On 1/14/14 10:16 AM, Pavel Stehule wrote:
> 2014/1/14 Florian Pflug <fgp@phlo.org>
>
>> So if we really want to change this, I think we need to have a
>> LANGUAGE_VERSION attribute on functions. Each time a major postgres release
>> changes the behaviour of one of the procedural languages, we'd increment
>> that language's version, and enable the old behaviour for all functions
>> tagged with an older one.
>>
>
> I dislike this proposal
>
> too enterprise, too complex, too bad - after ten years we can have a ten
> language versions and it helps nothing.
At this point I'm almost tempted to agree with Florian -- I'm really
hoping we could change PL/PgSQL for the better over the next 10 years,
but given the backwards compatibility requirements we have, this seems
like an absolute nightmare. Not saying we need a version number we can
keep bumping every release, but something similar.
> return back to topic
>
> a) there is agreement so we like this functionality as plpgsql option
>
> b) there is no agreement so we would to see this functionality as default
> (or in near future)
>
> @b is a topic, that we should not to solve and it is back again and again.
> Sometimes we have success, sometimes not. Temporal GUC is not enough
> solution for some people.
>
> So my result - @a can be implement, @b not
FWIW, I would like to have this behaviour even if it's not the default
(that was my original proposal anyway). It could help catch a lot of
bugs in testing, and for important code it could be even turned on in
production (perhaps on a per-function basis). Maybe even globally, idk.
> I am working on plpgsql_debug extension and I am thinking so I am able
> (with small change in plpgsql executor) implement this check as extension.
> So it can be available for advanced users (that will has a knowledge about
> additional plpgsql extensions).
While this sounds cool, running a modified postgres locally seems a lot
easier. I already have to do that for PL/PgSQL development because of
the reasons I outlined in the thread "PL/PgSQL, RAISE and error context".
Regards,
Marko Tiikkaja