David Rowley <dgrowleyml(at)gmail(dot)com> writes:
> I'm not following why my in method if the planner chooses a poor plan
> is any different from the planner choosing a poor plan for anything
> else because the stats or costs aren't a good reflection of reality.
Well, we hope we'll usually pick the right thing. But upthread
you were complaining that any change might be a regression, or
at least that's how I read it.
regards, tom lane