Re: Redundant sub query triggers slow nested loop left join

From: Tom Lane
Subject: Re: Redundant sub query triggers slow nested loop left join
Date: ,
Msg-id: 13922.1177199288@sss.pgh.pa.us
(view: Whole thread, Raw)
In response to: Redundant sub query triggers slow nested loop left join  ("henk de wit")
List: pgsql-performance


"henk de wit" <> writes:
> Naively I would say that a planner
> would have to be smart enough to see this by itself?

We got rid of direct tests for redundant WHERE clauses a long time ago
(in 7.4, according to some quick tests I just made).  They took a lot
of cycles and almost never accomplished anything.

Since you have two redundant tests, the selectivity is being
double-counted, leading to a too-small rows estimate and a not very
appropriate choice of join plan.

FWIW, CVS HEAD does get rid of the duplicate conditions for the common
case of mergejoinable equality operators --- but it's not explicitly
looking for duplicate conditions, rather this is falling out of a new
method for making transitive equality deductions.

            regards, tom lane


pgsql-performance by date:

From: Tom Lane
Date:
Subject: Re: Odd problem with planner choosing seq scan
From: Ulrich Cech
Date:
Subject: Re: Large objetcs performance