Model-Based Analysis of Adoption Factors for Software Visualization Tools in Corrective Maintenance

Several classification models exist for software visualization (SoftVis) tools. Such models can be used to compare the features provided by several tools to the requirements of a given use case, in the process of selecting optimally fitting tools. However, it is not easy to predict how such models will perform when used to predict the adoption of SoftVis tools at large, especially for tools which were not considered during the model design. Here, we consider an existing classification model that aims to provide generic guidelines for comparing SoftVis tools for corrective maintenance (CM) based of their features perceived as desirable by users. Although this model explicitly captures several such features, it is not evident that tools that fit the model will be found effective by users in practice. This paper tests the above hypothesis by presenting a comparative evaluation of four software visualization (SoftVis) tools used in CM. The tools were selected to fit well the desirable criteria captured by the model under evaluation. Four independent groups of professional software developers participated in the evaluation, each group using a different tool to solve the same CM task on a real-world code base under typical industry conditions. The results show matches between the features described by the model as highly desirable and and those observed in practice to be essential for tool acceptance, e.g. IDE integration, extended search capabilities, multiple views, scalability, and the need for both dynamic and static visualizations; weakly relevant features, e.g. the commercial tool status; and features which do not influence acceptance, e.g. 3D and animation. Besides showing the correlation between the classification model and observed practice, our study further refines the model’s criteria seen as important for industrial acceptance of software visualization tools.