Verification of properties of interactive components from their executable code

In this paper we describe how an executable model of interactive software can be exploited to allow programmers or specifiers to express properties that will be automatically checked on the components they create or reuse. The djnn framework relies on a theoretical model of interactive software in which applications are described in their totality as hierarchies of interactive components, with no additional code. This includes high level components, but also the graphics, behaviors, computations and data manipulations that constitute them. Because of this, the structure of the application tree provides significant insights in the nature and behavior of components. Pattern recognition systems can then be used to express and check simple properties, such as the external signature of a component, its internal flows of control, or even the continued visibility of a component on a display. This provides programmers with solutions for checking their components, ensuring non-regression, or working in a contract-oriented fashion with other UI development stakeholders.