Code-Imp: a tool for automated search-based refactoring

Abstract

Manual refactoring is tedious and error-prone, so it is natural to try to automate this process as much as possible. Fully automated refactoring usually involves using metaheuristic search to determine which refactorings should be applied to improve the program according to some tness function, expressed in terms of standard software quality metrics. Code-Imp (Combinatorial Optimisation for Design Im- provement) is such an automated refactoring platform for the Java language. It can apply a range of refactorings, sup- ports several search types, and implements over 25 software quality metrics which can be combined in various ways to form a tness function. The original goal of the Code-Imp project was to investigate the use of automated refactoring to improve software quality as expressed by a contemporary metrics suite. In this paper we present a technical overview of the Code- Imp implementation, and summarise three active research strands involving Code-Imp: refactoring for testability, met- rics exploration, and multi-level design improvement.