pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: John A Meinel <john@arbash-meinel.com>
To: Guido Neitzer <guido.neitzer@pharmaline.de>
Cc: pgsql-performance@postgresql.org
Subject: Re: Prepared statement not using index
Date: Mon, 12 Sep 2005 00:58:34 -0400
Message-ID: <43250AFA.7010900@arbash-meinel.com> (raw)
In-Reply-To: <08E57775-5E88-42B0-926D-21CAFB21CF24@pharmaline.de>
References: <08E57775-5E88-42B0-926D-21CAFB21CF24@pharmaline.de>

Guido Neitzer wrote:
> Hi.
>
> I have an interesting problem with the JDBC drivers. When I use a
> select like this:
>
> "SELECT t0.aktiv, t0.id, t0.ist_teilnehmer, t0.nachname, t0.plz,
> t0.vorname FROM public.dga_dienstleister t0 WHERE t0.plz  like
> ?::varchar(256) ESCAPE '|'" withBindings: 1:"53111"(plz)>
>
> the existing index on the plz column is not used.
>
> When I the same select with a concrete value, the index IS used.
>
> I use PostgreSQL 8.0.3 on Mac OS X and the JDBC driver 8.0-312 JDBC 3.
>
> After a lot of other things, I tried using a 7.4 driver and with  this,
> the index is used in both cases.
>
> Why can this happen? Is there a setting I might have not seen?
> Something I do wrong?
>
> cug

I've had this problem in the past. In my case, the issue was that the
column I was searching had a mixed blend of possible values. For
example, with 1M rows, the number 3 occurred 100 times, but the number
18 occurred 700,000 times.

So when I manually did a search for 3, it naturally realized that it
could use an index scan, because it had the statistics to say it was
very selective. If I manually did a search for 18, it switched to
sequential scan, because it was not very selective (both are the correct
plans).

But if you create a prepared statement, parameterized on this number,
postgres has no way of knowing ahead of time, whether you will be asking
about 3 or 18, so when the query is prepared, it has to be pessimistic,
and avoid worst case behavior, so it choses to always use a sequential scan.

The only way I got around this was with writing a plpgsql function which
used the EXECUTE syntax to dynamically re-plan part of the query.

Hope this makes sense. This may or may not be your problem, without
knowing more about you setup. But the symptoms seem similar.

John
=:->

Attachments:

  signature.asc (application/pgp-signature, 252B) download

view thread (7+ messages)  latest in thread

Message-ID: <43250AFA.7010900@arbash-meinel.com>
Permalink:  ../43250AFA.7010900@arbash-meinel.com/
Also on:    postgresql.org/message-id/43250AFA.7010900@arbash-meinel.com

 · 

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: john@arbash-meinel.com, guido.neitzer@pharmaline.de
  Subject: Re: Prepared statement not using index
  In-Reply-To: <43250AFA.7010900@arbash-meinel.com>

* 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