pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Federico Simonetti (Liveye) <federico@liveye.net>
Cc: pgsql-performance@postgresql.org
Subject: Re: Query performance on ILIKE with AND operator...
Date: Mon, 10 Oct 2005 10:58:15 -0400
Message-ID: <20804.1128956295@sss.pgh.pa.us> (raw)
In-Reply-To: <434A6C2D.9070207@liveye.net>
References: <434A6C2D.9070207@liveye.net>

"Federico Simonetti (Liveye)" <federico@liveye.net> writes:
> I'm encountering a quite strange performance problem.

The problem stems from the horrid misestimation of the number of rows
fetched from detail0009:

> "                    ->  Seq Scan on detail0009  (cost=0.00..20500.11
> rows=26 width=1005) (actual time=453.000..5983.000 rows=53588 loops=1)"
> "                          Filter: ((txcontenttype ~~* '%html%'::text)
> AND ((vchost)::text ~~* '%www.%'::text))"

When the planner is off by a factor of two thousand about the number of
rows involved, it's not very likely to produce a good plan :-(

In the OR case the rowcount estimate is 6334, which is somewhat closer
to reality (only about a factor of 10 off, looks like), and that changes
the plan to something that works acceptably well.

Assuming that this is web-log data, the prevalence of www and html
together is hardly surprising, but PG's statistical mechanisms will
never realize it.  Not sure about a good workaround.  Does it make
sense to combine the two conditions into one?
	(vchost || txcontenttype) ilike '%www.%html%'

			regards, tom lane



view thread (3+ messages)  latest in thread

Message-ID: <20804.1128956295@sss.pgh.pa.us>
Permalink:  ../20804.1128956295@sss.pgh.pa.us/
Also on:    postgresql.org/message-id/20804.1128956295@sss.pgh.pa.us

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: tgl@sss.pgh.pa.us, federico@liveye.net
  Subject: Re: Query performance on ILIKE with AND operator...
  In-Reply-To: <20804.1128956295@sss.pgh.pa.us>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox