| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> |
| Cc: | Zeugswetter Andreas SB <ZeugswetterA(at)wien(dot)spardat(dot)at>, "'hackers(at)postgresql(dot)org'" <hackers(at)postgreSQL(dot)org> |
| Subject: | Re: AW: question about index cost estimates |
| Date: | 2000-05-18 16:29:03 |
| Message-ID: | 24065.958667343@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
> So how about a kind of sample scanning in a separate ANALYZE
> command ?
I do think it'd be a good idea to separate out the stats-gathering
into an ANALYZE command that can be invoked separately from VACUUM.
For one thing, there's no good reason to hold down an exclusive
lock on the target table while we are gathering stats.
But that doesn't answer the question: how can we measure the extent
to which a table is in the same order as an index? And even that
is too one-dimensional a concept to apply to r-tree indexes, for
example. What do we do for r-trees?
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Alessio Bragadini | 2000-05-18 16:33:43 | The New Slashdot Setup (includes MySql server) |
| Previous Message | Hiroshi Inoue | 2000-05-18 16:15:28 | RE: AW: question about index cost estimates |