Re: BUG #8598: Row count estimates of partial indexes - Mailing list pgsql-bugs

From Kevin Grittner
Subject Re: BUG #8598: Row count estimates of partial indexes
Date
Msg-id 1384702450.20685.YahooMailNeo@web162902.mail.bf1.yahoo.com
Whole thread Raw
In response to BUG #8598: Row count estimates of partial indexes  (marko@joh.to)
List pgsql-bugs
"marko@joh.to" <marko@joh.to> wrote:=0A=0A>=A0=A0 "index_transactions_trans=
action_balance_details" btree=0A> (transactionid)=0A> WHERE NOT processed A=
ND accountbalancesdailyid IS NOT NULL=0A>=0A> However, querying with the WH=
ERE clause completely ignores the=0A> pg_class.reltuples value for the inde=
x:=0A=0AWhat is the pg_class.reltuples value for the index?=0A=0A> =3D# EXP=
LAIN ANALYZE SELECT * FROM transactions WHERE NOT processed=0A> AND account=
balancesdailyid IS NOT NULL;=0A=0A> Index Scan using index_transactions_tra=
nsaction_balance_details=0A> on transactions=A0 (cost=3D0.00..3883160.47 ro=
ws=3D66259403 width=3D130)=0A> (actual time=3D0.033..18.268 rows=3D13962 lo=
ops=3D1)=0A=0A> .. which makes for some silly joins when this index is part=
 of a=0A> larger query.=0A>=0A> Is this expected on 9.1?=A0 Has this been f=
ixed in more recent=0A> versions?=0A=0APlease provide a little more informa=
tion:=0A=0ASELECT version();=0ASELECT name, current_setting(name), source=
=0A=A0 FROM pg_settings=0A=A0 WHERE source NOT IN ('default', 'override');=
=0A=0AWhat VACUUM or ANALYZE commands are run outside of autovacuum?=0A=0A-=
-=0AKevin Grittner=0AEDB: http://www.enterprisedb.com=0AThe Enterprise Post=
greSQL Company

pgsql-bugs by date:

Previous
From: marko@joh.to
Date:
Subject: BUG #8598: Row count estimates of partial indexes
Next
From: Tom Lane
Date:
Subject: Re: BUG #8598: Row count estimates of partial indexes