On 11/26/2008 03:20 PM, Scott Marlowe wrote:
> On Tue, Nov 25, 2008 at 10:22 PM, Clemens Schwaighofer
> <clemens.schwaighofer@tequila.jp> wrote:
>> On 11/26/2008 02:15 PM, Scott Marlowe wrote:
>>> Are there more rows in the 8.2 table you're testing on? Or is the
>>> whole table small enough to fit on a few pages?
>> I highly doubt that. I have right now in one of the DBs I transfered
>> tables from ~100.000 down to ~40.000 rows that all join together. I
>> somehow really doubt that fit in a few pages.
>
> Right, with more rows, and choosing fewer, pgsql will go for an index
> scan. if choosing a good %, the seq scan.
okay, then I think I get it. One of my test queries was actually
selecting a very big chunk (90%) of the data, so it makes sense the
planner chooses seq scan over index scan here.
>> So, I am fine when I trust the Postgresql planner :) Because speed wise
>> I see no difference that 8.3 would be slower than 8.2
>
> Well, the planner's not perfect. Some off corner cases can catch it
> out, or if your database isn't analyzed after a lot of changes it may
> make an uninformed decision. But most the time it makes the right, or
> close enough, decision.
Well, I have autovacuum turned on, so this should hopefully keep the
planner up to date.
Anyway, thanks a lot for your help.
--
[ Clemens Schwaighofer -----=====:::::~ ]
[ IT Engineer/Manager ]
[ E-Graphics Communications, TEQUILA\ Japan IT Group ]
[ 6-17-2 Ginza Chuo-ku, Tokyo 104-8167, JAPAN ]
[ Tel: +81-(0)3-3545-7706 Fax: +81-(0)3-3545-7343 ]
[ http://www.tequila.jp ]