Verified Parallel String Matching in Haskell

Abstract

In this paper, we prove correctness of parallelizing a string matcher using Haskell as a theorem prover. We use refinement types to specify correctness properties, Haskell terms to express proofs and Liquid Haskell to check correctness of proofs. First, we specify and prove that a class of monoid morphisms can be parallelized via parallel monoid concatenation. Then, we encode string matching as a morphism to get a provably correct parallel transformation. Our 1839LoC prototype proof shows that Liquid Haskell can be used as a fully expressive theorem prover on realistic Haskell implementations.