Fabric and TurboModule support

[naftalibeder]

This is a discussion on our Fabric/Turbo module migration.
See PR #2819 for POC and concrete details.

mfazekas

---

This is an outline of what it would take to upgrade to Fabric and TurboModules (the new React Native renderer and native module interface).

I think it's too big of a job for just me, so I'd like to see if there's any interest in collaborating on it.

cc @mfazekas

• Update dependencies
  • Update the RNMapbox library from 0.67.0 (link)
  • Update the Example project from 0.68.1 (link)
• Create spec to support code generation
  • TurboModule, e.g. NativeRNMapbox.ts (link)
  • Fabric components, e.g. CameraNativeComponent.ts (link)
• Migrate codebase to new API
  • Migrating findNodeHandle / getting a HostComponent (link)
  • Migrating .measure*() Not used?
  • Migrating off setNativeProps (link)
  • Move the call to requireNativeComponent to a separate file (link)
  • Migrating from dispatchViewManagerCommand to codegenNativeCommands (link)
  • Update native method interfaces (link)
• Enable native code generation
  • Android (link)
  • iOS (link)

[systemlevel]

hi @naftalibeder, let's talk a bit more. Would you be interested in being sponsored to undertake this work? Would that free up your time to dedicate to this?

[naftalibeder]

Thanks for the offer! I would certainly prefer to just collaborate - do you have any availability at all? If not, I could probably pull off 4-5 hours a week, if you want to contact me at my email (in profile). Either way we should loop in @mfazekas so we don't do this in a way that contradicts his philosophy for the repo.

[mfazekas]

@naftalibeder I agree that we should try adopting TurboModules.

• I'm ok if that is v10 only. (so we don't need to deal with MapLibre or old MapboxGL)
• Ideally we should start with a simple component, and continue based on that.

Also there is https://docs.expo.dev/modules/module-api/ as I understood they also promise fabric/new architecture support but with much less ceremony (Swift + Kotlin DSL instead of C++ generated code) - https://www.youtube.com/watch?v=te7bwfNsRKg&t=4217s
But I'm not sure how mature, but sounds promising and much simpler than the TurboModule system

[alaindimabuyo]

commenting so I can be notified, for now I cant use the v10 of mapbox because of the fabric issue on android still using maplibre on android and mapbox v10 on ios so please dont drop maplibre if v10 is not as stable as it is on the new architecture thank you!

[janbaykara]

@mfazekas that PR is exciting! Curious what timelines look like on this and what could unblock progress, whether that be sponsorship or something else.

[mfazekas]

@janbaykara yes having my time spent on this sponsored, would be a great help, I cannot really justify spending my free time on this. Also last time I've checked the react-native codegen has various limitations we'd have to work around that.

[janbaykara]

OK good to know. It would be cool if people or orgs who are willing to sponsor this work would shout out in the comments and maybe raise some awareness that this is a fundable project. CC'ing @systemlevel who previously sounded like they might be willing to given their last comment. I wonder if the Mapbox team would also be interested in contributing funds?

[mfazekas]

As an update #3022, was merged into main. 10.1* releases should include initial fabric support. MapView should work but Sources/Layers and a lot of other stuff aren't going to work yet. Thanks much @j-piasecki