Rod Taylor <rbt(at)rbt(dot)ca> writes:
> I would think that a simple optimization would be to push off evaluation
> of a subplan whenever possible by re-arranging AND statements.
Thought we did that already ... [ checks code ] ... hmph, it looks like
this is done correctly for quals attached to a scan node, but not for
quals attached to a join node. Oops.
regards, tom lane