url-parser/url.js

url parser that works uniformly on all javascript platforms

!function(root){
  var punycode
    , freeDefine = typeof define == 'function' && typeof define.amd == 'object' && define.amd && define
    , freeExports = typeof exports == 'object' && exports
    , freeModule = typeof module == 'object' && module
    , freeRequire = typeof require == 'function' && require

    /** Highest positive signed 32-bit float value */
    , maxInt = 2147483647 // aka. 0x7FFFFFFF or 2^31-1

    /** Bootstring parameters */
    , base = 36
    , tMin = 1
    , tMax = 26
    , skew = 38
    , damp = 700
    , initialBias = 72
    , initialN = 128 // 0x80
    , delimiter = '-' // '\x2D'

    /** Regular expressions */
    , regexPunycode = /^xn--/
    , regexNonASCII = /[^ -~]/ // unprintable ASCII chars + non-ASCII chars
    , regexSeparators = /\x2E|\u3002|\uFF0E|\uFF61/g // RFC 3490 separators

    /** Error messages */
    , errors = {
      'overflow': 'Overflow: input needs wider integers to process',
      'not-basic': 'Illegal input >= 0x80 (not a basic code point)',
      'invalid-input': 'Invalid input'
    }

    /** Convenience shortcuts */
    , baseMinusTMin = base - tMin
    , floor = Math.floor
    , stringFromCharCode = String.fromCharCode

    /** Temporary variable */
    , key

  /*--------------------------------------------------------------------------*/

  /**
   * A generic error utility function.
   * @private
   * @param {String} type The error type.
   * @returns {Error} Throws a `RangeError` with the applicable error message.
   */
  function error(type) {
    throw RangeError(errors[type]);
  }

  /**
   * A generic `Array#map` utility function.
   * @private
   * @param {Array} array The array to iterate over.
   * @param {Function} callback The function that gets called for every array
   * item.
   * @returns {Array} A new array of values returned by the callback function.
   */
  function map(array, fn) {
    var length = array.length;
    while (length--) {
      array[length] = fn(array[length]);
    }
    return array;
  }

  /**
   * A simple `Array#map`-like wrapper to work with domain name strings.
   * @private
   * @param {String} domain The domain name.
   * @param {Function} callback The function that gets called for every
   * character.
   * @returns {Array} A new string of characters returned by the callback
   * function.
   */
  function mapDomain(string, fn) {
    return map(string.split(regexSeparators), fn).join('.');
  }

  /**
   * Creates an array containing the decimal code points of each Unicode
   * character in the string. While JavaScript uses UCS-2 internally,
   * this function will convert a pair of surrogate halves (each of which
   * UCS-2 exposes as separate characters) into a single code point,
   * matching UTF-16.
   * @see `punycode.ucs2.encode`
   * @see <http://mathiasbynens.be/notes/javascript-encoding>
   * @memberOf punycode.ucs2
   * @name decode
   * @param {String} string The Unicode input string (UCS-2).
   * @returns {Array} The new array of code points.
   */
  function ucs2decode(string) {
    var output = [],
        counter = 0,
        length = string.length,
        value,
        extra;
    while (counter < length) {
      value = string.charCodeAt(counter++);
      if ((value & 0xF800) == 0xD800 && counter < length) {
        // high surrogate, and there is a next character
        extra = string.charCodeAt(counter++);
        if ((extra & 0xFC00) == 0xDC00) { // low surrogate
          output.push(((value & 0x3FF) << 10) + (extra & 0x3FF) + 0x10000);
        } else {
          output.push(value, extra);
        }
      } else {
        output.push(value);
      }
    }
    return output;
  }

  /**
   * Creates a string based on an array of decimal code points.
   * @see `punycode.ucs2.decode`
   * @memberOf punycode.ucs2
   * @name encode
   * @param {Array} codePoints The array of decimal code points.
   * @returns {String} The new Unicode string (UCS-2).
   */
  function ucs2encode(array) {
    return map(array, function(value) {
      var output = '';
      if (value > 0xFFFF) {
        value -= 0x10000;
        output += stringFromCharCode(value >>> 10 & 0x3FF | 0xD800);
        value = 0xDC00 | value & 0x3FF;
      }
      output += stringFromCharCode(value);
      return output;
    }).join('');
  }

