Message-ID: <51E082D8.4030503@consistentstate.com>
Date: Fri, 12 Jul 2013 16:27:36 -0600
From: Brian Fehrle <brianf@consistentstate.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64;
 rv:17.0) Gecko/20130311 Thunderbird/17.0.4
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: Trying to eliminate union and sort
References: <51DEF924.1040809@consistentstate.com>
 <WM!193bd65d81fa7c39e56918332a81c274363f4bb6c8de7151691ea39398f8f9117cf6aec5b0fd2887b97b4b5a921c1aa8!@asav-1.01.com>
 <51DF51DA.7010509@agliodbs.com>
In-Reply-To: <51DF51DA.7010509@agliodbs.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

On 07/11/2013 06:46 PM, Josh Berkus wrote:
> Brian,
>
>> 3. I'm trying to eliminate the union, however I have two problems.
>> A) I can't figure out how to have an 'or' clause in a single join that
>> would fetch all the correct rows. If I just do:
>> LEFT OUTER JOIN table2 t2 ON (t2.real_id = t.id OR t2.real_id =
>> t.backup_id), I end up with many less rows than the original query. B.
>>
>> I believe the issue with this is a row could have one of three
>> possibilities:
>> * part of the first query but not the second -> results in 1 row after
>> the union
>> * part of the second query but not the first -> results in 1 row after
>> the union
>> * part of the first query and the second -> results in 2 rows after the
>> union (see 'B)' for why)
>>
>> B) the third and fourth column in the SELECT will need to be different
>> depending on what column the row is joined on in the LEFT OUTER JOIN to
>> table2, so I may need some expensive case when logic to filter what is
>> put there based on whether that row came from the first join clause, or
>> the second.
> No, it doesn't:
>
> SELECT t.id,
> 	t.mycolumn1,
> 	table3.otherid as otherid1,
> 	table3a.otherid as otherid2,
> 	t.mycolumn2
> FROM t
> 	LEFT OUTER JOIN table2
> 	   ON ( t.id = t2.real_id OR t.backup_id = t2.real_id )
> 	LEFT OUTER JOIN table3
> 	   ON ( t.typeid = table3.id )
>          LEFT OUTER JOIN table3 as table3a
>             ON ( table2.third_id = table3.id )
> WHERE t.external_id IN ( ... )
> ORDER BY t.mycolumn2, t.id
I tried this originally, however my resulting rowcount is different.

The original query returns 9,955,729 rows
This above one returns 7,213,906

As for the counts on the tables:
table1      3,653,472
table2      2,191,314
table3    25,676,589

I think it's safe to assume right now that any resulting joins are not 
one-to-one

- Brian F


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance