Re: [v9.2] LEAKPROOF attribute of FUNCTION (Re: [v9.2] Fix Leaky View Problem)

2012/2/14 Robert Haas <robertmhaas(at)gmail(dot)com>:
> On Mon, Feb 13, 2012 at 7:51 AM, Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp> wrote:
>> I rebased the patch due to the updates of pg_proc.h.
>>
>> Please see the newer one. Thanks,
>
> Thanks, committed.  I think, though, that some further adjustment is
> needed here, because you currently can't do ALTER FUNCTION ... NO
> LEAKPROOF, which seems unacceptable.  It's fairly clear why not,
> though: you get a grammar conflict, because the parser allows this:
>
> create or replace function z() returns int as $$select 1$$ language
> sql set transaction not deferrable;
>
> However, since that syntax doesn't actually work, I'm thinking we
> could just refactor things a bit to reject that at the parser stage.
> The attached patch adopts that approach.  Anyone have a better idea?
>
I could not find out where is the origin of grammer conflicts, although
it does not conflict with any options within ALTER FUNCTION.

Do you think the idea of ALTER ... NOT LEAKPROOF should be
integrated within v9.2 timeline also?

> I also think we ought to stick create_function_3 into one of the
> parallel groups in the regression tests, if possible.  Can you
> investigate that?
>
Not yet. This test does not have dependency with other tests,
so, I'm optimistic to run create_function_3 concurrently.

Thanks,
--
KaiGai Kohei <kaigai(at)kaigai(dot)gr(dot)jp>