On Fri, Jun 13, 2003 at 15:12:39 -0500,
"Jim C. Nasby" <jim@nasby.net> wrote:
> On Fri, Jun 13, 2003 at 10:37:46AM -0500, Bruno Wolff III wrote:
> > On Thu, Jun 12, 2003 at 17:26:11 -0500,
> > "Jim C. Nasby" <jim@nasby.net> wrote:
> > > On Thu, Jun 12, 2003 at 12:08:15PM +0800, Ang Chin Han wrote:
> > >
> > > This becomes really important whenever pgsql gains the ability to use
> > > multiple indexes per table (someone smack me if it can do this now and I
> >
> > SMACK!
>
> You can use multiple indexes per query plan node? IE: if you have
> indexes on a and b, you can use both indexes when you do WHERE a=blah
> and b=foo? That's what I was reffering to...
I think what you are describing doesn't make much sense. On any scan
you are only going to using one index (possibly over several columns).
But while processing a query you might be making use of several indexes
on a table. For example the following query should would fast if
appropiate indexes exist:
select (select col1 from tab order by col1),
(select col2 from tab order by col2)