| From: | Robert Haas <robertmhaas(at)gmail(dot)com> |
|---|---|
| To: | Brian Ceccarelli <bceccarelli(at)net32(dot)com> |
| Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, "pgsql-bugs(at)postgresql(dot)org" <pgsql-bugs(at)postgresql(dot)org> |
| Subject: | Re: BUG #5611: SQL Function STABLE promoting to VOLATILE |
| Date: | 2010-08-11 15:32:52 |
| Message-ID: | AANLkTin3UGUxW++=mDRZqOeWe4siCYJiKUUdx0Lp3CLO@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-bugs |
On Wed, Aug 11, 2010 at 11:01 AM, Brian Ceccarelli
<bceccarelli(at)net32(dot)com> wrote:
> My complaint remains. That inlined function f_return_ver_id_4() is a STABLE function, inlined or not. Postgres now calls it multiple times during the transaction, even though the arguments to f_return_ver_id_4() have not changed.
>
> STABLE no longer means STABLE. This behavior is killing my performance. I am getting 500% to 30000% increase in latency.
We've never guaranteed that, and almost certainly never will. Marking
a function STABLE means that the planner is *allowed to assume* that
the results won't change for a given set of arguments, not that it is
*required to prevent* it from being called multiple times with the
same set of arguments.
You can certainly prevent the function from being inlined, though
(perhaps, by writing it in PL/pgsql).
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Brian Ceccarelli | 2010-08-11 15:50:25 | Re: BUG #5611: SQL Function STABLE promoting to VOLATILE |
| Previous Message | Tom Lane | 2010-08-11 15:18:26 | Re: BUG #5611: SQL Function STABLE promoting to VOLATILE |