KeyEvent::text should be None for backspace

[rparrett]

Tested Winit 0.29 and bdeb257 with Chrome 120 and Firefox 121.

MacOS 13.6.2 (native) produces Some("\u{8}"), and from a glance at the code, it looks like Windows is probably doing the same. Linux doesn't seem as clear and I'm unable to test.

It would be ideal if this behavior was the same on all platforms.

[kchibisov]

Use https://docs.rs/winit/latest/winit/keyboard/enum.NamedKey.html#method.to_text if you want text repr of NamedKey.

[rparrett]

Okay.

It looks like I may have misread the docs, and the behavior on MacOS is the (and I assume windows) is buggy? KeyEvent::text should only be Some for "real characters"? Am I understanding that right?

My issue is that this behavior is not consistent across platforms.

[daxpedda]

I guess reopening because of platform inconsistencies.

[daxpedda]

On X11 it seems to be the same issue: pressing backspace returns text: Some("\u{8}").
@kchibisov I'm unsure what the supposed behavior should be at this point, but it seems to me that your initial assessment is correct. Meaning this has to be fixed in apparently most backends.

[kchibisov]

I guess reopening because of platform inconsistencies.

you can, it's just different issue to the originally reported one.

@kchibisov I'm unsure what the supposed behavior should be at this point, but it seems to me that your initial assessment is correct. Meaning this has to be fixed in apparently most backends.

The issue is a bit more complicated since there's generally no control on how NamedKey values text is filled and the end caveats. Setting always None for it also won't work the way the current input is done, because named key could be at the end of compose input thus it'll have a valid input.

[rparrett]

Is there some documentation / standard that is governing this behavior that you can point me to? The distinction between what KeyCode::text and Key::to_text are providing seems arbitrary and confusing.

[kchibisov]

I'd at least remove text from events having to_text, but I'm not sure how I feel about remove text in general from NamedKey variants.