Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Intro to DAX Patterns

•Download as PPTX, PDF•
•
E
Eric Bragas

An introductory session to DAX and common analytic patterns that we've built and used in enterprise environments. This session was originally presented at SQL Saturday Silicon Valley 2016.

1 of 30
Intro to DAX Patterns
 Eric Bragas – MCP, PSM I  DesignMind - Business Intelligence Consultant
Agenda  DAX  Data Analysis Expressions  Notation  Functions  Evaluation Contexts  Measure Patterns  Calculated Column Patterns
Data Analysis Expressions (DAX)  What is DAX?  DAX is a language that allows us to write dynamic expressions for relataional constructs, using familiar functions  Powerful dynamic data analysis tool for relational data  Expressions can traverse relationships!  Available in PowerPivot, PowerBI, and SSAS Tabular  Classic “Import Mode”  What isn’t DAX?  NOT a programming language
DAX Notation Table ‘Table’ Column ‘Table’[Column] Function fn ( <arguments> ) Measure Measure Name := SUM ( ‘Table’[Column] ) Calculated Column Column Name = SUM ( ‘Table’[Column] )
Measures  A measure is a formula/expression comprising functions applied to columns and tables  Reusable aggregation evaluated differently, depending on how you use it  Measures can be nested
Functions  Logical  IF( logical test, <value if true>, <value if false> )  SWITCH( <expression>, <value>, <result>, … ) – evaluates an expression against a list of values, and returns the result corresponding to the first matching value  TRUE() – returns logical true  Aggregate  SUM( <column> ) – adds all the numbers in a column  DIVIDE( <numerator>, <denominator>, [, <alternateresult>] ) – basic division; optional value returned  Statistical  MAX( <column> ) – returns the largest numeric value in a <column>  MIN( <column> ) – returns the smallest value in a <column>  Text  BLANK() – returns a blank  Filter  FILTER(<table>, <filter>) – returns a table representing a subset of another table or expression  ALL( <table> | <column>) – returns all the rows in a table, ignoring any filter context  VALUES(<table or column>) – returns one column of the distinct values from the specified column or table  CALCULATE( <expression>, <filter1>, <filter2>, … ) – evaluates an expression in a context that is modified by the specified filters
Functions (cont’d)  Date and Time  TODAY() – returns the current date  NOW() – returns the current date and time in datetime format  Time Intelligence*  DATESBETWEEN(<dates>, <start date>, <end date>) – returns a table of <dates> starting with the <start date> and continues until the <end date>.  NEXTDAY(<dates>) – returns a table that contains a column with the next dates following each of the <dates> passed  FIRSTDATE(<dates>) – returns the first date in the context of the specified column of dates  LASTDATE(<dates>) – returns the last date in the context of the specified column of dates  SAMEPERIODLASTYEAR(<dates>) – returns a table with a column of dates shifted one year back for each of the <dates> specified  LASTNONBLANK( <column>, <expression> ) – returns last value in the <column> where the <expression> returns blank  FIRSTNONBLANK( <column>, <expression> ) – returns the first value in the <column> where the <expression> returns blank *Requires Date Table
Evaluation Contexts  Evaluation Contexts:  Filter Context  Four types of filter context: 1. Row Selection 2. Column Selection 3. Slicer Selection 4. Filter Selection  Defines the subset of data a measure is calculated using aka “Which rows are selected based on which attribute values?”  Applied before anything else  Row Context  All the columns in the Current Row “DAX is simple, it’s not easy, but it’s simple” - Alberto Ferrari
Basic Evaluation Context - Demo TotalSales = SUM ( Sales[SalesAmount] )
Anatomy of a Filter Context - PivotTable Rows Slicer
Anatomy of a Filter Context - Chart Filter Slicer
Measure Patterns BASIC CUMULATIVE YTD YEAR OVER YEAR SEMI-ADDITIVE DISCONNECTED SLICERS
Basic Measures  SUM( ‘Sales’[SalesAmount] )  AVERAGE( ‘Inventory’[InventoryOnHand] )  MIN( ‘Weather’[Temp] )  MAX ( ‘Weather’[Temp] )  Etc.
