Object Metrics for Aspect Systems: Limiting Empirical Inference Based on Modularity

In empirical comparisons of Aspect-Oriented (AO) to ObjectOriented (OO) systems, system properties such as understandability, maintainability, reusability, and testability have often been inferred from other metrics, such as lines of code, sites of change, and modularity. However, in traditional OO metrics suites such system properties are assessed separately from these measures. We applied OO metrics for comparing an AO with an OO system and found that improved modularity did not imply improvements in the system properties listed above. This paper illustrates the appropriateness of using OO metrics for the empirical analysis of systems, and presents evidence that in subsequent studies of AO systems certain system properties should be assessed separately from modularity.