[news.eclipse.modeling.gmf] Re: Links for references?

[Miles Parker <milesparker@xxxxxxxxx>]

Excellent, that worked perfectly!

That was actually the very first thing I tried, but I couldn't get the 
model to validate. (Diagnosis of Mapping  Link mapping without 'Domain 
Element' class does not use 'Containment Feature' and 'Source 
Feature'). The secret was "(leave all other features blank)" -- I had 
been specifying both the source meta feature and its target opposite. 
Of course, EMF handles the bi-direcitonal link automatically.



On 2007-01-25 09:36:42 -0500, "Boris Blajer" <boris.blajer@xxxxxxxxxxx> said:

> Hi Miles,
>
> Reference-based links are supported. Just leave the Domain Model 
> Element empty for the link mapping and specify the correct Link Meta 
> Feature (leave all other features blank).
> The code will be generated for reference-based link without an 
> underlying model element.
>
> Best regards,
> Boris Blajer
>
>
> "Miles Parker" <milesparker@xxxxxxxxx> wrote in message 
> news:ep8h66$72u$1@xxxxxxxxxxxxxxxxxxxx
> > Apologize is this has been addressed already but I couldn't find anything.
> >
> > I'd like to support links for refernces between nodes without having an 
> > explicit domain object cooresponding to the link. For example, imagine 
> > the State model without the transition (this of course assumes that you 
> > don't have a many-to-many relation or have actual transition state you 
> > care about).
> >
> > I've managed to support the creation of the link by overiding 
> > *TypeLinkCreateCommand.doElementCreation, [with support for doUndo and 
> > doRedo), ignoring the super object and simply adding the references to 
> > the source and target node objects.
> >
> > But am now having lot's of second thoughts about wheree to go from 
> > here. For example, all of the edit pieces are generated for whatever my 
> > "imaginary" Link doman element is. I'm not even sure how one would 
> > manage delting the domain element. When I do an undo, it works fine, 
> > but after a save, if I try to try that the diagram is deleted (!!) with 
> > no error reported. etc..
> >
> > So not asking to solve all of these problems for me ;) but I'm really 
> > wondering wether this is even a reasonable thing to do?