Towards a framework for automatic correction of anti-patterns

Abstract

One of the biggest concerns in software maintenance is design quality; poor design hinders software maintenance and evolution. One way to improve design quality is to detect and correct anti-patterns (i.e., poor solutions to design and implementation problems), for example through refactorings. There are several approaches to detect anti-patterns, that rely on metrics and structural properties. However, finding a specific solution to remove anti-patterns is a challenging task as candidate refactorings can be conflicting and their number very large, making it costly. Hence, development teams often have to prioritize the refactorings to be applied on a system. In addition to this, refactoring is risky, since non-experienced developers can change the behaviour of a system, without a comprehensive test suite. Therefore, there is a need for tools that can automatically remove anti-patterns. We will apply meta-heuristics to propose a technique for automated refactoring that improves design quality.