修改@Aaron Bertrand的功能

CREATE FUNCTION [dbo].[SplitString]
(
    @List NVARCHAR(MAX),
    @Delim VARCHAR(255),
    @Idx int
)
RETURNS NVARCHAR(1000)
AS
BEGIN
    DECLARE @ValueTable TABLE(String NVARCHAR(50), Ind int)
    DECLARE @Value NVARCHAR(50)
    BEGIN
    INSERT INTO @ValueTable
    SELECT Value, idx FROM
        (SELECT [Value], idx = RANK() OVER (ORDER BY n) FROM 
              ( 
                SELECT n = Number, 
                [Value] = LTRIM(RTRIM(SUBSTRING(@List, [Number],
                CHARINDEX(@Delim, @List + @Delim, [Number]) - [Number])))
                FROM    
                        (SELECT Number = ROW_NUMBER() OVER (ORDER BY name)
                         FROM sys.all_objects) AS x
                WHERE Number <= LEN(@List)
                AND SUBSTRING(@Delim + @List, [Number], LEN(@Delim)) = @Delim
              ) AS y
          ) AS R WHERE idx = @Idx
    SET @Value = (SELECT String FROM @ValueTable)
    END
    RETURN @Value
END
GO

I realize this is a really old question, but starting with SQL Server 2016 there are functions for parsing JSON data that can be used to specifically address the OP's question--and without splitting strings or resorting to a user-defined function. To access an item at a particular index of a delimited string, use the JSON_VALUE function. Properly formatted JSON data is required, however: strings must be enclosed in double quotes " and the delimiter must be a comma ,, with the entire string enclosed in square brackets [].

DECLARE @SampleString NVARCHAR(MAX) = '"Hello John Smith"';
--Format as JSON data.
SET @SampleString = '[' + REPLACE(@SampleString, ' ', '","') + ']';
SELECT 
    JSON_VALUE(@SampleString, '$[0]') AS Element1Value,
    JSON_VALUE(@SampleString, '$[1]') AS Element2Value,
    JSON_VALUE(@SampleString, '$[2]') AS Element3Value;

输出

Element1Value         Element2Value       Element3Value
--------------------- ------------------- ------------------------------
Hello                 John                Smith

(1 row affected)

下面的示例使用递归CTE

更新18.09.2013

CREATE FUNCTION dbo.SplitStrings_CTE(@List nvarchar(max), @Delimiter nvarchar(1))
RETURNS @returns TABLE (val nvarchar(max), [level] int, PRIMARY KEY CLUSTERED([level]))
AS
BEGIN
;WITH cte AS
 (
  SELECT SUBSTRING(@List, 0, CHARINDEX(@Delimiter,  @List + @Delimiter)) AS val,
         CAST(STUFF(@List + @Delimiter, 1, CHARINDEX(@Delimiter, @List + @Delimiter), '') AS nvarchar(max)) AS stval, 
         1 AS [level]
  UNION ALL
  SELECT SUBSTRING(stval, 0, CHARINDEX(@Delimiter, stval)),
         CAST(STUFF(stval, 1, CHARINDEX(@Delimiter, stval), '') AS nvarchar(max)),
         [level] + 1
  FROM cte
  WHERE stval != ''
  )
  INSERT @returns
  SELECT REPLACE(val, ' ','' ) AS val, [level]
  FROM cte
  WHERE val > ''
  RETURN
END

演示SQLFiddle

你可以在SQL用户定义函数解析带分隔符的字符串中找到有用的解决方案(来自代码项目)。

你可以使用这个简单的逻辑:

Declare @products varchar(200) = '1|20|3|343|44|6|8765'
Declare @individual varchar(20) = null

WHILE LEN(@products) > 0
BEGIN
    IF PATINDEX('%|%', @products) > 0
    BEGIN
        SET @individual = SUBSTRING(@products,
                                    0,
                                    PATINDEX('%|%', @products))
        SELECT @individual

        SET @products = SUBSTRING(@products,
                                  LEN(@individual + '|') + 1,
                                  LEN(@products))
    END
    ELSE
    BEGIN
        SET @individual = @products
        SET @products = NULL
        SELECT @individual
    END
END

