ANALYSING THE EFFECTS OF REFACTORING ON SOFTWARE QUALITY ATTRIBUTES

Abstract

Bad-smell indicates code-design flaws and poor software quality that weaken software design and inversely affects software development. It also works as a catalyst for bugs and failures in the software-system. Refactoring methods are used by software practitioners as corrective actions for bad-smells. Refactoring is not only limited to removing bad-smells but it does have a strong correlation with quality attributes. Countless studies are present in the literature that studies the effect of refactoring methods on software quality attributes. It is said to improve certain aspects of quality. Also, refactoring is a costly activity and the problem relies upon the fact that there are over seventy refactoring-methods available in literature and multiple refactoring methods can be used to nullify the effect of a particular bad-smell. So, it becomes very difficult to apply refactoring on complete source-code and almost impossible if software size is dramatically large. Thus, there arises a need for prioritizing classes in some way. This study aims to first provide a systematic literature review on the correlation of refactoring-methods and bad-smells and their improvement on internal as well as external quality attributes and second, it comes up with a way to apply refactoring to only severely affected classes to improve the overall software quality. The systematic literature review helps software developers in identifying the commonly prevalent bad smell, their possible refactoring solution and effect of those refactoring methods on software quality attributes and guide the researchers in conducting future research. In the end, a framework is proposed that detects a small subset of classes from the entire v

source-code instantly require refactoring. This prioritization of classes is based on two factors-severity of the presence of bad-smells and object-oriented characteristics. The approach is evaluated on eight open-source software-systems written in Java using ten most common bad-smells and six widely known Chidamber & Kemerer metrics. Both these factors help in calculating a new metric Quality Depreciation Index Rule (QDIR) that exposes those classes that are highly affected by bad-smells and demand an immediate refactoring solution. Results of the empirical study indicate that QDIR is an effective metric to remove bad-smells in an environment of stringent time constraints and limited cost, making the maintenance of software-system easier and effective with enhanced software quality.