Smartphones and tablets with rich graphical user interfaces (GUIs) are becoming increasingly popular. Hundreds of thousands of specialized applications, called apps, are already available for these mobile platforms, and the number of newly released apps continues to increase. The complexity of these apps lies often in the user interface, with data processing either minor, or delegated to a backend component. A similar situation exists in applications using the software-as-a-service architecture, where the client-side component consists mostly of user interface code. Testing such applications predominantly involves GUI testing. Existing automatic techniques for testing these interfaces either require a priori models of the interface and are thus hard to use, or operate blindly by sending random user events to the application and are typically unable to test the application in satisfactory depth.
This project investigates automatic GUI testing techniques that systematically explore the state space of an application without requiring an a priori defined model. One insight behind this project is that the automatic construction of a model of the user interface and the testing of the interface are tasks that can cooperate in a mutually beneficial way. Furthermore, a guiding principle throughout this research is to design algorithms that operate with abstractions and heuristics that are simple enough to be understood by humans who do not necessarily understand the internals of the tested app. Such algorithms are easier to comprehend and to incorporate into a wholistic test process that includes automated techniques, such as the ones developed in this project, and manual testing and guidance. The techniques developed in this project benefit directly programmers for these apps, and indirectly the numerous users of mobile and web applications.
