Analyzing the effects of refactorings on bad smells

• Autor(a) principal: Cleiton Silva Tavares
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/38392
• topic: Refactoring; Bad smell; Refactoring impacts; Refactoring effects; Computação – Teses; Engenharia de Software – Teses; Software - Refatoração– Teses; Bad smell – Teses

Refactoring aims to increase software systems' maintainability by improving their structure without changing their behavior, may applied to remove bad smells. Even with tools to assist refactoring, many developers do not trust their solutions, claiming that some studies show that refactoring can introduce new bad smells into the source code. However, we have not found a complete catalog that states when this may occur. To investigate this subject deeply, the goal of this dissertation is to evaluate the effects of refactoring on bad smells. Specifically, we want to know if and what refactoring removes bad smells or introduces them. To achieve this goal, we conducted a Systematic Literature Review (SLR) to identify the relationship between refactorings and bad smells proposed by Fowler. We also conducted an empirical study with eight software systems applying five refactorings to analyze their effects on ten bad smells with the assist of five tools. As a result of the empirical study, we present, through the data studied, which bad smells tend to be removed or introduced by the automated refactoring process. In the SLR, we found 20 papers showing the direct relationship between 31 refactorings and 16 bad smells. We produced a catalog showing these relationships, and we also showed a contrast with relationships discussed by Fowler. We identified that the most discussed relationship in the literature is between Move Method and Feature Envy. The SLR also revealed different refactoring strategies than those discussed by Fowler for dealing with bad smells. In the empirical study, we observed that refactoring generated decrease, increase, and neutral variations in the number of bad smells. Unlike Fowler's definition, we surprisingly found that the number of bad smells decrease was the lowest compared to cases of increase and neutral variations. In an additional analysis, we contrast the results found in the two studies carried out, classifying, validating and complementing them.