A Framework for Dependent Types and Effects

We generalise Levy's call-by-push-value (CBPV) to dependent type theory, to gain a better understanding of how to combine dependent types with effects. We define a dependently typed extension of CBPV, dCBPV-, and show that it has a very natural small-step operational semantics, which satisfies subject reduction and (depending on the effects present) determinism and strong normalization, and an elegant categorical semantics, which - surprisingly - is no more complicated than the simply typed semantics. We have full and faithful translations from a dependently typed version of Moggi's monadic metalanguage and of a call-by-name (CBN) dependent type theory into dCBPV- which give rise to the expected operational behaviour. However, it turns out that dCBPV- does not suffice to encode call-by-value (CBV) dependent type theory or the strong (dependent) elimination rules for positive connectives in CBN-dependent type theory. To mend this problem, we discuss a second, more expressive system dCBPV+, which additionally has a principle of Kleisli extension for dependent functions. We obtain the desired CBV- and CBN-translations of dependent type theory into dCBPV+. It too has a natural categorical semantics and operational semantics. However, depending on the effects we consider, we may lose uniqueness of typing, as the type of a computation may become more specified as certain effects are executed. This idea can be neatly formalized using a notion of subtyping. We hope that the theoretical framework of this paper on the one hand provides at least a partial answer to the fundamental type theoretic question of how one can understand the relationship between computational effects and dependent types. On the other hand, we hope it can contribute a small-step towards the ultimate goal of an elegant fully fledged language for certified effectful programming.