InvAASTCluster: On Applying Invariant-Based Program Clustering to Introductory Programming Assignments

Due to the vast number of students enrolled in Massive Open Online Courses (MOOCs), there has been an increasing number of automated program repair techniques focused on introductory programming assignments (IPAs). Such state-of-the-art techniques use program clustering to take advantage of previous correct student implementations to repair a given new incorrect submission. Usually, these repair techniques use clustering methods since analyzing all available correct student submissions to repair a program is not feasible. The clustering methods use program representations based on several features such as abstract syntax tree (AST), syntax, control flow, and data flow. However, these features are sometimes brittle when representing semantically similar programs. This paper proposes InvAASTCluster, a novel approach for program clustering that takes advantage of dynamically generated program invariants observed over several program executions to cluster semantically equivalent IPAs. Our main objective is to find a more suitable representation of programs using a combination of the program's semantics, through its invariants, and its structure, through its anonymized abstract syntax tree. The evaluation of InvAASTCluster shows that the proposed program representation outperforms syntax-based representations when clustering a set of different correct IPAs. Furthermore, we integrate InvAASTCluster into a state-of-the-art clustering-based program repair tool and evaluate it on a set of IPAs. Our results show that InvAASTCluster advances the current state-of-the-art when used by clustering-based program repair tools by repairing a larger number of students' programs in a shorter amount of time.