Improving Software Design by Mitigating Code Smells

Abstract

Improving design is a key research challenge during software maintenance. It aims at identifying and mitigating structural characteristics that may cause future problems (code smells). The existing approaches lack in considering the developer’s expertise and penalty consideration during inheritance usage. Therefore, this paper targets identifying three key code smells viz refused bequest, shotgun surgery, and duplicated code. The refused bequest detection approach is based on probabilistic penalized inheritance usage and domain closeness among methods. The shotgun surgery is detected using combined use of co-change and usage pattern based functional relations. Finally, the duplicated code is based on the abstract tree representation and identification of longest common subsequence between the inorder/ postorder traversals. The experimental results uphold the feasibility of the proposed approach.