KP, could you post diagram? 
I have read this superstructure sentence, but without example and explanation it doesn't help a lot
BTW, if you choose other ends, you can get not classA with the left classes, but classC with the rest (classC-classA, classC-classB & etc)
Hi SF_lt (can we have a more human moniker for you?
),
If you are familiar with ERD or DB design, the best way to think about the Lozenge (in my view) is as an Intersector Table, with a mandatory, singular, foreign key to each of the associated Classes (as Tables). Therefore the UML statement KP (another non-human?
) quoted merely states that if you run a query over the lozenge where you supply values for each of the foreign keys but one, you'll get the set of rows where the value of the remaining foreign key varies. The quote goes on to say that if the end is ordered or unique, this will apply the same constraints on the set of values returned by the query. Does that help?
Also, one little specification on constraints: what I miss, I miss feature to add constraint for the dotted line from association class to association line - there is no way to attach constraint or smth else to this part of association. Maybe UML2 metamodel prohibits it...
The dotted line is not a
real connector. It is merely trying to reinforce the idea that the Association Class
IS the Association line. It is a
rendering artifact, not a real UML element. Therefore you can't attach any metadata to it. That's why Thomas said use the Constraints tab on the AssociationClass or Lozenge.
FWIW, I can't reinforce enough that Sparx needs to understand that the lozenge shape is merely a rendering of the AssociationClass (like say the Boundary, Control and Entity shapes from the Analysis toolkit). Also, that the AssociationClass defined by UML is merely binary example of the N-ary Association. So a binary Association may be depicted by an Association line by itself, an Association line with associated AssociationClass or an Association lozenge with two AssociationEnds - they are all the same thing!
Accordingly, you should be able to do everything you can do to a Class to the Lozenge!
jeshaw, I agree, that road doesn't need to know, what cars use it (what about women? :-D )
in my example, channel passes signals from sender to receivers, also signals travel through the channel to reach their's destination. Also not all signals use one or other channel - there may be many channels.
But still, thanks for your point
By Signal here, do you mean Signal Type? If you mean Signal, what is the domain identifier that allows you to specify which signal instance you are associating. From my intersector table example above, you'll see you need to provide the foreign key to identify each AssociationEnd. IF you can't find one, then it's an indication the model is wrong.
It strikes me that this would be a good example to work through for AssociationClasses and lozenges etc. Like Jim, I'm not so sure there is an N-ary relationship here.
To determine how to model the rules, we need to know the rules. Could you give us a short definition of each of the Classes involved: Sender, Receiver, Channel and Signal? Without them, we can't really provide the direction you are seeking. Nothing fancy (as they say in the competitions - 25 words or less
)
If you've been following some of my postings, you'll see I strongly favour a technique called "Verbalisation" - where you "speak out" the model. If the words sound wrong, the model is wrong!
HTH,
Paolo
[size=0]©2005 Paolo Cantoni, -Semantica-[/size]
