Programming Style Analysis with Recurrent Neural Network to Automatic Pull Request Approval
Lucas Roque, Altino Dantas, Celso G. Camilo-Junior · 2019
Although recognized as important, programming style is one aspect commonly neglected by developers. However, follow the pattern of programming presents in a project may be useful to understand and maintain the system. Usually, the companies build their own guidelines for coding bug fixes or new features. Nonetheless, developing this set of rules is not a simple task, and there are even inconsistencies in the specialized literature. Therefore, this paper proposes a new approach to programming style analysis, using a recurrent neural network (RNN) that learns the programming style presents in a project and determines whether a piece of code submitted to it, follows the project's pattern. A study on three real projects was conducted and demonstrated the promising of the approach, by revealing the RNN capability to recognize the programming style pattern of each project.