Re: Ideas about a better API for postgres_fdw remote estimates

On Mon, Jul 6, 2020 at 11:28:28AM -0400, Stephen Frost wrote:
> > Yeah, thinking about it as a function that inspects partial planner
> > results, it might be useful for other purposes besides postgres_fdw.
> > As I said before, I don't think this necessarily has to be bundled as
> > part of postgres_fdw. That still doesn't make it part of EXPLAIN.
>
> Providing it as a function rather than through EXPLAIN does make a bit
> more sense if we're going to skip things like the lookups you mention
> above. I'm still inclined to have it be a part of core rather than
> having it as postgres_fdw though. I'm not completely against it being
> part of postgres_fdw... but I would think that would really be
> appropriate if it's actually using something in postgres_fdw, but if
> everything that it's doing is part of core and nothing related
> specifically to the postgres FDW, then having it as part of core makes
> more sense to me. Also, having it as part of core would make it more
> appropriate for other tools to look at and adding that kind of
> inspection capability for partial planner results could be very
> interesting for tools like pgAdmin and such.

I agree the statistics extraction should probably be part of core.
There is the goal if FDWs returning data, and returning the data
quickly. I think we can require all-new FDW servers to get improved
performance. I am not even clear if we have a full understanding of the
performance characteristics of FDWs yet. I know Tomas did some research
on its DML behavior, but other than that, I haven't seen much.

On a related note, I have wished to be able to see all the costs
associated with plans not chosen, and I think others would like that as
well. Getting multiple costs for a query goes in that direction.

--
Bruce Momjian <bruce(at)momjian(dot)us> https://momjian.us
EnterpriseDB https://enterprisedb.com

The usefulness of a cup is in its emptiness, Bruce Lee