Refactorings to evolve object-oriented systems with aspect-oriented concepts
Miguel P. Monteiro · Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) · 2005
Software engineering tools should support complete separation of concerns, by enabling the deployment of each different concern in its own unit of modularity. Unfortunately, current tools and languages – including those supporting the object-oriented programming paradigm – fail to provide a complete and effective support for the separation of all concerns. Undesirable phenomena such as code scattering and code tangling ensue. Aspect-Oriented Programming is a new programming paradigm capable of modularising crosscutting concerns. Aspect-oriented programming complements existing programming paradigms, including object-oriented programming, with constructs that provide a fuller separation of concerns. Refactoring is a technique to restructure program source code in order to improve its underlying design and style while preserving the externally observable behaviour. “Code smells” help to detect inadequate structures and designs, which are then gradually removed through refactoring processes. There is a prospect of aspect-oriented programming becoming a mainstream technology in the near future. This begs the question of how to deal with a large base of object-oriented legacy code when aspect-orientation becomes standard practice. AspectJ's backward compatibility with Java opens the way for refactoring existing Java applications to leverage the concepts and mechanisms of aspects. This requires a prior idea of good style for aspect-oriented source code, something yet to be developed and matured. This thesis contributes to the definition of a new style appropriate to aspect orientation. To this effect, this thesis documents a collection of novel refactorings enabling the extraction of crosscutting concerns from object-oriented legacy source code and the subsequent restructuring of the aspects thus obtained. In addition, this thesis presents a review of traditional object-oriented code smells so they can be used as indicators of latent aspects in object-oriented source code. Finally, this thesis proposes several novel aspect-oriented code smells. We validate the refactorings through an illustrative refactoring process.