On Tue, 14 Sep 2004, Krzysztof Czaja wrote:
since a minor tweak in any of the g_ files could break everything...(mathieu has seemed interested in pure-Tk GUI objects so i would hope or suspect the process is integrated one day..)
I had started working on Model-View object relationships in Pd. The way I implement it in IMPD is that the Model is a C object and the View is a Tcl object that is as much as possible a clone of the first one and updated by recopying its fields from the Model.
The problem with doing such work is that it completely obsoletes any old kind of GUI object API in such a way that it becomes very boring and/or silly to make the effort to keep compatibility with the old system. So, I thought, I should take that occasion to make the most useful possible changes all at once.
That new system wouldn't be pure-Tk but would be rather close to it; and possibly that in such a context a kind of stub class could be created, that does the minimum possible handling in C in a generic way so that subclasses of it may be created by only adding Tcl code.
Actually I have a prototype of that but it's been on hold for ~4 months because I have more GridFlow-related inspiration these days. I stopped working on it just before working out the Observer/Observable relationship and the new pd_changed() function that decouple C and Tcl one step further.
Mathieu Bouchard http://artengine.ca/matju