association class and generalization bug

[SF_lt]

in EA 6.780 you can't make association class to be a descendant of the another class ;-)

[KP]

in EA 6.780 you can't make association class to be a descendant of the another class ;-)

...only of another association class :-/  I'll try to get that fixed for 781. Until then, Tools | Options | Diagram | Strict UML Syntax.

Neil  

[Paolo F Cantoni]

...only of another association class :-/  I'll try to get that fixed for 781. Until then, Tools | Options | Diagram | Strict UML Syntax.

Neil  

Neil,
While you're at it, can you also add the ability to generalize Associations?

As per my treatise on AssociationClasses, if you can Generalize an AssociationClass, you have to be able to Generalize an Association (and by implication Generalize and AssociationClass to an Association).

BTW:  Unchecking Strict UML syntax still won't all it.

Paolo

[SF_lt]

Until then, Tools | Options | Diagram | Strict UML Syntax.

Neil, strict UML syntax on allows the same name classes in the same package - this shouldn't be I guess.

Paolo, I saw somewhere generalizable association diagrams, but could you provide good example on this? Right now can't think any case, when this would be needed

[KP]

Neil, strict UML syntax on allows the same name classes in the same package - this shouldn't be I guess.

UML states that elements must be distinguishable by name and type within the same namespace. Somebody (I think it was Bruce aka Sargasso) recently suggested that if two elements aren't distinguishable then it just means they are the same element. I'm happy to go along with that.

Neil

[sargasso]

  You betcha it was me.  And I still stand by it.  It's one of the more gotcha features of UML.  An element of type X with the name Namespace.Name is the same thing as a completely different X called Namespace.Name that appears in the same model.

The problem is that EA doesn't recognise it as such.

IMO, either OMG has to define more exsplicitly () rules for mulitple representation in a model,  (or at least explain why they use it in their metamodels but don't define it), or EA should pre-empt them and give us a sane and recognizable way to represent the same thing multiple times in a model.

Let me go further out on a limb - according to the "Old Bore's Almanac" (me) - .NET 2 "partial classes" are going to become a major EA problem in 2006.   Try rev-engineering some and see what I mean - multiple classes that are all the same thing!

I, for one, would like to see this issue as the one big thing fixed this year .


bruce

[Hans]

Bruce,

if only you could convince MS to have aspx-code part of a WebForm (Page) class as is the case for Windows.Forms and persuade them to take a leave from partial classes, *that* would be the major step!

Regards
Hans