A weighted call graph approach for finding relevant components in source code

Abstract

The reuse or maintenance of existing code requires understanding how the program is implemented and how the architecture is organized. Small modification of open source requests partial understanding. However, it``s not easy because open source doesn``t have design document which maps the feature to source code. I propose a method to find the components relevant to a feature in open source code. My solution is a static approach to recovery the traceability between requirement specification and source codes. A premise of my work is that open source follows GNU C coding standard. The coding standard suggests not to use abbreviation of identifier and to use meaningful identifiers in source code. To recover these traceability, my approach build and analyze the weighted call graph, which is built using the similarity values obtained from the vector space information retrieval(IR). Through analysis of this weighted call graph, the possibly relevant functions are finally determined. I use two analysis methods: core function analysis and weight analysis. To evaluate my idea, I do experimental study of my approach and compare and discuss the result of my approach with the existing approach.