Re: Query planner plans very inefficient plans - Mailing list pgsql-performance

From Tom Lane
Subject Re: Query planner plans very inefficient plans
Date
Msg-id 16644.1056999926@sss.pgh.pa.us
Whole thread Raw
In response to Query planner plans very inefficient plans  ("Robert Wille" <a2om6sy02@sneakemail.com>)
List pgsql-performance
"Robert Wille" <a2om6sy02@sneakemail.com> writes:
> select * from image natural join ancestry where ancestorid=1000000 and
> (state & 7::bigint) = 0::bigint;

The planner is not going to have any statistics that allow it to predict
the number of rows satisfying that &-condition, and so it's unsurprising
if its off-the-cuff guess has little to do with reality.

I'd recommend skipping any cute tricks with bit-packing, and storing the
state (and any other values you query frequently) as its own column, so
that the query looks like

select * from image natural join ancestry where ancestorid=1000000 and
state = 0;

ANALYZE should be able to do a reasonable job with a column that has 8
or fewer distinct values ...

            regards, tom lane

pgsql-performance by date:

Previous
From: "Robert Wille"
Date:
Subject: Query planner plans very inefficient plans
Next
From: Sean Chittenden
Date:
Subject: Re: Query planner plans very inefficient plans