Tsql 在一个表的列中分解csv以插入到新表中

如何拆分如下所示的值： 'some:1234，dumb:4321，thing:6534，some:65432，did:43287'

为此：

value1    value2
some      1234
dumb      4321
thing     6534
someone   65432
did       43287

用于插入到新表中

下面的代码显示了一些我想要实现的东西。我已经为此工作了大约一天了，现在把头发拔出来

我还包括了表和函数定义以及几行数据，以说明我希望实现的目标

-- Shamelessly copied this function from the CTE String Split here: https://sqlperformance.com/2012/07/t-sql-queries/split-strings
CREATE FUNCTION dbo.splitStrings -- Just being trying to be complete so adding this function since it is not an mssql built-in
(
   @List       NVARCHAR(MAX),
   @Delimiter  NVARCHAR(255)
)
RETURNS @Items TABLE (Item NVARCHAR(4000))
WITH SCHEMABINDING
AS
BEGIN
   DECLARE @ll INT = LEN(@List) + 1, @ld INT = LEN(@Delimiter);

   WITH a AS
   (
       SELECT
           [start] = 1,
           [end]   = COALESCE(NULLIF(CHARINDEX(@Delimiter, 
                       @List, 1), 0), @ll),
           [value] = SUBSTRING(@List, 1, 
                     COALESCE(NULLIF(CHARINDEX(@Delimiter, 
                       @List, 1), 0), @ll) - 1)
       UNION ALL
       SELECT
           [start] = CONVERT(INT, [end]) + @ld,
           [end]   = COALESCE(NULLIF(CHARINDEX(@Delimiter, 
                       @List, [end] + @ld), 0), @ll),
           [value] = SUBSTRING(@List, [end] + @ld, 
                     COALESCE(NULLIF(CHARINDEX(@Delimiter, 
                       @List, [end] + @ld), 0), @ll)-[end]-@ld)
       FROM a
       WHERE [end] < @ll
   )
   INSERT @Items SELECT [value]
   FROM a
   WHERE LEN([value]) > 0
   OPTION (MAXRECURSION 0);

   RETURN;
END
GO

-- Primary table where the messed up data currently resides
create table dbo.things
(
    Id int identity(1,1) not null,
    User int not null,
    Values varchar(256)
)
-- This is the bad data
insert into things(USER,Values) values
{100,'some:1234,dumb:4321,thing:6534,someone:65432,did:43287'},
{101,'this:31234,is:43221,so wrong:65482'}

-- This is where the split up data will reside in the future
create table dbo.propertableforvalues
(
        ThingId int not null,
        ValueID int not null,
    ValueName varchar(256) not null
)

-- This doesn't come close to working, but 'looks' like what I'm trying to achieve
insert into dbo.propertableforvalues
select
(
    select 
        t.Id,
        outerSplit.Name,
        outerSplit.ValueId
    from 
        (select -- one 'things' record should create n rows based on the number of items in the comma split of the 'Values' string
            innerT.Name,
                        innerT.ValueId
         from 
                        dbo.splitStrings(
                        (select
                            ValueId,
                            ValueName
                        from
                                dbo.splitStrings(t.Values, ',') -- Split first based on the comma
                                    , ':')) as innerSplit                               -- Split a second time on the colon
        ) as outerSplit
)
from
    things t
where
    t.Values is not null
    and t.Values <> ''

---

一种选择是将字符串转换为XML，然后简单地解析XML

范例

编辑-通知数据位于表中

---

在这种情况下，我更喜欢这种方法：

这是你的绳子

DECLARE @csv VARCHAR(100)='some:1234,dumb:4321,thing:6534,someone:65432,did:43287';

某些替换将此字符串转换为XML：

SELECT CAST('<x><y>' + REPLACE(REPLACE((SELECT @csv AS [*] FOR XML PATH('')),',','</y></x><x><y>'),':','</y><y>') + '</y></x>' AS XML);

这是我们可以用XML方法查询的结果

<x>
  <y>some</y>
  <y>1234</y>
</x>
<x>
  <y>dumb</y>
  <y>4321</y>
</x>
<x>
  <y>thing</y>
  <y>6534</y>
</x>
<x>
  <y>someone</y>
  <y>65432</y>
</x>
<x>
  <y>did</y>
  <y>43287</y>
</x>

试一试：

DECLARE @csv VARCHAR(100)='some:1234,dumb:4321,thing:6534,someone:65432,did:43287';

SELECT x.value('y[1]','nvarchar(max)') AS Item
      ,x.value('y[2]','nvarchar(max)') AS [Value]
FROM (SELECT CAST('<x><y>' + REPLACE(REPLACE((SELECT @csv AS [*] FOR XML PATH('')),',','</y></x><x><y>'),':','</y><y>') + '</y></x>' AS XML)) t(casted)
CROSS APPLY t.casted.nodes('/x') A(x);

使现代化 表数据也是如此

DECLARE @things TABLE
(
    Id int identity(1,1) not null,
    [User] int not null,
    [Values] varchar(256)
)
insert into @things([USER],[Values]) values
(100,'some:1234,dumb:4321,thing:6534,someone:65432,did:43287'),
(101,'this:31234,is:43221,so wrong:65482');

