Improving reverse engineering models with test-case related knowledge

Abstract

In order to be effective in software evolution activities, the quality of reverse engineering models should be characterized and improved in various aspects, ranging from completeness to abstraction level and semantic content. This entails the combination of multiple knowledge sources, ranging from source code to user documents, and human expert knowledge. The products of testing activities can be used to create links or assign evidence weights to links between these sources, having different degrees of formalization, abstraction level and reliability; this applies also to the results of past system understanding efforts, occurring during maintenance tasks, which can be referred to test cases. This paper reports on the results of systematic investigations in this area which were conducted in the context of the Docket project, leading to a method that combines static analysis, information filtering and enrichment steps, and criteria to exploit the knowledge productivity of test cases and dynamic analysis.