Enabling Feature Location for API Method Recommendation and Usage Location
IEEE ACCESS(2019)
摘要
Given a new feature request during software evolution, developers are used to employing existing third-party libraries and APIs for implementation. However, it is usually non-trivial to find suitable APIs and to decide where to use these APIs in the original software. In this paper, we develop an approach for recommending API methods and usage locations through mining various software repositories. First, we analyze software repositories and use the feature location technique to localize feature-related files as API usage locations. Then, we utilize the feature-related files and API libraries to identify potential API methods for the implementation of the incoming feature request. We evaluate our approach on 5000 feature requests selected from five Java projects, and the results demonstrate that our approach can achieve 29.7% and 9.6% improvement in terms of Hit@5 and Hit@10 in recommending API methods, compared with the existing approach. For API usage localization, our approach can get MAP, MRR, Hit@1, HIt@5, and Hit@10 values with 0.293, 0.434, 0.303, 0.602, and 0.685, respectively.
更多查看译文
关键词
Feature implementation,API recommendation,API usage location,feature location,mining software repositories
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络