Question

有没有人知道我能用以下格式规范化字符串的最佳方法： -

'(20111026,1000,34.10)(20111027,1000,44.10)(20111028,1000,54.10)(20111029,1000,64.10)(20111030,1000,74.10)'

分为5行，3列？

|Date       |Time     |Amount|
-------------------------------
|2011-10-26 |10:00:00 |34.10 |
|2011-10-27 |10:00:00 |44.10 |
|2011-10-28 |10:00:00 |54.10 |
|2011-10-29 |10:00:00 |64.10 |
|2011-10-30 |10:00:00 |74.10 |

我已经设法使用字符串解析器和分隔符'）（'获取行和'，'再次获取列。但是当我对700万字符串执行此操作时，db会爆炸。

这是我到目前为止的SQL：

DECLARE @Text VARCHAR(500) = '(20111026,1000,34.10)(20111027,1000,44.10)(20111028,1000,54.10)(20111029,1000,64.10)(20111030,1000,74.10)'

SELECT  
   TRY_CONVERT(DATE, [1]) AS StartDate ,
   CAST(TRY_CONVERT(TIME(0), DATEADD(HOUR, ( [2] / 100 ) % 100, DATEADD(MINUTE, ( [2] / 1 ) % 100, CAST('00:00' AS TIME)))) AS VARCHAR(8)) AS StartTime ,
   TRY_CONVERT(NUMERIC(16, 6), [3]) AS Amount
FROM    
   (SELECT 
       X.Ordinal AS RoNum ,
       Y.Ordinal AS ColNum ,
       REPLACE(Y.StringValue, '(', '') AS Value
    FROM 
       dbo.ParseString(@Text, ')(') X
    CROSS APPLY 
       dbo.ParseString(StringValue,',') Y
    WHERE 
       NOT Y.StringValue = '') AS SRC 
PIVOT
( MIN(Value) FOR ColNum IN ( [1], [2], [3] ) ) AS PVT;

解析字符串函数： -

CREATE FUNCTION [dbo].[ParseString]
   (@String VARCHAR(500), @Delimiter CHAR(1))
RETURNS TABLE
AS
   RETURN
     (WITH Results AS
        (SELECT 1 AS Ordinal,
            LTRIM(LEFT(@String, CHARINDEX(@Delimiter, @String + @Delimiter)-1)) AS StringValue,
            CONVERT(VARCHAR(500), RIGHT(@String + @Delimiter, LEN(@String) - CHARINDEX(@Delimiter, @String+@Delimiter) + 1)) AS Remaining
    UNION ALL
    SELECT  Ordinal+1,
            LTRIM(LEFT(Remaining, CHARINDEX(@Delimiter, Remaining)-1)),
            RIGHT(Remaining, LEN(Remaining) - CHARINDEX(@Delimiter, Remaining))
    FROM    Results
    WHERE   LEN(Remaining) > 0)
 SELECT Ordinal,
        StringValue
 FROM   Results
)

非常感谢任何帮助。

Answer 1

没有必要为这样的事情分配字符串。您发布的格式几乎正确使用表值构造函数。使用替换来在（）已经很好地包装的值集之间粘贴逗号意味着你可以使用一些动态的sql轻松地完成这项操作。

declare @String varchar(max) = '(20111026,1000,34.10)(20111027,1000,44.10)(20111028,1000,54.10)(20111029,1000,64.10)(20111030,1000,74.10)'
declare @SQL nvarchar(max)

set @SQL = 'select * from (VALUES ' + REPLACE(@String, ')(', '),(') + ')N (col1, col2, col3)'

select @SQL
exec sp_executesql @SQL

使用sql server将字符串规范化为rows \ columns

1 个答案: