[Uml-devel] news on 2.crap

[Andrew Sutton]

just kidding. i have good news :)

although its taken me around 6 months and 4 iterations of design and 
implementation, i've finally achieved a zen-like understanding of reflection 
and model/metamodel relationships - well, i understand most of it anyways. 
okay, maybe just a decent chunk of it.

so what. what's that mean? well, it took me writing an implementation - or a 
good chunk of one - of the MOF, reflective interfaces and the type library to 
get my wrapped around it. as it stands right now my reflection, type library, 
and MOF implementation covers about 130 classes and about 12,000 lines of 
code. pending the implementation of about a dozen more classes and a 
restructuring of the build environment, i'm going to post the initial 
contribution to 2.0 in the near future - i'm aiming for sunday, but that 
could change if i run into some real sticky issues.

so, what's this mean for people wanting to work on 2.0? well, basically its a 
giant code review. there's plenty of places where i need exceptions thrown 
and checked, memory leaks plugged and the like - maybe reference counting for 
RefObject wouldn't hurt either, but if we follow the basic memory management 
rules, it shouldn't be too much of an issue. also, people could start work on 
the gui framework - it should basically be the MVC (model view controller). 
people could also work on real implementations of the interfaces for MOF - 
i've just stubbed them out for now. for example, those implementations can 
now be safely derived from QObject (for signals and slots), although i didn't 
do it myself (yet).

after i get this stuff done, i'm moving on to the serialization/persistence 
frameworks - particularly XMI. i figured out the generic recipe for almost 
unassisted XMI parsing so i should be able to build the XMI serialization 
stuff in short order - maybe a week or two. after that, i'm writing the 
metamodel generator so we can generate the entire UML implementation. i hope 
that by the time i finish the XMI parser and the document abstraction, we 
should have hammered out all the implementation details for metamodels.

speaking of UML, i found something pretty funny. the XMI specification of UML 
doesn't comply to the MOF standards. they're instantiating the abstract 
Model:DataType class to define structures and enumerations for UML. oops. i 
guess that means we have to tweak the file to be compliant with what we know. 
oh well.

andy