sql server 查询 - 在 n 列之后进行透视

Posted 2023-04-12

【中文标题】sql server 查询 - 在 n 列之后进行透视

【英文标题】：sql server query - pivot after n columns

【发布时间】：2013-02-01 11:46:33

【问题描述】：

我有一个包含 N 列的表格：

TABLE:
TB_SAMPLE

COLUMNS:

VALUE_A
COUNT_A
VALUE_B
COUNT_B
VALUE_C
COUNT_C
VALUE_D
COUNT_D
VALUE_E
COUNT_E
VALUE_F
COUNT_F
VALUE_G
COUNT_G
VALUE_H
COUNT_H
VALUE_I
COUNT_I

我想这样返回：

'Value', 'Count'
A, 1 -- this is the first and the second column
B, 3 -- this is the third and the fourth column
C, 4 -- this is the fifth and sixth column

等等。我已经尝试使用 UNION ALL 子句进行多项选择，但是由于我要返回很多记录，所以速度不是那么快。有没有更好的方法来“透视”这个表的 2 列？

谢谢。

【问题讨论】：

如何计算每个列或值 A 的计数？

你能从你的表中发布一些示例数据吗？很难说出想要的结果是什么。

@techdo 这是另一个保存此信息的应用程序

@bluefeet 例如，我在一个表中有 30 列。我想每两列切换一次。 （第一列和第二列连续，第三和第四列在第二列，依此类推）。

【参考方案1】：

这也可以使用CROSS APPLY 和VALUES 语句来执行：

SELECT vals, counts
FROM temp t
CROSS APPLY 
(
    VALUES
        (value_a, count_a),
        (value_b, count_b),
        (value_c, count_c),
        (value_d, count_d),
        (value_e, count_e),
        (value_f, count_f),
        (value_g, count_g),
        (value_h, count_h)
) x (vals, counts);

见SQL Fiddle with Demo。

这里有一篇文章解释了这是如何完成的（需要注册）：

http://www.sqlservercentral.com/articles/CROSS+APPLY+VALUES+UNPIVOT/91234/

【讨论】：

不错的解决方案。虽然，您发布的链接需要注册，这有点痛苦......

【参考方案2】：

您可以 UNPIVOT，然后 PIVOT 数据，它可能会表现得更好：

DECLARE @T TABLE 
(   VALUE_A INT, COUNT_A INT, VALUE_B INT, COUNT_B INT, VALUE_C INT, COUNT_C INT, VALUE_D INT, COUNT_D INT, VALUE_E INT, 
    COUNT_E INT, VALUE_F INT, COUNT_F INT, VALUE_G INT, COUNT_G INT, VALUE_H INT, COUNT_H INT, VALUE_I INT, COUNT_I INT
);
INSERT @T VALUES (1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1);

WITH D AS
(   SELECT  Category = SUBSTRING(B, 7, LEN(B)), [Type] = 'Value', A
    FROM    @T
            UNPIVOT 
            (   A
                FOR B IN ([VALUE_A], [VALUE_B], [VALUE_C], [VALUE_D], [VALUE_E], [VALUE_F], [VALUE_G], [VALUE_H], [VALUE_I])
            ) upvt
    UNION ALL
    SELECT  Category = SUBSTRING(B, 7, LEN(B)), [Type] = 'Count', A
    FROM    @T
            UNPIVOT 
            (   A
                FOR B IN ([COUNT_A], [COUNT_B], [COUNT_C], [COUNT_D], [COUNT_E], [COUNT_F], [COUNT_G], [COUNT_H], [COUNT_I])
            ) upvt
)
SELECT  *
FROM    D
        PIVOT
        (   SUM(A)
            FOR [Type] IN ([Value], [Count])
        ) pvt;

