| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Matthew Hagerty <mhagerty(at)voyager(dot)net> |
| Cc: | pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Query not using index, please explain. |
| Date: | 2001-03-08 20:19:43 |
| Message-ID: | 29777.984082783@sss.pgh.pa.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Matthew Hagerty <mhagerty(at)voyager(dot)net> writes:
> The query where the time_stamp < '03-01-2000' does not return any rows, the
> 04-01-2000 date does return rows. When I disable seqscan the query is
> almost instant, but with it on, it takes about 3 or 4 minutes. Why can't
> the query planner use the index in the later case?
It *can* (and did, in two of the three examples you gave). It just
doesn't think the indexscan is faster --- note the cost estimates.
Evidently the cost estimates are way off, probably because the estimated
number of selected rows is way off.
Have you done a VACUUM ANALYZE lately? Not that that will help if the
distribution of timestamps is highly irregular :-(. See the many past
discussions of these issues.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Richard Poole | 2001-03-08 20:38:33 | Re: Query not using index, please explain. |
| Previous Message | Tom Lane | 2001-03-08 20:04:22 | Re: WAL does not recover gracefully from out-of-disk-sp ace |