Evaluating the performance of model transformation styles with Maude

Rule-based programming has been shown to be very successful in many application areas. Two prominent examples are the specification of model transformations in model driven development approaches and the definition of structured operational semantics of formal languages. General rewriting frameworks such as Maude are flexible enough to allow the programmer to adopt and mix various rule styles. The choice between styles can be biased by the programmer’s background. For instance, experts in visual formalisms might prefer graph-rewriting styles, while experts in semantics might prefer structurally inductive rules. This paper evaluates the performance of different rule styles on a significant benchmark taken from the literature on model transformation. Depending on the actual transformation being carried out, our results show that different rule styles can offer drastically different performances. We point out the situations from which each rule style benefits to offer a valuable set of hints for choosing one style over the other.