Title: Nachricht
|
Hi
Martin, hi Wouter,
the
Hibernate cartridge also uses EJB session beans for the <<Service>>
classes. The <<Entity>> classes are implemented with Hibernate. This
is why you can use the web tier as-is (see my car rental sample). There is no
difference in the service API between the EJB and the Hibernate cartridge,
therefor, there is only one set of web tier classes in the car rental
sample. Of course, internally, there are two implementations of the service
session beans which behave differently because they talk to different entity
implementations.
You
asked in your question whether it makes sense to do a three-step generation
after separating different architectural aspects into different cartridges.
The truth is: you can even do a one-step generation using multiple
cartridges in the same AndroMDA run. However, only one cartridge should
respond to the same aspect (persistence, web workflow and so
on).
Example: The car rental system uses two AndroMDA runs: 1) with EJB,
Struts and Java cartridge activated, 2) with Hibernate, Struts and Java
cartridge activated.
HTH...
Matthias
hi Martin,
the cartridges are
already doing that and have been designed with this level of abstraction in
mind
I don't see for example how you can not plug the Hibernate
cartridge under your Struts instead of the EJB cartridge,... but since it is
another technology alltogether you will have to write some adjustments
to properly take care of transactions, security, caching etc... this is
inevitable since they do not support it (yet) in the same fashion
it is
indeed best to split up you project so that you can work on each module
separately, and you're right, the frontend always depends on the backend
-consider the services you expose an API towards the client developer- that's
normal unless you start writing independent interfaces between each tier, but
I don't see much advantage in doing so (will you actually use them for what
they have been written?)
I hope this clears it up for you
best
regards
Wouter.
MArtin Schumacher wrote:
Hi everyone,
during my diploma thesis we chose androMDA to support MDA in our company.
These days we exctract our architecture from the existing application. Now I
wanna create the cartidge for our architecture. Now we got the idea to
seperate the cartriges for web-, application- and persistance-layer and to
do a three-step-generation. Doing it this way, it is quite easy to change
the i.e. the persistance. Another advantage is, that every layer of the
architecture can be developed and maintained by its own. The disadvantage
is, that the cartriges depend to eachother...
Do you think its reasonable to do it this way?
Am I right, that the standard cartridges of andromda are already prepared to
be combined to eachother?
MArtin
|
