[Biojava-dev] AnnotationForwarder

[Thomas Down]

On Fri, Nov 01, 2002 at 01:59:12PM +1300, Schreiber, Mark wrote:
> Would it be a good idea to add a method to Annotatable called
> getAnnotationForwarder() as a reminder that the AnnotationForwarder
> needs to be implemented?
> 
> Obviously this would break lots of stuff so might not be a very
> attractive solution.

Please no.

The AbstractChangable/ChangeSupport/ChangeForwarder mechanisms
are a reasonable stab at providing a `general purpose' implementation
of the ChangeEvent framework, especially since Matthew refactored
them and thrashed them about a bit.  However, if you've got
events being forwarder around a fairly complex graph of objects,
the number of ChangeSupport and ChangeListener objects can 
rapidly become large.  Their existence also means that any object
which has listeners is effectively `locked' into memory.  This
just doesn't worth for big databases.  I've implemented a number
of systems which ignore these default implementations.  The most
obvious is the latest incarnation of the biojava-biosql.  This
has a ChangeHub object which fakes all the event forwarding
behind the scenes, even if an event is actually being forwarded
through an object which isn't currently loaded in memory.

While I welcome any attempt to identify and fix classes which are genuinely
missing necessary event-handling logic (and I know there are a number
of these), it should still be possible to write objects which use
non-standard event implementations (or, for that matter, no implementation
at all in cases where this underlying data isn't going to change).

     Thomas.