| Controlling Effects (1999) | |||||||||||||||
Abstract | |||||||||||||||
| Many computational effects, such as exceptions, state, or nondeterminism, can be conveniently specified in terms of monads. We investigate a technique for uniformly adding arbitrary such effects to ML-like languages, without requiring any structural changes to the programs themselves. Instead, we use monadic reflection, a new language construct for explicitly converting back and forth between representations of effects as behavior and as data. Using monadic reflection to characterize concisely all effects expressible with a given monad, we can give a precise meaning to the notion of simulating one effect by another, more general one. We isolate a simple condition allowing such a simulation, and in particular show that any monadic effect can be simulated by a continuation monad. In other words, under relatively mild assumptions on the base language (allowing formation of a suitably large answer type), control becomes a universal effect. Concluding the development, we show that this u... | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||