Basic Measure Demo – PowerPivot & PowerBI Asia Sales := CALCULATE ( [Total Sales], Geography[ContinentName] = "Asia" ) Asia Sales := CALCULATE ( [Total Sales], FILTER ( 'Geography', Geography[ContinentName] = "Asia" ) ) Boolean Table
Cumulative Total Measures  Aggregates values of a column for the currently selected date and all previous dates within the specified range  Can be used to derive balances from transactions eg.  Inventory Stock  Balances  Cumulative Balances  Does not require use of Time Intelligence functions
Cumulative Measure Demo - PowerBI Cumulative Energy Generated (Checked) = IF ( MIN ( 'Date'[Date] ) <= MAX ( ‘Output’[Date], ALL ( ‘Output’ ) ) , CALCULATE ( SUM ( 'Output'[Energy Generated] ), FILTER ( ALL ( 'Date'[Date] ), 'Date'[Date] <= MAX ( 'Date'[Date] ) ) ) )
Year-to-Date Total  TOTALYTD function applies the expression for all data from the start of the year to the currently selected date in the filter context Year To Date = TOTALYTD( <expression>, <dates> [, <filter>] [, <year end date>] )
Year-to-Date Total Demo - PowerBI Total Energy Generated YTD = TOTALYTD ( SUM ( 'Output'[Energy Generated] ), 'Date'[Date] )
Year Over Year  Use time intelligence to calculate an aggregate for the same period last year  “Last Year” measure is used to compare to “Current Year” measure, and/or to derive a measure of the change year-over-year
Year-Over-Year Demo – PowerBI Total Energy Generated Last Year = CALCULATE ( [Total Energy Generated], SAMEPERIODLASTYEAR ( 'Date'[Date] ) )
Semi-Additive Measures  Snapshot Fact Table with balance values, such as Inventory or Account Balances  These scenarios disallow us from summing across time  The solution is to sum across all attributes except for time by filtering for only a single point in time (eg. last date in the period)  Several functions allow you to adjust filter context to a single point in time, within the original context period  FIRSTDATE / LASTDATE  FIRSTNONBLANK / LASTNONBLANK  OPENING… / CLOSING…
Semi-Additive Measure Demo - PowerPivot Total On Hand Quantity LASTNONBLANK = CALCULATE ( SUM ( Inventory[OnHandQuantity] ), LASTNONBLANK ( 'Date'[Date], CALCULATE ( SUM ( Inventory[OnHandQuantity] ) ) ) )
Disconnected Slicers Allows you to use a slicer to modify measures  Measure Switching  Used to switch between a set of measure values in a container measure
Disconnected Slicers Demo - PowerBI Setup Steps: 1. Create/identify your target measures (eg. [Energy Exported] & [Energy Generated]) 2. Create disconnected table to use in slicer selection 3. Create background “value selection measure” using MAX() • Hide this! 4. Create SWITCH measure to use in visualizations
Calculated Column Patterns VALUE BINNING
Value Binning  Used to group similar values, or to bin values for analysis aka Histograms  Can bin values based on equality, or inequality comparisons with the SWITCH() function  Use Cases:  Age groups  Product Groups  Any kind of frequency distributions
Value Binning Demo - PowerBI Inventory Age (bin) = SWITCH ( TRUE (), 'Inventory'[DaysInStock] < 20, "0-20", 'Inventory'[DaysInStock] < 50, "21-50", 'Inventory'[DaysInStock] < 80, "51-80", 'Inventory'[DaysInStock] < 100, "81-100", "101+" )
Summary  DAX is dynamic because you can write measures that correctly evaluate under their current Evaluation Context  Filter Context  Row Context  Functions are the building blocks of our measures and perform a myriad of tasks eg. altering Context, aggregating, logical operations, time intelligence, etc  Time Intelligence functions require a Date table to operate  Understanding Tabular Data Modeling will go a long way towards helping your understanding of DAX
Thanks!

