Re: Functionscan estimates

Alvaro, Michael,

> > About a month ago I mentioned that I'd find that useful.  In a
> > followup, Christopher Kings-Lynne brought up the idea of a GUC
> > variable that could give hints about the expected row count.
>
> That seems pretty limited ... what happens if the query contains more
> that one SRF?

Yeah, I'd see that as a pretty bad idea too. I don't want to tell the planner
how many rows I expect "all functions" to return, I want to tell it how many
*one particular* function will return.

> Maybe issuing some sort of special call to the function (say, with
> some boolean in the call info struct) on which it returns planning data;
> thus the planner can call the function itself.  The hard part would be
> figuring out how to do it without breaking backwards compatibility with
> functions that don't know how to handle that.  (And how to do it in
> plpgsql).

Or in pl/perl, or pl/python, or plsh .... doesn't sound feasable.

My solution would be a lot simpler, since we could simply populate
pg_proc.proestrows with "1000" by default if not changed by the DBA. In an
even better world, we could tie it to a table, saying that, for example,
proestrows = my_table*0.02.

--
--Josh

Josh Berkus
Aglio Database Solutions
San Francisco