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     {
  17         parent::__construct(true); // always embedded
  18     }
  19
  20     /**
  21      * @param string $uri_string
  22      * @param HTMLPurifier_Config $config
  23      * @param HTMLPurifier_Context $context
  24      * @return bool|string
  25      */
  26     public function validate($uri_string, $config, $context)
  27     {
  28         // parse the URI out of the string and then pass it onto
  29         // the parent object
  30
  31         $uri_string = $this->parseCDATA($uri_string);
  32         if (strpos($uri_string, 'url(') !== 0) {
  33             return false;
  34         }
  35         $uri_string = substr($uri_string, 4);
  36         $new_length = strlen($uri_string) - 1;
  37         if ($uri_string[$new_length] != ')') {
  38             return false;
  39         }
  40         $uri = trim(substr($uri_string, 0, $new_length));
  41
  42         if (!empty($uri) && ($uri[0] == "'" || $uri[0] == '"')) {
  43             $quote = $uri[0];
  44             $new_length = strlen($uri) - 1;
  45             if ($uri[$new_length] !== $quote) {
  46                 return false;
  47             }
  48             $uri = substr($uri, 1, $new_length - 1);
  49         }
  50
  51         $uri = $this->expandCSSEscape($uri);
  52
  53         $result = parent::validate($uri, $config, $context);
  54
  55         if ($result === false) {
  56             return false;
  57         }
  58
  59         // extra sanity check; should have been done by URI
  60         $result = str_replace(array('"', "\\", "\n", "\x0c", "\r"), "", $result);
  61
  62         // suspicious characters are ()'; we're going to percent encode
  63         // them for safety.
  64         $result = str_replace(array('(', ')', "'"), array('%28', '%29', '%27'), $result);
  65
  66         // there's an extra bug where ampersands lose their escaping on
  67         // an innerHTML cycle, so a very unlucky query parameter could
  68         // then change the meaning of the URL.  Unfortunately, there's
  69         // not much we can do about that...
  70         return "url(\"$result\")";
  71     }
  72 }
  73
  74 // vim: et sw=4 sts=4