Greg Stark <gsstark(at)mit(dot)edu> writes:
> This is sort of interesting. It seems the time required for the optimizer to
> handle a query doubles with every layer of subquery it has to dig through.
Seems to be due to some brain fade in pull_up_subqueries --- the thing
was constructing truly monstrous rangetables due to poorly thought out
recursive modification of the parse tree :-(. Patch applied in CVS tip.
regards, tom lane