CrocoPat: A Tool for Efficient Pattern Recognition
          in Large Object-Oriented Programs

Dirk Beyer and Claus Lewerentz
Software Systems Engineering Research Group
Technical University Cottbus, Germany
Computer Science Reports I-04/2003, January 2003

Abstract:  Nowadays, software systems are too large to be
understandable by reading the source code. For reengineering
activities, methods and tools for automated design recovery are
needed. CrocoPat is a tool for efficient pattern-based design
analysis of object-oriented programs. Patterns can be flexibly
specified by expressions based on standard mathematics. The software
meta model is interpreted in terms of relations, and the patterns are
described by relational expressions over these relations. The tool
represents the abstract model of the program using a data structure
based on binary decision diagrams for performance improvement. The
representation is proved to allow for an efficient recognition also
for large systems up to 10'000 classes comprising several MLOC source
code.