15
« on: June 23, 2005, 06:13:04 am »
Paolo,
> You may care to view
I see there that you had a question there like mine, recently.
> EA (for some reason) makes a distinction between an Association with associationKind (Aggregation: feature) set to other than none. UML doesn't make that distinction.
Yes: that's the way the situation seemed to me. I found that confusing. I thought I should check it with you, especially since I'm not especially familiar with EA nor with UML.
> The current code generation process doesn't handle the creation of the destructor body to correctly implement shared versus composite aggregation, but one day, hopefully will.
I see. Perhaps for now you can do this by explicitly including the various species of [Boost] auto-pointer in the model, so that there's one-to-one mapping from the model to the [template] classes which implement who destroys what?
In any case I'm using C# so destructors aren't such an issue for me.
> For my part, I try to make sure I've got the aggregation right in the model and I ONLY use Associations, setting the Aggregation feature appropriately (on the appropriate end - as required). A bit more work, but safer. As you've seen, reverse engineering sets to Association as does XMI import (from most other tools).
Got it. Thanks for confirming that what I'm seeing is as-expected, and for telling me how you work-around that.
I found it especially confusing that the direction of the connector (not only the type of the connector) changes too: e.g. an aggegation from containee to container, or an association from container to containee.
Thanks again,
Christopher