From: | Ottó Havasvölgyi <havasvolgyi(dot)otto(at)gmail(dot)com> |
---|---|
To: | "Nicolas Barbier" <nicolas(dot)barbier(at)gmail(dot)com> |
Cc: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Eliminating unnecessary left joins |
Date: | 2007-04-21 08:55:07 |
Message-ID: | 34608c0c0704210155h32cf6c56h3411c86c0b5d8737@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
But then what about the null values? Perhaps unique + notnull is better?
Otto
2007/4/20, Nicolas Barbier <nicolas(dot)barbier(at)gmail(dot)com>:
>
> 2007/4/16, Ottó Havasvölgyi <havasvolgyi(dot)otto(at)gmail(dot)com>:
>
> > Eliminate the table T from the query/subquery if the following
> requirements
> > are satisfied:
> > 1. T is left joined
> > 2. T is referenced only in the join expression where it is left joined
> > 3. the left join's join expression is a simple equality expression like
> > T1.C1=T2.C2; T1!=T2 and (T==T1 or T==T2)
> > 4. the column of T in the join exression is the primary key of T
>
> Condition 4 should be: the column of T in the join expression is a key
> of T (i.e. it doesn't need to be the PK, a UNIQUE constraint would be
> enough).
>
> This process can be done recursively (implementation doesn't have to
> be recursive, of course), to eliminate whole sub-trees of the join
> tree.
>
> Nicolas
>
> --
> Nicolas Barbier
> http://www.gnu.org/philosophy/no-word-attachments.html
>
From | Date | Subject | |
---|---|---|---|
Next Message | sharath kumar | 2007-04-21 09:46:42 | functions to obtain query plan |
Previous Message | Zoltan Boszormenyi | 2007-04-21 08:49:46 | Re: [HACKERS] parser dilemma |