Protecting representation with effect encapsulation

Abstract

Representation exposure is a well-known problem in the object-oriented realm. Object encapsulation mechanisms have established a tradition for solving this problem based on a principle of reference containment. This paper proposes a novel type system which is based on a different principle, we call effect encapsulation, which confines side effects, rather than object references, according to an ownership structure. Compared to object encapsulation, effect encapsulation liberates us from the restriction on object referenceability and offers more flexibility. In this paper, we show that effect encapsulation can be statically type checked.