SELECT t.Id
      ,t.[User] 
      ,x.value('y[1]','nvarchar(max)') AS Item
      ,x.value('y[2]','nvarchar(max)') AS [Value]
FROM @things t
CROSS APPLY (SELECT CAST('<x><y>' + REPLACE(REPLACE((SELECT t.[Values] AS [*] FOR XML PATH('')),',','</y></x><x><y>'),':','</y><y>') + '</y></x>' AS XML)) A(casted)
CROSS APPLY A.casted.nodes('/x') B(x);

---

作为一种选择，我将不发布执行相同任务的xml方法，不破坏前面给出的答案

为了使此方法正常工作，@CSV字符串必须在开头和结尾用分隔符括起来：

DECLARE @csv VARCHAR(8000) = 'some:1234,dumb:4321,thing:6534,some one:65432,did:43287'

SET @csv = ',' + @csv + ','

首先，建立一个数字cte，也称为计数：

并找到separatordelimiter的索引，在我们的例子中，separatordelimiter是一个'

SELECT N
FROM TALLY
WHERE SUBSTRING(@csv,N,1) = ','

然后使用子字符串并将索引值为delimiter的CSV切片以获得第一部分，然后使用新的delimiter重复相同的操作，即“：”

---

只有一个提示：如果项作为属性名无效，或者字符串可能包含禁止的字符，这种方法可能会遇到麻烦。。。就像第二个例子中的错误…John，你看到提供的测试数据了吗？不知道OP的真实世界数据，但101，“这：31234，是：43221，所以错误：65482”将由于错误的空白而中断…@ Shnugo它感觉像JSON对象。但你是对的。如果错了，它就会失败。没错。。。这是用sql-server-2012标记的-遗憾的是没有JSON支持。。。仍然是一个非常圆滑的方法-@Shnugo谢谢但你的X/Y更安全

<x>
  <y>some</y>
  <y>1234</y>
</x>
<x>
  <y>dumb</y>
  <y>4321</y>
</x>
<x>
  <y>thing</y>
  <y>6534</y>
</x>
<x>
  <y>someone</y>
  <y>65432</y>
</x>
<x>
  <y>did</y>
  <y>43287</y>
</x>

DECLARE @csv VARCHAR(100)='some:1234,dumb:4321,thing:6534,someone:65432,did:43287';

SELECT x.value('y[1]','nvarchar(max)') AS Item
      ,x.value('y[2]','nvarchar(max)') AS [Value]
FROM (SELECT CAST('<x><y>' + REPLACE(REPLACE((SELECT @csv AS [*] FOR XML PATH('')),',','</y></x><x><y>'),':','</y><y>') + '</y></x>' AS XML)) t(casted)
CROSS APPLY t.casted.nodes('/x') A(x);

DECLARE @things TABLE
(
    Id int identity(1,1) not null,
    [User] int not null,
    [Values] varchar(256)
)
insert into @things([USER],[Values]) values
(100,'some:1234,dumb:4321,thing:6534,someone:65432,did:43287'),
(101,'this:31234,is:43221,so wrong:65482');

SELECT t.Id
      ,t.[User] 
      ,x.value('y[1]','nvarchar(max)') AS Item
      ,x.value('y[2]','nvarchar(max)') AS [Value]
FROM @things t
CROSS APPLY (SELECT CAST('<x><y>' + REPLACE(REPLACE((SELECT t.[Values] AS [*] FOR XML PATH('')),',','</y></x><x><y>'),':','</y><y>') + '</y></x>' AS XML)) A(casted)
CROSS APPLY A.casted.nodes('/x') B(x);

DECLARE @csv VARCHAR(8000) = 'some:1234,dumb:4321,thing:6534,some one:65432,did:43287'

SET @csv = ',' + @csv + ','

-- Build 10000 numbers.
;WITH
TENS (N)      AS (SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0),
THOUSANDS (N) AS (SELECT 1
                FROM TENS t1
                     CROSS JOIN TENS t2
                     CROSS JOIN TENS t3),
TALLY (N)     AS (SELECT ROW_NUMBER() OVER (ORDER BY (SELECT 0))
                FROM THOUSANDS)

SELECT N
FROM TALLY
WHERE SUBSTRING(@csv,N,1) = ','

DECLARE @csv VARCHAR(8000) = 'some:1234,dumb:4321,thing:6534,some one:65432,did:43287'

SET @csv = ',' + @csv + ','

-- Build 10000 numbers.
;WITH
TENS (N)      AS (SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0 UNION ALL
                  SELECT 0 UNION ALL SELECT 0),
THOUSANDS (N) AS (SELECT 1
                FROM TENS t1
                     CROSS JOIN TENS t2
                     CROSS JOIN TENS t3),
TALLY (N)     AS (SELECT ROW_NUMBER() OVER (ORDER BY (SELECT 0))
                FROM THOUSANDS)
--Split CSV values into columns
,SPLITTED AS (
            SELECT SUBSTRING(@csv,N+1,CHARINDEX(',',@csv,N+1)-N-1) AS ColName
            FROM TALLY
            WHERE N < LEN(@csv) AND SUBSTRING(@csv,N,1) = ','  )

--Split column values further into as multiple columns
SELECT Part1 = SUBSTRING(ColName,1,CHARINDEX(':', ColName,1)-1),
       Part2 = SUBSTRING(ColName,CHARINDEX(':', ColName,1)+1,LEN(ColName))
FROM SPLITTED