  /**
   * Converts a basic code point into a digit/integer.
   * @see `digitToBasic()`
   * @private
   * @param {Number} codePoint The basic (decimal) code point.
   * @returns {Number} The numeric value of a basic code point (for use in
   * representing integers) in the range `0` to `base - 1`, or `base` if
   * the code point does not represent a value.
   */
  function basicToDigit(codePoint) {
    return codePoint - 48 < 10
      ? codePoint - 22
      : codePoint - 65 < 26
        ? codePoint - 65
        : codePoint - 97 < 26
          ? codePoint - 97
          : base;
  }

  /**
   * Converts a digit/integer into a basic code point.
   * @see `basicToDigit()`
   * @private
   * @param {Number} digit The numeric value of a basic code point.
   * @returns {Number} The basic code point whose value (when used for
   * representing integers) is `digit`, which needs to be in the range
   * `0` to `base - 1`. If `flag` is non-zero, the uppercase form is
   * used; else, the lowercase form is used. The behavior is undefined
   * if flag is non-zero and `digit` has no uppercase form.
   */
  function digitToBasic(digit, flag) {
    //  0..25 map to ASCII a..z or A..Z
    // 26..35 map to ASCII 0..9
    return digit + 22 + 75 * (digit < 26) - ((flag != 0) << 5);
  }

  /**
   * Bias adaptation function as per section 3.4 of RFC 3492.
   * http://tools.ietf.org/html/rfc3492#section-3.4
   * @private
   */
  function adapt(delta, numPoints, firstTime) {
    var k = 0;
    delta = firstTime ? floor(delta / damp) : delta >> 1;
    delta += floor(delta / numPoints);
    for (/* no initialization */; delta > baseMinusTMin * tMax >> 1; k += base) {
      delta = floor(delta / baseMinusTMin);
    }
    return floor(k + (baseMinusTMin + 1) * delta / (delta + skew));
  }

  /**
   * Converts a basic code point to lowercase if `flag` is falsy, or to
   * uppercase if `flag` is truthy. The code point is unchanged if it's
   * caseless. The behavior is undefined if `codePoint` is not a basic code
   * point.
   * @private
   * @param {Number} codePoint The numeric value of a basic code point.
   * @returns {Number} The resulting basic code point.
   */
  function encodeBasic(codePoint, flag) {
    codePoint -= (codePoint - 97 < 26) << 5;
    return codePoint + (!flag && codePoint - 65 < 26) << 5;
  }

  /**
   * Converts a Punycode string of ASCII code points to a string of Unicode
   * code points.
   * @memberOf punycode
   * @param {String} input The Punycode string of ASCII code points.
   * @returns {String} The resulting string of Unicode code points.
   */
  function decode(input) {
    // Don't use UCS-2
    var output = [],
        inputLength = input.length,
        out,
        i = 0,
        n = initialN,
        bias = initialBias,
        basic,
        j,
        index,
        oldi,
        w,
        k,
        digit,
        t,
        length,
        /** Cached calculation results */
        baseMinusT;

    // Handle the basic code points: let `basic` be the number of input code
    // points before the last delimiter, or `0` if there is none, then copy
    // the first basic code points to the output.

    basic = input.lastIndexOf(delimiter);
    if (basic < 0) {
      basic = 0;
    }

    for (j = 0; j < basic; ++j) {
      // if it's not a basic code point
      if (input.charCodeAt(j) >= 0x80) {
        error('not-basic');
      }
      output.push(input.charCodeAt(j));
    }

    // Main decoding loop: start just after the last delimiter if any basic code
    // points were copied; start at the beginning otherwise.

    for (index = basic > 0 ? basic + 1 : 0; index < inputLength; /* no final expression */) {

      // `index` is the index of the next character to be consumed.
      // Decode a generalized variable-length integer into `delta`,
      // which gets added to `i`. The overflow checking is easier
      // if we increase `i` as we go, then subtract off its starting
      // value at the end to obtain `delta`.
      for (oldi = i, w = 1, k = base; /* no condition */; k += base) {

        if (index >= inputLength) {
          error('invalid-input');
        }

        digit = basicToDigit(input.charCodeAt(index++));

        if (digit >= base || digit > floor((maxInt - i) / w)) {
          error('overflow');
        }

        i += digit * w;
        t = k <= bias ? tMin : (k >= bias + tMax ? tMax : k - bias);

        if (digit < t) {
          break;
        }

        baseMinusT = base - t;
        if (w > floor(maxInt / baseMinusT)) {
          error('overflow');
        }

        w *= baseMinusT;

