pg.ddx.io  pgsql-admin@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Kris Kiger <kris@musicrebellion.com>
To: pgsql-admin@postgresql.org
Subject: Re: tsearch2 poor performance
Date: Mon, 27 Sep 2004 13:56:36 -0500
Message-ID: <41586264.5020302@musicrebellion.com> (raw)
References: <71E37EF6B7DCC1499CEA0316A2568328DC9DF0@loki.wc.globexplorer.net>
	<4158409E.8060109@musicrebellion.com>
	<Pine.GSO.4.58.0409272040090.14980@ra.sai.msu.su>
	<41584E49.5040301@musicrebellion.com>
	<Pine.GSO.4.58.0409272243140.14980@ra.sai.msu.su>

Yes, it is much better than no index of sequential scan.  We may just be 
looking at the best performance tsearch2 can offer on my machine.  

search_test=#  explain analyze SELECT count(q) FROM product, 
to_tsquery('oil') AS q  WHERE vector @@ q;
                                                            QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=67847264.50..67847264.50 rows=1 width=32) (actual 
time=83311.552..83311.555 rows=1 loops=1)
   ->  Nested Loop  (cost=12.50..67839764.50 rows=3000001 width=32) 
(actual time=0.204..81960.198 rows=226357 loops=1)
         Join Filter: ("outer".vector @@ "inner".q)
         ->  Seq Scan on product  (cost=0.00..339752.00 rows=3000000 
width=32) (actual time=0.100..27415.795 rows=3000000 loops=1)
         ->  Materialize  (cost=12.50..22.50 rows=1000 width=32) (actual 
time=0.003..0.006 rows=1 loops=3000000)
               ->  Function Scan on q  (cost=0.00..12.50 rows=1000 
width=32) (actual time=0.020..0.024 rows=1 loops=1)
 Total runtime: 83311.735 ms
(7 rows)

search_test=# explain analyze select count(*) from product where 
description like '% oil %';
                                                       QUERY PLAN
------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=347264.01..347264.01 rows=1 width=0) (actual 
time=39858.350..39858.353 rows=1 loops=1)
   ->  Seq Scan on product  (cost=0.00..347252.00 rows=4801 width=0) 
(actual time=0.100..38320.293 rows=226357 loops=1)
         Filter: (description ~~ '% oil %'::text)
 Total runtime: 39858.491 ms


>>Oleg,
>>
>>    Thanks for the help on this.
>>
>>    The query I used to return the 508 number is:
>>         SELECT * FROM stat('SELECT vector FROM product') ORDER BY ndoc
>>desc, word ;
>>
>>    Testing says, the more words I use, the faster the query is.  My
>>original search word, 'oil', appears in 226,357 documents 233,266 times.
>> As far as distinct words go, 'oil' is middle of the road for
>>occurences.  As it is set up now, the best search time I am getting on
>>this single word is roughly 22 seconds.
>>    
>>
>
>Does this time (22 seconds) is still better than seq. scan (no index)
>or standard 'LIKE' ?
>
>  
>





view thread (18+ messages)  latest in thread

Message-ID: <41586264.5020302@musicrebellion.com>
Permalink:  ../41586264.5020302@musicrebellion.com/
Also on:    postgresql.org/message-id/41586264.5020302@musicrebellion.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 from your mail client; the link fills in
  the headers below: reply via email
  Some clients, webmail especially, drop In-Reply-To and References from
  a link, and the reply then starts a new thread: use the mbox below.

  To: pgsql-admin@postgresql.org
  Cc: kris@musicrebellion.com
  Subject: Re: tsearch2 poor performance
  In-Reply-To: <41586264.5020302@musicrebellion.com>
  References: <71E37EF6B7DCC1499CEA0316A2568328DC9DF0@loki.wc.globexplorer.net> <4158409E.8060109@musicrebellion.com> <Pine.GSO.4.58.0409272040090.14980@ra.sai.msu.su> <41584E49.5040301@musicrebellion.com> <Pine.GSO.4.58.0409272243140.14980@ra.sai.msu.su> <41586264.5020302@musicrebellion.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