Employing issues and commits for in-code sentence based use case identification and remodularization

Abstract

Use case driven modularization improves code comprehension and maintenance and provides another view on software alongside object-oriented modularization. However, approaches enabling use case driven modularization require to modularize code manually. In this paper, we propose an approach to employing issues and commits for in-code sentence based use case identification and remodularization. The approach aims at providing use case based perspective on the existing code. The sentences of use case steps are compared to sentences of issue descriptions, while the sentences generated from the source code of issue commits are compared to sentences generated from the corresponding methods in source code in order to quantify the similarity between use case steps and methods in source code using different similarity calculation algorithms. The resulting level of similarity is used to remodularize source code according to use cases. We conducted a study on the OpenCart open source e-shop employing 16 use cases. The approach achieved the recall of 3.37% and precision of 75%. The success of the approach strongly depends on issues and commits assigned to them. The results would be better especially for the code that natively employs use case driven modularization.