[news.eclipse.modeling.gmt.amw] Re: [ATL] UserInteraction in Transformat

[Frédéric Jouault <frederic.jouault@xxxxxxxxxxxxxx>]

Hello,

> First of all, thanks for your help.

You are welcome.


> I read something about AMW, but I don't think that I would get it right
> with that.
>
> The problem of capturing an AMW Model with the UserDecisions would be,
> that for the same Input SQL2003Type the user could decide once to use
> ORCLTypeA and for the second ocurrence of the SQL2003Type to use
> ORCLTypeB. But with AMW the user could only match one ORCLType for the
> whole woven transformation.

If you weave your decisions on the metamodel, you would indeed have the 
problem that you mentioned: the decisions would apply in the same way to 
all the occurences of a given type.

However, if you weave your decisions to the source model (instead of 
metamodel), then you should be able to achieve one decision per occurence.

I am CCing the AMW newsgroup because this discussion is now about AMW ;-).


> Are there any other possibilities to capture the UserDecisions?
> Otherwise I think I have to move the whole transformation process to the
> Java layer.

There are other possibilities like specifying your own operations in 
Java, and call them from the ATL transformation. In such a situation, 
you do not need to write the whole transformation in Java, only the user 
decision part.

However, I do not consider such a scheme as a clean model engineering 
solution. My advice is still to explore the possibilities offered by 
AMW. I am indeed confident you can achieve what you want with it.


Regards,

Frédéric Jouault