Stable and predictable Voronoi treemaps for software quality monitoring

Abstract

Context: Voronoi treemaps can be used to effectively visualize software quality attributes of a given software system. Algorithms for computing Voronoi treemaps are non-deterministic making them unsuited for monitoring the development of such attributes over time. Objective: We adapt an existing sweep line algorithm to efficiently compute Voronoi treemaps and we introduce a novel algorithm that adds stability and predictability. Method: We introduce stable and predictable Voronoi treemaps based on additively weighted power Voronoi diagrams. We employ scaled Hilbert curves to place Voronoi sites in the plane, retaining the order in which sites are placed along the curve for easy comparison with revisions of the same software system. Results: Our algorithm achieves a predictable first good approximation of the final location of the sites. We show that our algorithm not only provides more stability, but also that because of better placement it needs fewer iterations to compute its result. As part of our implementation we introduce a visualization to show the difference between two versions of a software system. We also present a small case study in which we use a web based application that implements our work to investigate the usefulness of stability and predictability of visualizations. Conclusion: It is possible to achieve stable and predictable visualizations of software system attributes, while, as a pleasant side effect, decreasing the number of iterations necessary to arrive at the visualization.