Re: [PGSQL v8.2.5] Similar queries behave differently

"Gregory Stark" <stark(at)enterprisedb(dot)com> writes:

> "Reg Me Please" <regmeplease(at)gmail(dot)com> writes:
>
>> -> Seq Scan on tt_elem (cost=0.00..29.40 rows=1940 width=8)
>> (actual time=0.012..0.013 rows=1 loops=1)
>
> The discrepancy etween the estimated rows and actual rows makes me think
> you've not analyzed this table in a long time. It's probably best to analyze
> the whole database to have a consistent set of statistics and to catch any
> other old table stats.
>
> There could be other misestimations based due to Postgres limitations but
> first fix the out of date stats and re-post both plans.

Actually it's pretty clear there are some other bad estimations as well. You
should send along the view definition too.

And I would recommend you try it with a normal JOIN ON/USING instead of the
NATURAL JOIN. It's possible it's joining on some unexpected columns -- though
that doesn't really look like it's the case here.

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com