Bug 161506

Created attachment 287932 [details] Patch

Seems like a reasonable improvement; the decision on this should be made by an iOS keyboards expert.

Comment on attachment 287932 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=287932&action=review I like this idea. Can this be tested? > Source/WebKit2/WebProcess/WebPage/ios/WebPageIOS.mm:2517 > +static const char inputPatternNumber[] = "^(\\[0-9\\]|\\\\d)(\\+|\\*|\\{[0-9,]+\\})?$"; > +static const char inputPatternDecimal[] = "^(\\[0-9\\]|\\\\d)?(\\+|\\*|\\{[0-9,]+\\})?\\\\\\.(\\[0-9\\]|\\\\d)(\\+|\\*|\\{[0-9,]+\\})?$"; This is clever, but matching regular expressions with regular expressions seems like a step in the wrong direction. We were doing string comparison before, which wasn't the greatest, but could we at least continue to do string comparison?

I agree the regex isn't an ideal solution. I was trying to brainstorm something more flexible than string comparisons though, and this was the best I came up with. One complaint of the current string comparison whitelisting is that it makes it hard for people to use the pattern attribute for actual validations. If you want to collect a 5-digit number, the best pattern would be "\d{5}", but that causes iOS to opt-out of the number pad keyboard because it doesn't match the whitelist. It's not obvious what the string comparison should be for a decimal pad, since the naïve "[0-9\.]*" would actually allow multiple decimal points to validate. Suggested patterns from online resources include "\d*\.\d*" or "[0-9]*\.[0-9]*", but both of those make the decimal point required.

The testing of the regular expression with another regular expression seems much more useful to me than extreme limits that are there now. I would like to use patterns like this: \d+(\.\d*)? // decimal number \d+(\.\d{2})? // dollars with optional cents ((\d+)?(\.\d{2})?)+ // optional dollar digits with optional cents Etc. Or maybe an expression for validating IP addresses.

Comment on attachment 287932 [details] Patch Patches that have been up for review since 2016 are almost certainly too stale to be relevant to trunk in their current form. If this patch is still important please rebase it and post it for review again.

As an alternative to this (and of trying to use pattern to determine the correct keyboard in general), it might be better to add support for the `inputmode` attribute which would allow authors to specify a hint for which keyboard type to use: https://html.spec.whatwg.org/multipage/interaction.html#input-modalities%3A-the-inputmode-attribute There was a recent intent to ship for Blink/Chrome: https://groups.google.com/a/chromium.org/forum/#!searchin/blink-dev/inputmode/blink-dev/MAHQkH4vvUQ/i0yCobrMCQAJ

I’ve created a bug for adding `inputmode` support. https://bugs.webkit.org/show_bug.cgi?id=183621

inputmode support was just merged, and provides a way to get a decimal keyboard via inputmode="decimal", so I'll close this as a duplicate of #183621 *** This bug has been marked as a duplicate of bug 183621 ***