Multi‐view city‐based approach for code‐smell evolution visualisation

Abstract

Code smells are indicators of inappropriate and possibly harmful design decisions that could lead to issues in the comprehensibility and maintainability of software systems. To avoid such quality complications, understanding the presence and prioritising the removal of code smells are required. This study presents a visualisation approach to help better understanding the evolutional characteristics of code smells presented in the different versions of the software system. The core of the visualisation approach is the metaphor of buildings and building blocks. An overall framework for detecting, categorising and visualising code smells is proposed. Three types of code smells were considered in this study. The considered code smells are God Class, Long Method and Type Checking. The applicability of the proposed approach is demonstrated by evaluating several versions of an open-source java software and visualising the detected code smells. Additionally, a pilot experimental study is conducted to empirically assure the usefulness of the proposed visualisations.