Reverse Engineering of Design Patterns from Java Source Code

Recovering design patterns can enhance existing source code analysis tools by bringing program understanding to the design level. This paper presents a new, fully automated pattern detection approach. The new approach is based on our reclassification of the GoF patterns by their pattern intent. We argue that the GoF pattern catalog classifies design patterns in the forward-engineering sense; our reclassification is better suited for reverse engineering. Our approach uses lightweight static program analysis techniques to capture program intent. This paper also describes our tool, PINOT, that implements this new approach. PINOT detects all the GoF patterns that have concrete definitions driven by code structure or system behavior. Our tool is faster, more accurate, and targets more patterns than existing pattern detection tools. PINOT has been used successfully in detecting patterns in Java AWT, JHotDraw, Swing, Apache Ant, and many other programs and packages