Ripple: A Test-Aware Architecture Modeling Framework

Abstract

Different architecture views can be used to address concerns of different stakeholders. While architecture models have been used to guide software detailed design, development, and maintenance, no existing work has incorporated information generated in testing activities into architecture models for providing testing guidance. In this paper, we present Ripple, the framework for constructing test-aware DRSpace modeling to simultaneously reveal dynamic test coupling and static structural dependencies among source files in a software system. Ripple first mines from dynamic test coverage reports to extract traceability links between source files and test cases. It then encodes testing information into DRSpaces and leverages the DRH algorithm to cluster source files into independent test modules. To evaluate Ripple, we conducted a pilot study on a component of Hadoop. The study shows that Ripple has the potential to provide guidance for various stakeholders in making test-related decisions.