Experiments in Run-Time Model Extraction

Reverse engineering and software evolution are probably today among the most challenging research areas in software engineering. They are also very rich fields for applying model engineering techniques. However, in order to define a sound model-driven reverse engineering framework, we need to understand more clearly the conditions under which various abstract models may be extracted from legacy systems and from other various software assets. The operation of extracting a model from a system is still very poorly understood. As part of several projects that have been performed in the INRIA ATLAS team, we have been extracting various kinds of models from various kinds of legacy systems like COBOL, Java, Smalltalk, Visual Basic, etc. Taking stock on this, we study in this paper some of the conditions that would allow partial automation of modelbased reverse engineering. More precisely we propose a method for metamodel-driven model extraction. The proof of concept is based on a recent experiment using Visual Basic 9.0 and several previous experiments done with the Squeak version of the Smalltalk language. Besides metamodel-driven model extraction, one of the research contributions of this work is to show the feasibility of extracting models not only from static systems but from dynamic systems as well. In certain ideal conditions, the models extracted from both situations may also be jointly used.