Thread: plpythonu and return void

I recently migrated a database from Postgres 7.4.1 to Postgres 8.03. The only
problem I have run into is that a plpythonu function that returns void will
not run under 8.03. The error message states that a plpython function cannot
return void. I got around it by having it return text and not actually
returning anything. I have searched for an explanation and can't seem to
locate one. Obviously I have missed something I just don't know what? Any
suggestions would be appreciated.
--
Adrian Klaver
aklaver@comcast.net

Adrian Klaver <aklaver@comcast.net> writes:
> I recently migrated a database from Postgres 7.4.1 to Postgres 8.03. The only
> problem I have run into is that a plpythonu function that returns void will
> not run under 8.03. The error message states that a plpython function cannot
> return void. I got around it by having it return text and not actually
> returning anything. I have searched for an explanation and can't seem to
> locate one.

The explanation is doubtless here:

2004-09-19 19:38  tgl

    * src/pl/plpython/plpython.c: Add defenses against plpython
    functions being declared to take or return pseudotypes.  Not sure
    why I neglected to add these checks at the same time I added them
    to the other PLs, but it seems I did.

It's certainly possible to weaken this test to allow "void" again,
but shouldn't there be corresponding changes elsewhere to ensure that
the behavior is actually sensible?

            regards, tom lane

On Monday 08 August 2005 08:02 pm, Tom Lane wrote:
> Adrian Klaver <aklaver@comcast.net> writes:
> > I recently migrated a database from Postgres 7.4.1 to Postgres 8.03. The
> > only problem I have run into is that a plpythonu function that returns
> > void will not run under 8.03. The error message states that a plpython
> > function cannot return void. I got around it by having it return text and
> > not actually returning anything. I have searched for an explanation and
> > can't seem to locate one.
>
> The explanation is doubtless here:
>
> 2004-09-19 19:38  tgl
>
>     * src/pl/plpython/plpython.c: Add defenses against plpython
>     functions being declared to take or return pseudotypes.  Not sure
>     why I neglected to add these checks at the same time I added them
>     to the other PLs, but it seems I did.
>
> It's certainly possible to weaken this test to allow "void" again,
> but shouldn't there be corresponding changes elsewhere to ensure that
> the behavior is actually sensible?
>
>             regards, tom lane
Thanks for the explanation. One of these days I will remember that the source
is my friend.
--
Adrian Klaver
aklaver@comcast.net