Alfred Perlstein <bright(at)wintelcom(dot)net> writes:
>> Reimplementation of UNION/INTERSECT/EXCEPT.
> Does this mean that in the next release EXCEPT will be a lot faster?
> Will I probably be able to drop my "NOT EXISTS" hacks that I've
> been using?
UNION/INTERSECT/EXCEPT are now all basically a sort phase and a
unique-filter phase, with minor variations on what the unique filter
thinks it should output. So the cost should be O((M+N) log (M+N)) for
M+N input tuples, as opposed to O(M*N) for the old INTERSECT and
EXCEPT code.
I didn't do anything to change EXISTS ...
regards, tom lane