pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Frank van Vugt <ftm.van.vugt@foxi.nl>
To: pgsql-performance@postgresql.org
Subject: why is a constraint not 'pushed down' into a subselect when this subselect is using a 'group by' ??
Date: Tue, 15 Jun 2004 08:49:29 +0200
Message-ID: <200406150849.29470.ftm.van.vugt@foxi.nl> (raw)

Hi all,

I noticed the following. Given two tables, just simply articles and their 
packages:

	article(id int)
	package( id int, article_id int, amount)

When taking the minimum package for articles given some constraint on the 
article table

	select 
		article.id, 
		p_min 
	from 
		article, 
		(select 
			article_id, 
			min(amount) as p_min 
		from 
			package 
		group by 
			article_id
		) as foo 
	where 
		article.id = foo.article_id and 
		article.id < 50;

The query plan shows

                                       QUERY PLAN
----------------------------------------------------------------------------------------
 Hash Join  (cost=730.11..808.02 rows=1 width=8)
   Hash Cond: ("outer".article_id = "inner".id)
   ->  Subquery Scan foo  (cost=726.10..781.74 rows=4451 width=8)
         ->  HashAggregate  (cost=726.10..737.23 rows=4451 width=8)
               ->  Seq Scan on package  (cost=0.00..635.40 rows=18140 width=8)
   ->  Hash  (cost=4.01..4.01 rows=1 width=4)
         ->  Index Scan using article_pkey on article  (cost=0.00..4.01 rows=1 
width=4)
               Index Cond: (id < 50)

So, a seq scan on the complete package table is done, although it seems the 
planner could have deduced that only the part that has 'article_id < 50' 
would be relevant, since 'article.id < 50' and 'article.id = foo.article_id'

If I ditch the group by, then this contraint does get pushed into the 
subselect :

	select 
		article.id, 
		p_min 
	from 
		article, 
		(select 
			article_id, 
			amount as p_min 
		from 
			package
		) as foo 
	where 
		article.id = foo.article_id and 
		article.id < 50;


                                    QUERY PLAN
-----------------------------------------------------------------------------------
 Nested Loop  (cost=0.00..14.22 rows=5 width=8)
   ->  Index Scan using article_pkey on article  (cost=0.00..4.01 rows=1 
width=4)
         Index Cond: (id < 50)
   ->  Index Scan using package_idx3 on package  (cost=0.00..10.16 rows=4 
width=8)
         Index Cond: ("outer".id = package.article_id)


So it seems the math-knowledge is there ;)
I'm wondering about what I'm overlooking here. When I take the first query and 
add the article constraint 'manually', I get:

	select 
		article.id, 
		p_min 
	from 
		article, 
		(select 
			article_id, 
			min(amount) as p_min 
		from 
			package 
		where 
			article_id < 50 
		group by 
			article_id
		) as foo 
	where 
		article.id = foo.article_id and 
		article.id < 50;

                                          QUERY PLAN
----------------------------------------------------------------------------------------------
 Nested Loop  (cost=0.00..9.65 rows=1 width=8)
   Join Filter: ("outer".id = "inner".article_id)
   ->  Index Scan using article_pkey on article  (cost=0.00..4.01 rows=1 
width=4)
         Index Cond: (id < 50)
   ->  Subquery Scan foo  (cost=0.00..5.62 rows=1 width=8)
         ->  GroupAggregate  (cost=0.00..5.61 rows=1 width=8)
               ->  Index Scan using package_idx3 on package  (cost=0.00..5.60 
rows=2 width=8)
                     Index Cond: (article_id < 50)


which would have been a nice plan for the first query ;)


Could someone shine a light on this, please?



-- 
Best,




Frank.




view thread (5+ messages)  latest in thread

Message-ID: <200406150849.29470.ftm.van.vugt@foxi.nl>
Permalink:  ../200406150849.29470.ftm.van.vugt@foxi.nl/
Also on:    postgresql.org/message-id/200406150849.29470.ftm.van.vugt@foxi.nl

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: ftm.van.vugt@foxi.nl
  Subject: Re: why is a constraint not 'pushed down' into a subselect when this subselect is using a 'group by' ??
  In-Reply-To: <200406150849.29470.ftm.van.vugt@foxi.nl>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox