Re: [rgui-dev] Jumping Right In

[Tom Sawyer]

On Tue, 2002-08-06 at 16:26, Massimiliano Mirra wrote:
> It's not really instance variables that we're after.  They represent
> the state of an object, not what we want to expose of it to the
> outside.  That's what getter methods are for.  So, your current way of
> dynamically adding code to *methods* might have implementation
> drawbacks but is conceptually the right one or one of the right ones.

well, i just worked out my idea for binding on instance variables. and
it works! obviously we need to discuss this a bit more per the above to
make sure this is the best way to proceed, but it does deal with all the
short comings i have mentioned of only being able to bind on accessor
methods. i realize that the accessor methods are intended for the
outside exposure and the instance variables are not, but like i was
saying, they expose the the object's data with only procedural
programming in mind, not the automagical (declarative) like we're trying
to do --that's the source of the short comings.

so my work on this seems to be coalescing into a design pattern in
itself. which is pretty cool. don't know what to call it. perhaps the
Automate Pattern? its similar to the Observable Pattern, but unlike
Observable it does not require any explict calls be added to the
"observed" object. in light of this, i wonder if it this pattern should
be implemented as a module or not. if i implement it as a module then
one would have to add an include to every model object one wishes to
"Automate".  otherwise in would just be an include that modifes the
Object class, just as the current GUtopIa #bind method does.

thoughts?

~transami

p.s. just did a little search. turns out that my Automate Pattern (for a
lack of a better name) is a subtype of the Observer with Method
Combination Pattern. So that bodes well.