Data Structures

___________________________________________________________________

Not an AVL

Level
0

4
3

2
5

The height of the left subtree of node 6 is three whereas the height of the right subtree
is one. Therefore the difference is 2. The balanced condition is not satisfactory.
Therefore, it is not an AVL tree.
Lets give this condition a formal shape that will become a guiding principle for us
while creating a tree. We will try to satisfy this condition during the insertion of a
node in the tree or a deletion of a node from the tree. We will also see later how we
can enforce this condition satisfactorily on our tree. As a result, we will get a tree
whose structure will not be like a singly linked list.
The definition of height of a tree is:

The height of a binary tree is the maximum level of its leaves (also called the
depth).

The height of a tree is the longest path from the root to the leaf. This can also be
calculated as the maximum level of the tree. If we have to calculate the height of
some node, we should start counting the levels from that node.
The balance of a node is defined as:

The balance of a node in a binary tree is defined as the height of its left subtree
minus height of its right subtree.

Here, for example, is a balanced tree whose each node has an indicated balance of 1,
0, or 1.

Page 225 of 519

CS301 Data Structures

___________________________________________________________________
-1
1
0
0

0
-1

0
0

0
0

0
0

0
0

In this example, we have shown the balance of each node instead of the data item. In
the root node, there is the value -1. With this information, you know that the height of
the right subtree at this node is one greater than that of the left subtree. In the left
subtree of the root, we have node with value 1. You can understand from this example
that the height of the right subtree at this node is one less than the height of the left
subtree. In this tree, some nodes have balance -1, 0 or 1. You have been thinking that
we have to calculate the balance of each node. How can we do that? When we create
a tree, there will be a need of some information on the balance factor of each node.
With the help of this information, we will try to balance the tree. So after getting this
balance factor for each node, we will be able to create a balance tree even with the
sorted data. There are other cases, which we will discuss, in the next lecture. In short,
a balance tree with any kind of data facilitates the search process.

Page 226 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 20
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.4

Summary

AVL Tree
Insertion in AVL Tree
Example (AVL Tree Building)

We will continue the discussion on AVL tree in this lecture. Before going ahead, it
will be better to recap things talked about in the previous lecture. We built a balanced
search tree (BST) with sorted data. The numbers put in that tree were in increasing
sorted order. The tree built in this way was like a linked list. It was witnessed that the
use of the tree data structure can help make the process of searches faster. We have
seen that in linked list or array, the searches are very time consuming. A loop is
executed from start of the list up to the end. Due to this fact, we started using tree data
structure. It was evident that in case, both the left and right sub-trees of a tree are
almost equal, a tree of n nodes will have log2 n levels. If we want to search an item in
this tree, the required result can be achieved, whether the item is found or not, at the
maximum in the log n comparisons. Suppose we have 100,000 items (number or
names) and have built a balanced search tree of these items. In 20 (i.e. log 100000)
comparisons, it will be possible to tell whether an item is there or not in these 100,000
items.

AVL Tree
In the year 1962, two Russian scientists, Adelson-Velskii and Landis, proposed the
criteria to save the binary search tree (BST) from its degenerate form. This was an
effort to propose the development of a balanced search tree by considering the height
as a standard. This tree is known as AVL tree. The name AVL is an acronym of the
names of these two scientists.
An AVL tree is identical to a BST, barring one difference i.e. the height of the left
and right sub-trees can differ by at most 1. Moreover, the height of an empty tree is
defined to be (1).
Keeping in mind the idea of the level of a tree, we can understand that if the root of a
tree is at level zero, its two children (subtrees) i.e. nodes will be at level 1. At level 2,
there will be 4 nodes in case of a complete binary tree. Similarly at level 3, the
number of nodes will be 8 and so on. As discussed earlier, in a complete binary tree,
the number of nodes at any level k will be 2k. We have also seen the level order
traversal of a tree. The term height is identical to the level of a tree. Following is the
figure of a tree in which level/height of nodes is shown.
Page 227 of 519

CS301 Data Structures

___________________________________________________________________
level
5
2
1

4
3

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

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

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

Fig 20.1: levels of nodes in a tree

Here in the figure, the root node i.e. 5 is at the height zero. The next two nodes 2 and
8 are at height (or level) 1. Then the nodes 1, 4 and 7 are at height 2 i.e. two levels
below the root. At the last, the single node 3 is at level (height) 3. Looking at the
figure, we can say that the maximum height of the tree is 3. AVL states that a tree
should be formed in such a form that the difference of the heights (maximum no of
levels i.e. depth) of left and right sub-trees of a node should not be greater than 1. The
difference between the height of left subtree and height of right subtree is called the
balance of the node. In an AVL tree, the balance (also called balance factor) of a node
will be 1,0 or 1 depending on whether the height of its left subtree is greater than,
equal to or less than the height of its right subtree.
Now consider the tree in the figure 20.1. Its root node is 5. Now go to its left subtree
and find the deepest node in this subtree. We see that node 3 is at the deepest level.
The level of this deepest node is 3, which means the height of this left subtree is 3.
Now from node 5, go to its right subtree and find the deepest level of a node. The
node 7 is the deepest node in this right subtree and its level is 2. This means that the
height of right subtree is 2. Thus the difference of height of left subtree (i.e. 3) and
height of right subtree (i.e. 2) is 1. So according to the AVL definition, this tree is
balanced one. But we know that the AVL definition does not apply only to the root
node of the tree. Every node (non-leaf or leaf) should fulfill this definition. This
means that the balance of every node should be 1, 0 or 1. Otherwise, it will not be an
AVL tree.
Now consider the node 2 and apply the definition on it. Lets see the result. The left
subtree of node 2 has the node 1 at deepest level i.e. level 2. The node 2, itself, is at
level 1, so the height of the left subtree of node 2 is 2-1 i.e. 1. Now look at the right
subtree of node 2. The deepest level of this right subtree is 3 where the node 3 exists.
The height of this right subtree of node 2 will be 3 1 = 2 as the level of node 2 is 1.
Now the difference of the height of left subtree (i.e. 1) and height of the right subtree
(i.e. 2) is 1. We subtract the height of left subtree from the height of the right subtree
and see that node 2 also fulfills the AVL definition. Similarly we can see that all other
nodes of the tree (figure 20.1) fulfill the AVL definition. This means that the balance
of each node is 1, 0 or 1. Thus it is an AVL tree, also called the balanced tree. The
following figure shows the tree with the balance of each node.

Page 228 of 519

CS301 Data Structures

___________________________________________________________________

level

-1
0

1
0

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

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

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

Fig 20.2: balance of nodes in an AVL

Lets consider a tree where the condition of an AVL tree is not being fulfilled. The
following figure shows such a tree in which the balance of a node (that is root node 6)
is greater than 1. In this case, we see that the left subtree of node 6 has height 3 as its
deepest nodes 3 and 5 are at level 3. Whereas the height of its right subtree is 1 as the
deepest node of right subtree is 8 i.e. level 1. Thus the difference of heights (i.e.
balance) is 2. But according to AVL definition, the balance should be1, 0 or 1. As
shown in the figure, this node 6 is only the node that violates the AVL definition (as
its balance is other than 1, 0 and -1). The other nodes fulfill the AVL definition. We
know that to be an AVL tree, each node of the tree should fulfill the definition. Here
in this tree, the node 6 violates this definition so this is not an AVL tree.
level

-1
0

0
0

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

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

4
0

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

Fig 20.3: not an AVL tree

Page 229 of 519

CS301 Data Structures

___________________________________________________________________
From the above discussion, we encounter two terms i.e. height and balance which can
be defined as under.

Height
The height of a binary tree is the maximum level of its leaves. This is the same
definition as of depth of a tree.
Balance
The balance of a node in a binary search tree is defined as the height of its left subtree
minus height of its right subtree. In other words, at a particular node, the difference in
heights of its left and right subtree gives the balance of the node.
The following figure shows a balanced tree. In this figure the balance of each node is
shown along with. We can see that each node has a balance 1, 0 or 1.
6

-1
1
0

4
0

1 10

0 3

12

-1 14

0 11

0 13

15

16

17

Fig 20.4: A balanced binary tree

Here in the figure, we see that the balance of the root (i.e. node 6) is 1. We can find
out this balance. The deepest level of the left subtree is 3 where the nodes 1 and 3 are
located. Thus the height of left subtree is 3. In the right subtree, we see some leaf
nodes at level 3 while some are found at level 4. But we know that the height of the
tree is the maximum level. So 4 is the height of the right subtree. Now we know that
the balance of the root node will be the result of height of left subtree minus the
height of right subtree. Thus the balance of the root node is 3 4 = -1. Similarly we
can confirm the balance of other nodes. The confirmation of balance of the other
nodes of the tree can be done. You should do it as an exercise. The process of height
computation should be understood as it is used for the insertion and deletion of nodes
in an AVL tree. We may come across a situation, when the tree does not remain
balanced due to insertion or deletion. For making it a balanced one, we have to carry
out the height computations.

Page 230 of 519

CS301 Data Structures

___________________________________________________________________
While dealing with AVL trees, we have to keep the information of balance factor of
the nodes along with the data of nodes. Similarly, a programmer has to have
additional information (i.e. balance) of the nodes while writing code for AVL tree.

Insertion of Node in an AVL Tree

Now lets see the process of insertion in an AVL tree. We have to take care that the
tree should remain AVL tree after the insertion of new node(s) in it. We will now see
how an AVL tree is affected by the insertion of nodes.
We have discussed the process of inserting a new node in a binary search tree in
previous lectures. To insert a node in a BST, we compare its data with the root node.
If the new data item is less than the root node item in a particular order, this data item
will hold its place in the left subtree of the root. Now we compare the new data item
with the root of this left subtree and decide its place. Thus at last, the new data item
becomes a leaf node at a proper place. After inserting the new data item, if we
traverse the tree with the inorder traversal, then that data item will become at its
appropriate position in the data items. To further understand the insertion process,
lets consider the tree of figure 20.4. The following figure (Fig 20.5) shows the same
tree with the difference that each node shows the balance along with the data item.
We know that a new node will be inserted as a leaf node. This will be inserted where
the facility of adding a node is available. In the figure, we have indicated the positions
where a new node can be added. We have used two labels B and U for different
positions where a node can be added. The label B indicates that if we add a node at
this position, the tree will remain balanced tree. On the other hand, the addition of a
node at the position labeled as U1, U2 .U12, the tree will become unbalanced. That
means that at some node the difference of heights of left and right subtree will
become greater than 1.
-1
1
0

0
U1

U2

4
0

2
0

U3

1 10

5
B

-1 14

0 11

U4

0
U5

12

0
U6

U7

0 13
B

0
U8

16

U9

15

0
U10

17

U11

U12

Fig 20.5: Insertions and effect in a balanced tree

Page 231 of 519

CS301 Data Structures

___________________________________________________________________

By looking at the labels B, U1, U2 .U12, we conclude some conditions that will
be implemented while writing the code for insert method of a balanced tree.
We may conclude that the tree becomes unbalanced only if the newly inserted node

Is a left descendent of a node that previously had a balance of 1

(in the figure 20.5 these positions are U1, U2 ..U8)
Or is a descendent of a node that previously had a balance of 1
(in the tree in fig 20.5 these positions are U9, U10, U11 and U12)

The above conditions are obvious. The balance 1 of a node indicates that the height of
its left subtree is 1 more than the height of its right subtree. Now if we add a node to
this left subtree, it will increase the level of the tree by 1. Thus the difference of
heights will become 2. It violates the AVL rule, making the tree unbalanced.
Similarly the balance 1 of a node indicates that the right subtree of this node is one
level deep than the left subtree of the node. Now if the new node is added in the right
subtree, this right subtree will become deeper. Its depth/height will increase as a new
node is added at a new level that will increase the level of the tree and the height.
Thus the balance of the node, that previously has a balance 1, will become 2.
The following figure (Fig 20.6) depicts this rule. In this figure, we have associated the
new positions with their grand parent. The figure shows that U1, U2, U3 and U4 are
the left descendents of the node that has a balance 1. So according to the condition,
the insertion of new node at these positions will unbalance the tree. Similarly the
positions U5, U6, U7 and U8 are the left descendents of the node that has a balance 1.
Moreover we see that the positions U9, U10, U11 and U12 are the right descendents
of the node that has balance 1. So according to the second condition as stated earlier,
the insertion of a new node at these positions would unbalance the tree.
-1
1
0

0
U1

U2

4
0

2
0 3

U3

5
B

U5

-1 14

10

0 11

U4

0 7

0
U6

12

U7

0 13
B

0
U8

16

U9

15

U10

17

U11

U12

Fig 20.6: Insertions and effect in a balanced tree

Page 232 of 519

CS301 Data Structures

___________________________________________________________________

Now lets discuss what should we do when the insertion of a node makes the tree
unbalanced. For this purpose, consider the node that has a balance 1 in the previous
tree. This is the root of the left subtree of the previous tree. This tree is shown as
shaded in the following figure.
-1
1
0

0
U1

U2

4
0

2
0

U3

1 10

5
B

-1 14

0 11

U4

0
U5

12

0
U6

U7

0 13
B

0
U8

16

U9

15

0
U10

17

U11

U12

Fig 20.7: The node that has balance 1 under

We will now focus our discussion on this left subtree of node having balance 1 before
applying it to other nodes. Look at the following figure (Fig 20.8). Here we are
talking about the tree that has a node with balance 1 as the root. We did not mention
the other part of the tree. We indicate the root node of this left subtree with label A. It
has balance 1. The label B mentions the first node of its left subtree. Here we did not
mention other nodes individually. Rather, we show a triangle that depicts all the
nodes in subtrees. The triangle T3 encloses the right subtree of the node A. We are
not concerned about the number of nodes in it. The triangles T1 and T2 mention the
left and right subtree of the B node respectively. The balance of node B is 0 that
describes that its left and right subtrees are at same height. This is also shown in the
figure. Similarly we see that the balance of node A is 1 i.e. its left subtree is one level
deep than its right subtree. The dotted lines in the figure show that the difference of
depth/height of left and right subtree of node A is 1 and that is the balance of node A.

CS301 Data Structures

___________________________________________________________________

Now considering the notations of figure 20.8, lets insert a new node in this tree and
observe the effect of this insertion in the tree. The new node can be inserted in the tree
T1, T2 or T3. We suppose that the new node goes to the tree T1. We know that this
new node will not replace any node in the tree. Rather, it will be added as a leaf node
at the next level in this tree (T1). The following figure (fig 20.9) shows this
phenomenon.

A
B

1
T3

T1

T2

1
2

new
Fig 20.9: Inserting new node in AVL tree
Due to the increase of level in T1, its difference with the right subtree of node A (i.e.
T3) will become 2. This is shown with the help of dotted line in the above figure. This
difference will affect the balances of node A and B. Now the balance of node A
becomes 2 while balance of node B becomes 1. These new balances are also shown in
the figure. Now due to the balance of node A (that is 2), the AVL condition has been
violated. This condition states that in an AVL tree the balance of a node cannot be
other than 1, 0 or 1. Thus the tree in fig 20.9 is not a balanced (AVL) tree.
Now the question arises what a programmer should do in case of violation of AVL
condition .In case of a binary search tree, we insert the data in a particular order. So

Page 234 of 519

CS301 Data Structures

___________________________________________________________________
that at any time if we traverse the tree with inorder traversal, only sorted data could be
obtained. The order of the data depends on its nature. For example, if the data is
numbers, these may be in ascending order. If we are storing letters, then A is less than
B and B is less than C. Thus the letters are generally in the order A, B, C . This
order of letters is called lexographic order. Our dictionaries and lists of names follow
this order.
If we want that the inorder traversal of the tree should give us the sorted data, it will
not be necessary that the nodes of these data items in the tree should be at particular
positions. While building a tree, two things should be kept in mind. Firstly, the tree
should be a binary tree. Secondly, its inorder traversal should give the data in a sorted
order. Adelson-Velskii and Landis considered these two points. They said that if we
see that after insertion, the tree is going to be unbalanced. Then the things should be
reorganized in such a way that the balance of nodes should fulfill the AVL condition.
But the inorder traversal should remain the same.
Now lets see the example of tree in figure 20.9 and look what we should do to
balance the tree in such a way that the inorder traversal of the tree remains the same.
We have seen in figure 20.9 that the new node is inserted in the tree T1 as a new leaf
node. Thus T1has been modified and its level is increased by 1. Now due to this, the
difference of T1 and T3 is 2. This difference is the balance of node A as T1 and T3
are its left and right subtrees respectively. The inorder traversal of this tree gives us
the result as given below.
T1 B T2 A T3
Now we rearrange the tree and it is shown in the following figure i.e. Fig 20.10.

0
A
0

T1
T2

T3

Fig 20.10: Rearranged tree after inserting a new

By observing the tree in the above figure we notice at first that node A is no longer
the root of the tree. Now Node B is the root. Secondly, we see that the tree T2 that
was the right subtree of B has become the left subtree of A. However, tree T3 is still
the right subtree of A. The node A has become the right subtree of B. This tree is
balanced with respect to node A and B. The balance of A is 0 as T2 and T3 are at the

Page 235 of 519

CS301 Data Structures

___________________________________________________________________
same level. The level of T1 has increased due to the insertion of new node. It is now
at the same level as that of T2 and T3. Thus the balance of B is also 0. The important
thing in this modified tree is that the inorder traversal of it is the same as in the
previous tree (fig 10.9) and is
T1 B T2 A T3
We see that the above two trees give us data items in the same order by inorder
traversal. So it is not necessary that data items in a tree should be in a particular node
at a particular position. This process of tree modification is called rotation.

Example (AVL Tree Building)

Lets build an AVL tree as an example. We will insert the numbers and take care of
the balance of nodes after each insertion. While inserting a node, if the balance of a
node becomes greater than 1 (that means tree becomes unbalance), we will rearrange
the tree so that it should become balanced again. Lets see this process.
Assume that we have insert routine (we will write its code later) that takes a data item
as an argument and inserts it as a new node in the tree. Now for the first node, lets
say we call insert (1). So there is one node in the tree i.e. 1. Next, we call insert (2).
We know that while inserting a new data item in a binary search tree, if the new data
item is greater than the existing node, it will go to the right subtree. Otherwise, it will
go to the left subtree. In the call to insert method, we are passing 2 to it. This data
item i.e. 2 is greater than 1. So it will become the right subtree of 1 as shown below.
1
2
As there are only two nodes in the tree, there is no problem of balance yet.
Now insert the number 3 in the tree by calling insert (3). We compare the number 3
with the root i.e.1. This comparison results that 3 will go to the right subtree of 1. In
the right subtree of 1 there becomes 2. The comparison of 3 with it results that 3 will
go to the right subtree of 2. There is no subtree of 2, so 3 will become the right
subtree of 2. This is shown in the following figure.
1

-2
2

Lets see the balance of nodes at this stage. We see that node 1 is at level 0 (as it is the
root node). The nodes 2 and 3 are at level 1 and 2 respectively. So with respect to the

Page 236 of 519

CS301 Data Structures

___________________________________________________________________
node 1, the deepest level (height) of its right subtree is 2. As there is no left subtree of
node 1 the level of left subtree of 1 is 0. The difference of the heights of left and right
subtree of 1 is 2 and that is its balance. So here at node 1, the AVL condition has
been violated. We will not insert a new node at this time. First we will do the rotation
to make the tree (up to this step) balanced. In the process of inserting nodes, we will
do the rotation before inserting next node at the points where the AVL condition is
being violated. We have to identify some things for doing rotation. We have to see
that on what nodes this rotation will be applied. That means what nodes will be
rearranged. Some times, it is obvious that at what nodes the rotation should be done.
But there may situations, when the things will not be that clear. We will see these
things with the help of some examples.
In the example under consideration, we apply the rotation at nodes1 and 2. We rotate
these nodes to the left and thus the node 1 (along with any tree if were associated with
it) becomes down and node 2 gets up. The node 3 (and trees associated with it, here is
no tree as it is leaf node) goes one level upward. Now 2 is the root node of the tree
and 1 and 3 are its left and right subtrees respectively as shown in the following
figure.

-2

2
2
1

3
Non AVL Tree

AVL tree after applying rotation

We see that after the rotation, the tree has become balanced. The figure reflects that
the balance of node 1, 2 and 3 is 0. We see that the inorder traversal of the above tree
before rotation (tree on left hand side) is 1 2 3. Now if we traverse the tree after
rotation (tree on right hand side) by inorder traversal, it is also 1 2 3. With respect to
the inorder traversal, both the traversals are same. So we observe that the position of
nodes in a tree does not matter as long as the inorder traversal remains the same. We
have seen this in the above figure where two different trees give the same inorder
traversal. In the same way we can insert more nodes to the tree. After inserting a node
we will check the balance of nodes whether it violates the AVL condition. If the tree,
after inserting a node, becomes unbalance then we will apply rotation to make it
balance. In this way we can build a tree of any number of nodes.

Page 237 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 21
Reading Material
Data Structures and Algorithm Analysis in C++
4.4, 4.4.1

Chapter. 4

Summary

AVL Tree Building Example

Cases for Rotation

AVL Tree Building Example

This lecture is a sequel of the previous one in which we had briefly discussed about
building an AVL tree. We had inserted three elements in the tree before coming to the
end of the lecture. The discussion on the same example will continue in this lecture.
Lets see the trees figures below:

-2
2

3
F ig 2 1 .1 : in s e r t(3 ) s in g le le ft r o ta tio n

Fig 21.2: insert(3)

Page 238 of 519

CS301 Data Structures

___________________________________________________________________
Node containing number 2 became the root node after the rotation of the node having
number 1. Note the direction of rotation here.
Lets insert few more nodes in the tree. We will build an AVL tree and rotate the node
when required to fulfill the conditions of an AVL tree.
To insert a node containing number 4,we will, at first, compare the number inside the
root node. The current root node is containing number 2. As 4 is greater than 2, it will
take the right side of the root. In the right subtree of the root, there is the node
containing number 3. As 4 is also greater than 3, it will become the right child of the
node containing number 3.
2

4
Fig 21.3: insert(4)
Once we insert a node in the tree, it is necessary to check its balance to see whether it
is within AVL defined balance. If it is not so, then we have to rotate a node. The
balance factor of the node containing number 4 is zero due to the absence of any left
or right subtrees. Now, we see the balance factor of the node containing number 3. As
it has no left child, but only right subtree, the balance factor is 1. The balance factor
of the node containing number 1 is 0. For the node containing number 2, the height of
the left subtree is 1 while that of the right subtree is 2. Therefore, the balance factor
of the node containing number 2 is 1 2 = -1. So every node in the tree in fig. 21.3
has balance factor either 1 or less than that. You must be remembering that the
condition for a tree to be an AVL tree, every nodes balance needs not to be zero
necessarily. Rather, the tree will be called AVL tree, if the balance factor of each
node in a tree is 0, 1 or 1. By the way, if the balance factor of each node inside the
tree is 0, it will be a perfectly balanced tree.

-2

Fig 21.4: insert(5)

Next, we insert a node containing number 5 and see the balance factor of each node.
The balance factor for the node containing 5 is 0. The balance factor for node
containing 4 is 1 and for the node containing 3 is -2. The condition for AVL is not
Page 239 of 519

CS301 Data Structures

___________________________________________________________________
satisfied here for the node containing number 3, as its balance factor is 2. The
rotation operation will be performed here as with the help of an arrow as shown in the
above Fig 21.4. After rotating the node 3, the new tree will be as under:
2

Fig 21.5: insert(5)

You see in the above figure that the node containing number 4 has become the right
child of the node containing number 2. The node with number 3 has been rotated. It
has become the left child of the node containing number 4. Now, the balance factor
for different nodes containing numbers 5, 3 and 4 is 0. To get the balance factor for
the node containing number 2, we see that the height of the left subtree containing
number 2 is 1 while height of the right subtree is 2. So the balance factor of the node
containing number 2 is 1. We saw that all the nodes in the tree above in Fig 21.5
fulfill the AVL tree condition.
If we traverse the tree Fig 21.5, in inorder tree traversal, we get:
1 2 3 4 5
Similarly, if we traverse the tree in inorder given in Fig 21.4 (the tree before we had
rotated the node containing number 3), following will be the output.
1 2 3 4 5
In both the cases above, before and after rotation, we saw that the inorder traversal of
trees gives the same result. Also the root (node containing number 2) remained the
same.
See the Fig 21.4 above. Considering the inorder traversal, we could arrange the tree in
such a manner that node 3 becomes the root of the tree, node 2 as the left child of
node 3 and node 1 as the left child of the node 2. The output after traversing the
changed tree in inorder will still produce the same result:
1 2 3 4 5
While building an AVL tree, we rotate a node immediately after finding that that the
node is going out of balance. This ensures that tree does not become shallow and
remains within the defined limit for an AVL tree.
Lets insert another element 6 in the tree. The figure of the tree becomes:

Page 240 of 519

CS301 Data Structures

___________________________________________________________________
2

-2

5
6

Fig 21.6: insert(6)

The newly inserted node 6 becomes the right child of the node 5. Usually, after the
insertion of a node, we will find out the node factor for each node and rotate it
immediately. This is carried out after finding the difference out of limit. The balance
factor for the node 6 is 0, for node 5 is 1 and 0 for node 3. Node 4 has 1 balance
factor and node 1 has 0. Finally, we check the balance factor of the root node, node 2,
the left subtrees height is 1 and the right subtrees height is 3. Therefore, the balance
factor for node 2 is 2, which necessitates the rotation of the root node 2. Have a look
on the following figure to see how we have rotated the node 2.
4
2

5
3

Fig 21.7: insert(6)

Now the node 4 has become the root of the tree. Node 2, which was the root node, has
become the left child of node 4. Nodes 5 and 6 are still on their earlier places while
remaining the right child and sub-child of node 4 respectively. However, the node 3,
which was left child of node 4, has become the right child of node 2.
Now, lets see the inorder traversal of this tree:
1 2 3 4 5 6
You are required to practice this inorder traversal. It is very important and the basic
point of performing the rotation operation is to preserve the inorder traversal of the
tree. There is another point to note here that in Binary Search Tree (BST), the root
node remains the same (the node that is inserted first). But in an AVL tree, the root
node keeps on changing.
In Fig 21.6: we had to traverse three links (node 2 to node 4 and then node 5) to reach
the node 6. While after rotation, (in Fig 21.7), we have to traverse the two links (node
Page 241 of 519

CS301 Data Structures

___________________________________________________________________
4 and 5) to reach the node 6. You can prove it mathematically that inside an AVL tree
built of n items; you can search up to 1.44log2n levels to find a node inside. After this
maximum number of links traversal, a programmer will have success or failure, as
1.44log2n is the maximum height of the AVL tree. Consider the BST case, where we
had constructed a linked list. If we want to build a BST of these six numbers, a linked
list structure is formed. In order to reach the node 6, we will have to traverse five
links. In case of AVL tree, we had to traverse two links only.
Lets add few more items in the AVL tree and see the rotations performed to maintain
AVL characteristics of the tree.
4
2

-2

5
3

Fig 21.8: insert(7)

Node 7 is inserted as the right child of node 6. We start to see the balance factors of
the nodes. The balance factors for node 7, 6 are 0 and 1 respectively. As the balance
factor for node 5 is 2, the rotation will be performed on this node. After rotation, we
get the tree as shown in the following figure.
4
2

Fig 21.9: insert(7)

After the rotation, node 5 has become the left child of node 6. We can see in the Fig
21.9 that the tree has become the perfect binary tree. While writing our program, we
will have to compute the balance factors of each node to know that the tree is a
perfectly balanced binary tree. We find that balance factor for all nodes 7, 5, 3, 1, 6, 2
and 4 is 0. Therefore, we know that the tree is a perfect balanced tree. Let see the
inorder traversal output here:
1 2 3 4 5 6 7
It is still in the same sequence and the number 7 has been added at the end.

Page 242 of 519

CS301 Data Structures

___________________________________________________________________
4
2

16
Fig 21.10: insert(16)

We have inserted a new node 16 in the tree as shown in the above Fig 21.10. This
node has been added as the right child of the node 7. Now, lets compute the balance
factors for the nodes. The balance factor for nodes 16, 7, 5, 3, 1, 6, 2 and 4 is either 0
or 1. So this fulfills the condition of a tree to be an AVL. Lets insert another node
containing number 15 in this tree. The tree becomes as given in the figure below:
4
2

-2
16

15
Fig 21.11: insert(15)

Next step is to find out the balance factor of each node. The factors for nodes 5 and 16
are 0 and 1 respectively. This is within limits of an AVL tree but the balance factor
for node 7 is 2. As this is out of the limits of AVL, we will perform the rotation
operation here. In the above diagram, you see the direction of rotation. After rotation,
we have the following tree:

Page 243 of 519

CS301 Data Structures

___________________________________________________________________
4
2

16

Fig 21.12: insert(15)

15

Node 7 has become the left child of node 16 while node 15 has attained the form of
the right child of node 7. Now the balance factors for node 15, 7 and 16 are 0, -1 and
2 respectively. Note that the single rotation above when we rotated node 7 is not
enough as our tree is still not an AVL one. This is a complex case that we had not
encountered before in this example.

Cases of Rotation
The single rotation does not seem to restore the balance. We will re-visit the tree and
rotations to identify the problem area. We will call the node that is to be rotated as
(node requires to be re-balanced). Since any node has at the most two children, and a
height imbalance requires that s two sub-trees differ by two (or 2), the violation
will occur in four cases:
1. An insertion into left subtree of the left child of .
2. An insertion into right subtree of the left child of .
3. An insertion into left subtree of the right child of .
4. An insertion into right subtree of the right child of .
The insertion occurs on the outside (i.e., left-left or right-right) in cases 1 and 4.
Single rotation can fix the balance in cases 1 and 4.
Insertion occurs on the inside in cases 2 and 3 which a single rotation cannot fix.

Page 244 of 519

CS301 Data Structures

___________________________________________________________________

kk22

k
k11

kk11

k
k22
Z

Level n-2

X
Z

Level n-1
new
Level n

new
Fig 21.13: Single right rotation to fix case 1

We have shown, single right notation to fix case 1. Two nodes k2 and k1 are shown in
the figure, here k2 is the root node (and also the node, k1 is its left child and Z
shown in the triangle is its right child. The nodes X and Y are the left and right
subtrees of the node k1. A new node is also shown below to the triangle of the node X,
the exact position (whether this node will be right or left child of the node X) is not
mentioned here. As the new node is inserted as a child of X that is why it is called an
outside insertion, the insertion is called inside if the new node is inserted as a child of
the node Y. This insertion falls in case 1 mentioned above, so by our definition above,
single rotation should fix the balance. The k2 node has been rotated single time
towards right to become the right child of k1 and Y has become the left child of k2. If
we traverse the tree in inorder fashion, we will see that the output is same:
X k1 Y k2 Z
Consider the the figure below:

kk11

k1
kk22

k2

k
k12

Level n-2

Level n-1

Level n
Fig 21.14: Single left rotation to fix case 4

In this figure (Fig 21.14), the new node has been inserted as a child node of Z, that is
why it is shown in bigger size covering the next level. Now this is an example of case
4 because the new node is inserted below the right subtree of the right child of the
Page 245 of 519

CS301 Data Structures

___________________________________________________________________
root node (). One rotation towards should make it balanced within limits of AVL
tree. The figure on the right is after rotation the node k1 one time towards left. This
time node Y has become the right child node of the node k1.
In our function of insertion in our code, we will do insertion, will compute the balance
factors for nodes and make rotations.
Now, we will see the cases 2 and 3, these are not resolved by a single rotation.

k1

k2

k1

kk12

kk22
Z

Level n-2

Level n-1

Level n
new

new
Fig 21.15: Single right rotation fails to fix case 2

We see here that the new node is inserted below the node Y. This is an inside
insertion. The balance factor for the node k2 became 2. We make single rotation by
making right rotation on the node k2 as shown in the figure on the right. We compute
the balance factor for k1, which is 2. So the tree is still not within the limits of AVL
tree. Primarily the reason for this failure is the node Y subtree, which is unchanged
even after making one rotation. It changes its parent node but its subtree remains
intact. We will cover the double rotation in the next lecture.
It is very important that you study the examples given in your text book and try to
practice the concepts rigorously.

Page 246 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 22
Reading Material
Data Structures and Algorithm Analysis in C++
4.4.2

Chapter. 4

Summary

Cases of rotations
Left-right double rotation to fix case 2
Right-left double rotation to fix case 3
C++ Code for avlInsert method

Cases of rotations
In the previous lecture, we discussed how to make insertions in the AVL tree. It was
seen that due to the insertion of a node, the tree has become unbalanced. Resultantly,
it was difficult to fix it with the single rotation. We have analyzed the insertion
method again and talked about the node. The new node will be inserted at the left or
right subtree of the s left child or at the left or right subtree of the s right child.
Now the question arises whether the single rotation help us in balancing the tree or
not. If the new node is inserted in the left subtree of the s left child or in the right
subtree of s right child, the balance will be restored through single rotation.
However, if the new node goes inside the tree, the single rotation is not going to be
successful in balancing the tree.
We face four scenarios in this case. We said that in the case-1 and case-4, single
rotation is successful while in the case-2 and case-3 single rotation does not work.
Lets see the tree in the diagram given below.
Single right rotation fails to fix case 2.

k2

k1

k1

k2
Z

Level n-2

Level n-1

X
Y

Level n
new

new

Page 247 of 519

CS301 Data Structures

___________________________________________________________________
In the above tree, we have node as k2, which has a left child as k1. Whereas X and Y
are its left and right children. The node k2 has a right child Z. Here the newly inserted
node works as the left or right child of node Y. Due to this insertion, one level is
increased in the tree. We have applied single rotation on the link of k1 and k2. The
right side tree in the figure is the post-rotation tree. The node k1 is now at the top
while k2 comes down and node Y changes its position. Now if you see the levels of
the node, these are seen same. Have a look on the level of the node i.e. k1 which
reflects that the difference between the left and right side levels is still 2. So the single
rotation does not work here.
Lets see how we can fix that problem. A fresh look on the following diagram will
help us understand the problem.

k2
k1
Z
X
Y
Here k2 is the root node while k1 and Z are the right and left children respectively.
The new node is inserted under Y so we have shown Y in a big triangle. The new node
is inserted in the right subtree of k1, increasing its level by 1. Y is not empty as the
new node was inserted in it. If Y is empty, the new node will be inserted under k1. It
means that Y has a shape of a tree having a root and possibly left and right subtrees.
Now view the entire tree with four subtrees connected with 3 nodes. See the diagram
below.

Page 248 of 519

CS301 Data Structures

___________________________________________________________________

k3
k1
D
k2
A
B

We have expanded the Y and shown the root of Y as K2, B and C are its left and right
subtrees. We have also changed the notations of other nodes. Here, we have A, B, C
and D as subtrees and k1, k2 and k3 as the nodes. Lets see where the new node is
inserted in this expanded tree and how can we restore its balance. Either tree B or C is
two levels deeper than D. But we are not sure which one is deeper. The value of new
node will be compared with the data in k2 that will decide that this new node should
be inserted in the right subtree or left subtree of the k2. If the value in the new node is
greater than k2, it will be inserted in the right subtree i.e. C. If the value in the new
node is smaller than k2, it will be inserted in the left subtree i.e. B. See the diagram
given below:

Page 249 of 519

CS301 Data Structures

___________________________________________________________________

k3
k1
k2
D
A
B

1
2

new

new

New node inserted at either of the two spots

We have seen the both possible locations of the new node in the above diagram. Lets
see the difference of levels of the right and left subtrees of the k3. The difference of B
or C from D is 2. Therefore the expansion of either of B or C, due to the insertion of
the new node, will lead to a difference of 2. Therefore, it does not matter whether the
new node is inserted in B or C. In both of the cases, the difference becomes 2. Then
we try to balance the tree with the help of single rotation. Here the single rotation
does not work and the tree remains unbalanced. To re-balance it, k3 cannot be left as
the root. Now the question arises if k3 cannot become root, then which node will
become root? In the single rotation, k1 and k3 were involved. So either k3 or k1 will
come down. We have two options i.e. left rotation or right rotation. If we turn k1 into
a root, the tree will be still unbalanced. The only alternative is to place k2 as the new
root. So we have to make k2 as root to balance the tree. How can we do that?
If we make k2 the root, it forces k1 to be k2s left child and k3 to be its right child.
When we carry out these changes, the condition is followed by the inorder traversal.
Lets see the above tree in the diagram. In that diagram, the k3 is the root and k1 is its
left child while k2 is the right child of k1. Here, we have A, B, C and D as subtrees.
You should know the inorder traversal of this tree. It will be A, k1, B, k2, C, k3 and D
where A, B, C and D means the complete inorder traversal of these subtrees. You
should memorize this tree traversal.
Page 250 of 519

CS301 Data Structures

___________________________________________________________________
Now we have to take k2 as the root of this tree and rearrange the subtrees A, B, C and
D. k1 will be the left child of k2 while k3 is going to be its right child. Finally if we
traverse this new tree, it will be the same as seen above. If it is not same, it will mean
that there is something wrong in the rotation. We have to find some other solution.
Now lets see how we can rotate our tree so that we get the desired results.

Left-right double rotation to fix case 2

We have to perform a double rotation to achieve our desired results. Lets see the
diagram below:
Left-right double rotation to fix case 2.

k3

k3
Rotate left

k1

k2

k2

k1
D
A
B

C
1

B
2

new

new

new

A
new

On the left side, we have the same tree with k3 as its root. We have also shown the
new nodes as new and new i.e. the new node will be attached to B or C. At first, we
will carry out the left rotation between k1 and k2. During the process of left rotation,
the root k1 comes down and k2 goes up. Afterwards, k1 will become the left child of
k2 and the left subtree of k2 i.e. B, will become the right subtree of k1. This is the
single rotation. You can see the new rotated tree in the above figure. It also shows that
the B has become the right child of the k1. Moreover, the new node is seen with the B.
Now perform the inorder traversal of this new rotated tree. It is A, k1, B, k2, C, k3 and
D. It is same as witnessed in case of the inorder traversal of original tree. With this
single rotation, the k2 has gone one step up while k1 has come down. Now k2 has
become the left child of k3. We are trying to make the k2 the root of this tree. Now
what rotation should we perform to achieve this?
Now we will perform right rotation to make the k2 the root of the tree. As a result, k1
and k2 have become its left and right children respectively. The new node can be
inserted with B or C. The new tree is shown in the figure below:

Page 251 of 519

CS301 Data Structures

___________________________________________________________________

k3

k2

Rotate right

k2

k1

k3

k1

B
new

new

new

new

Now lets see the levels of new and new. Of these, one is the new node. Here you can
see that the levels of new, new i.e. A and D are the same. The new tree is now a
balanced one. Lets check the inorder traversal of this tree. It should be the same as
that of the original tree. The inorder traversal of new tree is A, k1, B, k2, C, k3 and D,
which is same as that of the original tree.
This is known as double rotation. In double rotation, we perform two single rotations.
As a result, the balance is restored and the AVL condition is again fulfilled. Now we
will see in which order, the double rotation is performed? We performed a left
rotation between k1 and k2 link, followed by a right rotation.

Right-left double rotation to fix case 3

In case, the node is inserted in left subtree of the right child, we encounter the same
situation as discussed above. But here, we will perform right rotation at first before
going for a left rotation. Lets discuss this symmetric case and see how we can apply
double rotation here. First we perform the right rotation.

Page 252 of 519

CS301 Data Structures

___________________________________________________________________

Right-left double rotation to fix case 3.

Rotate right

k1

k1

k3

k2

A
k2

k3
D

Here k1 is the root of the tree while k3 is the right child of the k1. k2 is the inner child.
It is the Y tree expanded again here and the new node will be inserted in the k2s right
subtree C or left subtree B. As we have to transform the k2 into the root of the tree, so
the right rotation between the link k2 and k3 will be carried out. As a result of this
rotation, k2 will come up and k3 will go down. The subtree B has gone up with the k2
while subtree C is now attached with the k3. To make the k2 root of the tree, we will
perform the left rotation between then k1 and k2. Lets see this rotation in the figure
below:

Rotate left

k1

k2

k2

k1

k3

A
k3
B

A
C

In the above figure at the right side, we have the final shape of the tree. You can see
that k2 has become the root of the tree. k1 and k3 are its left and right children
respectively. While performing the inorder traversal, you will see that we have
preserved our inorder traversal.
We have started this activity while building an example tree. We inserted numbers in
it. When the balance factor becomes more than one, rotation is performed. During this
process, we came at a point when single rotation failed to balance the tree. Now there
is need to perform double rotation to balance the tree that is actually two single
rotations. Do not take double rotation as some complex function, it is simply two

Page 253 of 519

CS301 Data Structures

___________________________________________________________________
single rotations in a special order. This order depends on the final position of the new
node. Either the new node is inserted at the right subtree of the left child of node or
at the left subtree of the right child of node. In first case, we have to perform leftright rotation while in the second case, the right-left rotation will be carried out.
Lets go back to our example and try to complete it. So far, we have 1, 2, 3, 4, 5, 6, 7
and 16 in the tree and inserted 15 which becomes the left child of the node 16. See the
figure below:

4
2
1

6
3

7 k1

16 k2
X
(null
Y

Z
(null

15
Here we have shown X, Y and Z in case of the double rotation. We have shown Y
expanded and 15 is inside it. Here we will perform the double rotation, beginning
with the right rotation first.

Page 254 of 519

CS301 Data Structures

___________________________________________________________________

4
2
1

6
3

k1
7

k3
16
k2
15

Rotate right

We have identified the k1, k2 and k3 nodes. This is the case where we have to perform
right-left double rotation. Here we want to promote k2 upwards. For this purpose, the
right rotation on the link of k2 and k3 i.e. 15 and 16 will be carried out.

4
2
1

6
3

5
Rotate left

k1
7
k2
15
16

k3

The node 15 now comes up while node 16 has gone down. We have to promote k2 to
the top and k3 and k1 will become its right and left children respectively. Now we will
Page 255 of 519

CS301 Data Structures

___________________________________________________________________
perform left rotation on the link of k1 and k2 i.e. 7 and 15. With this left rotation, 15
goes up and 7 and 16 become its left and right children respectively.

4
2

6
k2

15
k1
7

k3
16

Here we have to check two things. At first, the tree is balanced or not i.e. the AVL
condition is fulfilled or not. Secondly we will confirm that the inorder traversal is
preserved or not. The inorder traversal should be the same as that of the inorder
traversal of original tree. Lets check these two conditions. The depth of the left
subtree of node 4 is 2 while the depth of the right subtree of node 4 is three.
Therefore, the difference of the levels at node 4 is one. So the AVL condition is
fulfilled at node 4. At node 6, we have one level on it left side while at the right side
of node 6, there are two levels. As the difference of levels is one, therefore node 6 is
also balanced according to the AVL condition. Similarly other nodes are also
fulfilling the AVL condition. If you see the figure above, it is clear that the tree is
balanced.
We are doing all this to avoid the link list structure. Whenever we perform rotation on
the tree, it becomes clear from the figure that it is balanced. If the tree is balanced, in
case of searching, we will not have to go very deep in the tree. After going through
the mathematical analysis, you will see that in the worst case scenario, the height of
the tree is 1.44 log2 n. This means that the searching in AVL is logarithmic. Therefore
if there are ten million nodes in an AVL tree, its levels will be roughly as log2(10
million) which is very few. So the traversal in an AVL tree is very simple.
Lets insert some more nodes in our example tree. We will perform single and double
rotations, needed to make the tree balanced. The next number to be inserted is 14. The
position of node 14, according to the inorder traversal, is the right child of 7. Lets see
this in the diagram as:

Page 256 of 519

CS301 Data Structures

___________________________________________________________________

4
k1
2
1

6
3

k3
15

5
k2

Rotate right

16
14

The new node 14 is inserted as the right child of 7 that is the inner subtree of 15. Here
we have to perform double rotation again. We have identified the k1, k2 and k3. k2
has to become the root of this subtree. The nodes k1 and k3 will come down with their
subtrees while k2 is going to become the root of this subtree. After the right rotation
the tree will be as:

4
k1
2
1

6
3

Rotate left

k2

15
14

k3

16
Page 257 of 519

CS301 Data Structures

___________________________________________________________________
With the right rotation, k2 has come one step up while k3 has been turned into the
right child of k2 but k1 is still up. Now we will perform a left rotation on the link of
k1 and k2 to make the k2 root of this subtree. Now think that after this rotation and
rearrangement of node what will be the shape of the tree.
After the double rotation, the final shape of the tree will be as:

4
k2
2

7
k1

k3
15

6
5

14

16

k2 has become the root of the subtree. k1 has attained the role of the left child of k2
and k3 has become the right child of the k2. The other nodes 5, 14 and 16 have been
rearranged according to the inorder traversal. The node 7 has come up where as node
6 and 15 have become its left and right child. Now just by viewing the above figure, it
is clear that the tree is balanced according to the AVL condition. Also if we find out
its inorder traversal, it should be the same as the inorder traversal of original tree. The
inorder traversal of the above tree is 1, 2, 3, 4, 5, 6, 7, 14, 15, and 16. This is in sorted
order so with the rotations the inorder traversal is preserved.
Lets insert some more numbers in our tree. The next number to be inserted is 13.

Page 258 of 519

CS301 Data Structures

___________________________________________________________________

Rotate left

6
5

15
14

16

13
We have to perform single rotation here and rearrange the tree. It will look like as:

15

2
1

14

6
3

16

13

The node 7 has become the root of the tree. The nodes 4, 2, 1, 3, 6, 5 have gone to its
left side while the nodes 15, 14, 13, 16 are on its right side. Now try to memorize the
tree which we build with these sorted numbers. If you remember that it looks like a
link list. The root of that tree was 1. After that we have its right child as 2, the right
child of 2 as 3, then its right child 4 and so on up to 16. The shape of that tree looks
exactly like a linked list. Compare that with this tree. This tree is a balanced one. Now
if we have to traverse this tree for search purposes, we have to go at the most three
levels.
Now you must be clear why we need to balance the trees especially if we have to use
the balanced search trees. While dealing with this AVL condition, it does not matter

Page 259 of 519

CS301 Data Structures

___________________________________________________________________
whether the data, provided to a programmer is sorted or unsorted. The data may be
sorted in ascending order or descending order. It may contain alphabets or names in
any order. We will keep our tree balanced during the insertion and tree will be
balanced at each point. Our tree will not be balanced in the end, it will be balanced
with each insertion. It will not be completely balanced. At the maximum, if we pick
any node, the difference in the levels of its right subtree and left subtree will not be
more than 1.
Now if we have 9 or 10 nodes in the tree and take log of this, it will be near 3.
Therefore our tree has three levels after that there are its leaf nodes. Please keep this
in mind that originally we have thought BST as an abstract data type. We have
defined operations on it like insert, remove and the major method was find. The find
method takes a data item and searches the tree. It will also show that this data item
exists or not in the tree. We also right findMin and findMax methods. In case of a
BST, we can find the minimum and maximum value in it. The minimum will be the
left most node of the tree while the right most node of BST will give the maximum
value. You can confirm it by applying this on the above tree.

C++ Code for avlInsert method

Now lets see the C++ code of avlinsert method. Now we have to include this
balancing procedure in the insert method. We have already written this insert method
which takes some value and adds a node in the tree. That procedure does not perform
balancing. Now we will include this balancing feature in our insert method so that the
newly created tree fulfills the AVL condition.
Here is the code of the function.
/* This is the function used to insert nodes satisfying the AVL condition.*/
TreeNode<int>* avlInsert(TreeNode<int>* root, int info)
{
if( info < root->getInfo() ){
root->setLeft(avlInsert(root->getLeft(), info));
int htdiff = height(root->getLeft()) height(root->getRight());
if( htdiff == 2 )
if( info < root->getLeft()->getInfo() ) // outside insertion case
root = singleRightRotation( root );
else // inside insertion case
root = doubleLeftRightRotation( root );
}
else if(info > root->getInfo() ) {
root->setRight(avlInsert(root->getRight(),info));
int htdiff = height(root->getRight()) height(root->getLeft());
if( htdiff == 2 )
if( info > root->getRight()->getInfo() )
root = singleLeftRotation( root );
else
root = doubleRightLeftRotation( root );
}
Page 260 of 519

CS301 Data Structures

___________________________________________________________________
// else a node with info is already in the tree. In
// case, reset the height of this root node.
int ht = Max(height(root->getLeft()), height(root->getRight()));
root->setHeight( ht + 1 ); // new height for root.
return root;
}
We have named the function as avlInsert. The input arguments are root node and the
info is of the type int. In our example, we are having a tree of int data type. But of
course, we can have any other data type. The return type of avlInsert method is
TreeNode<int>*. This info will be inserted either in the right subtree or left subtree of
root. The first if statement is making this decision by comparing the value of info and
the value of root. If the new info value is less than the info value of root, the new node
will be inserted in the left subtree of root. In the next statement, we have a recursive
call as seen in the following statement.
avlInsert(root->getLeft(), info)
Here we are calling avlInsert method again. We are passing it the first node of the left
subtree and the info value. We are trying to insert this info in the left subtree of root if
it does not already exist in it. In the same statement, we are setting the left of root as
the return of the avlInsert method. Why we are doing setLeft? As we know that this is
an AVL tree and the new data is going to be inserted in the left subtree of root. We
have to balance this tree after the insertion of the node. After insertion, left subtree
may be rearranged to balance the tree and its root can be changed. You have seen the
previous example in which node 1 was the root node in the start and in the end node 7
was its root node. During the process of creation of that tree, the root nodes have been
changing. After the return from the recursive call in our method, the left node may be
different than the one that was before insertion. The complete call is as:
root->setLeft(avlInsert(root->getLeft(), info));
Here we are inserting the new node and also getting the root of the left subtree after
the insertion of the new node. If the root of the left subtree has been changed after the
insertion, we will update our tree by assigning this new left-node to the left of the
root.
Due to the insertion of the new node if we have rearranged the tree, then the balance
factor of the root node can be changed. So after this, we check the balance factor of
the node. We call a function height to get the height of left and right subtree of the
root. The height function takes a node as a parameter and calculates the height from
that node. It returns an integer. After getting the heights of left and right subtrees, we
take its difference as:
int htdiff = height(root->getLeft()) height(root->getRight());
Now if the difference of the heights of left and right subtrees is greater than 1, we
have to rebalance the tree. We are trying to balance the tree during the insertion of

Page 261 of 519

CS301 Data Structures

___________________________________________________________________
new node in the tree. If this difference is 2, we will perform single rotation or double
rotation to balance the tree. We have to deal with one of the four cases as discussed
earlier. This rotation will be carried out with only one node where the balance factor
is 2. As a result, the tree will be balanced. We do not need to rotate each node to
balance the tree. This is characteristic of an AVL tree.
We have seen these rotations in the diagrams. Now we have to perform these ones in
the code. We have four cases. The new node is inserted in the left or right subtree of
the left child of node or the new node is inserted in the left or right subtree of the
right child of node. How can we identify this in the code? We will identify this with
the help of the info value of the new node.
We can check whether the difference (htdiff) is equal to 2 or not. We have if condition
that checks that the info is less than the value of left node of the current node (
node). If this condition is true, it shows that it is the case of outside insertion. The
node will be inserted as the left-most child. We have to perform single rotation. If the
else part is executed, it means that this is the case of inside insertion. So we have to
perform double rotation. We are passing node to the rotation function. When a
programmer performs these rotations, he gets another node that is assigned to root.
This is due to the fact that the root may change during the rotations.
We encounter this case when the new node is inserted in the left subtree of the root.
The new node can be inserted in the right subtree of the root. With respect to
symmetry, that code is also similar. You can easily understand that. In the code, we
have if-else condition that checks whether the info is greater than the value of root.
Therefore, it will be inserted in the right subtree. Here again, we made a recursive call
to avlInsert by passing it the first node of the right subtree of the root. When this
recursive call is completed, we may have to perform rotations. We set the right node
of the root as the node returned by the avlInsert. After this, we check the balance
factor. We calculate the difference as:
int htdiff = height(root->getRight()) height(root->getLeft());
In the previous case, we have subtracted the height of right tree from left tree. Here
we have subtraction the height of left tree from the right tree. This is due to the fact
that we want to avoid the -ve value. But it does not matter whether we get the -ve
value. We can take its absolute value and test it. If the value of htdiff is 2 , we have to
identify the case. We check that the new node is to be inserted in the right subtree of
the right child of node or in the left subtree of the right child of node. If the info is
greater than the value of the right child of root, it is the case 4. Here we will restore
the balance of the tree by performing single rotation. If the else part is executed, we
have to perform the double rotation. These rotation routines will return the root of the
rearranged tree that will be set as a root.
In the end, we have the third case. This is the case when the info is equal to the root. It
means that the data is already in the tree. Here we will readjust the height if it is
needed. We will take the maximum height between the left and right subtree and set
this height plus one as the height of the root. We have added one to the height after
adding one node to the tree. In the end, we return the root. We have included the

Page 262 of 519

CS301 Data Structures

___________________________________________________________________
balancing functionality in our tree and it will remain balanced.
We have not seen the code of rotation routines so far. The code of these methods is
also very simple. We have to move some of pointers. You can easily understand that
code. We will discuss this in the next lecture. Please see the code of single and double
rotation in the book and try to understand it.

