Am 02.08.2006 um 21:12 schrieb Mike Schrag:
>> I agree that if *I* had written EM, I'd probably have used WO
>> classes, but that is just because I am lazy.
> I honestly don't know (well, laziness aside :) ) ... We have a lot
> of capability for fixing problems with our own version. Honestly,
> other than SQL generation (which we hook into WO to do, but at
> large performance cost with dynamic classloader generation), most
> of EOModeler is just wrapping a data structure. The EOF EOModel
> frameworks also have a very coarse event notification system that
> would have made building a live-updating UI very frustrating -- you
> would basically end up having to write wrappers for all the model
> objects, at which point, what else is there really other than file-
> loading/saving, which is pretty small.
Which was was popped into my mind 10 seconds after I hit send:) So
I'm not only lazy, but also slow...
You *do* get a long way without private API but some things would be
rather painful. And actually this is also the reason why EOM is such
a mess (UI update wise).
Also, plugging into the system to get you own subclasses loaded is
rather a pain (I tried. Hard.), in particular for subclasses of
attributes and relationships. Someone posted some very screwy code on
the wonder lists which involved serializing and deserializing to get
even simple things done...
So lets put this to rest. It's very good the way it is now and I
don't suppose that EOFs model format will change during WWDC (or in
the next 5 years).
When someone actually comes forth and fleshes out the RuleEditor (not
me, I did enough work on RuleModeler), we might think about it again
but this currently is not in the cards.
Cheers, Anjo
This archive was generated by hypermail 2.0.0 : Wed Aug 02 2006 - 15:55:45 EDT