Using Change Context with Statistical Learning for API Code Recommendation

Learning and remembering how to use APIs is hard. While codecompletion tools list all the API methods available on a given object, reading through a long list of API method names and their associated documentation is tedious, and users can be easily overloaded with too many suggestions. While several researchers proposed techniques for recommending APIs, their accuracy is low. We present a novel API recommendation approach that taps into the predictive power of repetitive code changes. Our approach and tool, APIREC, is based on statistical learning from fine-grained code changes and from the context in which those changes were made. We trained APIREC on 43M changes from 100K commits in open-source projects. Our empirical evaluation shows that APIREC correctly recommends an API method in the first position 59% of the time, and it recommends the correct API method in the top 5 positions 75% of the time. This is a significant improvement over the best in class state of the art recommender by a factor of 2.4x for the first position, and 2.2x for the top 5 positions, respectively.