An Automated Framework for Recommending Program Elements to Novices (N)

Abstract

Novice programmers often learn programming by implementing well-known algorithms. There are several challenges in the process. Recommendation systems in software currently focus on programmer productivity and ease of development. Teaching aides for such novice programmers based on recommendation systems still remain an under-explored area. In this paper, we present a general framework for recognizing the desired target for partially-written code and recommending a reliable series of edits to transform the input program into the target solution. Our code analysis is based on graph matching and tree edit algorithms. Our experimental results show that efficient graph comparison techniques can accurately match two portions of source code and produce an accurate set of source code edits. We provide details on implementation of our framework, which is developed as a plugin for Java in Eclipse IDE.