More Related Content

Intro to DAX Patterns

  • 1. Intro to DAX Patterns
  • 2.  Eric Bragas – MCP, PSM I  DesignMind - Business Intelligence Consultant
  • 3. Agenda  DAX  Data Analysis Expressions  Notation  Functions  Evaluation Contexts  Measure Patterns  Calculated Column Patterns
  • 4. Data Analysis Expressions (DAX)  What is DAX?  DAX is a language that allows us to write dynamic expressions for relataional constructs, using familiar functions  Powerful dynamic data analysis tool for relational data  Expressions can traverse relationships!  Available in PowerPivot, PowerBI, and SSAS Tabular  Classic “Import Mode”  What isn’t DAX?  NOT a programming language
  • 5. DAX Notation Table ‘Table’ Column ‘Table’[Column] Function fn ( <arguments> ) Measure Measure Name := SUM ( ‘Table’[Column] ) Calculated Column Column Name = SUM ( ‘Table’[Column] )
  • 6. Measures  A measure is a formula/expression comprising functions applied to columns and tables  Reusable aggregation evaluated differently, depending on how you use it  Measures can be nested
  • 7. Functions  Logical  IF( logical test, <value if true>, <value if false> )  SWITCH( <expression>, <value>, <result>, … ) – evaluates an expression against a list of values, and returns the result corresponding to the first matching value  TRUE() – returns logical true  Aggregate  SUM( <column> ) – adds all the numbers in a column  DIVIDE( <numerator>, <denominator>, [, <alternateresult>] ) – basic division; optional value returned  Statistical  MAX( <column> ) – returns the largest numeric value in a <column>  MIN( <column> ) – returns the smallest value in a <column>  Text  BLANK() – returns a blank  Filter  FILTER(<table>, <filter>) – returns a table representing a subset of another table or expression  ALL( <table> | <column>) – returns all the rows in a table, ignoring any filter context  VALUES(<table or column>) – returns one column of the distinct values from the specified column or table  CALCULATE( <expression>, <filter1>, <filter2>, … ) – evaluates an expression in a context that is modified by the specified filters
  • 8. Functions (cont’d)  Date and Time  TODAY() – returns the current date  NOW() – returns the current date and time in datetime format  Time Intelligence*  DATESBETWEEN(<dates>, <start date>, <end date>) – returns a table of <dates> starting with the <start date> and continues until the <end date>.  NEXTDAY(<dates>) – returns a table that contains a column with the next dates following each of the <dates> passed  FIRSTDATE(<dates>) – returns the first date in the context of the specified column of dates  LASTDATE(<dates>) – returns the last date in the context of the specified column of dates  SAMEPERIODLASTYEAR(<dates>) – returns a table with a column of dates shifted one year back for each of the <dates> specified  LASTNONBLANK( <column>, <expression> ) – returns last value in the <column> where the <expression> returns blank  FIRSTNONBLANK( <column>, <expression> ) – returns the first value in the <column> where the <expression> returns blank *Requires Date Table
  • 9. Evaluation Contexts  Evaluation Contexts:  Filter Context  Four types of filter context: 1. Row Selection 2. Column Selection 3. Slicer Selection 4. Filter Selection  Defines the subset of data a measure is calculated using aka “Which rows are selected based on which attribute values?”  Applied before anything else  Row Context  All the columns in the Current Row “DAX is simple, it’s not easy, but it’s simple” - Alberto Ferrari
  • 10. Basic Evaluation Context - Demo TotalSales = SUM ( Sales[SalesAmount] )
  • 11. Anatomy of a Filter Context - PivotTable Rows Slicer
  • 12. Anatomy of a Filter Context - Chart Filter Slicer
  • 13. Measure Patterns BASIC CUMULATIVE YTD YEAR OVER YEAR SEMI-ADDITIVE DISCONNECTED SLICERS
  • 14. Basic Measures  SUM( ‘Sales’[SalesAmount] )  AVERAGE( ‘Inventory’[InventoryOnHand] )  MIN( ‘Weather’[Temp] )  MAX ( ‘Weather’[Temp] )  Etc.
