Translating quality-driven code change selection to an instance of multiple-criteria decision making

Abstract

Context: The definition and assessment of software quality have not converged to a single specification. Each team may formulate its own notion of quality and tools and methodologies for measuring it. Software quality can be improved via code changes, most often as part of a software maintenance loop. Objective: This manuscript contributes towards providing decision support for code change selection given a) a set of preferences on a software product’s qualities and b) a pool of heterogeneous code changes to select from. Method: We formulate the problem as an instance of Multiple-Criteria Decision Making, for which we provide both an abstract flavor and a prototype implementation. Our prototype targets energy efficiency, technical debt and dependability. Results: This prototype achieved inconsistent results, in the sense of not always recommending changes reflecting the decision maker’s preferences. Encouraged from some positive cases and cognizant of our prototype’s shortcomings, we propose directions for future research. Conclusion: This paper should thus be viewed as an imperfect first step towards quality-driven, code change-centered decision support and, simultaneously, as a curious yet pragmatic enough gaze on the road ahead.