Solving the Flowgraphs Case with Eclectic

The TTC 2013 Flow Graphs case [3] proposes the analysis of Java programs, conforming to the JaMoPP meta-model [2], by transforming them into a language-independent meta-model which represents the structure of the program and includes information about control and data flows. This solution makes use of the Eclectic transformation tool [1] to solve the four proposed tasks. The fourth task has been addressed using METADEPTH [4] to create a small DSL. 1 Eclectic is a transformation tool designed as a family of model transformation languages, that is, a set of transformation languages each one specifically designed to address a specific transformation concern, as well as some composition mechanisms for their combination. The objective of this solution is thus to show how it is possible to address a non-trivial transformation task, such as this case, using several languages and how this approach has the potential of improving modularity and readability. Eclectic currently provides the following languages: i) a mapping language for establishing one-toone and one-to-many correspondences, ii) a target-oriented language with object notation and explicit rule calls, iii) a traversal language based on in the idea of attributed grammars, iv) a pattern matching language which used object-notation, and v) a lower-level scripting language, which also plays the role of scheduling language. Languages i, ii and iv do not allow complex expressions, but these need to be encoded in navigation libraries, written in the scripting language. In principle, the combination of these languages permits covering many model transformation scenarios, in a more intentional way than using a general purpose transformation language. Addressing case studies could allow this intuition to be evaluated in practice. The solution of this case has used the mapping language, the attribution language, the pattern matching language, the scripting language and a navigation module. The target-oriented language is not needed because it is typically useful for synthesis tasks, but the case only involves mappings and analysis tasks. In Eclectic every language is compiled to an intermediate representation, called IDC. It provides primitive instructions for model manipulation. Then, IDC is compiled to the Java Virtual Machine (JVM) bytecode format. In this way, all Eclectic languages share the same execution infrastructure. The composition mechanisms are implemented at the IDC level. There is also a runtime library, which provides datatypes (e.g., immutable lists), a model manager (i.e., EMF and METADEPTH are supported), etc.