  • 15. Basic Measure Demo – PowerPivot & PowerBI Asia Sales := CALCULATE ( [Total Sales], Geography[ContinentName] = "Asia" ) Asia Sales := CALCULATE ( [Total Sales], FILTER ( 'Geography', Geography[ContinentName] = "Asia" ) ) Boolean Table
  • 16. Cumulative Total Measures  Aggregates values of a column for the currently selected date and all previous dates within the specified range  Can be used to derive balances from transactions eg.  Inventory Stock  Balances  Cumulative Balances  Does not require use of Time Intelligence functions
  • 17. Cumulative Measure Demo - PowerBI Cumulative Energy Generated (Checked) = IF ( MIN ( 'Date'[Date] ) <= MAX ( ‘Output’[Date], ALL ( ‘Output’ ) ) , CALCULATE ( SUM ( 'Output'[Energy Generated] ), FILTER ( ALL ( 'Date'[Date] ), 'Date'[Date] <= MAX ( 'Date'[Date] ) ) ) )
  • 18. Year-to-Date Total  TOTALYTD function applies the expression for all data from the start of the year to the currently selected date in the filter context Year To Date = TOTALYTD( <expression>, <dates> [, <filter>] [, <year end date>] )
  • 19. Year-to-Date Total Demo - PowerBI Total Energy Generated YTD = TOTALYTD ( SUM ( 'Output'[Energy Generated] ), 'Date'[Date] )
  • 20. Year Over Year  Use time intelligence to calculate an aggregate for the same period last year  “Last Year” measure is used to compare to “Current Year” measure, and/or to derive a measure of the change year-over-year
  • 21. Year-Over-Year Demo – PowerBI Total Energy Generated Last Year = CALCULATE ( [Total Energy Generated], SAMEPERIODLASTYEAR ( 'Date'[Date] ) )
  • 22. Semi-Additive Measures  Snapshot Fact Table with balance values, such as Inventory or Account Balances  These scenarios disallow us from summing across time  The solution is to sum across all attributes except for time by filtering for only a single point in time (eg. last date in the period)  Several functions allow you to adjust filter context to a single point in time, within the original context period  FIRSTDATE / LASTDATE  FIRSTNONBLANK / LASTNONBLANK  OPENING… / CLOSING…
  • 23. Semi-Additive Measure Demo - PowerPivot Total On Hand Quantity LASTNONBLANK = CALCULATE ( SUM ( Inventory[OnHandQuantity] ), LASTNONBLANK ( 'Date'[Date], CALCULATE ( SUM ( Inventory[OnHandQuantity] ) ) ) )
  • 24. Disconnected Slicers Allows you to use a slicer to modify measures  Measure Switching  Used to switch between a set of measure values in a container measure
  • 25. Disconnected Slicers Demo - PowerBI Setup Steps: 1. Create/identify your target measures (eg. [Energy Exported] & [Energy Generated]) 2. Create disconnected table to use in slicer selection 3. Create background “value selection measure” using MAX() • Hide this! 4. Create SWITCH measure to use in visualizations
  • 27. Value Binning  Used to group similar values, or to bin values for analysis aka Histograms  Can bin values based on equality, or inequality comparisons with the SWITCH() function  Use Cases:  Age groups  Product Groups  Any kind of frequency distributions
  • 28. Value Binning Demo - PowerBI Inventory Age (bin) = SWITCH ( TRUE (), 'Inventory'[DaysInStock] < 20, "0-20", 'Inventory'[DaysInStock] < 50, "21-50", 'Inventory'[DaysInStock] < 80, "51-80", 'Inventory'[DaysInStock] < 100, "81-100", "101+" )
  • 29. Summary  DAX is dynamic because you can write measures that correctly evaluate under their current Evaluation Context  Filter Context  Row Context  Functions are the building blocks of our measures and perform a myriad of tasks eg. altering Context, aggregating, logical operations, time intelligence, etc  Time Intelligence functions require a Date table to operate  Understanding Tabular Data Modeling will go a long way towards helping your understanding of DAX

Editor's Notes

