From: | Gregory Stark <stark(at)enterprisedb(dot)com> |
---|---|
To: | "W(dot)Alphonse HAROUNY" <wharouny(at)gmail(dot)com> |
Cc: | <pgsql-sql(at)postgresql(dot)org> |
Subject: | Re: Index usage in bitwise operation context |
Date: | 2007-09-13 22:55:51 |
Message-ID: | 87myvqnpjc.fsf@oxford.xeocode.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-sql |
"W.Alphonse HAROUNY" <wharouny(at)gmail(dot)com> writes:
> Question:
> --------------
> I have an SQL request similar to:
>
> SELECT ..... FROM TBL1, TBL2 WHERE
> <inner join between TBL1 and TBL2 is True> AND
> TBL1.CATEGORY & TBL2.CATEGORY <> 0 //-- where & is the AND bitwise
> operator
>
> Qst:
> 1/ IS the above SQL request will use the INDEX [categoryGroup] defined on
> TBL1 and TBL2 ?
No, & isn't an indexable operator for any of the standard indexing methods.
You could create 32 partial indexes on some other key with clauses like
WHERE tbl1.category & 0100... = 0100...
But I don't think that would be useful for a join clause in any case.
Second idea, you could create an expression index on
tbl1 (category & 0100... = 0100...,
category & 0010... = 0010...,
category & 0001... = 0001...,
...)
Again I don't see that it's going to be used for a join condition.
Lastly, you could look for a GIST index method for varbit which would be
superior to both of the above tactics. I'm still not sure it would be able to
handle a join clause though, but maybe?
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
From | Date | Subject | |
---|---|---|---|
Next Message | Stanislas de Larocque | 2007-09-14 08:31:39 | Optimize querry sql |
Previous Message | W.Alphonse HAROUNY | 2007-09-13 12:34:32 | Index usage in bitwise operation context |