Horje

MS SQL Server - Microsoft SQL Server - PIVOT / UNPIVOT
Start
  Simple PIVOT & UNPIVOT (T-SQL)

Below is a simple example which shows average item's price of each item per weekday.

First, suppose we have a table which keeps daily records of all items' prices.

CREATE TABLE tbl_stock(item NVARCHAR(10), weekday NVARCHAR(10), price INT);

INSERT INTO tbl_stock VALUES 
('Item1', 'Mon', 110), ('Item2', 'Mon', 230), ('Item3', 'Mon', 150), 
('Item1', 'Tue', 115), ('Item2', 'Tue', 231), ('Item3', 'Tue', 162), 
('Item1', 'Wed', 110), ('Item2', 'Wed', 240), ('Item3', 'Wed', 162), 
('Item1', 'Thu', 109), ('Item2', 'Thu', 228), ('Item3', 'Thu', 145), 
('Item1', 'Fri', 120), ('Item2', 'Fri', 210), ('Item3', 'Fri', 125),
('Item1', 'Mon', 122), ('Item2', 'Mon', 225), ('Item3', 'Mon', 140),
('Item1', 'Tue', 110), ('Item2', 'Tue', 235), ('Item3', 'Tue', 154),
('Item1', 'Wed', 125), ('Item2', 'Wed', 220), ('Item3', 'Wed', 142);

The table should look like below:

+========+=========+=======+
|   item | weekday | price |
+========+=========+=======+
|  Item1 |    Mon  |   110 |
+--------+---------+-------+
|  Item2 |    Mon  |   230 |
+--------+---------+-------+
|  Item3 |    Mon  |   150 |
+--------+---------+-------+
|  Item1 |    Tue  |   115 |
+--------+---------+-------+
|  Item2 |    Tue  |   231 |
+--------+---------+-------+
|  Item3 |    Tue  |   162 |
+--------+---------+-------+
|          . . .           |
+--------+---------+-------+
|  Item2 |    Wed  |   220 |
+--------+---------+-------+
|  Item3 |    Wed  |   142 |
+--------+---------+-------+

In order to perform aggregation which is to find the average price per item for each week day, we are going to use the relational operator PIVOT to rotate the column weekday of table-valued expression into aggregated row values as below:

SELECT * FROM tbl_stock
PIVOT ( 
    AVG(price) FOR weekday IN ([Mon], [Tue], [Wed], [Thu], [Fri])
) pvt;

Result:

+--------+------+------+------+------+------+
|  item  |  Mon |  Tue |  Wed |  Thu |  Fri |
+--------+------+------+------+------+------+
|  Item1 |  116 |  112 |  117 |  109 |  120 |
|  Item2 |  227 |  233 |  230 |  228 |  210 |
|  Item3 |  145 |  158 |  152 |  145 |  125 |
+--------+------+------+------+------+------+

Lastly, in order to perform the reverse operation of PIVOT, we can use the relational operator UNPIVOT to rotate columns into rows as below:

SELECT * FROM tbl_stock
PIVOT ( 
    AVG(price) FOR weekday IN ([Mon], [Tue], [Wed], [Thu], [Fri])
) pvt
UNPIVOT ( 
    price FOR weekday IN ([Mon], [Tue], [Wed], [Thu], [Fri])
) unpvt;

Result:

+=======+========+=========+
|  item |  price | weekday |
+=======+========+=========+
| Item1 |    116 |     Mon |
+-------+--------+---------+
| Item1 |    112 |     Tue |
+-------+--------+---------+
| Item1 |    117 |     Wed |
+-------+--------+---------+
| Item1 |    109 |     Thu |
+-------+--------+---------+
| Item1 |    120 |     Fri |
+-------+--------+---------+
| Item2 |    227 |     Mon |
+-------+--------+---------+
| Item2 |    233 |     Tue |
+-------+--------+---------+
| Item2 |    230 |     Wed |
+-------+--------+---------+
| Item2 |    228 |     Thu |
+-------+--------+---------+
| Item2 |    210 |     Fri |
+-------+--------+---------+
| Item3 |    145 |     Mon |
+-------+--------+---------+
| Item3 |    158 |     Tue |
+-------+--------+---------+
| Item3 |    152 |     Wed |
+-------+--------+---------+
| Item3 |    145 |     Thu |
+-------+--------+---------+
| Item3 |    125 |     Fri |
+-------+--------+---------+
  Dynamic PIVOT

One problem with the PIVOT query is that you have to specify all values inside the IN selection if you want to see them as columns. A quick way to circumvent this problem is to create a dynamic IN selection making your PIVOT dynamic.

