Automatically Generating Learning Setups for GUI- based Programs through Annotation Processing

The main challenge for widespread adaptation of active automata learning is the effort required to design and implement application specific learning setups. In this paper, we propose an abstraction of GUI elements by dividing them into three categories: widgets, containers and windows. We analyse at how events are handled in Java, JavaFX and how the MVC pattern is structured. With this information in hand we propose a Mealy machine based behaviour model. Learning GUI-based programs has a few challenges: Event handler accessibility, event space transmutations and the order of events. We solve the event handler accessibility by performing checks, event space transmutations by storing the event handlers dynamically and the order of events by analysing the validity of queries through an NFA-ε with as language all possible event sequences at that given moment. With our research, we have paved the way for active automata learning of GUI-based programs.