Re: [mdt-uml2.dev] UML 2.5 progress

[Ed Willink <ed@xxxxxxxxxxxxx>]

Hi Kenn

Eventually I did get URI mappings to work. Temporarily, and in order to 
allow me to use the UML 2.5 models as transformation sources, the 
org.eclipse.ocl.examples.uml25 plugin installs the relevant 
registrations to allow the UML 2.5 namespaces to be just used. (This is 
only on the edw/417663 branch at present; won't be in M2, might be in M3 
examples, may be only in M3 tests).

In doing this I discovered ExtendedMetadata and so 
XMI252UMLResourceFactoryImpl has an initialize method that provides all 
the missing standalone registrations. UMLResourcesUtil.init would 
therefore be much better structured as just calls to 
XMI***2UMLResourceFactoryImpl.init for each known version. This would 
make it clear exactly how to add another version. Dangerous but there 
could also be a catch all for all unknown namespaces.

(It really is naff that there is no implementation of the MOF 
package(s) and so it all happens dynamically after an Internet timeout 
or worse, prohibiting installation of the MOF package under yet another 
namespace.)

    Regards

        Ed Willink


On 24.09.2013 19:57, Kenn Hussey wrote:
> Thanks for all the legwork, Ed. This will be instrumental as we look
> to adopt UML 2.5 at Eclipse, hopefully in the Luna release timeframe.
>
> Unfortunately, I don't think the disproportionately small number of
> UML2 users in our situation (i.e., those working on future versions 
> of
> the specification) justifies investing in a new mechanism for mapping
> OMG URIs from the future to UML2 URIs from the present. :(
>
> Have you tried registering URI mappings in the URI converter map?
>
> Cheers,
>
> Kenn
>
> Sent from my iPhone
>
> > On Sep 24, 2013, at 12:23 PM, Ed Willink <ed@xxxxxxxxxxxxx> wrote:
> >
> > Hi
> >
> > I have reviewed the latest UML 2.5 XMI and successfully used UML and 
> > OCL validation so that a few difficulties, but unfortunately not all, 
> > are being resolved 'editorially' before the final XMI for UML 2.5.
> >
> > It looks like the final UML XMI will have two known bugs, the first 
> > of which UML validation detects:
> >
> >    ExtensionEnd::/lower has a bad cardinality override; fix is to 
> > change multiplicity to [1..1]
> >    Lifeline::interaction_uses_share_lifeline has an OCL semantic 
> > error; fix is to perhaps change "select" to "exists"
> >
> > Similarly the UMLDI XMI will have two known bugs, the first of which 
> > UML validation detects:
> >
> >    There is diamond inheritance for which the two inheritance paths 
> > provide 11 conflicts; fix re-redefine at the diamond join.
> >    UMLAssociationOrConnectorOrLinkShape::edge_instancespec has an 
> > OCL semantic error; fix is perhaps to change "e.modelElement" to "e".
> >
> > The ability to validate the UML 2.5 with Eclipse UML '2.4.1' is very 
> > encouraging. It is however really tedious to have to change all the 
> > namespaces by XMI text edit. Is there no way we can support a 
> > user-defined mapping from e.g. 
> > http://www.omg.org/spec/UML/20131001/PrimitiveTypes to 
> > PrimitiveTypes.uml? With the UML metamodel and serialization 
> > increasingly stable, use of Eclipse tooling for UML version N to 
> > support N+1 may become increasingly practical and useful.
> >
> >    Regards
> >
> >        Ed
> >
> >
> > _______________________________________________
> > mdt-uml2.dev mailing list
> > mdt-uml2.dev@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/mdt-uml2.dev