Search Based Path and Input Data Generation for Web Application Testing

Test case generation for web applications aims at ensuring full coverage of the navigation structure. Existing approaches resort to crawling and manual/random input generation, with or without a preliminary construction of the navigation model. However, crawlers might be unable to reach some parts of the web application and random input generation might not receive enough guidance to produce the inputs needed to cover a given path. In this paper, we take advantage of the navigation structure implicitly specified by developers when they write the page objects used for web testing and we define a novel set of genetic operators that support the joint generation of test inputs and feasible navigation paths. On a case study, our tool Subweb was able to achieve higher coverage of the navigation model than crawling based approaches, thanks to its intrinsic ability of generating inputs for feasible paths and of discarding likely infeasible paths.