Bug 57723 – Spinbuttons on [type="number"] etc. with [step="any"] do not work

• Component: Forms
• Status: RESOLVED FIXED

[alexander farkas]

1. simply add <input type="number" step="any" />
2. Try to use the spinbuttons

The first mousepress will change the value to 0 after that, the spinnbutton doesn't work. I would suggest, if step is "any" the button "spins" the value with the default step (like it would if no step is configured).

[Kent Tamura]

> I would suggest, if step is "any" the button "spins" the value with the default step (like it would if no step is configured).

That sounds good.

[Joseph Pecoraro]

We could probably then fix up RangeInputType::handleKeydownEvent
which has the following comment:

    // FIXME: We can't use stepUp() for the step value "any". So, we increase
    // or decrease the value by 1/100 of the value range. Is it reasonable?

[Shinya Kawanaka]

Created attachment 99054 [details]
Patch

[Kent Tamura]

Comment on attachment 99054 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=99054&action=review

> Source/WebCore/ChangeLog:17
> +        * html/HTMLInputElement.cpp:
> +        (WebCore::HTMLInputElement::applyStep):
> +        (WebCore::HTMLInputElement::getAllowedValueStepWithDecimalPlacesWithAllowingStepAny):
> +        (WebCore::HTMLInputElement::stepUp):
> +        (WebCore::HTMLInputElement::stepDown):
> +        (WebCore::HTMLInputElement::stepUpFromRenderer):
> +        * html/HTMLInputElement.h:

You should write summaries of changes in each of files/functions.

> Source/WebCore/html/HTMLInputElement.cpp:380
> +void HTMLInputElement::applyStep(double count, bool allowsStepAny, ExceptionCode& ec)

Do not introduce bool parameter.  We prefer enum to bool strongly.
e.g.
enum AnyStepHandling { RejectAny, AnyIsDefaultStep };

> Source/WebCore/html/HTMLInputElement.cpp:443
> +    ASSERT(step);
> +    double defaultStep = m_inputType->defaultStep();
> +    double stepScaleFactor = m_inputType->stepScaleFactor();
> +    if (!isfinite(defaultStep) || !isfinite(stepScaleFactor))
> +        return false;

This part is duplicate with getAllowedValueStepWithDecimalPlaces().  It's not good.

The difference between this function and getAllowedValueStepWithDecimalPlaces() is only "any" handling. I think it's ok to add the enum parameter to getAllowedValueStepWithDecimalPlaces().

[Shinya Kawanaka]

Created attachment 99260 [details]
Patch

[Kent Tamura]

Comment on attachment 99260 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=99260&action=review

> Source/WebCore/html/HTMLInputElement.h:67
> +    bool getAllowedValueStepWithDecimalPlaces(AnyStepHandling, double*, unsigned*) const;

Can you make getAllowedValeuStepWithDecimalPlaces() and AnyStepHandling private?

[Shinya Kawanaka]

Created attachment 99670 [details]
Patch

[Kent Tamura]

Comment on attachment 99670 [details]
Patch

Great patch.  Thank you!

[WebKit Review Bot]

Comment on attachment 99670 [details]
Patch

Clearing flags on attachment: 99670

Committed r90385: <http://trac.webkit.org/changeset/90385>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Kent Tamura]

(In reply to comment #2)
> We could probably then fix up RangeInputType::handleKeydownEvent
> which has the following comment:
> 
>     // FIXME: We can't use stepUp() for the step value "any". So, we increase
>     // or decrease the value by 1/100 of the value range. Is it reasonable?

I'm not sure if assuming "any" 1.0 for type=range is better.  We have another issue about keyboard operation with the default step, Bug 63574.