A novel method on software structure evaluation

The software is becoming much more complex and large, thus it is difficult to examine and evaluate the structure of software containing millions of lines of code and thousands of functions or objects. In large-scale software, the structure is one of the most important factors for people to design, develop and maintain trusted software. In this paper, we concentrate on dynamic analysis instead of static analysis and the Calling Network is proposed to represent the dynamic structure of software. Then, the community and hierarchical structure of the software are detected from the Calling Network using Newman Fast algorithm and Bi-Breadth-First Search algorithm. Finally, the Partition Matching Ratio is proposed to evaluate the rationality of software structure by comparing the theoretical community and hierarchical structure with the physical partition. In the experiments, we evaluate two Java applications, JPetStore and Jforum, and find the structure of JPetStore is better than Jforum which is same as the users' evaluation.