Re: call the same pl/pgsql procedure twice in the same connection

Bruce Momjian wrote:
> Jan Wieck wrote:
> > Bruce Momjian wrote:
> > > Jan Wieck wrote:
> > > > Bruce Momjian wrote:
> > > > >
> > > > > Jan, instead of doing cache invalidation to fix temporary tables, can we
> > > > > disable cached plans for functions that use temporary tables?
> > > >
> > > > I was thinking of a different approach. Enhancing the SPI
> > > > manager to detect if a plan uses temporary objects and to
> > > > remember the original querystring in the SPI_plan. Having
> > > > callbacks when temp object beeing destroyed into the SPI
> > > > manager, causing it to reparse and plan on the next call to
> > > > SPI_execp() would do it for everything that uses SPI.
> > >
> > > I was merely proposing that preventing caching of functions ueing temp
> > > tables may be easier than trying to invalidation them on temp table
> > > destruction.
> >
> > It's neat to say "preventing caching of functions using ...",
> > now tell in detail how you detect that a function "is" using
> > a temp table? No, I don't mean how "you" can detect it, how
> > can the PL/pgSQL parser or executor detect it. And when do
> > you detect it? Remember that PL/pgSQL has delayed SPI
> > preparation?
> >
> > Second, it doesn't really look smart to me to prevent saving
> > of all query plans just because one of them uses a temp
> > table.
>
> Well, I assume you could spin through the plan at save time and look at
> every relation reference to see if it is a temp table.

Bruce,

you are really good at making things sound simple by stopping
half way through.

Your assumption is exactly my "detect if a plan uses
temporary objects". No go on, what exactly are you doing with
the information that the plan uses a temp table?

And BTW, since you are the one who invented temp tables, how
does someone detect a temp table?

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck(at)Yahoo(dot)com #