Generating Java unit tests with AI planning
Eddie Dingels, Timothy Fraser, Alexander J. Quinn · 2007
Although several efforts have applied search-based approaches to the generation of unit tests, none have demonstrated that these approaches scale beyond small programs. Our experiment compares the ability of three general-purpose AI planners to generate unit tests for Java classes with varying sizes and semantics. Although the GraphPlan planner handled most of our test classes, none of our planners could handle them all. We suggest a number of alternate search-based approaches and conclude that further experiments are needed to evaluate them.