To: Ludwig Lim <lud_nowhere_man@yahoo.com>
Cc: PostgreSQL Mailing List <pgsql-sql@postgresql.org>
Subject: Re: Question on OUTER JOINS. 
In-reply-to: <20030628061618.19440.qmail@web80510.mail.yahoo.com> 
References: <20030628061618.19440.qmail@web80510.mail.yahoo.com>
Comments: In-reply-to Ludwig Lim <lud_nowhere_man@yahoo.com>
	message dated "Fri, 27 Jun 2003 23:16:18 -0700"
Date: Sat, 28 Jun 2003 11:43:47 -0400
Message-ID: <18085.1056815027@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Ludwig Lim <lud_nowhere_man@yahoo.com> writes:
>    Is there way of rewritting :

>    SELECT  a.status,
>            employee_id
>    FROM   permission a LEFT JOIN 
>           ( SELECT * FROM employee WHERE employee_id
> =5) as b ON (a.status = b.status)
>    WHERE status='test'

> into a query that has no subselect in the FROM clause.

In this particular case you could do

    FROM   permission a LEFT JOIN 
           employee b ON (a.status = b.status AND b.employee_id=5)
    WHERE status='test'

which AFAICS would give the same answers.  As Bruno points out,
you can't move qualification conditions up and down past outer
joins without changing the answers in general.  But the above
change is okay: either way, A rows that don't match to a B row
with employee_id=5 will be emitted with nulls instead.

However, in this particular case I don't see why you're bothering.
If the sub-SELECT can be flattened, the planner will generally do
it for you.  I'd expect the first form of the query to give the
same plan (in pre-7.4 releases, maybe even a better plan) as the
second.

Perhaps you're showing us an oversimplified version of your
real problem?

			regards, tom lane