Page 263 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 23
Reading Material
Data Structures and Algorithm Analysis in C++
4.4.1, 4.4.2

Chapter. 4

Summary

Single Right Rotation

Single Left Rotation
Double Right-Left Rotation
Double Left- Right Rotation
Deletion in AVL Tree
Cases of Deletion in AVL Tree

We demonstrated how to write the insert procedure for an AVL tree in the previous
lecture. The single and double rotation call methods were also discussed at length.
While inserting a node if we see that the tree is going to be unbalanced, it means that
the AVL condition is being violated. We also carried out the balancing of the tree by
calling single or double rotation. Now lets see the code of these routines i.e. single
and double rotation.

Single Right Rotation

At first, we will discuss the code of SingleRightRotation function. Its argument is
TreeNode, which is given the name k2. The reason of using k2 is that it alongwtih k1
etc was used in the earlier discussion. After the rotation, this function will return a
pointer, as tree will be re-organized. Resultantly, its root will be changed. Following
is the code of this SingleRightRotation function.
TreeNode<int>* singleRightRotation(TreeNode<int>* k2)
{
if( k2 == NULL ) return NULL;
// k1 (first node in k2's left subtree) will be the new root
TreeNode<int>* k1 = k2->getLeft();
// Y moves from k1's right to k2's left
k2->setLeft( k1->getRight() );
k1->setRight(k2);
// reassign heights. First k2
int h = Max(height(k2->getLeft()), height(k2->getRight()));
k2->setHeight( h+1 );
// k2 is now k1's right subtree
Page 264 of 519

CS301 Data Structures

___________________________________________________________________
h = Max( height(k1->getLeft()), k2->getHeight());
k1->setHeight( h+1 );
return k1;
}
In the function, at first, we check whether k2 is NULL. If it is NULL, the function is
exited by returning NULL. If it is not NULL, the rotation process starts. The figure
below depicts the single right rotation process. In this diagram, we see that k2 has
been shown as the root node to this function. We see that k1 is the left subtree of k2.

k2
k1

Z
Y

X
Fig 23.1: Single Right Rotation

We are going to apply single right rotation on the link between k1 and k2. The node
k1 will be the new root node after rotation. So we get its value by the following
statement
TreeNode <int>* k1 = k2 -> getLeft() ;
Due to the single right rotation, k2 has come down, resulting in the upward movement
of k1. The tree Y has to find its new place. The place of the trees X and Z remains
intact. The change of place of Y is written in the code as below.
k2->setLeft( k1->getRight() );
In the above statement, we get the right child of k1 (i.e. k1 -> getRight() ) i.e. Y and
pass it to setLeft function of k2. Thus it becomes the left child of k2. By the statement
k1 -> setRight (k2) ;
We set the node k2 as right child of k1. Now after these three statements, the tree has
been transformed into the following figure.

Page 265 of 519

CS301 Data Structures

___________________________________________________________________

k1
k2
X

Fig 23.2: Tree after rotation

From the above figure, it is reflected that k1 is now the root of the tree while k2 is its
right child. The Y, earlier the right subtree of k1 (see fig 23.1), has now become the
right subtree of k2. We see that the inorder traversal of this tree is X k1 Y k2 Z. It is
the same as of the tree before rotation in fig 23.1 i.e. X k1 Y k2 Z.
Now we set the heights of k1 and k2 in this re-arranged tree. In the code, to set the
height of k2, we have written
int h = Max(height(k2->getLeft()) , height(k2->getRight()));
k2->setHeight( h+1 );
Here we take an integer h and assign it a value. This value is the maximum height
among the two subtrees i.e. height of left subtree of k2 and height of right subtree of
k2. Then we add 1 to this value and set this value as the height of k2. We add 1 to h as
the height of root is one more than its child.
Similarly we set the height of k1 and get the heights of its left and right subtrees
before finding the higher value of these two by the Max function. Afterwards, we add
1 to this value and set this value as the height of k1. The following two statements
perform this task.
h = Max( height(k1->getLeft()), k2->getHeight());
k1->setHeight( h+1 );
In the end, we return the root node i.e. k1. Thus the right single rotation is completed.
In this routine, we saw that this routine takes root node of a tree as an argument and
then rearranges it, resets the heights and returns the new root node.
Height Function
In the above SingleRightRotation, we used the height routine. This routine calls the
getHeight function for the argument passed to it and returns the value got from that
function. If there is an empty tree, then by definition its height will be -1. So in this
routine, we return the value -1 if the argument passed to it is NULL. Following is the
code of this routine.

Page 266 of 519

CS301 Data Structures

___________________________________________________________________
int height( TreeNode<int>* node )
{
if( node != NULL ) return node->getHeight();
return 1;
}

Single Left Rotation

This rotation is almost similar to the single right rotation. We start from k1. k1 is the
root while k2 is its right child. Due to the left rotation, the right child k2 of the root k1
will become the root. The k1 will go down to the left child of k2. The following figure
shows that we have to change the positions of some pointers. The condition of the tree
before and after single left rotation is also seen.
k1

k2
k2

k1

Z
Y

Fig 23.3: Single Left Rotation

Following is the code of the function performing the single left rotation.
TreeNode<int>* singleLeftRotation( TreeNode<int>* k1 )
{
if( k1 == NULL ) return NULL;
// k2 is now the new root
TreeNode<int>* k2 = k1->getRight();
k1->setRight( k2->getLeft() ); // Y
k2->setLeft( k1 );
// reassign heights. First k1 (demoted)
int h = Max(height(k1->getLeft()), height(k1->getRight()));
k1->setHeight( h+1 );
// k1 is now k2's left subtree
h = Max( height(k2->getRight()), k1->getHeight());
k2->setHeight( h+1 );
return k2;

Page 267 of 519

CS301 Data Structures

___________________________________________________________________
}
In the above code, it is evident that if the node passed to the function is not NULL
(that means k1 is not NULL), we assign to k2 the value of k1 -> getRight (). Now in
the rotation, Y tree i.e. the left subtree of k2, becomes the right subtree of k1. We
write it in the next statement as.
k1->setRight( k2->getLeft() );
In this statement, we get the left child of k2 i.e. Y and set it as the right child of k1.
Afterwards, k1 is set as the left child of k2. The following statement reflects this
process.
k2->setLeft( k1 );
After single left rotation, the k2 becomes the root node after going up. However, k1
goes down as the left child of k2. The tree Y, earlier the left child of k2 has, now
become the right child of k1.X and Z continue to remain at the previous positions.
The tree has been transformed with changes into the heights of k1 and k2. We reassign the heights to these nodes in line with the process adopted in the right single
rotation. At first, we adjust the height of k1. We get the heights of left and right
subtrees of k1. The greater of these is taken and assigned to an int h. Then we add 1 to
this value and set it as the height of k1. The following two lines of the code execute
this task.
int h = Max(height(k1->getLeft()), height(k1->getRight()));
k1->setHeight( h+1 );
Similarly, we adjust the height of k2 by getting the greater one among the heights of
its right subtree and its left subtree. Taking k1 as its left subtree, we get the height of
k1. We add 1 and set it as the height of k2. Following are the statements that perform
this task.
h = Max( height(k2->getRight()), k1->getHeight());
k2->setHeight( h+1 );
Finally, k2 is returned as the root of the tree.

Double Right-Left Rotation

As obvious from the nomenclature, in the double rotation, we at first carry out the
right rotation before going ahead with the left rotation. Lets say that we pass it the
node k1. Look at the following figure to see what are the nodes k1, k2 and k3. The
figure 23.4 shows the first step (A) of double right-left rotation that is the single right
rotation. It shows the rearranged tree on the right side in the figure.

Page 268 of 519

CS301 Data Structures

___________________________________________________________________

k1

k1

k2

k3
A

A
k2
D
B

k3

Fig 23. 4: Double Right-Left Rotation (A)

In the left portion of the above figure, we see that k1 is the root of the tree. k3 is the
right child of k1 while k2 is the left child of k3. A, B, C and D are trees. We carry out
the right rotation between the link of k3 and k2. In the code, it is shown that if k1 is
not NULL, we go ahead and perform the rotation. The code of this double rotation is
given below.
TreeNode<int>* doubleRightLeftRotation(TreeNode<int>* k1)
{
if( k1 == NULL ) return NULL;
// single right rotate with k3 (k1's right child)
k1->setRight( singleRightRotation(k1->getRight()));
// now single left rotate with k1 as the root
return singleLeftRotation(k1);
}
We perform the right rotation with the help of k3. Now think about the single right
rotation. We are at k3 and its left child k2 will be involved in the rotation. In the code,
for the single right rotation, we have written
k1->setRight( singleRightRotation(k1->getRight()));
Here, we are passing the right child of k1 (i.e. k3) to the single right rotation. The
function singleRightRotation itself will find the left child of k3 to perform the single
right rotation. The singleRightRotation will take k3 downward and bring up its left

Page 269 of 519

CS301 Data Structures

___________________________________________________________________
child i.e. k2. The left subtree of k2 i.e. C, now becomes the right subtree of k3. Due to
this single right rotation, the tree will be transformed into the one, shown in the right
part of the figure 23.4.
After this we apply the single left rotation on this tree that is got from the single right
rotation. This single left rotation is done by passing it k1. In this rotation, k1 goes
down and k2 i.e. the right child of k1, becomes the root. We return the result of single
left rotation. The following statement in the above code performs this task.
return singleLeftRotation(k1);
Thus the tree gets its final form i.e. after the double right-left rotation. The following
figure shows the single left rotation and the final form of the tree.

k2

k1
k1

k2

k3

A
k3
A

Fig 23.5: Double right-left rotation (B)

In the above figure, it was witnessed that the double rotation consists of few
statements. Thats why, we have had written the routines for single right and left
rotations. In the double right-left rotation, we just carried out a single right rotation
and the single left rotation to complete the double right-left rotation.

Double Left-Right Rotation

While performing the double left-right rotation, we simply carry out the single left
rotation at first. It is followed by the single right rotation. Following is the code of the
double Left-Right rotation.
TreeNode<int>* doubleLeftRightRotation(TreeNode<int>* k3)
{
if( k3 == NULL ) return NULL;
// single left rotate with k1 (k3's left child)
k3->setLeft( singleLeftRotation(k3->getLeft()));

Page 270 of 519

CS301 Data Structures

___________________________________________________________________
// now single right rotate with k3 as the root
return singleRightRotation(k3);
}
In the above code, we receive the argument node as k3. As we have to do the left
rotation first, it is applied to the left child of k3. From the figure below (Fig 23.6), we
see that the left child of k3 is k1. However, when we send k1 to the
singleLeftRotation it will take the right child of k1 i.e. k2. This process will rotate the
link between k1and k2. The tree formed as a result of this single left rotation, is
shown at the right side in the figure 23.6.

k3

k3
k2

k1

D
D

k1

k2

A
A
B

C
Fig 23.6: Double left-right rotation (A)

In this new rearranged tree, k2 comes up while k1 goes down, becoming the left
subtree of k2. The subtree B, earlier left subtree of k2, now becomes the right subtree
of k1.
The second step in the double left-right rotation is to apply the single right rotation on
this new tree. We will carry out the single right rotation on k3. The pictorial
representation of this single right rotation on k3 is given in the figure below.

k3

k2

k2

k1

k3

k1
C
A

B
Fig 23.7: Double left-right rotation (B)
Page 271 of 519

CS301 Data Structures

___________________________________________________________________

In the code, we wrote the statement for this as below.

return singleRightRotation(k3);
We pass k3 to the singleRightRotation. It is, internally applied to the left child of k3
ie. k2. Now due to the right rotation, k2 will come up and k3 will become its right
child. The node k1 will remain the left child of k2. The subtree C that was the right
subtree of k2, will now be the left subtree of k3.
By now, the discussion on the insert routine has been completed. We have written the
insert routine so that data item could be passed to it and inserted at a proper position
in the tree. After inserting a node, the routine checks the balance or height factors and
does the left or right rotation if needed and re-balances the tree.
Note that a property of the AVL tree is that while inserting a node in an AVL tree if
we have to balance the tree then we have to do only one single right or left rotation or
one double rotation to balance the tree. There is no such situation that we have to do a
number of rotations. We do the one single or one double rotation at the node whose
balance has become 2 after inserting a node.

Deletion in AVL Tree

The deletion of a data item from a data structure has always been difficult whatever
data structure we use. The deletion is the inverse of insertion. In deletion there is a
given value x and an AVL tree T. We delete the node containing the value x and
rebalance the tree if it becomes unbalance after deleting the node. We will see that the
deletion of a node is considerably more complex than the insertion of a node. It is
complex in the sense that we may have to do more than one rotations to rebalance the
tree after deleting a node. We will discuss the deletion case by case and will see that
about what points we have to take care in the deletion process.
We are not going to write the code for deletion here. If we have to use AVL tree
routines or class, the deletion and insertion routines of AVL tree are available in
standard library. We can use these routines in our program. We have no need to write
these routines. But here we discuss these to understand their functionality.
We know that insertion in a height-balanced tree requires at most one single rotation
or one double rotation. We do this rotation at the node whose balance violates the
AVL condition. We can use rotations to restore the balance when we do a deletion. If
the tree becomes unbalance after deleting a node then we use rotations to rebalance it.
We may have to do a rotation at every level of the tree. Thus in the worst case of
deletion we have to do log2 N rotations. As log2 N is the number of levels of a tree of
N nodes.
Lets consider an example of deleting a node from a tree. In this example, we will
discuss the worst case of deletion that is we have to do rotation at each level after
deleting a node. Look at the following figure i.e. Fig 23.8. In this figure the root of
the tree is node N and we have expanded only the left subtree of this node. The right
subtree of it is indicated by a triangle. We focus on the left subtree of N. The balance
Page 272 of 519

CS301 Data Structures

___________________________________________________________________
of each non-leaf node of the left subtree of N is 1. This means that at every non-leaf
node the depth/height of left subtree is one shorter than the height of right subtree. For
example look at the node C. The left subtree of C is one level deep where as its right
subtree is two levels deep. So balance of it is 1 2 = -1. If we look at node I its left
subtree has height 2 as there are two levels where nodes G and H exists. The right
subtree of I has number of levels (i.e. height) 3 where exists the nodes K, L and M
respectively. Thus the balance of I is 2 3 = -1. Similarly we can see that other nodes
also have the balance 1. This tree is shown in the following figure.

N
F
C
A

I
D

G
E

K
H

L
M

Fig 23.8: Tree under consideration

Here in this tree, the deletion of node A from the left subtree causes the worst case of
deletion in the sense that we have to do a large number of rotations. Now we delete
the node A from the tree. The effect of this deletion is that if we consider the node C
the height of its left subtree is zero now. The height of the right subtree of C is 2.
Thus the balance of C is 2 now. This makes the tree unbalance. Now we will do a
rotation to make the tree balanced. We rotate the right subtree of C that means the link
between C and D so that D comes up and C goes down. This is mentioned in the
figure below.

Page 273 of 519

CS301 Data Structures

___________________________________________________________________

N
F
C

I
D

G
E

K
H

L
M

Fig 23.9: Single left rotation after deleting node A

After this rotation the tree has transformed into the following figure. Now D becomes
the left child of F and C becomes the left child of D.

N
F
D
C

I
E

K
H

L
M

Fig 23.10: Tree (Fig 23.9) after rotation

By looking at the inorder traversal of the tree, we notice that it is preserved. The
inorder traversal of the tree before rotation (i.e. fig 23.8) is C D E F G H I J K L M N.
Page 274 of 519

CS301 Data Structures

___________________________________________________________________
Now if we traverse the tree after rotation (i.e. fig 23.9) by inorder traversal we get C
D E F G H I J K L M N, which is the same as it was before rotation.
After this rotation we see that the tree having D as root is balanced. But if we see the
node F we notice that the height of its left subtree is 2 and height of its right subtree is
4. Thus the balance of F is 2 (or 2 if we take the absolute value) now. Thus the tree
becomes unbalance. So we have to do rotation again to balance the tree. The whole
left subtree of F is shorter so we do the left rotation on the link of F and I (in the tree
in fig 23.9) to bring F down and I up so that the difference of height could be less.
After rotation the tree gets the new form that is shown in the figure below.

N
I
F
D
C

K
G

J
H

L
M

Fig 23.11: Tree (fig 23.10) after rotation

Here we see that the nodes G and H, which were in the left subtree of I, now become
the right subtree of F. We see that the tree with I as root is balanced now. Now we
consider the node N. We have not expanded the right subtree of N. Although we have
not shown but there may be nodes in the right subtree of N. If the difference of
heights of left and right subtree of N is greater than 1 then we have to do rotation on
N node to balance the tree.
Thus we see that there may be such a tree that if we delete a node from it we have to
do rotation at each level of the tree. We notice that we have to do more rotations in
deletion as compared to insertion. In deletion when we delete a node we have to
check the balance at each level up to the root. We do rotation if any node at any level
violates the AVL condition. If nodes at a level do not violate AVL condition then we
do not stop here we check the nodes at each level and go up to the root. We know that
a binary tree has log2 N levels (where N is total number of nodes) thus we have to do
log2 N rotations. We have to identify the required rotations that mean we have to
identify that which one rotation out of the four rotations (i.e. single left rotation,
single right rotation, double right-left rotation and double left-right rotation) we have
to do. We have to identify this at each level.

Page 275 of 519

CS301 Data Structures

___________________________________________________________________
We can summarize the deletion procedure as under.
Delete the node as in binary search tree. We have seen in the discussion of deleting a
node from a BST that we have three cases, which we discussed as follows
Case I: The node to be deleted is the leaf node i.e. it has no right or left child. It is
very simple to delete such node. We make the pointer in the parent node pointing to
this node as NULL. If the memory for the node has been dynamically allocated, we
will release it.
Case II: The node to be deleted has either left child (subtree) or right child (subtree).
In this case we bypass the node in such a way that we find the inorder successor of
this node and then link the parent of the node to be deleted to this successor node.
Thus the node was deleted.
Case III: The node to be deleted has both the left and right children (subtree). This is
the most difficult case. In this case we find the inorder successor of the node. The left
most node of the right subtree will be the inorder successor of it. We put the value of
that inorder successor node into the node to be deleted. After it we delete the inorder
successor node recursively.
In deletion in AVL tree, we delete the node as we delete it in a BST. In third case of
deletion in BST we note that the node deleted will be either a leaf or have just one
subtree (that will be the right subtree as node deleted is the left most subtree so it
cannot have a left subtree). Now we are talking about deletion in an AVL tree. Since
this is an AVL tree so if the deleted node has one subtree that subtree contains only
one node. Why it is so? Think about its reason and answer.
After deleting the node we traverse up the tree from the deleted node checking the
balance of each node at each level up to the root. Thus the deletion in AVL tree is like
the deletion in BST except that here in AVL tree we have to rebalance the tree using
rotations.

Cases of Deletion in AVL Tree

Now lets consider the cases of deletion that will help to identify what rotation will be
applied at what point.
There are 5 cases to consider. Lets go through the cases graphically and determine
what actions are needed to be taken. We will not develop the C++ code for
deleteNode in AVL tree. This is left as an exercise.
Case 1a:
The first case is that the parent of the deleted node had a balance of 0 and the node
was deleted in the parents left subtree.
In the following figure (Fig 23.12) the left portion shows the parent node with a
horizontal line in it. This horizontal line indicates that the left and right subtrees of
this node have the same heights and thus the balance of this node is 0. When we
delete a node from its left subtree then height of its right subtree becomes larger than
the left subtree. The right portion in the figure shows this. We are showing a symbol
instead of balance of node inside the node. The notation (symbol) in the node

Page 276 of 519

CS301 Data Structures

___________________________________________________________________
indicates that the height of left subtree is shorter than the height of right subtree of the
node.

Delete on
Fig 23.12: Deletion Case 1a
this side
Now the action that will be taken in this case is that, change the balance of the parent
node and stop. No further effect on balance of any higher node. There is no need of
rotation in this case. Thus it is the easiest case of deletion.
Lets consider an example to demonstrate the above case.
Consider the tree shown in the following figure i.e. Fig 23.13. This is a perfectly
balanced tree. The root of this tree is 4 and nodes 1, 2 and 3 are in its left subtree. The
nodes 5, 6 and 7 are in the right subtree.

-------------------------------------- 0

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

----- 2

Fig 23.13: Tree under consideration

Consider the node 2. We have shown the balance of this node with a horizontal line,
which indicates that the height of its left subtree is equal to that of its right subtree.
Similarly we have shown the balance of node 4.
Now we remove the node 1 which is the left subtree of node 2. After removing the left
child (subtree) of 2 the height of left subtree of 2 is 0. The height of right subtree of 2
is 1 so the balance of node 2 becomes 1. This is shown in the figure by placing a

Page 277 of 519

CS301 Data Structures

___________________________________________________________________
sign that is down ward to right side, which indicates that height of right subtree is
greater than left subtree. Now we check the node 4. Here the height of left subtree of
it is still 2. The height of its right subtree is also 2. So balance of the node 4 is 0 that is
indicated by the small horizontal line (minus sign) in the figure below. Here we dont
need to change the balance of 4.

0
---------------------------------------

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

4
2

2
-------------------------

Fig 23.14: Tree before and after deleting node 1

Case 1b:
This case is symmetric to case 1a. In this case, the parent of the deleted node had a
balance of 0 and the node was deleted in the parents right subtree. The following
figure shows that the balance of the node was zero as left and right subtree of it have
the same heights.

Delete on
this side
Fig23.15: Deletion Case 1b

After removing the right child the balance of it changes and it becomes 1, as the
height of its left subtree is 1 greater than the height of right subtree. The action
performed in this case is the same as that in case 1a. That is change the balance of the
parent node and stop. No further effect on balance of any higher node. The previous
example can be done for this case only with the change that remove the right child of
node 2 i.e. 3.

Page 278 of 519

CS301 Data Structures

___________________________________________________________________
Case 2a:
This is the case where the parent of the deleted node had a balance of 1 and the node
was deleted in the parents left subtree. This means that the height of left subtree of
the parent node is 1 greater than the height of its right subtree. It is shown in the left
portion of the figure below.

Delete on
this side
Fig 23.16: Deletion Case 2a
Now if we remove a node from the left subtree of this node then the height of left
subtree will decrease by 1 and get equal to the height of right subtree. Thus the
balance of this node will be zero. In this case, the action that we will perform to
balance the tree is that change the balance of the parent node. This deletion of node
may have caused imbalance in higher nodes. So it is advised to continue up to the root
of the tree. We will do rotation wherever the balance of the node violates the AVL
condition.

Page 279 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 24
Reading Material
Data Structures and Algorithm Analysis in C++
4.4

Chapter. 4

Summary

Deletion in AVL Tree

Other Uses of Binary Trees

Deletion in AVL Tree

At the end of last lecture, we were discussing about deleting a node from an AVL
tree. There are five cases to consider while deleting a node of an AVL tree. When a
node is deleted, the tree can become unbalanced. We calculate the balance factor of
each node and perform rotation for unbalanced nodes. But this rotation can prolong to
the root node. In case of insertion, only one nodes balance was adjusted as we saw in
previous lectures but in case of deletion, this process of rotation may expand to the
root node. However, there may also be cases when we delete a node and perform no
or one rotation only.
Now, we will see the five cases of deletion. A side note is that we are not going to
implement these cases in C++ in this lecture, you can do it yourself as an exercise
with the help of the code given inside your text book. In this lecture, the emphasis
will be on the deletion process and what necessary actions we take when a node is
required to be deleted from an AVL tree. Actually, there are two kinds of actions
taken here, one is deletion and the other one is the rotation of the nodes.
Case 1a: The parent of the deleted node had a balance of 0 and a node was deleted in
the parents left subtree.

Delete on
this side

Fig 24.1
In the left tree in the Fig 24.1, the horizontal line inside the tree node indicates that
the balance is 0, the right and left subtrees of the node are of equal levels. Now, when
a node is deleted from the left subtree of this node, this may reduce by one level and
cause the balance of the right subtree of the node to increase by 1 relatively. The
Page 280 of 519

CS301 Data Structures

___________________________________________________________________
balance of the node in favor of the right subtree is shown by a triangular knob tilted
towards right. Now, the action required in this case to make the tree balanced again is:
Change the balance of the parent node and stop. There is no further effect on
balance of any higher node.
In this case, the balance of the tree is changed from 0 to 1, which is within the
defined limits of AVL tree, therefore, no rotation is performed in this case.
Below is a tree in which the height of the left subtree does not change after deleting
one node from it.
0

Fig 24.2

The node 4 is the root node, nodes 2 and 6 are on level 1 and nodes 1, 3, 5, 7 are
shown on level 2. Now, if we delete the node 1, the balance of the node 2 is tilted
towards right, it is 1. The balance of the root node 4 is unchanged as there is no
change in the number of levels within right and left subtrees of it. Similarly, there is
no change in the balances of other nodes. So we dont need to perform any rotation
operation in this case.
Lets see the second case.
Case 1b: the parent of the deleted node had a balance of 0 and the node was deleted
in the parents right subtree.

Delete on
this side
Fig 24.3
On the left of Fig 24.3, the tree is within the balance limits of AVL. After a node is
deleted from the right subtree of it. The balance of the tree is tilted towards left as
shown in the right tree show in the Fig 24.3. Now, we see what action will be
required to make the tree balanced again.

Page 281 of 519

CS301 Data Structures

___________________________________________________________________
Change the balance of the parent node and stop. No further effect on balance of any
higher node (same as 1a).
So in this case also, we dont need to perform rotation as the tree is still an AVL (as
we saw in the Case 1a). It is important to note that in both of the cases above, the
balance of the parent node was 0. Now, we will see the cases when the balance of the
parent node is not 0 previously.
Case 2a: The parent of the deleted node had a balance of 1 and the node was deleted
in the parents left subtree.

Delete on
this side
Fig 24.4
In the Fig 24.4 above,
the tree on the left contains the balance factor as 1, which
means that the left subtree of the parent node is one level more than the number of
levels in the right subtree of it. When we delete one node from the left subtree of the
node, the height of the left subtree is changed and the balance becomes 0 as shown in
the right side tree of Fig 24.4. But it is very important understand that this change of
levels may cause the change of balance of higher nodes in the tree i.e.
Change the balance of the parent node. May have caused imbalance in higher nodes
so continue up the tree.
So in order to ensure that the upper nodes are balanced, we calculate their balance
factors for all nodes in higher levels and rotate them when required.
Case 2b: The parent of the deleted node had a balance of -1 and the node was deleted
in the parents right subtree.
Similar to the Case 2a, we will do the following action:
Change the balance of the parent node. May have caused imbalance in higher nodes
so continue up the tree.

Page 282 of 519

CS301 Data Structures

___________________________________________________________________
Now, we see another case.
Case 3a:The parent had balance of -1 and the node was deleted in the parents left
subtree, right subtree was balanced.

Fig 24.5
As shown in the left tree in Fig 24.5, the node A is tilted towards right but the right
subtree of A (node B above) is balanced. The deleted node lies in the left subtree of
the node A. After deletion, the height of the left subtree is changed to h-1 as depicted
in the right tree of above figure. In this situation, we will do the following action:
Perform single rotation, adjust balance. No effect on balance of higher nodes so stop
here.

Single rotate

i
Node A has become the left subtree of node B and node 2 left subtree of node B has
become the right subtree of node A. The balance of node B is tiled towards left and
balance of node A is tilted towards right but somehow, both are within AVL limits.
Hence, after a single rotation, the balance of the tree is restored in this case.

Page 283 of 519

CS301 Data Structures

___________________________________________________________________
Case 4a: Parent had balance of -1 and the node was deleted in the parents left
subtree, right subtree was unbalanced.

Fig 24.7
In the last case 3a, the right subtree of node A was balanced. But in this case, as
shown in the figure above, the node C is tilted towards left. The node to be deleted
lies in the left subtree of node A. After deleting the node the height of the left subtree
of node A has become h-1. The balance of the node A is shown tilted towards right by
showing two triangular knobs inside node A. So what is the action here.
Double rotation at B. May have affected the balance of higher nodes, so continue up
the tree.
double
rotate

Fig 24.8
Node A, which was the root node previously, has become the left child of the new
root node B. Node C, which was the right child of the root node C has now become
the right child of the new root node B.

Page 284 of 519

CS301 Data Structures

___________________________________________________________________
Case 5a: The parent had balance of -1 and the node was deleted in the parents left
subtree, right subtree was unbalanced.

Fig 24.9
In the figure above, the right tree of the node B has a height of h-1 while the right
subtree is of height h. When we remove a node from the left subtree of node A, the
new tree is shown on the right side of Fig 24.9. The subtree 1 has height h-1 now,
while subtrees 2 and 3 have the same heights. So the action we will do in this case is:
Single rotation at B. May have effected the balance of higher nodes, so continue up
the tree. single
rotate

Fig 24.10

These were the five cases of deletion of a node from an AVL tree. Until now, we are
trying to understand the concept using the figures. You might have noticed the phrase
continue up the tree in the actions above. How will we do it? One way is to maintain
the pointer of the parent node inside each node. But often the easiest method when we
go in downward direction and then upward is recursion. In recursion, the work to be
done later is pushed on to the stack and we keep on moving forward until at a certain
point we back track and do remaining work present in the stack. We delete a node
when we reach at the desired location and then while traversing back, do the rotation
operation on our way to the root node.
Symmetrical to case 2b, we may also have cases 3b, 4b and 5b. This should not be a
problem in doing it yourself.

Page 285 of 519

CS301 Data Structures

___________________________________________________________________

Other Uses of Binary Trees

A characteristic of binary trees is that the values inside nodes on the left of a node are
smaller than the value in the node. And the values inside the nodes on the right of a
node are greater than the value in the node. This is the way a binary tree is
constructed.
Whatever is the size of the tree, the search is performed after traversing upto log2n
levels maximum.
We have observed that the binary tree becomes a linked list and it can become
shallow. The AVL conditions came into picture to control the height balance of a
binary tree. While searching in an AVL tree, in the worst case scenario we have to
search 1.44 log2n levels. For searches, binary and AVL trees are the most efficient but
we do have some other kinds of trees that will be studied later.
Lets see what could be some other uses of binary trees, we start our discussion with
Expression Trees.
Expression Trees
Expression trees, the more general parse trees and abstract syntax trees are significant
components of compilers.
We already know about compilers that whenever we write our program or code in
some computer language like C++ or Java. These programs are compiled into
assembly language or byte code (in case of Java). This assembly language code is
translated converted into machine language and an executable file is made by another
program called the assembler.
By the way, if you have seen your syllabus, you might have seen that there is a
dedicated subject on compilers. We study in detail about compilers in that course. For
this course, we will see expression or parse trees.
We will take some examples of expression trees and we will not delve into much
depth of it rather that would be an introduction to expression trees.
(a+b*c)+((d*e+f)*g)
+

Fig 24.11

Page 286 of 519

CS301 Data Structures

___________________________________________________________________
You can see the infix expression above (a + b * c) + ( (d * e + f) * g), it is represented
in the tree form also.
You can see from bottom of the tree that the nodes b and c in the nodes are present at
the same level and their parent node is multiplication (*) symbol. From the expression
also, we see that the b and c are being multiplied. The parent node of a is + and right
subtree of + is b*c. You might have understood already that this subtree is depicting
a+b*c. On the right side, node d and e are connected to the parent *. Symbol + is the
parent of * and node f. The expression of the subtree at node + is d*e+f. The parent of
node + is * node, its right subtree is g. So expression of the subtree at this node * is
(d*e+f)*g). The root node of the tree is +.
These expression trees are useful in compilers and in spreadsheets also, they are
sometimes called parse trees.
Parse Tree in Compilers
See the expression tree of expression A := A + B * C below. We are adding B and C,
adding the resultant in A and then finally assigning the resultant to A.
A := A + B * C

<assign
<id
A

<expr

<expr

<term

<term
<factor>
<id>
Expression grammar
<assign>
<id> := <expr>
<id>
A|B|C
<expr>
<expr> + <term> | <term>
<term>
<term> * <factor> | <factor>

<term

<factor>

<factor>

<id

<id

B
Fig 24.12

The root node in the parse tree shown above is <assign>.

The assignment statement (<assign>) has three parts. On the left of it, there is always
an identifier (single or an array reference) called l-value. The l-value shown in the
tree above is <id> and the identifier is A in the expression above. The second part of
assignment statement is assignment operator (= or :=). One the right of assignment
operator lies the third part of assignment statement, which is an expression. In the
expression A := A + B * C above , the expression after assignment operator is A + B *
C. In the tree, it is represented by the node <expr>. The node <expr> has three
subnodes: <expr>, + and <term>. <expr>s further left subtree is <expr>, <term>,
<factor>, <id> and then finally is B. The right subchild <term> has further subnodes

Page 287 of 519

CS301 Data Structures

___________________________________________________________________
as <term>, * and <factor>. <factor> has <id> as subchild and <id> has C.
Note the nodes in gray shade in the tree above form A = A + B * C.
Compiler creates these parse trees. We will see how to make these trees, when we will
parse any language tree like C++. Parsing mean to read and then extract the required
structure. A compiler parses a computer language like C++ to form parse trees.
Similarly, when we do speech recognition. Each sentence is broken down into a
certain structure in form of a tree. Hand writing recognition also involves this. The
tablet PCs these days has lot of implementation of parse trees.
Parse Tree for an SQL Query
Lets see another example of parse trees inside databases. The parse trees are used in
query processing. The queries are questions to the databases to see a particular kind of
data. Consider you have a database for a video store that contains data of movies and
artists etc. You are querying that database to find the titles of movies with stars born
in 1960. The language used to query from the database is called SQL (Structured
Query Language), this language will be dealt in depth in the databases course. The
tables lying in this movies database are:
StarsIn(title, year, starName)
MovieStar(name, address, gender, birthdate)
The following SQL query is used to retrieve information from this database:
SELECT title
FROM StarsIn, MovieStar
WHERE starName = name AND birthdate LIKE %1960 ;
This query is asking to retrieve the titles of those movies from StarsIn and MovieStar
tables where the birthdate of an actor is 1960. We see in query processing a tree is
formed as shown in the figure below:
< Query >

SELECT

<SelList>
<Attribute>
title

<FromList>

FROM
<RelName>

WHERE

<FromList>

<Condition>
AND

<RelName>

StarsIn

MovieStar
Condition

Condition
<Attribute>

<Attribute>

<Attribute>

LIKE

<Pattern>

Fig 24.13
setName

name

birthdate

%1960

The root node is Query. This node is further divided into SELECT, <SelList>,
FROM, <FromList>, WHERE and <Condition> subnodes. <SelList> will be an
Attribute and finally a title is reached. Observe the tree figure above, how the tree is

Page 288 of 519

CS301 Data Structures

___________________________________________________________________
expanded when we go in the downward direction. When the database engine does the
query process, it makes these trees. The database engine also performs query
optimization using these trees.
Compiler Optmization
Lets see another expression tree here:
Common subexpression:
(f+d*e) + ((d*e+f)*g)

*
d

+
e

*
d

g
f

Fig 24.14

The root node is +, left subtree is capturing the f+d*e expression while the right
subtree is capturing (d*e+f)*g.
Normally compilers has intelligence to look at the common parts inside parse trees.
For example in the tree above, the expressions f+d*e and d*e+f are same basically.
These common subtrees are called common subexpressions. To gain efficiency,
instead of calculating the common subexpressions again, the compilers calculates
them once and use them at other places. The part of the compiler that is responsible to
do this kind of optimization is called optimizer.
See the figure below, the optimizer (part of compiler) will create the following graph
while performing the optimization. Because both subtrees were equivalent, it has
taken out one subtree and connected the link from node * to node +.

Page 289 of 519

CS301 Data Structures

___________________________________________________________________
(Common Subexpression:
(f+d*e) + ((d*e+f)*g)

g
Graph!

e
Fig 24.15

This figure is not a tree now because it has two or more different paths to reach a
node. Therefore, this has become a graph. The new connection is containing a
directed edge, which is there in graphs.
Optimizer uses the expressions trees and converts them to graphs for efficiency
purposes. You read out from your book, how the expression trees are formed, what
are the different methods of creating them.

Page 290 of 519

CS301 Data Structures

___________________________________________________________________

Lecture No. 25
Reading Material
Data Structures and Algorithm Analysis in C++
4.2.2, 10.1.2

Chapter. 4, 10

Summary

Expression tree
Huffman Encoding

Expression tree
We discussed the concept of expression trees in detail in the previous lecture. Trees
are used in many other ways in the computer science. Compilers and database are two
major examples in this regard. In case of compilers, when the languages are translated
into machine language, tree-like structures are used. We have also seen an example of
expression tree comprising the mathematical expression. Lets have more discussion
on the expression trees. We will see what are the benefits of expression trees and how
can we build an expression tree. Following is the figure of an expression tree.
+

*
d

In the above tree, the expression on the left side is a + b * c while on the right side,
we have d * e + f * g. If you look at the figure, it becomes evident that the inner nodes
contain operators while leaf nodes have operands. We know that there are two types
of nodes in the tree i.e. inner nodes and leaf nodes. The leaf nodes are such nodes
which have left and right subtrees as null. You will find these at the bottom level of
the tree. The leaf nodes are connected with the inner nodes. So in trees, we have some
inner nodes and some leaf nodes.
In the above diagram, all the inner nodes (the nodes which have either left or right
child or both) have operators. In this case, we have + or * as operators. Whereas leaf
Page 291 of 519

CS301 Data Structures

___________________________________________________________________
nodes contain operands only i.e. a, b, c, d, e, f, g. This tree is binary as the operators
are binary. We have discussed the evaluation of postfix and infix expressions and
have seen that the binary operators need two operands. In the infix expressions, one
operand is on the left side of the operator and the other is on the right side. Suppose, if
we have + operator, it will be written as 2 + 4. However, in case of multiplication, we
will write as 5*6. We may have unary operators like negation (-) or in Boolean
expression we have NOT. In this example, there are all the binary operators.
Therefore, this tree is a binary tree. This is not the Binary Search Tree. In BST, the
values on the left side of the nodes are smaller and the values on the right side are
greater than the node. Therefore, this is not a BST. Here we have an expression tree
with no sorting process involved.
This is not necessary that expression tree is always binary tree. Suppose we have a
unary operator like negation. In this case, we have a node which has (-) in it and there
is only one leaf node under it. It means just negate that operand.
Lets talk about the traversal of the expression tree. The inorder traversal may be
executed here.
+

*
d

Inorder traversal yields: a+b*c+d*e+f*g

We use the inorder routine and give the root of this tree. The inorder traversal will be
a+b*c+d*e+f*g. You might have noted that there is no parenthesis. In such
expressions when there is addition and multiplication together, we have to decide
which operation should be performed first. At that time, we have talked about the
operator precedence. While converting infix expression into postfix expression, we
have written a routine, which tells about the precedence of the operators like
multiplication has higher precedence than the addition. In case of the expression 2 + 3
* 4, we first evaluate 3 * 4 before adding 2 to the result. We are used to solve such
expressions and know how to evaluate such expressions. But in the computers, we
have to set the precedence in our functions.
We have an expression tree and perform inorder traversal on it. We get the infix form
of the expression with the inorder traversal but without parenthesis. To have the

Page 292 of 519

CS301 Data Structures

___________________________________________________________________
parenthesis also in the expressions, we will have to do some extra work.
Here is the code of the inorder routine which puts parenthesis in the expression.
/* inorder traversal routine using the parenthesis */
void inorder(TreeNode<int>* treeNode)
{
if( treeNode != NULL ){
cout << "(";
inorder(treeNode->getLeft());
cout << ")";
cout << *(treeNode->getInfo());
cout << "(";
inorder(treeNode->getRight());
cout << ")";
}
}
This is the same inorder routine used by us earlier. It takes the root of the tree to be
traversed. First of all, we check that the root node is not null. In the previous routine
after the check, we have a recursive call to inorder passing it the left node, print the
info and then call the inorder for the right node. Here we have included parenthesis
using the cout statements. We print out the opening parenthesis (before the recursive
call to inorder. After this, we close the parenthesis. Then we print the info of the node
and again have opening parenthesis and recursive call to inorder with the right node
before having closing parenthesis in the end. You must have understood that we are
using the parenthesis in a special order. We want to put the opening parenthesis
before the start of the expression or sub expression of the left node. Then we close the
parenthesis. So inside the parenthesis, there is a sub expression.
On executing this inorder routine, we have the expression as ( a + ( b * c )) + ((( d * e
) + f ) * g ).

Page 293 of 519

CS301 Data Structures

___________________________________________________________________
+

*
c

*
d

Inorder: ( a + ( b * c )) + ((( d * e ) + f ) * g )
We put an opening parenthesis and start from the root and reach at the node a. After
reaching at plus (+), we have a recursive call for the subtree *. Before this recursive
call, there is an opening parenthesis. After the call, we have a closing parenthesis.
Therefore the expression becomes as (a + ( b * c )). Similarly we recursively call the
right node of the tree. Whenever, we have a recursive call, there is an opening
parenthesis. When the call ends, we have a closing parenthesis. As a result, we have
an expression with parenthesis, which saves a programmer from any problem of
precedence now.
Here we have used the inorder traversal. If we traverse this tree using the postorder
mode, then what expression we will have? As a result of postorder traversal, there
will be postorder expression.
+

*
d

Postorder traversal: a b c * + d e * f + g * +
This is the same tree as seen by us earlier. Here we are performing postorder traversal.
In the postorder, we print left, right and then the parent. At first, we will print a.

Page 294 of 519

CS301 Data Structures

___________________________________________________________________
Instead of printing +, we will go for b and print it. This way, we will get the postorder
traversal of this tree and the postfix expression of the left side is a b c * + while on
the right side, the postfix expression is d e * f + g * +. The complete postfix
expression is a b c * + d e * f + g * +. The expression undergoes an alteration with
the change in the traversal order. If we have some expression tree, there may be the
infix, prefix and postfix expression just by traversing the same tree. Also note that in
the postfix form, we do not need parenthesis.
Lets see how we can build this tree. We have some mathematical expressions while
having binary operators. We want to develop an algorithm to convert postfix
expression into an expression tree. This means that the expression should be in the
postfix form. In the start of this course, we have seen how to covert an infix
expression into postfix expression. Suppose someone is using a spreadsheet program
and typed a mathematical expression in the infix form. We will first convert the infix
expression into the postfix expression before building expression tree with this postfix
expression.
We already have an expression to convert an infix expression into postfix. So we get
the postfix expression. In the next step, we will read a symbol from the postfix
expression. If the symbol is an operand, put it in a tree node and push it on the stack.
In the postfix expression, we have either operators or operands. We will start reading
the expression from left to right. If the symbol is operand, make a tree node and push
it on the stack. How can we make a tree node? Try to memorize the treeNode class.
We pass it some data and it returns a treeNode object. We insert it into the tree. A
programmer can also use the insert routine to create a tree node and put it in the tree.
Here we will create a tree node of the operand and push it on the stack. We have been
using templates in the stack examples. We have used different data types for stacks
like numbers, characters etc. Now we are pushing treeNode on the stack. With the
help of templates, any kind of data can be pushed on the stack. Here the data type of
the stack will be treeNode. We will push and pop elements of type treeNode on the
stack. We will use the same stack routines.
If symbol is an operator, pop two trees from the stack, form a new tree with operator
as the root and T1 and T2 as left and right subtrees and push this tree on the stack. We
are pushing operands on the stacks. After getting an operator, we will pop two
operands from the stack. As our operators are binary, so it will be advisable to pop
two operands. Now we will link these two nodes with a parent node. Thus, we have
the binary operator in the parent node.
Lets see an example to understand it. We have a postfix expression as a b + c d e + *
*. If you are asked to evaluate it, it can be done with the help of old routine. Here we
want to build an expression tree with this expression. Suppose that we have an empty
stack. We are not concerned with the internal implementation of stack. It may be an
array or link list.
First of all, we have the symbol a which is an operand. We made a tree node and push
it on the stack. The next symbol is b. We made a tree node and pushed it on the stack.
In the below diagram, stack is shown.

Page 295 of 519

CS301 Data Structures

___________________________________________________________________

Our stack is growing from left to right. The top is moving towards right. Now we
have two nodes in the stack. Go back and read the expression, the next symbol is +
which is an operator. When we have an operator, then according to the algorithm, two
operands are popped. Therefore we pop two operands from the stack i.e. a and b. We
made a tree node of +. Please note that we are making tree nodes of operands as well
as operators. We made the + node parent of the nodes a and b. The left link of the
node + is pointing to a while right link is pointing to b. We push the + node in the
stack.

ab+cde+**

+
a

If symbol is an operator, pop two trees from the stack, form a new tree with operator
as the root and T1 and T2 as left and right subtrees and push this tree on the stack.

Actually, we push this subtree in the stack. Next three symbols are c, d, and e. We
made three nodes of these and push these on the stack.

Page 296 of 519

CS301 Data Structures

___________________________________________________________________
ab+cde+**

Next we have an operator symbol as +. We popped two elements i.e. d and e and
linked the + node with d and e before pushing it on the stack. Now we have three
nodes in the stack, first + node under which there are a and b. The second node is c
while the third node is again + node with d and e as left and right nodes.

ab+cde+**

The next symbol is * which is multiplication operator. We popped two nodes i.e. a
subtree of + node having d and e as child nodes and the c node. We made a node of *
and linked it with the two popped nodes. The node c is on the left side of the * node
and the node + with subtree is on the right side.

Page 297 of 519

CS301 Data Structures

___________________________________________________________________
ab+cde+**

c
+

The last symbol is the * which is an operator. The final shape of the stack is as under:
ab+cde+**

c
+

In the above figure, there is a complete expression tree. Now try to traverse this tree
in the inorder. We will get the infix form which is a + b * c * d + e. We dont have
parenthesis here but can put these as discussed earlier.

Page 298 of 519

CS301 Data Structures

___________________________________________________________________
This is the way to build an expression tree. We have used the algorithm to convert the
infix form into postfix form. We have also used stack data structure. With the help of
templates, we can insert any type of data in the stack. We have used the expression
tree algorithm and very easily built the expression tree.
In the computer science, trees like structures are used very often especially in
compilers and processing of different languages.

Huffman Encoding
There are other uses of binary trees. One of these is in the compression of data. This is
known as Huffman Encoding. Data compression plays a significant role in computer
networks. To transmit data to its destination faster, it is necessary to either increase
the data rate of the transmission media or simply send less data. The data compression
is used in computer networks. To make the computer networks faster, we have two
options i.e. one is to somehow increase the data rate of transmission or somehow send
the less data. But it does not mean that less information should be sent or transmitted.
Information must be complete at any cost.
Suppose you want to send some data to some other computer. We usually compress
the file (using winzip) before sending. The receiver of the file decompresses the data
before making its use. The other way is to increase the bandwidth. We may want to
use the fiber cables or replace the slow modem with a faster one to increase the
network speed. This way, we try to change the media of transmission to make the
network faster. Now changing the media is the field of electrical or communication
engineers. Nowadays, fiber optics is used to increase the transmission rate of data.
With the help of compression utilities, we can compress up to 70% of the data. How
can we compress our file without losing the information? Suppose our file is of size 1
Mb and after compression the size will be just 300Kb. If it takes ten minutes to
transmit the 1 Mb data, the compressed file will take 3 minutes for its transmission.
You have also used the gif images, jpeg images and mpeg movie files. All of these
standards are used to compress the data to reduce their transmission time.
Compression methods are used for text, images, voice and other types of data.
We will not cover all of these compression algorithms here. You will study about
algorithms and compression in the course of algorithm. Here, we will discuss a
special compression algorithm that uses binary tree for this purpose. This is very
simple and efficient method. This is also used in jpg standard. We use modems to
connect to the internet. The modems perform the live data compression. When data
comes to the modem, it compresses it and sends to other modem. At different points,
compression is automatically performed. Lets discuss Huffman Encoding algorithm.
Huffman code is method for the compression of standard text documents. It makes
use of a binary tree to develop codes of varying lengths for the letters used in the
original message. Huffman code is also part of the JPEG image compression scheme.
The algorithm was introduced by David Huffman in 1952 as part of a course
assignment at MIT.
Now we will see how Huffman Encoding make use of the binary tree. We will take a
Page 299 of 519

CS301 Data Structures

___________________________________________________________________
simple example to understand it. The example is to encode the 33-character phrase:
"traversing threaded binary trees"
In the phrase we have four words including spaces. There is a new line character in
the end. The total characters in the phrase are 33 including the space. We have to send
this phrase to some other computer. You know that binary codes are used for
alphabets and other language characters. For English alphabets, we use ASCII codes.
It normally consists of eight bits. We can represent lower case alphabets, upper case
alphabets, 0,1,9 and special symbols like $, ! etc while using ASCII codes.
Internally, it consists of eight bits. If you have eight bits, how many different patterns
you can be have? You can have 256 different patterns. In English, you have 26 lower
case and 26 upper case alphabets. If you have seen the ASCII table, there are some
printable characters and some unprintable characters in it. There are some graphic
characters also in the ASCII table.
In our example, we have 33 characters. Of these, 29 characters are alphabets, 3 spaces
and one new line character. The ASCII code for space is 32 and ASCII code for new
line character is 10. The ASCII value for a is 95 and the value of A is 65. How many
bits we need to send 33 characters? As every character is of 8 bits, therefore for 33
characters, we need 33 * 8 = 264. But the use of Huffman algorithm can help send the
same message with only 116 bits. So we can save around 40% using the Huffman
algorithm.
Lets discuss how the Huffman algorithm works. The algorithm is as:

List all the letters used, including the "space" character, along with the
frequency with which they occur in the message.
Consider each of these (character, frequency) pairs to be nodes; they are
actually leaf nodes, as we will see.
Pick the two nodes with the lowest frequency. If there is a tie, pick randomly
amongst those with equal frequencies.
Make a new node out of these two, and turn two nodes into its children.
This new node is assigned the sum of the frequencies of its children.
Continue the process of combining the two nodes of lowest frequency till the
time only one node, the root is left.

Lets apply this algorithm on our sample phrase. In the table below, we have all the
characters used in the phrase and their frequencies.
Original text:
traversing threaded binary trees
size:
33
newline)

characters

Letters :

Frequency

NL

(space

and

Page 300 of 519

CS301 Data Structures

___________________________________________________________________
SP
a
b
d
e
g
h

:
:
:
:
:
:
:

3
3
1
2
5
1
1

The new line occurs only once and is represented in the above table as NL. Then
white space occurs three times. We counted the alphabets that occur in different
words. The letter a occurs three times, letter b occurs just once and so on.
Now we will make tree with these alphabets and their frequencies.
2 is equal to sum of
the frequencies of the

a
3

e
5

t
3
d
2

i
2

n
2

r
5
2

s
2
N
L

b
1

g
1

SP
3
h
1

v
1

y
1

CS301 Data Structures

___________________________________________________________________

CS301 Data Structures

___________________________________________________________________

Lecture No. 26
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 4
4.4.2

Summary

Hoffman Encoding
Mathematical Properties of Binary Trees

Huffman Encoding
We will continue our discussion on the Huffman encoding in this lecture. In the
previous lecture, we talked about the situation where the data structure binary tree
was built. Huffman encoding is used in data compression. Compression technique is
employed while transferring the data. Suppose there is a word-document (text file)
that we want to send on the network. If the file is, say, of one MB, there will be a lot
of time required to send this file. However, in case of reduction of size by half
through compression, the network transmission time also get halved. After this
example, it will be quite easy to understand the Hoffman encoding to compress a text
file.
We know that Huffman code is a method for the compression of standard text
documents. It makes use of a binary tree to develop codes of varying lengths for the
letters used in the original message. Huffman code is also a part of the JPEG image
compression scheme. David Huffman introduced this algorithm in the year 1952 as
part of a course assignment at MIT.
In the previous lecture, we had started discussing a simple example to understand
Huffman encoding. In that example, we were encoding the 32-character phrase:
"traversing threaded binary trees". If this phrase were sent as a message in a network
using standard 8-bit ASCII codes, we would have to send 8*32= 256 bits. However,
the Huffman algorithm can help cut down the size of the message to 116 bits.
In the Huffman encoding, following steps are involved:
1. List all the letters used, including the "space" character, along with the
frequency with which they occur in the message.
2. Consider each of these (character, frequency) pairs as nodes; these are actually
leaf nodes, as we will see later.
3. Pick two nodes with the lowest frequency. If there is a tie, pick randomly
amongst those with equal frequencies
4. Make a new node out of these two and develop two nodes as its children.
5. This new node is assigned the sum of the frequencies of its children.
6. Continue the process of combining the two nodes of lowest frequency till the
time, only one node, the root, is left.
In the first step, we make a list of all letters (characters) including space and end line
character and find out the number of occurrences of each letter/character. For example

Page 306 of 519

CS301 Data Structures

___________________________________________________________________
we ascertain how many times the letter a is found in the file and how many times b
occurs and so on. Thus we find the number of occurrences (i.e. frequency) of each
letter in the text file.
In the step 2, we consider the pair (i.e. letter and its frequency) as a node. We will
consider these as leaf nodes. Afterwards, we pick two nodes with the lowest
frequency in the list. If there are more than one pairs of same frequency, we will
choose a pair randomly amongst those with equal frequencies.
Suppose, in a file, the letter a occurs 50 times and b and c five times each. Here,
b and c have the lowest frequency. We will take these two letters as leaf nodes and
build the tree from these ones. As fourth step states, we make a new node as the
parent of these two nodes. The b and c are its children. In the fifth step, we assign
to this new node the frequency equal to the sum of the frequencies of its children.
Thus a three-node tree comes into existence. This is shown in the following figure.
10

a, 5

b, 5

Fig 26.1:
We continue this process of combining the two nodes of lowest frequency till the
time, only one node i.e. the root is left.
Now we come back to our example. In this example, there is a text string as written
below.
traversing threaded binary trees
The size of this character string is 33 (it includes 3 space characters and one new line
character). In the first step, we perform the counting of different characters in the
string manually. We do not assign a fake or zero frequency to a letter that is not
present in the string. A programmer may be concerned only with the characters/letters
that are present in the text. We see that the letters and their frequencies in the above
text is as given below.
Character frequency character
NL
1
I
SP
3
n
A
3
r
B
1
s
D
2
t
E
5
v
G
1
y
H
1
Table 1: Frequency table

frequency
2
2
5
2
3
3
1

In the second step, we make nodes of these pairs of letters and frequencies. The

Page 307 of 519

CS301 Data Structures

___________________________________________________________________
following figure (fig 26.2) depicts the letters as nodes. We have written the frequency
of each letter with the node. The nodes have been categorized with respect to the
frequencies for simplicity. We are going to build the tree from downside i.e. from the
lowest frequency.
Now according to third step, two nodes of lowest frequency are picked up. We see
that nodes NL, b, g, h, v and y have the frequency 1. We randomly choose the nodes v
and y. now, as the fourth step, we make a new node and join the leaf nodes v and y to
it as its children. We assign the frequency to this new (parent) node equal to the sum
of the frequencies of its children i.e. v and y. Thus in the fifth step; the frequency of
this new node is 2. We have written no letter in this node as shown in the figure
below.

2 is equal to sum
of the frequencies of
the two children
d

a
3

e
5

t
3
d
2

i
2

n
2

r
5

s
2

SP
3

2
N
L

b
1

g
1

h
1

v
1

y
1

Fig 26.2: Nodes with their respective frequencies

Now we continue this process with other nodes. Now we join the nodes g and h as
children of a new node. The frequency of this node is 2 i.e. the sum of frequencies of
g and h. After this, we join the nodes NL and b. This also makes a new node of
frequency 2. Thus the nodes having frequency 1 have joined to the respective parent
nodes. This process is shown in the following figure (Fig 26.3).

Page 308 of 519

CS301 Data Structures

___________________________________________________________________

e
5

t
3

a
3

i
2

d
2

n
2

s
2

2
N
L

r
5

2
b
1

g
1

2
h
1

v
1

SP
3
y
1

Fig 26.3: joining leaf nodes to create new nodes

Now we come to the nodes with a frequency 2. Here we

join the pair of nodes d and i and also the pair n and s.
Resultantly, the two nodes coming into being after joining
have the frequency 4. This frequency is the sum of the
frequencies of their children. Now, we will bring the nodes,
a and t together. The parent node of a and t has the
frequency 6 i.e. the sum of a and t. These new nodes are
shown in the following figure.

6
a
3

t
3
d
2

4
i
2

n
2

e
5

4
s
2

2
N
L

r
5

2
b
1

g
1

SP
3

2
h
1

v
1

y
1

Fig 26.4: joining different nodes to create new nodes

Now we consider these new nodes as inner nodes and

build the tree upward towards the root. Now we take the
node SP and join it with the node that is the parent of v
Page 309 of 519

CS301 Data Structures

___________________________________________________________________

and y. The resultant node has frequency 5 as the

frequencies of its children are 2 and 5 respectively. Now
we join these nodes of higher frequencies. In other words,
the node r is joined with the newly created node of
frequency 5 that is on the left side of node r in the figure
26.5. Thus a new node of frequency 10 is created. We join
the node e and the node of frequency 4 on its right side.
Resultantly, a new node of frequency 9 comes into
existence. Then we join the nodes having frequency 4 and
create a new node of frequency 8. The following figure
shows the nodes created so far.

6
a
3

t
3
d
2

4
i
2

n
2

e
5

4
s
2

2
N
L

10

r
5

2
b
1

g
1

2
h
1

v
1

SP
3
y
1

Fig 26.5: joining of nodes to build a tree

Now we will join the nodes of frequency 6 and 8 to create the node of frequency 14
and join the nodes of frequency of 9 and 10 to develop a new node of frequency of 19.
At the end, we make the root node with the frequency 33 and it comprises nodes of
frequency 14 and 19. Thus the tree is completed and shown in the following figure.

Page 310 of 519

CS301 Data Structures

___________________________________________________________________

33

19

14

6
4

t
3

a
3

d
2

4
i
2

n
2

e
5

4
s
2

2
N
L

10

r
5

2
b
1

g
1

2
h
1

v
1

SP
3
y
1

Fig 26.6: Hoffman encoding tree

Now we will perform other steps of Hoffman encoding and develop characterencoding scheme needed for data compression.
To go ahead, we have to do the following steps.

Start at the root. Assign 0 to left branch and 1 to the right branch.
Repeat the process down the left and right subtrees.
To get the code for a character, traverse the tree from the root to the character
leaf node and read off the 0 and 1 along the path.

We start from the root node of the tree and assign the value 0 to the left branch and 1

Page 311 of 519

CS301 Data Structures

___________________________________________________________________
to the right branch. Afterwards, we repeat this value assigning at nodes in left and
right subtrees. Thus all the paths have a value 0 or 1 as shown in the following figure.
We will develop the code with the help of these path values.

33
1

19

14
1

a
3

t
3

d
2

i
2

10

n
2

s
2

0
0

N
L

e
5

r
5

b
1

g
1

5
0

h
1

2
0

v
1

y
1

SP
3

Fig 26.7: Hoffman encoding tree with path values

In the last step, we get the code for the characters. To get
the code for a character, there is need of traversing the
tree from the root to the character leaf node and read off
the 0 and 1 along the path. We start from the root and go
to the letter of leaf node following the edges. 0 and 1 are
written down in the order in which they come in this
traversal to leaf node. For example, we want the code of
Page 312 of 519

CS301 Data Structures

___________________________________________________________________

letter d. To reach d from the root; we have to go through

the nodes of frequency 14. This path has value 0. Here, 0
will be the first digit in the code of d. From 14, we go to
node of frequency 8. This link (from 14 to 8) has value 1.
Thus the second digit in code is 1. From node of
frequency 8, we go to node of frequency 4 to its left side.
This side has value 0, meaning that 0 is the third digit of
code. From 4, we finally go to d and in this link we get the
value 0. Thus we see that to reach at d from the root, we
have gone through the branches 0, 1, 0 and 0. Thus, the
code of letter d is 0100. Similarly the code of i is 0101.
The same way, we find the code for each letter in the tree.
The following table shows the letters and their
correspondent codes.
characte code
character
r
NL
10000
i
SP
1111
n
a
000
r
b
10001
s
d
0100
t
e
101
v
g
10010
y
h
10011
Table 2: Hoffman code table

code
0101
0110
110
0111
001
11100
11101

We know that every character is stored in the computer in binary format. Each
character has a code, called ASCII code. The ASCII code of a character consists of
ones and zeros i.e. it is in binary form. ASCII code is of eight bits. Normally, we
remember the decimal value of the characters. For example, letter A has decimal
value 65. We can easily convert the decimal value into binary one in bit pattern. We
need not to remember these values. ASCII value of any character can be found from
the ASCII table. The ASCII code of each character is represented in eight bits with
different bit patterns.
Here in the example, we assign a code of our own (i.e. Hoffman code) to the letters
that are in the message text. This code also consists of ones and zeros. The above
table shows the characters and their Hoffman code. Now we come back to the
message text from which we developed the tree and assigned codes to the letters in
the text.
Look at the table (Table 2) shown above. Here we notice that the code of letters is of
variable length. We see that letters with higher frequency have shorter code. There are
some codes with a length five that are the codes of NL, b, g, h, v and y. Similarly we
see that the letters SP, d, i, n and s have codes of length four. The codes of the
Page 313 of 519

CS301 Data Structures

___________________________________________________________________
remaining letters have length three. If we look at the frequency table (Table 1) of
these letters, we see that the letters with some higher frequency like a, e, r, and t, have
the code of shorter length, whereas the letters of lower frequency (like NL, b, g, h, v
and y) have codes of larger length.
We see in the table of the Hoffman codes of the letters that there will be need of 5, 4
or in some codes only 3 bits to represent a character, whereas in ASCII code, we need
8 bits for each character. Now we replace the letters in the text message with these
codes. In the code format i.e. in the form of ones and zeros, the message becomes as
under.
Our original message was
traversing threaded binary trees
The encoded form of the message is as under
t

00111000011100101110011101010110100101111001
100111101010000100101010011111000101010110000
110111011111001110101101011110000
We see that there are only ones and zeros in the code of message. Some letters have
been shown with their corresponding code. The first three bits 001 are for letter t.
The next three bits 110 are for letter r. After this the letter a has three bits i.e. 000.
Next to it is the letter v that has five bits. These bits are 11100 (shaded in the
figure). Similarly we have replaced all the letters of the message with their
corresponding Hoffman code. The encoded message is shown above.
Lets compare this Hoffman encoded message with the encoding of message with
ASCII code. The ASCII code encoding means that if we have encoded this message
with 8 bits per character, the total length of message would be 264. As there are 33
characters, so the total length is 33 x 8 = 264. But in the Hoffman encoded message
(written in above table), there are only 120 bits. This number of bits is 54% less than
the number of bits in ASCII encoding form. Thus we can send the message with
almost half of the original length to a receiver.
Here the Hoffman encoding process comes to an end. In this process, we took a
message, did the frequency count of its characters and built a tree from these
frequencies. From this tree, we made codes of letters consisting of ones and zeros
only. Then with the help of these codes, we did the data compression. In this process
we saw that less data is required for the same message.
Now an important thing to be noted is regarding the tree building. We have built the
tree with our choices of nodes with same and different frequencies. Now if we have
chosen the nodes to join in different way, the tree built would be in a different form.
This results in the different Hoffman code for the letters. These will be in ones and
zeros but with different pattern. Thus the encoded message would have different
codes for the letters. Now the question arises how does the receiver come to know
that what code is used for what letter? The answer is very simple that the sender has

Page 314 of 519

CS301 Data Structures

___________________________________________________________________
to tell the receiver that he is using such codes for letters. One way to tackle this
problem is that the sender sends the tree built through the use of frequencies, to the
receiver to decode the message. The receiver keeps a copy of this tree. Afterwards,
when the sender sends a message, the receiver will match it with respect to bit pattern
with the tree to see that what letter the sender has sent. The receiver will find the letter
for the first 2, 3, 4 or what numbers of bits match to a letter that it has in the tree as
the receiver also has a copy of the tree. We know that a tree has a root, inner nodes
and leaf node(s). The leaf node is a node whose left and right links is NULL. An inner
node has a left or right or both children. Now consider that the receiver has the same
tree data structure that the sender used to construct the codes. The sender has sent the
message that we have discussed above. The sender sends the 122 bits encoded
message to the receiver. The receiver will take the first bit of message and being on
the root of the tree, it will decide that on what side this bit will go. If the bit is zero, it
will go to the left of the root. However, if the bit is one, it will go to the right side of
the root before reaching to the child node. The next bit will go to the next level of the
tree to the left or right side depending on zero or one. Thus the traversal will go one
level down on receiving a bit. If we (the receiver) are on a path of the tree where the
leaf node is at level 6, it cannot reach the leaf node unless the receiver receives 6 bits.
We consider the case of letter e whose code was of three bits. In this case, we go to
the first level by first bit and the second bit takes us to the third level. Finally on
reaching the third level with the third bit, we get the leaf node. We know that this
node is letter e as the sender has sent us (as receiver) the whole tree with the
characters in the nodes. Thus after traversing the three bits, the receiver has confirmed
that it has received the letter e. Now on receiving the fourth bit, the receiver will go
back to the root and continue to choose the path i.e. on zero it will go to the left and
on one it will go to right. This way, it will reach the leaf node. The character at that
node will be the next character of the message. The bit pattern that was comprised of
following these links in the tree is extracted from the message. On the next bit, the
receiver again goes to the root node and the previous procedure is repeated to find the
next character. This way, it decodes the whole message.
The compression is very useful technique especially for communication purposes.
Suppose that we have to send a file of one Mb. Here each line in the file can be
compressed to 50 or 60 percent. Thus the file of one MB will be compressed to half
MB and can be sent more easily.
There is one more thing about this tree. When we started to build the tree from leaf
nodes i.e. bottom-up build of tree, we saw that there were choices for us to choose any
two leaf nodes to join. In our example, we chose them at random. The other way to do
it is the priority queue. We have seen the example of bank simulation. There we used
a priority queue for the events. We know that in a priority queue, the elements do not
follow the FIFO (first in first out) rule. But the elements have their position in the
queue with respect to a priority. In the example of bank simulation, in the Event
Queue, we remove the element from the queue that was going to occur first in the
future. We can use priority queue here in such a way that we put the letters in the
queue with respect to their frequencies. We will put and remove letters from the
queue with respect to their frequency. In this priority queue, the character with lowest
frequency is at the start of the queue. If two characters have the same frequency, these
will be one after the other in the queue. The character with the larger frequency will
be in the last of the queue. Now we take two frequencies from the queue and join
them to make a new node. Suppose that the nodes that we joined have frequency 1

Page 315 of 519

CS301 Data Structures

___________________________________________________________________
and 2 respectively. So the frequency of the new node will be 3. We put this new node
in the queue. It takes its position in the queue with respect to its frequency as we are
using the priority queue. It is evident in procedure that we proceed to take two nodes
form the queue. These are removed and their parent node goes back to the queue with
a new frequency. This procedure goes on till the time the queue is empty. The last
node that becomes in the queue in the result of this procedure is the root node. This
root node has frequency 33 if we apply this procedure to our previous example.
Lets talk about some general things related to Hoffman encoding. We use modems to
connect to Internet. These modems do the compression. The modem has a
compression feature. The modem has a chip that performs the task of compression.
When we give a sentence of say 80 characters to the modem to send, the modem
makes a Hoffman encoded tree of these characters. Then it will make codes from this
tree. We know that there is also a modem on the other end that will decode this
message. The sender modem will send the tree structure to the receiver. Now the
sender modem will send the data in compressed form. The receiving modem will
decode this compressed data by using the Hoffman encoded tree. Now the question
arises, will these codes be useful for other messages? In our example message the
letter y has lower frequency and code of five bits. It may happen that in some
messages y has higher frequency, needing code of less number of bits. To solve this
problem, the modems (sender and receiver) revise the codes after a certain time
period or after a particular number of bytes. In this revision, they build a new
Hoffman tree and exchange it to use it for communication for the next time period or
for the next fixed number of bytes.
There is some other compression techniques/algorithms for compression. The zip
routines like winzip and the jpeg, mpeg and other image formatting routines use
different algorithms for compression. We will read the compression algorithm in
detail in the course of Algorithms.

Mathematical Properties of Binary Trees

There are some mathematical properties of binary trees, which are actually theorems.
We will not prove these here. Most of these properties will be studied in some other
courses where we will prove them as theorem. Here we are going to talk about some
properties, much needed in the next topic about trees.
The first property is that a binary tree of N internal nodes has N+1 external nodes. We
are familiar with the term binary tree and internal node. The term external node is a
new one. To understand the external nodes, look at the following figure.

Page 316 of 519

CS301 Data Structures

___________________________________________________________________

In this figure, the nodes with value i.e. A, B, C, D, E, F and G are the internal nodes.
Note that the leaf nodes are also included in internal nodes. In the figure, we see that
the right pointer of B is NULL. Similarly, the left pointer of D is NULL. The square
nodes in the figure are the NULL nodes. There is no data in these nodes as these are
NULL pointers. However these are the positions where the nodes can exist. These
square nodes are the external nodes. Now we see in the figure that the internal nodes
(leaf nodes are also included) are 9 and the external nodes (NULL nodes indicated by
squares) are 10 (i.e. 9 + 1). Hence it is the property that we have stated. We will see
the usage of this property in the upcoming lectures.

Page 317 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 27
Reading Material
Data Structures and Algorithm Analysis in C++
4.3

Chapter. 4

Summary

Properties of Binary Tree

Threaded Binary Trees
Adding Threads During Insert
Where is Inorder Successor?
Inorder Traversal

Properties of Binary Tree

By the end of the last lecture, we were having a discussion about the properties of the
binary trees. Let us recall, that I told you about a property of the binary trees
regarding relationship between internal nodes and external nodes i.e. If the number of
internal nodes is N, the number of external nodes will be N+1. Today I am going to
discuss another property of the binary trees, which together with the previous lecture,
will give us a start into a new topic. Let me have your attention to the second property
of the binary trees.
Property
A binary tree with N internal nodes has 2N links, N-1 links to internal nodes and N+1
links to external nodes.
Please recall that the first property dealt with the relationship between internal and
external nodes. This property is dealing with the relationship of links to the internal
nodes.
Now, what is a link? As you might already have understood, a link is that line, which
we draw between two nodes in a tree. Internally we use pointers in C++ to realize
links. In pictorial sketches, however, we use a line to show a link between the two
nodes. The property defines, that if you have a binary tree with Nodes, how many
links, it will have between the internal nodes (remember, it includes the leaf nodes),
and how many links it will have between the external nodes. We have not been
showing any links between the external nodes in the diagrams. These are, in fact, null
pointers. That means, these are the links, which we will show with the help of the
square nodes. Let us see a binary tree, on the basis of which, we will further explore
this property. In the following figure, the binary tree is shown again, which, in
addition to the normal links between the internal nodes, also contains external nodes
as squares and the external links as lines going to those squares.

Page 318 of 519

CS301 Data Structures

___________________________________________________________________
Internal links: 8
External links: 10

A
Internal link

C
external link

Fig 27.1

Now if you count the total number of links in the diagram between internal and
external nodes, it will be 2N. Remember, we are talking about links and not nodes. In
this tree, we have 9 nodes marked with capital letters, 8 internal links and 10 external
links. Adding the both kinds of links, we get 18, which is exactly 2 x 9.
As discussed already that these properties are mathematical theorems and can
therefore be proven mathematically. Let us now prove this property as to how do we
get 2N links in a binary tree with N internal nodes.
Property
A binary tree with N internal nodes has 2N links, N-1 links to internal nodes and N+1
links to external nodes.
In every rooted tree, each node, except the root, has a unique parent.
Every link connects a node to its parents, so there are N-1 links connecting
internal nodes.
Similarly each of the N+1 external nodes has one link to its parents.
Thus N-1+N+1=2N links.
In the previous lectures, I told you about the important property of the trees, that they
contain only one link between the two nodes. I had also shown you some structures,
which did not follow this property and I told you, that those were graphs.

Threaded Binary Trees

In many applications binary tree traversals are carried out repeatedly.
The overhead of stack operations during recursive calls can be costly.
The same would true if we use a non-recursive but stack-driven traversal
procedure
It would be useful to modify the tree data structure which represents the binary
Page 319 of 519

CS301 Data Structures

___________________________________________________________________
tree so as to speed up, say, the inorder traversal process: make it "stack-free".
You must be remembering that there were four traversing methods of binary trees:
preorder, inorder, postorder and levelorder. First three preorder, inorder and
postorder were implemented using recursion. Those recursive routines were very
small, 3 to 4 lines of code and they could be employed to traverse a tree of any size.
We also traversed BST in inorder to retrieve the information in sorted order. We
employed stacks in recursive implementations. Although, recursive routines are of
few lines but when recursion is in action, recursive stack is formed that contains the
function calls. We also explicitly used stack for inorder non-recursive traversal. When
the calling pattern of recursive and non-recursive stack based routines were
compared, the calling pattern of both of the routines were similar.
Suppose that we have a BST that is traversed again and again for some operations of
find or print. Due to lot of recursive operations, the stack size keeps on growing. As a
result, the performance is affected. To overcome this performance bottleneck, we can
use non-recursive method but stack-driven traversal will again be an issue. The push
and pop operations of stack for insertion and retrieval will again take time. So is there
a way to do traversal without using a stack neither of implicit function call stack nor
explicit. The same idea is presented in the last bullets above that leads to threaded
binary trees:
It would be useful to modify the tree data structure which represents the binary
tree so as to speed up, say, the inorder traversal process: make it "stack-free".
The idea in the above statement is to modify the tree data structure to speed up and
make it stack-free. Now, we see what kind of modification is required in the binary
trees.
Oddly, most of the pointer fields in our representation of binary trees are NULL!
Since every node (except the root) is pointed to, there are only N-1 non-NULL
pointers out of a possible 2N (for an N node tree), so that N+1 pointers are
NULL.
We know that all the leaf node pointers are NULL. Each node of the tree contains the
data part, two pointer variables for left and right nodes links. But these pointer
variables are used when the node has further child nodes. We know that in a binary
tree the total number of links are 2N including both internal and external and the
number of NULL pointers is N+1.

Page 320 of 519

CS301 Data Structures

___________________________________________________________________
Internal nodes: 9
External nodes: 10

A
Internal node

F
external node

Fig 27.2
In the figure above, the tree is the same as shown in Fig 27.1. The square nodes
shown in this figure are external nodes. Thinking in terms of pointers all the pointers
of these nodes are NULL or in other words they are available to be used later. We
recognize these nodes as leaf nodes. Besides that, what can we achieve using them is
going to be covered in Threaded Binary Trees.
The threaded tree data structure will replace these NULL pointers with pointers
to the inorder successor (predecessor) of a node as appropriate.
We are creating a new data structure inside the tree and when the tree will be
constructed, it will be called a threaded binary tree. The NULL pointers are replaced
by the inorder successor or predecessor. That means while visiting a node, we can tell
which nodes will be printed before and after that node.
We'll need to know whenever formerly NULL pointers have been replaced by non
NULL pointers to successor/predecessor nodes, since otherwise there's no way to
distinguish those pointers from the customary pointers to children.
This is an important point as we need to modify our previous logic of identifying leaf
nodes. Previously the node with left and right nodes as NULL was considered as the
leaf node but after this change the leaf node will contain pointers to predecessor and
successor. So in order to identify that the pointers has been modified to point to their
inorder successor and predecessor, two flags will be required in the node. One flag
will be used for successor and other for predecessor. If both the pointers were NULL,
left pointer variable will be used to point inorder predecessor, the flag for this will be
turned on and the right pointer variable will be used to keep inorder successor and the
flag will be turned on once the successor address is assigned.

Adding Threads During Insert

Page 321 of 519

CS301 Data Structures

___________________________________________________________________

14

15

18

20

16
Fig 27.3

If we print the above tree in inorder we will get the following output:
14 15 18 20
In the above figure, the node 14 contains both left and right links. The left pointer is
pointing to a subtree while the right subtree is pointing to the node 15. The node 15s
right link is towards 18 but the left link is NULL but we have indicated it with a
rounded dotted line towards 14. This indicates that the left pointer points to the
predecessor of the node.
Below is the code snippet for this logic.
t->L = p->L; // copy the thread
t->LTH = thread;
t->R = p; // *p is successor of *t
t->RTH = thread; p->L = t; // attach the new leaf
p->LTH = child;

Lets insert a new node in the tree shown in the above figure. The Fig 27.4 indicates
this new insertion.

Page 322 of 519

CS301 Data Structures

___________________________________________________________________

14

15

18

20
16

Fig 27.4
The new node 16 is shown in the tree. The left and right pointers of this new node are
NULL. As node 16 has been created, it should be pointed to by some variable. The
name of that variable is t. Next, we see the location in the tree where this new node
with number 16 can be inserted. Clearly this will be after the node 15 but before node
18. As a first step to insert this node in the tree as the left child of the node 18, we did
the following:
1. t->L = p->L; // copy the thread
2. t->LTH = thread;
3. t->R = p; // *p is successor of *t
4. t->RTH = thread;
5. p->L = t; // attach the new leaf
6. p->LTH = child;
As the current predecessor of node 18 is 15. After node 16 will be inserted in the tree,
it will become the inorder predecessor of 18, therefore, in the first line of the code t>L = p->L, left pointer of node 18 (pointed to by pointer p) is assigned to the left
pointer of node 16 (pointer to by pointer t).

Page 323 of 519

CS301 Data Structures

___________________________________________________________________

14

15

18

p
1

20
16

t
2

Fig 27.5
In the next line of code t->LTH = thread, the left flag is assigned a variable thread
that is used to indicate that it is on.

14

15

18

p
1

20
t

16

2
Fig 27.6
In the third line of code, t->R = p, 18 being the successor of node 18, its pointer p is
assigned to the right pointer (t->R) of node 16.

Page 324 of 519

CS301 Data Structures

___________________________________________________________________

14

15

18

p
1

20
16 4

2
Fig 27.7
Next line, t->RTH = thread contains flag turning on code.

14

15

18
5

20
16 4

2
Fig 27.8
In the next line p->L = t, the node 16 is attached as the left child of the node 18.

Page 325 of 519

CS301 Data Structures

___________________________________________________________________

14

15

18

20
16 4

2
Fig 27.9
The flag is truned on in the last line, p->LTH = child.
If we insert few more nodes in the tree, we have the tree as given below:

14
15

4
3

18

16

20

5
Fig 27.10
Above given is a BST and you have seen many BSTs before, which are not thread
binary trees. Without the threads, it is clear from the figure that there are number of
links present in the tree that are NULL. We have converted the NULLs to threads in
this tree.
Lets do inorder non-recursive traversal of the tree. We started at 14 then following
the left link came to 4 and after it to 3. If we use recursion then after the call for node
3 is finished (after printing 3), it returns to node 4 call and then 4 is printed using the
recursive call stack. Here we will print 3 but will not stop. As we have used threads,
we see the right pointer of node 3 that is not NULL and pointing to its successor node
Page 326 of 519

CS301 Data Structures

___________________________________________________________________
4, we go to 4 and print it. Now which node is inorder successor of node 4. It is node 5.
From node 4, we traversed to right child of it node 9. From node 9, we went to node 7
and then finally node 5. Now, this node 5 is a leaf node. Previously, without using
threads, we could identify leaf nodes, whose both pointers left and right were NULL.
In this case, using threads, as discussed above, we set the pointers and turn the flags
on when a pointer left or right is set to its predecessor or successor. After printing
node 5, we traverse its right thread and go to node 7. In this fashion, whole of the tree
can be traversed without recursion.
Now, lets see some code:
TreeNode* nextInorder(TreeNode* p)
{
if(p->RTH == thread)
return(p->R);
else {
p = p->R;
while(p->LTH == child)
p = p->L;
return p;
}
}
Above given is a routine nextInorder, which gives the inorder successor of a node
passed in parameter pointer p. Now what it does is:
If the RTH flag of the p node (the node passed in the parameter) is thread then it will
return the node being pointed by the right thread (p->R), which would be its inorder
successor. Otherwise, it does the following.
It goes to the right node of p and starts pointing it using the same p pointer. From
there it keeps on moving (in a loop fashion) towards the left of the node as long as the
statement p->LTH == child is true. After this loop is terminated, the node p is
returned.
Next, we see this pictorially.

Where is Inorder Successor?

Page 327 of 519

CS301 Data Structures

___________________________________________________________________
Inorder successor of 4.

14
15

4
3

18

16

20

5
Fig 27.11
We are at node 4 and want to find its inorder successor. If you remember the delete
operation discussed in the previous lecture, where we searched for the inorder
successor and found it to be the left-most node in the right subtree of the node.
Inorder successor of 4.

14
15

4
3

18

16

20

5
Fig 27.12
Left most node in right subtree of 4

In this figure, the right subtree of 4 is starting from node 9 and ending at node 5. Node
5 is the left most node of it and this is also the inorder successor of node 4. We cannot
go to node 5 directly from node 4, we go to node 9 first then node 7 and finally to
node 5.

Page 328 of 519

CS301 Data Structures

___________________________________________________________________
Inorder successor of 9.

14
15

4
3

18

16

20

Follow right thread to 14.

5
Fig 27.13
We move from node 9 to node 5 following the normal tree link and not thread. As
long as the normal left tree link is there of a node, we have set the LTH flag to child.
When we reach at node 5, the left link is a thread and it is indicated with a flag. See
the while loop given in the above routine again:
while(p->LTH == child)
p = p->L;
return p;

Page 329 of 519

CS301 Data Structures

___________________________________________________________________

Inorder Traversal
Now by using this routine, we try to make our inorder traversal procedure that is nonrecursive and totally stack free.
If we can get things started correctly, we can simply call nextInorder repeatedly
(in a simple loop) and move rapidly around the tree inorder printing node labels
(say) - without a stack.

14

15

4
3

18

16

20

5
Fig 27.14
The pointer p is pointing to the root node of the tree. If we start traversing from this
node and pass this root node pointer to our routine nexInorder above, it will create a
problem.
We see the routine again to see the problem area clearly:
TreeNode* nextInorder(TreeNode* p)
{
if(p->RTH == thread)
return(p->R);
else {
p = p->R;
while(p->LTH == child)
p = p->L;
return p;
}
}
In the first part, it is checking for the RTH flag to be set to thread, which is not the
case for the root node. The control will be passed to the else part of the routine. In
else part, in the very first step, we are moving towards right of root that is to node 15.

Page 330 of 519

CS301 Data Structures

___________________________________________________________________

14
15

4
3

P?

18

16

20

5
Fig 27.15
If we call nextInorder with the root of the binary tree, we're going to have some
difficulty. The code won't work at all the way we want.
Note that in this tree inorder traversal, the first number we should print is 3 but now
we have reached to node 15 and we dont know how can we reach node 3. This has
created a problem. In the lecture, we will make a small change in this routine to cater
to this situation i.e. when a root node pointer is passed to it as a parameter. After that
change the routine will work properly in case of root node also and it will be nonrecursive, stack free routine to traverse the tree.

Page 331 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 28
Reading Material
Data Structures and Algorithm Analysis in C++
6.3.1

Chapter. 6

Summary

Inorder traversal in threaded trees

Complete Binary Tree

Inorder traversal in threaded trees

Discussion on the inroder traversal of the threaded binary tree will continue in this
lecture. We have introduced the threads in the tree and have written the nextInorder
routine. It is sure that the provision of the root can help this routine perform the
inorder routine properly. It will go to the left most node before following the threads
to find the inorder successors. The code of the routine is given below:
/* The inorder routine for threaded binary tree */
TreeNode* nextInorder(TreeNode* p){
if(p->RTH == thread) return(p->R);
else {
p = p->R;
while(p->LTH == child)
p = p->L;
return p;
}
}
When we apply this routine on the sample tree, it does not work properly because the
pointer that points to the node goes in the wrong direction. How can we fix this
problem? Lets review the threaded binary tree again:

Page 332 of 519

CS301 Data Structures

___________________________________________________________________

14
4

p
15

9
7

18
16

20

In the above figure, we have a binary search tree. Threads are also seen in it. These
threads points to the successor and predecessor.
Our nextInoder routine, first of all checks that the right pointer of the node is thread.
It means that it does not point to any tree node. In this case, we will return the right
pointer of the node as it is pointing to the inorder successor of that node. Otherwise,
we will go to some other part. Here we will change the value of pointer p to its right
before running a while loops as long as the left pointer is the node. That means the
left child is not a thread. We move to the left of the pointer p and keep on doing so till
the time the left pointer becomes a thread.
We will pass the root of the tree to the nextInorder routine. The pointer p is pointing
to the node 14 i.e. the root node. As the right pointer of the node 14 is not a thread, so
the pointer p will move to the node 15 as shown below:

Page 333 of 519

CS301 Data Structures

___________________________________________________________________

14
4

p?

15

9
7

18
16

20

Here we want the inorder traversal. It is obvious from the above figure that 15 is not
the first value. The first value should be 3. This means that we have moved in the
wrong direction. How this problem can be overcome? We may want to implement
some logic that in case of the root node, it is better not to go towards the right side.
Rather, the left side movement will be appropriate. If this is not the root node, do as
usual. It may lend complexities to our code. Is there any other way to fix it? Here we
will use a programming trick to fix it.
We will make this routine as a private member function of the class so other classes
do not have access to it. Now what is the trick? We will insert a new node in the tree.
With the help of this node, it will be easy to find out whether we are on the root node
or not. This way, the pointer p will move in the correct direction.
Lets see this trick. We will insert an extra node in the binary tree and call it as a
dummy node. This is well reflected in the diagram of the tree with the dummy node.
We will see where that dummy node has been inserted.

Page 334 of 519

CS301 Data Structures

___________________________________________________________________

dummy

14
4

15

18

16

20

This dummy node has either no value or some dummy value. The left pointer of this
node is pointing to the root node of the tree while the right pointer is seen pointing
itself i.e. to dummy node. There is no problem in doing all these things. We have put
the address of dummy node in its right pointer and pointed the left thread of the left
most node towards the dummy node. Similarly the right thread of the right-most node
is pointing to the dummy node. Now we have some extra pointers whose help will
make the nextInorder routine function properly.
Following is a routine fastInorder that can be in the public interface of the class.
/* This routine will traverse the binary search tree */
void fastInorder(TreeNode* p)
{
while((p=nexInorder(p)) != dummy)
}

cout << p->getInfo();

This routine takes a TreeNode as an argument that make it pass through the root of the
tree. In the while loop, we are calling the nextInorder routine and pass it p. The
pointer returned from this routine is then assigned to p. This is a programming style of
C. We are performing two tasks in a single statement i.e. we call the nextInorder by
passing it p and the value returned by this routine is saved in p. Then we check that
the value returned by the nextInorder routine that is now actually saved in p, is not a
dummy node. Then we print the info of the node. This function is called as:
fastInorder(dummy);
We are not passing it the root of the tree but the dummy node. Now we will get the
correct values and see in the diagrams below that p is now moving in the right
direction. Lets try to understand this with the help of diagrams.
Page 335 of 519

CS301 Data Structures

___________________________________________________________________
First of all we call the nextInorder routine passing it the dummy node.

dummy

p
14
4

15

9
7

18
16

20

The pointer p is pointing to the dummy node. Now we will check whether the right
pointer of this node is not thread. If so, then it is advisable to move the pointer
towards the right pointer of the node. Now we will go to the while loop and start
moving on the left of the node till the time we get a node with the left pointer as
thread. The pointer p will move from dummy to node 14. As the left pointer of node
14 is not thread so p will move to node 4. Again the p will move to node 3. As the left
pointer of p is thread, the while loop will finish here. This value will be returned that
is pointing to node 3. The node 3 should be printed first of all regarding the inorder
traversal. So with the help of our trick, we get the right information.
Now the while loop in the fastInorder will again call the nextInorder routine. We
have updated the value of p in the fastInorder that is now pointing to the node 3. This
is shown in the figure below:

Page 336 of 519

CS301 Data Structures

___________________________________________________________________
dummy

14
4

15

9
7

18
16

20

According to the code, we have to follow the right thread of the node 3 that is
pointing to the node 4. Therefore p is now pointing to the node 4. Here 4 is inorder
successor of 3. So the pointer p has moved to the correct node for inorder traversal.
As the right pointer of the node 4 is a link, p will move to node 9. Later, we will go on
the left of nodes and reach at the node 5. Looking at the tree, we know that the inorder
successor of the node 4 is node 5. In the next step, we will get the node 7 and so on.
With the help of threads and links, we are successful in getting the correct inorder
traversal. No recursive call has been made so far. Therefore stack is not used. This
inorder traversal will be faster than the recursive inorder traversal. When other classes
use this routine, it will be faster. We have not used any additional memory for this
routine. We are using the null links and putting the values of thread in it. This routine
is very simple to understand. In the recursive routines, we have to stop the recursion
at some condition. Otherwise, it will keep on executing and lead to the aborting of our
program.

Complete Binary Tree

We have earlier discussed the properties of the binary trees besides talking about the
internal and external nodes theorem. Now we will discuss another property of binary
trees that is related to its storage before dilating upon the complete binary tree and the
heap abstract data type.
Here is the definition of a complete binary tree:

A complete binary tree is a tree that is completely filled, with the possible
exception of the bottom level.
The bottom level is filled from left to right.

You may find the definition of complete binary tree in the books little bit different
from this. A perfectly complete binary tree has all the leaf nodes. In the complete
binary tree, all the nodes have left and right child nodes except the bottom level. At

Page 337 of 519

CS301 Data Structures

___________________________________________________________________
the bottom level, you will find the nodes from left to right. The bottom level may not
be completely filled, depicting that the tree is not a perfectly complete one. Lets see a
complete binary tree in the figure below:

A
B

In the above tree, we have nodes as A, B, C, D, E, F, G, H, I, J. The node D has two

children at the lowest level whereas node E has only left child at the lowest level that
is J. The right child of the node E is missing. Similarly node F and G also lack child
nodes. This is a complete binary tree according to the definition given above. At the
lowest level, leaf nodes are present from left to right but all the inner nodes have both
children. Lets recap some of the properties of complete binary tree.

A complete binary tree of height h has between 2h to 2h+1 1 nodes.

The height of such a tree is

the tree.

Because the tree is so regular, it can be stored in an array. No pointers are

necessary.

log2N

where N is the number of nodes in

We have taken the floor of the log2 N. If the answer is not an integer, we will take the
next smaller integer. So far, we have been using the pointers for the implementation
of trees. The treeNode class has left and right pointers. We have pointers in the
balance tree also. In the threaded trees, these pointers were used in a different way.
But now we can say that an array can be stored in a complete binary tree without
needing the help of any pointer.
Now we will try to remember the characteristics of the tree. 1) The data element can
be numbers, strings, name or some other data type. The information is stored in the
node. We may retrieve, change or delete it. 2) We link these nodes in a special way
i.e. a node can have left or right subtree or both. Now we will see why the pointers are
being used. We just started using these. If we have some other structure in which trees
can be stored and information may be searched, then these may be used. There should
be reason for choosing that structure or pointer for the manipulation of the trees. If we

Page 338 of 519

CS301 Data Structures

___________________________________________________________________
have a complete binary tree, it can be stored in an array easily.
The following example can help understand this process. Consider the above tree
again.

A
B

9 10 11 12 13 14

We have seen an array of size 15 in which the data elements A, B, C, D, E, F, G, H, I,

J have been stored, starting from position 1. The question arises why we have stored
the data element this way and what is justification of storing the element at the 1st
position of the array instead of 0th position? You will get the answers of these very
shortly.
The root node of this tree is A and the left and right children of A are B and C. Now
look at the array. While storing elements in the array, we follow a rule given below:

For any array element at position i, the left child is at 2i, the right child is at
(2i +1) and the parent is at floor(i/2).

In the tree, we have links between the parent node and the children nodes. In case of
having a node with left and right children, stored at position i in the array, the left
child will be at position 2i and the right child will be at 2i+1 position. If the value of i
is 2, the parent will be at position 2 and the left child will be at position 2i i.e. 4 .The
right child will be at position 2i+1 i.e. 5. You must be aware that we have not started
from the 0th position. It is simply due to the fact if the position is 0, 2i will also
become 0. So we will start from the 1st position, ignoring the 0th.
Lets see this formula on the above array. We have A at the first position and it has two
children B and C. According to the formula the B will be at the 2i i.e. 2nd position and
C will be at 2i+1 i.e. 3rd position. Take the 2nd element i.e. B, it has two children D

Page 339 of 519

CS301 Data Structures

___________________________________________________________________
and E. The position of B is 2 i.e. the value of i is 2. Its left child D will be at positon
2i i.e. 4th position and its right child E will be at position 2i+1 i.e. 5. This is shown in
the figure below:

9 10 11 12 13 14

If we want to keep the trees data in the array, the children of B should be at the
position 4 and 5. This is true. We can apply this formula on the remaining nodes also.
Now you have understood how to store trees data in an array. In one respect, we are
using pointers here. These are not C++ pointers. In other words, we have implicit
pointers in the array. These pointers are hidden. With the help of the formula, we can
obtain the left and right children of the nodes i.e. if the node is at the ith position, its
children will be at 2i and 2i+1 position. Lets see the position of other nodes in the
array.
As the node C is at position 3, its children should be at 2*3 i.e. 6th position and 2*3+1
i.e. 7th position. The children of C are F and G which should be at 6th and 7th position.
Look at the node D. It is at position 4. Its children should be at position 8 and 9. E is
at position 5 so its children should be at 10 and 11 positions. All the nodes have been
stored in the array. As the node E does not have a right child, the position 11 is empty
in the array.

9 10 11 12 13 14

You can see that there is only one array going out of E. There is a link between the
parent node and the child node. In this array, we can find the children of a node with
the help of the formula i.e. if the parent node is at ith position, its children will be at 2i
and 2i+1 position. Similarly there is a link between the child and the parent. A child
can get its parent with the help of formula i.e. if a node is at ith position, its parent
will be at floor(i/2) position. Lets check this fact in our sample tree. See the diagram
below:

Page 340 of 519

CS301 Data Structures

___________________________________________________________________

1 A
2

3
B

C
5

4
D
8

10

9
H

7
G

9 10 11 12 13 14

Level Order Numbers & Array index

Consider the node J at position is 10. According to the formula, its parent should be at
floor(10/2) i.e. 5 which is true. As the node I is at position 9, its parent should be at
floor(9/2) i.e. 4. The result of 9/2 is 4.5. But due to the floor, we will round it down
and the result will be 4. We can see that the parent of I is D which is at position 4.
Similarly the parent of H will be at floor(8/2). It means that it will be at 4. Thus we
see that D is its parent. The links shown in the figure depict that D has two children H
and I. We can easily prove this formula for the other nodes.
From the above discussion we note three things. 1) We have a complete binary tree,
which stores some information. It may or may not be a binary search tree. This tree
can be stored in an array. We use 2i and 2i+1 indexing scheme to put the nodes in the
array. Now we can apply the algorithms of tree structure on this array structure, if
needed.
Now lets talk about the usage of pointers and array. We have read that while
implementing data structures, the use of array makes it easy and fast to add and
remove data from arrays. In an array, we can directly locate a required position with
the help of a single index, where we want to add or remove data. Array is so
important that it is a part of the language. Whereas the data structures like tree, stack
and queue are not the part of C or C++ language as a language construct. However we
can write our classes for these data structures. As these data structures are not a part
of the language, a programmer can not declare them directly. We can not declare a
tree or a stack in a program. Whereas we can declare an array directly as int x []; The
array data type is so efficient and is of so common use that most of the languages
support it. The compiler of a language handles the array and the programmer has to
do nothing for declaring and using an array.
We have built the binary trees with pointers. The use of pointers in the memory
Page 341 of 519

CS301 Data Structures

___________________________________________________________________
requires some time. In compilers or operating system course, we will read that when a
program is stored in the memory and becomes a process, the executable code does not
come in the memory. There is a term paging or virtual memory. When a program
executes, some part of it comes in the memory. If we are using pointers to go to
different parts of the program, some part of the code of program will be coming
(loading) to memory while some other may be removed (unloading) from the
memory. This loading and unloading of program code is executed by a mechanism,
called paging. In Windows operating system, for this virtual memory (paging
mechanism), a file is used, called page file. With the use of pointers, this process of
paging may increase. Due to this, the program may execute slowly. In the course of
Operating System and Compilers, you will read in detail that the usage of pointers can
cause many problems.
So we should use arrays where ever it can fulfill our requirements. The array is a very
fast and efficient data structure and is supported by the compiler. There are some
situations where the use of pointers is beneficial. The balancing of AVL tree is an
example in this regard. Here pointers are more efficient. If we are using array, there
will be need of moving a large data here and there to balance the tree.
From the discussion on use of pointers and array, we conclude that the use of array
should be made whenever it is required. Now it is clear that binary tree is an
important data structure. Now we see that whether we can store it in an array or not.
We can surely use the array. The functions of tree are possible with help of array.
Now consider the previous example of binary tree. In this tree, the order of the nodes
that we maintained was for the indexing purpose of the array. Moreover we know the
level-order traversal of the tree. We used queue for the level-order of a tree. If we do
level-order traversal of the tree, the order of nodes visited is shown with numbers in
the following figure.

1 A
2

3
B

C
5

4
D
8

10

9
H

7
G

9 10 11 12 13 14

In the above figure, we see that the number of node A is 1. The node B is on number 2
Page 342 of 519

CS301 Data Structures

___________________________________________________________________
and C is on number 3. At the next level, the number of nodes D, E, F and G are 4, 5, 6
and 7 respectively. At the lowest level, the numbers 8, 9 and 10 are written with nodes
H, I and J respectively. This is the level-order traversal. You must remember that in
the example where we did the preorder, inorder and post order traversal with
recursion by using stack. We can do the level-order traversal by using a queue. Now
after the level-order traversal, lets look at the array shown in the lower portion of the
above figure. In this array, we see that the numbers of A, B, C and other nodes are the
same as in the level-order traversal. Thus, if we use the numbers of level-order
traversal as index, the values are precisely stored at that numbers in the array. It is
easy for us to store a given tree in an array. We simply traverse the tree by level-order
and use the order number of nodes as index to store the values of nodes in the array. A
programmer can do the level-order traversal with queue as we had carried out in an
example before. We preserve the number of nodes in the queue before traversing the
queue for nodes and putting the nodes in the array. We do not carry out this process,
as it is unnecessarily long and very time consuming. However, we see that the levelorder traversal directly gives us the index of the array depending upon which data can
be stored.

Page 343 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 29
Reading Material
Data Structures and Algorithm Analysis in C++
6.3

Chapter. 6

Summary

Complete Binary Tree

Heap
Max Heap
Insertion in a Heap

Complete Binary Tree

In the previous lecture, we talked about the ways to store a complete binary tree in an
array. The 2i and 2i+1 scheme were employed to store the link of the parent to
children and link of children to parent. Through this link, a programmer can go to the
children of a node. We know that array is a very efficient data structure. In a number
of languages, it is found as a built-in data type. Now the question arises if we can
store the binary tree in an array, why there should be the use of pointers? It is very
simple that an array is used when the tree is a complete binary tree. Array can also be
used for the trees that are not complete binary trees. But there will be a problem in
this case. The size of the array will be with respect to the deepest level of the tree
according to the traversal order. Due to incomplete binary tree, there will be holes in
the array that means that there will be some positions in the array with no value of
data. We can understand it with a simple example. Look at the following figure where
we store a complete binary tree in an array by using 2i and 2i+1 scheme. Here we
stored the nodes from A to J in the array at index 1 to 10 respectively.

Page 344 of 519

CS301 Data Structures

___________________________________________________________________

1 A
2

3
B

C
5

4
D

10

7
G

9 10 11 12 13 14

Figure 29.1: Complete Binary Tree

Suppose that this tree is not complete. In other words, B has no right subtree that
means E and J are not there. Similarly we suppose that there is no right subtree of A.
Now the tree will be in the form as shown in the following figure (29.2).

1 A
2
B
4
D
8

9
H

I
A

D
3

9 10 11 12 13 14

Figure 29.2: Not a complete binary tree

Page 345 of 519

CS301 Data Structures

___________________________________________________________________
In this case, the effort to store this tree in the array will be of no use as the 2i and 2i+1
scheme cannot be applied to it. To store this tree, it may be supposed that there are
nodes at the positions of C, F, G, E and J (that were there in previous figure). Thus we
transform it into a complete binary tree. Now we store the tree in the array by using 2i
and 2i +1 scheme. Afterwards, the data is removed from the array at the positions of
the imaginary nodes (in this example, the nodes are C, F, G, E and J). Thus we notice
that the nodes A, B and H etc are at the positions, depicting the presence of a
complete binary tree. The locations of C, f, G, E and J in the array are empty as
shown in the following figure.

D
3

9 10 11 12 13 14

Now imagine that an incomplete binary tree is very deep. We can store this tree in the
array that needs to be of large size. There will be holes in the array. This is the
wastage of memory. Due to this reason, it is thought that if a tree is not completely
binary, it is not good to store it into an array. Rather, a programmer will prefer to use
pointers for the storage.
Remember that two things are kept into view while constructing a data structure that
is memory and time. There should such a data structure that could ensure the running
of the programs in a fast manner. Secondly, a data structure should not use a lot of
memory so that a large part of memory occupied by it does not go waste. To manage
the memory in an efficient way, we use dynamic memory with the help of pointers.
With the use of pointers only the required amount of memory is occupied.
We also use pointers for complex operations with data structure as witnessed in the
deletion operation in AVL tree. One of the problems with arrays is that the memory
becomes useless in case of too many empty positions in the array. We cannot free it
and use in other programs as the memory of an array is contiguous. It is difficult to
free the memory of locations from 50 to 100 in an array of 200 locations. To manage
the memory in a better way, we have to use pointers.
Now we come to a new data structure, called heap.

Heap
Heap is a data structure of big use and benefit. It is used in priority queue. Recall the
example of bank simulation. In that case, we used event-based queues. We put the
events that were going to happen in a special queue i.e. priority queue. This priority
queue does not follow the FIFO rule. We put the elements in the queue at the end but
later got the element with respect to its priority. We get the element that is going to
occur first in the future. In that example, we implemented the priority queue with
arrays. It was seen that when we insert an element in the queue, the internally used
data was sorted in the array. Thus the event with minimum time of occurrence
becomes at first position in the sorted array. We get the event with minimum time
first. After the removal of the element, a programmer shifts the array elements to left.
When we insert a new element, the array is sorted again. As there, in the bank
example, were five or six events, the use of array fulfilled our requirement. We need
not to have other data type that makes the queue efficient. The array is efficient but
sorting is an expensive procedure. It may be complex and time-consuming. Secondly,
Page 346 of 519

CS301 Data Structures

___________________________________________________________________
we have to shift elements while adding or removing them from the array. Thus the
implementation of priority queue with an array is not efficient in terms of time. There
is an alternate that is called heap. The use of priority queue with the help of heap is a
major application. The priority queue is itself a major data structure, much-used in
operating systems. Similarly priority queue data structure is used in network devices
especially in routers. Heap data structure is also employed in sorting algorithms.
There are also other uses of heap.
Lets discuss the heap data structure with special reference to its definition,
The definition of heap is that it is a complete binary tree that conforms to the heap
order.
In this definition there is a term heap order. The heap order is a property that states
that in a (min) heap for every node X, the key in the parent is smaller than (or equal
to) the key in X. In other words, the parent node has key smaller than or equal to both
of its children nodes. This means that the value in the parent node is less than the
value on its children nodes. It is evident from the definition that we implement the
heap by complete binary tree. It can also be implemented by some other method. But
normally, we implement heap with complete binary tree. We know that in a binary
search tree, the value of a node is greater than the values in its left subtree and less
than the values in its right subtree. The heap has a variation to it. In heap, the value of
a node is less than the values of its left and right children. The values in left and right
children may be more or less with each other but these will be greater than the value
in their parent node.
Consider the tree shown in the following figure.

13
21

16

24

65

31

26

19

68

32
Figure 29.3: A min heap

This is a complete binary tree. Now consider the values (numbers) in the nodes. This
is not a binary search tree. If we carry out the inorder traversal, the result will be
65, 24, 26, 21, 32, 31, 13, 19, 16, 68.
We can see that it is not in a sorted order as got while traversing a binary search tree.

Page 347 of 519

CS301 Data Structures

___________________________________________________________________
So it is not a binary search tree. Its simply a complete binary tree.
Now we see the heap order in this tree. We start from node having value 13. Its left
child has value 21 while the value possessed by the right child is 16. So this is in line
with the heap order property. Now we come to node 21. The values in its left and
right child are 24 and 31 respectively. These values are greater than 21. Thus, it also
fulfills the heap order property. Now we consider the node having value 16. The value
in left child node of it is 19 and value in right child node is 68. So the value of parent
node (i.e. 16) is less than the values of its children nodes. Similarly we can see other
nodes. The node 24 is less than its children that are 65 and 26 respectively. And the
node 31 is less than its child i.e. 32. Now for this tree, three things have been proved.
First, this is a binary tree. Secondly it is a complete binary tree. Thirdly it fulfills the
heap order property i.e. the value of the parent node is less than that of its left and
right child nodes. This property is valid for every node of the tree. The leaf nodes
have no child so there is no question of heap property. The heap shown in the figure
above is a min heap. In the min heap, the value of parent node is less than the values
in its child nodes. Thus in a min heap, the value of the root node is the smallest value
in the tree.
Now consider the tree in the following figure. This is a complete binary tree. This is
neither a search tree, nor a heap.

13

9
21

65

19

31

26

16

68

32
Figure 29.4: Not a heap

Look at the node having value 19. The values in its left and right child nodes are 16
and 68 respectively. Thus the value of left child (i.e. 16) is less than that of the
parent. So it is not a heap. If it were a heap or min heap, the value 16 should have
been parent and the value 19 should have its child. Due to this violation, (the value of
child is less than that of the parent) it is not a heap (min heap).
Max Heap
We can also make a max heap. In max heap, each node has a value greater than the
value of its left and right child nodes. Moreover, in this case, the value of the root
node will be largest and will become lesser at downward levels. The following figure
shows a max heap.

Page 348 of 519

CS301 Data Structures

___________________________________________________________________

73
52

63

40

25

31

26

27

57

13
Figure 29.5: A max heap

Consider the min heap again. By removing the root from the min heap, we get the
smallest value. Now if the remaining values adjust themselves to again form a heap,
the minimum value among these remaining values will get on the top. And if we
remove this value, there will be the second minimum value of the heap. The
remaining values again form a heap and there will be the smallest number among
these on the top. Thus we will get the third minimum number. If we continue this
process of getting the root node and forming a heap of remaining values, the numbers
will get in ascending order. Thus we get the sorted data. By putting data in heap and
getting it in a particular way, we achieve a sorting procedure. This way, a programmer
gets sorted data.
Now suppose that the numbers in heap are priorities or events. If we build the max
heap, the number with greater priority will get on the top. Now if we get data from the
heap, the data on top will be gotten first. As it is max heap, the data on top will be the
largest value.

Insertion in a Heap
Now lets discuss the insertion of a value in a min or max heap. We insert a value in a
min heap and rearrange it in such a way that the smallest value gets on the top.
Similarly, if we are inserting a value in a max heap, the largest value goes on the top.
To discuss insertion in min heap, consider the following existing heap. This is the
same heap discussed earlier. As it is a complete binary tree, we keep it in an array. In
the figure, the level order of numbers of the nodes is also seen that describes the
position of nodes (index number) in the array.

Page 349 of 519

CS301 Data Structures

___________________________________________________________________

13

21
4

24

65

26

10

31

16

19 6

68

32

13 21 16 24 31 19 68 65 26 32
0

9 10 11 12 13 14

Figure 29.6: An existing heap

Now we add a new node to this existing heap. Suppose the node that we want to add
has a value 14. We will add this node at a position in such a way that the tree should
remain complete binary one. Following this guideline, this node will be added on
right side of node 31. In the array, the node 31 is presently at position 5. Now
according to the formula of 2i and 2i+1, the left child of 31 will be at positions 10 that
already exists. The right child of 31 (that is 14 now) will be at position 11 in the array.
This addition of node in the tree and value in the array is shown in the following
figure. We have shown the node link and position in the array with dotted lines. By
this, it is shown that the node of value 14 may become here. We did not actually put it
there. We have to put it at such position so that the heap should remain intact.

Page 350 of 519

CS301 Data Structures

___________________________________________________________________

13

21
4

24

65

26

10

31

19 6
11

32

16
7

68

14

13 21 16 24 31 19 68 65 26 32 14
0

9 10 11 12 13 14

Figure 29.7: inserting new value in a heap

With the addition of new node, we see that this is still a complete binary tree and can
be stored in an array. There are no holes in the array, leading to no wastage of the
positions.
Now preservation of the heap order is needed. We are discussing the min heap in this
example. In the min heap, the smallest element is at the top (root). And at each level,
every node has a smaller value than that of its children nodes. Now we have to put the
node 14 at its proper position with respect to heap order. Thus heap is being
implemented with complete binary tree with the help of an array. We have to find the
final; position of node 14 in the tree and array preserving the heap order. A look on
the tree in the previous figure shows that if we put node 14 (shown with dotted line)
at that position, the heap order will not be preserved. By considering the node 31, we
see that the value of its right child (i.e. 14) is less than it. Thus heap property is
violated. To preserve heap property, the node 14 should be up and node 31 should be
down. If node 14 is at position of 31 and it is compared with its parent that is 21, we
come to know that node 21 should also be at down position. How do we take node 14

Page 351 of 519

CS301 Data Structures

___________________________________________________________________
to its proper position? To achieve this objective, we have to take node 21 and 31
down and move the node 14 up. Now lets see how we can do this. In the following
figure, the position is seen where a new node can be added with an empty circle
(hole). The value of this node is not shown. However, the new inserted value may go
to that position. Similarly in the array we show that there will be some value at
position 11.

13

21
4

24

65

26

10

31

16

19 6

68

11

32

13 21 16 24 31 19 68 65 26 32
0

9 10 11 12 13 14

Figure 29.8: inserting new value in a heap

Now we compare the new value to be inserted (that is 14) at position 11 with its
parent node. The parent node of newly inserted node is 31 that is greater than the new
value. This is against the (min) heap property. According to the heap property, the
node 14 may be the parent of 31 but not its child. So we move the node 31 to the
position 11 and the new node to the position 5 that was earlier the position of 31. This
technique is also employed in the array. Thus we get the array and tree in the form as
shown in the following figure.

CS301 Data Structures

___________________________________________________________________

Now if the new node comes at the position 5, its parent (that is node 21 at position 2)
is again greater than it. This again violates the heap property because the parent (i.e.
21) is greater than the child whose value is 14. So we bring the node 21 down and
take the new node up. Thus the node 21 goes to the position 5 and the new node
attains the position 2 in the tree and array as shown in the following figure.
1

13

65

24

26

10

32

21

19 6
11

16
7

68

31

Page 353 of 519

CS301 Data Structures

___________________________________________________________________

16 24 21 19 68 65 26 32 31

13
0

9 10 11 12 13 14

Figure 29.10: inserting new value in a heap

To interchange the positions of values in the array, only these values are swapped
which is very easy process. We can do this easily in arrays. Now if the new value 14
comes at the position 2, the heap property will be preserved due to the fact that the
parent of this node i.e. the node of value 13 is less than its child (the new node that is
14). Thus the final position of new node 14 is determined and we put it here i.e. at
position 2. This is shown in the following figure.
1

13

14
4

24

65

26

10

21

19 6
11

32

16
7

68

31

13 14 16 24 21 19 68 65 26 32 31
0

9 10 11 12 13 14

Figure 29.11: inserting new value in a heap

It is clear by now that the tree after insertion of a new value follows the heap
property.
We see that there is an algorithm for inserting a new value in the heap. If we have a
heap stored in an array as a complete binary tree, a new value is put in the array at a

Page 354 of 519

CS301 Data Structures

___________________________________________________________________
position so that it can hold preserving the property of complete binary tree. In our
example, it is the position 11 where new value may be put. We know that the parent
of a node is at position floor (i / 2). Thus we find the position of parent node of the
new node and compare this new value with that. If the parent node has a value greater
than this new value (i.e. its child), the heap property is violated. To maintain this
property, we exchange these values. Now at the new position we find the parent of
that position and compare the value with the value at that position. Thus the array is
traversed by level-order. After a comparison, we go one level up in the tree. And we
go to upward by looking at the parent node of the newly inserted node level by level.
We stop at the position where the value of the parent node is less than the value of its
child i.e. the node to be inserted. To insert a node, it is necessary to find the position
of the node before putting the value there. This is efficient and fast way as actual
values are not exchange in this case. We simply move the data. Under the second
method, it can also be done with exchanges. Lets do this for our previous heap. Look
at the following figure. Here we put the new node 14 at its position. It is the right
child of node 31. We have already seen that this is the position where a new node can
be added so that the tree remains complete binary tree. In the array, it is at position
11.
1
2

21
5

24

65

26

13

10

31

19 6
11

32

16
7

68

14

13 14 16 24 21 19 68 65 26 32 31
0

9 10 11 12 13 14

Figure 29.12: inserting new value in a heap

Now we check the heap order and compare node 14 with its parent that is node 31. It
is seen that this child node i.e. 14 is less than its parent i.e. 31. Thus the heap order
property is violated. We exchange the node 14 and 31due to which the node 14
becomes the parent and node 31 turns into its child. That is why, the node 31 is now

Page 355 of 519

CS301 Data Structures

___________________________________________________________________
at position 11 and node 14 is at position 5. After this exchange, the tree remains
complete binary one as we only have exchanged the values of nodes. The following
array representation shows this.

13 21 16 24 14 19 68 65 26 32 31
0

9 10 11 12 13 14

After this we compare the node 14 with its new parent. The new parent of node 14
can be found by the formula of floor (i / 2). The parent node of 14 will be at position
floor (5/2) that is position 2. We can see that the node at position 2 is 21. Thus 21 is
greater than its child i.e. 14, violating the heap property. So we again exchange these
values. The node 14 now becomes the parent of 21 and 21 gets its child. In the array,
the nodes 14 and 21 are at positions 2 and 5 respectively. The array representation of
it is as below.

13 14 16 24 21 19 68 65 26 32 31
1

9 10 11 12 13 14

Now we compare this node 14 with its new parent i.e. 13. Here the heap property
stands preserved due to the fact that the parent node i.e. 13 is less than its child node
i.e. 14. So this tree is a heap now. The following figure shows this heap and array
representation of it.

13

14
4

65

24

26

10

32

21

19 6
11

16
7

68

31
Page 356 of 519

CS301 Data Structures

___________________________________________________________________

13 14 16 24 21 19 68 65 26 32 31
0

9 10 11 12 13 14

Figure 29.13: inserting new value in a heap

Suppose we want to add another node to this heap. This new node has value 15. As
the heap is a complete binary tree, so this new node will be added as left child of node
19. In the array, we see that the position of 19 is 6 so the position of its left child will
be (by formula of 2i) 6 x 2 that is 12. This is shown in the following figure.

13

14
4

24

65

26

10

21

32

11

31

19

16
7

68

15 12

13 14 16 24 21 19 68 65 26 32 31 15
0

9 10 11 12 13 14

Figure 29.14: inserting new value in a heap

Here the new node 15 is less than its parent node (i.e. 19). To preserve the heap
property, we have to exchange these values. Thus the value 15 goes to the position 6
and value 19 attains the position 12 as shown in the array in figure below.

Page 357 of 519

CS301 Data Structures

___________________________________________________________________

13 14 16 24 21 15 68 65 26 32 31 19
0

9 10 11 12 13 14

Now we compare the value 15 with its new parent i.e.16. Here again the parent (16) is
greater than its child (15). So to preserve the heap property we need to exchange these
values. After the exchange, the value 15 is now at position 3 and value 16 is seen
position 6. The following figure shows this step.

13

14
4

24

65

26

10

21

32

11

31

16

15
7

68

19 12

13 14 15 24 21 16 68 65 26 32 31 19
0

9 10 11 12 13 14

Figure 29.15: inserting new value in a heap

The figure also shows that the node 15 is greater than its new parent i.e. 13. In other
words, the parent node is less than its child. So the heap property has been preserved,
warranting no exchange of values. The node 15 has gone to its proper position .All the
nodes in this tree now follow the heap order. Thus it is a heap in which a new node
has been inserted.

Page 358 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 30
Reading Material
Data Structures and Algorithm Analysis in C++
6.3

Chapter. 6

Summary

Inserting into a Min-Heap

Deleting from a Min-Heap
Building a Heap

Inserting into a Min-Heap

In the previous lecture, we discussed about the heap data structure besides touching
upon the concept of storage of complete binary tree in an array. For the study of
parent-child relationship, the 2i and 2i+1 scheme was used. Then we changed our
focus to heap and found it best for the priority queues. In the previous lecture, we did
not really discuss the uses of heap. Rather, most of the discussion remained focused
on insertion method in the binary tree employed at the time of implementation with
the help of an array. After inserting a new element in the array and by moving few
elements, a programmer can have minimum or maximum heap. In case of minimum
heap, the minimum value in the tree is always in the root of the tree. However, in case
of maximum heap, the maximum value in the tree lies in the root node.
When we insert a new element in the tree implemented with the help of an array, we
insert element at the last position (of the array). Due to this insertion at the end, the
heap order may be violated. Therefore, we start moving this element upwards. While
moving upward, this element may reach at the root of the tree. It is important to note
that only one branch of the tree is affected because of this movement in the upward
direction. This process, being of localized nature, will not disturb the entire tree.
To recap, see the figure Fig 30.1, where we are inserting an element 15.

Page 359 of 519

CS301 Data Structures

___________________________________________________________________

14

24

65

13

Insert (15)
with exchange

10

26

21

11

32

12

31

16

19

68

15

13

14

16

24

21

19

68

65

26

32

31

15

10

11

12

13

14

Fig 30.1

The new element 15 is inserted at the last array position 12. Being a complete binary
tree, the next new node will be the left child of node 19. As node 19 is at array
position 6 (or level order traversal), its left child will be 6 * 2 = 12th position node or
the value at 12th position in the array. Now, we see where we will have to carry out
the exchange operation. As the parent of 15, the number 19 is greater than it, the first
exchange will be among 19 and 15 as shown in the above figure. After exchange, the
new figure is shown in Fig 30.2.
1

Insert (15)
with exchange
2

14

24

26

21

10

65

13

11

32

12

31

16

15

68

19

13

14

16

24

21

15

68

65

26

32

31

19

10

11

12

13

14

Fig 30.2

You can see that both the elements have exchanged positions i.e. 19 has come down
and 15 gone up. But number 15 is still less than its parent 16, so we will have another
exchange operation.

Page 360 of 519

CS301 Data Structures

___________________________________________________________________

14

24

26

21

10

65

13

Insert (15)
with exchange

11

32

12

31

15

16

68

19

13

14

15

24

21

16

68

65

26

32

31

19

10

11

12

13

14

Fig 30.3

Now the new parent of 15 is 13 which is less than it. Therefore, the exchange
operation will stop here as 15 has found its destination in the tree. This new tree is not
violating any condition of heap order as witnessed before insertion of 15. It has
become min-heap again while maintaining its status as a complete binary tree.
1

Insert (15)
with exchange
2

14

24

10

65

13

26

21

11

32

12

31

15

16

68

19

13

14

15

24

21

16

68

65

26

32

31

19

10

11

12

13

14

Fig 30.4

A view of the path shows that the number 15 has gone through to find its final
destination in the tree. It has only affected the right sub-trees left branch that was
containing 16, 19 and 15. It has not affected any other branch. This is not a binary

Page 361 of 519

CS301 Data Structures

___________________________________________________________________
search tree. Here we are only fulfilling one condition that the parent value should be
less than that of its two-children. These exchanges will not take long as we did
mathematically that a tree containing N nodes; can go to log2N level maximum. You
know, when we built binary tree for balancing, it had turned into a linked list. Here
the case is different. the number of levels in complete binary tree will be around
log2N while building a complete binary tree.
Now, we should be clear that a new element can go up to what maximum level in the
tree. If the number is greater than all parent nodes, it will be already at its destination
place, needing no exchange. If we insert a number smaller than all parent nodes, the
exchange operation reaches to the root of the tree. But this exchange operations
intensity does not increase from the one we saw in our previous case of linked lists.

Deleting from a Min-Heap (deleteMin)

Now, we will see how the delete operation is executed in the heap. We do a lot of
insert and delete (removal from the data structure) operations especially when heap is
used in the priority queue. Normally the delete operation is a bit complex. However,
in this case, it is quite easy-to-do.
We want to write a function deleteMin(), which will find and delete the minimum
number from the tree.
Finding the minimum is easy; it is at the top of the heap.
See the heap below. It is a min-heap. Now the question arises where the minimum
number will be lying. It is clear from the definition of min-heap that it should be in
the root of the tree. So finding the minimum is very easy.
1

65

14

24

10

26

13

32

21

11

19

16

68

31

Fig 30.5
To understand it better, consider the tree form first, instead of thinking about the
array. Remember, we have implemented the complete binary tree with the help of an
array for the sake of efficiency. It is not necessary to implement it this way. We can
also implement it with pointers. However, heap is normally implemented through
array.
Coming back to the deletion, it is necessary to understand that
Deletion (or removal) causes a hole which needs to be filled.

Page 362 of 519

CS301 Data Structures

___________________________________________________________________
deleteMin()

24

65

14

10

26

21

11

32

16

19

68

31

Fig 30.6
In the above figure, we have deleted the root node, resulting in a hole at the root
position. To maintain it as a complete binary tree, we have to fill this hole. Now, we
will identify the appropriate candidates to fill this hole. As the parent is the minimum
as compared to its right and left children, the appropriate candidates are both right and
left children of the root node. Both children will be compared and the smaller one will
take the vacant place. In the above figure, the children 14 and 16 are candidates for
the vacant position. Being the minimum, 14 will be put in the hole.
deleteMin()

14

65

24

10

26

32

21

11

19

16

68

31

Fig 30.7

In the Fig 30.7, we can see that the 14 has been placed at the root position. However,
the vacuum at the previous position of 14 has created a hole. To fill this hole, the
same logic is applied i.e. we compare both right and left children of the hole and let
the minimum take over the place. Here 24 and 21 are compared. Being the smaller
one, 21 is placed at the vacant position. This way we get the latest figure of the tree as
seen in the following figure. (Fig 30.8.)

Page 363 of 519

CS301 Data Structures

___________________________________________________________________
deleteMin()

21

24

65

14

10

26

11

32

16

19

68

31

Fig 30.8

The hole is now created at previous position of 21. This time, the children- 32 and 31
are compared and being smaller, 31 takes over the place and the tree becomes as
shown in the figure Fig 30.9.
deleteMin()

65

21

24

10

26

14

32

31

19

16

68

11

Fig 30.9
The hole has been transferred from top to the bottom. It has reached at such a position
where it can be deleted.

Page 364 of 519

CS301 Data Structures

___________________________________________________________________
deleteMin(): heap size is reduced by 1

65

14
3

21

24

31

19

16

68

10

26

32
Fig 30.10

While using array to store complete binary tree, we can free array element at the end.
From the above figure Fig 30.10, the last array index 11 is no more there and the node
has been deleted.
We saw that the data element was inserted at the bottom of the tree and moved
upwards while comparing it with the parents (following the i/2 scheme) until its
destination was found. Similarly, when we deleted a node, the hole was produced.
While following the definition of min-heap, the hole kept on moving from top to the
bottom (following the 2i or 2i+1 scheme).

Page 365 of 519

CS301 Data Structures

___________________________________________________________________

Building a Heap (buildHeap)

Now, we will see how a heap can be made and in which peculiar conditions, it should
be built. We may have the data (for example, numbers) on hand, needed to used to
construct the tree .These data elements may be acquired one by one to construct the
tree. A programmer may face either the situations. If we consider priority queue, it is
normally empty initially. Events are inserted in the queue as these are received. You
can also consider priority queue in another application where data can be inserted into
the queue or taken out at one time.
Lets say we have a large unsorted list of names and want to sort it. This list can be
sorted with the help of the heap sort algorithm. When we l construct the heap or
complete binary tree of the list, the smallest name (considering alphabet a as the
smallest) in the list will take the root node place in the tree. The remaining names will
take their places in the nodes below the root node according to their order.
Remember, we are not constructing binary search tree but a min-heap to sort the data.
After the construction of the min-heap from all the names in the list, we start taking
the elements out (deleting) from the heap in order to retrieve the sorted data. The first
element that is taken out would be the smallest name in the heap. The remaining heap
after taking out the deleted node will be reduced by one element in size and the next
name will be at the root position. It is taken out to further reduce the tree size by one.
This way, the continuation of the process of taking out the elements from the heap
will ultimately lead to a situation when there is no more node left in the tree.
While keeping in view the time consumption factor, can we find a better way than
this? We may change the data structure or algorithm for that. There are data structures
which are more efficient than heap. But at the moment, we are focusing the heap data
structure. We will see, what we can do algorithmically to improve the performance,
having all the data simultaneously to construct the tree.
Following are some of the important facts about building a heap.
Suppose we are given as input N keys (or items) to build a heap of the keys.
Obviously, this can be done with N successive inserts.
Each call to insert will either take unit time (leaf node) or log2N (if new key
percolates all the way up to the root).
The worst time for building a heap of N keys could be Nlog2N.
It turns out that we can build a heap in linear time.
Suppose we have a method percolateDown(p) that moves down the key in node p
downwards.
This is what happens in deleteMin.We have used the word keys in the first point.
This is the same concept as that of the database keys, used to identify information
uniquely. Using the example of telephone directory (also discussed in a number of
previous lectures), we see that it contains the person name (first, middle and last),
address (street address, district and postal code) and telephone number. All these
information together form a data record. This set of information fields (record) will
always be together. The key item here can be the unique name of the person or the
telephone number.
Consider that we have N number of items i.e. names. One way is to call insert for
every element one by one. For N elements, it will be N number of calls obviously.
Being an iterative process, this insertion technique can be executed in a loop fashion.
Lets talk about the time taken to insert a node in the tree. It depends on the value to
be inserted and the values already present in the tree. The number of exchanges can
be variable. Therefore, the insertion time is also variable. Each call will be a unit time
Page 366 of 519

CS301 Data Structures

___________________________________________________________________
(in case of leaf node) or log2N (if new key percolates all the way up to the root node).
For the whole tree, the worst time for N keys would be Nlog2N.
Next point tells about the possibility of reducing the time from Nlog2N and it can turn
out to be a linear time. This is our target and we want to achieve it. Lets talk about
the ways to achieve the linear time.
The word percolate above means something is going upward from downward. This
terminology is coming from coffee maker, where the hot water moves upwards. This
phenomenon happens at the time of insertion. On the contrary, in delete operation, the
hole moves towards the bottom of the tree. You must be remembering that we
discussed about insertion in the heap without making exchanges and by finding the
destination position first before making any change. Similarly in case of deletion, the
hole moves downward while the data moves upwards. Lets write a method for the
delete operation percolateDown(p). This method will find which child of the hole will
be moved upwards and the movement of hole to the bottom of the tree.
We will see, using this method, how can we built a heap that will be take lesser time
than Nlog2N.
Initial data (N = 15)

65

31

32

26

21

19

68

13

24

15

14

16

70

12

10

11

12

13

14

15

Fig 30.11

As shown in the figure Fig 30.11, we are given a list of 15 elements and asked to
construct a heap of them. The numbers are stored in an array, starting from index
(position) 1. This start of array from position 1 is to make the tree construction easier.
It will be clear in a moment. Contrary to the previous situation when array was used
to store heap or complete binary tree, we should now think what would be the picture
of complete binary tree out of it. It may seem complex. But actually it is quite easy.
You start from the very first element in the array i.e. the root of the tree. The children
of root are present at 2i and 2i+1, which in this case, are at positions 2(1) = 2 and
2(1)+1=3. The children nodes for the node at position 2 are at positions 2(2)=4 and
2(2)+1=5. Similarly, the children nodes for the node at position 3 are at positions 6
and 7. Apply this logic to the whole of this array and try to construct a tree yourself.
You should build a tree as given in the figure Fig 30.12.

Page 367 of 519

CS301 Data Structures

___________________________________________________________________
Initial data (N = 15)

65

26

13

31

10

24

21

11

15

19

13

12

14

32

16

14

68

70

15

65

31

32

26

21

19

68

13

24

15

14

16

70

12

10

11

12

13

14

15

12

Fig 30.12

Is this tree binary one? Yes, every node has only two left and right children.
Is it a complete binary tree? It surely is as there is no node that has missing left or
right child.
The next question is; is it a min-heap. By looking at the root node and its children, as
the root node is containing 65 and the children nodes are containing 31 and 32, you
can abruptly say that no, it is not a min-heap.
How can we make min-heap out of it? We may look for percolateDown(p).method to
convert it into a min-heap. As discussed above, the percolateDown(p).moves the node
with value p down to its destination in the min-heap tree orders. The destination of a
node, can be its next level or maximum the bottom of the tree (the leaf node). The
node will come down to its true position (destination) as per min-heap order and other
nodes (the other data) will be automatically moving in the upward direction.
The general algorithm is to place the N keys in an array and consider it to be an
unordered binary tree.
The following algorithm will build a heap out of N keys.
for( i = N/2; i > 0; i-- )
percolateDown(i);
A close look on the above loop shows that it is starts from N/2 position and goes
down to 0. The loop will be terminated, once the position 0 is reached. Inside, this
loop is a single function call percolateDown(i). We will try to understand this loop
using figures.

Page 368 of 519

CS301 Data Structures

___________________________________________________________________
i = 15/2 = 7

65

31

26

24

21

10

13

Why i = n/2?

11

15

19

14

13

12

14

32

16

68

15

70

12

i
0

65

31

32

26

21

19

68

13

24

15

14

16

70

12

10

11

12

13

14

15

Fig 30.13

You can see the top left of the figure Fig 30.13. It is given i = 15/2 = 7. This is the
initial value of i .The value in the array at position 7 is 68. In our discussion, we will
interchangeably discuss array and binary tree. The facts arising in the discussion
would be applicable to both. Considering this position 7, we will try to build minheap below that. We know that for position 7, we have children at positions 2(7)=14
and 2(7)+1=15. So as children of the number 68 (which is the value at position 7), we
have 70 and 12. After applying the percolateDown(i), we get the tree as shown in Fig
30.14.
i = 15/2 = 7

65

26

13

31

10

24

21

11

15

19

13

12

14

32

16

14

12

70

15

68

i
0

65

31

32

26

21

19

12

13

24

15

14

16

70

68

10

11

12

13

14

15

Fig 30.14

You can see in the figure that 12 has moved upward and 68 has gone down. Now,
what about this little tree starting from 12 and below it is min-heap. Surely, it is.
Next, we go for second iteration in the loop, the value of i is decremented and it
becomes 6.
Page 369 of 519

CS301 Data Structures

___________________________________________________________________
65

26

13

31

10

24

21

11

15

19

14

13

12

14

32

15

70

16

12

68

i
65

31

32

26

21

19

12

13

24

15

14

16

70

68

10

11

12

13

14

15

Fig 30.15

The node at position 6 is 19. Here the method percolateDown(i) is applied and we get
the latest tree as shown in Fig 30.16.
i=5

65

26

13

31

10

24

21

11

15

13

12

14

32

16

14

19

12

70

15

68

i
0

65

31

32

26

21

12

13

24

15

14

16

19

70

68

10

11

12

13

14

15

Fig 30.16

The node 5 has come upward while the node 19 has moved downward. Its value of
has decremented. It is now ready for next iteration. If we see the positions of i in the
tree, we can see that it is traveling in one level of the tree, which is the second last
level.
The question might have already arisen in your minds that why did we start i by N/2
instead of N. You think about it and try to find the answer. The answer will be given
in the next lecture. We will continue with this example in the next lecture. You are
strongly encouraged to complete it yourself before the next lecture.

Page 370 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 31
Reading Material
Data Structures and Algorithm Analysis in C++
6.3.3, 6.3.4

Chapter. 6

Summary

BuildHeap
Other Heap methods
C++ Code

BuildHeap
In the previous lecture, we discussed about the BuildHeap method of heap abstract
data structure. In this handout, you are going to know why a programmer adopts this
method. Suppose we have some data that can be numbers or characters or in some
other form and want to build a min-Heap or max-Heap out of it. One way is to use the
insert() method to build the heap by inserting elements one by one. In this method, the
heap property will be maintained. However, the analysis shows that it is NlogN
algorithm i.e. the time required for this will be proportional to NlogN. Secondly, if we
have all the data ready, then it will be better to build a heap at once as it is a better
option than NlogN.
In the delete procedure, when we delete the root, a new element takes the place of
root. In case of min-heap, the minimum value will come up and the larger elements
will move downwards. In this regard, percolate procedures may be of great help. The
percolateDown procedure will move the smaller value up and bigger value down.
This way, we will create the heap at once, without calling the insert() method
internally. Lets revisit it again:

The general algorithm is to place the N keys in an array and consider it to be

an unordered binary tree.
The following algorithm will build a heap out of N keys.
for ( i = N/2; i > 0; i-- )
percolateDown(i);

Suppose, there are N data elements (also called as N Keys). We will put this data in
an array and call it as a binary tree. As discussed earlier, a complete binary tree can be
stored in an array. We have a tree in an array but it is not a heap yet. It is not
necessary that the heap property is satisfied. In the next step, we apply the algorithm.
Why did we start the i from N/2? Lets apply this algorithm on the actual data. We
will use the diagram to understand the BuildHeap. Consider the diagram given below:

Page 371 of 519

CS301 Data Structures

___________________________________________________________________

65

2 31

3 32

4 26

9 24 10 15

13

6 19

21

11 14

12 16

7 68

13 5 14 70

15 12

65 31 32 26 21 19 68 13 24 15 14 16 5 70 12
0

9 10 11 12 13 14 15

In the above diagram, there is an array containing the elements as 65, 31, 32, 26, 21,
19, 68, 13, 24, 15, 14, 16, 5, 70, 12. We have all the data in this array. The zeroth
element of the array is empty. We kept it empty intentionally to apply the 2i, 2i+1
scheme. We will take this array as a complete binary tree. Lets take the first element
that is 65, applying the 2i and 2i+1 formula. This element has 31 and 32 as its
children. Take the second element i.e. 32 and use the formula 2i and 2i+1. Its children
are 26 and 21. We can take the remaining elements one by one to build the tree. The
tree is shown in the above figure. This tree is not a min-heap as it has 65 as root
element while there are smaller elements like 13, 15, 5 as its children. So, this being a
binary tree has been stored in an array.
Lets think about the above formula, written for the heap building. What will be the
initial value of i. As the value of N is 15, so the value of i (i=N/2) will be 7 (integer
division). The first call to percolateDown will be made with the help of the value of i
as 7. The element at 7th position is 68. It will take this node as a root of subtree and
make this subtree a minimum heap. In this case, the left child of node 68 is 70 while
the right child of node 68 is 12. The element 68 will come down while the element 12
moves up.

Page 372 of 519

CS301 Data Structures

___________________________________________________________________

65

2 31

3 32

4 26

9 24 10 15

13

6 19

21

12 16

11 14

7 12

13 5 14 70

15 68

65 31 32 26 21 19 12 13 24 15 14 16 5 70 68
0

9 10 11 12 13 14 15

Look at this subtree with 12 as the root node. Its left and right children are smaller
than root. This subtree is a minimum heap. Now in the loop, the value of i is
decreased by 1, coming to the mark of 6. The element at 6th position is 19 that has
been passed to percolateDown method. It will convert this small tree into a minimum
heap. The element 19 is greater than both of its children i.e.16 and 5. One of these
will move up. Which node should move up? The node 5 will move up while node 19
will move down. In the next repetition, the value of i becomes 5, i.e. the element 21.
We apply the percolateDown() method again and convert it into minimum heap. Now
the subtree with node 26 will be converted into a minimum heap.

65

2 31

3 32

4 13

9 24 10 15

26

6 5

14

11 21

12 16

7 12

13 19 14 70

15 68

65 31 32 13 14 5 12 26 24 15 21 16 19 70 68
0

9 10 11 12 13 14 15
Page 373 of 519

CS301 Data Structures

___________________________________________________________________
Here the value of i is 3. The element at 3rd position is 32. You can see that the level of
the node has also been changed. We have moved one level up. Now the
percolateDown method will take this node as the root of the tree. It needs to move the
minimum value at this position. The minimum value is 5 so the values 32 and 5 are
going to be exchanged with each other. Now 32 becomes the parent of node 16 and
node 19 which are smaller values. Therefore, it will further move down to node 16.
The dotted line in the above figure shows the path of nodes that will be replaced. This
subtree has been converted into the min-heap.
Now, the value of i is 2 while the element is 31. It will change its position with 13
first and then with 24. So the node 31 will move down and become a leaf node. You
have seen that some nodes are moving up and some moving downwards. This
phenomenon is known as percolate. Therefore we have named this method as
percolateDown as it makes big values move down and smaller values upward.
Finally the value of i is 1. The element at first position is 65. As it is not the smallest
value, so it will move down and the smallest value will move up. It will move to the
position 7 as it is greater than 68 and 70. The node 5 will move to the top of the tree.
The final figure is shown below:

2 13

3 12

4 24

9 31 10 15

26

6 16

14

11 21

12 32

7 65

13 19 14 70

15 68

5 13 12 24 14 16 65 26 31 15 21 32 19 70 68
0

9 10 11 12 13 14 15

Is this a minimum heap? Does the heap property satisfy or not? Lets analyze this
tree. Start from the root of the tree. Is the value of the root is smaller than that of its
left and right children? The node 5 is smaller than the node 13 and node 12. Check
this on the next level. You will see that the values at each node are smaller than its
children. Therefore it satisfies the definition of the min-heap.
Now we have to understand why we have started the for loop with i = N/2. We can
start i from N. As i moves from level to level in the tree. If we start i from N, it will
Page 374 of 519

CS301 Data Structures

___________________________________________________________________
start from the right most leaf node. As long as i is l less than N/2 it will remain on leaf
nodes level. Is the leaf node alone is min-heap or not? Yes it is. As it does not have
any left or right child to compare that it is smaller or not. All the leaf nodes satisfy the
min-heap definition. Therefore we do not need to start with the leaf nodes. We can do
that but there is no reason to do that. For efficiency purposes we start i from N/2. This
is the one level up to the leaf node level. We applied the percolateDown method on
each node at this level. Due to this some nodes move down and some nodes move up.
We apply this on each level and reaches at the root. In the end we have a minimum
heap.
If we have data available, the use of BuildHeap method may be more appropriate as
compared to insert method. Moreover, BuildHeap method takes less time. How can
we prove that the BuildHeap method takes less time? Some mathematical analysis can
help prove it. As insert method is NlogN and BuildHeap method should be better than
that.

Other Heap Methods

Lets have a look on some more methods of heap and see the C++ codes.
decreaseKey(p, delta)
This method lowers the value of the key at position p by the amount delta. Since
this might violate the heap order, so it (the heap) must be reorganized with percolate
up (in min-heap) or down (in max-heap).
This method takes a pointer to the node that may be the array position as we are
implementing it as an array internally. The user wants to decrease the value of this
node by delta. Suppose we have a node with value 17 and want to decrease it by 10.
Its new value will be 10. By decreasing the value of a node, the heap order can be
violated. If heap order is disturbed, then we will have to restore it. We may not need
to build the whole tree. The node value may become smaller than that of the parents,
so it is advisable to exchange these nodes. We use percolateUp and percolateDown
methods to maintain the heap order. Here the question arises why we want to decrease
the value of some node? The major use of heap is in priority queues. Priority queues
are not FIFO or LIFO. The elements are taken out on some key value, also known as
priority value. Suppose we have some value in the priority queue and want to
decrease its priority. If we are using heap for the priority queue, the priority of the
some elements can be decreased so that it could be taken out later and some other
element that now has higher priority will be taken out first. You can find many real
life examples to understand why we need to increase or decrease the priority of
elements.
One of these examples is priorities of processes. You will read about this in the
operating system course. You can see the priorities of the processes using the task
manager. Task manager can be activated by pressing Ctrl+Alt+Del. Under the process
tab, we may see the priority of each process. Normally, there is one processor in a
computer and there is lot of processes running in it. How can we decide which
process should be given the CPU time and how much time should be given to this

Page 375 of 519

CS301 Data Structures

___________________________________________________________________
process? We will cover all of these topics in the operating system course. Suppose we
have 25 processes with different priorities. How can we decide which process should
be given the CPU time? We gave the CPU to the highest priority process for one
second. We can schedule this by using the priority queue. For some reason, we
decrease the priority of some process. Now its turn will come later. We may increase
the priority of some process so that it should be given the CPU time. Therefore, we
adopt the increase and decrease methods. If we are implementing the priority queue
by using the heap, the increase or decrease method of priority queue will use the
increase and decrease method of heap internally. Using this method, the turn of the
element will be according to the new priority.
increaseKey(p, delta)
This method is the opposite of decreaseKey. It will increase the value of the element
by delta. These methods are useful while implementing the priority queues using
heap.
remove(p)
This method removes the node at position p from the heap. This is done first by
decreaseKey(p, ) and then performing deleteMin(). First of all, we will decrease the
value of the node by and call the method deleteMin. The deleteMin method deletes
the root. If we have a min-heap, the root node contains the smallest value of the tree.
After deleting the node, we will use the percolateDown method to restore the order of
the heap.
The user can delete any node from the tree. We can write a special procedure for this
purpose. Here we will use the methods which are already available. At first, the
decreaseKey method will be called and value of node decreased by . It will result in
making the value of this node smallest of all the nodes. If the value is in integers, this
node will have the smallest integer. Now this node has the minimum value, so it will
become the root of the heap. Now we will call the deleteMin() and the root will be
deleted which is the required node. The value in this node is not useful for us. The is
a mathematical notation. It is not available in the C++. Actually we want to make the
minimum possible value of this node supported by the computer.

C++ Code
Now we will look at the C++ code of the Heap class. The objects of Heap may be got
from this factory class. This class will contain the methods including those discussed
earlier and some new ones. Heap is used both in priority queues and sorting.
We have some public and private methods in the class. Lets have a look on the code.
/* The heap class. This is heap.h file */
template <class eType>
class Heap
{
Page 376 of 519

CS301 Data Structures

___________________________________________________________________
public:
Heap( int capacity = 100 );
Void insert( const eType & x );
Void deleteMin( eType & minItem );
Const eType & getMin( );
Bool isEmpty( );
Bool isFull( );
int Heap<eType>::getSize( );
private:
int currentSize; // Number of elements in heap
eType* array; // The heap array
int capacity;
void percolateDown( int hole );
};
We may like to store different type of data in the heap like integers, strings, floatingpoint numbers or some other data type etc. For this purpose, template is used. With
the help of template, we can store any type of object in the heap. Therefore first of all
we have:
template <class eType>
Here eType will be used as a type parameter. We can use any meaningful name. Then
we declare the Heap class. In the public part of the class, there is a constructor as
given below.
Heap( int capacity = 100 );
We have a parameter capacity in the constructor. Its default value is 100. If we call it
without providing the parameter, the capacity will be set to 100. If we call the
constructor by providing it some value like 200, the capacity in the Heap object will
be 200.
Next we have an insert method as:
void insert( const eType & x );
Here we have a reference element of eType which is of constant nature. In this
method, we will have the reference of the element provided by the caller. The copy of
element is not provided through this method. We will store this element in the Heap.
Similarly we have a delete method as:
void deleteMin( eType & minItem );

Page 377 of 519

CS301 Data Structures

___________________________________________________________________
This method is used to delete the element from the heap.
If we want to know the minimum value of the heap, the getMin method can be useful.
The signatures are as:
const eType & getMin( );
This method will return the reference of the minimum element in the Heap. We have
some other methods in the Heap class to check whether it is empty or full. Similarly,
there is a method to check its size. The signatures are as:
bool isEmpty( );
bool isFull( );
int Heap<eType>::getSize( );
When will Heap become full? As we are implementing the Heap with the help of an
array, the fixed data type, so the array may get full at some time. This is the
responsibility of the caller not to insert more elements when the heap is full. To
facilitate the caller, we can provide the methods to check whether the heap is full or
not. We can call the getSize() method to ascertain the size of the Heap.
In the private part, we have some data elements and methods. At first, we have
currentSize element. It will have the number of elements in the heap. Then we have
an array of eType. This array will be dynamically allocated and its size depends on the
capacity of the Heap. We have one private method as percolateDown.
This is our .h file, now we will see the implementation of these methods that is in our
.cpp file. The code is as under:
/* heap.cpp file */
#include "Heap.h
template <class eType>
Heap<eType>::Heap( int capacity )
{
array = new eType[capacity + 1];
currentSize=0;
}
/* Insert item x into the heap, maintaining heap order. Duplicates
are allowed. */
template <class eType>
bool Heap<eType>::insert( const eType & x )
{
if( isFull( ) ) {
cout << "insert - Heap is full." << endl;
return 0;

Page 378 of 519

CS301 Data Structures

___________________________________________________________________
}
// Percolate up
int hole = ++currentSize;
for(; hole > 1 && x < array[hole/2 ]; hole /= 2)
array[ hole ] = array[ hole / 2 ];
array[hole] = x;
}
template <class eType>
void Heap<eType>::deleteMin( eType & minItem )
{
if( isEmpty( ) ) {
cout << "heap is empty. << endl;
return;
}
minItem = array[ 1 ];
array[ 1 ] = array[ currentSize-- ];
percolateDown( 1 );
}
We include the heap.h file before having the constructor of the heap. If we do not
provide the value of capacity, it will be set to 100 that is default. In the constructor,
we are dynamically creating our array. We have added 1 to the capacity of the array
as the first position of the array is not in use. We also initialize the currentSize to zero
because initially Heap is empty.
Next we have insert method. This method helps insert item x into the heap, while
maintaining heap order. Duplicates are allowed. Inside the insert method, we will call
the isFull() method. If the heap is full, we will display a message and return 0. If the
heap is not full, we will insert the element in the heap and take a variable hole and
assign it currentSize plus one value. Then we will have a for loop which will be
executed as long as hole is greater than 1. This is due to the fact that at position one,
we have root of the array. Secondly, the element which we want to insert in the array
is smaller than the array[hole/2]. In the loop, we will assign the array[hole/2] to
array[hole]. Then we divide the hole by 2 and again check the loop condition. After
exiting from the loop, we assign x to the array[hole]. To understand this insert
process, you can get help from the pictorial diagrams of the insert, discussed earlier.
We have a complete binary tree stored in an array and placed this new element at the
next available position in the array and with respect to tree it was left most leaf node.
This new node may have smaller value so it has to change its position. If we are
building the min-heap, this value should be moved upward. We will compare this
with the parent. If the parent is bigger, it will move down and child will move up.
Using the array notation, the parent of a child is at i/2 if child is at i position. We will
first try to find the final position of this new node and exchange the values. You might
want to remember the example of insert discussed in some earlier lecture. We have
shown a hole going up. Now the swapping function generally contains three
statements and is not an expensive operation. But if we perform swapping again and
again, it may cost some time. Therefore we will first find the final position of the new

Page 379 of 519

CS301 Data Structures

___________________________________________________________________
node and then insert it at that position. In the for loop we are finding the final
position of the node and the hole is moving upwards. In the statement array[hole] =
array[hole/2]; we are moving the parent down till the time final position of the new
node is achieved. Then we insert the node at its final position. It is advisable to
execute this code only for actual data and see how it works.
The next method is deletMin. First of all, it calls the isEmpty() method. If heap is
empty, it will display a message and return. If the heap is not empty, it will delete the
node. As the minimum value lies at the first position of the array, we save this value
in a variable and store the currentSize at the first position of the array. At the same
time, we reduce the currentSize by one as one element is being deleted. Then we call
the percolateDown method, providing it the new root node. It will readjust the tree.
We put the last element of the array which is the right most leaf node of the tree as the
root element will be deleted now. With this operation, the heap order can be
disturbed. Therefore we will call the percolateDown method which has the ability to
readjust the tree. This method will make the tree as min-heap again. There are some
more methods which we will cover in the next lecture.

Page 380 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 32
Reading Material
Data Structures and Algorithm Analysis in C++
6.3

Chapter. 6

Summary

perculateDown Method
getMin Method
buildHeap Method
buildHeap in Linear Time
Theorem

In the previous lecture, we defined a heap class and discussed methods for its
implementation. Besides, the insert and delete procedures were talked about. Here, we
will continue with discussion on other methods of the heap class. C++ language code
will be written for these procedures so that we could complete the heap class and use
it to make heap objects.
We are well aware of the deleteMin method. In this method, array[1] element ( a root
element) is put in the minItem variable. In the min heap case, this root element is the
smallest among the heap elements. Afterwards, the last element of the array is put at
the first position of the array. The following two code lines perform this task.
minItem = array[ 1 ];
array[ 1 ] = array[ currentSize-- ];

perculateDown Method
Then we call the perculateDown method. This is the same method earlier employed in
the build heap process. We passed it the node, say i, form where it starts its function
to restore the heap order. Lets look at this perculateDown method. It takes the array
index as an argument and starts its functionality from that index. In the code, we give
the name hole to this array index. Following is the code of this method.
// hole is the index at which the percolate begins.
template <class eType>
void Heap<eType>::percolateDown( int hole )
{
int child;
eType tmp = array[ hole ];
for( ; hole * 2 <= currentSize; hole = child )
{
child = hole * 2;
if( child != currentSize && array[child+1] < array[ child ] )
Page 381 of 519

CS301 Data Structures

___________________________________________________________________
child++; // right child is smaller
if( array[ child ] < tmp )
array[ hole ] = array[ child ];
else break;
}
array[ hole ] = tmp;
}
In the code of this function, it declares an integer variable named child before putting
the value at the index hole of the array in tmp variable. It is followed by a for loop,
the core of this method. The termination condition of this for loop is hole * 2 <=
currentSize; We know the 2i (or i * 2) formula that gives us index position of the left
child of i. In the code given above, we are using hole instead of i. This variable i.e.
hole is the loop variable. In other words, the termination condition means that the
loop will continue as long as the left child (hole * 2) is less than or equal to the
current size (currentSize) of the heap. Before looking at the third condition in which
we change the value of loop variable, lets have a view of the iteration of for loop.
In the body of the for loop, we assign the value hole * 2 to the child variable. The
left child of an index i is located at the index 2i. So it is obvious that the child has the
index number of the left child of hole. Then there is an if statement. In this statement,
we check two conditions combined with && (logical AND) operator. This if
statement returns TRUE only if both the conditions are TRUE. The first condition in
this if statement checks that the child (index) should not be equal to the currentSize
while in the second condition, we compare the value at index child + 1 with the value
at index child. We check whether the value at child + 1 is less than the value at child.
The value at index child is the left child of its parent as we have found it by the
formula of 2i (here we used hole * 2). Thus the value at child + 1 will be the right
child of the parent (i.e. hole). Actually, we are comparing the left and right children of
hole. If both the conditions are TRUE, it means that right child is less than the left
child. Resultantly, we increment the child index by 1 to set it to the right child. Now
again in an if statement, we compare the value at index child (which is left or right
child depending upon our previous check in if statement) with the tmp value. If value
at child is less than the tmp value, we will put the value of index child at the index
hole. Otherwise, if tmp value is less than the child, we exit for loop by using the break
statement. Thus the value of child comes to the position of the parent (i.e. hole). The
hole gets downward and child goes upward. The for loop continues, bringing the
hole downward and setting it to its proper position. When the hole reaches its
position, the for loop exits (it may exit by meeting the break statement) and we put
the tmp value in the array at the index hole. In this method, at first, the final position
of hole is determined before putting value in it. Here the perculateDown procedure
ends.
Now the question arises why we are bringing the hole (index) downward? Why dont
we exchange the values? We can execute the process of exchanging values. We will
use swap procedure to do this exchange. The swap process is carried out in three
statements. We have to do this swapping in the for loop. It is time-consuming process.
Contrast to it, in the perculateDown method, we execute one statement in for loop
instead of three swap statements. So it is better to use one statement instead of three
statements. This increases the efficiency with respect to time.

Page 382 of 519

CS301 Data Structures

___________________________________________________________________

getMin Method
We discussed this method in the previous lectures while talking about the interface of
heap. Under this method, the minimum value in the heap is determined. We just try
that element and do not remove it. It is like the top method of stack that gives us the
element on the top of the stack but do not remove it from the stack. Similarly, the
getMin method gives us the minimum value in the heap, which is not deleted from the
heap. This method is written in the following manner.
template <class eType>
const eType& Heap<eType>::getMin( )
{
if( !isEmpty( ) )
return array[ 1 ];
}
Now we will discuss the buildHeap method.

buildHeap Method
This method takes an array along with its size as an argument and builds a heap out of
it. Following is the code of this method.
template <class eType>
void Heap<eType>::buildHeap(eType* anArray, int n )
{
for(int i = 1; i <= n; i++)
array[i] = anArray[i-1];
currentSize = n;
for( int i = currentSize / 2; i > 0; i-- )
percolateDown( i );
}
In the body of this method, at first, there is a for loop that copies the argument array
(i.e. anArray) into our internal array, called array. We do this copy as the array that
this method gets as an argument starts from index zero. But we want that the array
should start from index 1 so that the formula of 2i and 2i +1 could be used easily. In
this for loop, we start putting values in the internal array from index 1. Afterwards,
we set the currentSize variable equal to the number of the elements got as an
argument. Next is the for loop that we have already seen while discussing the build of
heap. In the previous lecture, we used N for number of elements but here, it will be
appropriate to use the currentSize, which is also the number of elements. We start this
loop from the value i = currentSize / 2 and decrement the value of i by 1 for each
iteration and execute this loop till the value of i is greater than 0. In this for loop, we
call the method perculateDown to find the proper position of the value given as an
argument to this method. Thus we find the position for each element and get a heap.
Then there are three small methods. The isEmpty method is used to check whether the
heap is empty. Similarly the isFull method is used to check if the heap is full. The
getSize method is used to get the size of the heap. These three methods i.e. isEmpty,
isFull ans getSize are written below.

Page 383 of 519

CS301 Data Structures

___________________________________________________________________
//isEmpty method
template <class eType>
bool Heap<eType>::isEmpty( )
{
return currentSize == 0;
}
//isFull method
template <class eType>
bool Heap<eType>::isFull( )
{
return currentSize == capacity;
}
//getSize method
template <class eType>
int Heap<eType>::getSize( )
{
return currentSize;
}

buildHeap in Linear Time

We have seen that buildHeap takes an array to make a heap out of it. This process of
making heap (buildHeap algorithm) works better than N log2N. We will prove it
mathematically. Although our course is Data Structures, yet we have been discussing
the efficiency of the data structures from the day one of this course. We talked about
how efficiently a data structure uses memory storage and how much it is efficient
with respect to time. In different cases, mathematics is employed to check or compare
the efficiency of different data structures.
Here, we will show that the buildHeap is a linear time algorithm and is better than N
log2N algorithm which is not of linear nature. Linear algorithm is such a thing that if
we draw its graph, it will be a straight line with some slope. Whereas the graph of a
non-linear algorithm will be like a curve. We will discuss it in detail later. To prove
the superiority of the buildHeap over the N log2N, we need to show that the sum of
heights is a linear function of N (number of nodes). Now we will a mathematical
proof that proves that the buildHeap algorithm is better than N log2N. Now consider
the following theorem.

Theorem

According to this theorem, For a perfect binary tree of height h containing 2h +1 1

nodes, the sum of the heights of nodes is 2h +1 1 (h +1), or N-h-1.
This theorem is about heights of tree. Lets try to understand the concept of height. If
we start from the root node and go to the deepest leaf node of the tree, the number of
links we pass to go to the deepest leaf node is called the height of the tree. We have
been using the term depth and level for it also. Height can also be measured with
reference to the deepest leaf node at height zero and going upward to the root node.
Normally we start from the root and go to the deepest level to determine the height. A
perfect binary tree of height h has total number of nodes N equal to 2h +1 1. Now

Page 384 of 519

CS301 Data Structures

___________________________________________________________________
according to the theorem, the sum of heights of nodes will be equal to N - h 1. Lets
prove it mathematically.
In a perfect binary tree, there are 20 nodes at level 0, 21 nodes at level 1, 22 nodes at
level 2 and so on. Generally, level n has 2n nodes. Now in terms of height, in a perfect
binary tree, if h is the height of the root node, there are 20 nodes at height h, 21 nodes
at height h-1, 22 nodes at height h-2 and so on. In general, there are 2i nodes at height
h-i. Following figure shows a perfect binary tree with the number of nodes at its
heights.

h : 20 nodes

h -1: 21 nodes

h -2: 22 nodes

h -3: 23 nodes

Figure 32.1: Perfect Binary Tree

In the above figure, we can see that there are 21 nodes at height h-1and 23 nodes at
height h-3. The same is true about the number of nodes at other heights.
By making use of this property, we find the sum of heights of all the nodes
mathematically in the following manner. Suppose S is the sum of the heights of all the
nodes. Then
S = 2i ( h - i ), for i = 0 to h 1
(1)
S = h + 2(h-1) + 4(h-2) + 8(h-3)+ .. + 2h-1
Now multiplying by 2 the both sides of equation
2S =
2h + 4(h-1) + 8(h-2) + 16(h-3)+ .. + 2h (2)
Now subtracting the equation 2 from equation 1
-S = h 2 4 8 16 2h-1 2h
S = -h + 2 + 4 + 8 + 16+ .. + 2h-1 +2h
S = (2h+1 1) - (h+1)
Here S is the sum of heights. Thus it proves the theorem.
As stated earlier, in a perfect binary tree of height h, the total number of nodes N is
(2h+1 1). So by replacing it in the above equation, we get S = N ( h + 1).
Since a binary complete tree has nodes between 2h and 2h+1, the equation
S = (2h+1 1) - (h+1)
Can also be written as
Page 385 of 519

CS301 Data Structures

___________________________________________________________________
S ~ N Log2 (N+1)
In this equation, N stands for (2h+1 1) while Log2 (N+1) has taken the place of
(h+1). This equation is in term of N (number of nodes) and there is no h i.e. height in
this equation. Moreover, it also shows that with N getting larger, the log2(N +1) term
becomes insignificant and S becomes a function of N.
Suppose if N has a value 1000000, the value of log2 (1000000) will be approximately
20. We can see that it is insignificant with the value of N i.e. 1000000. Thus we can
say that S is approximately equal to N. This insignificance of log N shows that when
we build a heap from N elements, the values move up and down to maintain heap
order. Thus in buildHeap case, the values traverse the tree up and down. Now the
question arises what will be the maximum number of these traversals? This means
that if every node has to go up and down, what will be the maximum number of level
or height. We have proved that the sum of heights (i.e. S) is approximately equal to
the total number of nodes (i.e. N) as N becomes larger. Now in the buildHeap, the
maximum number of up down movement of values is actually the sum of heights i.e.
S. This S is the upper limit of number of movements in buildHeap. As this S is equal
to N, so this upper limit is N and there is not any log term involved in it. Thus
buildHeap is a linear time application.
Now lets prove the previous theorem with a non-mathematical method. We know
that the height of a node in the tree is equal to the number of edges on the longest
downward path to a leaf. So if we want to find the height of a node, it will be
advisable to go to the deepest leaf node in the tree by following the links (edges) and
this path (the number of links traveled) to the leaf node is the height of that node. The
height of a tree is the height of its root node. After these definitions, consider the
following two points that prove the theorem.

For any node in the tree that has some height h, darken h tree edges
Go down the tree by traversing left edge then only right edges.
There are N 1 tree edges, and h edges on right path, so number of darkened
edges is N 1 h, which proves the theorem.

