Abstract

A proof checking system may support syntax that is more convenient for users than its `official' language. For example LEGO (a typechecker for several systems related to the Calculus of Constructions) has algorithms to infer some polymorphic instantiations (e.g. pair 2 true instead of pair nat bool 2 true) and universe levels (e.g. Type instead of Type(4)). Users need to understand such features, but do not want to know the algorithms for computing them. In this note I explain these two features by non-deterministic operational semantics for "translating" implicit syntax to the fully explicit underlying formal system. The translations are sound and complete for the underlying type theory, and the algorithms (which I will not talk about) are sound (not necessarily complete) for the translations. This note is phrased in terms of a general class of type theories. The technique described has more general application. 1 Introduction Consider the usual formal system, !, for simp...

Citations