Scott Marlowe <smarlowe@g2switchworks.com> writes:
> Tom, would that help the planner make better choices for this kind of
> query?
I think the problem was that he had
select ... from a, b full join c on ... where ...
where table b is big and you only need a few rows from it, so it really
needs to be joined last, but the above forced doing it first. It wasn't
clear to me why he wanted the full join at all (in fact, you could see
from the plan that the planner had been able to reduce it to a left join
because there were WHERE clauses that'd discard one set of null-extended
rows anyway). Without knowing that, it's hard to say whether there's
another way to get what he wants.
regards, tom lane