摘要:Context-aware applications, whose behavior reactively depends on the time-varying status of the surrounding environment - such as network connection, battery level, and sensors - are getting more and more pervasive and important. The term "context-awareness" usually suggests prompt reactions to context changes: as the context change signals that the current execution cannot be continued, the application should immediately abort its execution, possibly does some clean-up tasks, and suspend until the context allows it to restart. Interruptions, or asynchronous exceptions, are useful to achieve context-awareness. It is, however, difficult to program with interruptions in a compositional way in most programming languages because their support is too primitive, relying on synchronous exception handling mechanism such as try-catch.
We propose a new domain-specific language ContextWorkflow for interruptible programs as a solution to the problem. A basic unit of an interruptible program is a workflow, i.e., a sequence of atomic computations accompanied with compensation actions. The uniqueness of ContextWorkflow is that, during its execution, a workflow keeps watching the context between atomic actions and decides if the computation should be continued, aborted, or suspended. Our contribution of this paper is as follows; (1) the design of a workflow-like language with asynchronous interruption, checkpointing, sub-workflows and suspension; (2) a formal semantics of the core language; (3) a monadic interpreter corresponding to the semantics; and (4) its concrete implementation as an embedded domain-specific language in Scala.
关键词:workflow; asynchronous exception; checkpoint; monad; embedded domain specific language