Why those queries do not utilize indexes? - Mailing list pgsql-performance

From Artimenko Igor
Subject Why those queries do not utilize indexes?
Date
Msg-id 20040827192911.72205.qmail@web11905.mail.yahoo.com
Whole thread Raw
Responses Re: Why those queries do not utilize indexes?  (Dennis Bjorklund <db@zigo.dhs.org>)
Re: Why those queries do not utilize indexes?  (Christopher Kings-Lynne <chriskl@familyhealth.com.au>)
List pgsql-performance
Hi everybody!

Here is my queries:

1. explain SELECT * FROM messageinfo WHERE user_id = CAST( 20000 AS BIGINT ) and msgstatus = CAST(
0 AS smallint );

2. explain SELECT * FROM messageinfo WHERE messageinfo.user_id = 20000::int8 and msgstatus =
0::smallint;

In both cases Explain command shows:
1. Sequential search and very high cost if set enable_seqscan to on;
Seq scan on messageinfo ( cost=0.00..24371.30, rows =36802 )

2. Index scan but even bigger cost if set enable_seqscan to off;
Index �messagesStatus� on messageinfo ( Cost=0.00..27220.72, rows=36802 )

messageinfo table has 200 records which meet this criteria and 662420 in total:

CREATE TABLE messageinfo
(
  user_id int8 NOT NULL,
  msgstatus int2 NOT NULL DEFAULT (0)::smallint,
  receivedtime timestamp NOT NULL DEFAULT now(),
  �
  msgread bool DEFAULT false,
  �
  CONSTRAINT "$1" FOREIGN KEY (user_id) REFERENCES users (id) ON UPDATE CASCADE ON DELETE CASCADE,
  )
WITH OIDS;

CREATE INDEX msgstatus
  ON messageinfo
  USING btree
  (user_id, msgstatus);

CREATE INDEX "messagesStatus"
  ON messageinfo
  USING btree
  (msgstatus);

CREATE INDEX msgread
  ON messageinfo
  USING btree
  (user_id, msgread);

CREATE INDEX "receivedTime"
  ON messageinfo
  USING btree
  (receivedtime);


MY QUESTIONS ARE:

1.    Should I afraid of high cost indexes? Or query will still be very efficient?

2.    Postgres does not use the index I need. For my data sets it�s always msgstatus index is
narrowest compare with �messagesStatus� one. Is any way to �enforce� to use a particular index?
What�s the logic when Postgres chooses one index compare with the other.

3.    I can change db structure to utilize Postgres specifics if you can tell them to me.

4.    Also, originally I had �messagesStatus� index having 2 components ( �msgstatus�, �user_id� ).
But query SELECT * FROM messageinfo WHERE msgstatus = 0 did not utilize indexes in this case. It
only worked if both index components are in WHERE part. So I have to remove 2-nd component
�user_id� from messagesStatus index even I wanted it. Is any way that where clause has only 1-st
component but index is utilized?

Igor Artimenko



__________________________________
Do you Yahoo!?
Yahoo! Mail Address AutoComplete - You start. We finish.
http://promotions.yahoo.com/new_mail

pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: Equivalent praxis to CLUSTERED INDEX?
Next
From: Greg Stark
Date:
Subject: Re: Equivalent praxis to CLUSTERED INDEX?