Some general questions about architecture compared to RN

[etodanik]

I've been building an OOT platform on React Native for a while and would love to play around with Lynx since it looks to me that it could be architected slightly better when it comes to layering.

I have a few clarifications about some things that are typically pain points.

• Are you facilitating communication between the JS and the C++ native parts of Lynx via a seralization/deserialization message bus or are you binding C++/JS directly on the JS engine level? (e.g old RN style architecture vs new RN style architecture)
• What layout engine do you use for non-web platforms? (RN uses yoga, Flutter has its' on monolithic layout engine)
• On RN one of the biggest pain points of making a new platform is the fact that almost none of the C++ is compile-able on its' own (it all assumes Android as a target), do you have a "platform agnostic" C++ build process?
• On RN the ultimately biggest pain point is how the packaging happens. They have a big JS/TS package set that essentially needs to be composed with your own OOT platform ... It leads to terrible DX because you end up having to literally copy files at build time (react-native-windows does it), and kind of merge your own code with lots of their code. RN for windows ended up just making a JS script that copies their own sources & RN's original sources into the same folder before bundling, but that's a huge PITA. Any pointers to how you're planning to organize the built-ins (view/page/e.t.c) and layer them for custom platforms to implement?

I wanna give a shot to implementing an Unreal Engine lynx platform with native Slate widgets for views e.t.c

[lybvinci]

Firstly, thank you for your feedback.

I will now address the questions mentioned above:

1. We use C++ bindings directly on the JavaScript engine to avoid data serialization/deserialization.
2. We use our own layout engine.
3. We use ninja-build to compile our C++ code, allowing you to easily use targets for compilation.
4. Each target platform has its own macro definitions. When compiling for a specific target platform, we only package the target view code.

We hope you can implement this on Unreal Engine successfully