#14 Should support AssociationClass properly

[Simon Wright]

In UML, there's no such thing as an associative class; an AssociationClass has both Association and Class aspects.

The translation should reflect this (the initial ArgoUML «association-class-name» approach is very clumsy).

[Simon Wright]

• Description has changed:

Diff:

--- old
+++ new
@@ -1,3 +1,3 @@
 In UML, there's no such thing as an associative class; an AssociationClass has both Association and Class aspects.

-The translation should reflect this (the initial ArgoUML <<association-class-name>> approach is very clumsy).
+The translation should reflect this (the initial ArgoUML &#171;association-class-name&#187; approach is very clumsy).

• Group: --> (none)

[Simon Wright]

Now using «association-class-naming», which might be considered an improvement :-)

[Alex Proudfoot]

I noticed in your House Management example that this approach puts the association name on the class and the class name in a tag, which doesn't appear on the class diagram. Since the class name is (should be) dictated by an analysis of the problem domain, would it not be better to use the reverse approach? I don't know if the stereotype would be required in this circumstance. If it were, maybe it should be called << association-class >> .

[Simon Wright]

In this case, the use of an AssociationClass was enforced by the formalism, not by the application (it's an Mc:Mc relationship, must have an associative object). If it was the application that required it it might make more sense to name the associative object visibly and use «association-class-naming» and the tag {association-name=new-name}; perhaps with a comment on the associative class … no, ArgoUML won't let you link a comment to an association or an associative class, just to a class (or perhaps a data type). Huh.

[Alex Proudfoot]

Sounds good. If ColdFrame were to allow both, the analyst/designer would have flexibility to indicate whether the AssociationClass has significance in the problem domain or only exists to formalize the relationship. :)

PS. I see what you mean about the non-attachment of the note. How irritating.

[Alex Proudfoot]

Oh, silly me. I didn't realise that the association-name tag was already available. That solves my problem showing the class name on the class diagram when modelling interactions then.