Empirical Validation of a Dynamic Coupling Metric

Most object-oriented coupling measures proposed in the literature deal with coupling at the static class level. Measuring dynamic object coupling on the other hand gives us the potential for greater insight into system structure and a comparison of the architectural aspects of different systems. In previous work, we theoretically validated a Dynamic Coupling Metric (DCM) and showed, with the aid of a small example, how a system with meta-level characteristics contained less coupling than a corresponding system without these features. In this paper, we build on the theoretical foundations of our DCM and investigate, in more detail, an empirical validation of the metric. A profiler was developed to collect dynamic coupling data represented by message exchanges between objects at runtime; two system models with different architectural characteristics providing the same functionality were then considered. Empirical validation of the metric showed it to be a useful vehicle for comparing the runtime coupling characteristics of software systems. Results also showed that for examples covering various types of couplings, reflective systems separating the base level from the meta-level exhibited significantly less coupling when compared with corresponding systems with a single-level architecture.