      }

      out = output.length + 1;
      bias = adapt(i - oldi, out, oldi == 0);

      // `i` was supposed to wrap around from `out` to `0`,
      // incrementing `n` each time, so we'll fix that now:
      if (floor(i / out) > maxInt - n) {
        error('overflow');
      }

      n += floor(i / out);
      i %= out;

      // Insert `n` at position `i` of the output
      output.splice(i++, 0, n);

    }

    return ucs2encode(output);
  }

  /**
   * Converts a string of Unicode code points to a Punycode string of ASCII
   * code points.
   * @memberOf punycode
   * @param {String} input The string of Unicode code points.
   * @returns {String} The resulting Punycode string of ASCII code points.
   */
  function encode(input) {
    var n,
        delta,
        handledCPCount,
        basicLength,
        bias,
        j,
        m,
        q,
        k,
        t,
        currentValue,
        output = [],
        /** `inputLength` will hold the number of code points in `input`. */
        inputLength,
        /** Cached calculation results */
        handledCPCountPlusOne,
        baseMinusT,
        qMinusT;

    // Convert the input in UCS-2 to Unicode
    input = ucs2decode(input);

    // Cache the length
    inputLength = input.length;

    // Initialize the state
    n = initialN;
    delta = 0;
    bias = initialBias;

    // Handle the basic code points
    for (j = 0; j < inputLength; ++j) {
      currentValue = input[j];
      if (currentValue < 0x80) {
        output.push(stringFromCharCode(currentValue));
      }
    }

    handledCPCount = basicLength = output.length;

    // `handledCPCount` is the number of code points that have been handled;
    // `basicLength` is the number of basic code points.

    // Finish the basic string - if it is not empty - with a delimiter
    if (basicLength) {
      output.push(delimiter);
    }

    // Main encoding loop:
    while (handledCPCount < inputLength) {

      // All non-basic code points < n have been handled already. Find the next
      // larger one:
      for (m = maxInt, j = 0; j < inputLength; ++j) {
        currentValue = input[j];
        if (currentValue >= n && currentValue < m) {
          m = currentValue;
        }
      }

      // Increase `delta` enough to advance the decoder's <n,i> state to <m,0>,
      // but guard against overflow
      handledCPCountPlusOne = handledCPCount + 1;
      if (m - n > floor((maxInt - delta) / handledCPCountPlusOne)) {
        error('overflow');
      }

      delta += (m - n) * handledCPCountPlusOne;
      n = m;

      for (j = 0; j < inputLength; ++j) {
        currentValue = input[j];

        if (currentValue < n && ++delta > maxInt) {
          error('overflow');
        }

        if (currentValue == n) {
          // Represent delta as a generalized variable-length integer
          for (q = delta, k = base; /* no condition */; k += base) {
            t = k <= bias ? tMin : (k >= bias + tMax ? tMax : k - bias);
            if (q < t) {
              break;
            }
            qMinusT = q - t;
            baseMinusT = base - t;
            output.push(
              stringFromCharCode(digitToBasic(t + qMinusT % baseMinusT, 0))
            );
            q = floor(qMinusT / baseMinusT);
          }

          output.push(stringFromCharCode(digitToBasic(q, 0)));
          bias = adapt(delta, handledCPCountPlusOne, handledCPCount == basicLength);
          delta = 0;
          ++handledCPCount;
        }
      }

      ++delta;
      ++n;

    }
    return output.join('');
  }

  /**
   * Converts a Punycode string representing a domain name to Unicode. Only the
   * Punycoded parts of the domain name will be converted, i.e. it doesn't
   * matter if you call it on a string that has already been converted to
   * Unicode.
   * @memberOf punycode
   * @param {String} domain The Punycode domain name to convert to Unicode.
   * @returns {String} The Unicode representation of the given Punycode
   * string.
   */
  function toUnicode(domain) {
    return mapDomain(domain, function(string) {
      return regexPunycode.test(string)
        ? decode(string.slice(4).toLowerCase())
        : string;
    });
  }

  /**
   * Converts a Unicode string representing a domain name to Punycode. Only the
   * non-ASCII parts of the domain name will be converted, i.e. it doesn't
   * matter if you call it with a domain that's already in ASCII.
   * @memberOf punycode
   * @param {String} domain The domain name to convert, as a Unicode string.
   * @returns {String} The Punycode representation of the given domain name.
   */
  function toASCII(domain) {
    return mapDomain(domain, function(string) {
      return regexNonASCII.test(string)
        ? 'xn--' + encode(string)
        : string;
    });
  }

