Certified derivative-based parsing of regular expressions

Abstract

We describe the formalization of a certified algorithm for regular expression parsing based on Brzozowski derivatives, in the dependently typed language Idris. The formalized algorithm produces a proof that an input string matches a given regular expression or a proof that no matching exists. A tool for regular expression based search in the style of the well known GNU grep has been developed with the certified algorithm, and practical experiments were conducted with this tool.