用户定义的函数替换 WHERE col IN(...)

Question

我已经创建了一个用户定义的函数，以通过包含 'WHERE col IN (...)' 的查询来获得性能，例如这种情况：

SELECT myCol1, myCol2
FROM myTable
WHERE myCol3 IN (100, 200, 300, ..., 4900, 5000);

查询是从 Web 应用程序生成的，并且在某些情况下要复杂得多。 函数定义如下：

CREATE FUNCTION [dbo].[udf_CSVtoIntTable]
(
  @CSV VARCHAR(MAX),
  @Delimiter CHAR(1) = ','
)
RETURNS 
@Result TABLE 
(
    [Value] INT
)
AS
BEGIN

  DECLARE @CurrStartPos SMALLINT;
  SET @CurrStartPos = 1;
  DECLARE @CurrEndPos SMALLINT;
  SET @CurrEndPos = 1;
  DECLARE @TotalLength SMALLINT;

  -- Remove space, tab, linefeed, carrier return
  SET @CSV = REPLACE(@CSV, ' ', '');
  SET @CSV = REPLACE(@CSV, CHAR(9), '');
  SET @CSV = REPLACE(@CSV, CHAR(10), '');
  SET @CSV = REPLACE(@CSV, CHAR(13), '');

  -- Add extra delimiter if needed
  IF NOT RIGHT(@CSV, 1) = @Delimiter
    SET @CSV = @CSV + @Delimiter;

  -- Get total string length 
  SET @TotalLength = LEN(@CSV);

  WHILE @CurrStartPos < @TotalLength
  BEGIN

    SET @CurrEndPos = CHARINDEX(@Delimiter, @CSV, @CurrStartPos);

    INSERT INTO @Result
    VALUES (CAST(SUBSTRING(@CSV, @CurrStartPos, @CurrEndPos - @CurrStartPos) AS INT));

    SET @CurrStartPos = @CurrEndPos + 1;

  END

    RETURN 

END

该函数旨在像这样使用（或作为 INNER JOIN）：

SELECT myCol1, myCol2
FROM myTable
WHERE myCol3 IN (
    SELECT [Value] 
    FROM dbo.udf_CSVtoIntTable('100, 200, 300, ..., 4900, 5000', ',');

在我的情况下，是否有人对我的功能或其他提高性能的方法有一些优化想法？ 有没有我遗漏的缺点？

我正在使用 MS SQL Server 2005 Std 和 .NET 2.0 框架。

Answer 1

我不确定性能是否会提高，但我会将其用作内部连接并远离内部选择语句。

Answer 2

在 WHERE 子句或（更糟糕的）子查询中使用 UDF 是自找麻烦。优化器有时会做对，但经常会出错，并为查询中的每一行评估一次函数，这是您不想要的。

如果您的参数是静态的（它们看起来是静态的）并且您可以发出多语句批处理，我会将您的 UDF 的结果加载到表变量中，然后对表变量使用连接来进行过滤。这应该更可靠地工作。

Answer 3

那个循环会影响性能！

创建一个这样的表：

CREATE TABLE Numbers
(
    Number  int   not null primary key
)

具有包含值 1 到 8000 左右的行并使用此函数：

CREATE FUNCTION [dbo].[FN_ListAllToNumberTable]
(
     @SplitOn  char(1)       --REQUIRED, the character to split the @List string on
    ,@List     varchar(8000) --REQUIRED, the list to split apart
)
RETURNS
@ParsedList table
(
    RowNumber int
   ,ListValue varchar(500)
)
AS
BEGIN

/*
DESCRIPTION: Takes the given @List string and splits it apart based on the given @SplitOn character.
             A table is returned, one row per split item, with a columns named "RowNumber" and "ListValue".
             This function workes for fixed or variable lenght items.
             Empty and null items will be included in the results set.

