node/lib/url.js

// Copyright Joyent, Inc. and other Node contributors.
//
// Permission is hereby granted, free of charge, to any person obtaining a
// copy of this software and associated documentation files (the
// "Software"), to deal in the Software without restriction, including
// without limitation the rights to use, copy, modify, merge, publish,
// distribute, sublicense, and/or sell copies of the Software, and to permit
// persons to whom the Software is furnished to do so, subject to the
// following conditions:
//
// The above copyright notice and this permission notice shall be included
// in all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN
// NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
// OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
// USE OR OTHER DEALINGS IN THE SOFTWARE.

exports.parse = urlParse;
exports.resolve = urlResolve;
exports.resolveObject = urlResolveObject;
exports.format = urlFormat;

// 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]+$/,
    // RFC 2396: characters reserved for delimiting URLs.
    delims = ['<', '>', '"', '`', ' ', '\r', '\n', '\t'],
    // RFC 2396: characters not allowed for various reasons.
    unwise = ['{', '}', '|', '\\', '^', '~', '[', ']', '`'].concat(delims),
    // Allowed by RFCs, but cause of XSS attacks.  Always escape these.
    autoEscape = ['\''],
    // Characters that are never ever allowed in a hostname.
    // Note that any invalid chars are also handled, but these
    // are the ones that are *expected* to be seen, so we fast-path
    // them.
    nonHostChars = ['%', '/', '?', ';', '#']
      .concat(unwise).concat(autoEscape),
    nonAuthChars = ['/', '@', '?', '#'].concat(delims),
    hostnameMaxLen = 255,
    hostnamePartPattern = /^[a-zA-Z0-9][a-z0-9A-Z-]{0,62}$/,
    hostnamePartStart = /^([a-zA-Z0-9][a-z0-9A-Z-]{0,62})(.*)$/,
    // 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.href) return url;

  var out = {},
      rest = url;

  // cut off any delimiters.
  // This is to support parse stuff like "<http://foo.com>"
  for (var i = 0, l = rest.length; i < l; i++) {
    if (delims.indexOf(rest.charAt(i)) === -1) break;
  }
  if (i !== 0) rest = rest.substr(i);


  var proto = protocolPattern.exec(rest);
  if (proto) {
    proto = proto[0];
    var lowerProto = proto.toLowerCase();
    out.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);
      out.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) {
      // there *may be* an auth
      var hasAuth = true;
      for (var i = 0, l = nonAuthChars.length; i < l; i++) {
        var index = rest.indexOf(nonAuthChars[i]);
        if (index !== -1 && index < atSign) {
          // not a valid auth.  Something like http://foo.com/bar@baz/
          hasAuth = false;
          break;
        }
      }
      if (hasAuth) {
        // pluck off the auth portion.
        out.auth = rest.substr(0, atSign);
        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) {
      out.host = rest.substr(0, firstNonHost);
      rest = rest.substr(firstNonHost);
    } else {
      out.host = rest;
      rest = '';
    }

    // pull out port.
    var p = parseHost(out.host);
    if (out.auth) out.host = out.auth + '@' + out.host;
    var keys = Object.keys(p);
    for (var i = 0, l = keys.length; i < l; i++) {
      var key = keys[i];
      out[key] = p[key];
    }

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

    // validate a little.
    if (out.hostname.length > hostnameMaxLen) {
      out.hostname = '';
    } else {
      var hostparts = out.hostname.split(/\./);
      for (var i = 0, l = hostparts.length; i < l; i++) {
        var part = hostparts[i];
        if (!part) continue;
        if (!part.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
          }
          out.hostname = validParts.join('.');
          break;
        }
      }
    }
    // hostnames are always lower case.
    out.hostname = out.hostname.toLowerCase();

    out.host = ((out.auth) ? out.auth + '@' : '') +
        (out.hostname || '') +
        ((out.port) ? ':' + out.port : '');
    out.href += out.host;
  }

  // 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);
    }

    // Now make sure that delims never appear in a url.
    var chop = rest.length;
    for (var i = 0, l = delims.length; i < l; i++) {
      var c = rest.indexOf(delims[i]);
      if (c !== -1) {
        chop = Math.min(c, chop);
      }
    }
    rest = rest.substr(0, chop);
  }


