Re: [Modeling-users] Notification Framework v0.5 released

[Sebastien B. <sbi...@us...>]

  Hi Kodi and all,

Kodi wrote:
> I've tested the new feature and it is possible now to add an observer
> with different methods for different Notifications.
> But it should be pointed out that one has to define a value for the
> key 'NOTIFICATION_CENTER_MULTIPLE_CALLBACKS_PER_OBSERVER' in the
> environment, in order to enable the new behaviour.

  Well, yes, that was a terse announcement. However the documentation
was updated --I've just updated it a bit more so that it's clear where
the reference documentation stands, e.g. in module NotificationCenter
cf. http://modeling.sourceforge.net/API/Notification-API/index.html

  I also forgot to mention that there is a new script
tests/compare_perfs.py for comparing the performance of the framework
when the new feature is en-/dis-abled. It is ~2.4x slower on my machine
when enabled (the modeling framework does not need it to be activated,
but of course behaves well when it is).

I wrote:
> > This release fixes two important bugs implying:
> >   - addObserver() not reporting when an object tries to register two or
> >     more different callbacks,
> >     - postNotification() copying the observers back to its own list,
> >     making it grow and causing degradation of performance each time it
> >     is called.
> >   The modeling framework itself is not affected by these changes and
> >   did not even triggered the bugs. However and since it may take
> >   advantages of these changes in the future, you'll probably want to
> >   upgrade your local copy.

  Sorry but I was a bit distracted when writing this: on the contrary
  the MDL is likely to trigger the second bug. It is revealed when a
  given notification is listened by both generic and specific observers,
  which is at least the case for MDL's GlobalIDChangedNotification
  (listened by EditingContext [generic] and DatabaseContext [specific]).

  Hence it is now *strongly recommended* that you upgrade your local
  copy with v0.5. Sorry for the wrong comment.

        Cheers,

-- S=E9bastien.