March 4, 2026
Fast regex, faster comment wars
RE#: how we built the fastest regex engine in F#
F#’s regex rocket goes open source—fans cheer while Microsoft, Haskell and light-mode squabbles erupt
TLDR: An F# team open-sourced RE#, a blazing-fast regex tool that adds useful “AND” and “NOT” matching without slowing down. Comments celebrate the speed, gripe about Microsoft ties and light-mode readability, and point to Haskell rivals—making a nerdy benchmark win feel like a full-on community rumble.
The devs behind RE# just open-sourced a turbocharged regex engine—think “find text” on rocket fuel—and the comment section turned into a victory parade with a side of drama. Fans called it “beautiful work” and rushed to the interactive web app to try those headline-grabbing extras: explicit NOT (aka complement), AND (intersection), and clever lookarounds, all while staying fast. That’s rare. The project even earned a serious academic nod at POPL 2025, but the crowd came for the vibes.
Strongest takes? F# loyalists basked in glory while tossing shade at the Microsoft association—one commenter claimed F# would beat Haskell in popularity “if it wasn’t Microsoft.” Meanwhile, a nitpick brigade arrived with a very internet complaint: light mode made some code unreadable (they’re not wrong), proving that even the fastest engine can’t outrun UI gripes. Old-schoolers flexed Usenet nostalgia and loved that RE# makes “does not contain X” practical, while another commenter dropped receipts linking Haskell work tied to a POPL 2024 paper—cue friendly rivalry over who did it first and best. The overall mood? Excited, a little petty, and extremely online. It’s the perfect storm: academic cred, real-world speed, and a comments section ready to rumble.
Key Points
- •RE# is an open-sourced F# regex engine described in a POPL 2025 paper and claimed to be the fastest on extensive benchmarks.
- •It supports union, intersection, complement, and context-aware lookarounds while preserving O(n) search-time complexity.
- •RE# aims to combine expressiveness with linear-time guarantees, addressing limitations of Thompson-style and backtracking engines.
- •Its design is inspired by SRM credential scanning (2019) and the .NET NonBacktracking engine (2023).
- •Brzozowski derivatives are a core mechanism, and the engine enforces leftmost-longest matching semantics rather than PCRE defaults.