Composite Structure Diagrams, etc.

[Dave.B]

Is there any model integrity maintained by EA WRT to a class model that explicitly composes a class from other classes and a composite structure diagram of that class that seeks to give each of the compositional classes a named role?

E.g. {Class model in words:} A Car is composed of an Engine and it has 2 "front" Wheel and 2 "rear" Wheel. The Engine is associated to 2 Wheel by an "axle".
The composite structure diagram for the structured Car class can show that it has an e:engine part and two independent references (dashed boxes) to front:Wheels[2] and rear:Wheels[2] parts. Moreover the connection between the e:engine part and the front:Wheels[2] part can be made explicit as a:axle.
(This is taken from Tom Weilkiens Book "System Engineering with SysML/UML", fig's 3-24 and 3-26.)

Now it is easy to draw this in EA as a class diagram and a composite structure diagram, although the realisation of the class diagram's "axle" association as the CSD's a:axle connection is awkward.

However, EA does not seem to maintain any integrity between the class model's associations (and thereby the classes links) and the CSD. This is typified by the fact that parts must be manually determined to be "reference" parts (dashed borders) by setting the parts custom properties as opposed to letting the class models association aggregation status drive this automatically (anything other than a black diamond composition is a reference part).

In SysML I think that you are supposed to use the Block Definition Diagrams to show what blocks are aggregated together and then to use an Internal Block Diagram to reveal that compositional arrangement of these aggregated blocks. I'm assuming that UML's structured classes work the same way. In this respect it seems that EA show be able to check and even enforce the models integrity. Or do I have this all wrong?

Finally as an aside, what is the difference between a structured class on a class diagram and a structured class on a CSD? Does the CSD simply act as a hierarchical diagram for revealing a class as a structured class?

Regards
Dave B.

[«Midnight»]

I like the way the book approaches this (and other) issues too!

The answer is probably that EA just hasn't gotten there. I hope that the reality is that EA is not there yet and that it will happen. There's not been too much activity on the SysML side of EA since the add-in first came out, so whether we'll see what we need is anyone's guess.

Whether the support for composition tracking (for lack of a better term) would need to be embedded in (EA's) core UML engine is another matter. If it would, then I don't know if this would be in the cards any time soon. There are several of these 'high value' features that Sparx seems to be immune from considering...

[Of course, that's just my opinion, so take it with the appropriate seasoning.]

David