  /*--------------------------------------------------------------------------*/

  /** Define the public API */
  punycode = {
    /**
     * A string representing the current Punycode.js version number.
     * @memberOf punycode
     * @type String
     */
    'version': '1.2.0',
    /**
     * An object of methods to convert from JavaScript's internal character
     * representation (UCS-2) to decimal Unicode code points, and back.
     * @see <http://mathiasbynens.be/notes/javascript-encoding>
     * @memberOf punycode
     * @type Object
     */
    'ucs2': {
      'decode': ucs2decode,
      'encode': ucs2encode
    },
    'decode': decode,
    'encode': encode,
    'toASCII': toASCII,
    'toUnicode': toUnicode
  };

  var url = {};
  url.parse = urlParse;
  url.resolve = urlResolve;
  url.resolveObject = urlResolveObject;
  url.format = urlFormat;
  url.Url = Url;

  function Url() {
    this.protocol = null;
    this.slashes = null;
    this.auth = null;
    this.host = null;
    this.port = null;
    this.hostname = null;
    this.hash = null;
    this.search = null;
    this.query = null;
    this.pathname = null;
    this.path = null;
  }

  // Reference: RFC 3986, RFC 1808, RFC 2396

  // define these here so at least they only have to be
  // compiled once on the first module load.
  var protocolPattern = /^([a-z0-9.+-]+:)/i,
      portPattern = /:[0-9]*$/,

      // RFC 2396: characters reserved for delimiting URLs.
      // We actually just auto-escape these.
      delims = ['<', '>', '"', '`', ' ', '\r', '\n', '\t'],

      // RFC 2396: characters not allowed for various reasons.
      unwise = ['{', '}', '|', '\\', '^', '`'].concat(delims),

      // Allowed by RFCs, but cause of XSS attacks.  Always escape these.
      autoEscape = ['\''].concat(unwise),
      // Characters that are never ever allowed in a hostname.
      // Note that any invalid chars are also handled, but these
      // are the ones that are *expected* to be seen, so we fast-path
      // them.
      nonHostChars = ['%', '/', '?', ';', '#'].concat(autoEscape),
      nonAuthChars = ['/', '@', '?', '#'].concat(delims),
      hostnameMaxLen = 255,
      hostnamePartPattern = /^[a-z0-9A-Z_-]{0,63}$/,
      hostnamePartStart = /^([a-z0-9A-Z_-]{0,63})(.*)$/,
      // protocols that can allow "unsafe" and "unwise" chars.
      unsafeProtocol = {
        'javascript': true,
        'javascript:': true
      },
      // protocols that never have a hostname.
      hostlessProtocol = {
        'javascript': true,
        'javascript:': true
      },
      // protocols that always have a path component.
      pathedProtocol = {
        'http': true,
        'https': true,
        'ftp': true,
        'gopher': true,
        'file': true,
        'http:': true,
        'ftp:': true,
        'gopher:': true,
        'file:': true
      },
      // protocols that always contain a // bit.
      slashedProtocol = {
        'http': true,
        'https': true,
        'ftp': true,
        'gopher': true,
        'file': true,
        'http:': true,
        'https:': true,
        'ftp:': true,
        'gopher:': true,
        'file:': true
      },
      querystring = require('querystring');

  function urlParse(url, parseQueryString, slashesDenoteHost) {
    if (url && typeof(url) === 'object' && url instanceof Url) return url;

    var u = new Url;
    u.parse(url, parseQueryString, slashesDenoteHost);
    return u;
  }

  Url.prototype.parse = function(url, parseQueryString, slashesDenoteHost) {
    if (typeof url !== 'string') {
      throw new TypeError("Parameter 'url' must be a string, not " + typeof url);
    }

    var rest = url;

    // trim before proceeding.
    // This is to support parse stuff like "  http://foo.com  \n"
    rest = rest.trim();

    var proto = protocolPattern.exec(rest);
    if (proto) {
      proto = proto[0];
      var lowerProto = proto.toLowerCase();
      this.protocol = lowerProto;
      rest = rest.substr(proto.length);
    }

    // figure out if it's got a host
    // user@server is *always* interpreted as a hostname, and url
    // resolution will treat //foo/bar as host=foo,path=bar because that's
    // how the browser resolves relative URLs.
    if (slashesDenoteHost || proto || rest.match(/^\/\/[^@\/]+@[^@\/]+/)) {
      var slashes = rest.substr(0, 2) === '//';
      if (slashes && !(proto && hostlessProtocol[proto])) {
        rest = rest.substr(2);
        this.slashes = true;
      }
    }

    if (!hostlessProtocol[proto] &&
        (slashes || (proto && !slashedProtocol[proto]))) {
      // there's a hostname.
      // the first instance of /, ?, ;, or # ends the host.
      // don't enforce full RFC correctness, just be unstupid about it.

