Method name recommendation based on source code depository and feature matching

Quality of method names is critical for the readability and maintainability of program. However, it is difficult for software engineers, especially non-English speaking, inexperienced engineers, to propose high quality method names. To address this issue, this paper proposes an approach to recommend method names. First, a method corpus is constructed from open source applications. For a given method f to be named, similar methods are retrieved from the method corpus. Names of these retrieved methods are divided into phrases, and features of these methods are extracted as well. A mapping between these phrases and features is also created to derive a list of candidate phrases and features for the method to be named. These phrases are finally constructed into candidate method names. The proposed approach is evaluated on 1 430 methods in open source applications. Evaluation results suggest that 22.7 percent of recommended method names are the same as original ones, and 57.9 percent has the same or almost the same keywords as original ones.