Análise da modularidade de características aspectuais de tecnologias para implementar linhas de produtos de software

Abstract

Software Product Lines (SPL) aim to generate software products belonging to a specific domain, by means of the systematic reuse of software artifacts. Thence, the approach identifies common features, which allows the development of products with standard behavior; and variable features, which allows the variation in product behavior and the composition of specific products. The critical point of SPL is the modularization of software artifacts to be coherent, well-defined, independent, and combinable. For this, the feature set development uses variability management technologies, which support configurators for the automated generation of configurations and derivation of the LPS products. In this work, we proposed to investigate the modularity of an SLP developed with three variability management technologies: i) Aspect Oriented; ii) Feature Oriented; and iii) Aspectual Feature Modules. Thus, we selected three coupling metrics (Dependency In, Dependency Out and Structural Feature Coupling) and three cohesion metrics (External-ratio Feature Dependency, Internal-ratio Feature Dependency and Lack of Concern-based Cohesion). We analyzed these metrics using the Friedman significance test (with a significance level of 5%). Firstly, we analyzed which metric individually, and as a result, there was a significant difference between the technologies Aspect Oriented and Aspectual Feature Modules, for all metrics. After, we analyzed the set of coupling metrics and the set of cohesion metrics. However, there was no significant difference among the three variability management technologies, using the set of cohesion and coupling metrics.