几乎所有其他答案都是替换正在分割的字符串，这浪费了CPU周期并执行不必要的内存分配。

我在这里介绍了一种更好的进行字符串拆分的方法:http://www.digitalruby.com/split-string-sql-server/

代码如下:

SET NOCOUNT ON

-- You will want to change nvarchar(MAX) to nvarchar(50), varchar(50) or whatever matches exactly with the string column you will be searching against
DECLARE @SplitStringTable TABLE (Value nvarchar(MAX) NOT NULL)
DECLARE @StringToSplit nvarchar(MAX) = 'your|string|to|split|here'
DECLARE @SplitEndPos int
DECLARE @SplitValue nvarchar(MAX)
DECLARE @SplitDelim nvarchar(1) = '|'
DECLARE @SplitStartPos int = 1

SET @SplitEndPos = CHARINDEX(@SplitDelim, @StringToSplit, @SplitStartPos)

WHILE @SplitEndPos > 0
BEGIN
    SET @SplitValue = SUBSTRING(@StringToSplit, @SplitStartPos, (@SplitEndPos - @SplitStartPos))
    INSERT @SplitStringTable (Value) VALUES (@SplitValue)
    SET @SplitStartPos = @SplitEndPos + 1
    SET @SplitEndPos = CHARINDEX(@SplitDelim, @StringToSplit, @SplitStartPos)
END

SET @SplitValue = SUBSTRING(@StringToSplit, @SplitStartPos, 2147483647)
INSERT @SplitStringTable (Value) VALUES(@SplitValue)

SET NOCOUNT OFF

-- You can select or join with the values in @SplitStringTable at this point.

这里的大多数解决方案使用while循环或递归cte。我保证，如果你可以使用空格以外的分隔符，基于集合的方法会更好:

CREATE FUNCTION [dbo].[SplitString]
    (
        @List NVARCHAR(MAX),
        @Delim VARCHAR(255)
    )
    RETURNS TABLE
    AS
        RETURN ( SELECT [Value], idx = RANK() OVER (ORDER BY n) FROM 
          ( 
            SELECT n = Number, 
              [Value] = LTRIM(RTRIM(SUBSTRING(@List, [Number],
              CHARINDEX(@Delim, @List + @Delim, [Number]) - [Number])))
            FROM (SELECT Number = ROW_NUMBER() OVER (ORDER BY name)
              FROM sys.all_objects) AS x
              WHERE Number <= LEN(@List)
              AND SUBSTRING(@Delim + @List, [Number], LEN(@Delim)) = @Delim
          ) AS y
        );

示例用法:

SELECT Value FROM dbo.SplitString('foo,bar,blat,foo,splunge',',')
  WHERE idx = 3;

结果:

----
blat

您还可以将需要的idx作为参数添加到函数中，但我将把它作为练习留给读者。

您不能仅使用SQL Server 2016中添加的本地STRING_SPLIT函数来实现这一点，因为不能保证输出将按照原始列表的顺序呈现。换句话说，如果你传递3,6,1结果可能是这个顺序，但它可能是1,3,6。我已经在这里请求社区的帮助来改进内置功能:

请帮助改进STRING_SPLIT

有了足够的定性反馈，他们可能会考虑做出以下改进:

STRING_SPLIT不是特性完整的

更多关于拆分函数，为什么(和证明)while循环和递归cte不能扩展，以及更好的替代方案，如果拆分字符串来自应用层:

以正确的方式拆分字符串-或者退而求其次的方式 拆分字符串:后续 分割字符串:现在使用更少的T-SQL 比较字符串分割/连接方法 处理一个整数列表:我的方法 拆分整数列表:另一个汇总 更多关于拆分列表的内容:自定义分隔符、防止重复和维护顺序 在SQL Server中删除字符串中的重复项

在SQL Server 2016或更高版本上，你应该看看STRING_SPLIT()和STRING_AGG():

性能惊喜和假设:STRING_SPLIT() STRING_SPLIT()在SQL Server 2016:后续#1 STRING_SPLIT()在SQL Server 2016:后续#2 SQL Server v.Next: STRING_AGG()性能 使用SQL Server的新STRING_AGG和STRING_SPLIT函数解决老问题