Question

I have 2 simplified tables (all columns are varchar). Some rows in T1_TAB for F2 contain multiple values separated by ;, some do not have separators at all as shown below (sometimes ; might also appear at the beginning and/or at the end). F2 in T2_TAB always has a single value.

I need to be able to pull rows from ether table based on single selection from one table and likeliness on F2 columns.

T1_TAB

F0   |  F2
--------------
1          ;30
2       ;10;20;30
3          ;20;30;
4          10

T2_TAB

F1   |  F2
-------------
100       10    
200       20      
300       30

I can do:

SELECT T1.F0
FROM T1_TAB T1 
LEFT JOIN T2_TAB T2
ON T2.F2 LIKE '%' + T1.F2 + '%'
WHERE T2.F1 = '200'

This would bare result:

2
3

Now, I need to do the opposite. For instance:

Based on condition WHERE T1.F0 = 3, I need to pull from T2 rows with F1 equals 200 and 300 respectively. I guess I need to somehow split ;20;30; by a “;” and do the loop to match each value separately at run-time disregarding blank tokens.

Answer 1

You can create a function which converts the semicolon-separated string into a table of values:

create function CreateTableFromList(@Values varchar(1000))
returns @table table (id int)
as
begin
declare @p int = 1, @q int = 1, @n int = len(@Values)

while @p<@n begin
  set @q = CHARINDEX(';',@Values,@p)
  if @q=0 set @q=@n + 1
  if @q > @p 
    insert into @table values (cast(substring(@Values,@p,@q-@p) as int))
  set @p= @q + 1
end
return
end

and then

select T2.F1 
  from T2
where T2.F2 in (
  select ID 
    from T1 
         cross apply dbo.CreateTableFromList(T1.F2) 
  where T1.F0=3
  )