2008/12/27 Tom Lane <tgl@sss.pgh.pa.us>:
> "Robert Haas" <robertmhaas@gmail.com> writes:
>> Unfortunately, if we don't want to add an explicit iswindowable flag
>> (and I understand that that's ugly), then I think this is the way to
>> go. It's a shame that people will have to make code changes, but
>> inventing a fake AggState object just to get around this problem
>> sounds worse. The array_agg code is new and the fact that it doesn't
>> follow the design pattern should be considered a bug in that code
>> rather than a justification for an ugly workaround.
>
> Well, array_agg may be new but it's simply a re-implementation of a
> design pattern that existed in contrib/intagg since 7.3 or so. I have
> no problem with fixing array_agg --- what I'm wondering about is who
> has copied intagg before.
We agree that the best solution for ten core aggregates is to rewrite
them to support or not support WindowAgg, so the care for third party
aggregates copied from intagg is nothing but announcing that the
behavior is changing. -- if we had better alternative we should do it,
but it seems to me that there's no way not to break the non-core
aggregates.
SInce at t least you must compile the modules again on 8.4 release,
compiling time warnings or something is the best announcing for now.
Or any other suggestions?
Regards,
--
Hitoshi Harada