SQL Server UDF 的列别名错误

Question

我正在尝试创建一个存储的 sql server 函数，该函数将返回一个中值表，我可以将其连接回另一个表，因此：

CREATE FUNCTION [dbo].getmedian (@varPartionBy1 int,@varPartionBy2 int, @varForTieBreak int, @varForMeasure int)
    RETURNS TABLE
    AS
    RETURN 
    (
        SELECT
           @varPartionBy1,
           @varPartionBy2,
           AVG(@varForMeasure) 
        FROM
        (
           SELECT
              @varPartionBy1,
              @varPartionBy2,
              ROW_NUMBER() OVER (
                 PARTITION BY @varPartionBy1, @varPartionBy2
                 ORDER BY @varForMeasure ASC, @varForTieBreak ASC) AS RowAsc,
              ROW_NUMBER() OVER (
                 PARTITION BY @varPartionBy1, @varPartionBy2 
                 ORDER BY @varForMeasure ASC, @varForTieBreak  DESC) AS RowDesc
            from 
            [fakename].[dbo].[temptable] bp

        ) bp
        WHERE 
           RowAsc IN (RowDesc, RowDesc - 1, RowDesc + 1)
        GROUP BY @varPartionBy1, @varPartionBy2

    )

去

这将返回错误：“消息 8155，级别 16，状态 2，过程 getmedian，第 25 行 没有为 'bp' 的第 1 列指定列名。”——我猜我不明白如何在 UDF 的上下文中为列分配表别名。

我应该怎么做才能修复错误？

这是我的第一个 USF，因此我感谢您在解决主要问题时提供的任何其他有用的设计见解。感谢您的帮助！

Answer 1

您有SELECT @varPartionBy1, @varPartionBy2 的地方需要为它们分配列名。您可以直接分配它们，例如SELECT @varPartionBy1 AS varPartionBy1 或SELECT varPartionBy1 = @varPartionBy1，也可以在表中指定别名) bp(varPartionBy1, varPartionBy2,...

正确的函数可能是

CREATE FUNCTION [dbo].getmedian (@varPartionBy1 int,@varPartionBy2 int, @varForTieBreak int, @varForMeasure int)
RETURNS TABLE
AS
RETURN 
(
    SELECT
       varPartionBy1,
       varPartionBy2,
       AVG(@varForMeasure) AS AvgVarForMeasure
    FROM
    (
       SELECT
          @varPartionBy1 AS varPartionBy1,
          @varPartionBy2 As varPartionBy1,
          ROW_NUMBER() OVER (
             PARTITION BY @varPartionBy1, @varPartionBy2
             ORDER BY @varForMeasure ASC, @varForTieBreak ASC) AS RowAsc,
          ROW_NUMBER() OVER (
             PARTITION BY @varPartionBy1, @varPartionBy2 
             ORDER BY @varForMeasure ASC, @varForTieBreak  DESC) AS RowDesc
        from 
        [fakename].[dbo].[temptable] bp

    ) bp
    WHERE 
       RowAsc IN (RowDesc, RowDesc - 1, RowDesc + 1)
    GROUP BY varPartionBy1, varPartionBy2

)

Answer 2

好吧，一旦你解决了语法问题，你的问题就不会结束。一方面，你不能真正做你想做的事（将变量传递给PARTITION BY 和ORDER BY 子句）。这些只是被视为常量，因此ROW_NUMBER() 将被任意应用。

观察这里发生了什么：

DECLARE @foo SYSNAME = N'name';

SELECT name, foo = @foo, -- this is just a constant value, not a column
  varASC  = ROW_NUMBER() OVER (ORDER BY @foo ASC), 
  varDESC = ROW_NUMBER() OVER (ORDER BY @foo DESC), 
  colASC  = ROW_NUMBER() OVER (ORDER BY name ASC), 
  colDESC = ROW_NUMBER() OVER (ORDER BY name DESC) 
FROM sys.objects --WHERE is_ms_shipped = 0
ORDER BY varASC;

部分结果：

name  foo   varASC  varDESC  colASC  colDESC
----  ----  ------  -------  ------  -------
t1    name       1        1       1      100
t2    name       2        2       2       99
t3    name       3        3       3       98
t4    name       4        4       4       97
t5    name       5        5       5       96
------ only column that deviates ----^^^^^^^

@foo 的变量值在每一行上都是相同的，因此，按此划分和排序完全没有意义。