On Mon, Jan 20, 2014 at 2:45 PM, Florian Pflug <fgp@phlo.org> wrote:
* EXPLAIN VERBOSE ANALYZE now shows the max. number of forward aggregate transitions per row and aggregate. It's a bit imprecise, because it doesn't track the count per aggregate, but it's still a good metric for how well the inverse transition functions work. If the number is close to one, you know that very few rescans are happening.
I've not looked at this yet and I don't think I'll have time tonight, but it sounds interesting. I guess it might be quite nice to have a way to see this especially with the way the numeric stuff works, it might be actually pretty hard to otherwise know how many inverse transition "failures" there had been. Do you think it's also worth tracking the inverse transition failures too?
I've merged this patch but I attempted to get it into a bit more of a ready state by moving the code out into a helper function the same way as the other explain stuff is done. I've not touched explain before so do let me know if I've made it worse.