angular.js/src/ng/urlUtils.js

'use strict';
// NOTE:  The usage of window and document instead of $window and $document here is
// deliberate.  This service depends on the specific behavior of anchor nodes created by the
// browser (resolving and parsing URLs) that is unlikely to be provided by mock objects and
// cause us to break tests.  In addition, when the browser resolves a URL for XHR, it
// doesn't know about mocked locations and resolves URLs to the real document - which is
// exactly the behavior needed here.  There is little value is mocking these out for this
// service.
var urlParsingNode = document.createElement("a");
/*
Matches paths for file protocol on windows,
such as /C:/foo/bar, and captures only /foo/bar.
*/
var windowsFilePathExp = /^\/?.*?:(\/.*)/;
var originUrl = urlResolve(window.location.href, true);


/**
 *
 * Implementation Notes for non-IE browsers
 * ----------------------------------------
 * Assigning a URL to the href property of an anchor DOM node, even one attached to the DOM,
 * results both in the normalizing and parsing of the URL.  Normalizing means that a relative
 * URL will be resolved into an absolute URL in the context of the application document.
 * Parsing means that the anchor node's host, hostname, protocol, port, pathname and related
 * properties are all populated to reflect the normalized URL.  This approach has wide
 * compatibility - Safari 1+, Mozilla 1+, Opera 7+,e etc.  See
 * http://www.aptana.com/reference/html/api/HTMLAnchorElement.html
 *
 * Implementation Notes for IE
 * ---------------------------
 * IE >= 8 and <= 10 normalizes the URL when assigned to the anchor node similar to the other
 * browsers.  However, the parsed components will not be set if the URL assigned did not specify
 * them.  (e.g. if you assign a.href = "foo", then a.protocol, a.host, etc. will be empty.)  We
 * work around that by performing the parsing in a 2nd step by taking a previously normalized
 * URL (e.g. by assigning to a.href) and assigning it a.href again.  This correctly populates the
 * properties such as protocol, hostname, port, etc.
 *
 * IE7 does not normalize the URL when assigned to an anchor node.  (Apparently, it does, if one
 * uses the inner HTML approach to assign the URL as part of an HTML snippet -
 * http://stackoverflow.com/a/472729)  However, setting img[src] does normalize the URL.
 * Unfortunately, setting img[src] to something like "javascript:foo" on IE throws an exception.
 * Since the primary usage for normalizing URLs is to sanitize such URLs, we can't use that
 * method and IE < 8 is unsupported.
 *
 * References:
 *   http://developer.mozilla.org/en-US/docs/Web/API/HTMLAnchorElement
 *   http://www.aptana.com/reference/html/api/HTMLAnchorElement.html
 *   http://url.spec.whatwg.org/#urlutils
 *   https://github.com/angular/angular.js/pull/2902
 *   http://james.padolsey.com/javascript/parsing-urls-with-the-dom/
 *
 * @function
 * @param {string} url The URL to be parsed.
 * @description Normalizes and parses a URL.
 * @returns {object} Returns the normalized URL as a dictionary.
 *
 *   | member name   | Description    |
 *   |---------------|----------------|
 *   | href          | A normalized version of the provided URL if it was not an absolute URL |
 *   | protocol      | The protocol including the trailing colon                              |
 *   | host          | The host and port (if the port is non-default) of the normalizedUrl    |
 *   | search        | The search params, minus the question mark                             |
 *   | hash          | The hash string, minus the hash symbol
 *   | hostname      | The hostname
 *   | port          | The port, without ":"
 *   | pathname      | The pathname, beginning with "/"
 *
 */
function urlResolve(url, base) {
  var href = url,
      pathname;

  if (msie) {
    // Normalize before parse.  Refer Implementation Notes on why this is
    // done in two steps on IE.
    urlParsingNode.setAttribute("href", href);
    href = urlParsingNode.href;
  }

  urlParsingNode.setAttribute('href', href);

  /*
   * In Windows, on an anchor node on documents loaded from
   * the filesystem, the browser will return a pathname
   * prefixed with the drive name ('/C:/path') when a
   * pathname without a drive is set:
   *  * a.setAttribute('href', '/foo')
   *   * a.pathname === '/C:/foo' //true
   *
   * Inside of Angular, we're always using pathnames that
   * do not include drive names for routing.
   */

  pathname = removeWindowsDriveName(urlParsingNode.pathname, url, base);
  pathname = (pathname.charAt(0) === '/') ? pathname : '/' + pathname;

  // urlParsingNode provides the UrlUtils interface - http://url.spec.whatwg.org/#urlutils
  return {
    href: urlParsingNode.href,
    protocol: urlParsingNode.protocol ? urlParsingNode.protocol.replace(/:$/, '') : '',
    host: urlParsingNode.host,
    search: urlParsingNode.search ? urlParsingNode.search.replace(/^\?/, '') : '',
    hash: urlParsingNode.hash ? urlParsingNode.hash.replace(/^#/, '') : '',
    hostname: urlParsingNode.hostname,
    port: urlParsingNode.port,
    pathname: pathname
  };
}


/**
 * Parse a request URL and determine whether this is a same-origin request as the application document.
 *
 * @param {string|object} requestUrl The url of the request as a string that will be resolved
 * or a parsed URL object.
 * @returns {boolean} Whether the request is for the same origin as the application document.
 */
function urlIsSameOrigin(requestUrl) {
  var parsed = (isString(requestUrl)) ? urlResolve(requestUrl) : requestUrl;
  return (parsed.protocol === originUrl.protocol &&
          parsed.host === originUrl.host);
}

function removeWindowsDriveName (path, url, base) {
  var firstPathSegmentMatch;

  //Get the relative path from the input URL.
  if (url.indexOf(base) === 0) {
    url = url.replace(base, '');
  }

  /*
   * The input URL intentionally contains a
   * first path segment that ends with a colon.
   */
  if (windowsFilePathExp.exec(url)) {
    return path;
  }

  firstPathSegmentMatch = windowsFilePathExp.exec(path);
  return firstPathSegmentMatch ? firstPathSegmentMatch[1] : path;
}