Thread: Why is my index not used

I have a table, currently at about 150.000 rows, with a btree index on a
field named 'misscnt'. ~138k of the rows have this fields set as null, ~12k
as 0, ~350 as 1, ~200 as 2 and ~150 as 3 (these are the only values used).

Still, I get the following:

# explain select * from cam where misscnt>=1;
NOTICE:  QUERY PLAN:

Seq Scan on cam  (cost=0.00..3609.59 rows=46896 width=66)

Why do postgre think it will get over 46k rows from this query? If I try a
VACUUM ANALYZE on the table, I get:

# explain select * from cam where misscnt>=1;
NOTICE:  QUERY PLAN:

Seq Scan on cam  (cost=0.00..3874.01 rows=50347 width=66)

So now it actually thinks it will get _more_ rows:

# select count(*) from cam where misscnt>=1;count 
-------  692

If I use "set enable_seqscan=false;" and try "select * from cam where
misscnt>=1;" the index is used, and the query executes quite a bit faster.
However, shouldn't Postgre be able to do this automatically?

This is using PostgreSQL 7.1.2.


Thanks.

---
Kristian Eide

try to cast your argument, for example, if misscnt is of type bigint, try:

explain select * from cam where misscnt>=1::int8;


Best regards,
Chris

At 19:54 +0100 02/18/2002, Kristian Eide wrote:
>I have a table, currently at about 150.000 rows, with a btree index on a
>field named 'misscnt'. ~138k of the rows have this fields set as null, ~12k
>as 0, ~350 as 1, ~200 as 2 and ~150 as 3 (these are the only values used).
>
>Still, I get the following:
>
># explain select * from cam where misscnt>=1;
>NOTICE:  QUERY PLAN:
>
>Seq Scan on cam  (cost=0.00..3609.59 rows=46896 width=66)
>
>Why do postgre think it will get over 46k rows from this query? If I try a
>VACUUM ANALYZE on the table, I get:
>
># explain select * from cam where misscnt>=1;
>NOTICE:  QUERY PLAN:
>
>Seq Scan on cam  (cost=0.00..3874.01 rows=50347 width=66)
>
>So now it actually thinks it will get _more_ rows:
>
># select count(*) from cam where misscnt>=1;
>  count
>-------
>    692
>
>If I use "set enable_seqscan=false;" and try "select * from cam where
>misscnt>=1;" the index is used, and the query executes quite a bit faster.
>However, shouldn't Postgre be able to do this automatically?
>
>This is using PostgreSQL 7.1.2.
>
>
>Thanks.
>
>---
>Kristian Eide
>
>
>
>---------------------------(end of broadcast)---------------------------
>TIP 2: you can get off all lists at once with the unregister command
>     (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)


-- 
Chris Ruprecht
Network grunt and bit pusher extraordinaíre

_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

"Kristian Eide" <kreide@online.no> writes:
> Why do postgre think it will get over 46k rows from this query?

Not sure (and you didn't give the pg_statistic data that might tell us
why).  But it's no longer a very interesting question; we will not be
fixing 7.1 optimizer bugs anymore.  Please try it in 7.2 and see if you
get better results.
        regards, tom lane