Towards Viewpoint-driven Visual Analysis for Effective Architecture Recovery
In the context of software architecture recovery, visual representations can allow architects or developers to perceive and comprehend the recovered information faster. Many visualizations address various facets of the complex information space, but there are no systematic approaches for the architect to efficiently evaluate candidate visualizations for suitability, or to explore ways to extend them. The resultant ad-hoc selection from available generic visualizations leads to ineffective comprehension and analysis. We propose a viewpoint-driven approach towards tackling this problem. From a comparison of the conceptual models of architectural viewpoints and information visualization, we observe that focusing on the concerns framed by a viewpoint can guide the suitable selection of visualization tasks. The viewpoint description also serves a source for the definition of the visualization dataset. Starting from a seed visualization, a simple measure called the task/concern coverage allows the architecture visualization designer to decide if the resulting visualization can express all the desired concerns. This approach is illustrated with an example of the Polyptychon hierarchical dependencies visualization.