Translating proofs from an impredicative type system to a predicative one

TL;DR

This paper introduces an algorithm and tool for translating proofs from impredicative to predicative type systems, enabling proof reuse across different proof assistants like Coq and Agda, especially when impredicativity is non-essential.

Contribution

It presents a novel translation algorithm using universe polymorphism and unification solving, implemented in the Predicativize tool, facilitating proof sharing between impredicative and predicative systems.

Findings

Successfully translated non-trivial proofs like Bertrand's Postulate and Fermat's Little Theorem

The translation handles many proofs that do not rely on impredicativity

The algorithm is partial but effective in practical scenarios

Abstract

As the development of formal proofs is a time-consuming task, it is important to devise ways of sharing the already written proofs to prevent wasting time redoing them. One of the challenges in this domain is to translate proofs written in proof assistants based on impredicative logics, such as Coq, Matita and the HOL family, to proof assistants based on predicative logics like Agda, whenever impredicativity is not used in an essential way. In this paper we present an algorithm to do such a translation between a core impredicative type system and a core predicative one allowing prenex universe polymorphism like in Agda. It consists in trying to turn a potentially impredicative term into a universe polymorphic term as general as possible. The use of universe polymorphism is justified by the fact that mapping an impredicative universe to a fixed predicative one is not sufficient in most cases. During the algorithm, we need to solve unification problems modulo the max-successor algebra on universe levels. But, in this algebra, there are solvable problems having no most general solution. We however provide an incomplete algorithm whose solutions, when it succeeds, are most general ones. The proposed translation is of course partial, but in practice allows one to translate many proofs that do not use impredicativity in an essential way. Indeed, it was implemented in the tool Predicativize and then used to translate semi-automatically many non-trivial developments from Matita's arithmetic library to Agda, including Bertrand's Postulate and Fermat's Little Theorem, which were not available in Agda yet.