We will explain these two points with the help of figures.

Consider the perfect binary tree, shown in the figure below.

Page 386 of 519

CS301 Data Structures

___________________________________________________________________

Figure 32.2: Marking the left edges for height 1 nodes

There are no values shown in the nodes. We are concerned with the heights of nodes
while the values do not matter. We mark the left edges of the nodes having height 1.
The leaf nodes are at height 0, so the one level above nodes has height 1, the left
edges of which are marked (shown light gray in above figure).
Now we mark the first left edge and the subsequent right edge up to the leaf node for
each node at height 2. These marked edges are shown in the figure below with gray
color.

Figure 32.3: Marking the first left edge and the subsequent
right edge for height 2 nodes

Similarly we go to the nodes at height 3, mark the first left edge and the subsequent
right edge up to the leaf node. Thus we mark one left and then two right edges for the
nodes at height 3. This is shown in the following figure. The edges marked in this step
are seen as the dark lines.

Page 387 of 519

CS301 Data Structures

___________________________________________________________________

Figure 32.4: Marking the first left edge and the subsequent two right
edges for height 3 nodes

Now we reach at the root whose height is 4. We mark the first left edge and the
subsequent three right edges for this node. This is shown in the figure below.

Figure 32.5: Marking the first left edge and the subsequent three
right edges for height 4 nodes
Now consider the following figure. In this figure, we show all the marked edges (that
we have marked in the previous steps) in gray and the non-marked edges are shown
with dotted lines.

Page 388 of 519

CS301 Data Structures

___________________________________________________________________

Figure 32.6: Marked and unmarked edges

The marked links are the ones through which the data can move up and down in the
tree. We can move the data in any node at a height to a node at other height following
the marked links. However, for the movement of the data in the root node to right side
of it, we can have opposite of the above figure. The opposite of the above figure can
be drawn by symmetry of the above steps. That means we first mark the right edge
and then the subsequent left edges. This will give us the figure of marked links in
which we can move to the right subtree of root.
Now lets sort out different aspects of the tree shown in above figure. We know that
in case of tree of N nodes, there will be N 1. Now in the tree above there are 31
nodes that means n = 31, so the number of edges is 31 - 1 = 30. The height of the tree
is 4. Height is represented by the letter H. so H = 4. The number of dotted edges (that
were not marked) in the tree is 4 that is the same as the height of the tree. Now we put
these values in the formula for the sum of the heights of the nodes. We know the
formula i.e.
S=NH1
By putting values in this formula, we get
S = 31 4 1 = 26
If we count the darkened edges (marled links) in the above tree, it is also 26 that is
equal to the sum of heights. Thus with the help of these figures, the theorem earlier
proved mathematically, is established by a non-mathematical way.

Page 389 of 519

CS301 Data Structures

___________________________________________________________________
Deleted: 36
Deleted: In
Deleted: discussed a lot about
buildHeap.

Data Structures

Deleted: If we have all the data

available then

Lecture No. 33
Reading Material
Data Structures and Algorithm Analysis in C++
6.3, 8.1

Chapter. 6, 8

Summary

Priority Queue Using Heap

The Selection Problem
Heap Sort
Disjoint Set ADT
Equivalence Relations

Priority Queue Using Heap

As discussed in the previous lecture, we generally prefer to employ the buildHeap to
construct heap instead of using insert if we have all the required data. buildHeap is
optimized as compared to insert method as it takes lesser time than Nlog2N . In the
previous discussion, we had proved a theorem that if the number of links in the tree
are counted, the sum of all is N-h-1. We have been iterating a lot of times that the best
use of heap is priority queues implementation. Lets try to develop a class of priority
queue with the help of a heap. You are very familiar with the concept of bank
simulation. While explaining the bank simulation, we had used a priority queue that
was implemented by using an array. Now, we will implement the same queue with the
help of a heap. For this purpose, the code will be modified so that heap is used in
place of array. The interface (.h file) will remain the same. However, the
implementation (.cpp file) will be changed. Lets see the following cpp code, which
also shows the change in the .cpp file:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

#include Event.cpp
#include Heap.cpp
#define PQMAX 30
class PriorityQueue
{
public:
PriorityQueue()
{
heap = new Heap <Event> ( PQMAX );
};
~PriorityQueue()
{

Page 390 of 519

Deleted: we can use buildHeap to

construct heap. buildHeap is optimized as
compared to insert method because it
takes lesserthan Nlog2N time. We also
proved a theorem that if the number of
linkes in the tree are counted then the sum
of heights of the nodes is N-h-1. We have
been iteratinglot of times that the best use
of heap is implementing priority queues.
Lets try to develop a class of priority
queue using heap. You might be
remembering, when we did bank
simulation, weused a priority queue but
that was implementedusing an array.
Now, we will implement the same queue
using a heap. We will modify the same
code to use heap instead of any array. The
interface (.h file) will remain the same but
the implementation (.cpp file) will be
changed. As .cpp file will change, so lets
see the .cpp code below:

The first line is including a file Event.cpp,

this is containing all the events used for
simulation. We are including .cpp files
here as we have used templates of C++.
At the second line, we have included
heap.cpp. In the third line, we are
defining a constant PQMAX, declaring
the maximum size of the priority queue to
be 30. In line 4, class PriorityQueue is
declared. public keyword is given at line
6 indicating that all class members below
will be of public scope. Line 7 is starting
the class constructors definition. At line
9, a new heap object is being created, this
object is a collection of Event type
objects and the number of elements in the
collection is PQMAX. The address
(pointer) of the newly created heap is
stored in the heap object pointer. The
heap is a private pointer variable of
PriorityQueue class. Next comes the
destructor of the class, where we are
deleting (deallocating the allocated
resources) the pointer variable heap. Next
isthe remove method that was sometimes,
named as dequeue in our previous
lectures. remove method is returning an
Event pointer. Inside remove, the first
statement at line 19 is an if-condition;
which is checking whether heap is not
empty. If the condition is true (i.e. the
heap is not empty), a local variable (local
to the block) Event* variable e is
declared. In the next line at line 22, we
are calling deleteMin method to delete
(remove) an element from the heap, the
removed element is assigned to the
passed in parameter pointer variable e.
You might have noticed already that in
this version of deleteMin, the parameter is
being passed by pointer, in the previous
implementation we used reference. In
... the
[1]

CS301 Data Structures

___________________________________________________________________
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
1.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49. };

delete heap;
};

Formatted: Bullets and Numbering

Event * remove()
{
if( !heap->isEmpty() )
{
Event * e;
heap->deleteMin( e );
return e;
}
cout << "remove - queue is empty." << endl;
return (Event *) NULL;
};

Formatted: Bullets and Numbering

int insert(Event * e)
{
if( !heap->isFull() )
{
heap->insert( e );
return 1;
}
cout << "insert queue is full." << endl;
return 0;
};

Formatted: Bullets and Numbering

int full(void)
{
return heap->isFull();
};
int length()
{
return heap->getSize();

Formatted: Bullets and Numbering

};

Formatted: Bullets and Numbering

The first line has a file Event.cpp that contains all the events used for simulation. We
are including .cpp files here as done in case of templates of C++. In the second line,
there is heap.cpp while a constant PQMAX has been defined in third line, declaring
the maximum size of the priority queue to be 30. In line 4, class PriorityQueue is
declared. public keyword is given at line 6 indicating that all class members below
will be of public scope. Line 7 starts the class constructors definition while in the line
9; a new heap object is being created. This object is a collection of Event type objects
and the number of elements in the collection is PQMAX. The address (pointer) of the
newly created heap is stored in the heap object pointer. The heap is a private pointer
variable of PriorityQueue class. Now there is the destructor of the class, where we

Page 391 of 519

CS301 Data Structures

___________________________________________________________________
are deleting (deallocating the The first line is including a file Event.cpp, this is
containing all the events used for simulation. At the second line, we have included
heap.cpp. In the third line, we are defining a constant PQMAX, declaring the
maximum size of the priority queue to be 30. In line 4, class PriorityQueue is
declared. public keyword is given at line 6 indicating that all class members below
will be of public scope. Line 7 is starting the class constructors definition. At line 9,
a new heap object is being created, this object is a collection of Event type objects and
the number of elements in the collection is PQMAX. The address (pointer) of the
newly created heap is stored in the heap object pointer. The heap is a private pointer
variable of PriorityQueue class. Next comes the destructor of the class, where we are
deleting (deallocating the allocated resources) the pointer variable heap. Next is the
remove method that was sometimes, named as dequeue in our previous lectures.
remove method is returning an Event pointer. Inside remove, the first statement at line
19 is an if-condition; which is checking whether heap is not empty. If the condition is
true (i.e. the heap is not empty), a local variable (local to the block) Event* variable e
is declared. In the next line at line 22, we are calling deleteMin method to delete
(remove) an element from the heap, the removed element is assigned to the passed in
parameter pointer variable e. You might have noticed already that in this version of
deleteMin, the parameter is being passed by pointer, in the previous implementation
we used reference. In the next line, the retrieved value is returned, the function returns
by returning the pointer variable e. But if the heap was empty when checked at line
19, the control is transferred to the line 25. At line 25, a message is displayed to show
that the heap is empty and the next line returns a NULL pointer.
It is important to understand one thing that previously when we wrote the array based
remove method. We used to take an element out from the start of the array and shifted
the remaining elements to left. As in this implementation, we are using heap,
therefore, all the responsibilities of maintaining the heap elements lie with the heap.
When the deleteMin() is called, it returns the minimum number.
We now observe the insert method line by line. It is accepting a pointer type
parameter of type Event. As the heap may become full, if we keep on inserting
elements into it, so the first line inside the insert function is checking whether the
heap has gone full. If the heap is not full, the if-block is entered. At line 33 inside the
if-block, an element that is passed as a parameter to insert method of heap is inserted
into the queue by calling the insert method of heap. This insert call will internally
perform percolate up and down operations to place the new element at its correct
position. The returned value 1 indicates the successful insertion in the queue. If the
heap has gone full, a message is displayed i.e. insert queue is full. Note that we have
used the word queue, not heap in that message. It needs to be done this way.
Otherwise, the users of the PriorityQueue class are unaware of the internal
representation of the queue (whether it is implemented using a heap or an array). Next
line is returning 0 while indicating that the insert operation was not successful. Below
to insert, we have full() method. This method returns an int value. Internally, it is
calling isFull() method of heap. The full() method returns whatever is returned by the
isFull() method of heap. Next is length() method as the size of heap is also that of the
queue. Therefore, length() is internally calling the getSize() method of heap.
In this new implementation, the code is better readable than the PriorityQueues

Page 392 of 519

Formatted: Bullets and Numbering

CS301 Data Structures

___________________________________________________________________
implementation with array. While implementing the PriorityQueue with an array, we
had to sort the internal array every time at each insertion in the array. This new
implementation is more efficient as the heap can readjust itself in log2N times. Gain in
performance is the major benefit of implementing PriorityQueue with heap as
compared to implementation with array.
There are other significant benefits of the heap that will be covered in this course time
to time. At the moment, we will have some common example usages of heap to make
you clear about other uses of heap. The heap data structure will be covered in the
Algorithms course also.

The Selection Problem

Given a list of N elements (numbers, names etc.) which can be totally ordered and
an integer k, find the kth smallest (or largest) element.
Suppose, we have list of N names (names of students or names of motor vehicles or a
list of numbers of motor vehicles or list of roll numbers of students or id card
numbers of the students, whatever). However, we are confronting the problem of
finding out the kth smallest element. Suppose we have a list of 1000 numbers and
want to find the 10th smallest number in it. The sorting is applied to make the
elements ordered. After sorting out list of numbers, it will be very easy to find out any
desired smallest number.
th

One way is to put these N elements in an array and sort it. The k smallest of these
th
is at the k position.
It will take Nlog2N time ,in case we use array data structure. Now, we want to see if it
is possible to reduce the time from Nlog2N by using some other data structure or by
improving the algorithm? Yes, we can apply heap data structure to make this
operation more efficient.
A faster way is to put the N elements into an array and apply the buildHeap
algorithm on this array.
Finally, we perform k deleteMin operations. The last element extracted from the
heap is our answer.
The buildHeap algorithm is used to construct a heap of given N elements. If we
construct min-heap, the minimum of the N elements, will be positioned in the root
node of the heap. If we take out (deleteMin) k elements from the heap, we can get the
Kth smallest element. BuildHeap works in linear time to make a min or a max-heap.
The interesting case is k = N/2 as it is also is known as the median.
In Statistics, we take the average of numbers to find the minimum, maximum and
median. Median is defined as a number in the sequence where the half of the numbers
are greater than this number while the remaining half are smaller ones. Now, we can
come up with the mechanism to find median from a given N numbers. Suppose, we
want to compute the median of final marks of students of our class while the
maximum aggregate marks for a student are 100. We use the buildHeap to construct a
heap for N number of students. By calling deleteMin for N/2 times, the minimum
marks of the half number students will be taken out. The N/2th marks would be the
Page 393 of 519

Deleted: array,
Deleted: .
Deleted: Clearly then this new
implementation is more efficient because
the heap can readjusts itself in log2N
times. Gain in performance is the major
benefit of implementing PriorityQueue
using heap over implementing with array
and very important for you to remember
as a student of Data Structures.
There are other significant benefits of
using heap and there are number of other
uses of heap that would be covered in this
course time to time. At the moment, we
willsome common example usages of
heap in order to make you clear about
other uses of heap. The heap data
structure will be covered in the
Algorithms course also.

The Selection Problem

<#>Given a list of N elements (numbers,

names etc.), which can be totally ordered,
and an integer k, find the kth smallest (or
largest) element.

Suppose we have list of N names (names

of students or names of motor vehicles or
a list of numbers of motor vehicles or list
of roll numbers of students or id card
numbers of the students, whatever). The
problem is to findout the kth smallest
element. So lets say we have a list of
1000 numbers and we want to find the
10th smallest number in it. The sorting is
applied in order to make the elements
ordered. After sorting out list of numbers
(in any order ascending or decending,
lets do in ascending in this as we are
finding the smallest number), it will be
very easy to findany desired smallest
number.

<#>One way is to put these N elements in

th

an array and sort it. The k smallest of

th

these is at the k position.

It will take Nlog2N time in case we use

array data structure. Now, we want to see
if it is possible to reduce the time from
Nlog2N by using some other data
structure or by improving the algorithm?
Yes, we can apply heap data structure in
order to make this operation more
efficient.

<#>A faster way is to put the N elements

into an array and apply the buildHeap
algorithm on this array.
<#>Finally, we perform k deleteMin
operations. The last element extracted
from the heap is our answer.

The buildHeap algorithm is used to ... [2]

Formatted: Bullets and Numbering

Formatted: Bullets and Numbering
Formatted: Bullets and Numbering
Formatted: Bullets and Numbering

CS301 Data Structures

___________________________________________________________________
median of the marks of our class. The alternate methods are there to calculate median.
However, we are discussing the possible uses of heap. Lets see another use of heap.

Heap Sort

To take the 100th minimum element out from the min-heap, we will call deleteMin to
take out 1st element, 2nd element, 3rd element. Eventually we will call deleteMin 100th
time to take out our required 100th minimum element. Suppose, if the size of the heap
is 100 elements, we have taken out all the elements out from the heap. Interestingly
the elements are sorted in ascending order. If somehow, we can store these numbers,
lets say in an array, all the elements sorted (in ascending order in this min-heap case)
can be had.
Hence,
If k = N, and we record the deleteMin elements as they come off the heap. We will
have essentially sorted the N elements.

Formatted: Bullets and Numbering

Later in the course, we will fine-tune this idea to obtain a fast sorting algorithm
called heapsort.
We conclude our discussion on the heap here. However, it will be discussed in the
forthcoming courses. At the moment, let see another Abstract Data Type.

Formatted: Bullets and Numbering

Disjoint Set ADT

Before actually moving to an Abstract Data Type (ADT), we need to see what that
ADT is, how it works and in which situations it can be helpful. We are going to cover
Disjoint Set ADT. Firstly, we will have its introduction, examples and later the ways
of its implementation.
Suppose we have a database of people.
We want to figure out who is related to whom.Initially, we only have a list of
people, and information about relations is obtained by updating the form Haaris
is related to Saad.Lets say we have a list of names of all people in a locality but
are not aware of their relationships to each other. After having the list of all people,
we start getting some information about their relationships gradually. For example,
Ali Abbas is the son of Abbas.
The situation becomes interesting when we have relationships like Ali Abbas is first
cousin of Ayesha Ali (i.e. fathers of both are brothers) but Ayesha Ali has other
immediate cousins also from her mothers side. Therefore, other immediate cousins of
Ayesha Ali also get related to Ali Abbas despite the fact that they are not immediate
to him.
So as we keep on getting relationship details of the people, the direct and indirect
relationships can be established.

Formatted: Bullets and Numbering

Key property: If Haaris is related to Saad and Saad is related to Ahmad, then
Haaris is related to Ahmad.
See the key property lines first part above Harris is related to Saad and Saad is
related to Ahmad. Suppose we have a program to handle this list of people and their
relationships. After providing all the names Harris, Saad and Ahmad to that
program and their relationships, the program might be able to determine the
remaining part Harris related to Ahmad.

Formatted: Bullets and Numbering

Page 394 of 519

CS301 Data Structures

___________________________________________________________________
The same problem (the intelligence required in the program) is described in the
sentence below:
Once we have relationships information, it will be easy for us to answer queries
like Is Haaris related to Ahmad?
To answer this kind of queries and have that intelligence in our programs, Disjoint Set
ADT is used. Before going for more information about Disjoint Set ADT , we see
another application of this ADT in image analysis. This problem is known as Blob
Coloring.
Blob Coloring
A well-known low-level computer vision problem for black and white images is the
following:
Put together all the picture elements (pixels) that belong to the same "blobs", and give
each pixel in each different blob an identical label.
You must have heard of robots that perform certain tasks automatically. How do they
know from the images provided to them that in which direction should they move?
They can catch things and carry them. They do different things the way human beings
do. Obviously, there is a software working internally in robots body, which is doing
all this controlling part and vision to the robot. This is very complex problem and
broad area of Computer Science and Electrical Engineering called Robotics
(Computer Vision in particular).

Page 395 of 519

Formatted: Bullets and Numbering

CS301 Data Structures

___________________________________________________________________
Consider the image below:

``
Fig 33.1
This image is black and white, consisting of five non-overlapping black colored
regions of different shapes, called blobs. These blobs are two elliptics- n and u shaped
(two blobs) and one arc at the bottom. We can see these five blobs. How can robot
identify them? So the problem is:
We want to partition the pixels into disjoint sets, one set per blob.

Formatted: Bullets and Numbering

If we make one set per blob, there will be five sets for the above image. To
understand the concept of disjoint sets, we can take an analogy with two sets- A and B
(as in Mathematics) where none of the elements inside set A is present in set B. The
sets A and B are called disjoint sets.
Another problem related to the Computer Vision is the image segmentation problem.
See the image below on the left of an old ship in gray scales.

Fig 33.2
We want to find regions of different colors in this picture e.g. all regions in the picture
of color black and all regions in the image of color gray. The image on the right
represents the resultant image.
Different scanning processes carried out in hospitals through MRI (Magnetic

Page 396 of 519

CS301 Data Structures

___________________________________________________________________
Resonance Imaging), CAT Scan or CT Scan views the inner parts of the whole body
of human beings. These scanned images in gray scales represent organs of the human
body. All these are applications of Disjoint Set ADT.

Equivalence Relations
Lets discuss some Mathematics about sets. You might have realized that
Mathematics is handy whenever we perform analysis of some data structure.
A binary relation R over a set S is called an equivalence relation if it has
following propertiesReflexivity: for all element x S, x R x
2. Symmetry: for all elements x and y, x R y if and only if y R x
3. Transitivity: for all elements x, y and z, if x R y and y R z then x R z
The relation is related to is an equivalence relation over the set of people.
You are advised to read about equivalence relations yourself from your text books or
from the internet.

Page 397 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 34
Reading Material
Data Structures and Algorithm Analysis in C++
8.1, 8.2, 8.3

Chapter. 8

Summary

Equivalence Relations
Disjoint Sets
Dynamic Equivalence Problem

Equivalence Relations
We will continue discussion on the abstract data structure, disjointSets in this
lecture with special emphasis on the mathematical concept of Equivalence Relations.
You are aware of the rules and examples in this regard. Lets discuss it further and
define the concept of Equivalence Relations.
A binary relation R over a set S is called an equivalence relation if it has following
properties:
1. Reflexivity: for all element x S, x R x
2. Symmetry: for all elements x and y, x R y if and only if y R x
3. Transitivity: for all elements x, y and z, if x R y and y R z then x R z
The relation is related to is an equivalence relation over the set of people. This is an
example of Equivalence Relations. Now lets see how the relations among people
satisfy the conditions of Equivalence Relation. Consider the example of Haris, Saad
and Ahmed. Haris and Saad are related to each other as brother. Saad and Ahmed are
related to each other as cousin. Here Haris is related to Saad and Saad is related
to Ahmed. Lets see whether this binary relation is Equivalence Relation or not. This
can be ascertained by applying the above mentioned three rules.
First rule is reflexive i.e. for all element x S, x R x. Suppose that x is Haris so Haris
R Haris. This is true because everyone is related to each other. Second is Symmetry:
for all elements x and y, x R y if and only if y R x. Suppose that y is Saad. According
to the rule, Haris R Saad if and only if Saad R Haris. If two persons are related, the
relationship is symmetric i.e. if I am cousin of someone so is he. Therefore if Haris is
brother of Saad, then Saad is certainly the brother of Haris. The family relationship is
symmetric. This is not the symmetric in terms of respect but in terms of relationship.
The transitivity is: for all elements x, y and z. If x R y and y R z, then x R z. Suppose
x is Haris, y is Saad and z is Ahmed. If Haris is related to Saad, Saad is related to
Ahmed. We can deduce that Haris is related to Ahmed. This is also true in

Page 398 of 519

CS301 Data Structures

___________________________________________________________________
relationships. If you are cousin of someone, the cousin of that person is also related to
you. He may not be your first cousin but is related to you.
Now we will see an example of binary relationship that is not based on equivalence
relationship. The relationship is not an equivalence relation. We will prove this by
applying the three rules. The first rule is reflexive. It is reflexive, since x x, as x is
not less than x but surely is equal to x. Lets check the transitive condition. Since x y
and y z implies x z., it is also true. However it is not symmetric as x y does not
imply y x. Two rules are satisfied but symmetric rule does not. Therefore is not an
equivalence relation.
Lets see another example of binary relationship that is also equivalence relation. This
is from the electric circuit domain. Electrical connectivity, where all connections are
by metal wires, is an equivalence relation. You can make the circuit diagram on the
paper including transistor, resistors and capacitors etc. These parts are connected to
each other with the metal wire. Now lets apply the rules of equivalence relations on
it. It is clearly reflexive, since the component is connected to itself. In a circuit, a
transistor is connected to itself. It is symmetric due to the fact that if component a is
connected to component b, then b must be electrically connected to a. Suppose we
have two capacitors a and b. If capacitor a is connected to b, capacitor b is also
connected to a. It is also transitive. If component a is connected to component b and b
is connected to c, then a is connected to c. This is also true in electrical connectivity.
All the three rules of equivalence relations satisfy in this case. Therefore this is
equivalence relation.

Disjoint Sets
In the beginning of the lecture, it was told that equivalence relationship partitioned the
set. Suppose we have a set of elements and a relation which is also equivalence
relation. Lets see what it does mathematically to the set. An equivalence relation R
over a set S can be viewed as a partitioning of S into disjoint sets. Here we have an
equivalence relationship which satisfies the three conditions. Keep in mind the
examples of family relationships or electrical circuits. Here the second point is that
each set of the partition is called an equivalence class of R (all elements that are
related).
Consider the diagram below. We have a set in the shape of an ellipse.

This set has been partitioned into multiple sets. All these parts are disjoint sets and
belong to an equivalence class.
Page 399 of 519

CS301 Data Structures

___________________________________________________________________
Lets discuss an example to understand it. Suppose there are many people around you.
How can we separate those who are related to each other in this gathering? We make
groups of people who are related to each other. The people in one group will say that
they are related to each other. Similarly we can have many groups. So every group
will say that they are related to each other with family relationship and there is no
group that is related to any other group. The people in the group will say that there is
not a single person in the other group who is related to them. There is a possibility
that a boy from one group marries to a girl from the other group. Now a relation has
been established between these two groups. The people in both groups are related to
each other due to this marriage and become a bigger family. With the marriages
people living in different families are grouped together. We will do operations like
this in case of disjoint sets by combining two sets. You must be aware of the union
and intersection operations in sets.
Every member of S appears in exactly one equivalence class. We have divided a set
into many disjoint sets. One member of the set can appear in only one equivalence
class. Keeping in mind the example of family relations, if a person belongs to a group
he cannot go to some other group. The second point is to decide if a R b. For this
purpose, we need only to check whether a and b are in the same equivalence class.
Consider the example of pixels. If a pixel p1 is in a region and want to know the
relation of pixel p2 with it, there is only the need to confirm that these two pixels are
in the same region. Lastly, this provides a strategy to solve the equivalence problem.
With the help of second point we can get help to solve the equivalence problem. So
far, we have not seen the disjoint sets and its data structure.

Dynamic Equivalence Problem

Lets talk about the data structure. Suppose we have a set and converted into disjoint
sets. Now we have to decide if these disjoint sets hold equivalence relation or not.
Keep in mind the example of family relations in which we had divided the people into
groups depending upon their relations. Suppose we have to decide that two persons
belong to the same family or not. You have to make this decision at once. We have
given a set of people and know the binary relation between them i.e. a person is a
cousin of other, a person is brother of other etc. So the relation between two persons
is known to us. How we can take that decision? What is the data available to us? We
have people (with their names) and the binary relation among them. So we have
names and the pair of relations. Think that Haris and Ahmed are related to each other
or not i.e. they belong to the same family or not. How can we solve this? One way to
find out the relationship is given below.
If the relation is stored as a two-dimensional array of booleans, this can be done in
constant time. The problem is that the relation is usually not explicitly defined, but
shown in implicit terms. Suppose we are given 1000 names of persons and relations
among them. Here we are not talking about the friendship as it is not a family relation.
We are only talking about the family relations like son-father, brother-brother, cousincousin, brother-sister, etc. Can we deduce some more relations from these given
relations? Make a two dimensional array, write the names of persons as the columns
and rows headings. Suppose Haris is the name of first column and first row. Saad is
the name of 2nd col and 2nd row. Similarly Ahmed, Omar, Asim and Qasim are in the

Page 400 of 519

CS301 Data Structures

___________________________________________________________________
3rd, 4th, 5th, and 6th columns and rows respectively. The names are arranged in the
same way in columns and rows.
H
a
a
r
i
s
Haaris
Saad
Ahmed
Omar
Asim
Qasim

S
a
a
d

A
h
m
e
d

O
m
a
r

A
s
i
m

Q
a
s
i
m

T
T

Now visit each cell of matrix. If two persons are related, store T(true) in the
corresponding column and row. As Haris and Saad are related to each other, so we
take the row of Haris and column of Saad and store T in that cell. Also take the row of
Saad and column of Ahmed and mark it with T. By now, we do not know that Haris
and Ahmed are related so the entry in this regard will be false. We cannot write
T(True) here as this relationship has not be stated explicitly. We can deduce that but
cannot write it in the matrix. Take all the pairs of relations and fill the matrix. As
Haris is related to Haris so we will write T in the Haris row and Haris column.
Similarly, the same can be done in the Saad row and Saad column and so on. This will
be square matrix as its rows and columns are equal. All of the diagonal entries are T
because a person is related to himself. This is a self relation. Can we find out that
Haris and Ahmed are related? We can find the answer of this question with the help
of two relations between Haris & Saad and Saad & Ahmed. Now we can write T in
the Haris row and Ahmed column. The problem is that this information was not given
initially but we deduced this from the given information. We have 1000 people so the
size of the matrix will be 1000 rows * 1000 columns. Suppose if we have 100,000
people, then the size of the array will be 100,000*100,000. If each entry requires one
byte then how much storage will be required to store the whole array? Suppose if we
have one million people, the size will be 10 to the power 14. Do we have as much
memory available in the computer? We have one solution but it is not efficient in
terms of memory. There is always need of efficient solution in terms of space and
time. For having a fast algorithm, we can compromise on the space. If we want to
conserve memory, a slow algorithm can be used. The algorithm we discussed is fast

Page 401 of 519

CS301 Data Structures

___________________________________________________________________
but utilized too much space.
Lets try to find out some more efficient solution of this problem. Consider another
example. Suppose the equivalence relation is defined over the five-element set {a1,
a2, a3, a4, a5}. These elements can be the name of people, pixels, electrical
components etc. How many pairs we can have in this set? We have a pair a1-a1, a1a2, a1-a3, a1-a4, a1-a5, a2-a2, a2-a3 etc. The pair a1-a2 and a2-a1 are equal due to
symmetric. We also find some self-pairs (i.e. a1-a1, a2-a2 etc). There are 25 pairs of
elements, each of which is related or not (30 pairs 5 self-pairs = 25). These are the
total pairs and we may not have as much relations. What will be size of array with
five elements? The size will be 5*5 = 25. We are also given relations i.e.

a1 R a2,
a3 R a4,
a5 R a1,
a4 R a2,

We have five people in the example and their relation is given in four pairs. If two
persons are brothers, then cousin of one brother is also the cousin of other. We can
find out this information with the help of a matrix. We would like to be able to infer
this information quickly.
We made nodes of each element. These five nodes are as a1, a2, a3, a4, a5.

a1
a5

a2

a3

a4

As a1 R a2, so we link the nodes a1 and a2. Similarly a3 R a4, these two nodes are
also linked. Following this pattern, it is established that a5 R a1 and a4 R a2. So we
connect these nodes too. It is clear from the figure that all of these nodes are related to
each other. If they are related to each other as cousin, then all of these five persons
belong to the same family. They need to be in the same set. They are not in disjoint
sets. Is a3 R a5? You can easily say yes. How you get this information? This relation
is not provided in the given four relations. With the above tree or graph we can tell
that a3 is related to a5. We can get the information of relationship between different
persons using these nodes and may not need the matrix.
We want to get the information soon after the availability of the input data. So the
data is processed immediately. The input is initially a collection of n sets, each with
one element. Suppose if we have 1000 people, then there will be need of 1000 sets

Page 402 of 519

CS301 Data Structures

___________________________________________________________________
having only one person. Are these people related to each other? No, because every
person is in different set. This initial representation is that all relations (except
reflexive relations) are false. We have made 1000 sets for 1000 people, so only the
reflexive relation (every person is related to himself) is true. Now mathematically
speaking, each set has a different element so that Si Sj = which makes the sets
disjoint. A person in one set has no relation with a person in another set, therefore
there intersection is null. Now here we have 1000 sets each containing only one
person. Only the reflexive relation is true and all the 1000 sets are disjoint. If we take
intersection of any two sets that will be null set i.e. there is no common member in
them.
Sometimes, we are given some binary relations and are asked to find out the
relationship between two persons. In other words, we are not given the details of
every pair of 1000 persons. The names of 1000 persons and around 50 relations are
provided. With the help of these 50 relations, we can find out more relations between
other persons. Such examples were seen above while find out the relationship with the
help of graph.
There are two permissible operations in these sets i.e. find and union. In the find
method, we are given one element (name of the person) and asked to find which set it
belongs to. Initially, we have 1000 sets and asked in which set person 99 is? We can
say that every person is in a separate set and person 99 is in set 99. When we get the
information of relationships between different persons, the process of joining the sets
together can be started. This is the union operation. When we apply union operation
on two sets, the members of both sets combined together and form a new set. In this
case, there will be no duplicate entry in the new sets as these were disjoint. The
definitions of find and union are:

Find returns the name of the set (equivalence class) that contains a given
element, i.e., Si = find(a)
Union merges two sets to create a new set Sk = Si U Sj.

We give an element to the find method and it returns the name of the set. The method
union groups the member of two sets into a new set. We will have these two
operations in the disjoint abstract data type. If we want to add the relation a R b, there
is need to see whether a and b are already related. Here a and b may be two persons
and a relation is given between them. First of all we will see that they are already
related or not. This is done by performing find on both a and b to check whether they
are in the same set or not. At first, we will send a to the find method and get the name
of its set before sending b to the find method. If the name of both sets is same, it
means that these two belong to the same set. If they are in the same set, there is a
relation between them. We did not get any useful information with this relation. Lets
again take the example of the Haris and Saad. We know that Haris is the brother of
Saad, so they can be placed into a new set. Afterwards, we get the information that
Saad is the brother of Haris.
Now the question arises, if they are not in the same set, what should we do? We will
not waste this information and apply union which merges the two sets a and b into a
new set. This information will be helpful later on. We keep on building the database

Page 403 of 519

CS301 Data Structures

___________________________________________________________________
with this information. Lets take the example of pixels. Suppose we have two pixels
and are told that these two belong to the same region, have same color or these two
pixels belong to the liver in CT scan. We will keep these two pixels and put them in
one set. We can name that set as liver. We will use union operation to merge two sets.
The algorithm to do this is frequently known as Union/Find for this reason.
There are certain points to note here. We do not perform any operations comparing
the relative values of set elements. That means that we are not considering that one
person is older than the other, younger or one electrical component is bigger than
other etc. We merely require knowledge of their location, i.e., which set an element,
belongs to.
Lets talk about the algorithm building process. We can assume that all elements are
numbered sequentially from 1 to n. Initially we have Si = {i} for i = 1 through n. We
can give numbers to persons like person1, person2 etc. Consider the example of jail.
In a jail, every prisoner is given a number. He is identified by a number, not by the
name. So you can say that persons in a cell of jail are related. We will use these
numbers to make our algorithm run faster. We can give numbers to every person,
every pixel, each electrical component etc. Under this scheme, the member of S10 is
number 10. That number 10 can be of a person, a pixel or an electrical component.
We are concerned only with number.
Secondly, the name of the set returned by find is fairly arbitrary. All that really
matters is that find(x) = find(y) if and only if x and y are in the same set. We will now
discuss a solution to the union/find problem that for any sequence of at most m finds
and up to n-1 unions will require time proportional to (m + n). We are only storing
numbers in the sets and not keeping any other relevant information about the
elements. Similarly we are using numbers for the names of the sets. Therefore we may
want that the method find just returns the number of the set i.e. the person 10 belongs
to which set number. The answer may be that the person 10 belongs to set number 10.
Similarly the set number of person 99 may be set 99. So if the set numbers of two
persons is not equal, we can decide that these two persons do not belong to the same
set. With the help of this scheme, we will develop our algorithm that will run very
fast. In the next lecture, we will discuss this algorithm and the data structure in detail.

Page 404 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 35
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 8
8.2, 8.3

Summary

Dynamic Equivalence Problem

Example 1
Parent Array
o Initialization
o Find (i)
o Union (i, j)
Example 2
o Initialization
o union operation
o find Operation
Running Time Analysis

Before talking about the data structure implementation in detail, we will recap the
things discussed in the previous lecture. The concepts that came under discussion in
the last lecture included the implementation of the disjoint set. It was observed that in
case of data elements, we assign a unique number to each element to make sets from
these numbers. Moreover, techniques of merger and searching in these sets by using
the operations union and find were, talked about. It was seen that if we send a set item
or a number to find method, it tells us the set of which this item is a member. It will
return the name of the set. This name of set may also be a number. We talked about
the union method in which members of two sets join together to form a new set. The
union method returns the name or number of this new set.
Now we will discuss the data structure used to implement the disjoint sets, besides
ascertaining whether this data structure implements the find and union operations
efficiently.

Dynamic Equivalence Problem

We are using sets to store the elements. For this purpose, it is essential to remember
which element belongs to which set. We know that the elements in a set are unique
and a tree is used to represent a set. Each element in a tree has the same root, so the
root can be used to name the set. The item (number) in the root will be unique as the
set has unique values of items. We use this root as the name of set for our
convenience. Otherwise, we can use any name of our choice. So the element in the
root (the number) is used as the name of the set. The find operation will return this

Page 405 of 519

CS301 Data Structures

___________________________________________________________________
name. Due to the presence of many sets, there will be a collection of trees. In this
collection, each tree will be representing one set. If there are ten elements, we will
have ten sets initially. Thus there will be ten trees at the beginning. In general, we
have N elements and initially there will be N trees. Each tree will have one element.
Thus there will be N trees of one node. Now here comes a definition for this
collection of trees, which states that a collection of trees is called a forest.
The trees used for the sets are not necessarily binary trees. That means it is not
necessary that each node should have a maximum of two children nodes. Here a node
may have more than two children nodes.
To execute the union operation in two sets, we merge the two trees of these sets in
such a manner that the root of one tree points to the root of other. So there will be one
root, resulting in the merger of the trees. We will consider an example to explain the
union operation later.
In the find operation, when we call find (x), it helps us to know which set this x
belongs to. Internally, we find this x in a tree in the forest. When this x is found in a
tree the find returns the number at root node (the name of the set) of that tree.
Example 1
Now lets consider an example to understand this phenomenon. Suppose, we have
developed a data structure and apply the union and find operations on it. There are
eight elements in this data structure i.e.1 to 8. These may be the names of people to
which we have assigned these numbers. It may be other distinct eight elements. We
will proceed with these numbers and make a set of each number. The following figure
shows these numbers in different sets.

Figure 35.1: Eight elements, initially in different sets

Now we carry out union operation on the sets 5 and 6. The call union(5,6) means,
merge the sets 5 and 6 and return the new set developed due to the merger of the sets5 and 6. In the following figure, we see this union. We put the set 6 below set 5,
which join together.

Figure 35.2: After union (5, 6)

Page 406 of 519

CS301 Data Structures

___________________________________________________________________
After the merger, the new set contains two members 5 and 6. Now the question arises
what will be the name of this new set. In the above union operation, the set 6 becomes
the node of 5. It may be in reverse i.e. 5 could be a node of 6. In the above figure, we
put 6 as node of 5. Moreover the arrow joining these is from 6 to 5. In the new set
formed due to the merger of 5 and 6, it seems that 5 has some superiority. So the
name of the new set is 5. We passed two arguments 5 and 6 to the union function.
And the union made 6 a member of 5. Thus, if we pass S1 and S2 to the union
function, the union will make S2 a member of S1. And the name of the new set will
be S1. That means the name of first argument will be the name of the new set.
Now we call union (7,8). After this call, 7 and 8 form a new set in which 8 is a
member of 7 and the name of this new set is 7 (that is the first argument in the call).
In other words, 7 is root and 8 its child. This is shown in the following figure.

Figure 35.3: After union (7, 8)

Now we call the union function by passing it set 5 and set 7 as arguments i.e. we call
union (5,7). Here the sets 5 and 7 have two members each. 5 and 6 are the members
of 5 and similarly the two members of 7 are 7 and 8. After merging these sets, the
name of the new set will be 5 as stated earlier that the new set will be named after the
first argument. The following figure (figure 35.4) shows that now in the set 5, there
are four members i.e. 5, 6, 7 and 8.

Figure 35.4: After union (5, 7)

We can see that there are four unique set members in this set (i.e. 5).
We will demonstrate the union operation with the help of another example. Suppose,
we have made another union that is union (3,4). This call merges the set 4 with the set
3. The following figure shows this.

Page 407 of 519

CS301 Data Structures

___________________________________________________________________

Figure 35.5: After union (3, 4)

In this figure, we see that there are four sets that are 1, 2, 3 and 5. Now we unite the
sets 4 and 5 and make a call union (4,5). The following figure shows the outcome of
this call. In this figure, the set 5 points to 3 whereas we made a call union (4,5).

Figure 35.6: After union (4, 5)

So we conclude that it is not necessary that the caller should send the roots to union. It
is necessary that the union function should be such that if a caller sends elements of
two sets, it should find the sets of those elements, merge them and return the name of
new set. Thus our previous call i.e. union (4,5) was actually carried out in the way
that first the union finds the root of 4 that is 3. Then it looks for 5 that itself is a root
of its set. After this, it merges both the trees (sets). This merger is shown in the above
figure i.e. Figure 35.6.
Up to now, we have come to know that the formation of this tree is not like a binary
tree in which we go down ward. Moreover, a binary tree has left and right children
that are not seen yet in the tree we developed. This is a tree like structure with some
properties.

Page 408 of 519

CS301 Data Structures

___________________________________________________________________
Lets talk about these properties.
Here we see that typical tree traversal (like inorder, preorder or postorder) is not
required. So there is no need for pointers to point to the children. Instead, we need a
pointer to parent, as its an up-tree. We call it up-tree due to the fact that it is such a
tree structure in which the pointers are upward. These parent pointers can be stored in
an array. Here we have to keep (and find) pointer to the parent of a node unlike a
binary tree in which we keep pointer to child nodes. In the array, we will set the
parent of root to 1. We can write it as

Parent[i] = -1

// if i is the root

Now we will keep these tree structures (forest) in an array in the same manner. With
the merger of the trees, the parents of the nodes will be changed. There may be nodes
that have no parent (we have seen this in the previous example). For such nodes, we
will keep 1 in the array. This shows that this node has no parent. Moreover, this
node will be a root that may be a parent of some other node.
Now we will develop the algorithm for find and union. Lets consider an example to
see the implementation of this disjoint set data structure with an array.

Parent Array
Initialization
We know that at the start we have n elements. These n elements may be the original
names of the elements or unique numbers assigned to them. Now we take an array and
with the help of a for loop, keep these n elements as root of each set in the array.
These numbers are used as the index of the array before storing 1 at each location
from index zero to n. We keep 1 to indicate a number as root. In code, this for loop
is written as under.
for ( i = 0; i < n ; i ++)
Parent [i] = -1 ;

Find ( i )
Now look at the following code of a loop. This loop is used
to find the parent of an element or the name of the set that
contains that element.
// traverse to the root (-1)
for(j=i; parent[j] >= 0; j=parent[j])
;
return j;
In

this loop, i is an argument passed to the find function.

The execution of the loop starts from the value, passed to
the find function. We assign this value to a variable j and
Page 409 of 519

CS301 Data Structures

___________________________________________________________________

check whether its parent is greater than zero. It means

that it is not 1. If it is greater than zero, its parent exists,
necessitating the re-initialization of the j with this parent of
j for the next iteration. Thus this loop continues till we find
parent of j (parent [j]) less than zero (i.e. -1). This means
that we come to the root before returning this number j.
Union ( i, j )
Now lets see the code for the function of union. Here we
pass two elements to the function union. The union finds
the roots of i and j. If i and j are disjoint sets, it will merge
them. Following is the code of this function.
root_i = find(i);
root_j = find(j);
if (root_i != root_j)
parent[root_j] = root_i;
In the code, we see that at first it finds the root of tree in which i exists by the find(i)

method and similarly finds the root of the set containing j

by find(j). Then there is a check in if statement to see
whether these sets (roots) are same or not. If these are
not the same, it merges them in such a fashion that the
root i is set as the parent of root j. Thus, in the array
where the value of root j exists, the value of root i
becomes there.
Example 2
To understand these concepts, lets consider an example.
We re-consider the same previous example of eight
numbers and see how the initialization, union and find
work.
Initialization
In the following figure (figure 35.7), we have shown the
initialization step. Here we make an array of eight
locations and have initialized them with 1 as these are
the roots of the eight sets. This 1 indicates that there is
no parent of this number. We start the index of array from

Page 410 of 519

CS301 Data Structures

___________________________________________________________________

1 for our convenience. Otherwise, we know that the index

of an array starts from zero.
1

-1

-1

-1

-1

-1

-1

-1

-1

Figure 35.7: Eight elements, initially in different sets

Union Operation
Now we come to the union operation. First of all, we call
union (5,6). This union operation forms an up-tree in
which 5 is up and 6 down to it. Moreover 6 is pointing to 5.
In the array, we put 5 instead of 1 at the position 6. This
shows that now the parent of 6 is 5. The other positions
have 1 that indicates that these numbers are the roots of
some tree. The only number, not a root now is 6. It is now
the child of 5 or in other words, its parent is 5 as shown in
the array in the following figure.

-1

-1

-1

-1

-1

-1

-1

Figure 35.8: After union (5, 6)

Page 411 of 519

CS301 Data Structures

___________________________________________________________________

Now we carry out the same process (i.e. union) with the
numbers 7 and 8 by calling union (7,8). The following
figure represents this operation. Here we can see that the
parent of 8 is 7 and 7 itself is still a root.
1

-1

-1

-1

-1

-1

-1

Figure 35.9: After union (7, 8)

Now we execute the union (5,7). The following figure

represents the tree and array status after the performance
of this union operation.

Page 412 of 519

CS301 Data Structures

___________________________________________________________________
1

8
-1

-1

-1

-1

-1

Figure 35.10: After union (5, 7)

The tree in the figure shows that 7 points to 5 now. In the

array, we see that the value at position 7 is 5. This means
that the parent of 7 is 5 now. Whereas the parent of 5 is
still 1 that means it is still a root. Moreover, we can see
that the parent of 8 is 7 as before.
Afterwards, we call union (3,4). The effect of this call is
shown in the following figure. Here in the array at position
4, there is 3 instead of 1. This shows that the parent of 4
is 3.
1

8
-1

-1

-1

-1

Figure 35.11: After union (3, 4)

Page 413 of 519

CS301 Data Structures

___________________________________________________________________

By looking at the array only, we can know that how many

trees are there in the collection (forest). The number of
trees will be equal to the number of 1 in the array. In the
above figure, we see that there are four 1 in the array so
the number of trees is four and the trees in the figure
confirms this. These four trees are with the roots 1, 2 3
and 5 respectively.
Now we carry out another union operation. In this
operation, we do the union of 4 and 5 i.e. union (4, 5).
Here the number 4 is not a root. Rather, it is an element of
a set. We have already discussed that the union operation
finds the set (root) of the element passed to it before
putting together those sets. So here the union finds the
set containing 4 that is 3. The element 5 itself is a name
(root) of a set. Now the union operation merges the two
sets by making the second set a member of the first set.
In this call (union (4, 5)), the first set is 3 (that is found for
the element 4) and second is 5. So the union joins 5 with
3 while 5 is pointing to 3. Due to this union operation, now,
in the array, there is 3 at position 5 instead of 1. The 3 at
this position 5 indicates that the parent of 5 is 3. The
following figure shows the tree and array after the
operation union (4,5).

Page 414 of 519

CS301 Data Structures

___________________________________________________________________
1

-1

-1

-1

Figure 35.12: After union (4, 5)

These were the union operations while using parent array.

Now lets look at the find operation.
Find Operation
To understand the find operation, we make a call find (8)
in the above forest. This call means that the caller wants
to know the set in which number 8 is lying. We know that
the root of the tree is also the name of the set. By looking
at the trees in the figure below, we come to know that 8 is
in the tree with 3 as root. This means 8 is in set 3.

Page 415 of 519

CS301 Data Structures

___________________________________________________________________
1

-1

-1

-1

Figure 35.13: find (8)

Thus from the figure, we find the set containing 8. This is

being implemented with the parent array to ultimately find
it in the array. For this find operation in the array, we have
discussed the algorithm in which we used a for loop. This
for loop starts from the position that given to the find
function. Here it is 8. The condition in the for loop was that
as long as parent [j] is greater than zero, set this parent [j]
to j. Now we execute the loop with the value of j equal to
8. First of all, the loop goes to position 8 and looks for the
value of parent of 8. This value is 7, which sets the value
of j to 7 and goes to position 7. At that position, the value
of parent of 7 is 5. It goes to position 5. At position 5, the
value of parent is 3. So the loop sets the value of j equal
to 3 and goes to the position 3. Here it finds that the
parent of 3 is 1 i.e. less than zero so the loop ends. This
position 3 is the root and name of the set as parent of it is
1. Thus the name of the set that contains 8 is 3. The find
will return 3 which mean that 8 is a member of set 3.
Similarly, in the above array, we can execute the
operation find (6). This will also return 3 as the loop
execution will be at the positions 6 before going to 5 and

Page 416 of 519

CS301 Data Structures

___________________________________________________________________

finally to 3. It will end here, as the parent of 3 is 1. Thus

find (6) returns 3 as the set that contains 6.
Running Time analysis
Now we will discuss how the implementation of disjoint set is better. We must
remember that while discussing the implementation of disjoint set, we talked about
Boolean matrix. This is a two dimensional structure in which the equivalence
relations is kept as a set of Boolean values. Here in the parent array, we also are
keeping the same information. The union will be used when the two items are related.
In the two-dimensional matrix, we can easily find an item by its index. Now we are
using an array that is a single dimensional structure and seems better with respect to
space. We keep all the information in this array which is at first kept in a matrix. Now
think about a single dimension array versus a two dimensional array. Suppose we
have 1000 set members i.e. names of people. If we make a Boolean matrix for 1000
items, its size will be 1000 x 1000. Thus we need 1000000 locations for Boolean
values. In case of an array, the number of locations will be 1000. Thus this use of
array i.e.tree like structure is better than two-dimensional array in terms of space to
keep disjoint sets and doing union and find operations. Moreover, we do not use
pointers (addresses) that are used in C++. We use array indices as pointers. In case of
find, we follow some indices (pointers) to find the set containing a particular member.
From these points of discussion, we conclude that

union is clearly a constant time operation.

Running time of find(i) is proportional to the height of the tree containing
node i.
This can be proportional to n in the worst case (but not always)
Goal: Modify union to ensure that heights stay small

We will discuss these points in detail in the next lecture.

Page 417 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 36
Reading Material
Data Structures and Algorithm Analysis in C++
8.4, 8.5, 8.6

Chapter. 8

Summary

Running Time Analysis

Union by Size
Analysis of Union by Size
Union by Height
Sprucing up find
Timing with Optimization

Running Time Analysis

In the previous lecture, we constructed up trees through arrays to implement disjoint
sets. union and find methods along with their codes were discussed with the help of
figures. We also talked about the ways to improve these methods. As far as the union
operation is concerned, considering the parent array, it was related to changing the
index of the array. The root value of the merging tree was put into that. In case of
find, we have to go to the root node following the up pointers.
While studying the balanced trees, we had constructed a binary search tree from the
already sorted data. The resultant tree becomes a linked list with height N. Similar
kind of risks involved here. Suppose, we have eight members i.e.1, 2, 3, 4, 5, 6, 7 and
8 as discussed in an example in the previous lecture. Suppose their unions are
performed as union(1,2), union(2,3), union(3,4) and so on up to union(7,8). When the
trees of these unions are merged together, there will be a longer tree of height 8. If we
perform a find operation for an element that is present in the lowest node in the tree
then N-1 links are traversed. In this particular case, 8-1=7 links.
We know that with reduced tree (with lesser height), the time required for find
operation will also be reduced. We have to see whether it is possible to reduce the
size (in terms of height) of the resultant tree (formed after unions of trees). Therefore,
our goal is:
Goal: Modify union to ensure that heights stay small
We had already seen this goal in the last slide of our previous lecture, which is given
below in full:
union is clearly a constant-time operation.

Page 418 of 519

CS301 Data Structures

___________________________________________________________________
Running time of find(i) is proportional to the height of the tree containing node i.
This can be proportional to n in the worst case (but not always)
Goal: Modify union to ensure that heights stay small
You might be thinking of employing the balancing technique here. But it will not be
sagacious to apply it here. Due to the use of array, the balancing may prove an errorprone operation and is likely to decrease the performance too. We have an easier and
appropriate method, called Union by Size or Union by Weight.

Union by Size
Following are the salient characteristics of this method:
Maintain sizes (number of nodes) of all trees, and during union.
Make smaller tree, the subtree of the larger one.
Implementation: for each root node i, instead of setting parent[i] to -1, set it
to -k if tree rooted at i has k nodes.
This is also called union-by-weight.
We want to maintain the sizes of the trees as given in the first point. Until now, we
are not maintaining the size but only the parent node. If the value for parent in a node
is 1, then this indicates that the node is the root node. Consider the code of the find
operation again:
// find(i):
// traverse to the root (-1)
for(j=i; parent[j] >= 0; j=parent[j])
;
return j;
The terminating condition for loop is checking for non-negative values. At any point
when the value of parent[j] gets negative, the loop terminates. Note that the condition
does not specify exactly the negative number. It may be the number 1, 2 , 3 or
some other negative number that can cause the loop to be terminated. That means we
can also put the number of nodes (size) in the tree in this place. We can put the
number of nodes in a tree in the root node in the negative form. It means that if a tree
consists of six nodes, its root node will be containing 6 in its parent. Therefore, the
node can be identified as the root being the negative number and the magnitude (the
absolute value) of the number will be the size (the number of nodes) of the tree.
When the two trees would be combined for union operation, the tree with smaller size
will become part of the larger one. This will cause the reduction in tree size. That is
why it is called union-by-size or union-by-weight.
Lets see the pseudo-code of this union operation (quite close to C++ language). This
contains the logic for reducing tree size as discussed above:

Page 419 of 519

CS301 Data Structures

___________________________________________________________________
//union(i,j):
1. root1 = find(i);
2. root2 = find(j);
3.
if (root1 != root2)
4.
if (parent[root1] <= parent[root2])
5.
{
6.
// first tree has more nodes
7.
parent[root1] += parent[root2];
8.
parent[root2] = root1;
9.
}
10.
else
11.
{
12.
// second tree has more nodes
13.
parent[root2] += parent[root1];
14.
parent[root1] = root2;
15.
}
This operation performs the union of two sets; element is set and element js set. The
first two lines are finding the roots of the sets of both the elements i.e.i and j. Line 3 is
performing a check that the root1 is not equal to root2. In case of being unequal, these
are merged together. In next statement (line 4), the numbers in parent variables of
root nodes are compared. Note that, the comparison is not on absolute values. So the
tree with greater number of nodes would contain the lesser number mathematically.
The condition at line 4 will be true, if the tree of root1 is greater in size than that of
root2. In line 7, the size of the smaller tree is being added to the size of the greater
one. Line 8 is containing the statement that causes the trees to merge. The parent of
root2 is being pointed to root1. If the root2 tree is greater in size than the root1 tree,
the ifcondition at line 4 returns false and the control is transferred to else part of the
if-condition. The parent of root1 (the size of the root1 tree) is added in the parent of
root2 in line 13 and the root1 tree is merged into the root2 tree using the statement at
line 14.
Lets practice this approach using our previous example of array of eight elements.

