From: Tom Lane <tgl@sss.pgh.pa.us>
To: kimaidou <kimaidou@gmail.com>
cc: pgsql-performance@lists.postgresql.org
Subject: Re: Simple JOIN on heavy table not using expected index
In-reply-to: 
 <CAMKXKO4tUH=nTZ-g09=wng+wSxtkj7K8yePMkfTiOpW4zT4Jdg@mail.gmail.com>
References: 
 <CAMKXKO4tUH=nTZ-g09=wng+wSxtkj7K8yePMkfTiOpW4zT4Jdg@mail.gmail.com>
Comments: In-reply-to kimaidou <kimaidou@gmail.com>
	message dated "Fri, 09 Feb 2024 15:14:31 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1394632.1707491575.1@sss.pgh.pa.us>
Date: Fri, 09 Feb 2024 10:12:55 -0500
Message-ID: <1394633.1707491575@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1394633.1707491575%40sss.pgh.pa.us>
Precedence: bulk

kimaidou <kimaidou@gmail.com> writes:
> It seems PostgreSQL does not use the index parcelles_dep_idx on "dep" (text
> field), even if the corresponding number of lines for this WHERE clause is
> a smal subset of the entire data:
> approx 6M against 80M in total

6M out of 80M rows is not a "small subset".  Typically I'd expect
the planner to use an index-based scan for up to 1 or 2 percent of
the table.  Beyond that, you're going to be touching most pages
of the table anyway.

You can try reducing random_page_cost to favor indexscans, but
you might not find that the query gets any faster.

			regards, tom lane