CCEyes: An Effective Tool for Code Clone Detection on Large-Scale Open Source Repositories

Code clone is a common phenomenon in the rapid development of the software ecosystem and open source community, which can improve efficiency, while leading to many negative effects on software maintenance at the same time. In this paper, we propose CCEyes, an effective tool for code clone detection on GitHub. We build a semantic vector representations corpora database of large-scale open sourse codes by the deep learning-based method that can automatically learn program features to address the resource challenge. CCEyes returns the detection results according to the user's input, including similar clone code fragments, similarity analysis and repository address on GitHub. Among which the Recursive Autoencoder is designed to learn code representation and then the comparator network is employed for similarity evaluation. We conduct comprehensive experiments on the ability of the detection method, the effectiveness of database storage. Besides, we also compare CCEyes with other application tools and conduct a user study. The evalution results in that CCEyes outperforms state-of-the-art.