Big upgrade to roc-lang/unicode package with verion 3.0.0:
On my AMD Ryzen 7 9700X, I can process ASCII at 1.30 GB/s, multilingual text at 164 MB/s, emoji sequences at 210 MB/s, and the official conformance corpus at 146 MB/s.
That's amazing! :heart_eyes:
How does this performance compare to other language implementations?
Very favourably :wink:
I left a benchmark script in the repo that compares against a representative set of implementations I could find from C, Rust, and Go.
I migrated the Exercism exercises to the new Unicode library. It only required three changes:
Grapheme.split with Grapheme.ownedList(Str) instead of a Try(List(Str), _) so I removed some error checking.Easy peasy! :smile:
Perhaps owned isn't the clearest name. split was clearer imho.
Aurélien Geron said:
Perhaps
ownedisn't the clearest name.splitwas clearer imho
Id someone wants to make a PR I'd be happy to change it, and make a patch release
Last updated: Aug 12 2026 at 12:35 UTC