History Coupling Space: A New Model to Represent Evolutionary Relations

Abstract

During software evolution, files are usually changed together for accommodating modifications. Although co-change analysis has been widely adopted for difference studies, such as defect prediction, impact analysis, architectural relations identification etc., there has been little work characterizing co-changed files as a group and modeling the evolution of these files to represent software maintenance and evolution. In this paper, we present the concept of history coupling dependency, based on which, we propose a novel model, history coupling space (HCSpace), to link the co-changed files and represent how files are historically connected as a group. Our investigations on seven open source projects show that each HCSpace could be treated as a maintenance unit where the involved files are more likely to evolve together. The results also show that the HCSpaces have important impacts on a project's maintenance. In particular, we demonstrate that identified HCSpaces consumes a relatively large portion of maintenance effort spent on a project, and these HCSpaces are still actively evolving.