Horizontal decomposition of Prevayler

Aspect-Oriented Programming (AOP) is an emerging paradigm to modularize crosscutting concerns. A series of guidelines to refactor a software system into a common core and a set of variable functionalities have become known as Horizontal Decomposition (HD). In this paper we apply the HD principles to the Prevayler main memory database management system. The objective is to evaluate and refine these principles and to extract patterns of their use through a case study on a popular open-source software system. Our evaluation shows that HD reveals six crosscutting functionalities. The refactoring of these concerns yield 36 different configurations of the Prevayler system which were previously not possible. The refactoring also reduces the core Prevayler code size by 53%, demonstrates a decrease of coupling between core functionality components by 43%, and reduces the lack of cohesion of the core system by 71%. Given the heterogeneous nature of crosscutting displayed in Prevayler, the size and separation of concern metrics have not reduced for the overall refactored system, i.e., for the core composed with the aspects. A posterior analysis of the re-engineering process reveals 22 refactoring patterns that could be readily used by an automatic aspect refactoring tool.