To: Miernik <public@public.miernik.name>
cc: pgsql-performance@postgresql.org
Subject: Re: why query plan for the inner SELECT of WHERE x IN is wrong,
	but when run the inner query alone is OK?
In-reply-to: <20080809203435.7227.0.NOFFLE@turbacz.local> 
References: <20080809203435.7227.0.NOFFLE@turbacz.local>
Comments: In-reply-to Miernik <public@public.miernik.name>
	message dated "Sat, 09 Aug 2008 22:34:35 +0200"
Date: Sat, 09 Aug 2008 16:57:58 -0400
Message-ID: <11224.1218315478@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Miernik <public@public.miernik.name> writes:
> miernik=> EXPLAIN UPDATE cnts SET p0 = FALSE WHERE uid IN (SELECT uid FROM alog WHERE pid = 3452654 AND o = 1);
>                                           QUERY PLAN
> -----------------------------------------------------------------------------------------------
>  Nested Loop IN Join  (cost=0.00..3317.34 rows=1 width=44)
>    ->  Seq Scan on cnts  (cost=0.00..36.00 rows=2000 width=44)
>    ->  Index Scan using alog_uid_idx on alog  (cost=0.00..296.95 rows=1 width=4)
>          Index Cond: ((alog.uid)::integer = (cnts.uid)::integer)
>          Filter: ((alog.pid = 3452654::numeric) AND (alog.o = 1::numeric))
> (5 rows)

> But if I give him only the inner part, it makes reasonable assumptions
> and runs OK:

What's the results for

explain select * from cnts, alog where alog.uid = cnts.uid

?

If necessary, turn off enable_hashjoin and enable_mergejoin so we can
see a comparable plan.  I'm suspecting it thinks the condition on
uid is more selective than the one on the other index.

			regards, tom lane