[news.eclipse.modeling.m2m] Re: [ATL] Merging ATL transformations?

[Dennis Wagelaar <dennis.wagelaar@xxxxxxxxx>]

Matthias Bohlen schreef:
> On 2007-03-19 11:59:59 +0100, Dennis Wagelaar 
> <dennis.wagelaar@xxxxxxxxx> said:
>
> > Matthias Bohlen schreef:
> > > Hi all,
> > >
> > > I have several small re-usable ATL transformations which I'd like to 
> > > combine into one large transformation. The combined transformation 
> > > should have the union of all metamodels in the "create" list and 
> > > helpers in the "uses" list and the union of all transformation rules.
> >
> > Hi Matthias,
> >
> > (This is probably starting to bore you:) you can write several small 
> > transformation modules and then superimpose them in any order, 
> > provided that there's no rule overlap.
>
> No, it is not boring me at all. :-)
>
> Example: When I have transformations A, B and C, can I really have a 
> transformation D which simply contains the following additional line:
>
>    uses A, B, C;

Well, you won't have a physical transformation "D", but the result of 
superimposing A, B and C on top of eachother is equivalent.

I suppose it's best to experiment a bit with superimposition, just to 
get to know what it can do...

EMF meta-models can be "merged", or rather linked, by defining a new 
meta-model that references elements from the other meta-models. For 
example, you can make an "Import" abstract meta-class, that inherits 
from at least one meta-class from each meta-model you want to reference. 
ATL then loads the whole network of linked EMF meta-models whenever you 
load the top-level meta-model.

Combine that with the solution I previously wrote about of referring to 
your meta-model with always the same symbolic name (e.g. "MM"), and you 
can combine rules on different (parts of the) meta-models within one 
transformation.

> This would be really easy! Should this work today, already?

It should work with ATL as it sits in CVS, yes.

> Cheers,
> Matthias at andromda.org

--
Regards,
Dennis