Offer a "strict" mode for Unicode

[tilgovi]

The library currently counts characters reported by the length of the Node.textContent property which is a JavaScript String.

We could offer a strict mode wherein the offsets are calculated by using punycode and String.prototype.normalize to count unicode symbols.

[nickstenning]

One option that's made available to us by ES2015 is use of the string iterator to split a string on codepoints. I haven't thought this through fully, but I think it should be possible to just replace the two references to

node.nodeValue.length

with something like

[...node.nodeValue].length

(that syntax will result in an Array of strings each representing a single codepoint).

My reading of https://kangax.github.io/compat-table/es6/ is that this would currently be supported natively by:

• Chrome 46+
• Firefox 36+
• Safari 9+
• Edge 12+

(No IE support at all for string symbol iterators...)

It also looks like Babel may support mapping this to ES5, although I'm also not sure about that...

[nickstenning]

This may also imply the creation of lots of garbage for collection by the runtime in the form of these temporary arrays, but... ¯_(ツ)_/¯