Aggregation Modelling

[Cosmo Nova]

Hi all, I was trying to draw a class diagram, where a class owns some other classes... and thus I applied the aggregation relations. I got TWO questions about this aggregation:

1. When I generated code from the class diagram, it created a poiner as member variable in the owner class. I got the same result no matter it's an aggregation or composition. How can I differentiate them in code generation?

2. When I use reverse engineering to generate class diagram, from the code just generated, I got a different diagram... The main problem is I lost ALL the aggregation relations, and they becomes normal "association" between the two classes.. The code reverse engineering supposed to be very useful when I modified the code and update the diagram through the code. How can I preserve the aggregation relations through settings or whatever?

EA is a great tool! Thanks a lot!

[«Midnight»]

Hey Cosmo,

There was some very good information on EA's implementation of these relationships quite a while back. The threads in question were started - I think - by Paolo Cantoni (in one of his incarnations, you'll have to look him up).

One point was that aggregation and composition relationships worked a bit differently depending on how you created them. This appeared to be a throwback to the UML 1.x days of EA. [I do not know if or to what extent this has been addressed. It might not be your problem here. So read on...]

Meanwhile, I have not done this for a while so I might not remember, try setting the Containment property of each end of the aggregation. You'll find this on the Target - and Source, but this is less likely to matter - Role tab of the Aggregation Properties dialog. [You can also set these through the API, and (I think) a script.]

If that does not work right away, grab a piece of your own source code - something you did not build with EA - and reverse engineer it into an EA project. Take a look at what properties EA sets, and to which values.

Any luck?

HTH, David

[jeshaw2]

Cosmo

You might also look at the navigation direction on the association.  Reversing what you see might help.

[Cosmo Nova]

Hey Cosmo,

There was some very good information on EA's implementation of these relationships quite a while back. The threads in question were started - I think - by Paolo Cantoni (in one of his incarnations, you'll have to look him up).

One point was that aggregation and composition relationships worked a bit differently depending on how you created them. This appeared to be a throwback to the UML 1.x days of EA. [I do not know if or to what extent this has been addressed. It might not be your problem here. So read on...]

Meanwhile, I have not done this for a while so I might not remember, try setting the Containment property of each end of the aggregation. You'll find this on the Target - and Source, but this is less likely to matter - Role tab of the Aggregation Properties dialog. [You can also set these through the API, and (I think) a script.]

If that does not work right away, grab a piece of your own source code - something you did not build with EA - and reverse engineer it into an EA project. Take a look at what properties EA sets, and to which values.

Any luck?

HTH, David

thx ur reply

I found this thread, but is it what u were trying to quot??
http://www.sparxsystems.com/cgi-bin/yabb/YaBB.cgi?num=1203597016/8#8

for the Containment property, what exactly it represents in a class modelling?  :-/

[«Midnight»]

Cosmo,

That's who I was referring to, but not the particular post.

Try Jim's idea first; it's quick and might do the trick.

If not, just play with containment a bit. If it does not solve your problem then you can leave the definition until later.

If you're still having problems, try importing some code that already does what you want, as I suggested above. Examine how EA models the result, and then try creating a model that is similar. Generate the code and see if that's what you wanted.

David

[Eve]

EA doesn't generate anything different for compositions or aggregations.  The pointer is because of the containment specified.

The reverse engineering sounds odd, but I don't really have enough information to know what you're talking about.