Multidimensional reporting with CROSS APPLY and PIVOT in MS SQL Server
In this post we’re going to demonstrate how to use PIVOT relational operator to transform data from table-valued into another table. As an example we will use simple Data Warehouse (DWH) that stores annual company reports for harvesting fruits. The goal is to display report showing annual reports of sold fruits for each year.
DWH schema
Our simple database stores information about annual reports of sold fruits grouped by companies:
Following query:
SELECT NAME, APPLE, GRAPE, YEAR
FROM dwh.dbo.HARVESTING_FRUITS
INNER JOIN dwh.dbo.COMPANY ON HARVESTING_FRUITS.COMPANY_ID = COMPANY.ID
returns data:
Transform reports to separate rows - (CROSS) APPLY
We are going to use CROSS APPLY operator to populate the same operation for each record from left side - HARVESTING_FRUITS. In below query we want to return pair of values for fruits reports based on year. Since we have known amount of fruit types, we will return concatenated strings, like:
- APPLES + YEAR
- GRAPES + YEAR
as FRUIT_YEAR, and return value for given report as AMOUNT.
SELECT COMPANY.NAME, FRUITS_BY_YEAR.*
FROM dwh.dbo.HARVESTING_FRUITS FRUITS
INNER JOIN dwh.dbo.COMPANY ON FRUITS.COMPANY_ID = COMPANY.ID
CROSS APPLY (
VALUES
(CONCAT('APPLES - ', YEAR), APPLE),
(CONCAT('GRAPES - ', YEAR), GRAPE)
) FRUITS_BY_YEAR (FRUIT_YEAR, AMOUNT)
The output is following:
Transform rows to columns - PIVOT
Since we have reports separated by fruit name and year, we can turn values from FRUIT_YEAR column into multiple columns. Help comes with PIVOT operator. PIVOT syntax requires to use aggregate function so for AMOUNT we can use MAX function to just get value. Columns will be displayed based on given order:
- [APPLES - 2015]
- [APPLES - 2016]
- [APPLES - 2017]
- [GRAPES - 2015]
- [GRAPES - 2016]
- [GRAPES - 2017]
SELECT *
FROM (
SELECT COMPANY.NAME, FRUITS_BY_YEAR.*
FROM dwh.dbo.HARVESTING_FRUITS FRUITS
INNER JOIN dwh.dbo.COMPANY ON FRUITS.COMPANY_ID = COMPANY.ID
CROSS APPLY (
VALUES
(CONCAT('APPLES - ', YEAR), APPLE),
(CONCAT('GRAPES - ', YEAR), GRAPE)
) FRUITS_BY_YEAR (FRUIT_YEAR, AMOUNT)
) COLLECTED_FRUITS
PIVOT (
MAX(AMOUNT)
FOR FRUIT_YEAR IN (
[APPLES - 2015], [APPLES - 2016], [APPLES - 2017],
[GRAPES - 2015], [GRAPES - 2016], [GRAPES - 2017]
)
) COMBINED_FRUITS
The final report is:
Summary
MS SQL Server comes with very useful operators that simplifies working with DWHs. Although operators syntax are easy, CROSS APPLY and PIVOT could be used for complex transformations. Script for example database creation you can find here.