PARAMETERS:
    @List      varchar(8000) --REQUIRED, the list to split apart
    @SplitOn   char(1)       --OPTIONAL, the character to split the @List string on, defaults to a comma ","


RETURN VALUES:
  a table, one row per item in the list, with a column name "ListValue"

TEST WITH:
----------
SELECT * FROM dbo.FN_ListAllToNumTable(',','1,12,123,1234,54321,6,A,*,|||,,,,B')

DECLARE @InputList  varchar(200)
SET @InputList='17;184;75;495'
SELECT
    'well formed list',LEFT(@InputList,40) AS InputList,h.Name
    FROM Employee  h
        INNER JOIN dbo.FN_ListAllToNumTable(';',@InputList) dt ON h.EmployeeID=dt.ListValue
    WHERE dt.ListValue IS NOT NULL

SET @InputList='17;;;184;75;495;;;'
SELECT
    'poorly formed list join',LEFT(@InputList,40) AS InputList,h.Name
    FROM Employee  h
        INNER JOIN dbo.FN_ListAllToNumTable(';',@InputList) dt ON h.EmployeeID=dt.ListValue

SELECT
    'poorly formed list',LEFT(@InputList,40) AS InputList, ListValue
    FROM dbo.FN_ListAllToNumTable(';',@InputList)

**/



/*this will return empty rows, and row numbers*/
INSERT INTO @ParsedList
        (RowNumber,ListValue)
    SELECT
        ROW_NUMBER() OVER(ORDER BY number) AS RowNumber
            ,LTRIM(RTRIM(SUBSTRING(ListValue, number+1, CHARINDEX(@SplitOn, ListValue, number+1)-number - 1))) AS ListValue
        FROM (
                 SELECT @SplitOn + @List + @SplitOn AS ListValue
             ) AS InnerQuery
            INNER JOIN Numbers n ON n.Number < LEN(InnerQuery.ListValue)
        WHERE SUBSTRING(ListValue, number, 1) = @SplitOn

RETURN

END /*Function FN_ListAllToNumTable*/

我有其他版本不返回空行或空行，只返回项目而不是行号等。查看标题注释以了解如何将其用作 JOIN 的一部分，这很多比 where 子句更快。

Answer 4

CLR 解决方案没有给我很好的性能，所以我将使用递归查询。所以这里是我将使用的 SP 的定义（主要基于 Erland Sommarskogs 的例子）：

CREATE FUNCTION [dbo].[priudf_CSVtoIntTable]
(
  @CSV VARCHAR(MAX),
  @Delimiter CHAR(1) = ','
)
RETURNS 
@Result TABLE 
(
    [Value] INT
)
AS
BEGIN

  -- Remove space, tab, linefeed, carrier return
  SET @CSV = REPLACE(@CSV, ' ', '');
  SET @CSV = REPLACE(@CSV, CHAR(9), '');
  SET @CSV = REPLACE(@CSV, CHAR(10), '');
  SET @CSV = REPLACE(@CSV, CHAR(13), '');

  WITH csvtbl(start, stop) AS 
  (
    SELECT  start = CONVERT(BIGINT, 1),
            stop = CHARINDEX(@Delimiter, @CSV + @Delimiter)
    UNION ALL
    SELECT  start = stop + 1,
            stop = CHARINDEX(@Delimiter, @CSV + @Delimiter, stop + 1)
    FROM csvtbl
    WHERE stop > 0
  )
  INSERT INTO @Result
  SELECT CAST(SUBSTRING(@CSV, start, CASE WHEN stop > 0 THEN stop - start ELSE 0 END) AS INT) AS [Value]
  FROM   csvtbl
  WHERE  stop > 0
  OPTION (MAXRECURSION 1000)

  RETURN 
END

Answer 5

感谢您的意见，我不得不承认我在开始工作之前做了一些糟糕的研究。我发现Erland Sommarskog在他的网页上写了很多这个问题，在你的回复和阅读他的页面后，我决定我会尝试做一个CLR来解决这个问题。

我尝试了递归查询，结果很好，但我还是会尝试 CLR 函数。