Model Differencing for Textual DSLs

The syntactic and semantic comparison of models is important for understanding and supporting their evolution. In this paper we present TMDIFF, a technique for semantically comparing models that are represented as text. TMDIFF incorporates the referential structure of a language, which is determined by symbolic names and language-specific scoping rules. Furthermore, it employs a novel technique for matching entities existing in source and target versions of a model, and finds entities that are added or removed. As a result, TMDIFF is fully language parametric, and brings the benefits of model differencing to textual languages.