An Experimental Prototype For Automatically Testing Student Programs Using Token Patterns

Automated systems and tools for assessing student programs are now commonly used for enhancing the teaching and learning of computer programming. However, many such systems employ rudimentary techniques in comparing program outputs when testing student programs for determining their correctness. These comparison techniques are typically inflexible and disallow even slight deviations of program output which a human assessor would normally tolerate. This may give rise to student frustrations and other undesirable pedagogical issues that can undermine the benefits of using these assessment tools. This paper presents an experimental prototype we have developed that adopts a token-pattern-based approach to accommodate more tolerant output comparisons in testing student programs, followed by a preliminary validation of the prototype by showing how it can be configured to handle the assessment of variants of program outputs.