Improving the Dataflow-Based Concern Identification Approach

Concern identification aims to identify the implementation of a functional concern in existing source code. The dataflow-based concern identification approach starts from a set of concern seeds and uses static dataflow information to extract the data skeleton of a functional concern. This paper builds upon previous work on dataflow-based concern identification and presents three improvements to the identification approach: the reduction of the search space for manual identification of concern seeds, the introduction of information sources as a mechanism to explicitly define concern boundaries and the separation of superimposed class roles. The paper also shows the impact of these improvements by comparing the results of the improved identification approach with previously published results on the JHotDraw open-source case-study.