A catalogue of lightweight visualizations to support code smell inspection

Abstract

Preserving the integrity of software systems is essential in ensuring future product success. Commonly, companies allocate only a limited budget toward perfective maintenance and instead pressure developers to focus on implementing new features. Traditional techniques, such as code inspection, consume many staff resources and attention from developers. Metrics automate the process of checking for problems but produce voluminous, imprecise, and incongruent results. An opportunity exists for visualization to assist where automated measures have failed; however, current software visualization techniques only handle the voluminous aspect of data but fail to address imprecise and incongruent aspects. In this paper, we describe several techniques for visualizing possible defects reported by automated inspection tools. We propose a catalogue of lightweight visualizations that assist reviewers in weeding out false positives. We implemented the visualizations in a tool called NOSEPRINTS and present a case study on several commercial systems and open source applications in which we examined the impact of our tool on the inspection process.