Reactive Method Dispatch for Context-Oriented Programming

We are on the verge of a post-PC revolution where computing power is shifting from laptops and desktops to smartphones, tablets and everyday objects such as eyeglasses and coffee machines. Those computing objects come equipped with a myriad of context sensors such as an accelerometer, a proximity sensor, a gyroscope, a GPS receiver and NFC technology. Context sensors enable the development of context-aware applications that continuously adapt their behaviour to match the context of use. In this dissertation, we refer to applications that are always prepared to promptly adapt their behaviour in reaction to context changes as reactive context-aware applications. Our research hypothesis is that developing reactive context-aware applications remains notoriously difficult because of the lack of suitable programming language abstractions coupled with the unpredictable nature of context changes. More concretely, current programming languages fall short of providing the appropriate support for developing context-aware applications that need to react promptly to a sudden context change – especially if such a context change occurs in the middle of an ongoing procedure execution. Such a context change may require an ongoing procedure execution to be promptly interrupted in order to prevent its execution from happening in a wrong context. The vision of this dissertation is to investigate novel programming language abstractions to facilitate the development of reactive context-aware applications. We propose a new programming language model called interruptible context-dependent executions, where a procedure execution is always constrained to happen only under a developer specified context condition. In this model, the execution of a context-dependent procedure can be seamlessly interrupted or resumed depending on whether the specified context condition is satisfied or not. The language runtime ensures that the execution state of a context-dependent procedure is automatically preserved between interruptions. The model features a new dispatching mechanism called reactive dispatching that continually takes into account new context changes to select the applicable procedures and to suspend or interrupt ongoing procedure executions. We present a proof-of-concept programming language called Flute that incorporates the interruptible context-dependent executions model. Flute has been implemented as a meta-interpreter on top of iScheme – a technological research artefact that we developed to enable the experimentation with novel language abstractions and features to ease the development of reactive context-aware applications. iScheme runs on the iOS mobile platform and as such it fosters experiments on real mobile devices.