For demonstration we will use a table Books in a Bookstore’s database. We assume that the table is quite de-normalised and has following columns

Table: Books
-----------------------------
BookId (Primary Key Column)
Name
Language
NumberOfPages
EditionNumber
YearOfPrint
YearBoughtIntoStore
ISBN
AuthorName
Price
NumberOfUnitsSold

Creation script for the table will be like:

CREATE TABLE [dbo].[BookList](
      [BookId] [int] NOT NULL,
      [Name] [nvarchar](100) NULL,
      [Language] [nvarchar](100) NULL,
      [NumberOfPages] [int] NULL,
      [EditionNumber] [nvarchar](10) NULL,
      [YearOfPrint] [int] NULL,
      [YearBoughtIntoStore] [int] NULL,
[NumberOfBooks] [int] NULL,
[ISBN] [nvarchar](30) NULL,
      [AuthorName] [nvarchar](200) NULL,
      [Price] [money] NULL,
      [NumberOfUnitsSold] [int] NULL,
 CONSTRAINT [PK_BookList] PRIMARY KEY CLUSTERED
(
      [BookId] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY]

GO

Now if we need to query on the database and figure out number of books in English, Russian, German, Hindi, Latin languages bought into the bookstore every year and present our output in a small report format, we can use PIVOT query like this

SELECT * FROM
  (
   SELECT YearBoughtIntoStore AS [Year Bought],[Language], NumberOfBooks
   FROM BookList
  ) sourceData 
 PIVOT
  (
  SUM(NumberOfBooks)
  FOR [Language] IN (English, Russian, German, Hindi, Latin)
  ) pivotrReport

Special case is when we do not have a full list of the languages, so we'll use dynamic SQL like below

DECLARE @query VARCHAR(4000)
DECLARE @languages VARCHAR(2000)
SELECT @languages =
        STUFF((SELECT DISTINCT '],['+LTRIM([Language])FROM [dbo].[BookList]
        ORDER BY '],['+LTRIM([Language]) FOR XML PATH('') ),1,2,'') + ']'
SET @query=
'SELECT * FROM
  (SELECT YearBoughtIntoStore AS [Year Bought],[Language],NumberOfBooks
   FROM BookList) sourceData
PIVOT(SUM(NumberOfBooks)FOR [Language] IN ('+ @languages +')) pivotrReport' EXECUTE(@query)
  Simple Pivot - Static Columns

Using Item Sales Table (opens new window) from Example Database (opens new window), let us calculate and show the total Quantity we sold of each Product.

This can be easily done with a group by, but lets assume we to 'rotate' our result table in a way that for each Product Id we have a column.

SELECT [100], [145]
  FROM (SELECT ItemId , Quantity
          FROM #ItemSalesTable
       ) AS pivotIntermediate
 PIVOT (   SUM(Quantity)
           FOR ItemId IN ([100], [145])
       ) AS pivotTable

Since our 'new' columns are numbers (in the source table), we need to square brackets []

This will give us an output like

100 145
45 18

Syntax

  • SELECT ``,
    [first pivoted column] AS ``,
    [second pivoted column] AS ``,
    ...
    [last pivoted column] AS ``
    FROM
    (`

    • Remarks

    Using PIVOT and UNPIVOT operators you transform a table by shifting the rows (column values) of a table to columns and vise-versa. As part of this transformation aggregation functions can be applied on the table values.

    		 

    Read Article
    https://develop.horje.com/learn/1434/reference

     

    Horje