[Uml-devel] bug 54573 and association roles

Jonathan Riddell jr at jriddell.org
Mon Dec 8 17:27:03 UTC 2003

On Tue, Dec 09, 2003 at 12:38:41AM +0100, Oliver Kellogg wrote:

> I tracked down bug 54573 to a problem with the irregular
> role assignments not being accounted for by the association
> line sorting code in AssociationWidget::updateAssociations().
> What I mean by "irregular role assignments":
> For generalizations, umbrello uses
> 	B --> A
> so B is the specialized class and A is the general class.
> However, for other associations (such as dependencies),
> umbrello uses
> 	A --> B
> I wonder why that is so - is it dictated by the UML standard?

I don't think so.  There may be a psychological difference,
generalisations are quite a different property from the other
associations (and the icon for generalisation is the other way around
too!).  But I can't think of any reason why it should be any different.

> I guess I'm not the only one who gets confused by this -
> the author of AssociationWidget::updateAssociations() most
> certainly did too :)
> So my question is, shall I make a global change to umbrello
> to use just one convention, i.e.
> 	A --> B,
> or shall I just add special-casing code to updateAssociations()
> to keep the different conventions apart?

If we're right and there is no real difference with generalisations
then the best thing to do would probably be to change it globally
(need to be careful with file compatibility I think) and the easiest
thing to do would probably be the special-case.


More information about the umbrello-devel mailing list