| From: | Marti Raudsepp <marti(at)juffo(dot)org> |
|---|---|
| To: | ach <alanchines(at)gmail(dot)com> |
| Cc: | "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org> |
| Subject: | Re: statistics target for columns in unique constraint? |
| Date: | 2013-05-13 22:10:06 |
| Message-ID: | CABRT9RBwmDxjq-tzLMh6rFX-suJYVQMiGSQAADsueg07rvB1dg@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
On Mon, May 13, 2013 at 6:01 PM, ach <alanchines(at)gmail(dot)com> wrote:
> what I'm wondering is, since
> the unique constraint already covers the whole table and all rows in
> entirety, is it really necessary for statistics to be set that high on
> those?
AFAIK if there are exact-matching unique constraints/indexes for a
query's WHERE clause, the planner will deduce that the query only
returns 1 row and won't consult statistics at all.
> Or does that only serve to slow down inserts to that table?
It doesn't slow down inserts directly. Tables are analyzed in the
background by autovacuum. However, I/O traffic from autovacuum analyze
may slow down inserts running concurrently.
Regards,
Marti
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Mike McCann | 2013-05-13 22:36:03 | Re: Hardware suggestions for maximum read performance |
| Previous Message | Tom Lane | 2013-05-13 20:33:19 | Re: RT3.4 query needed a lot more tuning with 9.2 than it did with 8.1 |