On Thu, Sep 24, 2009 at 8:22 AM, Jared Beck <jared@singlebrook.com> wrote:
> Should we try to improve statistics collection for that column
> (variableid) by using ALTER TABLE ... ALTER COLUMN ... SET STATISTICS?
It's worth a try, but I'm not sure it's going to help much. The LIKE
condition is hard for the planner to estimate at present.
> In other words, if the row estimate were perfect would we be likely
> to get a better plan? Or is that impossible to speculate on?
Good row estimates are the key to happiness, but I don't know whether
it will actually change the plan in this instance.
> Thanks again. Already you've been a big help. We love postgres and
> are very happy with our upgrade to 8.4 so far!
Glad to hear it.
...Robert