      // If there is an @ in the hostname, then non-host chars *are* allowed
      // to the left of the first @ sign, unless some non-auth character
      // comes *before* the @-sign.
      // URLs are obnoxious.
      var atSign = rest.indexOf('@');
      if (atSign !== -1) {
        var auth = rest.slice(0, atSign);

        // there *may be* an auth
        var hasAuth = true;
        for (var i = 0, l = nonAuthChars.length; i < l; i++) {
          if (auth.indexOf(nonAuthChars[i]) !== -1) {
            // not a valid auth.  Something like http://foo.com/bar@baz/
            hasAuth = false;
            break;
          }
        }

        if (hasAuth) {
          // pluck off the auth portion.
          this.auth = decodeURIComponent(auth);
          rest = rest.substr(atSign + 1);
        }
      }

      var firstNonHost = -1;
      for (var i = 0, l = nonHostChars.length; i < l; i++) {
        var index = rest.indexOf(nonHostChars[i]);
        if (index !== -1 &&
            (firstNonHost < 0 || index < firstNonHost)) firstNonHost = index;
      }

      if (firstNonHost !== -1) {
        this.host = rest.substr(0, firstNonHost);
        rest = rest.substr(firstNonHost);
      } else {
        this.host = rest;
        rest = '';
      }

      // pull out port.
      this.parseHost();

      // we've indicated that there is a hostname,
      // so even if it's empty, it has to be present.
      this.hostname = this.hostname || '';

      // if hostname begins with [ and ends with ]
      // assume that it's an IPv6 address.
      var ipv6Hostname = this.hostname[0] === '[' &&
          this.hostname[this.hostname.length - 1] === ']';

      // validate a little.
      if (!ipv6Hostname) {
        var hostparts = this.hostname.split(/\./);
        for (var i = 0, l = hostparts.length; i < l; i++) {
          var part = hostparts[i];
          if (!part) continue;
          if (!part.match(hostnamePartPattern)) {
            var newpart = '';
            for (var j = 0, k = part.length; j < k; j++) {
              if (part.charCodeAt(j) > 127) {
                // we replace non-ASCII char with a temporary placeholder
                // we need this to make sure size of hostname is not
                // broken by replacing non-ASCII by nothing
                newpart += 'x';
              } else {
                newpart += part[j];
              }
            }
            // we test again with ASCII char only
            if (!newpart.match(hostnamePartPattern)) {
              var validParts = hostparts.slice(0, i);
              var notHost = hostparts.slice(i + 1);
              var bit = part.match(hostnamePartStart);
              if (bit) {
                validParts.push(bit[1]);
                notHost.unshift(bit[2]);
              }
              if (notHost.length) {
                rest = '/' + notHost.join('.') + rest;
              }
              this.hostname = validParts.join('.');
              break;
            }
          }
        }
      }

      if (this.hostname.length > hostnameMaxLen) {
        this.hostname = '';
      } else {
        // hostnames are always lower case.
        this.hostname = this.hostname.toLowerCase();
      }

      if (!ipv6Hostname) {
        // IDNA Support: Returns a puny coded representation of "domain".
        // It only converts the part of the domain name that
        // has non ASCII characters. I.e. it dosent matter if
        // you call it with a domain that already is in ASCII.
        var domainArray = this.hostname.split('.');
        var newOut = [];
        for (var i = 0; i < domainArray.length; ++i) {
          var s = domainArray[i];
          newOut.push(s.match(/[^A-Za-z0-9_-]/) ?
              'xn--' + punycode.encode(s) : s);
        }
        this.hostname = newOut.join('.');
      }

      var p = this.port ? ':' + this.port : '';
      var h = this.hostname || '';
      this.host = h + p;
      this.href += this.host;

      // strip [ and ] from the hostname
      // the host field still retains them, though
      if (ipv6Hostname) {
        this.hostname = this.hostname.substr(1, this.hostname.length - 2);
        if (rest[0] !== '/') {
          rest = '/' + rest;
        }
      }
    }

    // now rest is set to the post-host stuff.
    // chop off any delim chars.
    if (!unsafeProtocol[lowerProto]) {

      // First, make 100% sure that any "autoEscape" chars get
      // escaped, even if encodeURIComponent doesn't think they
      // need to be.
      for (var i = 0, l = autoEscape.length; i < l; i++) {
        var ae = autoEscape[i];
        var esc = encodeURIComponent(ae);
        if (esc === ae) {
          esc = escape(ae);
        }
        rest = rest.split(ae).join(esc);
      }
    }


    // chop off from the tail first.
    var hash = rest.indexOf('#');
    if (hash !== -1) {
      // got a fragment string.
      this.hash = rest.substr(hash);
      rest = rest.slice(0, hash);
    }
    var qm = rest.indexOf('?');
    if (qm !== -1) {
      this.search = rest.substr(qm);
      this.query = rest.substr(qm + 1);
      if (parseQueryString) {
        this.query = querystring.parse(this.query);
      }
      rest = rest.slice(0, qm);
    } else if (parseQueryString) {
      // no query string, but parseQueryString still requested
      this.search = '';
      this.query = {};
    }
    if (rest) this.pathname = rest;
    if (slashedProtocol[proto] &&
        this.hostname && !this.pathname) {
      this.pathname = '/';
    }

    //to support http.request
    if (this.pathname || this.search) {
      var p = this.pathname || '';
      var s = this.search || '';
      this.path = p + s;
    }

    // finally, reconstruct the href based on what has been validated.
    this.href = this.format();
    return this;
  };

  // format a parsed object into a url string
  function urlFormat(obj) {
    // ensure it's an object, and not a string url.
    // If it's an obj, this is a no-op.
    // this way, you can call url_format() on strings
    // to clean up potentially wonky urls.
    if (typeof(obj) === 'string') obj = urlParse(obj);
    if (!(obj instanceof Url)) return Url.prototype.format.call(obj);
    return obj.format();
  }

  Url.prototype.format = function() {
    var auth = this.auth || '';
    if (auth) {
      auth = encodeURIComponent(auth);
      auth = auth.replace(/%3A/i, ':');
      auth += '@';
    }

    var protocol = this.protocol || '',
        pathname = this.pathname || '',
        hash = this.hash || '',
        host = false,
        query = '';

    if (this.host) {
      host = auth + this.host;
    } else if (this.hostname) {
      host = auth + (this.hostname.indexOf(':') === -1 ?
          this.hostname :
          '[' + this.hostname + ']');
      if (this.port) {
        host += ':' + this.port;
      }
    }

    if (this.query && typeof this.query === 'object' &&
        Object.keys(this.query).length) {
      query = querystring.stringify(this.query);
    }

    var search = this.search || (query && ('?' + query)) || '';

    if (protocol && protocol.substr(-1) !== ':') protocol += ':';

    // only the slashedProtocols get the //.  Not mailto:, xmpp:, etc.
    // unless they had them to begin with.
    if (this.slashes ||
        (!protocol || slashedProtocol[protocol]) && host !== false) {
      host = '//' + (host || '');
      if (pathname && pathname.charAt(0) !== '/') pathname = '/' + pathname;
    } else if (!host) {
      host = '';
    }

    if (hash && hash.charAt(0) !== '#') hash = '#' + hash;
    if (search && search.charAt(0) !== '?') search = '?' + search;

    pathname = pathname.replace(/[?#]/g, function(match) {
      return encodeURIComponent(match);
    });
    search = search.replace('#', '%23');

    return protocol + host + pathname + search + hash;
  };

  function urlResolve(source, relative) {
    return urlParse(source, false, true).resolve(relative);
  }

  Url.prototype.resolve = function(relative) {
    return this.resolveObject(urlParse(relative, false, true)).format();
  };

  function urlResolveObject(source, relative) {
    if (!source) return relative;
    return urlParse(source, false, true).resolveObject(relative);
  }

  Url.prototype.resolveObject = function(relative) {
    if (typeof relative === 'string') {
      var rel = new Url();
      rel.parse(relative, false, true);
      relative = rel;
    }

    var result = new Url();
    Object.keys(this).forEach(function(k) {
      result[k] = this[k];
    }, this);

    // hash is always overridden, no matter what.
    // even href="" will remove it.
    result.hash = relative.hash;

    // if the relative url is empty, then there's nothing left to do here.
    if (relative.href === '') {
      result.href = result.format();
      return result;
    }

    // hrefs like //foo/bar always cut to the protocol.
    if (relative.slashes && !relative.protocol) {
      // take everything except the protocol from relative
      Object.keys(relative).forEach(function(k) {
        if (k !== 'protocol')
          result[k] = relative[k];
      });

      //urlParse appends trailing / to urls like http://www.example.com
      if (slashedProtocol[result.protocol] &&
          result.hostname && !result.pathname) {
        result.path = result.pathname = '/';
      }

      result.href = result.format();
      return result;
    }

    if (relative.protocol && relative.protocol !== result.protocol) {
      // if it's a known url protocol, then changing
      // the protocol does weird things
      // first, if it's not file:, then we MUST have a host,
      // and if there was a path
      // to begin with, then we MUST have a path.
      // if it is file:, then the host is dropped,
      // because that's known to be hostless.
      // anything else is assumed to be absolute.
      if (!slashedProtocol[relative.protocol]) {
        Object.keys(relative).forEach(function(k) {
          result[k] = relative[k];
        });
        result.href = result.format();
        return result;
      }

      result.protocol = relative.protocol;
      if (!relative.host && !hostlessProtocol[relative.protocol]) {
        var relPath = (relative.pathname || '').split('/');
        while (relPath.length && !(relative.host = relPath.shift()));
        if (!relative.host) relative.host = '';
        if (!relative.hostname) relative.hostname = '';
        if (relPath[0] !== '') relPath.unshift('');
        if (relPath.length < 2) relPath.unshift('');
        result.pathname = relPath.join('/');
      } else {
        result.pathname = relative.pathname;
      }
      result.search = relative.search;
      result.query = relative.query;
      result.host = relative.host || '';
      result.auth = relative.auth;
      result.hostname = relative.hostname || relative.host;
      result.port = relative.port;
      // to support http.request
      if (result.pathname || result.search) {
        var p = result.pathname || '';
        var s = result.search || '';
        result.path = p + s;
      }
      result.slashes = result.slashes || relative.slashes;
      result.href = result.format();
      return result;
    }

    var isSourceAbs = (result.pathname && result.pathname.charAt(0) === '/'),
        isRelAbs = (
            relative.host ||
            relative.pathname && relative.pathname.charAt(0) === '/'
        ),
        mustEndAbs = (isRelAbs || isSourceAbs ||
                      (result.host && relative.pathname)),
        removeAllDots = mustEndAbs,
        srcPath = result.pathname && result.pathname.split('/') || [],
        relPath = relative.pathname && relative.pathname.split('/') || [],
        psychotic = result.protocol && !slashedProtocol[result.protocol];

    // if the url is a non-slashed url, then relative
    // links like ../.. should be able
    // to crawl up to the hostname, as well.  This is strange.
    // result.protocol has already been set by now.
    // Later on, put the first path part into the host field.
    if (psychotic) {
      result.hostname = '';
      result.port = null;
      if (result.host) {
        if (srcPath[0] === '') srcPath[0] = result.host;
        else srcPath.unshift(result.host);
      }
      result.host = '';
      if (relative.protocol) {
        relative.hostname = null;
        relative.port = null;
        if (relative.host) {
          if (relPath[0] === '') relPath[0] = relative.host;
          else relPath.unshift(relative.host);
        }
        relative.host = null;
      }
      mustEndAbs = mustEndAbs && (relPath[0] === '' || srcPath[0] === '');
    }

    if (isRelAbs) {
      // it's absolute.
      result.host = (relative.host || relative.host === '') ?
                    relative.host : result.host;
      result.hostname = (relative.hostname || relative.hostname === '') ?
                        relative.hostname : result.hostname;
      result.search = relative.search;
      result.query = relative.query;
      srcPath = relPath;
      // fall through to the dot-handling below.
    } else if (relPath.length) {
      // it's relative
      // throw away the existing file, and take the new path instead.
      if (!srcPath) srcPath = [];
      srcPath.pop();
      srcPath = srcPath.concat(relPath);
      result.search = relative.search;
      result.query = relative.query;
    } else if (relative.search !== null && relative.search !== undefined) {
      // just pull out the search.
      // like href='?foo'.
      // Put this after the other two cases because it simplifies the booleans
      if (psychotic) {
        result.hostname = result.host = srcPath.shift();
        //occationaly the auth can get stuck only in host
        //this especialy happens in cases like
        //url.resolveObject('mailto:local1@domain1', 'local2@domain2')
        var authInHost = result.host && result.host.indexOf('@') > 0 ?
                         result.host.split('@') : false;
        if (authInHost) {
          result.auth = authInHost.shift();
          result.host = result.hostname = authInHost.shift();
        }
      }
      result.search = relative.search;
      result.query = relative.query;
      //to support http.request
      if (result.pathname !== null || result.search !== null) {
        result.path = (result.pathname ? result.pathname : '') +
                      (result.search ? result.search : '');
      }
      result.href = result.format();
      return result;
    }

    if (!srcPath.length) {
      // no path at all.  easy.
      // we've already handled the other stuff above.
      result.pathname = null;
      //to support http.request
      if (result.search) {
        result.path = '/' + result.search;
      } else {
        result.path = null;
      }
      result.href = result.format();
      return result;
    }

    // if a url ENDs in . or .., then it must get a trailing slash.
    // however, if it ends in anything else non-slashy,
    // then it must NOT get a trailing slash.
    var last = srcPath.slice(-1)[0];
    var hasTrailingSlash = (
        (result.host || relative.host) && (last === '.' || last === '..') ||
        last === '');

    // strip single dots, resolve double dots to parent dir
    // if the path tries to go above the root, `up` ends up > 0
    var up = 0;
    for (var i = srcPath.length; i >= 0; i--) {
      last = srcPath[i];
      if (last == '.') {
        srcPath.splice(i, 1);
      } else if (last === '..') {
        srcPath.splice(i, 1);
        up++;
      } else if (up) {
        srcPath.splice(i, 1);
        up--;
      }
    }

    // if the path is allowed to go above the root, restore leading ..s
    if (!mustEndAbs && !removeAllDots) {
      for (; up--; up) {
        srcPath.unshift('..');
      }
    }

    if (mustEndAbs && srcPath[0] !== '' &&
        (!srcPath[0] || srcPath[0].charAt(0) !== '/')) {
      srcPath.unshift('');
    }

    if (hasTrailingSlash && (srcPath.join('/').substr(-1) !== '/')) {
      srcPath.push('');
    }

    var isAbsolute = srcPath[0] === '' ||
        (srcPath[0] && srcPath[0].charAt(0) === '/');

    // put the host back
    if (psychotic) {
      result.hostname = result.host = isAbsolute ? '' :
                                      srcPath.length ? srcPath.shift() : '';
      //occationaly the auth can get stuck only in host
      //this especialy happens in cases like
      //url.resolveObject('mailto:local1@domain1', 'local2@domain2')
      var authInHost = result.host && result.host.indexOf('@') > 0 ?
                       result.host.split('@') : false;
      if (authInHost) {
        result.auth = authInHost.shift();
        result.host = result.hostname = authInHost.shift();
      }
    }

    mustEndAbs = mustEndAbs || (result.host && srcPath.length);

    if (mustEndAbs && !isAbsolute) {
      srcPath.unshift('');
    }

    if (!srcPath.length) {
      result.pathname = null;
      result.path = null;
    } else {
      result.pathname = srcPath.join('/');
    }

    if (result.pathname !== null || result.search !== null) {
      result.path = (result.pathname ? result.pathname : '') +
                    (result.search ? result.search : '');
    }
    result.auth = relative.auth || result.auth;
    result.slashes = result.slashes || relative.slashes;
    result.href = result.format();
    return result;
  };

  Url.prototype.parseHost = function() {
    var host = this.host;
    var port = portPattern.exec(host);
    if (port) {
      port = port[0];
      if (port !== ':') {
        this.port = port.substr(1);
      }
      host = host.substr(0, host.length - port.length);
    }
    if (host) this.hostname = host;
  };

  /** Expose `url` */
  if (freeExports) {
    if (freeModule && freeModule.exports == freeExports) {
      freeModule.exports = url;
    } else {
      for (key in url) {
        url.hasOwnProperty(key) && (freeExports[key] = url[key]);
      }
    }
  } else if (freeDefine) {
    define('url', url);
  } else {
    root.url = url;
  }

}(this);