-1

-1

-1

-1

-1

-1

-1

-1

Eight elements, initially in different sets.

Fig 36.1

These are eight nodes containing initially 1 in the parent indicating that each tree
Page 420 of 519

CS301 Data Structures

___________________________________________________________________
contains only one node. Lets unite two trees of 4 and 6. The new tree will be like the
one shown in Fig 36.2.

6
Union(4,6)

-1

-1

-1

-2

-1

-1

-1

Fig 36.2

From the figure, you can see the node 6 has come down to node 4. The array also
depicts that the parent at position 4 is containing 2. The number of nodes has
become 2. The position 6 is set to 4 indicating that parent of 6 is 4. Next, we similarly
perform the union operation on nodes 2 and 3.

After union(2,3)

-1

-2

-2

-1

-1

-1

Fig 36.3

Lets perform the union operation further and merge the trees 1 and 4.

Page 421 of 519

CS301 Data Structures

___________________________________________________________________

union(1,4)

-2

-3

-1

-1

-1

Fig 36.4
1 was a single node tree (-1 in the array at position 1) and 4 was the root node of the
tree with two elements (-2 in the array at position 4) 4 and 6. As the tree with root 4 is
greater, therefore, node 1 will become part of it. Previously (when the union was not
based on weight), it happened contrary to it with second argument tree becoming the
part of the first tree.
In this case, the number of levels in the tree still remains the same (of two levels) as
that in the greater tree (with root 4). But we apply our previous logic i.e. the number
of levels of the trees would have been increased from two to three after the union.
Reducing the tree size was our goal of this approach.
In the next step, we merge the trees of 2 and 4. The size of the tree with root 2 is 2
(actually 2 in the parent array) while the size of the tree with root 4 is 3 (actually
3). So the tree of node 2 will be joined in the tree with root 4.

3
4

-5

-1

-1

-1

union(2,4)
Fig 36.5
The latest values in the array i.e. at position 4, have been updated to 5 (indicating the
size of the tree has reached 5). At position 2, the new value is 4, which indicates that

Page 422 of 519

CS301 Data Structures

___________________________________________________________________
the up node of 2 is 4.
Next, we perform the union(5,4) operation. We know that the size of the tree with
node 4 is 5 (-5 in the array) and node 5 is single node tree. As per our rules of union
by size, the node 5 will become part of the tree 4.

union(5,4)

-6

-1

-1

Fig 36.6
The updates inside the array can be seen as the value at position 4 has been changed
to 6 and new value at position 5 is 4. This is the tree obtained after applying the
union-by-size approach. Consider, if we had applied the previous method of union,
trees depth would have been increased more.
It seems that we have achieved our goal of reducing the tree size up to certain extent.
Remember, the benefit of reducing the tree height is to increase performance while
finding the elements inside the tree.

Analysis of Union by Size

If unions are done by weight (size), the depth of any element is never greater than
log2n.
By following the previous method of union where second tree becomes the part of the
first tree, the depth of the tree may extend to N. Here N is the number of nodes in the
tree. But if we take size into account and perform union by size, the depth of tree is
log2n maximum. Suppose the N is 100,000 i.e.for 100,000 nodes. The previous
methods may give us a tree with depth level as 100,000. But on the other hand,
log2100000 is approximately 20.Union-by-size gives us a tree of 20 levels of depth
maximum. So this is a significant improvement.
Mathematical proof of this improvement is very complex .We are not covering it here
but only providing the logic or reasoning in intuitive proof.

Page 423 of 519

CS301 Data Structures

___________________________________________________________________
Intuitive Proof:
Initially, every element is at depth zero.
When its depth increases as a result of a union operation (its in the smaller tree),
it is placed in a tree that becomes at least twice as large as before (union of two
equal size trees).
How often can each union be carried out? -- log2n times, because after at most
log2n unions, the tree will contain all n elements.

Union by Height
Alternative to union-by-size strategy: maintain heights,
During union, make a tree with smaller height a subtree of the other.
Details are left as an exercise.
This is an alternate way of union-by-size that we maintain heights of the trees and join
them based on their heights. The tree with smaller height will become part of the one
with greater height. This is quite similar with the union-by-size. In order to implement
Disjoint Set ADT, any of these solutions can work. Both the techniques i.e. union-bysize and union-by-height are equivalent, although, there are minor differences when
analyzed thoroughly.
Now, lets see what can we do for find operation to make it faster.

Sprucing up Find
So far we have tried to optimize union.
Can we optimize find?
Yes, it can be achieved by using path compression (or compaction).
Considering performance, we can optimize the find method. We have already
optimized the trees union operation and now we will see how can optimize the find
operation using the path compression.
During find(i), as we traverse the path from i to root, update parent entries for all
these nodes to the root.
This reduces the heights of all these nodes.
Pay now, and reap the benefits later!
Subsequent find may do less work.
To understand the statements above, lets see the updated code for find below:
find (i)
{
if (parent[i] < 0)
return i;

Page 424 of 519

CS301 Data Structures

___________________________________________________________________
else
return parent[i] = find(parent[i]);
}
We have modified the code in the body of the find. This implementation is of
recursive nature. parent[i] negative means that we have reached the root node.
Therefore, we are returning i. If this is not the root node, then find method is being
called recursively. We know that parent[i] may be positive or negative. But in this
case it will not be negative due to being an index. To understand this recursive call,
we elaborate it further with the help of figure. We call the find method with argument
1.

7
13

17

31

35

32

11

22

10

30

20

16

14

12

Find (1)

13

18

19

Fig 36.7

This tree is bigger as compared to the ones in our previous examples, it has been
constructed after performing some union operations on trees. Now we have called find
method with argument 1. We want to find the set with node 1. By following the
previous method of find, we will traverse the tree from 1 to 2, from 2 to 9, 9 to 4, 4 to
13 and finally from 13 to 7, which is the root node, containing the negative number.
So the find will return 7.
If we apply recursive mechanism given in the above code, the recursive call will start
from find(1). The parent (find(1)) will return 2. The recursive call for 2 will take the
control to node 9, then to 4, 13 and then eventually to 7. When find(7) is called, we
get the negative number and reach the recursion stopping condition. Afterwards, all
the recursive calls, on the stack are executed one by one. The call for 13, find(13)
returns 7. find(4) returns 7 because find(parent(4)) returns 7. That is why the link
between 4 and 13 has been dropped and a new link is established between 4 and 7 as
shown in the Fig 36.7. So the idea is to make the traversal path shorter from a node to
the root. The remaining recursive calls will also return 7 as the result of find
Page 425 of 519

CS301 Data Structures

___________________________________________________________________
operation. Therefore, for subsequent calls we establish the links directly from the
node to the root.

7
13

17

22

31

32

10

11

20

30

16

14

12

Find (1)

35

13

Fig 36.8

19

18

Similarly the node 2 is directly connected to the root node and the interlink between
the node 2 and 9 is dropped.

7
13

17

31

35

32

11

22

10

30

20

16

14

12

Find (1)

13

18

19

Fig 36.9

Page 426 of 519

CS301 Data Structures

___________________________________________________________________
The same will happen with 1.

7
13

17

32

10

11

20

30

16

14

12

Find (1)

35

22

31

13

Fig 36.10

19

18

7
13

17

31

35

32

11

22

10

30

20

16

14

12

F in d (1)

13

18

19

F ig 36.11

The union operation is based on size or weight but the reducing the in-between links
or path compression from nodes to the root is done by the find method. The find

Page 427 of 519

CS301 Data Structures

___________________________________________________________________
method will connect all the nodes on the path to the root node directly.
The path compression is depicted by the following tree:
f

d
Find(a)

Fig 36.12

If we have called find(a) then see the path from node a to the root node f. a is
connected to root node f through b, c, d and e nodes. Notice that there may be further
subtrees below these nodes. After we apply this logic of path compression for find
operation, we will have the tree as follows:
f

Find(a)

Fig 36.13

We see, how much is the impact on performance in find by a theorem.

Timing with Optimization

Theorem: A sequence of m union and find operations, n of which are find
operations, can be performed on a disjoint-set forest with union by rank (weight or
Page 428 of 519

CS301 Data Structures

___________________________________________________________________
height) and path compression in worst case time proportional to (m (n)).
(n) is the inverse Ackermanns function which grows extremely slowly. For all
practical purposes, (n) R 4.
Union-find is essentially proportional to m for a sequence of m operations, linear
in m.
There are number of things present in this theorem regarding analysis, which are
difficult to cover in this course. We sill study these in course of Algorithm Analysis.
At the moment, consider if there are m union operations out of which n are finds. The
average time for union-find will be linear in m and it will grow not quadratically or
exponentially. union and find operations using this data structure are very efficient
regarding both space and time.

Data Structures
Lecture No. 37
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 8

Summary

Review
Image Segmentation
Maze Example
Pseudo Code of the Maze Generation

Review
In the last lecture, we talked about union and find methods with special reference to
the process of optimization in union. These methods were demonstrated by reducing
size of tree with the help of the techniques-union by size or union by weight.

Page 429 of 519

CS301 Data Structures

___________________________________________________________________
Similarly the tree size was reduced through path optimization in the find method. This
was due to the fact that we want to reduce the tree traversal for the find method. The
time required by the find/union algorithm is proportional to m. If we have m union
and n find, the time required by find is proportional to m+n. Union is a constant time
operation. It just links two trees whereas in the find method tree traversal is involved.
The disjoint sets are increased and forest keeps on decreasing. The find operation
takes more time as unions are increased. But on the average, the time required by find
is proportional to m+n.

Now we will see some more examples of disjoint sets and

union/find methods to understand their benefits. In the
start, it was told that disjoint sets are used in image
segmentation. We have seen the picture of a boat in this
regard. We have also talked about the medical scanning
to find the human organs with the help of image
segmentation. Lets see the usage of union/find in image
segmentation.
Image segmentation is a major part of image processing. An image is a collection of
pixels. A value associated with a pixel can be its intensity. We can take images by
cameras or digital cameras and transfer it into the computer. In computer, images are
represented as numbers. These numbers may represent the gray level of the image.
The zero gray level may represent the black and 255 gray level as white. The numbers
between 0 and 255 will represent the gray level between black and white. In the color
images, we store three colors i.e. RGB (Red, Green, Blue). By combining these three
colors, new ones can be obtained.

Image Segmentation
In image segmentation, we will divide the image into different parts. An image may
be segmented with regard to the intensity of the pixels. We may have groups of pixels
having high intensity and those with pixels of low intensity. These pixels are divided
on the basis of their threshold value. The pixels of gray level less than 50 can be
combined in one group, followed by the pixels of gray level less between 50 and 100
in another group and so on. The pixels can be grouped on the basis of threshold for
difference in intensity of neighbors. There are eight neighbors of the pixel i.e. top,
bottom, left, right, top-left, top-right, bottom-left and bottom-right. Now we will see
the difference of the threshold of a pixel and its neighbors. Depending on the
difference value, we will group the pixels. The pixels can be grouped on the basis of
texture (i.e. a pattern of pixel intensities). You will study all these in detail in the
image processing subject.
Lets see an example. Consider the diagram below:

Page 430 of 519

CS301 Data Structures

___________________________________________________________________

It seems a sketch or a square of black, gray and white portions. These small squares
represent a pixel or picture element. The color of these picture elements may be
white, gray or black. It has five rows and columns each. This is a 5 * 5 image. Now
we will have a matrix of five rows and five columns. We will assign the values to the
elements in the matrix depending on their color. For white color, we use 0, 4 for black
and 2 for gray color respectively.
0
1
2
3
4

0
0
2
4
4
0

1
0
0
2
4
2

2
0
4
2
0
2

3
4
4
4
4
4

4
4
0
4
4
0

When we get the image from a digital device, it is stored in numbers. In the above
matrix, we have numbers from 0 to 4. These can be between 0 and 255 or may be
more depending upon the digital capturing device. Now there is a raw digital image.
We will apply some scheme for segmentation. Suppose we need the pixels having
value 4 or more and the pixels with values less than 4 separately. After finding such
pixels, put 1 for those that have values more than 4 and put 0 for pixels having less
than 4 values. We want to make a binary array by applying the threshold. Lets apply
this scheme on it.

0
1
2
3
4

0
0
0
1
1
0

1
0
0
0
1
0

2
0
1
0
0
0

3 4
1 1
1 0
1 1
1 1
1 0

We have applied threshold equal to 4. We have replaced all 4s with 1 and all the
values less than 4, have been substituted by zero. Now the image has been converted
into a binary form. This may represent that at these points, blood is present or not.
The value 4 may represent that there is blood at this point and the value less than 4
represents that there is no blood. It is very easy to a program for this algorithm.
Our objective was to do image segmentation. Suppose we have a robot which
captures an image. The image obtained contains 0, 2 and 4 values. To know the
Page 431 of 519

CS301 Data Structures

___________________________________________________________________
values above and below the threshold, it may use our program to get a binary matrix.
Similarly, for having the knowledge regarding the region of 1s and region of 0s, the
robot can use the algorithm of disjoint sets. In the figure above, you can see that most
of the 1s are on the right side of the matrix.
We will visit each row of the matrix. So far, there are 25 sets containing a single
element each. Keep in mind the example discussed in union/find algorithm. In the
zeroth row, we have three 0s. We can apply union on 0s. Yet we are interested in
1s. In the 3rd column, there is an entry of 1. Now the left neighbor of this element is
0 so we do not apply union here and move to the next column. Here, the entry is 1. Its
left neighbor is also 1. We apply union here and get a set of two elements. We have
traversed the first row completely and moved to the next row. There is 0 in the 0th and
1st column while in the 2nd column; we have an entry of 1. Its left neighbor is 0 so we
move to the next column where entry is also 1. We apply union here. There is a set of
1 at the top of this set also. So we take union of these two sets and combine them.
There is a set of four 1s, two from row 0 and two from row 1. In the next row, we
have 1, 0, 0, 1, 1. In the row 2 column 3, there is 1. At the top of this element, we
have 1. We apply union to this 1 and the previous set of four 1s. Now we have a set
of five 1s. In the next column again, there is 1, also included in the set as the sixth
element. In the row 3, we have 1 at the 0 column. There is 1 at the top of this so we
have set of these two 1s. The next element at the row 3 is again 1, which is included
in the left side set of 1s. Now we have another set of 1s having 3 elements. The 1s
at column 3 and column 4 are included in the set of six 1s making it a set of eight
elements. In the last row, we have single 1 at the column 3 which will be included
with the eight element set. So there are two sets of 1s in this matrix.
0
0
0
1
1
0

0
1
2
3
4

1
0
0
0
1
0

2
0
1
0
0
0

3
1
1
1
1
1

4
1
0
1
1
0

We have two disjoint sets

and it is clearly visible that
where are the entries of 1s
and 0s. This is the
situation when we have set
the threshold of 4 to obtain a binary image. We can change this threshold value.
Lets take the value 2 as threshold value. It means that if the value of pixel is 2 or
more than 2, replace it by 1 otherwise 0. The new image will be as:
0
1
2
3
4

0
0
1
1
1
0

1
0
0
1
1
1

2
0
1
1
0
1

3
1
1
1
1
1

4
1
0
1
1
0

We can apply the union scheme on it to find the region of 1s. Here we have a blob of
1.
0 1 2 3 4

Page 432 of 519

CS301 Data Structures

___________________________________________________________________
0
1
2
3
4

0
1
1
1
0

0
0
1
1
1

0
1
1
0
1

1
1
1
1
1

1
0
1
1
0

The region is shaded. The image has been segmented. With the help of union/find
algorithm, we can very quickly segment the image. The union/find algorithm does not
require much storage. Initially, we have 25 sets that are stored in an array i.e. the uptree. We do all the processing in this array without requiring any extra memory. For
the image segmentation, disjoint sets and union-find algorithm are very useful. In the
image analysis course, you will actually apply these on the images.

Maze Example
You have seen the maze game in the newspapers. This is a puzzle game. The user
enters from one side and has to find the path to exit. Most of the paths lead to blind
alley, forcing the user to go back to square one.

This can be useful in robotics. If the robot is in some room, it finds its path between
the different things. If you are told to build mazes to be published in the newspaper, a
new maze has to be developed daily. How will you do that? Have you done like this
before? Did anyone told you to do like this? Now you have the knowledge of disjoint
sets and union-find algorithm. We will see that the maze generation is very simple
with the help of this algorithm.
Lets take a 5 * 5 grid and generate a 5 * 5 maze. A random maze generator can use
union-find algorithm. Random means that a new maze should be generated every
time.
Consider a 5x5 maze:
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

Here we have 25 cells. Each cell is isolated by walls from the others. These cells are
numbered from 0 to 24. The line between 0 and 1 means that there is a wall between
them, inhibiting movement from 0 to 1. Similarly there is a wall between 0 and 5.
Take the cell 6, it has walls on 4 sides restricting to move from it to anywhere. The
internal cells have walls on all sides. We will remove these walls randomly to
establish a path from the first cell to the last cell.
Page 433 of 519

CS301 Data Structures

___________________________________________________________________
This corresponds to an equivalence relation i.e. two cells are equivalent if they can be
reached from each other (walls been removed so there is a path from one to the other).
If we remove the wall between two cells, then there can be movement from one cell to
the other. In other way, we have established an equivalence relationship between
them. This can be understood from the daily life example that when the separation
wall of two persons is removed, they become related to each other. In maze
generation, we will remove the walls, attaching the cells to each other in some
sequence. In the end, we will have a path from the start cell to the end cell.
First of all, we have to decide an entrance and an exit. From the entrance, we will
move into the maze and have to reach at the exit point. In our example, the entrance is
from the cell 0 and the exit will be at cell 24.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

How can we generate maze? The algorithm is as follows:

Randomly remove walls until the entrance and exit cells are in the same set.
Removal of the wall is the same as doing a union operation.
Do not remove a randomly chosen wall if the cells it separates are already in
the same set.

We will take cells randomly. It means that the probability of each cell is equal. After
selecting a cell, we will choose one of its surrounding walls. The internal cells have
four walls around them. Now we will randomly choose one wall. In this way, we will
choose the neighboring cell. Initially we have 25 sets. Now after removing a wall
between 2 cells, we have combined these two cells into one set. Here the union
method has been applied to combine these two cells. If these cells are already in the
same set, we will do nothing.
We will keep on randomly choosing the cells and removing the walls. The cells will
merge together. The elements of the set will keep on growing and at some point, we
may have the entrance cell (cell 0) and the exit cell (cell 24) in the same set. When the
entrance cell and the exit cell are in the same set, it means that we have a set in which
the elements are related to each other and there is no wall between them. In such a
situation, we can move from start to the end going through some cells of this set. Now
we have at least one path from entrance to exit. There may be other paths in which we
have the entrance cell but not the exit. By following this path, you will not reach at
the exit as these paths take you to the dead end. As these are disjoint sets, so unless
the entrance and exit cell are in the same set, you will not be able to find the solution
of the maze.

Pseudo Code of the Maze Generation

Lets take a look at the pseudo code of the maze generation. We will pass it to the size
Page 434 of 519

CS301 Data Structures

___________________________________________________________________
argument to make a maze of this size. We will use the entrance as 0 and exit as size-1
by default.
MakeMaze(int size) {
entrance = 0; exit = size-1;
while (find(entrance) != find(exit)) {
cell1 = randomly chosen cell
cell2 = randomly chosen adjacent cell
if (find(cell1) != find(cell2) {
knock down wall between cells
union(cell1, cell2)
}
}
}
After initializing the entrance and exit, we have a while loop. The loop will keep on
executing till the time the find(entrance) is not equal to find(exit). It means that the
loop will continue till the set returned by the find(entrance) and find(exit) are not
same. When the entrance and exit cells are in the same set, the loop will stop. If these
cells are not in the same set, we will enter into the loop. At first, we will randomly
choose a cell from the available cells e.g. from 0 to 24 from our example. We are not
discussing here how to randomly choose a cell in C++. There may be some function
available to do this. We store this value in the variable cell1 and choose randomly its
neighboring cell and store it in cell2. This cell may be its top, bottom, left or right
cell, if it is internal cell. If the cell is at the top row or top bottom, it will not have four
neighbors. In case of being a corner cell, it will have only two neighboring cells. Then
we try to combine these two cells into one set. We have an if statement which checks
that the set returned by find(cell1) is different than the set returned by find(cell2). In
this case, we remove the wall between them. Then we apply union on these two cells
to combine them into a set.
We randomly choose cells. By applying union on them, the sets are joined together
and form new sets. This loop will continue till we have a set which contains both
entrance and exit cell. This is a very small and easy algorithm. If we have the unionfind methods, this small piece of code can generate the maze. Lets take a pictorial
look at different stages of maze generation. You can better understand this algorithm
with the help of these figures. We are using the disjoint sets and union-find algorithm
to solve this problem.
Initially, there are 25 cells. We want to generate a maze from these cells.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

Apply the algorithm on it. We randomly choose a cell. Suppose we get cell 11. After
this, we randomly choose one of its walls. Suppose we get the right wall. Now we will

Page 435 of 519

CS301 Data Structures

___________________________________________________________________
have cell 11 in the variable cell1 and cell 12 in the variable cell2.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

By now, each cell is in its own set. Therefore find(cell 11) will return set_11 and
find(cell 12) will return set_12. The wall between them is removed and union is
applied on them. Now we can move from cell 11 to cell 12 and vice versa due to the
symmetry condition of disjoint sets. We have created a new set ( set_11 = {11,12} )
that contains cell 11 and cell 12 and all other cells are in their own cells.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

Now we randomly choose the cell 6 and its bottom wall. Now the find(cell 6) will
return set_6. The find(cell 11) will return set_11 which contains cell 11 and cell 12.
The sets returned by the two find calls are different so the wall between them is
removed and union method applied on them. The set set_11 now contains three
elements cell 11, cell 12 and cell 6. These three cells are joined together and we can
move into these cells.
Now we randomly select the cell 8. This cell is not neighbor of set_11 elements. We
can randomly choose the cell 6 again and its bottom wall. However, they are already
in the same set so we will do nothing in that case. We randomly choose the cell 8 and
its top wall (cell 3). As these two cells are in different sets, the wall between them is
removed so that union method could be applied to combine them into a set (set_8 =
{8, 3}).
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

Now we randomly choose the cell 14 and its top i.e. cell 9. Now we keep on
combining cells together but so far entrance and exit cells are not in the same set.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24
Page 436 of 519

CS301 Data Structures

___________________________________________________________________
We randomly choose cell 0 and its bottom wall so the cell 0 and cell 5 are combined
together.
0
5
10
15
20

1
6
11
16
21

2
7
12
17
22

3
8
13
18
23

4
9
14
19
24

If you keep on using this algorithm, the walls between cells will keep on vanishing,
resulting in the growth of set elements. At some point, the entrance and exit cell will
come into one set and there will be a path taking us from the start to exit. There may
be some other sets which do not contain the exit cell. Take it as a programming
exercise and do it. This is very interesting exercise. Find some information on the
internet regarding maze generation. In the next lecture, we will talk about the new
data type i.e. table.

Data Structures
Lecture No. 38
Summary

Table and Dictionaries

Operations on Table ADT
Implementation of Table
o Unsorted Sequential Array
o Sorted Sequential Array
Binary Search

We will discuss the concepts of Tables and Dictionaries in this lecture with special
reference to the operations on Table ADT and the implementation.
Tables and Dictionaries
The table, an abstract data type, is a collection of rows and columns of information.
From rows and columns, we can understand that this is like a two dimensional array.
But it is not always a two dimensional array of same data type. Here in a table, the

Page 437 of 519

CS301 Data Structures

___________________________________________________________________
type of information in columns may be different. The data type of first column may be
integer while the type of second column is likely to be string. Similarly there may be
other different data types in different columns. So the two-dimensional array used for
table will have different data types.
A table consists of several columns, known as fields. These fields are some type of
information. For example a telephone directory may have three fields i.e. name,
address and phone number. On a computer system, the user account may have fieldsuser ID, password and home folder. Similarly a bank account may have fields like
account number, account title, account type and balance of account.
Following is a table that contains three fields (We are calling columns as fields).
These three fields are name, address and phone. Moreover, there are three rows
shown in the table.
Name
Sohail Aslam
Imran Ahmad
Salman Akhtar

Address
50 Zahoor Elahi Rd, Gulberg-4, Lahore
30-T Phase-IV, LCCHS, Lahore
131-D Model Town, Lahore

Phone
567-4567
517-2349
756-5678

Figure 38.1: A table having three fields.

Each row of the table has a set of information in the three fields. The fields of a row
are linked to each other. The row of a table is called a record. Thus the above table
contains three records. It resembles to the telephone directory. There may be other
fields in the phone directory like fathers name or the occupation of the user. Thus the
data structure that we form by keeping information in rows and columns is called
table.
During the discussion on the concept of table, we have come across the word,
dictionary. We are familiar with the language dictionary, used to look for different
words and their meanings. But for a programmer, the dictionary is a data type. This
data is in the form of sets of fields. These fields comprise data items.
A major use of table is in Databases where we build and use tables for keeping
information. Suppose we are developing payroll software for a company. In this case,
a table for employees will be prepared. This table of employees has the name, id
number, designation and salary etc of the employees. Similarly there may be other
tables in this database e.g. to keep the leave record of the employees. There may be a
table to keep the personal information of an employee. Then we use SQL (Structured
Query Language) to append, select, search and delete information (records) from a
table in the database.
Another application of table is in compilers. In this case, symbol tables are used. We
will see in compilers course that symbol table is an important structure in a compiler.
In a program, each variable declared has a, type and scope. When the compiler
compiles a program, it makes a table of all the variables in the program. In this table,
the compiler keeps the name, type and scope of variables. There may be other fields
in this table like function names and addresses. In the compiler construction course,
we will see that symbol table is a core of compiler.
When we put the information in a table, its order will be such that each row is a
record of information. The word tuple is used for it in databases. Thus a row is called
a record or tuple. As we see there is a number of rows (records) in a table. And a
record has some fields.

Page 438 of 519

CS301 Data Structures

___________________________________________________________________
Now here are some things about tables.
To find an entry in the table, we only need to know the contents of one of the fields
(not all of them). This field is called key. Suppose we want to see information (of a
record) in dictionary. To find out a record in the dictionary, we usually know the
name of the person about whom information is required. Thus by finding the name of
the person, we reach a particular record and get all the fields (information) about that
person. These fields may be the address and phone number of the person. Thus in
telephone directory, the key is the name field. Similarly in a user account table, the
key is usually user id. By getting a user id, we can find its information that consists of
other fields from the table. The key should be unique so that there is only one record
for a particular value of key. In the databases, it is known as primary key. Thus, in a
table, a key uniquely identifies an entry. Suppose if the name is the key in telephone
book then no two entries in the telephone book have the same name. The key uniquely
identifies the entries. For example if we use the name imran ahmed in the following
table that contains three entries, the name imran ahmed is there only at one place in
the following table. In this record, we can also see the address and phone number of
imran ahmed.
Name
Sohail Aslam
Imran Ahmad
Salman Akhtar

Address
50 Zahoor Elahi Rd, Gulberg-4, Lahore
30-T Phase-IV, LCCHS, Lahore
131-D Model Town, Lahore

Phone
567-4567
517-2349
756-5678

Figure 38.2: A table having three fields.

Similarly if we use salman akhtar as key we find out one record for this key in the
table.
The purpose of this key is to find data. Before finding data, we have to add data. On
the addition of the data in the table, the finding of data will be carried through this
key. We may need to delete some data from the table. For example, in the employees
table, we want to delete the data of an employee who has left the company. To delete
the data from the table, the key i.e. name will be used. However in the employees
table, the key may be the employee id and not the name.

Operations on Table ADT

Now we see the operations (methods) that can be performed with the table abstract
data type.
insert
As the name shows this method is used to insert (add) a record in a table. For its
execution, a field is designated as key. To insert a record (entry) in the table, there is
need to know the key and the entry. The insert method puts the key and the other
related fields in a table. Thus we add records in a table.
find
Suppose we have data in the table and want to find some particular information. The
find method is given a key and it finds the entry associated with the key. In other
words, it finds the whole record that has the same key value as provided to it. For
Page 439 of 519

CS301 Data Structures

___________________________________________________________________
example, in employees table if employee id is the key, we can find the record of an
employee whose employee id is, say, 15466.
remove
Then there is the remove method that is given a value of the key to find and remove
the entry associated with that key from the table.

Implementation of Table
Lets talk about why and how we should implement a table. Our choice for
implementation of the Table ADT depends on the answers to the following.

How often entries are inserted, found and removed?

How many of the possible key values are likely to be used?
What is the likely pattern of searching for keys? Will most of the accesses be
to just one or two key values?
Is the table small enough to fit into the memory?
How long will the table exist?

In a table for searching purposes, it is best to store the key and the entry separately
(even though the keys value may be inside the entry). Now, suppose we have a
record of a person Saleem whose address is 124 Hawkers Lane while the phone
number is 9675846. Similarly we have another record of a person Yunus. The
address and phone fields of this person are 1 Apple crescent and 622455
respectively. For these records in the table, we will have two parts. One part is the
complete entry while the other is the key in which we keep the unique item of the
entry. This unique item is twice in a record one as part of the entry and the second in a
field i.e. the key field. This key will be used for searching and deletion purposes of
records. With the help of key, we reach at the row and can get other fields of it. We
also call TableNode to row of the table. Its pictorial representation is given below.

key
Saleem

entry
Saleem, 124 Hawkers Lane, 9675846

TableNod
Yunus Yunus, 1 Apple Crescent, 0044 1970

Figure 38.3: key and entry in table

Now we will discuss the implementation of table with different data structures. The
first implementation is the unsorted sequential array.

Unsorted Sequential Array

In this implementation, we store the data of table in an array such that TableNodes are
stored consecutively in any order. Each element of the row will have a key and entry
of the record.

Page 440 of 519

CS301 Data Structures

___________________________________________________________________
Now lets think how we can store the data of a table in an array. For a telephone
directory, there may be a field name i.e. a string. Then there is address, which is also
a string but of large size as compared to the name. The phone number may be a string
or integer. Each record in the directory has these three fields with different types.
How can we store this information in an array? Here comes the class. Suppose we
make a class Employee, with an ultimate aim of developing objects. An object of
Employee will contain the name, address, designation and salary. Similarly we can
make a class of PhoneDirectoryEntry that will have name, address and phone number.
We can add other fields to these classes. Thus we have Employee and
PhoneDirectoryEntry objects. Now we make an array of objects. In this array, every
object is a table entry. To insert a new object (of Employee or PhoneDirectoryEntry),
we add it to the back of the array. This insert operation in the array is fast as we
directly put the data at the last available position. Thus we add the data to the array as
soon as it is available. In other words, we dont have an order of the data. Now if
there are n entries in the array, the find operation searches through the keys one at a
time and potentially all of the keys to find a particular key. It has to go through all the
entries (i.e. n) if the required key is not in the table. Thus the time of searching will be
proportional to the number of entries i.e. n. Similarly the remove method also requires
time proportional to n. The remove method, at first, has to find the key, needed to be
removed. It consumes the time of find operation that is proportional to the number of
entries i.e. n. If the entry is found, the remove method removes it. Obviously it does
nothing if the entry is not found.
Here we see that in unsorted sequential array, the insertion of data is fast but the find
operation is slow and requires much time. Now the question arises if there is any way
to keep an array in which the search operation can be fast? The use of sorted
sequential array is a solution to this problem.

Sorted Sequential Array

We have studied in the tree section that binary search tree is used to search the
information rapidly. Equally is true about the sorted sequential array. But in this case,
we want to put the data in an array and not in a tree. Moreover, we want to search the
data in this array very fast. To achieve this objective, we keep the data in the array in
a sorted form with a particular order. Now suppose we are putting the information in
the telephoneDirectory table into an array in a sorted form in a particular order. Here,
for example, we put the data alphabetically with respect to name. Thus data of a
person whose name starts with a will be at the start of the table and the name
starting with b will be after all the names that start with a. Then after the names
starting with b there will be names starting with c and so on. Suppose a new name
starting from c needs to be inserted. We will put the data in the array and sort the
array so that this data can be stored at its position with respect to the alphabetic order.
Later in this course, we will read about sorting.
Lets talks about the insert, find and remove methods for sorted data.
insert
For the insertion of a new record in the array, we will have to insert it at a position in
the array so that the array should be in sorted form after the insertion. We may need to
shift the entries that already exist in the array to find the position of the new entry. For
example, if a new entry needs to be inserted at the middle of the array, we will have to
shift the entries after the middle position downward. Similarly if we have to add an

Page 441 of 519

CS301 Data Structures

___________________________________________________________________
entry at the start of the array, all the entries will be moved in the array to one position
right (down). Thus we see that the insert operation is proportional to n (number of
entries in the table). This means insert operation will take considerable time.
find
The find operation on a sorted array will search out a particular entry in log n time by
the binary search. The binary search is a searching algorithm. Recall that in the tree,
we also find an item in log n time. The same is in the case of sorted array.
remove
The remove operation is also proportional to n. The remove operation first finds the
entry that takes log n time. While removing the data, it has to shuffle (move) the
elements in the array to keep the sorted order. This shuffling is proportional to n.
Suppose, we remove the first element from the array, then all the elements of the
array have to be moved one position to left. Thus remove method is proportional to n.

Binary Search
The binary search is an algorithm of searching, used with the sorted data. As we have
sorted elements in the array, binary search method can be employed to find data in the
array. The binary search finds an element in the sorted array in log n time. If we have
100000 elements in the array, the log 1000000 will be 20 i.e. very small as compared
to 100000. Thus binary search is very fast.
The binary search is like looking up a phone number in the directory or looking up a
word in the dictionary. For looking a word in the dictionary, we start from the middle
in the dictionary. If the word that we are looking for comes before words on the page,
it shows that the word should be before this page. So we look in the first half.
Otherwise, we search for the word in the second half of the dictionary. Suppose the
word is in the first half of the dictionary, we consider first half for looking the word.
We have no need to look into the second half of the dictionary. Thus the data to be
searched becomes half in a step. Now we divide this portion into two halves and look
for the word. Here we again come to know that the word is in the first half or in the
second half of this portion. The same step is repeated with the part that contains the
required word. Finally, we come to the page where the required word exists. We see
that in the binary search, the data to be searched becomes half in each step. And we
find the entry very fast. The number of maximum steps needed to find an entry is log
n, where n is the total number of entries. Now if we have 100000 entries, the
maximum number of attempts (steps) required to find out the entry will be 20 (i.e. log
1000000).

Page 442 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 39
Reading Material
Data Structures and Algorithm Analysis in C++
10.4.2

Chapter. 10

Summary

Searching an Array: Binary Search

Binary Search - Example 1
Binary Search - Example 2
Binary Search - Example 3
Binary Search C++ Code
Binary Search Binary Tree
Binary Search - Efficiency
Implementation 3 (of Table ADT): Linked List
Implementation 4 (of Table ADT): Skip List
Skip List - Representation
Skip List - Higher Level Chains
Skip List - Formally

Page 443 of 519

CS301 Data Structures

___________________________________________________________________

Searching an Array: Binary Search

In the previous lecture, we had started discussion on Binary Search Tree algorithm.
The discussion revealed that if already sorted data is available, then it is better to
apply an algorithm of binary search for finding some item inside instead of searching
from start to the end in sequence. The application of this algorithm will help get the
results very quickly. We also talked about the example of directory of employees of a
company where the names of the employee were sorted. For efficient searching, we
constructed the binary search tree for the directory and looked for information about
an employee named Ahmed Faraz.
We also covered:
Binary search is like looking up a phone number or a word in the dictionary
Start in middle of book
If the name you're looking for, comes before names on the page, search in the first
half
Otherwise, look into the second half

Fig 39 1
The telephone directory is the quotable example to understand the way the binary
search method works.
In this lecture, we will focus on data structures for performing search operation.
Consider the data is present in an array as we discussed in the previous lecture. For
the first implementation, we supposed that the data is not sorted in the array. For
second implementation, we considered that the data inside the array is put in sorted
array. The advantage of the effort of putting the data in the array in sorted order pays
off when the searches on data items are performed.
Now, lets first see the algorithm (in pseudo code) for this operation below. It is
important to mention that this algorithm is independent of data type i.e. the data can
be of any type numeric or string.
if ( value == middle element )
value is found
else if ( value < middle element )

search left half of list with the same method

else
search right half of list with the same methodThe item we are searching for
in this algorithm is called value. The first comparison of this value is made with the
middle element of the array. If both are equal, it means that we have found our desired
search item, which is present in the middle of the array. If this is not the case, then the
value and the middle element are not the same. The else-if part of the algorithm is
computed, which checks if the value is less than the middle element. If so, the left half
part of the array is searched further in the same fashion (of logically splitting that half
Page 444 of 519

CS301 Data Structures

___________________________________________________________________
part of array into two further halves and applying this algorithm again). This search
operation can also be implemented using the recursive algorithm but that will be
discussed later. At the moment, we are discussing only the non-recursive algorithm.
The last part of the algorithm deals with the case when the value is greater than the
middle element. This processes the right half of the array with the same method.
Lets see this algorithm in action by taking an example of an array of elements:

Binary Search Example 1

Case 1: val == a[mid]
val = 10
low = 0, high = 8
mid = (0 + 8) / 2 = 4

a:

low

10 13 17 19 27
4
mid

8
high

Fig 39.2
You can see an array a in the Fig 39.2 with indexes 0 to 8 and values 1, 5, 7, 9, 10, 13,
17, 19 and 27. These values are our data items. Notice that these are sorted in
ascending (increasing) order.
You can see in the first line of case 1 that val = 10, which indicates that we are
searching for value 10. From second line, the range of data to search is from 0 to 8. In
this data range, the middle position is calculated by using a simple formula (low +
high)/2. In this case, it is mid =(0+8)/2=4. This is the middle position of the data
array. See the array in the above figure Fig 39.2, which shows that the item at array
position 4 is 10, exactly the value we are searching for. So, in this case, we have
found the value right away in the middle position of the array. The search operation
can stop here and an appropriate value can be returned back.
Lets see the case 2 now:

Binary Search Example 2

Case 2: val > a[mid]
val = 19
low = 0, high = 8
mid = (0 + 8) / 2 = 4
new low = mid + 1 = 5

Page 445 of 519

CS301 Data Structures

___________________________________________________________________

a:

low

10 13 17 19 27
4

mid new
low

8
high

Fig 39.3

The second case is about the scenario when value (val) is greater than the middle
value (a[mid]). The range of data items (low and high) is the same as that in case 1.
Therefore, the middle position (mid) is also the same. But the value (val) 19 is greater
than the value at the middle (mid) 10 of the array. As this array is sorted, therefore,
the left half of the array must not contain value 19. At this point of time, our
information about val 19 is that it is greater than the middle. So it might be present in
the right half of the array. The right half part starts from position 5 to position 8. It is
shown in Fig 39.3 that the new low is at position 5. With these new low and high
positions, the algorithm is applied to this right half again.
Now, we are left with one more case.

Binary Search Example 3

Case 3: val < a[mid]
val = 7
low = 0, high = 8
mid = (0 + 8) / 2 = 4
new high = mid - 1 =5

a:

low

10 13 17 19 27
4

new mid
high

8
high

Fig 39.4
The value to be searched (val) in this case is 7. The data range is the same starting
from low=0 to high=8. Middle is computed in the same manner and the value at the
middle position (mid) is compared with the val. The val is less than the value at mid
position. As the data is sorted, therefore, a value lesser than the one at mid position
should be present in the lower half (left half) of the array (if it is there). The left half
of the array will start from the same starting position low=0 but the high position is
going to be changed to mid-1 i.e. 3. Now, lets execute this algorithm again on this
left half.

Page 446 of 519

CS301 Data Structures

___________________________________________________________________
va l = 7

a:

a:

a:

10

13

17

19

27

10

13

17

19

27

10

13

17

19

27

F ig 3 9 .5

Firstly, we will compute the middle of 0 and 3 that results in 1 in this integer division.
This is shown in the top array in Fig 39.5. Now, the val 7 is compared with the value
at the middle position (mid) i.e.5. As 7 is greater than 5, we will process the right half
of this data range (which is positioned from low=0 to high=3). The right half of this
data range starts from position 2 and ends at position 3. The new data range is low=2
and high=3. The middle position (mid) is computed as (2+3)/2=2. The value at the
mid position is 7. We compare the value at mid position (7) to the val we are looking
for. These are found to be equal and finally we have the desired value.
Our desired number is found within positions- 0 to 8 at position 2. Without applying
this binary search algorithm, we might have performed lot more comparisons. You
might feel that finding this number 7 sequentially is easier as it is found at position 2
only. But what will happen in case we are searching for number 27. In that case, we
have to compare with each element present in the array to find out the desired
number. On the other hand, if this number 27 is searched with the help of the binary
search algorithm, it is found in third comparison.
Actually, we have already seen binary search while studying the binary search tree.
While comparing the number with the root element of the tree, we had come to know
that if the number was found smaller than the number in the root, we had to switch to
left-subtree of the root (ensuring that it cannot be found in the right subtree).
Now, lets see the C++ code for this algorithm:

Binary Search C++ Code

int isPresent(int *arr, int val, int N)
{
int low = 0;
int high = N - 1;
int mid;
while ( low <= high )
{
Page 447 of 519

CS301 Data Structures

___________________________________________________________________
mid = ( low + high )/2;
if (arr[mid] == val)
return 1;
// found!
else if (arr[mid] < val)
low = mid + 1;
else
high = mid - 1;
}
return 0; // not found
The name of the routine is isPresent, which expects an int pointer (an array in actual);
an int value val is required to be searched. Another int value is N that indicates the
maximum index value of the array. Inside the body of the function, int variables low,
high and mid are declared. low is initialized to 0 and high is initialized to N-1. while
loop is based on the condition that executes the loop until low <= high. Inside the
loop, very first thing is the calculation of the middle position (mid). Then comes the
first check inside the loop, it compares the val (the value being searched) with the
number at middle position (arr[mid]). If they are equal, the function returns 1. If this
condition returns false, it means that the numbers are unequal. Then comes the turn of
another condition. This condition (arr[mid] < val) is checking if the value at the
middle is less than the value being searched. If this is so, the right half of the tree is
selected by changing the position of the variable low to mid+1 and processed through
the loop again. If both of these conditions return false, the left half of the array is
selected by changing the variable high to mid-1. This left half is processed through
the loop again. If the loop terminates and the required value is not found, then 0 is
returned as shown in the last statement of this function.
You change this function to return the position of the value if it is found in the array
otherwise return 1 to inform about the failure. It is important to note that this
function requires the data to be sorted to work properly. Otherwise, it will fail.
This algorithm is depicted figurative in Fig 39.6.

Page 448 of 519

CS301 Data Structures

___________________________________________________________________

Binary Search Binary Tree

An entire sorted list

First half

First half

Second half

Second half

First half

First half

Second half

First half

Fig 39.6

The search divides a list into two small sub-lists till a sub-list is no more divisible.
You might have realized about the good performance of binary trees by just looking
at these if you remember the fully balanced trees of N items discussed earlier.

Binary Search - Efficiency

To see the efficiency of this binary search algorithm, consider when we divide the
array of N items into two halves first time.
After 1 bisection N/2
items
After 2 bisections N/4 = N/22 items
. . .
After i bisections N/2i =1 item
____________________________________
i = log2 N
First half contains N/2 items while the second half also contains around N/2 items.
After one of the halves is divided further, each half contains around N/4 elements. At
this point, only one of N/4 items half is processed to search further. If we carry out
three bisections, each half will contain N/8 items. Similarly for i bisections, we are
left with N/2i, which is at one point of time is only one element of the array. So we
have the equation here:
N/2i =1
Computing the value of i from this gives us:
i = log2 N
Which shows that after maximum log2 N bisections, either you will be successful in
finding your item or fail to do so.
This was our second implementation of table or dictionary abstract data type using

Page 449 of 519

CS301 Data Structures

___________________________________________________________________
sorted sequential array. As discussed at start of it that if we implement table or
dictionary abstract data type using an array, we have to keep the array elements in
sorted order. An easier way to sort them can be that whenever we want to insert an
element in the array, firstly we find its position (in sorted order) in the array and then
shift the right side (of that position) elements one position towards right to insert it. In
worst case, we might have to shift all the elements one position towards right just to
keep the data sorted so this will be proportional to N. Similarly the remove operation
is also proportional to N. But after keeping the data sorted, the search operation is
returned within maximum log2 N bisections.

Implementation 3 (of Table ADT): Linked List

We might also use linked list to implement the table abstract data type. We can keep
the data unsorted and keep on inserting the new coming elements to front in the list. It
is also possible to keep the data in sorted order. For that, to insert a new element in
the list, as we did for array, we will first find its position (in sorted order) in the list
and then insert it there. The search operation cannot work in the same fashion here
because binary search works only for arrays. Because the linked list may not be
contiguous in memory, normally its nodes are scattered through, therefore, binary
search cannot work with them.
TableNodes are again stored consecutively (unsorted or sorted)
insert: add to front; (1or n for a sorted list)
find: search through potentially all the keys, one at a time; (n for unsorted or for
a sorted list
remove: find, remove using pointer alterations; (n)
key

entry

and so on
Fig 39.7

Well, linked list is one choice to implement table abstract data type. For unsorted
elements, the insertion at front operation will take constant time. (as each element is
inserted in one go). But if the data inside the list is kept in sorted order then to insert a
new element in the list, the entire linked list is traversed through to find the
appropriate position for the element.
For find operation, all the keys are scanned through whether they are sorted or
unsorted. That means the time of find is proportional to N.

Page 450 of 519

CS301 Data Structures

___________________________________________________________________
For remove operation, we have to perform find first. After that the element is removed
and the links are readjusted accordingly.
We know that when we used sorted array, the find operation was optimized. Lets
compare the usage of array and linked list for table abstract data type. The fixed size
of the array becomes a constraint that it cannot contain elements more than that.
Linked list has no such constraint but the find operation using linked list becomes
slower. Is it possible to speed up this operation of find while using linked list? For
this, a professor of University of Maryland introduced a new data structure called skip
list. Lets discuss little bit about skip list.

Implementation 4 (of Table ADT): Skip List

Overcome basic limitations of previous lists
o Search and update require linear time
Fast Searching of Sorted Chain
Provide alternative to BST (binary search trees) and related tree structures.
Balancing can be expensive.
Relatively recent data structure: Bill Pugh proposed it in 1990.
Important characteristics of skip list are stated above. Now, we see skip list in bit
more detail.

Skip List - Representation

Can do better than n comparisons to find
element in chain of length n
head

tail
20

30

40

50

60

Fig 39.8
As shown in the figure. The head and tail are special nodes at start and end of the list
respectively. If we have to find number 60 in the list then we have no other choice but
starting from head traverse the subsequent nodes using the next pointer until the
required node is found or the tail is reached. To find 70 in the list, we will scan
through the whole list and then get to know that it is not present in it. The professor
Pugh suggested something here:

Page 451 of 519

CS301 Data Structures

___________________________________________________________________

Example: n/2 + 1 If we keep pointer to

middle element

head

20

30

40

tail

50

60

Fig 39.9
Firstly, we use two pointers head and tail. Secondly, the node in the middle has two
next pointers; one is the old linked list pointer leading to next node 50 and the second
is leading to the tail node. Additionally the head node also has two pointers, one is the
old liked list pointer pointing to the next node 20 and second one is pointing to the
middle elements next pointer, which is (as told above) further pointing to the tail
node.
Now, if we want to find element 60 in the above list. Is it possible to search the list in
relatively quick manner than the normal linked list shown in Fig 39.8? Yes, it is with
the help of the additional pointers we have placed in the list. We will come to the
middle of the list first and see that the middle element (40) is smaller than 60,
therefore the right half part of the list is selected to process further (as the linked list is
sorted). Isnt it the same we did in binary search? It definitely is.
What if we can add additional pointers (links) and boost the performance. See the
figure below.

Skip List - Higher Level Chains

head

tail

Level 1 & 2 chains

20

26

30

40

50

57

60

Fig 39.10

For general n, level 0 chain includes all elements

level 1 every other element, level 2 chain every fourth, etc.
level i, every 2i th elementLevel 0 chain is our old linked list chain as shown in
Fig 39. Level 1 is new chain added to contain the link of every other node (or
alternate node). Level 2 chain contains links to every 4th node. We keep on adding
levels of chains so that we generalize that for level i chain includes 2ith elements.
After adding these pointers, the skip list is no more our old linked list; it has become
sort of binary tree. Still, there are still few important things to consider.
Skip list contains a hierarchy of chains

Page 452 of 519

CS301 Data Structures

___________________________________________________________________
In general level i contains a subset of elements in level i-1Skip list becomes a
hierarchy of chains and every level contains a subset of element of previous level.
Using this kind of skip list data structure, we can find elements in log2n time. But the
problem with this is that the frequency of pointers is so high as compared to the size
of the data items that it becomes difficult to manage them. The insert and remove
operations on this kind of skip list become very complex because single insertion or
removal requires lot of pointers to readjust.
Professor Pugh suggested here that instead of doing leveling in powers of 2, it should
be done randomly. Randomness in skip lists is a new topic for us. Lets see a formal
definition of skip list.

Skip List - Formally

A skip list for a set S of distinct (key, element) items is a series of lists S0, S1 ,
, Sh such that
o Each list Si contains the special keys + and
o List S0 contains the keys of S in non-decreasing order Each list is a
subsequence
of
the
previous
one,
i.e.,
S0 S1 Sh
o List Sh contains only the two special keys
You are advised to study skip list from your text books. The idea of randomness is
new to us. We will study in the next lecture, how easy and useful becomes the skip
list data structure after employing randomness.

Page 453 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 40
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 10
10.4.2

Summary

Skip List
Skip List Search
Insertion in Skip List
Deletion from Skip List

In the previous lecture, we had started the discussion on the concept of the skip lists.
We came across a number of definitions and saw how the use of additional pointers
was effective in the list structures. It was evident from the discussion that a
programmer prefers to keep the data in the linked list sorted. If the data is not sorted,
we cannot use binary search algorithm. And the insert, find and remove methods are
proportional to n. But in this case, we want to use the binary search on linked list. For
this purpose, skip list can be used. In the skip list, there is no condition of the upper
limit of the array.

Skip List
A skip list for a set S of distinct (key, element) items is a series of lists S0, S1 , , Sh
such that
Each list Si contains the special keys + and -
List S0 contains the keys of S in non-decreasing order
Each list is a subsequence of the previous one, i.e.,
S0 S 1 S h
List Sh contains only the two special keys
Now lets see an example for the skip list. First of all, we have S0 i.e. a linked list.
We did not show the arrows in the list in the figure. The first and last nodes of S0
contain - and + respectively. In the computer, we can put these values by
max(int) and max(int). The values can also be used about which we are sure that these
will not be in the data items. However, for the sake of discussion to show these
values, the - and + are the best notations. We can see that - is less than any value
of data item while + is greater than any value of data item. If we insert any value
much ever, it is large the + will be greater than it. Moreover, we see that the
numbers in S0 are in the non- decreasing order. This S0 is the first list in which all
keys are present.
Now we will take some nodes from this list and link them. That will be not every
other node or every fourth node. It may happen this way. However, we will try that
Page 454 of 519

CS301 Data Structures

___________________________________________________________________
the node should not be every other or fourth node. It will be a random selection. Now
we see S1 i.e. a subset of S0. In S1 we selected the nodes 23, 31, 34 and 64. We have
chosen these nodes randomly with out any order or preference. Now from this S1, we
make S2 by selecting some elements of S1. Here we select only one node i.e. 31 for
the list S2. The additional pointer, here, has to move from - to 31 and from 31 to
+. Now the next list i.e. S3 will be subset of S2. As there is only one node in S2, so
in S3, there will only the special keys. In these lists, we use pointers to link all the
nodes in S0. Then with additional pointers, we linked these nodes additionally in the
other lists. Unlike 2i, there is not every other node in S1 or every fourth node in S2.
The following figure represents these lists.

S3

S2

S1

S0

31

23

12

23

26

31

34

31

34

64

56

64

78

Figure 40.1:

Now we have the list i.e. from S0 to S3. Actually, these list are made during insert
operation. We will see the insert method later. Lets first talk about the search
method.

Skip List Search

Suppose we have a skip list structure available. Lets see what is its benefit. We
started our discussion from the point that we want to keep linked list structure but do
not want to search n elements for finding an item. We want to implement the
algorithm like binary search on the linked list.

