Vivek Khera <khera@kcilink.com> writes:
> I, along with at least Erik, was thinking that the constraint
> expression would be evaluated to determine whether to include the
> partition in the final plan. Based on Tom's description, it is not
> the case: the planner basically proves that the constraint will be
> false. Until this was clarified, Tom's points totally confused the
> heck out of me.
> It would be amazingly wonderful if this distinction could be posted to
> the online docs. It will surely help future generations :-)
Feel free to send in a proposed doc patch. I'm not very clear on where
you think this should go or what it should say instead of what it does
say.
BTW, I always think of it the other way around: we're proving that the
WHERE condition must be false for any row meeting the check constraint.
regards, tom lane