Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> writes:
> BTW, should we cost push-down-able quals differently, say discount some
> fraction of the cost, to reflect the fact that they are cheaper to run?
> However, since the decision of which ones to push down depends on the
> cost, and the cost would depend on which ones we push down, it looks
> rather messy.
I don't think our cost model is anywhere near refined enough to make it
worth trying to distinguish that. (Frankly, I'm pretty skeptical of this
entire patch being worth the trouble...)
regards, tom lane