Rule-based Model Extraction from Source Code

In the context of an approach for reengineering legacy software systems at the architectural level, we present in this paper a reverse engineering methodology that uses a model defined as a type graph to represent source-code subject to a code categorization process. Two alternative methods for referencing the source code are discussed: native vs. graphical. To represent the code, the native representation uses the abstract syntax tree while the graphical uses a programming language metamodel. Two options regarding the way that the graph can relate to the source code reference model are also considered: association model vs. direct link. The extraction of the program representation, complying to the type graph, is based on rules that categorize source code according to its purpose. The techniques to address this process, such as the code categorization rules, are shown together with examples.