Re: choosing use an index or not - Mailing list pgsql-hackers

From Martijn van Oosterhout
Subject Re: choosing use an index or not
Date
Msg-id 20061214125614.GB24370@svana.org
Whole thread Raw
In response to choosing use an index or not  ("Felipe Rondon Rocha" <frondon@globo.com>)
List pgsql-hackers
On Thu, Dec 14, 2006 at 10:06:45AM -0300, Felipe Rondon Rocha wrote:
> Hi everyone,
>
> Do you know in which part of the file analyze.c, PostgreSQL decides
> to use or not an index? I´m making a test and I can´t trace the
> moment that he makes the choice.

It doesn't explicitly make the choice anywhere. The planner makes a set
of "paths" that can be used. So for each table a "path" is made for a
sequential scan and one or more "paths" for each index. For each a cost
is calcualted and whichever has the lowest cost is the one chosen.

It's quite dynamic, the choice also depends on what happens to the
result. There's no point using an index if the result is in the wrong
order for the join, for example.

Have a nice day,
--
Martijn van Oosterhout   <kleptog@svana.org>   http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.

pgsql-hackers by date:

Previous
From: Peter Eisentraut
Date:
Subject: Re: recovery.conf parsing problems
Next
From: Martijn van Oosterhout
Date:
Subject: Re: Operator class group proposal