On the classification of cyclic dependencies in Java programs : a thesis presented in partial fulfilment of the requirements for the degree of Master of Science in Computer Science at Massey University, Manawatū, New Zealand

Abstract

Software engineering guidelines and rules discourage cyclic dependencies between modules, yet empirical studies have shown that many software systems are burdened with them. This might indicate that not all cycles are as detrimental to software quality as previously thought. Clearly, a better understanding of the types of cyclic dependencies and their effect on software quality is required. As a first step in this direction, we look closely at the shapes formed by software dependency graphs containing cyclic dependencies. Such cyclic dependencies correspond to the concept of strongly connected components in graph theory. We propose an approach to classify strongly connected components according to their topologies. This allows us to distinguish between dense and sparse, symmetric and asymmetric structures. We extend on previous studies and investigate the relationship between cyclic dependencies and the package containment tree. We validate our approach with experiments based on a corpus of 103 open–source Java systems. We find that cyclic dependencies tend to form in branches of the package containment tree around parent packages that are not critical according to some researchers.