T-SQL Recipes之Dynamic PIVOT and UNPIVOT

Posted 2020-07-19

PIVOT

PIVOT在行转列的时候经常用到，最便捷的方式就是通过示例来理解它的作用。

示例1 Query to Return Select Product Data from AdventureWorks

SELECT  PRODUCT.Name AS product_name ,
        PRODUCT.Color AS product_color ,
        PRODUCT_INVENTORY.LocationID ,
        PRODUCT.ReorderPoint ,
        PRODUCT_INVENTORY.Quantity AS product_quantity
FROM    Production.Product PRODUCT
        LEFT JOIN Production.ProductInventory PRODUCT_INVENTORY ON PRODUCT.ProductID = PRODUCT_INVENTORY.ProductID;

View Code

结果：

  

如果我们想要product_coor 在列里面显示每个产品的数量呢？这时候PIVOT就出场了

示例2：Common Use of PIVOT to Report on Products by Color

WITH    PRODUCT_DATA
          AS ( SELECT   PRODUCT.Name AS product_name ,
                        PRODUCT.Color AS product_color ,
                        PRODUCT.ReorderPoint ,
                        PRODUCT_INVENTORY.Quantity AS product_quantity
               FROM     Production.Product PRODUCT
                        LEFT JOIN Production.ProductInventory PRODUCT_INVENTORY ON PRODUCT.ProductID = PRODUCT_INVENTORY.ProductID
             )
    SELECT  *
    FROM    PRODUCT_DATA PIVOT
( SUM(product_quantity) FOR product_color IN ( [Black], [Blue], [Grey],
                                               [Multi], [Red], [Silver],
                                               [Silver/Black], [White],
                                               [Yellow] ) ) PIVOT_DATA;

View Code

结果：

从SQL中可以看出PIVOT有两个步骤

• An aggregate function, which will aggregate if multiple values exist. In the initial
  SELECT statement that returns product data, there were many duplicate rows. This
  example uses SUM whenever this occurs, which will add up product quantities if there
  are multiple rows with the same product name.

• A value list for all values that will be changed from row data into column headers. In
  this case, the list is of colors from Product.Color .

PS:虽然解决了行转列的问题，但这个时候，我们应该知道color里面到底有多少条唯一的数据，如果在我们不知的情况下，如何解决呢？这个时候动态SQL就来了。

示例3：Common Use of PIVOT to Report on Products by Color

USE AdventureWorks2014;
GO

DECLARE @sql_command NVARCHAR(MAX);
DECLARE @sql_colors NVARCHAR(1000);

SET @sql_command = ‘
WITH    PRODUCT_DATA
          AS ( SELECT   PRODUCT.Name AS product_name ,
                        PRODUCT.Color AS product_color ,
                        PRODUCT.ReorderPoint ,
                        PRODUCT_INVENTORY.Quantity AS product_quantity
               FROM     Production.Product PRODUCT
                        LEFT JOIN Production.ProductInventory PRODUCT_INVENTORY ON PRODUCT.ProductID = PRODUCT_INVENTORY.ProductID
             )
    SELECT  *
    FROM    PRODUCT_DATA PIVOT
( SUM(product_quantity) FOR product_color IN (‘;
WITH    colorlist
          AS ( SELECT DISTINCT
                        Product.Color AS color_name
               FROM     Production.Product
               WHERE    Product.Color IS NOT NULL
             )
    SELECT  @sql_colors = ISNULL(@sql_colors, N‘‘) + N‘,‘
            + QUOTENAME(color_name)
    FROM    colorlist;

SET @sql_colors = STUFF(@sql_colors, 1, 1, ‘‘);

SET @sql_command = @sql_command + @sql_colors + N‘ )) PIVOT_DATA‘;

PRINT @sql_command;
EXEC sp_executesql @sql_command;

View Code

首先看一下打印出来的SQL：

WITH    PRODUCT_DATA
          AS ( SELECT   PRODUCT.Name AS product_name ,
                        PRODUCT.Color AS product_color ,
                        PRODUCT.ReorderPoint ,
                        PRODUCT_INVENTORY.Quantity AS product_quantity
               FROM     Production.Product PRODUCT
                        LEFT JOIN Production.ProductInventory PRODUCT_INVENTORY ON PRODUCT.ProductID = PRODUCT_INVENTORY.ProductID
             )
    SELECT  *
    FROM    PRODUCT_DATA PIVOT
( SUM(product_quantity) FOR product_color IN ( [Black], [Blue], [Grey],
                                               [Multi], [Red], [Silver],
                                               [Silver/Black], [White],
                                               [Yellow] ) ) PIVOT_DATA

View Code

结果：

从示例1到示例3，我们从中了解到如何把复杂的SQL慢慢分解出来，最后在组合在一起。

UNPIVOT

 

ds

ds