Size and Frequency of Class Change from a Refactoring Perspective

Abstract

A previous study by Bieman et al., investigated whether large, object-oriented classes were more susceptible to change than smaller classes. The measure of change used in the study was the frequency with which the features of a class had been changed over a specific period of time. From a refactoring perspective, the frequency of class change is of value But even for a relatively simple refactoring such as 'rename method', multiple classes may undergo minor modification without any net increase in class (and system) size. In this paper, we suggest that the combination of 'versions of a class and number of added lines of code ' in the bad code 'smell' detection process may give a better impression of which classes are most suitable candidates for refactoring; as such, effort in detecting bad code smells should apply to classes with a high growth rate as well as a high change frequency. To support our investigation, data relating to changes from 161 Java classes was collected. Results concluded that it is not necessarily the case that large classes are more change-prone than relatively smaller classes. Moreover, the bad code smell detection process is informed by using the combination of change frequency and class size as a heuristic.