Re: Initializing Datums for use with SPI_execute_plan

On 9/19/06, Martijn van Oosterhout <kleptog(at)svana(dot)org> wrote:
> On Tue, Sep 19, 2006 at 01:27:56PM -0400, Jack Orenstein wrote:
> > On 9/18/06, Jack Orenstein <jack(dot)orenstein(at)gmail(dot)com> wrote:
> > >---------- Forwarded message ----------
> > >From: Andrew - Supernews <andrew(at)supernews(dot)net>
> > ...
> > > Jack> I have an int8 that I need as a Datum for use with
> > > Jack> SPI_execute_plan.
> > >
> > >Int64GetDatum(your_variable)
> >
> > which then has to be pfree'd, correct?
>
> Possibly, it probably depends on the architechture. The memory is being
> allocated in a per-call context IIRC so it'll be freed at the end of
> the function anyway. I wouldn't worry about it.

Can you provide some guidance (or point to some documentation) on how
to manage memory? Is the idea that I should (must?) not pfree
palloc'ed memory from Int64GetDatum, but I should free anything I
allocate myself using palloc? Or not even that?

The C extension I'm writing, which uses the SPI, will be called
thousands or millions of times as part of a data conversion -- I do
have to worry about memory leaks. Once the conversion completes, I
won't use the function any longer. But I'd rather not leak memory and
have to do something drastic to reclaim it, such as bouncing
postgresql.

Jack