Visualizing Multivariate Attributes on Software Diagrams

Abstract

Software architecture diagrams and metrics are wellknown and heavily used in many areas in software engineering. However, they are rarely combined in one (visual) representation. Although there are some advances in this direction, there are also some limitations. In this research, we study how to overcome these limitations. Specifically, we are interested in visualizing metrics on several levels of detail (classes, methods, groups of classes) on UML-like diagrams in a scalable and intuitive way. We present the results obtained in the first three years of the PhD track and outline ongoing work.

The State of the Art in Visualizing Multivariate Networks

“…UML diagrams are node-link layouts with relevant attributes encoded directly on the node marks in the form of text, glyphs, or small embedded visualizations. The often hierarchical nature of structure data is well suited to the combined use of nodelink layout and overloading techniques, an approach optimized to encode group membership or hierarchical relations between elements [BT09].…”

Explanatory visualization of multidimensional prejections

“…This approach is relatively scalable, in the sense that one can encode a large number of attributes per node if the node size, in the visualization, is large enough. In the limit, for instance, nodes themselves can become full-fledged visualizations of multivariate data tables [27]. However, this implies a trade off between the number of nodes and the number of attributes per node that a given visualization can accommodate.…”