GMF: A Model Migration Case for the Transformation Tool Contest

Modeling languages and thus their metamodels are subject to evolution [2]. When a metamodel is adapted, existing models may no longer conform to the adapted metamodel and thus need to be migrated. Model migration is a special case of exogenous model transformation [7], since original and adapted metamodel are usually different from each other. However, the metamodel versions also share some similarity, as the metamodel is usually not completely changed during metamodel adaptation [14]. Consequently, migrating transformation definitions usually contain identity rules for the unchanged metamodel parts. To remove this boilerplate code, different approaches have been proposed [13]. Manual specification approaches—like Sprinkle’s language [14], MCL [8] and Epsilon Flock [11]— extend transformation languages so that they automatically copy model elements that are unaffected by metamodel adaptations. Operator-based approaches—like Ecoral [15] and COPE [4]—provide reusable operators that encapsulate recurring metamodel adaptations and model migrations. Metamodel matching approaches—like Cicchetti’s approach [1] and AML [3]—automatically derive a transformation definition from the difference between two metamodel versions. The existing approaches mostly use or extend existing model transformation languages and tools. To compare the different ways in which model migration can be defined, we propose a real-life case from the evolution of the Graphical Modeling Framework (GMF). The case is already well-researched, as it has been used in an empirical [5] and a comparative study [10]. It exhibits a number of differences to last year’s migration case [12]: (1) Most of the metamodel remains the same which is more typical for model migration. (2) The migration is more complex and thus requires more expressive transformation languages. (3) GMF provides a reference migrator implemented in Java which defines the migration semantics. (4) GMF provides a number of test cases which can be used for validating the solutions.