Abstract

In imperative languages I/O is realized through sequences of sideeffecting function applications/ procedure invocations. This seems to be a suitable way of specifying I/O since it coincides with an intuitive understanding of it as sequences of actions. In contrast, functional languages carefully have to avoid side-effects to sustain referential transparency. Many different solutions, such as dialogues, continuations, monads and uniqueness typing have been proposed. The I/O facilities of Sac are based on uniqueness typing. Instead of using an explicit type attribute as in Clean, unique types are introduced as special modules called classes. To provide a syntax as close as possible to that of imperative languages, we propose two new mechanisms to be included on top of classes in Sac: a call-by-reference mechanism and global objects. Although a combination of both mechanisms allows for very imperative-like notations, we can define a purely functional semantics. Thus we combine the advant...

Citations