Is there interest to use semantic scan code?

[stephanemagnenat]

I am working on a GUI library in Rust, aiming at making it backend-independent. So, my library defines abstract events, to be filled by the backend-specific data. For the scan code, I found the scancode crate, based on libSDL2, itself based on the USB HID standard. This seems a good way to encode scan codes, but unfortunately the scancode crate seems unmaintained, and I saw that winit currently uses a u32 for the scan code.

So I'm wondering whether there is interest to:

• use semantic scancode matching the USB HID standard, possibly based on the scancode crate,
• move this and the VirtualKeyCode struct into their own crate as part of rust-windowing.

[maroider]

* use semantic scancode matching the USB HID standard, possibly based on the `scancode` crate,

We're already working on implementing something like this (see #1806 and src/event.rs and src/keyboard.rs in #1788).

Do note that USB HID scancodes don't get forwarded to applications, as far as I'm aware.

move this and the VirtualKeyCode struct into their own crate as part of rust-windowing.

I have had thoughts along these lines myself, and we did consider using keyboard-types in Winit while discussing things in #753, but we decided to use our own enumerations and structs for the time being. There was also an attempt to pull out other types into a separate crate (#1298), but that effort has largely stalled.

I have been intending to bring this up again, but I'm currently trying to focus on implementing the API in #753/#1788.

[stephanemagnenat]

Thank you for the answer, and pointing out the keyboard-types crate. Hopefully at some point the ecosystem converges to a common crate with well-accepted tradeoffs :-)