Jeff Janes writes:
> I have a weird case where the planner doesn't choose the plan that it
> itself believes to be the fastest plan.
I poked into this a bit and found where things are going wrong. The
ideal plan for this case involves a nestloop with a parameterized inner
scan over an inheritanc
On Sat, Jul 6, 2013 at 10:52 AM, Pavel Stehule wrote:
> Hello
>
> pls, can you send EXPLAIN ANALYZE result?
Sure. I've put them on depesz. (I don't know how useful it will be,
as I'm worried about the internal inconsistency rather than the actual
run time):
enable_seqscan=on
http://explain.dep
Hello
pls, can you send EXPLAIN ANALYZE result?
Regards
Pavel Stehule
2013/7/6 Jeff Janes :
> I have a weird case where the planner doesn't choose the plan that it
> itself believes to be the fastest plan. If I disable seqscan, it then
> chooses a faster plan (faster in estimate and faster in
I have a weird case where the planner doesn't choose the plan that it
itself believes to be the fastest plan. If I disable seqscan, it then
chooses a faster plan (faster in estimate and faster in reality) than
the one chosen when all options were open to it. I can't figure out
how this can be any
