UC San Diego

Code deficiencies and bugs constitute an unavoidable part of software systems. In safety-critical systems, like aircrafts or medical equipment, even a single bug can lead to catastrophic impacts such as injuries or death. Formal verification can be used to statically track code deficiencies by proving or disproving correctness properties of a system. However, at its current state formal verification is a cumbersome process that is rarely used by mainstream developers, mostly because it targets non general purpose languages (e.g., Coq, Agda, Dafny).

We present Liquid Haskell, a usable program verifier that aims to establish formal verification as an integral part of the development process. Liquid Haskell naturally integrates the specification of correctness properties

as logical refinements of Haskell's types. Moreover, it uses the abstract interpretation framework of liquid types to automatically check correctness of specifications via Satisfiability Modulo Theories (SMT) solvers

requiring no explicit proofs or complicated annotations. Finally, the specification language is arbitrary expressive, allowing the user to write general correctness properties about their code, thus turning Haskell into a theorem prover.

Transforming a mature language --- with optimized libraries and highly tuned parallelism --- into a theorem prover enables us to verify a wide variety of properties on real world applications. We used Liquid Haskell to verify shallow invariants of existing Haskell code, e.g., memory safety of the optimized string manipulation library ByteString. Moreover, we checked deep, sophisticated properties of parallel Haskell code, e.g., program equivalence of a naive string matcher and its parallelized version. Having verified about 20K of Haskell code, we present how Liquid Haskell serves as a prototype verifier in a future where formal techniques will be used to facilitate, instead of hinder, software development.