library/HTMLPurifier/AttrDef/CSS/URI.php

   1 <?php
   2
   3 /**
   4  * Validates a URI in CSS syntax, which uses url('http://example.com')
   5  * @note While theoretically speaking a URI in a CSS document could
   6  *       be non-embedded, as of CSS2 there is no such usage so we're
   7  *       generalizing it. This may need to be changed in the future.
   8  * @warning Since HTMLPurifier_AttrDef_CSS blindly uses semicolons as
   9  *          the separator, you cannot put a literal semicolon in
  10  *          in the URI. Try percent encoding it, in that case.
  11  */
  12 class HTMLPurifier_AttrDef_CSS_URI extends HTMLPurifier_AttrDef_URI
  13 {
  14
  15     public function __construct() {
  16         parent::__construct(true); // always embedded
  17     }
  18
  19     public function validate($uri_string, $config, $context) {
  20         // parse the URI out of the string and then pass it onto
  21         // the parent object
  22
  23         $uri_string = $this->parseCDATA($uri_string);
  24         if (strpos($uri_string, 'url(') !== 0) return false;
  25         $uri_string = substr($uri_string, 4);
  26         $new_length = strlen($uri_string) - 1;
  27         if ($uri_string[$new_length] != ')') return false;
  28         $uri = trim(substr($uri_string, 0, $new_length));
  29
  30         if (!empty($uri) && ($uri[0] == "'" || $uri[0] == '"')) {
  31             $quote = $uri[0];
  32             $new_length = strlen($uri) - 1;
  33             if ($uri[$new_length] !== $quote) return false;
  34             $uri = substr($uri, 1, $new_length - 1);
  35         }
  36
  37         $uri = $this->expandCSSEscape($uri);
  38
  39         $result = parent::validate($uri, $config, $context);
  40
  41         if ($result === false) return false;
  42
  43         // extra sanity check; should have been done by URI
  44         $result = str_replace(array('"', "\\", "\n", "\x0c", "\r"), "", $result);
  45
  46         // suspicious characters are ()'; we're going to percent encode
  47         // them for safety.
  48         $result = str_replace(array('(', ')', "'"), array('%28', '%29', '%27'), $result);
  49
  50         // there's an extra bug where ampersands lose their escaping on
  51         // an innerHTML cycle, so a very unlucky query parameter could
  52         // then change the meaning of the URL.  Unfortunately, there's
  53         // not much we can do about that...
  54
  55         return "url(\"$result\")";
  56
  57     }
  58
  59 }
  60
  61 // vim: et sw=4 sts=4