A Visual Debugging Aid based upon Discriminative Graph Mining

Abstract

Why doesn’t my code work? Instructors for introductory programming courses frequently are asked that question. Often students understand the problem they are trying to solve well enough to specify a variety of input and output scenarios. However, they lack the ability to identify where the bug is occurring in their code. Mastering the use of a full-feature debugger can be difficult at this stage in their computer science education. But simply providing a hint as to where the problem lies may be sufficient to guide the student to add print statements or do a hand-trace focusing on a certain section of the code. Herein we present a software tool which, given a C++ program, some sample inputs, and respective expected outputs, uses discriminative graph mining to identify which lines in the program are most likely the source of a bug. Additionally, the particular operators (relational, logical, and arithmetic) that are used in the code may be considered in recommending where the bug may be. The tool includes a visual display of the control flow graph for each test case, allowing the user to step through the statements executed. Keywords-debugging; graph; data mining; visualization