Re: oversight in parallel aggregate

On 5 April 2016 at 13:09, David Rowley <david(dot)rowley(at)2ndquadrant(dot)com> wrote:
> On 5 April 2016 at 11:59, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>> One of my EDB colleagues, while in the process of refactoring some
>> unrelated Advanced Server code, discovered that (1) there's no way to
>> mark an aggregate as anything other than parallel-unsafe but (2) it
>> doesn't matter because has_parallel_hazard ignores Aggrefs anyway.
>> These mistakes cancel each other out (sorta) if all of your aggregates
>> happen to be parallel-safe, but otherwise not so much. Barring
>> objections, I intend to speedily apply the attached patch to fix this.
>
> Thanks for working on this. I should have noticed this myself...
>
> I had a quick look at this and I manged to make this happen;
>
> david=# create aggregate mysum(int) (sfunc=int4pl, combinefunc=int4pl,
> stype=int, parallel);
> server closed the connection unexpectedly
>
> I've attached a fix, which makes the code a bit more simple, and also
> inline with the other code in DefineAggregate().
>
> I think there was also a couple of missing syntax synopsis in the docs
> too. I've added those.

Another thought;

+ else if (IsA(node, Aggref))
+ {
+ Aggref *aggref = (Aggref *) node;
+
+ if (parallel_too_dangerous(func_parallel(aggref->aggfnoid), context))
+ return true;
+ }

Does this need to check the parallel flags on the transfn or serialfn?
these'll be executed on the worker process. Possibly we also need the
combinefn/deserialfn/finalfn to be checked too as I see that we do
generate_gather_paths() from set_append_rel_pathlist().

--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services