[Users] [SE-SCA] JBI considerations...and EIP toolbox integrated!

[vzurczak]

Sounds like there are mentions to the tooling.


fgardes wrote:
> For information, a little summary of the conclusion of the workshop initiated internally under the SCA on PEtALS.

Thanks for this. :)


fgardes wrote:
> 2-Today, we define component references at the design time (on the Eclipse SCA Designer) AND manually into the component Java implementation (with the annotation @Reference under an attribute with the same name of the reference into the .composite file.

In theory, use only one of the two should be enough.
But without the annotation, I think FraSCAti doesn't work correctly, probably because meta-model features lack. And if you declare things in the compposite, the SCA meta-model doesn't detect the references from the code. The second part will be soon available in STP. So references annotations will be enough. 

The other solution is interesting too as a modelling solution (for a full top-down approach with skeleton code generation).
Reference declarations should be available in FraSCAti. It is expected to be language independant and work even for languages who do not support annotations.
Annotations are only available in Java and widget implementations (the second one being only available in Tuscany). Other languages must be able to define references, and it goes through composite declaration or component types. If my remembers are good, the @ were required in FraSCAti because in SCA, declared reference names must match the name of the Java field. If it does not support, we will have to create a feature request.


fgardes wrote:
> if we want to manage dynamically EIP SCA Components, we have to be able to add this reference into the implementation dynamically too. 2 ways:
> At the Creation time. When you choose the name of the class and the interface to implement, it would be great to have an option (introspect references and add it as attributes). The implementation inherits from a provided by the EIP jar class which would have a method to catch all the references added.
> At the Packaging time. When you package the SCA su, a  registry would be added to the package, containing all the references to each component. The EIP jar class would use it to catch the references added.
> Or another way maybe of course ;)

Are you planning to use the SCA Designer for this, or rather use EID and project the design as an SCA application ?
In the case of the SCA Designer, let's make clear that having 50 references will result in a messy diagram. Clearly. For this kind of case, you will have the SCA Form editor (http://wiki.eclipse.org/STP/SCA_Component/SCA_Form_Editor), but it is not as visual as the Designer.
If you want to keep a visual editor for industrial use cases, keep EID (which needs committers to remain alive). Composites are clearly not expected to be 2 meter-high, with an infinite vertical scrolll bar.

Second, there is no simple way in SCA tools to graphically force reference ordering. Services and references do not even have their name displayed anymore on the diagrams (last two weeks upgrade). It was so messy otherwise. So, it's all about the user attention. But I got in touch with some people, here in EclipseCON, and they had interesting suggestions. I'd like to add a feature that generate the code skeleton (classes, SCA references, annotations if necessary) from a complete valid composite. So, basically, you would do your complete diagram, make some right-click actions, and got all your code generated, with some parts to complete by hand (interface methods to implement). That could help to prevent errors in reference ordering.

Now, there are some differences to apply for EIP.
Meaning that the applied pattern corresponds to a given library. See my second post below.


fgardes wrote:
> 3-The sorting of the references must be provided to be able to implement some EIP. So the order of creation (i.e. the order of the element node into the wsd:sequence into the .composite file) would have to be provided at the creation or packaging time.

OK... I guess the order is part of the EIP definitions (first one = first condition, ..., last one = default).




-------------------- m2f --------------------

Read this forum topic online here:
http://petals.ebmwebsourcing.com/forum/viewtopic.php?p=604#604

-------------------- m2f --------------------


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://forum-list.petalslink.org/pipermail/users/attachments/20110316/42d6ca72/attachment.htm>