Re: Avoiding Recheck Cond when using Select Distinct
| От | Jeff Janes |
|---|---|
| Тема | Re: Avoiding Recheck Cond when using Select Distinct |
| Дата | |
| Msg-id | CAMkU=1wo8qSLNut5ktkc4tpisOBVepTk=2BaFd4LmnkBLXh7jA@mail.gmail.com обсуждение исходный текст |
| Ответ на | Re: Avoiding Recheck Cond when using Select Distinct (Maciek Sakrejda <m.sakrejda@gmail.com>) |
| Список | pgsql-performance |
On Mon, Feb 25, 2013 at 10:11 AM, Maciek Sakrejda <m.sakrejda@gmail.com> wrote:
I don't think it is SUSET for performance reasons, as the ordinary user already has plenty of ways to shoot themselves (and fellow users) in the foot performance-wise. I think it was based on the idea that those tracking tools that the administrator has turned on, ordinary users may turn off. I think the other way around would be fine (if it is off for the server, the user can still turn it on for their session--and presumably also turn it off again if it is one only because they set it that way, not because the administrator set it that way), but I think that that behavior is not trivial to implement. I looked in the archives, but the SUSET nature of this doesn't seem to have been discussed.
Cheers,
Jeff
On Sat, Feb 23, 2013 at 3:53 PM, Jeff Janes <jeff.janes@gmail.com> wrote:Right, that's not supported right now, although given that the
> It would really help to have "explain (analyze, buffers)". Especially if
> you turn on track_io_timing, (although that part probably can't be done on
> Heroku, as it requires superuser access.)
superuser is primarily for performance considerations (right?),
perhaps we should find some way of exposing this.
I don't think it is SUSET for performance reasons, as the ordinary user already has plenty of ways to shoot themselves (and fellow users) in the foot performance-wise. I think it was based on the idea that those tracking tools that the administrator has turned on, ordinary users may turn off. I think the other way around would be fine (if it is off for the server, the user can still turn it on for their session--and presumably also turn it off again if it is one only because they set it that way, not because the administrator set it that way), but I think that that behavior is not trivial to implement. I looked in the archives, but the SUSET nature of this doesn't seem to have been discussed.
Cheers,
Jeff
В списке pgsql-performance по дате отправления: