RE: [mdt-papyrus.dev] Custom code and code generation

[Jerome BENOIS <jerome.benois@xxxxxxx>]

Hi Gabriel,

Please find below my response.

Le vendredi 09 janvier 2009 à 14:06 +0100, Gabriel Merin Cubero a
écrit :
> Hi everyone,
> 
> First of all I would like to say that the decision of changing the M2T
> language cannot be taken just because the virtues of the code editor.
> If we think that the Xpand2 editor has too many drawbacks maybe a
> collaboration with them would be interesting. A comprehensive
> comparative should be done with the advantages and disadvantages of
> each language. XPand2 already solves all the generative needs, their
> templates can be overridden and even the metamodels can be extended in
> a non intrusive way, does MTL improve or expand the functionality
> commented before?
I think MTL provide similar OO features than Xpand except AOP.
A little history, the C and Java code generators initially included in
PapyrusUML was wrote in Acceleo. And Acceleo is very similar as MTL.
Thus the question is :

Who already know Acceleo or MTL ? / Who already know Xpand in Papyrus
Team ?

The simplicity of the tool and the power of editor is very important for
me. Indeed, with a good tooling, we will be more productive to develop
templates.

Also, Eclipse Papyrus is focused on OMG standards (UML ans SysML), thus
it's seems logical to emphasize the usage of the MOF to Text Language
standard to develop Eclipse Papyrus.

> 
> Besides, we should really take into consideration that MTL is a very
> young language. In fact, it won't be "officially born" until Galileo.
> Therefore, using it now would be risky.
Besides, I agree MTL is a very young language and it's an incubation
project. Xpand is also an incubation project and the version use by
GMF.CodeGen is a fork of Xpand provided by org.eclipse.gmf.xpand*
plugins. And Xpand wasn't include in Ganymede. Also MTL is mostly
developped by my Obeo colleagues and i can be the bridge between Papyrus
team and MTL team to insure support for any problems.

> On top of all, using two languages to do the same (with all the
> learning involved) doesn't make too much sense and, I think we will
> all agree, it would be counterproductive to change all the GMF
> templates in order to use MTL. And, therefore, if we are not planning
> to alter the generative procedure of GMF it is necessary to learn
> XPand2 to customize the generative procedure.
So, I precise my proposal, it's just use MTL to override Xpand templates
to solve quickly our generation problems and not rewrite all GMF.Codegen
in MTL.

Cheers,
Jérôme.