Efficiency - SubQuery vs Distinct

Question

(Ora816/Win2K)
Which would be most efficient -
A non-correlated subquery, or a 'distinct' operator?
What if I were to have two levels of subquery (again non-correlated) - does that then become less or more efficient than the distinct?
Thanks

Answer 1

Consider two tables, tblDocument has a primary unique key of DocumentID. tblDocumentLookup has indexing on DocumentNo but it will not be unique in the table.
The version of the first derivation query using distinct would be:
select distinct d.DocumentID, d.VariousOtherFields
from tblDocument d, tblDocumentLookup dl
where d.DocumentID = dl.DocumentNo
and dl.Value = 'mystring';
versus
The version of the first derivation query using a non-correlated subquery would be:
select d.DocumentID, d.VariousOtherFields
from tblDocument d
where d.DocumentID in
(select dl.DocumentNo
from tblDocumentLookup dl
where dl.Value = 'mystring');
For the second derivation, consider a third table, tblLookup which has a primary unique key of LookupID. In the table tblDocumentLookup, the value LookupNo whilst indexed will not be unique in the table.
The version of the second derivation query using distinct would be:
select distinct d.DocumentID, d.VariousOtherFields
from tblDocument d, tblDocumentLookup dl, tblLookup l
where d.DocumentID = dl.DocumentNo and l.LookupID = dl.LookupNo
and dl.OtherValue = X
and l.Value = 'mystring';
versus
The version of the second derivation query using a non-correlated subquery would be:
select d.DocumentID, d.VariousOtherFields
from tblDocument d
where d.DocumentID in
(select dl.DocumentNo
from tblDocumentLookup dl
where dl.OtherValue = X
and dl.LookupNo in ( select l.LookupID
from tblLookup l
where l.Value = 'mystring' ));
I hope this helps explain. From my perspective I see distinct as being inefficient, and the subquery being inefficient, but my question really would be in this type of situation which will be the better of the two evils.
Thanks
Jason.