I've been kicking the tires of the new Doc.FloatInput and Doc.IntInput functions, provided as a solution to my original comments in issue #34. Thank you for that.

Your solution is superior to what I originally proposed, because it uses type="number" for the reasons you gave.

Testing on Windows 10 using Chrome 45, Firefox 42, Edge 20, and IE 11, it seems that, as you too observed, the various browsers differ on when they tell you the input has changed, and whether they tell you the actual string when the value is not currently a number, rather than just giving you a blank string.

There are other differences. Firefox echos the invalid text you enter, whereas Chrome suppresses invalid characters. Some browsers allow a '+' in front of the number, and IE 11 allows spaces before and after.

In order to prevent parse-format round trip changes, where "4." gets changed to "4", or "1e3" gets changed to "1000", as you type, CheckedInput<'T> stores the actual text in all three cases. But given the different behaviors of different browsers, the text serves no useful purpose to the caller in the Invalid and Blank cases.

With my original proposal, which used type="text", it made sense to have the text available in all cases, since all browsers would always give you the text, exactly as you typed it, with no suppression of characters. But now it seems the text is only reliable and therefore useful for the Valid case. It tells me whether the user entered "1e3" or "1000", and allows me to prime the field with either of these formats.

My suggestion is to use a private Var (or mutable value) which stores the text in all cases, but then push/pull the state to the caller's Var, which would only carry the text in the Valid case. The public union would be this:

1
2
3
4

type CheckedInput<'T> =
    | Valid of value: 'T * inputText: string
    | Invalid
    | Blank

As it stands today, I can simply ignore the text in the Invalid or Blank cases, but it is confusing to the programmer as to why it is there.

Two other things:

• I initialize my Var to be Blank "", but the input field shows as "0". This seems to be the case in all browsers.

• IntInputUncheched has some code that IntInput does not. Is that a mistake?

Thanks.