Multiscale and Multivariate Visualizations of Software Evolution

Software evolution visualization is a promising technique for assessing the software development process. We study how complex correlations of software evolution attributes can be made using multivariate visualization techniques. We use a combination of color and textures to depict up to four artifact attributes at the same time in one view using the same spatial layout. Next, we describe an interactive navigation method of the attribute space that can extend the correlation capabilities to four or more attributes. A second issue we address is how to use clustering to reduce the complexity of evolution visualizations. We propose two new methods, isometric and isorelevance, to generate relevant abstraction levels in a hierarchical clustering of software evolution artifacts. The isometric method generates partitions with similar size elements. The isorelevance method generates partitions with elements of similar relevance. We propose a novel widget, the cluster map, which visualizes all partitions in a clustering and supports users when making size/relevance compromises when choosing a partition. We illustrate the applicability of the proposed techniques with two usage scenarios on the evolution of two real-life industry size projects