Default to en-US if locale is not found

[oss92]

Fixed the isAlpha and isAlphanumeric methods to support locales that are not found by defaulting to en-US.

• Old code: If locale is en-GB for example, alpha[locale] will be undefined and will output an error Cannot read property 'test' of undefined.

validator.isAlpha = function (str, locale) {
        locale = locale || 'en-US';
        return alpha[locale].test(str);
};

• New code: If alpha[locale] is undefined, alpha['en-US'] will be used instead.

    validator.isAlpha = function (str, locale) {
        var regex = alpha[locale] || alpha['en-US'];
        return regex.test(str);
    };

[chriso]

Passing an invalid locale should be an error, or should at least cause it to return false like it does in the isMobilePhone() validator. Silently replacing an invalid locale with en-US will cause unexpected results.

The problem is with sequelize, which should not be passing true as the second argument (sequelize/sequelize#5377).

[oss92]

It was not really about sequelize. What about en-GB for example? Which will return an error even though it is the same character set as en-US.

The general case is that alphanumeric characters (POSIX/C locale) are 0-9A-Za-z. Also since it is already locale || 'en-US' which fallbacks to en-US is no locale was assigned then it makes sense that if the locale was not found to default to en-US.

Also your old code before localized alphanumerics and alphas was only en-US.

This will produce some problems with languages with unicode characters. (Chinese, Hebrew, Arabic etc..) but is performant on most other languages.

[chriso]

What about en-GB for example?

en-GB (en-AU etc) should be an explicit alias for en-US. I'd be happy to accept a PR to add additional locales/aliases.

diff --git i/validator.js w/validator.js
index c48fa9e..2662961 100644
--- i/validator.js
+++ w/validator.js
@@ -79,6 +79,9 @@
       , decimal = /^[-+]?([0-9]+|\.[0-9]+|[0-9]+\.[0-9]+)$/
       , hexcolor = /^#?([0-9A-F]{3}|[0-9A-F]{6})$/i;

+    alpha['en-GB'] = alpha['en-US']
+    alphaNumeric['en-GB'] = alphaNumeric['en-US']
+
     var ascii = /^[\x00-\x7F]+$/
       , multibyte = /[^\x00-\x7F]/
       , fullWidth = /[^\u0020-\u007E\uFF61-\uFF9F\uFFA0-\uFFDC\uFFE8-\uFFEE0-9a-zA-Z]/

The general case is that alphanumeric characters (POSIX/C locale) are 0-9A-Za-z.

Agreed. This is why when you don't specify a locale, it defaults to en-US. This was a backwards-compatible change from #477. If you want the default locale, don't specify one! Sequelize is calling validator.isAlphanumeric(string, true) which is invalid: the interface is clearly defined and true is not a locale.

it makes sense that if the locale was not found to default to en-US.

Completely disagree. You're suggesting that validator.isAlpha('foobar', 'zh-CN') should silently change zh-CN to en-US and then return true.

[oss92]

That was a fruitful conversation. Agreed 👍