Now a skip list with additional pointers is available. Lets

see how search will work with these additional pointers.
Suppose we want to search an item x. Then search
operation can be described as under.
We search for a key x in the following fashion:
We start at the first position of the top list
At the current position p, we compare x with y key(after(p))

x = y: we return element(after(p))
Page 455 of 519

CS301 Data Structures

___________________________________________________________________

x > y: we scan forward

x < y: we drop down
If we try to drop down past the bottom list, we return NO_SUCH_KEY

To search a key x, we start at the first position of the top list. For example, we are
discussing the top list is S3. We note the current position with p. We get the key in the
list after the current list (i.e. in which we are currently looking for key) by the key
(after(p)) function. We get this key as y. Then we compare the key to be searched for
i.e. x with this y. If x is equal to y then it means y is the element that we are searching
for so we return element(after(p)). If x is greater than y, we scan forward and look at
the next node. If x is less than y, we drop down and look in the down lists. No if we
drop down and past the bottom list, it means that the element (item) is not there and
we return NO_SUCH_KEY.
To see the working of this search strategy lets apply it on the skip list, already
discussed and shown in the figure 40.1. This figure shows four lists. Remember that
these four lists are actually in the one skip list and are made by the additional pointers
in the same skip list. There is not such situation thatS1 is developed by extracting the
data from S0 and S1 duplicates this data. Actually every node exists once and is
pointed by additional pointers. For example, the node 23 exists once but has two next
pointers. One is pointing to 26 while the other pointing to 31. In the same way, there
are three pointers in node 31, two are to the node 34 and the third is toward the +.

S3

S2

S1

S0

31

23

12

23

26

31

34

31

34

64

56

64

78

Figure 40.2: Search for 78

Suppose we want to search 78. We start from the first node of the top list i.e. S3. The
78 will be the current node and we denote it with p. Now we look at the value in the
node after p. In the figure, it is +. Now as the + is greater than 78, we drop down
to S2. Note in the figure that we drop down vertically. We dont go to the first
element p in the down list. This process of going down will be discussed later. Now
we drop from S3 to S2. This is our current pointer. Now we look for the value in the
next node to it. We see that this value is 31. Now 31 is less than 78, so we will do
scan forward. The next node is + that is obviously greater than 78. So we drop from
here and go to the list S1. In this list, the current position is 34. We compare 78 with
Page 456 of 519

CS301 Data Structures

___________________________________________________________________
this node. As 34 is less than 78, we scan forward in the list. The next node in the list
is 64 that is also less than 78. So we look at the next node and note that the next node
is + that is greater than 78. Due to this we drop down to list S0. Here we look at the
next node of 64 and find that this is 78. Thus at last we reach at the node that we are
searching for. Now if we look at the arrows that are actually the steps to find out the
value 78 in the skip list. Then we come to know that these are much less than the links
that we have to follow while starting from the -+ in the list S0. In S0 we have to
traverse the nodes 12, 23, 26, 31, 34, 44, 56, and 64 to reach at 78. This traversal
could be larger if there were more nodes before 31. But in our algorithm, we reach at
node 31 in one step. Thus we see that the search in the skip list is faster than the
search in a linear list. By the steps that we do in searching an element and see that this
search is like the binary search scheme. And we will see that this skip list search is
also log2 N as in binary search tree. Though the skip list is not a tree, yet its find
method works like the binary search in trees. As we know that find method is also
involved in the remove method, so remove method also becomes fast as the find
method is fast.
Insertion in Skip List
When we are going to insert (add) an item (x,0) into a skip list, we use a randomized
algorithm. Note that here we are sending the item in a pair. This is due to the fact that
we keep the data and the key separately to apply the find and remove methods on
table easily. In this pair, x is the key, also present in the record. The 0 denotes the data
(i.e. whole record). Now the randomized algorithm for insertion of this value is
described below.
The first step of the algorithm is that
We repeatedly toss a coin until we get tails, and we denote with i the number
of times the coin came up heads.
This first step describes that we toss a coin while knowing a 50 percent probability for
both head and tail. We keep a counter denoted with i to count the heads that come up.
Now if the head comes up, i becomes 1. We again toss the coin if again the head
comes up we add 1 to the counter i. We continue this counting until the tail comes up.
When tail comes up, we stop tossing and note the value of i.
After this, the second step of algorithm comes, stating that

If i > h, we add to the skip list new lists Sh+1, , Si +1, each containing only
the two special keys

Here we compare i (that is the count of heads came up) with h (that is the number of
list) if i is greater than or equal to h then we add new lists to the skip list. These new
lists are Sh+1, , Si +1. Suppose if i is 8 and h is 4 , we add additional lists S5, S6,
S7, S8 and S9. These lists initially will contain the only two special keys that means and +. The next steps are:

We search for x in the skip list and find the positions p0, p1 , , pi of the
items with largest key less than x in each list S0, S1, , Si
For j 0, , i, we insert item (x, o) into list Sj after position pj

Page 457 of 519

CS301 Data Structures

___________________________________________________________________
Now we will start from the left most node of the top list i.e. Si and will find the
position for the new data item. We find this position as we want to keep the data in
the list sorted.
Lets see this insert position with the help of figures. Suppose we have a skip list ,
shown in the figure below. Here are only three items 10, 23 and 36 in it. There are
also additional layers S1 and S2. These were made when 10, 23 and 36 were inserted.

S2

S1

S0

23

10

23

36

Figure 40.3:
Now we proceed with this list and insert a value in it. The value that we are going to
insert is 15. We have tossed the coin and figured out that the value of i is 2. As it is
randomized algorithm so in this case, the value of i has become 2. The value of I is
the count number of heads before the tail comes up by tossing the coin. From the
above figure, we see that the value of h is 2. As h and i are equal, we are not adding
S3, S4 and S5 to the list. Rather, we will apply the search algorithm. We start from
the left most node of the top list. We call this position p2. We label these positions for
their identification. We see that the item being added i.e.15 is less than the +, so we
drop down to list S1. We denote this step with p1. In this list, the next value is 23 that
is also greater than 15. So we again drop down and come in the list S0. Our current
position is still the first node in the list as we did not have any scan forward. Now the
next node is 10 that is less than 15. So we skip forward. We note this skip forward
with p0. Now after p0 the value in next node is 23 that is greater than 15. As we are in
the bottom list, there is no more drop down. So here is the position of the new node.
This position of new node is after the node 10 and before the node 23. We have
labeled the positions po, p1 and p2 to reach there. Now we add the value 15
additionally to the list that we have traversed to find the positions and labeled them to
remember. After this we add a list S3 that contains only two keys that are - and +,
as we have to keep such a list at the top of the skip list. The following figure shows
this insertion process.

Page 458 of 519

CS301 Data Structures

___________________________________________________________________
S3

S2

15

S1

15

23

S0

15

23

10

36

Figure 40.4: Insertion of 15

Here we see that the new inserted node is in three lists. The value that we insert must
will be in S0 as it is the data list in which all the data will reside. However, it is also
present in the lists that we have traversed to find its position. The value of i in the
insertion method is randomly calculated by tossing the coin. How can we toss a coin
in computers? There is routine library available in C and C++ that generates a random
number. We can give it a range of numbers to generate a number in between them.
We can ask it to give us only 0 and 1 and assign 1 to head and 0 to tail respectively.
Thus the count number of 1s will give us the number of heads that come up. We can
also use some other range like we can say that if the random number is less than some
fixed value (whatever we fixed) that it means head otherwise it will mean tail. The
algorithms that use random numbers are generally known as randomized algorithms.
So
A randomized algorithm performs coin tosses (i.e., uses random bits) to
control its execution
It contains statements of the type
b random()
if b <= 0.5 // head
do A
else
// tail
do B
Its running time depends on the outcome of the coin tosses, i.e, head or tail
In the randomized algorithm, we take a number from the random() function between 0
and 1. This number is up to one decimal place. However, we can keep it up to nay
decimal places. As stated above ,after getting a number we check its value. If this is
less than or equal to 0.5, we consider it as head and execute the process A. In our
algorithm, we increase the value of i by 1. However, if value is greater than 0.5, we
consider it as tail and do the process B. In our algorithm, the process B takes place
when we note the value of I and stop the tossing. We do this process of tossing in a
while loop. The while condition comes false when the tail (i.e. number greater than
0.5) comes. We cannot predict how many times this loop will execute as it depends
upon the outcome of the toss. It is also a random number. There may be only one or a
large number of heads before the tail comes up.
Page 459 of 519

CS301 Data Structures

___________________________________________________________________

Deletion from Skip List

In the remove method, we find the item to be removed with the find item and remove
it from the list. In the lists where ever this item has links, we bypass them. Thus, the
procedure is quite easy. Now lets talk about this method.
To remove an item with key x from a skip list, we proceed as follows:
We search for x in the skip list and find the positions p0, p1 , , pi of the items with
key x, where position pj is in list Sj. This means that we look for the links of the item
to be removed . We know that an item in the list has necessarily link in the list S0.
Moreover it may have links in other lists up to Si or say Sj. After this, we remove
positions p0, p1 , , pi from the lists S0, S1, , Si. We remove all the lists except
the list containing only the two special keys.
Lets consider the skip list shown in the figure below.

S3

S2

S1

S0

12

34

23

34

23

34

45

Figure 40.5 (a) : Remove key 34

S2

S1

S0

23

12

23

45

Figure 40.5 (b): After removal of key 34

Suppose we want to remove the node 34 from the list. When this node 34 was
inserted, it came not only in list S0 but there were its additional links o in S1 and S2

Page 460 of 519

CS301 Data Structures

___________________________________________________________________
lists respectively. The list S3 has only the two special keys. Now for finding 34, we
start from the top list. There is + as the next node, so we drop down to the list S2 as
+ is greater than 34. In S2, we are at node 34. Now we are at the point to remove
34. From here, we go to the remaining lists and reach list S0. We also label the links
being traversed. It is evident that we have labeled the links as p2, p1 and p0. Now we
remove the node 34 and change the pointers in the lists. We see that in S3, this was
the single node. After removing this node, there is only one link that is from - to
+. The list S3 already has link from - to +. Instead of keeping these two i.e. S2
and S3, we keep only S2. Thus we have removed the node 34 from the skip list. We
see that the remove method is simple. We dont have randomness and need not
tossing. The tossing and random number was only in the case of insertion. In the
remove method, we have to change some pointers.
Thus in this data structure, we need not to go for rotations and balancing like AVL
tree. In this data structure- Skip-list, we get rid of the limitation of arrays. This way,
the search gets very fast. Moreover, we have sorted data and can get it in a sorted
form.

Page 461 of 519

CS301 Data Structures

___________________________________________________________________

Data Structures
Lecture No. 41
Reading Material
Data Structures and Algorithm Analysis in C++
10.4.2

Chapter. 10

Summary

Review
Quad Node
Performance of Skip Lists
AVL Tree
Hashing
Examples of Hashing

Review
In the previous lecture, we studied three methods of skip list i.e. insert, find and
remove and had their pictorial view. Working of these methods was also discussed.
With the help of sketches, you must have some idea about the implementation of the
extra pointer in the skip list.
Lets discuss its implementation. The skip list is as under:
S

1
2

3
4

2
3

3
4

2
3

3
4

4
5

We have some nodes in this skip list. The data is present at 0, 1st and 2nd levels. The
actual values are 12, 23, 34 and 45. The node 34 is present in three nodes. It is not
necessary that we want to do the same in implementation. We need a structure with
next pointers. Should we copy the data in the same way or not? Lets have a look at
the previous example:

Page 462 of 519

CS301 Data Structures

___________________________________________________________________
head

tail

Tower Node

20

26

30

40

50

57

60

Here, the data is 20, 26, 30, 40, 50, 57, 60. At the lowest level, we have a link list. A
view of the node 26, node 40 and node 57 reveals that there is an extra next pointer.
The head pointer is pointing to a node from where three pointers are pointing at
different nodes.
We have seen the implementation of link list. At the time of implementation, there is
a data field and a next pointer in it. In case of doubly link list, we have a previous
pointer too. If we add an extra pointer in the node, the above structure can be
obtained. It is not necessary that every node contains maximum pointers. For
example, in the case of node 26 and node 57, there are two next pointers and the node
40 has three next pointers. We will name this node as TowerNode.
TowerNode will have an array of next pointers. With the help of this array of pointers,
a node can have multiple pointers. Actual number of next pointers will be decided by
the random procedure. We also need to define MAXLEVEL as an upper limit on
number of levels in a node. Now we will see when this node is created. A node is
created at a time of calling the insert method to insert some data in the list. At that
occasion, a programmer flips the coin till the time he gets a tail. The number of heads
represents the levels. Suppose we want to insert some data and there are heads for six
times. Now you know how much next pointers are needed to insert which data. Now
we will create a listNode from the TowerNode factory. We will ask the factory to
allocate the place for six next pointers dynamically. Keep in mind that the next is an
array for which we will allocate the memory dynamically. This is done just due to the
fact that we may require different number of next pointers at different times. So at the
time of creation, the factory will take care of this thing. When we get this node from
the factory, it has six next pointers. We will insert the new data item in it. Then in a
loop, we will point all these next pointers to next nodes. We have already studied it in
the separate lecture on insert method.
If your random number generation is not truly so and it gives only the heads. In this
case, we may have a very big number of heads and the Tower will be too big, leading
to memory allocation problem. Therefore, there is need to impose some upper limit on
it. For this purpose, we use MAXLEVEL. It is the upper limit for the number of next
pointers. We may generate some error in our program if this upper limit is crossed.
The next pointers of a node will point at their own level. Consider the above figure.
Suppose we want to insert node 40 in it. Its 0 level pointer is pointing to node 50. The
2nd pointer is pointing to the node 57 while the third pointer pointing to tail. This is
the case when we use TowerNode. This is one of the solutions of this problem.

Page 463 of 519

CS301 Data Structures

___________________________________________________________________

Quad Node
Lets review another method for this solution, called Quad node. In this method, we
do not have the array of pointers. Rather, there are four next pointers. The following
details can help us understand it properly.
A quad-node stores:
item
link to the node before
link to the node after
link to the node below
link to the node above

This will require copying of the key (item) at different levels. We do not have an
array of next pointers in it. So different ways are adopted to create a multilevel node
of skip list. While requiring six levels, we will have to create six such nodes and copy
the data item x in all of these nodes and insert these in link list structure. The
following figure depicts it well.
S3

S2

S1

S0

31

23

12

23

26

31

34

31

34

64

44

56

64

78

You can see next and previous and down and up pointers here. In the bottom layer, the

Page 464 of 519

CS301 Data Structures

___________________________________________________________________
down pointer is nil. Similarly the right pointers of right column are nil. In the top
layer, the top pointers are nil. You can see that the values 23, 34, and 64 are copied
two times and the value 31 is copied three times. What are the advantages of quad
node? In quad node, we need not to allocate the array for next pointers. Every list
node contains four pointers. The quad node factory will return a node having four
pointers in it. It is our responsibility to link the nodes up, bottom, left and right
pointers with other nodes. With the help of previous pointer, we can also move
backward in the list. This may be called as doubly skip list.

Performance of Skip Lists

Lets analyze this data structure and see how much time is required for search and
deletion process. The analysis is probability-based and needs lot of time. We will do
this in some other course. Lets discuss about the performance of the skip list
regarding insert, find and remove methods.
In a skip list, with n items the expected space used is proportional to n. When we
create a skip list, some memory is needed for its items. We also need memory to
create a link list at lowest level as well as in other levels. This memory space is
proportional to n i.e. number of items. For n items, we need n memory locations. The
items may be of integer data type. If we have 100 items, there will be need of 100
memory locations. Then we need space for next pointers that necessitates the
availability of 2n memory locations. We have different layers in our structure. We do
not make every node as towerNode neither we have a formula for this. We randomly
select the towerNode and their height. The next pointers can be up to maxLevel but
normally these will be few. We do not require pointers at each level. We do not need
20n or 30n memory locations. If we combine all these the value will be 15n to 20n.
Therefore the proportionality constant will be around 15 or 20 but it cant be n2 or n3.
If this is the case then to store 100 items we do need 1002 or 1003 memory locations.
There are some algorithms in which we require space in square or cubic times. This is
the space requirement and it is sufficient in terms of space requirements. It does not
demand too much memory.
Lets see the performance of its methods. The expected search, insertion and deletion
time is proportional to log n. It looks like binary tree or binary search tree (BST). This
structure is proposed while keeping in mind, the binary search tree. You have
witnessed that if we have extra nodes, search process can be carried out very fast. We
can prove it with the probabilistic analyses of mathematics. We will not do it in this
course. This information is available in books and on the internet. All the searches,
insertions and deletions are proportional to log n. If we have 100,000 nodes, its log n
will be around 20. We need 20 steps to insert or search or remove an element. In case
of insert, we need to search that this element already exists or not. If we allow
duplicate entries then a new entry would be inserted after the previous one. In case of
delete too, we need to search for the entry before making any deletion. In case of
binary search tree, the insertion or deletion is proportional to log n when the tree is a
balanced tree. This data structure is very efficient. Its implementation is also very
simple. As you have already worked with link list, so it will be very easy for you to
implement it.

AVL Tree
Page 465 of 519

CS301 Data Structures

___________________________________________________________________
The insertion, deletion and searches will be performed on the basis of key. In the
nodes, we have key and data together. Keep in mind the example of telephone
directory or employee directory. In the key, we have the name of the person and the
entry contains the address, telephone number and the remaining information. In our
AVL tree, we will store this data in the nodes. Though, the search will be on the key,
yet as we already noticed that the insert is proportional to log n. Being a balanced
tree, it will not become degenerated balance tree. The objective of AVL tree is to
make the binary trees balanced. Therefore the find will be log n. Similarly the time
required for the removal of the node is proportional to log n.
key

key

entry

entry

key

key

entry

entry

We have discussed all the five implementations. In some implementations, time

required is proportional to some constant time. In case of a sorted list, we have to
search before the insertion. However for an unsorted list, a programmer will insert the
item in the start. Similarly we have seen the data structure where insertions, deletions
and searches are proportional to n. In link list, insertions and deletions are
proportional to n whereas search is log n. It seems that log n is the lower limit and we
cannot reduce this number more.
Is it true that we cannot do better than log n in case of table? Think about it and send
your proposals. So far we have find, remove and insert where time varies between
constant and log n. It would be nice to have all the three as constant time operations.
We do not want to traverse the tree or go into long loops. So it is advisable to find the
element in first step. If we want to insert or delete, it should be done in one step. How
can we do that? The answer is Hashing.

Hashing
The hashing is an algorithmic procedure and a methodology. It is not a new data
structure. It is a way to use the existing data structure. The methods- find, insert and
remove of table will get of constant time. You will see that we will be able to do this
in a single step. What is its advantage? If we need table data structure in some
program, it can be used easily due to being very efficient. Moreover, its operations are
of constant time. In the recent lectures, we were talking about the algorithms and
procedures rather than data structure. Now we will discuss about the strategies and
Page 466 of 519

CS301 Data Structures

___________________________________________________________________
methodologies. Hashing is also a part of this.
We will store the data in the array but TableNodes are not stored consecutively. We
are storing the elements data in the TableNodes. You have seen the array
implementation of the Table data structure. We have also seen how to make the data
sorted. There will be no gap in the array positions whether we use the sorted or
unsorted data. This means that there is some data at the 1st and 2nd position of array
and then the third element is stored at the 6th position and 4th and 5th positions are
empty. We have not done like this before. In case of link list, it is non-consecutive
data structure with respect to memory.
In Hashing, we will internally use array. It may be static or dynamic. But we will not
store data in consecutive locations. Their place of storage is calculated using the key
and a hash function. Hash function is a new thing for you. See the diagram below:

array
index

hash
function

Key

We have a key that may be a name, or roll number or login name etc. We will pass
this key to a hash function. This is a mathematical function that will return an array
index. In other words, an integer number will be returned. This number will be in
some range but not in a sequence.
Keys and entries are scattered throughout the array. Suppose we want to insert the
data of our employee. The key is the name of the employee. We will pass this key to
the hash function which will return an integer. We will use this number as array
index. We will insert the data of the employee at that index.
key

entry

10

123

Page 467 of 519

CS301 Data Structures

___________________________________________________________________
The insert will calculate place of storage and insert in TableNode. When we get a new
data item, its key will be generated with the help of hash function to get the array
index. Using this array index, we insert the data in the array. This is certainly a
constant time operation. If our hash function is fast, the insert operation will also
rapid. It will take only one step to perform this.
Next we have find method. It will calculate the place of storage and retrieve the entry.
We will get the key and pass it to the hash function and obtain the array index. We get
the data element from that array position. If data is not present at that array position, it
means data is not found. We do not need to find the data at some other place. In case
of binary search tree, we traverse the tree to find the element. Similarly in list
structure we continue our search. Therefore find is also a constant time operation with
Hashing.
Finally, we have remove method. It will calculate the place of storage and set it to
null. That means it will pass the key to the hash function and get the array index.
Using this array index, it will remove the element.

Examples of Hashing
Lets see some examples of hashing and hash functions. With the help of these
examples you will easily understand the working of find, insert and remove methods.
Suppose we want to store some data. We have a list of some fruits. The names of
fruits are in string. The key is the name of the fruit. We will pass it to the hash
function to get the hash key.
Suppose our hash function gave us the following values:
HashCode ("apple")
=
hashCode ("watermelon") =
hashCode ("grapes")
=
hashCode ("cantaloupe")
=
hashCode ("kiwi")
=
hashCode ("strawberry")
=
hashCode ("mango")
=
hashCode ("banana")
=

5
3
8
7
0
9
6
2

Our hash function name is hashCode. We pass it to the string apple. Resultantly, it
returns a number 5. In case of watermelon we get the number 3. In case of grapes
there is number 8 and so on. Neither we are sending the names of the fruits in some
order to the function, nor is function returning the numbers in some order. It seems
that some random numbers are returned. We have an array to store these strings. Our
array will look like as:

Page 468 of 519

CS301 Data Structures

___________________________________________________________________
kiwi

0
1
2

banana

watermelon

4
5

apple

mango

cantaloupe

grapes

strawberry

We store the data depending on the indices got from the hashCode. The array size is
10. In case of apple, we get the index 5 from hashCode so apple is stored at array
index 5. As we are dealing with strings, so the array will be an array of strings. The
watermelon is at position 3 and so on every element is at its position. This array
will be in the private part of our data structure and the user will not know about it. If
our array is table then it will look like as under:
table[5]
table[3]
table[8]
table[7]
table[0]
table[9]
table[6]
table[2]

=
=
=
=
=
=
=
=

"apple"
"watermelon"
"grapes"
"cantaloupe"
"kiwi"
"strawberry"
"mango"
"banana"

We will store our data in the Table array using the string copy. The user is storing the
data using the names of the fruits and wants to retrieve or delete the data using the
names of fruits. We have used the array for storage purposes but did not store the data
consecutively. We store the data using the hash function which provides us the array
index. You can note that there are gaps in the array positions.
Similarly we will retrieve the data using the names of fruit and pass it to the
hashCode to get the index. Then we will retrieve the data at that position. Consider
the table array, it seems that we are using the names of fruits as indices.
table["apple"]
table["watermelon"]
table["grapes"]
table["cantaloupe"]
table["kiwi"]
Page 469 of 519

CS301 Data Structures

___________________________________________________________________
table["strawberry"]
table["mango"]
table["banana"]
We are using the array as table [apple], table [watermelon] and so on. We are not
using the numbers as indices here. Internally we are using the integer indices using
the hashCode. Here we have used the fruit names as indices of the array. This is
known as associative array. Now this is the internal details that we are thinking it as
associative array or number array.
Lets discuss about the hashCode. How does it work internally? We pass it to strings
that may be persons name or name of fruits. How does it generate numbers from
these? If the keys are strings, the hash function is some function of the characters in
the strings. One possibility is to simply add the ASCII values of the characters.
Suppose the mathematical notation of hash function is h. It adds all the ASCII values
of the string characters. The characters in a string are from 0 to length - 1. Then it will
take mod of this result with the size of the table. The size of the table is actually the
size of our internal array. This formula can be written mathematically as:

lengt 1

=
h( str ) str[i ] %TableSize
i =0

Example : h( ABC ) = (65 + 66 + 67) TableSize

Suppose we use the string ABC and try to find its hash value. The ASCII values of
A, B and C are 65, 66 and 67 respectively. Suppose the tableSize is 55. We will add
these three numbers and take mod with 55. The result (3.6) will be the hash value. To
represent character data in the computer ASCII codes are used. For each character we
have a different bit pattern. To memorize this, we use its base 10 values. All the
characters on the keyboard like $, %, have ASCII values. You can find the ASCII
table in you book or on the internet.
Lets see the C++ code of hashCode function.
int hashCode( char* s )
{
int i, sum;
sum = 0;
for(i=0; i < strlen(s); i++ )
sum = sum + s[i]; // ascii value
return sum % TABLESIZE;
}
The return type of hashCode function is an integer and takes a pointer to character. It
declares local variable i and sum, then initializes the sum with zero. We use the strlen
function to calculate the length of the string. We run a loop from 0 to length 1. In

Page 470 of 519

CS301 Data Structures

___________________________________________________________________
the loop, we start adding the ASCII values of the characters. In C++, characters are
stored as ASCII values. So we directly add s[i]. Then in the end, we take mod of sum
with TABLESIZE. The variable TABLESIZE is a constant representing the size of the
table.
This is the one of the ways to implement the hash function. This is not the only way
of implementing hash function. The hash function is a very important topic. Experts
have researched a lot on hash functions. There may be other implementations of hash
functions.
Another possibility is to convert the string into some number in some arbitrary base b
(b also might be a prime number). The formula is as:

length 1
i

h( str ) str[i] b %T
i =0

Example : h( ABC ) = (65b 0 + 66b1 + 67b 2)%T

We are taking the ASCII value and multiply it by b to the power of i. Then we
accumulate these numbers. In the end, we take the mod of this summation with
tableSize to get the result. The b may be some number. For example, we can take b as
a prime number and take 7 or 11 etc. Lets take the value of b as 7. If we want to get
the hash value of ABC using this formula:
H(ABC) = (65 * 7 ^0 + 66 * 7^1 + 67 * 7^2) mod 55 = 45
We are free to implement the hash function. The only condition is that it accepts a
string and returns an integer.
If the keys are integers, key%T is generally a good hash function, unless the data has
some undesirable features. For example, if T = 10 and all keys end in zeros, then
key%T = 0 for all keys. Suppose we have employee ID i.e. an integer. The employee
ID may be in hundreds of thousand. Here the table size is 10. In this case, we will take
mod of the employee ID with the table size to get the hash value. Here, the entire
employee IDs end in zero. What will be the remainder when we divide this number
with 10? It will be 0 for all employees. So this hash function cannot work with this
data. In general, to avoid situations like this, T should be a prime number.

Page 471 of 519

CS301 Data Structures

___________________________________________________________________

CS301 Data Structures

___________________________________________________________________

Summary

Collision
Linear Probing

In the previous lecture, we were discussing about the hash functions. The hash
algorithm depends on the hash function. The hash function generates the array index
to enable us to insert data in the table array. We have seen two examples of hash
functions in the previous lecture. Both the functions use the ASCII values of
characters to generate the index. Here the question arises how can we implement the
hash function in case of having integer data? We may have employee ID, user ID or
student ID as integers. Here we may take mod with some number or table size and the
result is used as an array index.
If the keys are integers then key%T is generally a good hash function unless the data
has some undesirable features. If we want to store the employee record, user record or
student record in the table, this can be done through hash function. We take the mod
of the value with the T. The value of T may be 10, 15 or 100 depending on the
requirements. There may be some problem. For example, if T = 10 and all keys end in
zeros, then key%T = 0 for all keys. The hash function gives 0 for all the keys, used as
array index. Now it is a problem. We cannot store our values as all the records have
same index i.e. 0. In general, to avoid such situations, T should be a prime number.
Internally, we have to store the data in the array and there is complete freedom to use
this array by taking the size of our own choice.
We have to store some data in the array. As the array is private, we will decide about
its size on our own. We will take the size of the array in prime numbers. To store 100
records, we will take prime number near 100. We will select this prime number as
MAXTABLESIZE. Then we will use this number in our hash function. This will help
resolve the problem arising due to situation where all keys end with 0. Using the
prime number, the values from the hash function will not be 0 for all the keys.
With the help of prime number, we cannot solve this problem completely. Similarly, it
cannot be made sure that the values from the hash function are unique for all the keys.
Sometimes, we may have same index for two different keys. This phenomenon is
known as collision i.e. the hash values are same of two different keys. How can we
solve this collision problem?

Collision
Collision takes place when two or more keys (data items) produce the same index.
Lets see the previous example of storing the names of fruits. Suppose our hash
function gives us the following values:

Page 473 of 519

CS301 Data Structures

___________________________________________________________________
hash("apple")
hash("watermelon")
hash("grapes") =
hash("cantaloupe")
hash("kiwi")
=
hash("strawberry")
hash("mango") =
hash("banana") =

= 5
= 3
8
= 7
0
= 9
6
2

kiwi

1
2

banana

watermelon

4
5

apple

mango

cantaloupe

grapes

strawberry

We store these data items in the respective index in the array. In the above example,
the index given by the hash function does not collide with any other entry. Suppose
we want to add another fruit honeydew in it. When honeydew is passed to the
hash function, we get the value 6 i.e.
hash("honeydew") = 6
This is not the responsibility of the hash function to see the data in the array before
generating the index. Hash function is generally a mathematical formula that takes the
keys and returns a number. It is responsibility of the caller to find its solution. We
have already mango at position 6. The user of our ADT calls the insert function
giving the value honeydew. We call the hash function to find out its index that
comes out to be 6. Now the problem is this that position 6 is already occupied. What
should we do to avoid it?
There are a lot of solutions of this problem. These solutions can be divided into two
main categories. One type of solutions is the changing of the hash function. Even with
the introduction of a new function, it is not guaranteed that there will be no collision
with future data. Second option is that we live with the collision and do something to
resolve it.
The definition of collision is:
When two values hash to the same array location, this is called a collision
We cannot say that the usage of this hash function will not result in collision
especially when the data is changing. Collisions are normally treated as a
phenomenon of first come, first served, the first value that hashes to the location
gets it. We have to find something to do with the second and subsequent values that
Page 474 of 519

CS301 Data Structures

___________________________________________________________________
hash to the same location.
As we have seen above in the example that at position 6, there is the data item mango
while another item honeydew is trying to acquire the same position. First come first
served means that the mango will remain at its position and we will do something for
honeydew. We will see three solutions for this problem:
Solution #1: Search for an empty location
Can stop searching when we find the value or an empty location.
Search must be wrap-around at the end.
Lets apply this solution on the above example. We have mango at position 6. Now
the position for honeydew is also 6. We find another empty location in the array. Keep
in mind that in hashing, we do not store data at consecutive positions. Rather, data is
scattered. We know that there are some empty locations in the array. We do not want
to refuse the user that we cannot store the data due to the occupation of the position
by some other item. Therefore, we will store the data item at the empty location. We
will see an example in a short while.
Solution #2: Use a second hash function
...and a third, and a fourth, and a fifth, ...
We have a primary hash function, we pass it to the string honeydew. It returns 6. As
the position 6 is occupied, so we call another hash function that is implemented in a
different way. This hash function will return an integer. If this location is empty, we
will store our data here and the problem is solved. But if the array location at the
index number returned by this function is also occupied. Then we will call the third
hash function. There is a possibility of collision so we will have three, four, five or
more hash functions. When there is a collision, we call these hash functions one by
one till the time, there is an index with empty location. We will store the data at this
empty location.
Solution #3: Use the array location as the header of a linked list of values that hash to
this location
In this solution, we will not store the data directly in the array. Our array will be an
array of pointers to TableNode. We will create a list node and store the data in it. The
array will have a pointer to the node. If we want to store some data at location 6, we
will store a pointer at location 6 that points to the node containing the data. When we
have another data at the position 6, we create a list node and attach it with the
previous node. There is a possibility of having a link list from each location of the
table. At least, it will have one node.
There are some more strategies for the resolution of this problem. But we will study
only the above-mentioned three solutions. Now we will see how we can implement
the methods of Table and Dictionary ADT using the hashing function. Firstly, we will
see how the insert, find and remove methods will work with the first solution.

Linear Probing
Page 475 of 519

CS301 Data Structures

___________________________________________________________________
The first solution is known as open addressing. When there is a collision, we try to
find some other place in our array. This approach of handling collisions is called open
addressing; it is also known as closed hashing. The word open is used with
addressing and the word closed is used with hashing. Be careful when naming these.
More formally, cells at h0(x), h1(x), h2(x), are tried in succession where
hi(x) = (hash(x) + f(i)) mod TableSize, with f(0) = 0.
Here hash is our hash function. If there is some collision, we add f(i) value to it before
taking its mod with TableSize. The function, f, is the collision resolution strategy.
We use f(i) = i, i.e., f is a linear function of i. Thus
location(x) = (hash(x) + i) mod TableSize
The function f can be any function. So our first implementation of f is that whatever
integer (i) we gave to it, it returns it back. That is f(i) is a linear function. We can
implement function f as it returns the square of i. That will be quadratic function.
Similarly, we can have cubic functions. We are free to implement it.
In the above example, we are given x for insertion. This can be string such as
honeydew. We call the hash(x) that returns a number. We will see that this location is
empty or not. In case of collision, we add i to this number. The value of i can be
0,1,2,3 etc. We will take the mod of this value with TableSize and try to store the data
there. The collision resolution strategy is called linear probing as it scans the array
sequentially (with wrap around) in search of an empty cell. Lets consider an example
to understand how it works. Keep in mind that when there is a collision, some other
location in the array is found. This is known as linear probing. Here we are trying to
probe the array linearly that where is the empty location.
Lets see an example. Here we have names of birds being stored in the array. We have
a larger array. We already have some data in the array that is as under:
...
141
142

robin

143

sparrow

144

hawk

145

seagull

146
147
148
...

bluejay
owl

Page 476 of 519

CS301 Data Structures

___________________________________________________________________
This is a part of array. There is some data before and after these locations. As shown
above, the locations 142, 143, 144, 145, 147 and 148 contains data. Suppose we want
to add seagull to this hash table. With respect to ADT, someone is using this Table or
Dictionary and calls the insert method of this Table. Now we have to use the seagull
as key and have to store its associated data. We will call the hashCode function to get
the array index and pass it the string seagull as:
hashCode(seagull) = 143
Now the location 143 is not empty and we have a bird, sparrow at this location. That
is table[143] is not empty. The seagull is different than sparrow i.e. table[143] !=
seagull. If it already exists and we are not allowing duplicates, we can say that the
data already exists. This is not the case here. We will check the next position, 143+1
(here i is 1) i.e. 144. At location 144, we have a bird hawk. So this position is also not
empty and not equal to seagull. Here also we have different data. Now we check the
next position 143+2 (here i is 2) i.e. 145. This location is empty so we store our data
at this location. We have two collisions here at 143 and 144 positions. We have found
the empty position in the array with single jumps. First we add 1, then 2 and so on.
Suppose we want to add some existing data. As collisions are happening here, so
when we get the index value from hash function, the data may not be at that position.
We have to follow this linear chain. Suppose you want to add hawk to this hash table.
The table is as under:
...
141
142

robin

143

sparrow

144

hawk

145

seagull

146
147
148
...

bluejay
owl

We know that hawk is already in the table. Also seagull is added in the table. We will
see that that data that we want to insert already exists or not. First of all, we call the
hash function that is hashCode as:
hashCode(hawk) = 143

Page 477 of 519

CS301 Data Structures

___________________________________________________________________
We use this value as the table index to insert this data in the table. The table[143] is
not empty. Then we check the data at position 143. We have stored sparrow here so
table[143] != hawk. At this point, we cannot decide whether the data item hawk
exists in the table or not. As we are doing linear probing and there are chances that
this data may exist at some other location. Therefore we will check it further taking
any final decision. Now we check the next position i.e. 144. It is also not empty but
the data here is hawk that is table[144] == hawk. We have two options here. One is
that hawk is already in the table, so do nothing. The other option is that user might
want to modify the data associated with hawk.
The size of our internal array is fixed and we store it in the constant TableSize. In
linear probing, at the time of collisions, we add one to the index and check that
location. If it is also not empty, we add 2 and check that position. Suppose we keep on
incrementing the array index and reach at the end of the table. We were unable to find
the space and reached the last location of the array. Now what does it mean? Is the
table completely filled? There are chances that we started from the position 143.
Suppose that the table size is 500. We were unable to find any empty location. Can
we say that the array locations before 143 are also not available? There is a possibility
that we find some empty spaces there. We will not stop at the end of the table and go
to the start of this table array. We will start linear probing from the start and if we find
some empty spaces, the data will be stored there. This process is called wrap around.
There is a chance that in wrap around, we could not find space and come back to the
143 position. The solution of this problem is to get a larger array. We may want to
create an array of size 1000 or some prime number larger than 500. Then copy all the
data in this new array. This new data may be inserted with the help of linear probing.
The other solution is that we may want to keep multiple arrays. In this case, when one
array is filled we go to the second array.
Lets see the example of wrap around process. Suppose we want to add cardinal to
this hash table. We will use the above table to insert this data. Also suppose that 148
is the last position of this table. We call the hashCode function as:
hashCode(cardinal) = 147
We get the array index as 147. This position is already filled in the table. The data at
this position is bluejay. So we will add 1 to the index and check the 148 position. This
position is also filled. We do not stop here and go back to the start of the array. We
cannot increment the array index further. We will treat the table as circular; after 148
comes 0. Now we will check the position 0. If it is occupied, we will see the position
1 then 2 and so on. We have seen the insert method using the linear probing. Now we
will see the find method.
Suppose we want to find hawk in this hash table. We call the hashCode as:
hashCode(hawk) = 143
The table[143] is not empty. However, the data at this position is not hawk. So with
the linear probing procedure, we add 1 to the index and check the 144 position. This
position is also not empty. We compare it with hawk that is true. We use the same

Page 478 of 519

CS301 Data Structures

___________________________________________________________________
procedure for looking things up in the table as generally done for inserting them.
Lets have a look on the delete procedure. If an item is placed in array[hash(key)+4],
the item just before it is deleted. How will probe determine that the hole does not
indicate the item is not in the array? We may have three states for each location as:
Occupied
Empty (never used)
Deleted (previously used)
Using the linear probe, we insert data where we get some empty space in the array.
Firstly, we try to insert it at the index position given by the hash function. If it is
occupied, we move to the next position and so on. In this way, we have a chain to
follow. Now if an element of the chain is deleted, how can we know that it was filled
previously. As we have seen in the above example that seagull collided twice before
getting a space in the array. Despite having a hash value of 143, it was stored at 145.
The three locations 143, 144 and 145 have different data values but are a part of a
chain. These three names collide with each other with respect to hash function.
Suppose we delete the 2nd name i.e. hawk due to some reason. Now there are two
names, which collide. A space has been created due to the deletion of hawk. If we try
to search seagull using the linear probing, there will be a hash value of 143. It is filled
but has different data. We move to the next position that is empty. How can linear
probing know that do not stop here because there was some data previously at this
position, which has not been deleted. And there is some data after this position, which
is part of this chain due to the collisions and that data is seagull.
To deal with such situations, we keep three states in our array. 1) It is empty (never
used) and has no data in it. 2) It is filled with some legal data and is occupied and 3) it
had some data which is now deleted and currently it is empty. We will use this state if
we have to go to next position using the linear probing.
One problem with linear probing technique is the tendency to form clusters. A
cluster is a group of items not containing any open slots. The bigger a cluster gets, the
more likely the new values will hash into the cluster, and make it even bigger.
Clusters cause efficiency to degrade. Suppose we want to add another bird in our
table. The hash function returns the index as 143. Now we have already three items
that collided at 143. The array locations at 143, 144, 145 are already occupied. So we
will insert this new data at 146 using the linear probing. The data is getting gathered
instead of scattering because linear probing inserts the data in the next position. It
seems as the normal use of the array in which we insert data in the array from first
position then next position and so on. It may depend on our data or our hash function.
This gathering of data is called clustering.
We were trying to store the data in the array in a constant time or in a single step.
Similarly the find and remove methods should be of constant time. That attribute has
now been lost. What should we do now? One of the solutions is quadratic probing.
Quadratic probing uses different formula:
Use F(i) = i2 (square of i) to resolve collisions
If hash function resolves to H and a search in cell H is inconclusive, try H +

Page 479 of 519

CS301 Data Structures

___________________________________________________________________
12, H + 22, H + 32,
In the quadratic probing when a collision happens we try to find the empty location at
index + 12. If it is filled then we add 22 and so on. Lets take the above example. We
want to insert seagull. The hash function generates the value 143 and this location is
already occupied. Now we will start probing. First of all, we add 12 in the 143 and
have 144. It is also not empty. Now we will add 22 in the 143 and have 147. If it is
also occupied, we will add 32 in 143 and we have 152. The data is now getting
scattered. Unfortunately, there are some problems with quadratic probing also.
Lets discuss the third solution that we will use the link list for the collision
resolution. Each table position is a linked list. When we are going to insert new data,
we will insert the keys and entries anywhere in the list (front easiest). It is shown in
the below diagram.

key entry

key entry

4
key entry

key entry

10

key entry
123
On the left side, we have vertical array that contains the pointers in it. When we insert
the first item, we attach a list node. When we have collision, the new item is inserted
in the start of the link list. Suppose an item is stored at position 4 and we have another
data, requiring the position 4. So there is a collision and we will insert the data at the
start of the list.

Page 480 of 519

CS301 Data Structures

___________________________________________________________________
Lets compare the link list and open addressing.

Advantages over open addressing:

Simpler insertion and removal
Array size is not a limitation
Disadvantage
Memory overhead is large if entries are small.

The problem in linear probing is that when our array is full what we should do. This
problem can be solved using the link list.
In the next lecture, we will continue our discussion on hashing. We will see an
animation. Hashing is very important methodology and can be used in data structures
besides Table and Dictionary.

Data Structures
Lecture No. 43
Reading Material
Data Structures and Algorithm Analysis in C++
5.4, 5.5, 5.6, 7.1

Chapter. 5, 7

Summary

Hashing Animation
Applications of Hashing
When Hashing is Suitable?
Sorting
Elementary Selection Algorithms
Selection Sort

Hashing Animation
In the previous lecture, we discussed about collision strategies in hashing. We studied

Page 481 of 519

CS301 Data Structures

___________________________________________________________________
three solutions, Linear Probing, Quadratic Probing and Linked List chaining. Hashing
is vast research field, which covers hash functions, storage and collision issues etc. At
the moment, we will see hashing in implementation of table ADT. Operations of
insert, delete and find are performed in constant time using this hashing strategy.
Constant time means the time does not increase with the increase in data volume.
However, if collisions start happening then the time does not remain constant.
Especially if we see linear probing, we had to insert the data by sorting the array
sequentially. Similar was the case with quadratic. In case of linked list, we start
constructing linked list that takes time and memory. But later we will see some
situations where hashing is very useful.
Today, we will study these three strategies of hash implementation using animations.
These animations will be provided to you in a Java program. It is important to
mention here that all the data structures and algorithms we have studied already can
be implemented using any of the languages of C/C++ or Java. However, it is an
important decision to choose the programming language because every language has
its strong area, where it has better application. Java has become very popular because
of its facilities for Internet. As you already know C++, therefore, Java is easy to learn
for you. The syntax is quite similar. If we show you the java code you will say it is
C++.
Lets see the hashing animation. This animation will be shown in the browser. This is
an applet written in java language. We will see linear probing, quadratic probing and
link list chaining in it. This is an example of how do we solve collision.
We have an array shown in four different columns. The size of the array is 100 and
the index is from 0 to 99. Each element of the array has two locations so we can store
200 elements in it. When we have first collision the program will use the 2nd part of
the array location. When there is a 2nd collision the data is stored using the linear
probing. At the top right corner we have hash function x and its definition is mod
100. That is when a number is passed to it, it will take mod with 100 and return the
result which is used as index of the array.
In this example we are using numbers only and not dealing with the characters. We
also have a statistical table on the right side. This program will generate 100 random
numbers and using the hash function it will store these in the array. The numbers will
be stored in the array at different locations. In the bottom left we have hashing
algorithms. We have chosen the linear probing. Now the program will try to solve this
problem using the linear probing. Press the run button to start it. It is selecting the
numbers randomly, dividing it by 100 and the remainder is the hash value which is
used as array index.
Here we have number 506. It is divided by 100 and the remainder is 6. It means that
this number will be stored at the sixth array location. Similarly we have a number
206, now its remainder is also 6. As location 6 is already occupied so we will store
206 at the 2nd part of the location 6. Now we have the number 806. Its remainder is
also 6. As both the parts of location 6 are occupied. Using the linear probing we will
store it in the next array location i.e. 7. If we have another number having the
remainder as 6, we will store it at the 2nd part of location 7. If we have number 807, its
remainder is 7. The location 7 is already occupied due to the linear probing. Therefore

Page 482 of 519

CS301 Data Structures

___________________________________________________________________
the number 807 will be stored using the linear probing in the location 8. Now you can
understand how the numbers are stored in the array. You can also see some clustering
effect in it. See the location 63. All the numbers having remainder as 63 are clustered
around location 63.
Lets change the collision resolution algorithm to quadratic probing. Run the
animation again. Now we have array size as 75 and the array is shown in three
columns. Each location of the array can store two numbers. In quadratic probing we
add square of one first i.e. 1 and then the square of two and so on in case of collisions.
Here we have used a different hash function. We will take the mod with 75. When the
both parts of the array location is filled we will use the quadratic probing to store the
next numbers. Analyze the numbers and see where the collisions have happened.
Lets see the animation using the linked list chaining. Now the hash function uses 50
to take mod with the numbers. So far pointers are not shown. When both parts of the
location are filled, we will see the link list appearing. We have four numbers having
remainder 0. The two numbers will be stored in the array and the next two will be
stored using the link list which is attached at the 0 location.
We are not covering the hashing topic in much depth here as it is done in algorithms
and analysis of algorithms domain. This domain is not part of this course. For the time
being, we will see the usage of hashing. For certain situations, table ADT can be used,
which internally would be using hashing.

Applications of Hashing
Lets see few examples of those applications where hashing is highly useful. The
hashing can be applied in table ADT or you can apply hashing using your array to
store and retrieve data.
Compilers use hash tables to keep track of declared variables (symbol table).
Compilers use hash tables in order to implement symbol tables. A symbol table is an
important part of compilation process. Compiler puts variables inside symbol table
during this process. Compiler has to keep track of different attributes of variables. The
name of the variable, its type, scope and function name where it is declared etc is put
into the symbol table. If you consider the operations on symbol table, those can be
insertion of variable information, search of a variable information or deletion of a
variable. Two of these insert and find are mostly used operations. You might have
understood already that a variable name will be parameter (or the key) to these
operations. But there is one slight problem that if you named a variable as x outside of
a code block and inside that code block, you declared another variable of the similar
type and name then only name cannot be the key and scope is the only differentiating
factor. Supposing that all the variables inside the program have unique names,
variable name can be used as the key. Compiler will insert the variable information by
calling the insert function and by passing in the variable information. It retrieves the
variable value by passing in the variable name. Well, this exercise is related to your
Compiler Construction course where you will construct you own language compiler.
Another usage of hashing is given below:
A hash table can be used for on-line spelling checkers if misspelling detection
Page 483 of 519

CS301 Data Structures

___________________________________________________________________
(rather than correction) is important, an entire dictionary can be hashed and
words checked in constant time.
You must have used spell checkers in a word processing program like MS Word. That
spell checker finds out mistakes, provides you correct options and prompts you to
choose any of the synonyms. You can also set the correct the words automatically.
Hashing can be used to find the spelling mistakes. For that you first take all the words
from the dictionary of spoken English and construct a hash table of those. To find the
spelling mistakes, you will take first word from the text that is being checked and
compare it with all the words present inside the hash table. If the word is not found in
the hash table then there is a high probability that the word is incorrect, although there
is a low probability that the word is correct but it is not present in the dictionary.
Based on the high probability a message can be displayed to the user of the
application that the word is wrong. MS Word does the same. As far the automatic
correct feature is concerned, it is another algorithm, which we are not going to discuss
here.
Lets see few more examples in this connection.
Game playing programs use hash tables to store seen positions, thereby saving
computation time if the position is encountered again.
Normal computer games are graphical, there are positions that are chosen by the
computer or by the player. Consider the game of chess where one player has chosen
one position again. Here we can use the positions of the pieces (64 pieces) at that time
as the key to store it in the hash table. If our program wants to analyze that if a player
has encountered the similar situation before, can pass in the positions of the pieces to
the function find. Inside the function, when the positions are hashed again then the
previously present index is returned, which shows that the similar situation has been
encountered before.
See another example below:
Hash functions can be used to quickly check for inequality if two elements hash
to different values they must be different.
Sometimes in your applications, you dont want to know which value is smaller or
bigger but you are only interested in knowing if they are equal or not. For this, we can
use hashing. If the two data items dont collide, then their hash values will be
different. Based on this two values are said to be unequal.
Above was the situation when hashing can be useful. You may like to know in what
circumstances hashing is not a good solution to apply.

When Hashing is Suitable?

Hash tables are very good if there is a need for many searches in a reasonably
stable table.
We have just seen the excellent example of reasonably stable hash table when we
discussed hash table for English dictionary. We had constructed a hash table of all the

Page 484 of 519

CS301 Data Structures

___________________________________________________________________
words inside dictionary and were looking for different words in it. So majorly, there
were frequent look up operations and insertions were in very minor frequency.
Hash tables are not so good if there are many insertions and deletions, or if table
traversals are needed in this case, AVL trees are better.
In some applications, it is required to frequently read and write data. In these kinds of
applications hash table might not be a good solution, AVL tree might be a good
option. But bear in mind that there are no hard and fast statistics to go for hash table
and then to AVL tree. You have to be a good software engineer to choose relevant
data structure.
Also, hashing is very slow for any operations which require the entries to be
sorted
o e.g. Find the minimum key
At times, you do other operations of insert, delete and find but additionally, you
require the data in sorted order or the minimum or maximum value. We have
discussed it many times that we insert data in the array of hash table without any sort
order and it is scattered through the array in such a fashion that there are holes in the
array. In these circumstances, the hash table is not useful. You might be remembering
from the animation we saw earlier on in this lecture that there was no real sequence of
filling of array. Some clusters were formed because of collision but there was no
order as such. So hashing is not really useful in these circumstances.
We are finishing with our discussion on hashing. The important thing is how we
thought about one data structure and internally we implemented in six different ways.
You must be remembering that as long as the interface of the data structures remains
the same, different internal implementations does not really matter from the client
perspective. Occasionally, somebody might be interested in knowing the internal
implementation of your data structure because that might be important for him in
order to use your data structure.
Lets move on to the next topic of Sorting. It is very vast topic and cannot be covered
in this course thoroughly.

Sorting
Sorting means to put the data in a certain order or sequence. We have discussed
sorting before in different scattered through topics in this course but it has not been
discussed so far as a separate topic. You must be remembering that when we traverse
the binary search tree in in-order way, the obtained data happens to be sorted.
Similarly, we saw other data structures, where we used to keep data in sorted order. In
case of min-heap if we keep on removing elements one by one, we get data in sorted
order.
Sorting is so useful that in 80-90% of computer applications, sorting is there in one
form or the other. Normally, sorting and searching go together. Lot of research has
been done on sorting; you can get lot of stuff on it from different sources. Very
efficient algorithms have already been developed for it. Moreover, a vast

Page 485 of 519

CS301 Data Structures

___________________________________________________________________
Mathematical analysis has been performed of these algorithms. If you want to expose
yourself, how these analyses are performed and what Mathematical tools and
procedures are employed for performing analysis then sorting is very useful topic for
you.

Sorting Integers
How to sort integers in this array?

20

10

10 20

Fig 43.1
We want to sort the numbers given in the above array. Apparently, this operation may
seem very simple. But think about it, if you are given a very large volume of data
(may be million of numbers) then you may realize that there has to be an efficient
mechanism to perform this operation. Firstly, lets put the problem in words:
We have a very large array of numbers. We want to sort the numbers inside the array
in ascending order such that the minimum number of the array will be the first
element of it and the largest element will be the last element at the end of the array.
Lets go to the algorithms of sorting. Point to be noted here that we are going to study
algorithms of sorting; we are not talking about data structures. Until now, you might
have realized that algorithms go along data structures. We use a data structure to
contain data and we use algorithms to perform certain operations or actions on that
data.

Elementary Sorting Algorithms

Selection Sort
Insertion Sort
Bubble Sort
These algorithms have been put as elementary because these are very simple. They
will act as our baseline and we will compare them with other algorithms in order to
find a better algorithm.

Selection Sort
Main idea:
o find the smallest element
o put it in the first position
Page 486 of 519

CS301 Data Structures

___________________________________________________________________
o find the next smallest element
o put it in the second position

And so on, until you get to the end of the list

This technique is so simple that you might have found it yourself already. You search
the whole array and find the smallest number. The smallest number is put on the first
position of the array while the previous element in this position is moved somewhere
else. Find the second smallest number and then put that number in the second position
in the array, again the previous number in that position is shifted somewhere else. We
keep on performing this activity again and again and eventually we get the array
sorted. This technique is called selection sort because we select elements for their
sorted positions.
In the next lecture, we will see how we can optimize this sorting operation. You read
about sorting in your textbooks and from the Internet.

