Analysis of Coupling Evolution on Open Source Systems

Software evolution is an intrinsic process of the software life cycle. The comprehension of this process is a central research topic in Software Engineering. It is widely accepted that as a software system evolves, its internal quality declines, and its complexity increases. However, there is a gap in the comprehension of how this process occurs in a fine-grained view. In this work, we apply a software metric approach to investigate how the internal quality of object-oriented software systems evolves in the aspect of coupling. More specifically, we analyze (i) how the coupling behavior may be described over the software evolution, (ii) how the coupling behavior affects the reusability and complexity of the systems, and (iii) the percentage of classes from the systems that directly impacts on the coupling evolution. The results and observations of this study are compiled in seven properties of coupling evolution, among which stand out: (i) the coupling behavior is better modeled by a cubic function, (ii) the coupling evolution tends to increase the complexity of the systems, (iii) the systems tend to be designed with a high level of complexity, and (iv) the coupling evolution is affected by a small group of classes.