| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Peter Geoghegan <pg(at)heroku(dot)com> |
| Cc: | Magnus Hagander <magnus(at)hagander(dot)net>, Noah Misch <noah(at)leadboat(dot)com>, pgsql-bugs <pgsql-bugs(at)postgresql(dot)org>, Marc-Olaf Jaschke <marc-olaf(dot)jaschke(at)s24(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com> |
| Subject: | Re: Re: Missing rows with index scan when collation is not "C" (PostgreSQL 9.5) |
| Date: | 2016-03-23 18:13:56 |
| Message-ID: | 12340.1458756836@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-bugs pgsql-hackers |
Peter Geoghegan <pg(at)heroku(dot)com> writes:
> What about the existing caller of strxfrm(), convert_string_datum()?
convert_string_datum is, and always has been, used only for planner
estimation purposes. We do not care if it sometimes gets inaccurate
answers. Even if it's as wrong as it can possibly be, that will only
affect planner estimates to the extent of wrongly interpolating between
the endpoints of a histogram bin, so that the effects are no worse than
about 1/statistics_target. And there are bigger limitations on the
accuracy of those estimates anyway, notably that we use the same stats
regardless of the collation that applies to a particular WHERE-clause
operator.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Peter Geoghegan | 2016-03-23 18:14:35 | Re: Re: Missing rows with index scan when collation is not "C" (PostgreSQL 9.5) |
| Previous Message | Magnus Hagander | 2016-03-23 18:09:53 | Re: Re: Missing rows with index scan when collation is not "C" (PostgreSQL 9.5) |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Alvaro Herrera | 2016-03-23 18:13:59 | Re: Rationalizing code-sharing among src/bin/ directories |
| Previous Message | Magnus Hagander | 2016-03-23 18:09:53 | Re: Re: Missing rows with index scan when collation is not "C" (PostgreSQL 9.5) |