Additional features for the QVTo language should be raised as OMG issues
at OMG.
Additional features for the QVTo tooling should be raised as Bugzillas,
but do not expect much response until we have migrated to underlying
Xtext-based tooling. We are very short of resources, so if you want to
help ...
Regards
Ed Willink
On 26/11/2012 13:36, Toni Siljamäki wrote:
> Hi there.
> Should any additional QVTO features be requested in Bugzilla?
> /Toni
>
Ciao.
Right now I'm not in the position (nor having the time/energy, at this point) to bang my head at some OMG-wall, trying make what I'm proposing to happen.
But I'm very interested in getting the very nice/powerful QVTO language/tooling even more competitive by introducing also M2T capabilities in a quite simple way.
The current QVTO language already supports 2 keywords _not_ listed in OMG's QVT spec, so who knows...
Perhaps it's a relatively "simple thing" to create the ultimate M2M+M2T language+tooling for EMF-/Ecore-based models.
Watch the Bugzilla space, where I soon will post a proposal on this thinking, just presented internally.
Perhaps all that's needed is a couple of skilled thesis guys having the "right" interest, and then magic may happen...
Regards/Toni
OMG has MOFM2T and Eclipse has the Acceleo implementation thereof.
I see zero prospect of a composite M2M+M2T based on the concrete syntax
QVTo and probably even less based on QVTc or QVTr.
MOFM2T is a much more promising starting point; you have a choice of
whether to add M2M to control-within-text or to text-within-control.
You could also look at Xtend which potentially offers
control-within-text-within-control, but when I tried it recently I was
very uncomfortable.
It seems to me that the basic ergonomics of mixed M2M and M2T are under
researched and so you should study these first and only then see whether
they can be an extension of an existing language. That done you might
then define an alternative concrete syntax for only slightly enhanced
abstract syntaxes of QVTo or MOFM2T or ....
Regards
Ed Willink
On 26/11/2012 18:56, Toni Siljamäki wrote:
> Ciao.
> Right now I'm not in the position (nor having the time/energy, at this
> point) to bang my head at some OMG-wall, trying make what I'm
> proposing to happen.
> But I'm very interested in getting the very nice/powerful QVTO
> language/tooling even more competitive by introducing also M2T
> capabilities in a quite simple way. :)
> The current QVTO language already supports 2 keywords _not_ listed in
> OMG's QVT spec, so who knows...
> Perhaps it's a relatively "simple thing" to create the ultimate
> M2M+M2T language+tooling for EMF-/Ecore-based models.
> Watch the Bugzilla space, where I soon will post a proposal on this
> thinking, just presented internally.
> Perhaps all that's needed is a couple of skilled thesis guys having
> the "right" interest, and then magic may happen... :)
> Regards/Toni
>
Thanx for your comment.
Have been doing M2M+M2T model compilers like-a-100%-working-time since before Y2K.
...using a model translaton technology available long before UML/MDA/MOF/EMF/Ecore whatever, supporting M2M+M2T, still available, soon semi-open-source...
I have a pretty good idea what I would appreciate to see in a soon Bugzilla-proposed QVTO+ and _before_ I retire.
1) I _have_ run/evaluated the MOFM2T-in-Eclipse you mention above, which is why I will post this QVTO+ suggestion on Bugzilla. (along with bug-reports on other tools)
2) We _can_ achieve a M2M+M2T half-measure-workaround ourselves using "black-boxing" in QVTO, but I would appreciate a more 1st-class solution.
Regards/Toni