Software Networks of Java Class and Application in Fault Localization

Complex networks are backbones of complex systems. A lot of empirical analysis demonstrates that software is kind of artificial complex systems that expose the small-world effects and follow scale-free degree distribution. Here analyzed is Java class complex networks construction in binary file with BCEL (Byte Code Engineering Library) of Apache Jakarta Project. This method needn´t source code to generate networks, so it can extract all java software´s inner structure to assist programmer to understand software macroscopic features. Existing lectures almost verified variety of software networks are complex networks, and some investigators have concluded software networks cannot yet produce factual instruction in software engineering. Here we utilize log information in Java networks to diagnose software fault and exception. Experiments show that software networks can not only visualize the software structure, but also really instruct software fault localization.