  1. How many people have worked with Excel formulas? How many people have worked with PowerPivot?
  2. Story: Introduced to data by my Dad (scary DBA types) Began learning Relational Databases and attending SQL Saturday’s Indexing internals – Kalen Delaney T-SQL – Kevin Boles Merge Operators – Ami Levin Developer/DBA Really interested in BI (DW) Realized my data analysis tool belt was lacking So I decided to present on DAX
  3. Introduction to DAX Introduction to Measure and Analysis concepts Introduction to Evaluation Context Introduction to Measure and Calculated Column Patterns (which happen to often alter Evaluation Context)
  4. “DAX is a language that allows us to write dynamic expressions for relational constructs, using familiar functions” What is DAX? though it shares functions with Excel formula language, it differs by being intended for analysis of relational data
  5. There are some minor differences between PowerBI and PowerPivot eg. the colon
  6. Review CALCULATE in detail
  7. We need to understand Evaluation Context so that we can begin altering it. Filter Context Row Context Iterator Functions… not going to cover these much
  8. Point out the parts of the measure syntax What sales does the measure sum? Sum of all sales Why do the numbers change when we add color to the rows? The measure still evaluates sum of all sales However now, it is operating under the context of a color, and can only sum sales for that color! The value of a formula depends on it’s context. What we put in a context filters the subset of data we can measure, and so we call it the Filter Context!
  9. Diagram View: Already setup data model ( imported tables ) Created relationships between tables; DAX can traverse these without an explicit join Data Grid: Created simple SUM Shows SUM of all Sales Add Continent to Rows; filter context changes Add ‘Product’[Product Class] to Columns; filter context changes Diagram View: Create SUM that adds Asia sales | CALCULATE ( SUM ( ‘Sales’[SalesAmount] ) [Total Sales], ‘Geography’[Continent] = “Asia” ) Show results Create SUM that only adds Asia sales; returns blank all others | CALCULATE ( [Total Sales], FILTER ( ‘Geography’, ‘Geography’[Continent] = “Asia” ) )
  10. Demo: show basic measures in PowerPivot then PowerBI Demo: CALCULATE with a simple argument against Geography
  11. To avoid calculating values for dates greater than the max date in the transactions table (‘Output’), add a check that the minimum ‘Date’[DateKey] <= maximum ‘Transactions’[Date] Can also replace the MAX(‘Transactions’[Date]) check with TODAY(), however this is not always valid as it assumes all data is “current”
  12. Paste to exclude future dates: IF ( MIN ( 'Date'[Date] ) <= CALCULATE ( MAX ( 'Output'[Date] ), ALL ( 'Output' ) ),
  13. TOTALYTD function applies the expression for all data from the start of the year to the currently selected date in the filter context
  14. TOTALYTD function applies the expression for all data from the start of the year to the currently selected date in the filter context
  15. Paste to exclude future dates: IF ( MIN ( 'Date'[Date] ) <= CALCULATE ( MAX ( 'Output'[Date] ), ALL ( 'Output' ) ), DIVIDE ( [Sales] – [Last Year Sales], [Last Year Sales] )
  16. Paste to exclude future dates: IF ( MIN ( 'Date'[Date] ) <= CALCULATE ( MAX ( 'Output'[Date] ), ALL ( 'Output' ) ),
  17. How can we alter the filter context to only a single point in time? Using CALCULATE to override the context First and Last Date will not return values - if the date a period ends on has no data * Time intelligence functions in PowerBI require the DateKey to be a Date data type * The expression argument in LASTNONBLANK must be wrapped in a CALCULATE, otherwise it will use the original filter context, and not the LASTNONBLANK column argument context
  18. Setup Target Measures ( [Energy Export] & [Energy Generated] ) User defined table ( disconnected ) Value selection measure ( MAX ( ‘disconnected_table’[id] ) ) Switching measure
  19. Demo: Inventory Aging
  20. Demo: Inventory Aging In this case, I am using the Inventory Semi-Additive Measure as well