# General (#1) - Transformation problems (#28) - Message List

Not entirely a mediniQVT question, more of a general QVT problem I'm having.

My situation is as follows: I want to perform a simple model-to-model transformation, where the input and output model use the same meta-model. Within this meta-model, I have the following construction:

`<a>`

`<b>`

`<one or more objects of type a>`

`</b>`

`<c>`

`<one or more objects of type a>`

`</c>`

`<d/>`

`</a>`

So: class a is the superclass of b, c and d. Now b and c *c*an have variables of type a, so, these may also be of b, c and d. I'm using rules like the following one:

relation b2b { checkonly domain in b1 : b{}; enforce domain out b2 : b{}; where { a2a(b1, b2); } }

This works for the top level classes a, b, c and d, but the 'variable objects' are not evaluated.

My question is: is the structure I'm using for my relation rules a good one for this type of problem, or should I use an other one? If it is correct, how can I convince QVT that the variables also should be evaluated?