Refactoring graphs: reasoning about refactoring over time

• Main Author: Aline Norberta de Brito
• Publication Date: 2023
• Format: Doctoral thesis
• Language: eng
• Source: Repositório Institucional da UFMG
• Download full: http://hdl.handle.net/1843/56371; https://orcid.org/0000-0002-7088-4935
• topic: Refactoring; Refactoring graphs; Mining software repositories; Software evolution; Computação – Teses; Engenharia de software – Teses; Grafos de refatoração – Teses; Mineração de repositórios de software – Teses

Frequently, practitioners refactor their code, producing new entities or changing the structure of existing ones. Sometimes, these transformations are performed in a constrained time frame. In other cases, they generate sequences of modifications performed over a long time period, a scenario not usually considered in the literature. In this context, the main goal of this Ph.D. thesis is to characterize and understand refactoring operations performed over time. For this purpose, we introduce a novel abstraction for reasoning about refactorings, named refactoring graphs. In this graph-based abstraction, nodes represent methods or functions, and edges refer to refactoring operations. We organize the research into three major working units. We start by defining the proposed abstraction, describing the elements and providing a set of scripts to detect refactoring graphs in real-world projects. Then, in the second working unit, we characterize about 1.5K refactoring subgraphs from popular Java and JavaScript projects hosted on GitHub. The results confirmed our hypothesis that refactorings are not only sole operations. There are also sequences of transformations performed over time. Besides, we also perform a qualitative study, in which we contact developers responsible by subgraphs describing large operations, aiming to identify their main reasons to perform such operations. Finally, in the last working unit, we explore applications of the proposed graph-based model. First, we evaluate refactoring tasks performed by undergraduate students from a Software Engineering course, relying on refactoring graphs to understand and inspect the operations. In other words, we assume a professor's perspective who needs to understand practical exercises on refactoring. We also propose a catalog of common refactorings performed over time, which we decide to call composite refactorings. We rely on refactoring graphs to document and illustrate instances of composites detected in an oracle with hundreds of operations and in the history of popular projects hosted on GitHub.