Porting the unicode library · contributing

Stream: contributing

Topic: Porting the unicode library

Aurélien Geron (Jun 29 2026 at 20:34):

I've finished migrating all the Roc code in the unicode library (draft PR #31), all the tests pass and all the examples work. :grinning:

That said, there are a few TODOs:

I had to workaround some compiler bugs (e.g., crash "${foo}" does not work): there will be a few local cleanups when the bugs are resolved.
A few builtin functions were missing (e.g., List.keep_oks and List.reverse) so I just added basic implementations for now, and will remove them once the builtins are available.
I changed CodePoint from CodePoint :: U32.{...} to CodePoint :: {cp : U32} to work around the fact that it wasn't possible to make the former work, but @Jared Ramirez solved this so I'll soon go back to CodePoint :: U32. Note that the original version used CodePoint :: InternalCP.{...} and InternalCP :: U32.{...}.
Migrating the non-Roc code (e.g., CI stuff and docs).

Btw, the use of Internal* types makes me wonder whether this is a satisfying & sufficient convention or whether there should be full intra-package visibility.

Richard Feldman (Jun 29 2026 at 20:56):

there's an alternative to Internal now, but I haven't tried it out yet

Richard Feldman (Jun 29 2026 at 20:56):

it's how we do Builtin.roc except you export each of the individual types - so you have one big type which you don't expose directly, and then inside that one big type each of its associated types can all reference each other

Richard Feldman (Jun 29 2026 at 20:57):

and then you expose those at the package boundary without exposing the outer type that they're all under

Richard Feldman (Jun 29 2026 at 20:57):

I haven't actually tried this though, and I'm not sure if the outer type ends up in error messages (it might; I'm not sure)

Richard Feldman (Jun 29 2026 at 20:58):

if you have time, would be great to have gh issues for the things you ran into that didn't work! :smiley:

Jared Ramirez (Jun 29 2026 at 21:28):

the PR with the Codepoint := U32 landed, so should be g2g now!

Luke Boswell (Jun 30 2026 at 07:32):

@Aurélien Geron do you mind if I push this along a little too?

I think it's fine for the test gen script to depend on roc-parser -- hopefully we can cut a release really soon.

I'm thinking of replacing the CI scripts with setup-roc and an updated release workflow etc and if everything is green we can cut a release. :grinning_face_with_smiling_eyes:

Aurélien Geron (Jun 30 2026 at 07:42):

Yes, thanks @Luke Boswell . I'm traveling for the next 4 weeks, so I won't be as available. If you want to jump in now, it's actually great timing. :folded_hands:

Luke Boswell (Jun 30 2026 at 07:43):

Great, if you dont mind I might mint a roc-parser release and then unicode too -- hopefully this evening

Aurélien Geron (Jun 30 2026 at 07:46):

Awesome, go for it! :tada:

Aurélien Geron (Jun 30 2026 at 08:26):

Richard Feldman said:

if you have time, would be great to have gh issues for the things you ran into that didn't work! :smiley:

Sure:

Richard Feldman (Jun 30 2026 at 11:06):

thank you, I really appreciate it! :smiley:

Anton (Jun 30 2026 at 11:26):

A few builtin functions were missing (e.g., List.keep_oks and List.reverse)

List.reverse is now called List.rev

Last updated: Jul 23 2026 at 13:15 UTC