[news.eclipse.modeling.gmt.amw] Re: [AMW] ATL and xmiID

[Brice Pajot <bpajot@xxxxxxxxxxx>]

Hi Marcos,

I used the second solution, since regarding the first solution, I do not 
know how to link the elements of B and C in W once the first ATL 
transformation is made.

The second solution using an endpoint rule did the work well. Even if it 
 is less elegant, I can not find an easy way to perform the first 
solution, but maybe I miss the point.

Thanks a lot for your help,

Regards,

Brice.

> Hi,
>
> This issue is probably related to the following post:
>
> http://www.eclipse.org/newsportal/article.php?id=186&group=eclipse.modeling.gmt.amw#186 
>
>
> When you set up the XMI:IDs, it takes the current value of the element C 
> in the output model. But, if the C element is not yet assigned to a 
> parent table, you can have an id similar to "//@columns.3". This value 
> changes when the parent table of the column is set up. However, the 
> "ref" has already the wrong value.
>
> There are two alternatives:
>
> - Split the transformation into 2 transformations. The first one only 
> produces A and C. The second one takes as input B and C and it creates 
> W. This is the most elegant solution, since you avoid reading values 
> from the output models.
>
> - The "ref" value should be set up only when all the elements of the 
> output models are already produced.
>
>
> Regards,
>
> Marcos.
>
> Brice Pajot wrote:
> > Hi,
> >
> > I try to use ATL to perform such kind of transformation:
> >
> > Input model : A
> > Output models : B, W, C
> >
> > where B and C are "linked" by the weaving model (for further 
> > information, B is a class metamodel with attribute elements, and C is 
> > a table with column elements, W is linking attributes to elements).
> >
> > Currently, I set the B.attributes and C.columns refs with target 
> > __xmiID__. However, whereas it was working perfectly, it began to fail 
> > when I modified the C meta-model (I just added a composite link on the 
> > column element), and thus the ATL transformation respectively. Failure 
> > is identified with broken links in the weaving navigator. When I look 
> > at the file in text mode, the references are indeed not uniform 
> > regarding the C elements (table columns).
> >
> > To illustrate the problem, here are the links before the changing of 
> > the C model:
> >   <rightModel xsi:type="ModelRefXMI" name="Data Table" 
> > ref="/FirstExample/SVF/Models/DataTable.ecore">
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.0"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.1"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.2"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.3"/>
> >   </rightModel>
> >
> > and after:
> >   <rightModel xsi:type="ModelRefXMI" name="Data Table" 
> > ref="/FirstExample/SVF/Models/DataTable.ecore">
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.0"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.1"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="/0/@columns.2"/>
> >     <ownedElementRef xsi:type="ElementRefXMI" ref="//@columns.3"/>
> >   </rightModel>
> >
> > I do not understand the problem. I also tried to set the __xmiID__: If 
> > I set it manually (for example, by setting it with the related input 
> > __xmiID__), the weaving model does work in the navigator. But this is 
> > of course not a solution! I also tried to use the 
> > setElementIDbyRefType, but without any success (maybe I do not use it 
> > in the correct manner?).
> >
> > Have someone any idea of the problem? I'm certainly doing something 
> > wrong but I can't figure out the problem.
> >
> > Brice.