Hey,
On my blog Ted asked about mixin selection overriding. Today this can be
done like this:
@Mixins(ListablePersonMixin.class)
interface PersonEntity
extends Listable, PersonEntity
{}
with a customization being:
@Mixins(CustomListablePersonMixin.class)
interface CustomPersonEntity
extends PersonEntity
{}
If both mixins implement the same interface, then the custom one will
have precedence.
Teds question was, however, that if you want to replace implementation
at assembly-time, how do you do it? The above works, but a more flexible
way would be to use something similar to .withConcerns() in the assembly.
So, instead one would do:
module.addEntity(PersonEntity.class).withMixins(CustomListablePersonMixin.class);
Pro's:
* Give us a very flexible and open-ended assembly-time mechanism
* No subclassing needed for customization. Everything can be done
programmatically
Con's:
* You can't look at the Composite type and know how it will work. We are
getting closer to AOP confusion here. But we've sort of already taken
that step with the .withConcerns(), so this would only be another
"logical step on the path to hell".
* You therefore NEED the visualizer to figure out what your system looks
like. The upside is that I think Tonny will do a great job with it!
The extreme case of all of this is that the Composite type ONLY declares
the interfaces and none of the mixins, which are then supplied using
.withMixins(). Superflexible and quite unreadable, but possible.
On a related note, I'm in another discussion thread on another forum
with Trygve Reenskaug (the MVC guy), and he wants to remove as much
polymorphism as possible, for the sake of readability (i.e. you see a
method and know what it will do, because there's only one
implementation). This feature would go against that desire.
WDYT? Is it worth it? What's the usecase?
/Rickard
_______________________________________________
qi4j-dev mailing list
[email protected]
http://lists.ops4j.org/mailman/listinfo/qi4j-dev
