Optimizing aspect-oriented product line architectures with search-based algorithms

Abstract

The adoption of Aspect-Oriented Product Line Architectures (AOPLA) brings many benefits to the software product line design. It contributes to improve modularity, stability and to reduce feature tangling and scattering. Improvements can also be obtained with a search based and multi-objective approach, such as MOA4PLA, which generates PLAs with the best trade-off between different measures, such as cohesion, coupling and feature modularization. However, MOA4PLA operators may violate the aspect-oriented modeling (AOM) rules, impacting negatively on the architecture understanding. In order to solve this problem, this paper introduces a more adequate representation for AOPLAs and a set of search operators, called SO4ASPAR (Search Operators for Aspect-Oriented Architectures). Results from an empirical evaluation show that the proposed operators yield better solutions regarding the fitness values, besides preserving the AOM rules.