Improving Software Modularity through Crosscutting Concern Extraction

Aspect-oriented programming (AOP) is a programming paradigm for improving the modularity of software systems by localizing crosscutting concerns in the system into aspects. Aspect-oriented refactorings extend AOP to legacy systems, by identifying and encapsulating existing crosscutting concerns through aspect-mining (discovery of crosscutting concerns) and aspect refactoring (semantic-preserving code transformation to extract the crosscutting code into aspects). However, not all the data obtained from aspectmining corresponds to crosscutting concerns, and existing aspect languages may not be capable of refactoring all crosscutting concerns into aspects. In this thesis, we describe an approach for extracting crosscutting concerns in a system to a form that is suitable for refactoring. This process includes identifying the presence of crosscutting code clusters in aspect-mining results; assessing if the concerns should be extracted using various metrics; and performing code transformation to extract the crosscutting clusters into standalone methods with a common method signature and parameters. The work also describes the ConcernExtractor, a software tool that implements the concern cluster extraction technique. We apply ConcernExtractor to assess and extract the crosscutting concerns in existing systems to evaluate the prevalence of crosscutting concerns that are refactorable, and the applicability of our approach for generating aspect refactoring opportunities.