On Fri, Jun 10, 2005 at 12:27:32 +0200,
Dawid Kuroczko <qnex42@gmail.com> wrote:
>
> In other words -- doing such a one row select means scanning the
> whole partitioned data. the primaryindex on logid is not used
> anywhere. Creating view (SELECT * UNION ALL SELECT * UNION ALL....)
> does not help either.
>
> I wonder -- is there any better solution? selecting highest from each
> table, unioning it and then selecting highest of the highest works
> well but it strikes me as 'not exactly the niciest implementation'
> (one of the reasons is that I have to remembed about it each time
> new table will be added to schema...).
>
> Any comments, hints?
I seem to remember people getting performance by doing union alls of selects
that explicity list the partition condition; as that will let the planner
speed up checks of tables that cannot result in matches.
