Visualization of Graphs and Trees for Software Analysis

A software architecture is an abstraction of a software system, which is indispensable for many software engineering tasks. Unfortunately, in many cases information pertaining to the software architecture is not available, outdated, or inappropriate for the task at hand. The RECONSTRUCTOR project focuses on software architecture reconstruction, i.e., obtaining architectural information from an existing system. Our research, which is part of RECONSTRUCTOR, focuses on interactive visualization and tries to answer the following question: How can users be enabled to understand the large amounts of information relevant for program understanding using visual representations? To answer this question, we have iteratively developed a number of techniques for visualizing software systems. A large number of these cases consists of hierarchically organized data, combined with adjacency relations. Examples are function calls within a hierarchically organized software system and correspondence relations between two different versions of a hierarchically organized software system. Hierarchical Edge Bundles (HEBs) are used to visualize adjacency relations in hierarchically organized data, such as the aforementioned function calls within a software system. HEBs significantly reduce visual clutter by visually bundling relations together. Massive Sequence Views (MSVs) are used in conjunction with HEBs to enable analysis of sequences of relations, such as function-call traces. HEBs are furthermore used to visually compare hierarchically organized data, e.g., two different versions of a software system. HEBs visually emphasize splits, joins, and relocations of subhierarchies and provide for interactive selection of sets of relations. Since HEBs require a hierarchy to perform the bundling, we present Force-Directed Edge Bundles (FDEBs) as an alternative to visually bundle relations together in the absence of a hierarchical component. FDEBs use a self-organizing approach to bundling in which edges are modeled as flexible springs that can attract each other. As a result, visual clutter is reduced and high-level edge patterns are better visible. Finally, in all these methods, a clear depiction of the direction of edges is important. We have therefore performed a separate study in which we evaluated ten representations (including the standard arrow) for depicting directed edges in a controlled user study.