COUPLING METRICS MODEL FOR PREDICTING THE MODULARITY OF ASPECT ORIENTED SOFTWARE

Abstract

The Aspect Oriented Software (AOS) paradigm emerged as a response to the limitations of Object-Oriented Programming (OOP), particularly its inability to modularize cross-cutting concerns effectively. While AOS improves modularity, analyzing its complexity remains a challenge, especially as software modifications increase complexity over time. Although AOS provides a robust framework for handling cross-cutting concerns, assessing its modularity is difficult because most existing metrics lack theoretical and empirical validation, making them unreliable for measuring AOS modularity. To address this issue, this study proposes four base metrics and two composite metrics for predicting the modularity of aspect-oriented software. The proposed base metrics are Method Calling Pointcut, Class Calling Pointcut, and Method Response for Advice and Variable Response for Advice while the composite metrics are Pointcut Complexity Metric (PCM) and Response for Advice Complexity Metric (RACM). These metrics were systematically derived using the Entity-Attribute-Metric-Tool (EAMT) model, ensuring precision and consistency by mapping AOS-specific entities to measurable attributes. Theoretical validation was conducted using Briand’s framework, confirming the soundness of the proposed metrics across key properties, including monotonicity, null value, merging of modules, and disjoint module additivity. A custom tool, the Aspect Oriented Software Metrics Tool (AOSMT), was developed to automate the computation of the proposed metrics. Experimental validation was conducted using a sample of 30 participants from the School of Computing and Information Technology. Participants analyzed AOS project files, provided modularity ratings, and their task completion times were recorded. The results showed strong positive correlations between the proposed metrics and modularity ratings: PCM (r = 0.811) and RACM (r = 0.794), both significant at p < 0.01. Similarly, task completion time was negatively correlated with PCM (r = -0.788) and RACM (r = -0.754), indicating that higher metric values were associated with faster analysis. ANOVA tests also confirmed the statistical significance of these relationships. In conclusion, the proposed coupling metrics are mathematically sound and empirically reliable for predicting modularity in AOS systems. The study recommends adoption of the metrics and the supporting tool (AOSMT) by software developers and researchers to support decision-making processes in modularity analysis, leading to improved software quality.