Empirical evaluation of a nesting testability transformation for evolutionary testing

摘要: Evolutionary testing is an approach to automating test data generation that uses evolutionary algorithm search a object's input domain for data. Nested predicates can cause problems testing, because information needed guiding the only becomes available as each nested conditional satisfied. This means process overfit early information, making it harder, and sometimes near impossible, satisfy constraints become apparent later in search. The article presents testability transformation allows evaluation of all conditionals at once. Two empirical studies are presented. first study shows form nesting handled prevalent practice. second how improves generation.