如果需要，您也可以通过在系统视图中查询列名来动态执行此操作。

附录

刚刚意识到这可以通过联接而不是 UNION 和 PIVOT 来实现，我想联接会在删除 PIVOT 操作时表现得更好。我仍然建议测试两者，看看哪个最适合您的数据。

WITH Val AS
(   SELECT  Category = SUBSTRING(B, 7, LEN(B)), A
    FROM    @T
            UNPIVOT 
            (   A
                FOR B IN ([VALUE_A], [VALUE_B], [VALUE_C], [VALUE_D], [VALUE_E], [VALUE_F], [VALUE_G], [VALUE_H], [VALUE_I])
            ) upvt
), Co AS
(   SELECT  Category = SUBSTRING(B, 7, LEN(B)), A
    FROM    @T
            UNPIVOT 
            (   A
                FOR B IN ([COUNT_A], [COUNT_B], [COUNT_C], [COUNT_D], [COUNT_E], [COUNT_F], [COUNT_G], [COUNT_H], [COUNT_I])
            ) upvt
)
SELECT  val.Category,
        [Value] = val.A,
        [Count] = co.A
FROM    val
        INNER JOIN co
            ON co.Category = val.Category;

【参考方案3】：

这是另一种方法（如果我正确理解了您的问题）

declare @table table
(
value_a varchar(10),
count_a int,
value_b varchar(10),
count_b int,
value_c varchar(10),
count_c int,
value_d varchar(10),
count_d int,
value_e varchar(10),
count_e int,
value_f varchar(10),
count_f int,
value_g varchar(10),
count_g int,
value_h varchar(10),
count_h int
)

Insert Into @table
Select  'value a',675,
        'value b',646,
        'value c',13,
        'value d',22,
        'value e',768,
        'value f',223,
        'value g',66,
        'value h',55


SELECT  vals,
        counts
FROM 
   (
    SELECT  *
   FROM @table
    ) p
UNPIVOT
   (vals FOR cols IN 
      ([value_a],[value_b],[value_c],[value_d], [value_e],[value_f],[value_g],[value_h])
)AS unpvtValues
Unpivot 
    (
        counts for counters in ([count_a],[count_b],[count_c], [count_d],[count_e],[count_f],[count_g],[count_h])
    ) as unpvtCounts
where Right(cols, 2) = right(counters, 2)

编辑

如果你想使用一些动态 SQL，那么你可以使用这个选项（这会在你的数据库中创建一个适当的表，而不是使用表变量）

create  Table  dbo.TempTable
(
value_a varchar(10),
count_a int,
value_b varchar(10),
count_b int,
value_c varchar(10),
count_c int,
value_d varchar(10),
count_d int,
value_e varchar(10),
count_e int,
value_f varchar(10),
count_f int,
value_g varchar(10),
count_g int,
value_h varchar(10),
count_h int
)

Insert Into dbo.TempTable
Select  'value a',675,
        'value b',646,
        'value c',13,
        'value d',22,
        'value e',768,
        'value f',223,
        'value g',66,
        'value h',55


Declare @ValueCols varchar(max),
        @CountCols varchar(max),
        @sql varchar(max)


Select @ValueCols = Coalesce(@ValueCols + ', ','') + '[' + Column_Name +']'
From INFORMATION_SCHEMA.COLUMNS
Where TABLE_NAME  = 'TempTable'
And COLUMN_NAME Like 'Value%'

Select @CountCols = Coalesce(@CountCols + ', ','') + '[' + Column_Name +']'
From INFORMATION_SCHEMA.COLUMNS
Where TABLE_NAME  = 'TempTable'
And COLUMN_NAME Like 'Count%'


Set @sql = 'SELECT  vals,
        counts
FROM 
   (
    SELECT  *
   FROM dbo.TempTable
    ) p
    UNPIVOT
   (vals FOR cols IN 
      (' + @ValueCols +')
)AS unpvtValues
Unpivot 
    (
        counts for counters in (' + @CountCols + ')
    ) as unpvtCounts
where Right(cols, 2) = right(counters, 2)'


Execute (@sql)

drop table TempTable