Test-driven Language Derivation with Graph Transformation-Based Dynamic Meta Modeling

Abstract

Deriving a new language L_B from an already existing one L_A is a typical task in domain-specific language engineering. Here, besides adjusting L_A's syntax, the language engineer has to modify the semantics of L_A to derive L_B's semantics. Particularly, in case of behavioral modeling languages, this is a difficult and error-prone task, as changing the behavior of language elements or adding behavior for new elements might have undesired side effects.

Therefore, we propose a test-driven language derivation process. In a first step, the language engineer creates example models containing the changed or newly added elements in different contexts. For each of these models, the language engineer also precisely describes the expected behavior. In a second step, each example model and its description of behavior is transformed into an executable test case. Finally, these test cases are used when deriving the actual semantics of L_B - at any time, the language engineer can run the tests to verify whether the changes he performed on L_A's semantics indeed produce the desired behavior.

In this paper, we illustrate the approach using our graph transformation-based semantics specification technique Dynamic Meta Modeling. This is once more an example where the graph transformation approach shows its strengths and appropriateness to support software engineering tasks as, e.g., model transformations, software specifications, or tool development.