That said, it seems reasonable to think that the optimizer should at
least consider doing a hash join on the subquery, in which case the
join between T2 and T3 could be materialized and then a hash-join
could be done using the predicate x1.j = t1.i.
I think the best thing would be to flatten the subquery into the
outer query so that the optimizer will see it as just another join.
BTW, in your example the subquery contains a Cartesian product. Is
that intentional?
- Jeff Lichtman
[EMAIL PROTECTED]
Check out Swazoo Koolak's Web Jukebox at
http://swazoo.com/
