Mike Bostock

Hello 
nice solution but due to compatibility of fromatToParts! what do you think about the following class

class UniParser {
    constructor(local) {
        this._decimal = new Intl.NumberFormat(local, { useGrouping: false }).format(9.9).charAt(1);
        this._numerals = new Intl.NumberFormat(local, { useGrouping: false }).format(9876543210).split('').reverse();
        this._regx = new RegExp('[' + this._numerals + ']', "g");

    }
    parse(string) {
        let self = this;
        return Number(string.replace(this._regx, function (match) {
            return self._numerals.indexOf(match);
        }).replace(self._decimal, '.'));
    }
}

Yaser Selim

ah I forgot to replace grouping symbol ! additional replace can solve it

If it works for you, then great! The ideas presented in this notebook are just a sketch (a hack), trying to adapt Intl.NumberFormat to support localized parsing. As I mentioned, this makes a number of assumptions and almost certainly doesn’t generalize to all locales, but maybe it’s better than nothing?

found it!           
  return (self._numerals.indexOf(match)>=0) ? self._numerals.indexOf(match) : '';

There is a bug in here when the group token is whitespace.

Repro test case:   numberParser('fr-CA', '1 999,99');   // returns NaN

This is because _group Regexp is constructed as /[ ]/, and what we really want in this case is probably /[\s]/

Mike Reinstein

Here is my fix:

    // modified slightly from his observable. because the group separator doesn't handle whitespace.
    // e.g., numberParser('fr-CA', '1 999,99')  // returns NaN
    // this is because the group regex is generated as /[ ]/ when we really want /[\s]/
    const groupVal = parts.find(d => d.type === 'group').value;

    const isWhiteSpace = /\s/.test(
        typeof groupVal === 'number' ? String.fromCodePoint(groupVal) : groupVal.charAt(0)
    )

    const _group = new RegExp(`[${isWhiteSpace ? '\\s' : groupVal}]`, 'g');