Data Structures
Lecture No. 44
Reading Material
Data Structures and Algorithm Analysis in C++

Chapter. 7
7.1, 7.2

Summary

Selection Sort
o Selection Sort analysis
Insertion Sort
o Insertion Sort Analysis
Bubble Sort
o Bubble Sort analysis
Summary
N log2 (N) Algorithms

This is the sequel of the previous lecture in which we talked about the sort algorithms.
Page 487 of 519

CS301 Data Structures

___________________________________________________________________
There are three elementary sorting methods being discussed in this hand out. These
are- selection sort, insertion sort and bubble sort. To begin with, we will talk about the
selection sort algorithm.

Selection Sort
Suppose we have an array with different numbers. For sorting it in an ascending
order, selection sorting will be applied on it in the following manner. We find the
smallest number in the array and bring it to the position 1. We do the same process
with the remaining part of the array and bring the smallest number among the
remaining array to the next position. This process continues till the time all the
positions of the array are filled with the elements. Thus the main idea of selection sort
is that
find the smallest element
put it in the first position
find the next smallest element in the remaining elements
put it in the second position

And so on, until we get to the end of the array
Lets understand this algorithm by considering an example with the help of figures.
Consider an array that has four numbers i.e. 19, 5, 7 and 12 respectively. Now we
want to sort this array in ascending order. To sort the array, selection algorithm will
be applied on it.

19

12

19

12

19

12

12

19

12

19

Figure 44.1: Selection Sort

The above pictorial representation explains the selection sort. It describes that at the

Page 488 of 519

CS301 Data Structures

___________________________________________________________________
start, we begin the search for the smallest number from the first position of the array
i.e. from the index zero. We find that 5 is the smallest number and bring it to the first
position i.e. index 0. Later, number 19 is put at the position that was occupied by
number 5. Thus in a sense, we swap these numbers. Now 5, the smallest number in
the array, has come at its final position i.e. index 0.
As index 0 has the proper number, so we start our search from position 2 i.e. index 1.
Now we look at the remaining elements 19, 7, 12 and find the smallest number among
them. Here 7 is the smallest so we change its position with 19 to bring 7 at its
position. Thus 5 and 7 have come at their final positions. Now there are two elements
are left behind i.e. 12 and 19. We search the smallest among these and find that 12 is
the smallest. So we swap it with 19 to bring it at index 2 i.e. its final position. Now
there is last element remaining and obviously it is at its position as there is no element
to compare with it. The array is now in the sorted form with ascending numbers.
The point to remember in the selection search is that at the beginning, we start search
for the smallest number from index 0 (i.e. first position). After it we start search from
the index 1 (i.e. position 2). After each search one number gets its final position so we
start the next search from the next position to it. Thus we do the multiple passes of the
array to sort it. First, we pass through the n elements of the array and search the n-1
elements and then n-2. Thus at last, we come to the single and last element of the
array.
Now lets see the code of this algorithm in C++.
void selectionSort(int *arr, int N)
{
int posmin, count, tmp ;
for (count=0;count<N;count++)
{
posmin = findIndexMin(arr, count, N) ;
tmp=arr[posmin] ;
arr[posmin]=arr[count] ;
arr[count]=tmp ;
}
}
int findIndexMin (int *arr, int start, int N)
{
int posmin=start ;
int index ;
for(index=start; index < N; index++)
if (arr[index]<arr[posmin])
posmin=index ;
return posmin ;
}
In the above code, we write the function selectionSort. This function takes an array of
integers as *arr and the size of array as N. There are local variables declared in
function as posmin, count and tmp. Then there is a for loop that starts from zero and

Page 489 of 519

CS301 Data Structures

___________________________________________________________________
goes to N-1. This is due to the fact that the index of array of N elements is from zero
to N-1. The condition count < N indicates that loop will execute as long as count is
less than N and will exit when count gets N. Now in the loop, we calculate the
posmin with a function i.e. findIndexMin. This findIndexMin method is written below
in the code. This routine or method works in the way that we pass to it the whole
array i.e. arr, the value of count (what it is in that execution of loop) and size of the
array i.e. N. This routine starts the search from the index equal to value of count and
goes to Nth position, returning the position of the minimum (smallest) element. Now
in the first routine, we get this position value in the variable posmin. Thus the code
line:
posmin = findIndexMin(arr, count, N) ;
gets the position of the smallest number returned by the method findIndexMin in the
variable posmin. After finding this position, we do the swapping of this posmin with
the count position. Thus we put the value of position posmin in the count position.
The findIndexMin routine is such that it takes arr, start and N as arguments. It starts
searching from the start position in the for loop, finds the position of the minimum
value and returns that position.
This sorting algorithm is also known as the in-place sorting algorithm as there is no
need of additional storage to carry out this sorting. The pictorial representation of the
swap action of this algorithm will be as under:
We want to sort the following array that has five elements. We start the search from
the first element and find that the smallest element is 5. As we have started our search
from index 0 (that is the count in our above code) so we swap 5 with the value at
index 0 i.e. 20. After this swap, 5comes at the position of 20 while 20 goes to the
position of 5. Afterwards, we start the search from index 1. We find that 7 is the
smallest number among the remaining elements. It swaps its position with 8. After
this, in the remaining three elements, 8 is the smallest. This number 8 swaps its
position with 20. At the end, 10 is the smallest number among 10 and 20. So there is
no need of swapping as 10 has already got its position.

Page 490 of 519

CS301 Data Structures

___________________________________________________________________

20

10

20

10

20

10

10

20

10

20

Figure 44.2: Swap Action (selection sorting)

Selection Sort Analysis

We have seen in the code for the algorithm that there are two loops. The loop in the
selectionSort method passes the array to search the smallest element and the second
loop that is in the findIndexMin method finds the position (index) of the smallest
value. To find the first smallest element, we have to go through the N elements of the
array. For the purpose of finding the second smallest element, we have to search the
N-1 elements. During this search process, we have to find two elements for the
second last smallest element. And obviously in the end, there is one element that is at
its proper position, necessitating no search. We have to do all these searches. These
are N, N-1, N-2 2, 1 for the first, second, third second last and last element
respectively. Now if we want to find the total searches, the addition of all these
searches together will help us get a sum total as given in the following equation.
Total searches = 1 + 2 + 3 + .+ (N-2) + (N-1) + N
= N (N+1) / 2
= (N2 + N) / 2
Suppose if N is 10, then according to this formula, the total searches will be (100 +
10) / 2 i.e. 55. If N is 100, the total searches will be (10000 + 100) / 2 i.e. 5050.
Similarly if N is 1 million, N2 is going to be very large as compared to N. Thus there
we can ignore N and can say that the time (total searches) is proportional to N2. This
means that larger the N, greater will be the performance of selection with respect to
N2.

Insertion Sort
Page 491 of 519

CS301 Data Structures

___________________________________________________________________
The main idea of insertion sort is
Start by considering the first two elements of the array data. If found
out of order, swap them
Consider the third element; insert it into the proper position among the
first three elements.
Consider the fourth element; insert it into the proper position among
the first four elements.

This algorithm is not something uncommon to the persons who know card playing. In
the game of cards, a player gets 13 cards. He keeps them in the sorted order in his
hand for his ease. A player looks at the first two cards, sorts them and keeps the
smaller card first and then the second. Suppose that two cards were 9 and 8, the player
swap them and keep 8 before 9. Now he takes the third card. Suppose, it is 10, then it
is in its position. If this card is of number 2, the player will pick it up and put it on the
start of the cards. Then he looks at the fourth card and inserts it in the first three cards
(that he has sorted) at a proper place. He repeats the same process with all the cards
and finally gets the cards in a sorted order. Thus in this algorithm, we keep the left
part of the array sorted and take element from the right and insert it in the left part at
its proper place. Due to this process of insertion, it is called insertion sorting.
Lets consider the array that we have used in the selection sort and sort it now with
the insertion sorting. The following figure shows the insertion sort of the array
pictorially.

19

12

12

19

12

19

12

19

Figure 44.3: Insertion Sort

The array consists of the elements 19, 12, 5 and 7. We take the first two numbers i.e.
19 and 12. As we see 12 is less than 19, so we swap their positions. Thus 12 comes at
index 0 and 19 goes to index 1. Now we pick the third number i.e. 5. We have to find
the position of this number by comparing it with the two already sorted numbers.

Page 492 of 519

CS301 Data Structures

___________________________________________________________________
These numbers are 12 and 19. We see that 5 is smaller than these two. So it should
come before these two numbers. Thus the proper position of 5 is index 0. To insert it
at index 0, we shift the numbers 12 and 19 before inserting 5 at index 0. Thus 5 has
come at its position. Now we pick the number 7 and find its position between 5 and
12. To insert 7 after 5 and before 12, we have to shift the numbers 12 and 19 to the
right. After this shifting, we put number 7 at its position. Now the whole array has
been sorted so the process stops here.
Following is the code of the insertion sort in C++.
void insertionSort(int *arr, int N)
{
int pos, count, val;
for(count=1; count < N; count++)
{
val = arr[count];
for(pos=count-1; pos >= 0; pos--)
if (arr[pos] > val)
arr[pos+1]=arr[pos];
else break;
arr[pos+1] = val;
}
}
In this sorting function, we start the process from index 0 and 1 and swap them.
Afterwards, we go to the third position and put it into its proper position. While
inserting a number in the sorted portion of the array, we have to shift the elements.
This shifting is an additional overhead of this sorting algorithm. Due to this shifting,
the sorting requires more time. This algorithm is also an in place algorithm as we
dont need any additional storage. At the most, we need some variables of local
nature, normally used in programs.
From the above code, we see that the name of this sorting routine is insertionSort. It
takes an array and its size as arguments. There are local variables pos, count and val
declared in this function. After this there is a for loop that starts from the count having
value equal to 1. Now we put the value of count index (i.e. arr[count]) in variable val.
This value has to find its place in the left sorted portion of the array. To find that
position we have to execute one more for loop. This loop starts from count-1 and goes
to down to zero. In the body of the loop we compare the value of val with the value at
current position in the array i.e. arr[pos]. If val is less than the arr[pos] i.e. value at
current index. It means that the val has to go to the left position of arr[pos].So we
shift the arr[pos] to right to create place for the new value i.e. val. When the loop
exits, we put this value val at arr[pos + 1]. Thus we have inserted the number in the
array at its proper position.
Following is the step by step explanation for the insertion sort of the above example
with same previous array.
First of all we take the first two elements 19 and 12. As 12 is less than 19, we do right

Page 493 of 519

CS301 Data Structures

___________________________________________________________________
shift on 19. And put 12 at its position i.e. index 0. Afterwards, we go to index 2.
There is 5 at this position. As we see that 5 is less than the other elements on the left
side of array, it has to come at the first position. To bring 5 to first position, the
number 12 and 19 has to be shifted to right. After this shifting, the position of index 0
becomes empty and we put 5 there. Finally, there is number 7. The position of 7 will
be between 5 and 12. For this, we have to shift 12 and 19 towards right so that the
place for 7 could be created. We put 7 at that place. Thus the array is sorted now.

Insertion Sort Analysis

Lets analyze that when the value of N increases. How much time for insertion sort
increases? In the code of insertion sort, we have seen that there are outer and inner
loops. Due to these two loops, we can understand that it is also like N2 algorithm. In
the sort process, there may be a situation that every iteration inserts an element at the
start of the array by shifting all sorted elements along. Now if we have to bring the
second element to its position, there will be need of shifting the first element. This
means that we have to shift one element. Similarly, for placing the third element at the
start position (we are discussing the worst case scenario in which at every iteration the
element has to go to the first position), we have to shift two elements. Thus we sum
up all the shiftings, the total becomes 2 + 3 + 4 +. + N-1 + N.
The summation can be written as follows.
Total = (2 + N ) (N -1) / 2
= O (N2)
From this expression, we see that when the value of N increases, the value of N2 will
dominate. It will increase significantly with respect to N. Thus we see that insertion
sort is also an N2 algorithm like selection sort.

Bubble Sort
The third sorting algorithm is bubble sort. The basic idea of this algorithm is that we
bring the smaller elements upward in the array step by step and as a result, the larger
elements go downward. If we think about array as a vertical one, we do bubble sort.
The smaller elements come upward and the larger elements go downward in the array.
Thus it seems a bubbling phenomenon. Due to this bubbling nature, this is called the
bubble sort. Thus the basic idea is that the lighter bubbles (smaller numbers) rise to
the top. This is for the sorting in ascending order. We can do this in the reverse order
for the descending order.
The steps in the bubble sort can be described as below
Exchange neighboring items until the largest item reaches the end of
the array
Repeat the above step for the rest of the array
In this sort algorithm, we do not search the array for the smallest number like in the
other two algorithms. Also we do not insert the element by shifting the other
elements. In this algorithm, we do pair-wise swapping. We will take first the elements
and swap the smaller with the larger number. Then we do the swap between the next
pair. By repeating this process, the larger number will be going to the end of the array
Page 494 of 519

CS301 Data Structures

___________________________________________________________________
and smaller elements come to the start of the array.
Lets try to understand this phenomenon with the help of figures how bubble sort
works. Consider the same previous array that has elements 19, 12, 5 and 7.
a:

a:

a:

a:

a:

19

12

19

12

12

19

12

19

12

19

a:

a:

a:

a:

a:

12

19

12

19

12

19

12

19

12

19

First of all, we compare the first pair i.e. 19 and 5. As 5 is less than 19, we swap these
elements. Now 5 is at its place and we take the next pair. This pair is 19, 12 and not
12, 7. In this pair 12 is less than 19, we swap 12 and 19. After this, the next pair is 19,
7. Here 7 is less than 19 so we swap it. Now 7 is at its place as compared to 19 but it
is not at its final position. The element 19 is at its final position. Now we repeat the
pair wise swapping on the array from index 0 to 2 as the value at index 3 is at its
position. So we compare 5 and 12. As 5 is less than 12 so it is at its place (that is
before 12) and we need not to swap them. Now we take the next pair that is 12 and 7.
In this pair, 7 is less than 12 so we swap these elements. Now 7 is at its position with
respect to the pair 12 and 7. Thus we have sorted the array up to index 2 as 12 is now
at its final position. The element 19 is already at its final position. Note that here in
the bubble sort, we are not using additional storage (array). Rather, we are replacing
the elements in the same array. Thus bubble sort is also an in place algorithm. Now as
index 2 and 3 have their final values, we do the swap process up to the index 1. Here,
the first pair is 5 and 7 and in this pair, we need no swapping as 5 is less than 7 and is
at its position (i.e. before 7). Thus 7 is also at its final position and the array is sorted.
Following is the code of bubble sort algorithm in C++.
void bubbleSort(int *arr, int N)
{
int i, temp, bound = N-1;
int swapped = 1;
while (swapped > 0 )

Page 495 of 519

CS301 Data Structures

___________________________________________________________________
{
swapped = 0;
for(i=0; i < bound; i++)
if ( arr[i] > arr[i+1] )
{
temp = arr[i];
arr[i] = arr[i+1];
arr[i+1] = temp;
swapped = i;
}
bound = swapped;
}
}
In line with the previous two sort methods, the bubbleSort method also takes an array
and size of the array as arguments. There is i, temp, bound and swapped variables
declared in the function. We initialize the variable bound with N1. This N-1 is our
upper limit for the swapping process. The outer loop that is the while loop executes as
long as swapping is being done. In the loop, we initialize the swapped variable with
zero. When it is not changed in the for loop, it means that the array is now in sorted
form and we exit the loop. The inner for loop executes from zero to bound-1. In this
loop, the if statement compares the value at index i and i+1. If I (element on left side
in the array) is greater than the element at i+1 (element on right side in the array) then
we swap these elements. We assign the value of i to the swapped variable that being
greater than zero indicates that swapping has been done. Then after the for loop, we
put the value of swapped variable in the bound to know that up to this index,
swapping has taken place. After the for loop, if the value of swapped is not zero, the
while loop will continue execution. Thus the while loop will continue till the time, the
swapping is taking place.
Now lets see the time complexity of bubble sort algorithm.

Bubble Sort Analysis

In this algorithm, we see that there is an outer loop and an inner loop in the code. The
outer loop executes N times, as it has to pass through the whole array. Then the inner
loop executes for N times at first, then for N-1 and for N-2 times. Thus its range
decreases with each of the iteration of the outer loop. In the first iteration, we do the
swapping up to N elements. And as a result the largest elements come at the last
position. The next iteration passes through the N-1 elements. Thus the part of the
array in which swapping is being done decreases after iteration. At the end, there
remains only one element where no swapping is required. Now if we sum up these
iterations i.e. 1 + 2 + 3 + + N-1 + N. Then this summation becomes N (1 + N)
/ 2 = O (N2). Thus in this equation, the term N2 dominates as the value of N increases.
It becomes ignorable as compared to N2. Thus when the value of N increases, the time
complexity of this algorithm increases proportional to N2.

Summary
Now considering the above three algorithms, we see that these algorithms are easy to
understand. Coding for these algorithms is also easy. These three algorithms are in
place algorithms. There is no need of extra storage for sorting an array by these
Page 496 of 519

CS301 Data Structures

___________________________________________________________________
algorithms. With respect to the time complexity, these algorithms are proportional to
N2. Here N is the number of elements. So we can see that as the value of N increases,
the performance time of these algorithms increases considerably as it is proportional
to N2. Thus these algorithms are expensive with respect to time performance. There
are algorithms that have the time complexity proportional to N log2 (N). The
following table shows the respective values of N2 and N log2(N) for some values of
N.

N2

10
100
100
10000
1000
1000000
10000
100000000
100000 10000000000
1000000
1E+12

N Log2 (N)
33.21
664.38
9965.78
132877.12
1660964.04
19931568.57

From this table we can see that for a particular value of N, the value of N2 is very
large as compared to the value of N log2 (N). Thus we see that the algorithms whose
time complexity is proportional to N2 are much time consuming as compared to the
algorithms the time complexity of which is proportional to N log2 (N). Thus we see
that the N log2 (N) algorithms are better than the N2 algorithms.

N log2 (N) Algorithms

Now lets see the algorithms that are N log2 (N) algorithms. These include the
following algorithms.
Merge Sort
Quick Sort
Heap Sort
These three algorithms fall under divide and conquer category. The divide and
conquer strategy is well known in wars. The philosophy of this strategy is , divide
your enemy into parts and then conquer these parts. To conquer these parts is easy,
as these parts cannot resist or react like a big united enemy. The same philosophy is
applied in the above algorithms. To understand the divide and conquer strategy in
sorting algorithm, lets consider an example. Suppose we have an unsorted array of
numbers is given below.
10

12

11

Now we split this array into two parts shown in the following figure.
10

12

11

Page 497 of 519

CS301 Data Structures

___________________________________________________________________
Now we have two parts of the array. We sort these parts separately. Suppose we sort
these parts with an elementary sort algorithm. These parts may be sorted in the
following manner.
4

10

12

11

After this we merge these two parts and get the sorted array as shown below.
2

10

11

12

Data Structures
Lecture No. 45
Reading Material
Data Structures and Algorithm Analysis in C++
7.6, 7.7

Chapter. 7

Summary

Divide and Conquer

Mergesort
mergeArrays
Mergesort and Linked Lists
Mergesort Analysis
Quicksort
Course Overview

Page 498 of 519

CS301 Data Structures

___________________________________________________________________

Divide and Conquer

In the previous lecture, we had started discussing three new sorting algorithms; merge
sort, quick sort and heap sort. All of these three algorithms take time proportional to
nlog2n. Our elementary three sorting algorithms were taking n2 time; therefore, these
new algorithms with nlog2n time are faster. In search operation, we were trying to
reduce the time from n to log2n.
Lets discuss these sorting algorithms; merge sort, quick sort and heap sort in detail.
We had started our discussion from divide and conquer rule where we also saw an
example. Instead of sorting a whole array, we will divide it in two parts, each part is
sorted separately and then they are merged into a single array.
S p lit th e lis t in to tw o p a rts

10

12

11

11

11

S o rt th e tw o p a rts

10

12

M e rg e th e tw o p a rts to g e th e r

10

12

F ig 4 5 .1

Let see few analysis to confirm the usefulness of the divide and conquer technique.
To sort the halves approximate time is (n/2)2+(n/2)2
To merge the two halves approximate time is n
So, for n=100, divide and conquer takes approximately:
= (100/2)2 + (100/2)2 + 100
= 2500 + 2500 + 100
(n2 = 10,000)
= 5100
We know that elementary three sorting algorithms were taking approximately n2 time.
Suppose we are using insertion sort of those elementary algorithms. We divide the list
into two halves then the time will be approximately (n/2)2+(n/2)2. The time required
for merging operation is approximately n. This operation contains a simple loop that
goes to n.
Suppose that n is 100. Considering if we apply insertion sort algorithm on it then the
time taken will be approximately (100)2 = 10000. Now, if we apply divide and
conquer technique on it. Then for first half approximate time will be (100/2)2.
Similarly for second half it will be (100/2)2. The merging approximate time will be
100. So the whole operation of sorting using this divide and conquer technique in
insertion sort will take around (100/2)2 + (100/2)2+100 = 5100. Clearly the time
spent (5100) after applying divide and conquer mechanism is significantly lesser than
Page 499 of 519

CS301 Data Structures

___________________________________________________________________
the previous time (10000). It is reduced approximately to half of the previous time.
This example shows the usefulness of divide and conquer technique.
By looking at the benefit after dividing the list into two halves, some further questions
arise:

Why not divide the halves in half?

The quarters in half?
And so on . . .
When should we stop?
At n = 1

So we stop subdividing the list when we reach to the single element level. This divide
and conquer strategy is not a thing, we have already prepared binary search tree on
the same lines. One side of the tree contains the greater elements than the root and
other part contains the smaller elements. Especially, when performing binary search
in an array, we had started our search from mid of it. Subdivided the array and kept on
comparing and dividing the array until we got success or failure. The subdivision
process may prolong to individual element of the array.

Recall Binary Search

Search

Search

Search

Fig 45.2
Hence, we used to perform binary search on the same lines of divide and conquer
strategy. Remember, we applied binary search on sorted array. From this one can
realize that sorting facilitates in searching operations.

Sort

Sort

Sort

Sort

Sort

Sort

Sort

Fig 45.3

Page 500 of 519

CS301 Data Structures

___________________________________________________________________
In figure 45.3, in order to sort the list, we have divided the list in the two parts and
each part is subdivided into further subparts. At end each part is consisting of either
single element or maximum two elements. If we have two numbers to sort, we can
compare them (or sort them) with single if statement. After sorting individual
subparts, we start merging them in the upward direction as shown in the figure Fig
45.4.

Combine

Combine

Combine

Fig 45.4
In Fig 45.4, we have four sorted smaller parts. We combine them to become two
sorted parts and two sorted parts are further combined or merged to become one
sorted list.

Mergesort
Mergesort is a divide and conquer algorithm that does exactly that.
It splits the list in half
Mergesorts the two halves
Then merges the two sorted halves together
Mergesort can be implemented recursively
Lets see the mergsort algorithm, how does that work.
The mergesort algorithm involves three steps:
o If the number of items to sort is 0 or 1, return
o Recursively sort the first and second halves separately
o Merge the two sorted halves into a sorted groupIf the data is consisting
of 0 or 1 element then there is nothing required to be done further. If the number of
elements is greater than 1 then apply divide and conquer strategy in order to sort
them. Divide the list into two halves and sort them separately using recursion. After
the halves (subparts) have been sorted, merge them together to get a list of sorted
elements.
We will discuss recursive sorting in a moment, before that lets see the merging
operation using pictures. We have two sorted array and another empty array whose
size is equal to the sum of sizes of two sorted arrays.

Page 501 of 519

CS301 Data Structures

___________________________________________________________________

10

12

11

2
Fig 45.5
You can see from Fig 45.5, array 1 on the top left is containing 4 elements, top right is
also containing 4 elements and both of them are sorted internally. Third array is
containing 4+4 = 8 elements.
Initially, very first elements (present at the starting index 0 of array) of both the arrays
are compared and the smaller of them is placed in the initial position of the third
array. You can see from Fig 45.5 that elements 4 and 2 are compared pointed to by
the indexes (actually arrays current indexes). The smaller of them is 2, therefore, it is
placed in the initial position in the third array. A pointer (the current index of array) is
also shown for third array that will move forward as the array is filled in. The smaller
number was from right array, therefore, its pointer is moved forward one position as
shown in Fig 45.6.

10

12

11

Fig 45.6
This time the numbers at current positions in both the arrays are compared. As 4 is
smaller of the two numbers, therefore, it is put in the third array and third arrays
pointer is moved one position forward. Also because this time, the number has been
chosen from left array, therefore, its pointer is also moved forward. The updated
figure is shown in Fig 45.7.

Page 502 of 519

CS301 Data Structures

___________________________________________________________________

10

12

11

Fig 45.7
Next, numbers 8 and 5 are compared. As 5 is smaller of the two, it is put in the third
array. The changed positions of pointers and the next comparison are shown in Fig
45.8.

10

12

11

Fig 45.8
Fig 45.9 has showed the situation after next comparison done in the similar manner.

10

12

11

Fig 45.9
By now, you must have understood how the merging operation works. Remember, we
do merging when we have two parts sorted already.
Consider this operation in terms of time and complexity. It is performed using a
simple loop that manipulates three arrays. An index is used for first array, which starts
from the initial position to the size of the array. Similar an index is used for second
array, which starts from the initial position and ends at the end of the array. A third
index is used for third array that sizes to the sum of the maximum values of both
previous indexes. This is simple single loop operation, which is not complex.

Page 503 of 519

CS301 Data Structures

___________________________________________________________________
Now, lets see sorting using recursion now pictorially.
Split the list in half.

10

Split the list in half.

10

10

12

11

Mergesort the left half.

Split the list in half.

Mergesort the left half.

12

Mergesort the left half.

Mergesort the right.

10

4
Fig 45.10

At the top of Fig 45.10 is an array, which is not sorted. In order to sort it using
recursive mechanism, we divide the array into two parts logically. Only logical
partitions of the array are required bases on size and index of the arrays. The left half
of the array containing 10,4,8 and 12 is processed further for sorting. So actually,
while calling the mergesort algorithm recursively, only that particular half of the array
is a passed as an argument. This is spitted further into two parts; 10, 4 and 8, 12. 10, 4
half is processed further by calling the mergesort recursively and we get both
numbers 10 and 4 as separate halves. Now these halves are numbers and they cannot
be subdivided therefore recursive call to mergesort will stop here. These numbers
(new halves) 10 and 4 are sorted individually, so they are merged. When they are
merged, they become as 4,10 as shown in Fig 45.11.

Page 504 of 519

CS301 Data Structures

___________________________________________________________________

10

12

10

12

11

Mergesort the right half. Merge the two halves.

10

12

Merge the two halves.

12
Fig 45.11

The recursive call to mergesort has done its work for left half that consisted of 10 and
4. Now, we apply the same technique (the recursive calls to mergesort) to the right
half that was consisting of 8 and 12. 8 and 12 are spitted into separate numbers as
indicated in Fig 45.11. Further division of them is not possible, therefore, they are
merged as shown in Fig 45.12.

10

12

10

12

11

Merge the two halves.

Mergesort the right half. Merge the two halves.

10

12

Fig 45.12
At this point in time, we have two parts one is containing elements as 4, 10 and other
as 8, 12. These two parts are merged. The merged half is shown in Fig 45.12. Note
that it is completely sorted as 4,8,10 and 12. This completes the operation on the left
half of the array. We do similar process with the right half now, which is consisting of
11, 2, 7 and 5.

Page 505 of 519

CS301 Data Structures

___________________________________________________________________
Mergesort the right half.

10

10

12

12

11

11

11

11

Fig 45.13
Similarly, left half of it 11,2 is processed recursively. It is divided into two halves and
we have two parts as two individual numbers. The recursive call for this part stops
here and the merging process starts. When the parts 11 and 2 are merged, they
become one sorted part as shown in Fig 45.14.
Mergesort the right half.

10

10

12

12

11

11

11

Fig 45.14
Now the same procedure is applied on right half consisting of 7 and 5. By applying
recursive mechanism, it is further subdivided into two individual number parts.

Page 506 of 519

CS301 Data Structures

___________________________________________________________________
Mergesort the right half.

10

10

12

11

11

12

11

Fig 45.15
The merging operation starts, the resultant is shown in the Fig 45.16. After merging,
the new part has become 5,7.
Mergesort the right half.

10

10

12

12

11

11

11

Fig 45.16
When 5,7 merged part is further merged with already merged part 2,11. The new half
becomes as 2,5,7,and 11 as shown in Fig 45.17.

Page 507 of 519

CS301 Data Structures

___________________________________________________________________
Mergesort the right half.

10

10

12

12

11

11

Fig 45.17
Now, we can merge the two biggest halves to get the array in sorted order. After
merging, the resulted sorted array is shown in Fig 45.18.
Merge the two halves.

10

11

12

Fig 45.18
Lets see the C++ code for this sorting algorithm. We are not going to write a class of
it but at the moment, we are only writing it as a procedure. We can use it standalone
or later on, we can also change it to make it a class member.
void mergeSort(float array[], int size)
{
int * tmpArrayPtr = new int[size];
if (tmpArrayPtr != NULL)
mergeSortRec(array, size, tmpArrayPtr);
else
{
cout << Not enough memory to sort list.\n);
return;
}
delete [] tmpArrayPtr;
}
void mergeSortRec(int array[], int size, int tmp[])
{
int i;
int mid = size/2;
if (size > 1)
{
mergeSortRec(array, mid, tmp);
mergeSortRec(array+mid, size-mid, tmp);
Page 508 of 519

CS301 Data Structures

___________________________________________________________________
mergeArrays(array, mid, array+mid, size-mid, tmp);
for (i = 0; i < size; i++)
array[i] = tmp[i];
}
}
The name of the procedure is mergeSort, it is accepting first argument as an array of
floats and as second argument as an int. The second parameter is containing the size
of the array. Inside the function, we have created a temporary array tmpArrayPtr
dynamically of the same size as the size passed in the second argument. This array
will be used for merging operation. In the next if statement if (tmpArrayPtr !=
NULL), allocation for dynamic array is being checked, if it is allocated then another
function mergeSortRec is called. Its first argument is floats array array , second
argument is int size, and third argument is tmpArrayPtr.
In mergeSortRec, we divided the size of the array by 2 to get the mid of the array. If
the size of the array is greater than 1 then inside the if-statement, mergeSortRec is
being called recursively two times. In first recursive call, mergeSortRec(array, mid,
tmp); we provided the array from index 0 to mid. For second recursive call,
mergeSortRec(array+mid, size-mid, tmp);we provided the array starting from one
position ahead of mid position to the end of the array, however, the size of the passed
array is size-mid. The first recursive call will keep on executing by further
subdividing the left part of the half every time until the size of the array reduces to 1.
Similarly, the second recursive call keeps on working on the right half of the array
until the size of the array reduces to 1. After both of these recursive calls will be
completed then the merging operation is performed by calling mergeArrays. At end,
inside the for loop, the sorted data is copied from tmp array to the actual array array.

mergeArrays
Lets see the merging of two array while using their indexes as i and j.

a:

15

28

30

aSize: 5

b:

10

14

22

43

50

bSize: 6

tmp:
Fig 45.19
We are doing the same operation of merging of arrays but with little more detail here.
One the left is an array a consisting of 2,5,15,28 and 30 and on the right is another
array b consisting of 6,10,14,22,43 and 50. The size of array a is 6 while bs size is 6.
We create a temporary array named tmp, whose size is 11. We have used indexes i, j
and k for arrays a, b and tmp respectively and each one of them has been initialized
with 0.

Page 509 of 519

CS301 Data Structures

___________________________________________________________________

a:

15

28

30

b:

i=0

10

14

22

43

50

j=0

tmp:
k=0

Fig 45.20

We compare the initial elements 3 and 6 of the arrays a and b. As 3 is smaller than 6,
therefore, it is put in the temporary array tmps starting location where k=0.

a:

15

28

30

b:

i=0

10

14

22

43

50

j=0

tmp:

3
k=0

Fig 45.21

i is incremented and now elements 5 and 6 are compared. 5 being smaller of these
take the place in the tmp array as shown in Fig 45.22.

a:

15

28

30

i=1

b:

10

14

22

43

50

j=0

tmp:

3
k=1

Fig 45.22

i is incremented again and it reaches the element 15. Now, elements 15 and 6 are
compared. Obviously 6 is smaller than 15, therefore, 6 will take the place in the tmp
array as shown in Fig 45.23.

Page 510 of 519

CS301 Data Structures

___________________________________________________________________

a:

15

28

30

b:

i=2

10

14

22

43

50

j=0

tmp:

k=2

Fig 45.23

Because this time the element was taken from right array with index j, therefore, this
time index j will be incremented. Keep an eye on the index k that is increasing after
every iteration. So this time, the comparison is made between 15 and 10. 10 being
smaller will take the place in tmp array as shown in Fig 45.24.

a:

15

28

30

b:

i=2

10

14

22

43

50

j=1

tmp:

k=3

10

Fig 45.24

Again, j will be incremented because last element was chosen from this array.
Elements 15 and 14 are compared. Because 14 smaller, so it is moved to tmp array in
the next position.

a:

15

28

30

b:

i=2

10

14

22

43

50

j=2

tmp:

3
k=4

10

14

Fig 45.25

With these movements, the value of k is increasing after each iteration. Next, 22 is
compared with 15, 15 being smaller of two is put into the tmp array as shown in Fig
45.26.

Page 511 of 519

CS301 Data Structures

___________________________________________________________________

a: 3

15

28

30

b:

i=2

10

14

22

43

50

j=3

tmp:

k=5

10

14

15

Fig 45.26

By now , you must have understood how this is working, lets see the next iterations
pictures below:

a: 3

15

28

30

b:

i=3

k=6

15

10

14

15

22

14

22

43

50

b:

10

14

22

43

50

22

43

50

Fig 45.27

28

30

i=3

j=4

tmp:

k=7

a: 3

10

j=3

tmp:

a: 3

15

10

14

15

22

28

b:

10

14

28

30

Fig 45.28

28

30

i=4

j=4

tmp:

3
k=8

10

14

15

22

Fig 45.29

Note that array a is all used up now, so the comparison operation will end here. All
the remaining elements of array b are incorporated in tmp array straightaway as
Page 512 of 519

CS301 Data Structures

___________________________________________________________________
shown in Fig 45.30. Hence, we have the resultant merged array tmp as shown in the
figure.

a: 3

15

28

30

b:

i=5

10

14

22

43

50

j=4
Done.

tmp:

k=9

10

14

15

22

28

30

43

50

Fig 45.30

Mergesort and Linked Lists

Merge sort works with arrays as well as linked lists. Now we see, how a linked list is
sorted. Suppose we have a singly linked list as shown in figure Fig 45.31. We can
divide the list into two halves as we are aware of the size of the linked list. Each half
is processed recursively for sorting. Both of the sorted resultant halves are merged
together.

Sort

Sort

M erge

Fig 45.31

Mergesort Analysis
As you have seen, we used an additional temporary array while performing the
merging operation. At the end of the merging operation, elements are copied from
temporary array to the original array. The merge sort algorithm is not an inplace
sorting algorithm because it requires an additional temporary array of the same size as
the size of the array under process for sorting. This algorithm is still a good sorting
algorithm, we see this fact from its analysis.

Page 513 of 519

CS301 Data Structures

___________________________________________________________________
Merging the two lists of size n /2:

O(n)

Merging the four lists of size n/4:

O(n)

Merging the n lists of size 1:

.
.
.

O(log2n)
times

O(n)

Fig 45.32
Mergesort is O(n log2 n)
Space?
The other sorts we have looked at (insertion, selection) are in-place (only require
a constant amount of extra space)
Mergesort requires O(n) extra space for merging
As shown in Fig 45.32, the array has been divided into two halves. We know that
merging operation time is proportional to n, as it is done in a single loop regardless of
the number of equal parts of the original array. We also know that this dividing the
array into halves is similar mechanism as we do in a binary tree and a complete or
perfect balance tree has log2n number of levels. Because of these two factors, the
merge sort algorithm is called nlog2n algorithm. Now, lets discuss about quick sort
algorithm, which is not only an nlog2n algorithm but also an inplace algorithm. As
you might have guessed, we dont need an additional array while using this algorithm.

Page 514 of 519

CS301 Data Structures

___________________________________________________________________

Quicksort
Quicksort is another divide and conquer algorithm.
Quicksort is based on the idea of partitioning (splitting) the list around a pivot or
split value.
Quicksort is also a divide and conquer algorithm. We see pictorially, how the quick
sort algorithm works. Suppose we have an array as shown in the figure Fig 45.33.
4

12

10

11

5
pivot value

Fig 45.33
We select an element from the array and call it the pivot. In this array, the pivot is the
middle element 5 of the array. Now, we swap this with the last element 3 of the array.
The updated figure of the array is shown in Fig 45.34.
4

12

10

11

low

high
5
pivot value

Fig 45.34
As shown in Fig 45.34, we used two indexes low and high. The index low is started
from 0th position of the array and goes towards right until n-1th position. Inside this
loop, an element that is bigger than the pivot is searched. The low index is
incremented further as 4 is less than 5.
4

12

10

low

11

high
5

pivot value

Fig 45.35
low is pointing to element 12 and it is stopped here as 12 is greater than 5. Now, we
start from the other end, the high index is moved towards left from n-1th position to 0.

Page 515 of 519

CS301 Data Structures

___________________________________________________________________
While coming from right to left, we search such an element that is smaller than 5.
Elements 7 and 11 towards left are greater than 5, therefore, the high pointer is
advanced further towards left. high index is stopped at the next position as next
element 2 is smaller than 5. Following figure Fig 45.36 depicts the latest situation.
4

12

10

low

11

high
5

pivot value

Fig 45.36
Both of the indexes have been stopped, low is stopped at a number 12 that is greater
than the pivot and high is stopped at number 2 that is smaller than the pivot. In the
next step, we swap both of these elements as shown in Fig 45.37.
4

10

low

12

11

high

5
pivot value

Fig 45.37
Note that our pivot element 5 is still there at its original position. We again go to
index low and start moving towards right, trying to find a number that is greater than
the pivot element 5. It immediately finds the next number 10 greater than 5. Similarly,
the high is moved towards left in search to find an element smaller than the pivot
element 5. The very next element 3 is smaller than 5, therefore, the high index stops
here. These elements 10 and 3 are swapped, the latest situation is shown in Fig 45.38.
4

low

10

12

11

high

5
pivot value

Fig 45.38
Now, in the next iteration both low and high indexes cross each other. When the high

Page 516 of 519

CS301 Data Structures

___________________________________________________________________
pointer crosses the low pointer, we stop it moving further as shown in Fig 45.39 and
swap the element at the crossing position (which is 8) with the pivot number as shown
in Fig 45.40.
4

10

12

11

12

11

high low
5
pivot value

Fig 45.39

10

high low

Fig 45.40
This array is not sorted yet but element 5 has found its destination. The numbers on
the left of 5 should be smaller than 5 and on right should be greater than it and we can
see in Fig 45.40 that this actually is the case here. Notice that smaller numbers are on
left and greater numbers are on right of 5 but they are not sorted internally.
Next, we recursively quick sort the left and right parts to get the whole array sorted.
4

42

Quicksort the left part

10
6

12

11

Quicksort the right part

Fig 45.41
Now, we see the C++ code of quick sort.
void quickSort(int array[], int size)
{
int index;
if (size > 1)
{
index = partition(array, size);
quickSort(array, index);
quickSort(array+index+1, size - index-1);
}
}

Page 517 of 519

CS301 Data Structures

___________________________________________________________________
int partition(int array[], int size)
{
int k;
int mid = size/2;
int index = 0;
swap(array, array+mid);
for (k = 1; k < size; k++){
if (array[k] < array[0]){
index++;
swap(array+k, array+index);
}
}
swap(array, array+index);
return index;
}
An array and its size are passed as arguments to the quickSort function. The function
declared a local variable index and the size of the array is checked in the next
statement. If the size of the array is more than 1 then the function does the recursive
calling mechanism to sort the array. It divides the array into two parts by choosing the
pivot element. In the subsequent calls, firstly the left side is sorted and then the right
side using the recursive mechanism. Quicksort algorithm is very elegant and it can
sort an array of any size efficiently. It is considered one of the best general purpose
algorithms for sorting. This normally is preferred sorting method being nlog2n and
inplace algorithm. You are advised to read more about this algorithm from your text
books and try to do it as an exercise.
Todays lecture being the last lecture of the course, lets have a short review of it.

Course Overview
We had started this course while keeping the objectives of data structures in mind that
appropriate data structures are applied in different applications in order to make them
work efficiently. Secondly, the applications use data structures that are not memory
hungry. In the initial stages, we discussed array data structure. After we found one
significant drawback of arrays; their fixed size, we switched our focus to linked list
and different other data structures. However, in the meantime, we started realizing the
significance of algorithms; without them data structures are not really useful rather I
should say complete.
We also studied stack and queue data structures. We implemented them with array
and linked list data structures. With their help, we wrote such applications which
seemed difficult apparently. I hope, by now, you must have understood the role of
stack in computers runtime environment. Also Queue data structure was found very
helpful in Simulations.
Later on, we also came across the situations when we started thinking that linear data
structures had to be tailored in order to achieve our goals with our work, we started
studying trees then. Binary tree was found specifically very useful. It also had a
degenerate example, when we constructed AVL trees in order to balance the binary
search tree. We also formed threaded binary trees. We studied union/find data
structure, which was an up tree. At that time, we had already started putting special

Page 518 of 519

CS301 Data Structures

___________________________________________________________________
importance on algorithms. We found one important fact that it is not necessary that
for every application we should use a new data structure. We formed new Abstract
Data Types (ADT) using the existing data structures. For dictionary or table data
structure, we majorly worked with ADTs when we implemented them in six different
ways. We also studied Skip List within the topic of Table ADT, which is a very recent
data structure. After that we discussed about Hashing. Hashing was a purely
algorithmic procedure and there was nothing much as a data structure.
In future, you will more realize the importance of algorithm. While solving your
problems, you will choose an algorithm to solve your problem and that algorithm will
bring along some data structure along. Actually, data structure becomes a companion
to an algorithm. For example, in order to build your symbol table while constructing
your own compiler, you will use hashing. For searches, trees will be employed.
One important fact here is that the data structures and algorithms covered in this
course are not complete in the sense that you dont need any other data structure
except them. One example is Graph, which is not discussed much in this course.
Graph data structures are primarily important from algorithmic point of view.
Now, you should examine yourself, what have you learned in this course. As a
software engineer, you have learned data structures and algorithmic skills to increase
your domain knowledge of design choices. You can apply these design choices in
order to resolve different design problems of your applications that you will come
across in your student and professional life.

Page 519 of 519

Page 390: [1] Deleted

Abutt

5/26/2003 8:59:00 PM

we can use buildHeap to construct heap. buildHeap is optimized as compared to insert

method because it takes lesserthan Nlog2N time. We also proved a theorem that if the
number of linkes in the tree are counted then the sum of heights of the nodes is N-h-1.
We have been iteratinglot of times that the best use of heap is implementing priority
queues. Lets try to develop a class of priority queue using heap. You might be
remembering, when we did bank simulation, weused a priority queue but that was
implementedusing an array. Now, we will implement the same queue using a heap. We
will modify the same code to use heap instead of any array. The interface (.h file) will
remain the same but the implementation (.cpp file) will be changed. As .cpp file will
change, so lets see the .cpp code below:
The first line is including a file Event.cpp, this is containing all the events used for
simulation. We are including .cpp files here as we have used templates of C++. At the
second line, we have included heap.cpp. In the third line, we are defining a constant
PQMAX, declaring the maximum size of the priority queue to be 30. In line 4, class
PriorityQueue is declared. public keyword is given at line 6 indicating that all class
members below will be of public scope. Line 7 is starting the class constructors
definition. At line 9, a new heap object is being created, this object is a collection of
Event type objects and the number of elements in the collection is PQMAX. The address
(pointer) of the newly created heap is stored in the heap object pointer. The heap is a
private pointer variable of PriorityQueue class. Next comes the destructor of the class,
where we are deleting (deallocating the allocated resources) the pointer variable heap.
Next isthe remove method that was sometimes, named as dequeue in our previous
lectures. remove method is returning an Event pointer. Inside remove, the first statement
at line 19 is an if-condition; which is checking whether heap is not empty. If the
condition is true (i.e. the heap is not empty), a local variable (local to the block) Event*
variable e is declared. In the next line at line 22, we are calling deleteMin method to
delete (remove) an element from the heap, the removed element is assigned to the passed
in parameter pointer variable e. You might have noticed already that in this version of
deleteMin, the parameter is being passed by pointer, in the previous implementation we
used reference. In the next line, the retrieved value is returned, the function returns by
returning the pointer variable e. But if the heap was empty when checked at line 19, the
control is transferred to the line 25. At line 25, a message is displayed to show that the
heap is empty and the next line returns a NULL pointer.
It is important to understand one thing that previously when we wrote the array based
remove method. We used to take an element out from the start of the array and shifted the
remaining elements to left. As in this implementation, we are using heap, therefore, all
the responsibilities of maintaining the heap elements lie with the heap. When the
deleteMin() is called, it returns the minimum number.
????
We see the insert method now line by line. It is accepting a pointer type parameter of
type Event. As the heap may become full if we keep on inserting elements into it,

therefore, the first line inside the insert function is checking if the heap has gone full. If
the heap is not full then the if-block is entered. At line 33 inside the if-block, an element
that is passed as a parameter to insert method of heap is inserted into the queue by calling
the insert method of heap. This insert call will internally perform percolate up and down
operations to place the new element at its correct position. The returned value 1 indicates
the successful insertion in the queue. If the heap has gone full then a message is
displayed that insert queue is full. Note that we have used the word queue not heap in
that message. It has to that way otherwise the users of the PriorityQueue class are
unaware of the internal representation of the queue (whether it is implemented using a
heap or an array). Next line is returning 0indicating that the insert operation was not
successful. Below to insert, we have full() method. This method returns an int value.
Internally, it is calling isFull() method of heap. The full() method returns whatever is
returned by the isFull() method of heap. Next is length() method, as the size of heap is
also the size of the queue, therefore, length() is internally calling the getSize() method of
heap.
In this new implementation the code is better readable than the PriorityQueues
implementation with array. You must be remembering when we implemented the
PriorityQueue with an array, at each insertion in the
Page 393: [2] Deleted

Abutt

5/26/2003 8:59:00 PM

Clearly then this new implementation is more efficient because the heap can readjusts
itself in log2N times. Gain in performance is the major benefit of implementing
PriorityQueue using heap over implementing with array and very important for you to
remember as a student of Data Structures.
There are other significant benefits of using heap and there are number of other uses of
heap that would be covered in this course time to time. At the moment, we willsome
common example usages of heap in order to make you clear about other uses of heap.
The heap data structure will be covered in the Algorithms course also.

The Selection Problem

Given a list of N elements (numbers, names etc.), which can be totally ordered, and
an integer k, find the kth smallest (or largest) element.
Suppose we have list of N names (names of students or names of motor vehicles or a list
of numbers of motor vehicles or list of roll numbers of students or id card numbers of the
students, whatever). The problem is to findout the kth smallest element. So lets say we
have a list of 1000 numbers and we want to find the 10th smallest number in it. The
sorting is applied in order to make the elements ordered. After sorting out list of numbers
(in any order ascending or decending, lets do in ascending in this as we are finding the
smallest number), it will be very easy to findany desired smallest number.
th

One way is to put these N elements in an array and sort it. The k smallest of these is
th
at the k position.

It will take Nlog2N time in case we use array data structure. Now, we want to see if it is
possible to reduce the time from Nlog2N by using some other data structure or by
improving the algorithm? Yes, we can apply heap data structure in order to make this
operation more efficient.
A faster way is to put the N elements into an array and apply the buildHeap algorithm
on this array.
Finally, we perform k deleteMin operations. The last element extracted from the heap
is our answer.
The buildHeap algorithm is used to construct a heap of given N elements. If we
constructed min-heap, the minimum of the N elements will be positioned in the root
node of the heap. If we take out (deleteMin) k elements from the heap, we can get the kth
smallest element. BuildHeap works in linear time to make a min or a max-heap.
The interesting case is k = N/2, since this is known as the median.
In Statistics, we take the average of numbers, find the minimum, maximum and median
as well. Median is defined as a number in the sequence where the half of the numbers are
greater than this number and half of the numbers are smaller. Now, we can come up with
our mechanism to find median from a given N numbers. Lets say , we want to compute
the median of final marks of students of our class. Further suppose that the maximum
aggregate marks for a student are 100. We use the buildHeap to construct a heap for N
number of students. By calling deleteMin for N/2 times, the minimum marks of the half
number students will be taken out. The N/2th marks would be the median of the marks of
our class. The alternates methods are there to calculate median, however, we are
discussing here about the possible uses of heap. Lets see another use of heap.

Heap Sort

As discussed above in The Selection Problem that to take the 100th minimum element out
from the min-heap, we will call deleteMin to take out 1st element, 2nd element, 3rd
element and eventually we will call deleteMin 100th time to take out our required 100th
minimum element. Suppose, if the size of the heap is 100 elements then we have taken
out all the elements out from the heap. Interestingly the elements are sorted in increasing
order. If somehow, we can store these taken out numbers, lets say in an array we can
have all the elements sorted (in ascending order in this min-heap case).
Hence,
If k = N, and we record the deleteMin elements as they come off the heap, we will
have essentially sorted the N elements.
Later in the course, we will refine this idea to obtain a fast sorting algorithm called
heapsort.
We finish our discussion above heap here but it will be discussed in the forthcoming
courses and a little bit when we will discuss about sort in the next lectures of this course.
At the moment, let see another Abstract Data Type.

Disjoint Set ADT

As always, before we actually move to an Abstract Data Type (ADT), we like to see what
that ADT is, how it works and in what situations it can be helpful. We are going to cover
Disjoint Set ADT. Firstly, we will have its introduction, examples and later on, we will
discuss about its implementation.
Suppose we have a database of people.
We want to figure out who is related to whom.Initially, we only have a list of people, and
information about relations is gained by updates of the form
Haaris is related to Saad.Lets say we have a list of names of all
people in a locality but we are not aware of their relationships to each other, if they are
related. After having the list of all people we start getting some information about their
relationships slowly, one relationship at a time, for example, Ali Abbas isson of Abbas.
The situation becomes interesting when we have relationships like Ali Abbas is first
cousin of Ayesha Ali (i.e. fathers of both are brothers) but Ayesha Ali has other
immediate cousins also from her mothers side, therefore, other immediate cousins of
Ayesha Ali also get related to Ali Abbas, although they are not immediate to Ali Abbas.
So as we keep on getting relationships details of those people, these direct and indirect
relationships can be establishded.
Key property: If Haaris is related to Saad and Saad is related to Ahmad, then Haaris is
related to Ahmad.
See the key property lines first part above Harris is related to Saad and Saad is related
to Ahmad. Suppose we have a program to handle this list of people and their
relationships. After we have provided all the name Harris, Saad and Ahmad to that
program and then this part of information about their relationships, the program might be
able to determine the remaining part Harris related to Ahmad.
The same problem (the intelligence required in the program) is describe in the sentence
below:
Once we have relationships information, we would like to answer queries like Is Haaris
related to Ahmad?
To answer this kind of queries and to have that intelligence in our programs Disjoint Set
ADT is used. Before we see what Disjoint Set ADT is, we see another application of this
ADT in image analysis. This problem is known as Blob Coloring.
Blob Coloring
A well-known low-level computer vision problem for black and white images is the
following:Gather together all the picture elements (pixels) that belong to the same
"blobs", and give each pixel in each different blob an identical label.
You must have heard of robots that perform certain tasks automatically. How do they
know from the images provided to them that what in which direction should they move,
they can catch things and carry them. They do different things like humans. Obviously,

there is a software working internal to robots body that is doing all this controlling part
and vision to the robot. This is very complex problem and broad area of Computer
Science and Electrical Engineering called Robotics (Computer Vision in particular).
Consider the image below:

``
Fig 33.1
This image is black and white, which consist of five non-overlapping black colored
regions of different shapes called blobs. These blobs are two elliptic, n and u shaped (two
blobs) and one arc at the bottom. From human eye we can see these five blobs but how a
robot can identify these. So the problem is:
We want to partition the pixels into disjoint sets, one set per blob.
Now if we make one set per blob then for five blobs in the image above, we have to have
five sets. To understand the concept of disjoint sets, we can take an analogy where we
have two sets A and B (as in Mathematics) where none of the elements inside set A is
present in set B. The sets A and B are called disjoint sets.
Another problem relatedthe Computer Vision is the image segmentation problem. See the
image below on the left of an old ship in gray scales.

Fig 33.2
We want to find regions of different colors in this picture, for example all regions in the

picture of color black, all regions in the image of color gray. The image on the right
represents the resultant image.
Different scanning done in hospitals mri scan, cat scan or ct scan scans the inner parts of
the whole body of humans. Those scans images in gray scales represent organs of the
human body. All these are application of Disjoint Set ADT.