[core] fix(AsyncControllableInput): handle compound composition events

[Athenodoros]

There's an issue with the AsyncControllableInput, where Korean input can lead to incorrect values being stored.

The Bug

The cause is the way that Korean handles keyboard events: letters in the Korean language are combined into symbols, which are bubbled up as CompositionEvents. For instance, when typing the characters ㄷ -> ㅣ -> ㄱ -> ㅏ, they would successively render as ㄷ, 디, 딕, 디가 - note that with the last input the value splits out into two characters, which manifests as one CompositionEvent (디) ending and then a new one (가) immediately starting. I put together some logging to show the flow (times in seconds, logging code here):

Currently, this gives incorrect results with AsyncControllableInput: because the value prop can be updated on a delay (eg. example from the docs), when the first composition ends the stored value (디) is blown away by the old state (딕), which results in the incorrect final result 닥가 (note the extra ㄱ in the first character).

The solution

As far as I can tell, there's no way in the onCompositionEnd callback to determine whether another CompositionEvent is about to start. If that's the case, the best solution I can think of is to manually wait to unlock the component for edits after some time delay - I've chosen 10ms, because in my experimentation it was always <=1ms before the next event started.

Checklist

• Includes tests
• Update documentation

Is there documentation I should add?

[palantirtech]

Thanks for your interest in palantir/blueprint, @Athenodoros! Before we can accept your pull request, you need to sign our contributor license agreement - just visit https://cla.palantir.com/ and follow the instructions. Once you sign, I'll automatically update this pull request.

[adidahiya]

docs preview: https://52033-71939872-gh.circle-artifacts.com/0/packages/docs-app/dist/index.html#core/components/text-inputs.input-group

first input group in that example uses asyncControl={true}. could you verify the bug is fixed with korean characters there @Athenodoros?

[Athenodoros]

Yep, looks good to me - thanks @adidahiya!