From: "Jan de Visser" <jdevisser@digitalfairway.com>
To: pgsql-performance@postgresql.org
Subject: Re: Beginner Question
Date: Mon, 9 Apr 2007 07:51:53 -0400
User-Agent: KMail/1.9.6
Cc: "s d" <s.d.sauron@gmail.com>
References: <24b53fa00704090209pcdd2c6cof487d6388c1743b8@mail.gmail.com>
In-Reply-To: <24b53fa00704090209pcdd2c6cof487d6388c1743b8@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
Message-Id: <200704090751.53347.jdevisser@digitalfairway.com>

On Monday 09 April 2007 05:09:53 s d wrote:
> Hi,
> I am trying to figure out how to debug a performance problem / use psql
> explain. The table in question is:
> # \d word_association;
>            Table "public.word_association"
>  Column |          Type          |     Modifiers
> --------+------------------------+--------------------
>  word1  | character varying(128) | not null
>  word2  | character varying(128) | not null
>  count  | integer                | not null default 0
> Indexes:
>     "word1_word2_comb_unique" unique, btree (word1, word2)
>     "word1_hash_index" hash (word1)
>     "word2_hash_index" hash (word2)
>     "word_association_count_index" btree (count)
>     "word_association_index1_1" btree (word1)
>     "word_association_index2_1" btree (word2)
>
> It has multiple indices since i wanted to see which one the planner chose=
s.
>
>
> # explain select * FROM word_association WHERE (word1 =3D 'bdss' OR
> word2 =3D 'bdss')  AND count >=3D 10;
>                                            QUERY PLAN
> -------------------------------------------------------------------------=
=2D-
>--------------------- Bitmap Heap Scan on word_association=20
> (cost=3D11.53..1192.09 rows=3D155 width=3D22) Recheck Cond: (((word1)::te=
xt =3D
> 'bdss'::text) OR ((word2)::text =3D 'bdss'::text))
>    Filter: (count >=3D 10)
>    ->  BitmapOr  (cost=3D11.53..11.53 rows=3D364 width=3D0)
>          ->  Bitmap Index Scan on word_association_index1_1
> (cost=3D0.00..5.79 rows=3D190 width=3D0)
>                Index Cond: ((word1)::text =3D 'bdss'::text)
>          ->  Bitmap Index Scan on word_association_index2_1
> (cost=3D0.00..5.67 rows=3D174 width=3D0)
>                Index Cond: ((word2)::text =3D 'bdss'::text)
> (8 rows)
>
> The questions:
> 1. i can undestand where the cost=3D11.53 came from but where did the
> 1192.09 come form? The values are in milli right ?
> 2. the query takes  in reality much longer than 1 second.
>
> In short, it feels like something is very wrong here (i tried vacuum
> analyze and it didn't do much diff).
> any ideas ?

You need an index on (word1, word2, count). In your current setup it will h=
ave=20
to scan all rows that satisfy word1 and word2 to see if count >=3D 10.

jan


=2D-=20
=2D-------------------------------------------------------------
Jan de Visser =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 jdevisser@digitalfair=
way.com

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 Baruk Khazad! Khazad ai-menu!
=2D-------------------------------------------------------------