fileformatpage2charclass.js

var SYSTEM = require('system'),
    HTTP = require('http-client'),
    OS = require('os');

var url = SYSTEM.args[1];

if (!url) {
    print('usage:');
    print('js ' + SYSTEM.args[0] + ' <fileformat.info URL>');
    OS.exit(1);
}

var fileContents = HTTP.read(url).decodeToString('utf-8'),
    charRegex = /\s*<td align="center"><a href="\/info\/unicode\/char\/.+?">U+(.+?)<\/a><\/td>\s*/g,
    matches = null,
    lastCharCode = null,
    classStartCharCode = null,
    classes = [];

var convertToUnicodeHex = function(charCode) {
    if (charCode > 0xFFFF) {
        charCode -= 0x10000;
        var highSurrogate = ((charCode & 0xFFC00) >> 10) + 0xD800;
        var lowSurrogate = (charCode & 0x3FF) + 0xDC00;
        return '\\u' + highSurrogate.toString(16).toUpperCase() + '\\u' + lowSurrogate.toString(16).toUpperCase();
    } else {
        var str = charCode.toString(16).toUpperCase();
        while (str.length < 4)
            str = '0' + str;
        return '\\u' + str;
    }
};

var charClassFromCharCodes = function(startCharCode, endCharCode) {
    if (startCharCode === endCharCode) {
        if (startCharCode < 0xFFFF)
            return convertToUnicodeHex(startCharCode);
        else
            return '"' + convertToUnicodeHex(startCharCode) + '"';
    } else {
        return convertToUnicodeHex(startCharCode) + '-' + convertToUnicodeHex(endCharCode);
    }
};

while (matches = charRegex.exec(fileContents)) {
    var charCode = parseInt(matches[1], 16);

    if (charCode > 0xFFFF) { // funky utf-16 escape sequence required
        if (classStartCharCode !== null) { // end the previous char class if it exists
            classes[classes.length - 1] += charClassFromCharCodes(classStartCharCode, lastCharCode) + ']';
            classStartCharCode = null;
        }

        classes.push(charClassFromCharCodes(charCode, charCode));
    }
    else if (classStartCharCode === null) {
        classStartCharCode = charCode;
        classes.push('[');
    }
    else if (charCode !== lastCharCode + 1) {
        classes[classes.length - 1] += charClassFromCharCodes(classStartCharCode, lastCharCode);
        classStartCharCode = charCode;
    }

    lastCharCode = charCode;
}

if (lastCharCode <= 0xFFFF && classStartCharCode !== null)
    classes[classes.length - 1] += charClassFromCharCodes(classStartCharCode, lastCharCode) + ']';

print(classes.join(' / '));