Survey of Code Completion using DeepLearning

Code completion systems are able to generate code automatically whendevelopers write code for system development. Developers use variousIDEs to write a program where IDEs refer to integrated develop-ment environments. Some of the available IDEs are Visual studio,IntelliJ idea, PyCharm, NetBeans, Eclipse, RubyMine, Aptana Studio3, PhpStorm, and WebStorm. Sometimes developers are stuck withprogramming for complex problems, code generation systems supportthem by suggesting the next word that may appear after the writ-ten word. Therefore, it is treated as an essential service feature inIDEs and a popular tool for coding assistance. A good code genera-tor should produce correct and efficient code. This paper investigatesthe current state of research in the code generation field using DeepLearning models. Classification based on various parameters is discussedin the paper. A comparison study is presented with respect to theirdataset, model implementations, and results obtained. A few researchquestions are also discussed which need to be addressed in future.