Joe Conway <mail(at)joeconway(dot)com> writes:
> I wonder if "[NOT] IN (subselect)" could be improved with a hash table in
> similar fashion to the hash aggregate solution Tom recently implemented?
It's being worked on ;-)
http://archives.postgresql.org/pgsql-hackers/2002-11/msg01055.php
Assuming I get this done, the conventional wisdom that "EXISTS
outperforms IN" will be stood on its head --- unless we add planner code
to try to reverse-engineer an IN from an EXISTS, which is something I'm
not really eager to expend code and cycles on.
regards, tom lane