[WebdriverBidi] Remove invalid code points

[jrandolf-2]

The removed values are not single code points. They are not relevant to the test.

[jgraham]

I believe the intent is that they're all single grapheme clusters. The spec has an open issue about this. Instead of removing them, I suggest moving them to a separate test that's marked as tentative.

[jrandolf-2]

I believe the intent is that they're all single grapheme clusters. The spec has an open issue about this. Instead of removing them, I suggest moving them to a separate test that's marked as tentative.

Could you please link the issue? Coming from UIEvents, this is not feasible given that key events can only contain single Unicode scalar values. Also, on all platforms, the maximum customizability provided by the OS to the user for keyboards is emitting single Unicode scalar values, not multiple.

I think we should still remove these tests. They are completely invalid and I would not say these tests will 'tentatively' be valid consider the above factors.

[jgraham]

Could you please link the issue? Coming from UIEvents, this is not feasible given that key events can only contain single Unicode scalar values.

It's just an inline issue: https://w3c.github.io/webdriver/#dfn-process-a-key-action

Also, on all platforms, the maximum customizability provided by the OS to the user for keyboards is emitting single Unicode scalar values, not multiple.

Well that's arguably not true; something like a virtual keyboard can happily emit arbitrarily complex strings and in particular it's pretty normal to have a single user input emit something like a flag emoji which is multiple unicode scalar values. Presumably in terms of UI events this maps down to composition events rather than simple key events, but it's absolutely a case we have to deal with somehow because something like send_keys("🇪🇺") has to do what you'd expect.

So I think the most useful thing to do here would be to work out how to model character composition in WebDriver. Some time ago I put up a proposal focused on explicit IME support at w3c/webdriver#1683. Review of that would be appreciated. But maybe we should also support cases that don't exactly match the IME model.

[jrandolf-2]

Could you please link the issue? Coming from UIEvents, this is not feasible given that key events can only contain single Unicode scalar values.

It's just an inline issue: https://w3c.github.io/webdriver/#dfn-process-a-key-action

Also, on all platforms, the maximum customizability provided by the OS to the user for keyboards is emitting single Unicode scalar values, not multiple.

Well that's arguably not true; something like a virtual keyboard can happily emit arbitrarily complex strings and in particular it's pretty normal to have a single user input emit something like a flag emoji which is multiple unicode scalar values. Presumably in terms of UI events this maps down to composition events rather than simple key events, but it's absolutely a case we have to deal with somehow because something like send_keys("🇪🇺") has to do what you'd expect.

So I think the most useful thing to do here would be to work out how to model character composition in WebDriver. Some time ago I put up a proposal focused on explicit IME support at w3c/webdriver#1683. Review of that would be appreciated. But maybe we should also support cases that don't exactly match the IME model.

Indeed we should work out some sort of composition model, however this PR is independent of that. More specifically, we should not provide the ability to send multi-code-point characters for the basic key actions.

Side note: Regarding composition, models are dependent on the OS and keyboard, so I think it'd be difficult to model a universal one for Webdriver. In Puppeteer, we solve this by just using "insertText" events on inputs. This assumes that users really only care about the "input" events that get created and nothing else such as composition events.