declare @t table(Id1 int, Id2 int)
insert @t values (100, 50)
insert @t values ( 120, 70)
insert @t values ( 70, 50)
insert @t values ( 34, 20)
insert @t values ( 50, 40)
insert @t values ( 40, 10)
;with a as
(
-- find all rows without parent <*>
select id2, id1 from @t t where not exists (select 1 from @t where t.id1 = id2)
union all -- recusive work down to lowest child while storing the parent id1
select t.id2 , a.id1
from a
join @t t on a.id2 = t.id1
)
-- show the lowest child for each row found in <*>
select id1, min(id2) id2 from a
group by id1
Resultat:
id1 id2
----------- -----------
34 20
100 10
120 10