| From: | mlw <markw(at)mohawksoft(dot)com> |
|---|---|
| To: | Thomas Lockhart <thomas(at)fourpalms(dot)org> |
| Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us>, Louis-David Mitterrand <vindex(at)apartia(dot)org>, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Index Scans become Seq Scans after VACUUM ANALYSE |
| Date: | 2002-04-17 14:31:21 |
| Message-ID: | 3CBD8739.6357196C@mohawksoft.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Thomas Lockhart wrote:
> Systems which have optimizing planners can *never* be guaranteed to
> generate the actual lowest-cost query plan. Any impression that Oracle,
> for example, actually does do that may come from a lack of visibility
> into the process, and a lack of forum for discussing these edge cases.
And here in lies the crux of the problem. It isn't a purely logical/numerical
formula. It is a probability estimate, nothing more. Currently, the statistics
are used to calculate a probable best query, not a guaranteed best query. The
presence of an index should be factored into the probability of a best query,
should it not?
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Christopher Kings-Lynne | 2002-04-17 14:31:30 | Re: problem with anoncvs? |
| Previous Message | Christopher Kings-Lynne | 2002-04-17 14:29:12 | Re: [SQL] 16 parameter limit |