动态反透视和拆分列 SQL Server 2012

Question

我有一个具有 20 列的 MarketOutput 表

[Region] [LOB]  [GWP 2013]  [GWP 2014]  [LR 2013]   [LR 2014]
-------------------------------------------------------------
North  Workers  38902.50     37,972,404   89             82

我想动态地将列更改为行。 Region和LOB是固定列，[GWP 2013]、[GWP 2014]、[LR 2013]、[LR 2014]是动态列。

明年他们将是[GWP 2015]，[LR 2015]

我想取消透视该列并将 [GWP 2014] 拆分为两列 [GWP], [2014]。

输出应该是这样的

Region  [LOB]      [Metrics] [Year] [Value]
--------------------------------------------------
North   Workers    GWP       2013         38902.50 
North   Workers    GWP       2014    37,972,404
North   Workers    LR        2013            89
North   Workers    LR        2014            82

您能否建议如何做到这一点？

我是 SQL Server 中枢轴的新手

我还想每次使用动态列表将输出插入到一个新表中

Answer 1

您只需要在执行UNPIVOT 之后拆分列，如下所示：

WITH Unpivoted 
AS
(
    SELECT region, lob, columns, value
    FROM Regions
    UNPIVOT
    (
       columns
       FOR value IN([GWP 2013] , [GWP 2014] ,
                    [LR 2013]  , [LR 2014] ,
                    [GWP 2015], [LR 2015])
    ) AS u
) 
SELECT 
  region, 
  lob,
  columns,
  CAST(CASE WHEN value LIKE 'GWP%' THEN REPLACE(value,'GWP ', '')
       WHEN value LIKE 'LR%' THEN REPLACE(value,'LR ', '')
  END AS INT) AS Year,
  CASE WHEN value LIKE 'GWP%' THEN 'GWP'
       WHEN value LIKE 'LR%' THEN 'LR'
  END AS Metrics
FROM Unpivoted;

当然你应该动态地做它以避免手动列出列并动态地做它：

DECLARE @cols AS NVARCHAR(MAX);
DECLARE @query AS NVARCHAR(MAX);

select @cols = STUFF((SELECT distinct ',' +
                        QUOTENAME(column_name)
                      FROM information_schema.columns
                      WHERE table_name = 'Regions'
                        AND COLUMN_NAME <> 'Region' 
                        AND COLUMN_NAME <> 'LOB'
                      FOR XML PATH(''), TYPE
                     ).value('.', 'NVARCHAR(MAX)') 
                        , 1, 1, '');


SELECT @query = 'WITH Unpivoted 
                AS
                (
                    SELECT region, lob, columns, value
                    FROM Regions
                    UNPIVOT
                    (
                       columns
                       FOR value IN('+ @cols + ')
                    ) AS u
                ) 
                SELECT 
                  region, 
                  lob,
                  columns,
                  CAST(CASE WHEN value LIKE ''GWP%'' THEN REPLACE(value,''GWP '', '''')
                       WHEN value LIKE ''LR%'' THEN REPLACE(value,''LR '', '''')
                  END AS INT) AS Year,
                  CASE WHEN value LIKE ''GWP%'' THEN ''GWP''
                       WHEN value LIKE ''LR%'' THEN ''LR''
                  END AS Metrics
                FROM Unpivoted';


EXECUTE(@query);

这应该可以正常工作，假设：

• [GWP 2013] , [GWP 2014] , [LR 2013] , [LR 2014] , [GWP 2015], [LR 2015], ... etc 的所有列都采用相同的格式（GWP 或 LR，然后是空格，然后是年份，以及

• 所有的列都是相同的数据类型int或十进制，如果数据类型不一样你应该在执行unpivot之前将它们全部转换为一种数据类型，否则你会得到一个错误。

• SQL Fiddle Demo

这会给你：

| region |     lob |  columns | Year | Metrics |
|--------|---------|----------|------|---------|
|  North | Workers |  38902.5 | 2013 |     GWP |
|  North | Workers | 37972404 | 2014 |     GWP |
|  North | Workers |       70 | 2015 |     GWP |
|  North | Workers |       89 | 2013 |      LR |
|  North | Workers |       82 | 2014 |      LR |
|  North | Workers |       80 | 2015 |      LR |

---

更新：

我使用FOR XML PATH('') .. 将所有值列表连接到一个字符串中，这是 SQL Server 中的一种解决方法。 @cols 的值将是字符串：[GWP 2013], [GWP 2014], ...。

如果您的字段数据类型不同，您必须在执行UNPVOT 之前对所有将在锚查询中取消透视的列进行强制转换，如下所示：

SELECT @query = 'WITH Unpivoted 
                AS
                (
                    SELECT region, lob, columns, value
                    FROM 
                    (
                       SELECT 
                         region,
                         lob,
                         CAST([GWP 2013] AS DECIMAL(10,2)) AS [GWP 2013],
                         CAST([GWP 2014] AS DECIMAL(10,2)) AS [GWP 2014],
                         ... etc
                       FROM Regions
                    ) AS t
                    UNPIVOT
                    (
                       columns
                       FOR value IN('+ @cols + ')
                    ) AS u
                ) 
                SELECT 
                  region, 
                  lob,
                  columns,
                  CAST(CASE WHEN value LIKE ''GWP%'' THEN REPLACE(value,''GWP '', '''')
                       WHEN value LIKE ''LR%'' THEN REPLACE(value,''LR '', '''')
                  END AS INT) AS Year,
                  CASE WHEN value LIKE ''GWP%'' THEN ''GWP''
                       WHEN value LIKE ''LR%'' THEN ''LR''
                  END AS Metrics
                FROM Unpivoted';

如果您发现手动为所有列编写强制转换很困难，您可以动态生成并附加它，例如：

DECLARE @colsCasted AS NVARCHAR(MAX);

select @colsCasted = STUFF((SELECT distinct ',' +
                        'CAST(' + QUOTENAME(column_name) + 'AS DECIMAL(10,2)) AS ' + QUOTENAME(column_name)
                      FROM information_schema.columns
                      WHERE table_name = 'Regions'
                        AND COLUMN_NAME <> 'Region' 
                        AND COLUMN_NAME <> 'LOB'
                      FOR XML PATH(''), TYPE
                     ).value('.', 'NVARCHAR(MAX)') 
                        , 1, 1, '');

然后在动态查询中将该值附加到它：

SELECT @query = 'WITH Unpivoted 
                    AS
                    (
                        SELECT region, lob, columns, value
                        FROM 
                        (
                          SELECT region, lob,
                          ' + @colsCasted + '
                          FROM Regions
                        ) AS t
                        UNPIVOT
                        (
                           columns
                           FOR value IN('+ @cols + ')
                        ) AS u
                    ) 
                    SELECT 
                      region, 
                      lob,
                      columns,
                      CAST(CASE WHEN value LIKE ''GWP%'' THEN REPLACE(value,''GWP '', '''')
                           WHEN value LIKE ''LR%'' THEN REPLACE(value,''LR '', '''')
                      END AS INT) AS Year,
                      CASE WHEN value LIKE ''GWP%'' THEN ''GWP''
                           WHEN value LIKE ''LR%'' THEN ''LR''
                      END AS Metrics
                    FROM Unpivoted';


    EXECUTE(@query);