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

  // finally, reconstruct the href based on what has been validated.
  out.href = urlFormat(out);

  return out;
}

// 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);

  var auth = obj.auth;
  if (auth) {
    auth = auth.split('@').join('%40');
    for (var i = 0, l = nonAuthChars.length; i < l; i++) {
      var nAC = nonAuthChars[i];
      auth = auth.split(nAC).join(encodeURIComponent(nAC));
    }
  }

  var protocol = obj.protocol || '',
      host = (obj.host !== undefined) ? obj.host :
          obj.hostname !== undefined ? (
              (auth ? auth + '@' : '') +
              obj.hostname +
              (obj.port ? ':' + obj.port : '')
          ) :
          false,
      pathname = obj.pathname || '',
      query = obj.query &&
              ((typeof obj.query === 'object' &&
                Object.keys(obj.query).length) ?
                 querystring.stringify(obj.query) :
                 '') || '',
      search = obj.search || (query && ('?' + query)) || '',
      hash = obj.hash || '';

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

  // only the slashedProtocols get the //.  Not mailto:, xmpp:, etc.
  // unless they had them to begin with.
  if (obj.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;

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

function urlResolve(source, relative) {
  return urlFormat(urlResolveObject(source, relative));
}

function urlResolveObject(source, relative) {
  if (!source) return relative;

  source = urlParse(urlFormat(source), false, true);
  relative = urlParse(urlFormat(relative), false, true);

  // hash is always overridden, no matter what.
  source.hash = relative.hash;

  if (relative.href === '') return source;

  // hrefs like //foo/bar always cut to the protocol.
  if (relative.slashes && !relative.protocol) {
    relative.protocol = source.protocol;
    return relative;
  }

  if (relative.protocol && relative.protocol !== source.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]) return relative;

    source.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 (relPath[0] !== '') relPath.unshift('');
      if (relPath.length < 2) relPath.unshift('');
      relative.pathname = relPath.join('/');
    }
    source.pathname = relative.pathname;
    source.search = relative.search;
    source.query = relative.query;
    source.host = relative.host || '';
    delete source.auth;
    delete source.hostname;
    source.port = relative.port;
    return source;
  }

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

  // 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.
  // source.protocol has already been set by now.
  // Later on, put the first path part into the host field.
  if (psychotic) {

    delete source.hostname;
    delete source.auth;
    delete source.port;
    if (source.host) {
      if (srcPath[0] === '') srcPath[0] = source.host;
      else srcPath.unshift(source.host);
    }
    delete source.host;

    if (relative.protocol) {
      delete relative.hostname;
      delete relative.auth;
      delete relative.port;
      if (relative.host) {
        if (relPath[0] === '') relPath[0] = relative.host;
        else relPath.unshift(relative.host);
      }
      delete relative.host;
    }
    mustEndAbs = mustEndAbs && (relPath[0] === '' || srcPath[0] === '');
  }

  if (isRelAbs) {
    // it's absolute.
    source.host = (relative.host || relative.host === '') ?
                      relative.host : source.host;
    source.search = relative.search;
    source.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);
    source.search = relative.search;
    source.query = relative.query;
  } else if ('search' in relative) {
    // just pull out the search.
    // like href='?foo'.
    // Put this after the other two cases because it simplifies the booleans
    if (psychotic) {
      source.host = srcPath.shift();
    }
    source.search = relative.search;
    source.query = relative.query;
    return source;
  }
  if (!srcPath.length) {
    // no path at all.  easy.
    // we've already handled the other stuff above.
    delete source.pathname;
    return source;
  }

  // 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 = (
      (source.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) {
    source.host = isAbsolute ? '' : srcPath.shift();
  }

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

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

  source.pathname = srcPath.join('/');


  return source;
}

function parseHost(host) {
  var out = {};
  var port = portPattern.exec(host);
  if (port) {
    port = port[0];
    out.port = port.substr(1);
    host = host.substr(0, host.length - port.length);
  }
  if (host) out.hostname = host;
  return out;
}