Improving Behavioral Design Pattern Detection through Model Checking

Recovering design pattern instances in a software system can help maintainers to understand its design and implementation. In this paper we present a fully automated design pattern recovery approach that analyzes the behavior of pattern instances both statically and dynamically. In particular, the proposed approach exploits model checking to statically verify the behavioral aspects of design pattern instances. To this end, we encode the properties defining the correct behavior of a pattern as LTL (Linear Temporal Logic) formulae and the sequence diagram representing the possible interaction traces among the objects involved in the candidate instances as PROMELA specifications. To verify whether the LTL properties are satisfied by the candidates we employ the SPIN model checking tool. The dynamic analysis of the pattern behavior is performed through a code instrumentation and monitoring phase applied on the candidate pattern instances. This phase allows us to obtain actual dynamic data during program execution, which is then used to verify its compliance to the pattern definition. The effectiveness of the proposed approach is shown by presenting and discussing the results obtained on JHotDraw and JRefactory.