[news.eclipse.modeling.mdt.uml2] Re: GeneralizationSet

[Kenn Hussey <Kenn.Hussey@xxxxxxxxx>]

Timothy,

Yes, the concept of generalization sets is indeed defined in terms of 
UML's notion of instances. However, I would not expect an object to be 
an instance of both Sub1 and Sub2 at the same time unless those 
classifiers are related by a generalization or (interface) realization 
relationship.

A covering generalization set means that all instances in the hierarchy 
are "covered" by the set. In your example, it would mean that all 
instances of Super (or of a specialization of Super) are directly or 
indirectly instances of either Super1 or Super2 (assuming the set 
includes Super1 and Super2).

Kenn


Timothy Marc wrote:
> Hey all,
>
> a more or less simple question concerning the GeneralizationSet. Let's 
> consider:
>
> class Super
> class Sub1 extends Super
> class Sub2 extends Super
>
> Concerning the GeneralizationSet (GS) mechanism, the spec mentioned, 
> that a GS will covering, if instances of Sub1 are also an instance of 
> Sub2. So far so good, but...
>
> IMHO, this instance relationsship deals with the UML-internal instance 
> specification mechanism, am i right? In Java, for example, any check 
> like (sub1 instanceof sub2) will fail, because there are not directly 
> compatible (only via an upcast).
>
> But in UML it seems possible, that an instance of a subclass can also be 
> an instance of any other corelating subclass of this GS, since any 
> instance specification can point to several classifiers.
>
> Is this correct?
>
> Thx
> Timothy