[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [stp-pmc] Fwd: [stp-dev] writing an introspector...
|
On 29 Aug 2006, at 13:40, Daniel Berg wrote:
The point I am talking about is providing intelligent content
assist "values" for tags.
What are the 'tags' here Dan - not sure exactly what you
mean.
This is only possible if you are processing the
actual source code and not just the binaries. The Tuscany
processor will be
using java.lang.Class to process the binaries. Within Eclipse we
should be
processing the annotation using AST.
Yes, agreed, we will need to process the source code to
make such an assist possible.
I agree that it may be easier to reuse the Tuscany code here but it
will not
be integrated well at all.
It may be a good place to start, so we can at least evaluate
exactly how we want to use the code, if at all. For a perfect
integration we would have to do a parallel implementation track
that is purposed directly to our codebase.
[deletia]
> There's two approaches I think -
>
> 1. let the tuscany code process the annotations and somehow
> consume the output
> 2. make an introspector that will process the annotations and
> populate the UI-specific model
>
> choice of approach will probably be governed by what the tuscany
> code does right now and how it could be reused.
<dcb>
I don't know what you mean by the UI-specific model in option 2. I
assume
it is the SCA model we have in core?.?
Yes that is what I mean.
I can guarantee that in option 1 the
Tuscany code can only update the actual serialized form of the SCA
XML files
and not the in-memory model (it is different than what is used in
Tuscany).
</dcb>
Yes - we can consume the SCA XML files that are generated from
Tuscany (if they are generated), but I there's also the possibility
that we could take the Tuscany *internal* model and transform it.
I don't know yet whether this is possible of course, I need to get
into the code base first.
> > I would also like to know which components within STP are blocked
> > by the models not being up to date.
>
> Nothing yet, but I'm sure this will change :)
<dcb>
I'm sure it will. I would like to see a plan that shows which
components
plan to use it and how.
</dcb>
We'll need something that will map composites to containers for
deployment,
so we'll need some code to allow the developer to construct that their
deployment profile over multiple connection profiles. This will involve
an introspection of the model elements, which is sensitive to their
structure. Also, we'll need to integrate popular service creation
methods,
including jaxws, tuscany, sca c++, sca php via the introspector
mechanism,
which again is sensitive to the structure of the models.
Let's do the planning together, shall we? I put out some very roughly
stated work items on stp-dev to pique peoples interest. I think that
we could break them down further on this list to produce a plan, what
do you think?
regards
--oh
