Testing Strategies for the Automated Grading of Student Programs

Abstract

Enrollments in introductory computer science courses are growing rapidly, thereby taxing scarce teaching resources and motivating the increased use of automated tools for program grading. Such tools commonly rely on regression testing methods from industry. However, the goals of automated grading differ from those of testing for software production. In academia, a primary motivation for testing is to provide timely and accurate feedback to students so that they can understand and fix defects in their programs. Testing strategies for program grading are therefore distinct from those of traditional software testing. This paper enumerates and describes a number of testing strategies that improve the quality of feedback for different types of programming assignments.