Pre-Implementation Method Name Prediction for Object-Oriented Programming

Method naming is a challenging development task in object-oriented programming. In recent years, several research efforts have been undertaken to provide automated tool support for assisting developers this task. general, literature approaches assume the availability of method implementation infer its name. Methods however are usually named before their implementations. work, we fill gap about name prediction by developing an approach that predicts names all methods be implemented within class. Our work considers class as input: overall intuition classes with semantically similar tend functionalities, and hence names. We first conduct large-scale empirical analysis on 258K+ from real-world projects validate our hypotheses. Then, propose hybrid big code-driven approach, Mario , predict based name: combine deep learning model heuristics summarized code analysis. Extensive experiments 22K+ yielded promising results: compared state-of-the-art code2seq (which leverages data), achieves comparable results terms F-score at token level prediction; additionally, outperforms level. further show significantly other baselines.