[news.eclipse.tools.emf] Re: On-the-fly EMOF-to-Ecore mapping

[Víctor Roldán Betancort <vroldan@xxxxxxxxxxxxxxxx>]

Ed,

I committed the corresponding bug (216750). You can find it at 
http://bugs.eclipse.org/bugs/show_bug.cgi?id=216750.

I didn't attach any .emof file (didn't know if it was appropriate), I 
could upload it or you could get them from UMLX sources.

Thank you very much for the feedback,
  Víctor Roldán.

Ed Merks escribió:
> Víctor,
>
> Comments below.
>
> Víctor Roldán Betancort wrote:
> > Hi Ed,
> >
> > the .emof files came from Ed Willink's UMLX project, having some QVT 
> > Core (i'm working on a Core implementation) transformation examples 
> > where input models conform to .emof metamodels.
> That makes sense.  He's the only person ever to have shown a direct 
> interest in EMOF.
> > The problem is that the wizard for genmodel creating shows up an error
> > when selecting the ".emof" file as source metamodel. For "rdbmsMM.emof"
> > something like this is shown:
> >
> > "The Ecore model file name for package 'rdbmsMM' must end in '.ecore'"
> I see.  That's not good.  Please open a bugzilla.  (I can reproduce this 
> without a test case.)
> > So, if I decide to rename the file as ".ecore" extension, EMF won't 
> > interpret correctly the XMI file, trying to load an ".ecore" file
> > instead of an ".emof", and then, all serialization differences between
> > ecore and emof show up (for instance, upperbound as "*" in EMOF, "-1" 
> > in Ecore").
> >
> > Maybe is just a matter of fixing the wizard.
> I think so.
> > If you want, I can send you the .emof files.
> >
> > Thanks!
> >
> > Ed Merks escribió:
> > > Víctor,
> > >
> > > Comments below.
> > >
> > > Víctor Roldán Betancort wrote:
> > > > Hi!
> > > >
> > > > I've been experimenting a bit with .emof files.
> > > I'm always curious if .emof files come from anything other than EMF.  
> > > Where did your EMOF models come from?
> > > > I was surprised how these kind of files could be opened without any 
> > > > EMOF metamodel registered. Taking a look at the source code I 
> > > > realized that there was an on-the-fly mapping during XMI serialization.
> > > Yes, the EMOF and Ecore models pretty near isomorphic.
> > > > I actually had some EMOF-based metamodels, but it was not possible 
> > > > to generate instances of that metamodel.
> > > Where did they come from?
> > > > Creating the genmodel failed.
> > > Did you validate the Ecore model that was produced?  The generator 
> > > typically does that for you, so I imagine it would be hard to 
> > > overlook the warnings/errors.  What exactly failed?
> > > > I think it's possible to be generated programmatically, but I'm aiming
> > > > to generate editors to create instances of my .emof based metamodel
> > > That should be possible.  I'd need to understand what isn't working.
> > > > My question is if it's really possible to generate the genmodel for 
> > > > an ..emof file, and in case there isn't, if there is any plan to 
> > > > give support to it.
> > > It should definitely be possible, as long as the resulting Ecore 
> > > model is well formed and complete.  Validation should find all the 
> > > problems.  I'll need more details.
> > > > Regards,
> > > >    Víctor.