Mapping features to source code based on execution trace

Abstract

One of the important activities in software maintenance is to understand the software system before implementing any change in the software. Mapping software features to their relevant source code is one aspect of software comprehension. In this paper, we propose a full dynamic feature location technique to map the source code to the relevant functional features. The proposed solution aims to use execution trace to do the required mapping. The developer first uses a scenario to exercise a feature to generate the required execution trace. Second, the developer formulates a text query to describe the feature of interest in a natural language, to retrieve and rank the relevant methods from the generated execution trace. Distance software metrics are proposed to measure the distance between the raked methods. The solution is evaluated by using the ArgoUML Java case study to show the effectiveness of the proposed solution. The results of the evaluation show the accuracy of allocating methods in the source code to the related feature is increased by using the proposed methodology.