Nice to see that it have been solved in the 2016 version, but for all of those that is not on that, here are two generalized and simplified versions of the methods above.
The XML-method is shorter, but of course requires the string to allow for the xml-trick (no 'bad' chars.)
XML-Method:
create function dbo.splitString(@input Varchar(max), @Splitter VarChar(99)) returns table as
Return
SELECT Split.a.value('.', 'VARCHAR(max)') AS Data FROM
( SELECT CAST ('<M>' + REPLACE(@input, @Splitter, '</M><M>') + '</M>' AS XML) AS Data
) AS A CROSS APPLY Data.nodes ('/M') AS Split(a);
Recursive method:
create function dbo.splitString(@input Varchar(max), @Splitter Varchar(99)) returns table as
Return
with tmp (DataItem, ix) as
( select @input , CHARINDEX('',@Input) --Recu. start, ignored val to get the types right
union all
select Substring(@input, ix+1,ix2-ix-1), ix2
from (Select *, CHARINDEX(@Splitter,@Input+@Splitter,ix+1) ix2 from tmp) x where ix2<>0
) select DataItem from tmp where ix<>0
Function in action
Create table TEST_X (A int, CSV Varchar(100));
Insert into test_x select 1, 'A,B';
Insert into test_x select 2, 'C,D';
Select A,data from TEST_X x cross apply dbo.splitString(x.CSV,',') Y;
Drop table TEST_X
XML-METHOD 2: Unicode Friendly (Addition courtesy of Max Hodges)
create function dbo.splitString(@input nVarchar(max), @Splitter nVarchar(99)) returns table as
Return
SELECT Split.a.value('.', 'NVARCHAR(max)') AS Data FROM
( SELECT CAST ('<M>' + REPLACE(@input, @Splitter, '</M><M>') + '</M>' AS XML) AS Data
) AS A CROSS APPLY Data.nodes ('/M') AS Split(a);