On Fri, 2007-08-31 at 16:42 -0400, Merlin Moncure wrote:
> On 8/31/07, Tom Lane <tgl@xxxxxxxxxxxxx> wrote:
> > "Merlin Moncure" <mmoncure@xxxxxxxxx> writes:
> > > At present, immutable functions are only treated as constants during a
> > > query, which is what we want (no problems with prepare).
> >
> > Uh, no, they'd be folded to constants at plan time, which is exactly
> > what Jeff doesn't want AFAICS.
>
> yikes! I did test this before I posted that, but I oversimplified it:
> I didn't move the func() to the where clause...do the subselect
> version defined as volatile seems the way to go. unfortunately this
> means you pay a small extra price for large result sets.
>
That sounds like a good solution to me. It looks like the planner is
able to optimize the queries, and the audit function is only called
once. It sounds like I may need to beware of future changes, however.
What is the small extra price for large tables though?
Thanks for the help!
Regards,
Jeff Davis
---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?
http://archives.postgresql.org/
