pg.ddx.io pgsql-sql@postgresql.org mailing list archive
help / color / mirror / Atom feedmany 'OR' in WHERE-condition
5+ messages / 2 participants
[nested] [flat]
* many 'OR' in WHERE-condition
@ 2007-04-20 05:13 A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:03 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
0 siblings, 1 reply; 5+ messages in thread
From: A. Kretschmer @ 2007-04-20 05:13 UTC (permalink / raw)
To: pgsql-sql
Hello,
a friend of mine has got a problem:
in a SELECT with a condition 'WHERE foo IN (1,2,3,4,...)' with a large
list within. The optimizer choose a rewrite to many OR's and a bitmap
index scan on the index, no problem, fast.
But if the list increase the planner choose an seq-scan. The amount of
values in the list is about 200 and the table has about 250.000 rows.
How can we avoid the seq-scan?
Andreas, thx in advance.
--
Andreas Kretschmer
Kontakt: Heynitz: 035242/47150, D1: 0160/7141639 (mehr: -> Header)
GnuPG-ID: 0x3FFF606C, privat 0x7F4584DA http://wwwkeys.de.pgp.net
^ permalink raw reply [nested|flat] 5+ messages in thread
* Re: many 'OR' in WHERE-condition
2007-04-20 05:13 many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
@ 2007-04-20 11:03 ` Richard Huxton <dev@archonet.com>
2007-04-20 11:29 ` Re: many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
0 siblings, 1 reply; 5+ messages in thread
From: Richard Huxton @ 2007-04-20 11:03 UTC (permalink / raw)
To: A. Kretschmer <andreas.kretschmer@schollglas.com>; +Cc: pgsql-sql
A. Kretschmer wrote:
> in a SELECT with a condition 'WHERE foo IN (1,2,3,4,...)' with a large
...
> But if the list increase the planner choose an seq-scan. The amount of
> values in the list is about 200 and the table has about 250.000 rows.
>
> How can we avoid the seq-scan?
Are you sure you'd want to? What plan have you found to be faster with
200 tests?
--
Richard Huxton
Archonet Ltd
^ permalink raw reply [nested|flat] 5+ messages in thread
* Re: many 'OR' in WHERE-condition
2007-04-20 05:13 many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:03 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
@ 2007-04-20 11:29 ` A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:59 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
0 siblings, 1 reply; 5+ messages in thread
From: A. Kretschmer @ 2007-04-20 11:29 UTC (permalink / raw)
To: pgsql-sql
am Fri, dem 20.04.2007, um 12:03:44 +0100 mailte Richard Huxton folgendes:
> A. Kretschmer wrote:
> >in a SELECT with a condition 'WHERE foo IN (1,2,3,4,...)' with a large
> ...
> >But if the list increase the planner choose an seq-scan. The amount of
> >values in the list is about 200 and the table has about 250.000 rows.
> >
> >How can we avoid the seq-scan?
>
> Are you sure you'd want to? What plan have you found to be faster with
> 200 tests?
It's not my problem - i found it in a forum (german)
http://pg-forum.de/showthread.php?t=1332
If the list contains up to 195 entrys -> up to 195 bitmap index scans
and BitmapOr -> 7.839ms
If the list contains 196 entrys -> seq-scan -> 5591.567ms
Andreas
--
Andreas Kretschmer
Kontakt: Heynitz: 035242/47150, D1: 0160/7141639 (mehr: -> Header)
GnuPG-ID: 0x3FFF606C, privat 0x7F4584DA http://wwwkeys.de.pgp.net
^ permalink raw reply [nested|flat] 5+ messages in thread
* Re: many 'OR' in WHERE-condition
2007-04-20 05:13 many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:03 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
2007-04-20 11:29 ` Re: many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
@ 2007-04-20 11:59 ` Richard Huxton <dev@archonet.com>
2007-04-20 12:20 ` Re: many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
0 siblings, 1 reply; 5+ messages in thread
From: Richard Huxton @ 2007-04-20 11:59 UTC (permalink / raw)
To: A. Kretschmer <andreas.kretschmer@schollglas.com>; +Cc: pgsql-sql
A. Kretschmer wrote:
>
> It's not my problem - i found it in a forum (german)
> http://pg-forum.de/showthread.php?t=1332
>
> If the list contains up to 195 entrys -> up to 195 bitmap index scans
> and BitmapOr -> 7.839ms
>
> If the list contains 196 entrys -> seq-scan -> 5591.567ms
It looks like it's getting the row estimate badly wrong.
"Seq Scan on t_datasets_searchindices_rel (cost=0.00..129053.50
rows=159277 width=4) (actual time=0.213..5590.435 rows=325 loops=1)
Probably worth increasing the statistics estimate for that column (ALTER
TABLE ... ALTER COLUMN ... SET STATISTICS=N)
It also looks like it's getting the cost estimates badly wrong. Probably
worth asking the original poster to read the tuning guide.
--
Richard Huxton
Archonet Ltd
^ permalink raw reply [nested|flat] 5+ messages in thread
* Re: many 'OR' in WHERE-condition
2007-04-20 05:13 many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:03 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
2007-04-20 11:29 ` Re: many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:59 ` Re: many 'OR' in WHERE-condition Richard Huxton <dev@archonet.com>
@ 2007-04-20 12:20 ` A. Kretschmer <andreas.kretschmer@schollglas.com>
0 siblings, 0 replies; 5+ messages in thread
From: A. Kretschmer @ 2007-04-20 12:20 UTC (permalink / raw)
To: pgsql-sql
am Fri, dem 20.04.2007, um 12:59:01 +0100 mailte Richard Huxton folgendes:
> A. Kretschmer wrote:
> >
> >It's not my problem - i found it in a forum (german)
> >http://pg-forum.de/showthread.php?t=1332
> >
> >If the list contains up to 195 entrys -> up to 195 bitmap index scans
> >and BitmapOr -> 7.839ms
> >
> >If the list contains 196 entrys -> seq-scan -> 5591.567ms
>
> It looks like it's getting the row estimate badly wrong.
>
> "Seq Scan on t_datasets_searchindices_rel (cost=0.00..129053.50
> rows=159277 width=4) (actual time=0.213..5590.435 rows=325 loops=1)
Uh oh, i have overlook this.
>
> Probably worth increasing the statistics estimate for that column (ALTER
> TABLE ... ALTER COLUMN ... SET STATISTICS=N)
Thx for your hints!
Andreas
--
Andreas Kretschmer
Kontakt: Heynitz: 035242/47150, D1: 0160/7141639 (mehr: -> Header)
GnuPG-ID: 0x3FFF606C, privat 0x7F4584DA http://wwwkeys.de.pgp.net
^ permalink raw reply [nested|flat] 5+ messages in thread
end of thread, other threads:[~2007-04-20 12:20 UTC | newest]
Thread overview: 5+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2007-04-20 05:13 many 'OR' in WHERE-condition A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:03 ` Richard Huxton <dev@archonet.com>
2007-04-20 11:29 ` A. Kretschmer <andreas.kretschmer@schollglas.com>
2007-04-20 11:59 ` Richard Huxton <dev@archonet.com>
2007-04-20 12:20 ` A. Kretschmer <andreas.kretschmer@schollglas.com>
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