Re: A *short* planner question - Mailing list pgsql-admin

From Nick Fankhauser
Subject Re: A *short* planner question
Date
Msg-id NEBBLAAHGLEEPCGOBHDGMEDFEMAA.nickf@ontko.com
Whole thread Raw
In response to Re: A *short* planner question  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: A *short* planner question  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-admin
> Hmm.  If the outer side of the nestloop were actually hitting 2696 rows
> on average, it wouldn't be "lightning fast".  So the planner's failure
> to choose this plan is probably due to this overestimate.

This may be a reasonable estimate- as you might have gathered, we're
matching people to court cases. The typical person (hopefully) only has one
case that they're involved with, but an attorney may be in the 1K-10K range
and a judge in 10k-100K. Since they all live in the same table, the estimate
may not be all that far off. In addition, since we're using a "like", typing
in only an initial char would get roughly 1/26 of the rows back- Does the
planner look at the number of leading chars provided when coming up with a
plan?

> still on 7.1.*, right?  It'd be interesting to know if 7.2 gets
> this right; it has more detailed stats and hopefully would make a better
> estimate of the number of matches.

7.2 on our development box is on the middle-priority TODO list. (2-4 weeks.)
I'll provide feedback when we get there.

As always, thanks for the help!

-Nick

--------------------------------------------------------------------------
Nick Fankhauser  nickf@ontko.com  Phone 1.765.935.4283  Fax 1.765.962.9788
Ray Ontko & Co.     Software Consulting Services     http://www.ontko.com/





pgsql-admin by date:

Previous
From: Tom Lane
Date:
Subject: Re: ALTER TABLE ... SET DEFAULT
Next
From: Tom Lane
Date:
Subject: Re: A *short* planner question