Cunha A, Macedo N, Guimarães T.
2014.
Target oriented relational model finding. 7th International Conference on Fundamental Approaches to Software Engineering - FASE. 8411
AbstractModel finders are becoming useful in many software engineering problems. Kodkod is one of the most popular, due to its support for relational logic (a combination of first order logic with relational algebra operators and transitive closure), allowing a simpler specification of constraints, and support for partial instances, allowing the specification of a priori (exact, but potentially partial) knowledge about a problem's solution. However, in some software engineering problems, such as model repair or bidirectional model transformation, knowledge about the solution is not exact, but instead there is a known target that the solution should approximate. In this paper we extend Kodkod's partial instances to allow the specification of such targets, and show how its model finding procedure can be adapted to support them (using both PMax-SAT solvers or SAT solvers with cardinality constraints). Two case studies are also presented, including a careful performance evaluation to assess the effectiveness of the proposed extension.
Macedo N, Cunha A, Pacheco H.
2014.
Towards a framework for multi-directional model transformations. 3rd International Workshop on Bidirectional Transformations - BX. 1133
AbstractThe Query/View/Transformation Relations (QVT-R) standard for bidirectional model transformation is notorious for its underspecified semantics. When restricted to transformations between pairs of models, most of the ambiguities and omissions have been addressed in recent work. Nevertheless, the application of the QVT-R language is not restricted to that scenario, and similar issues remain unexplored for the multidirectional case (maintaining consistency between more than two models), that has been overlooked so far.
In this paper we first discuss ambiguities and omissions in the QVT-R standard concerning the mutidirectional transformation scenario, and then propose a simple extension and formalization of the checking and enforcement semantics that clarifies some of them. We also discuss how such proposal could be implemented in our Echo bidirectional model transformation tool.
Ours is just a small step towards making QVT-R a viable language for bidirectional transformation in realistic applications, and a considerable amount of basic research is still needed to fully accomplish that goal.
