Code recommendation for Android development： how does it work and what can be improved

• Published in: Science China. Information sciences 2017-09, Vol.60 (9), p.145-158, Article 092111
• Main Authors: Wu, Junwei, Shen, Liwei, Guo, Wunan, Zhao, Wenyun
• Format: Article
• Language: English
• Subjects: Codes; Computer Science; Data mining; Information Systems and Communication Service; Research Paper; 代码; 可定制; 图像; 工具; 应用程序; 开发效率; 技术; 框架
• ISSN: 1674-733X; 1869-1919
• DOI: 10.1007/s11432-017-9058-0

Android applications are developed based on framework and are always pattern-based. For Android developers, they can be facilitated by code recommendation to ensure high development efficiency and quality. Existing research work has proposed several methods and tools to support recommendation in diverse ways. However, how code recommendation work in Android development and what can be further improved to better support Android development has not been clarified. To understand the reality, we conduct a qualitative review on current code recommendation techniques and tools reported in prime literature. The collected work is first grouped into three categories based on a multidimensional framework. Then the review is performed to draw a comprehensive image of the adoption of recommendation in Android development when meeting specific development requirements. Based on the review, we give out possible improvements of code recommendation from two aspects. First, a set of improvement suggestions are presented to enhance the ability of the state-of- the-art code recommendation techniques. Second, a customizable tool framework is proposed to facilitate the design of code recommendation tools and the tool framework is able to integrate the recommendation features more easily.