lib/htmlpurifier/HTMLPurifier/Lexer/DOMLex.php

   1 <?php
   2
   3 require_once 'HTMLPurifier/Lexer.php';
   4 require_once 'HTMLPurifier/TokenFactory.php';
   5
   6 /**
   7  * Parser that uses PHP 5's DOM extension (part of the core).
   8  *
   9  * In PHP 5, the DOM XML extension was revamped into DOM and added to the core.
  10  * It gives us a forgiving HTML parser, which we use to transform the HTML
  11  * into a DOM, and then into the tokens.  It is blazingly fast (for large
  12  * documents, it performs twenty times faster than
  13  * HTMLPurifier_Lexer_DirectLex,and is the default choice for PHP 5.
  14  *
  15  * @note Any empty elements will have empty tokens associated with them, even if
  16  * this is prohibited by the spec. This is cannot be fixed until the spec
  17  * comes into play.
  18  *
  19  * @note PHP's DOM extension does not actually parse any entities, we use
  20  *       our own function to do that.
  21  *
  22  * @warning DOM tends to drop whitespace, which may wreak havoc on indenting.
  23  *          If this is a huge problem, due to the fact that HTML is hand
  24  *          edited and you are unable to get a parser cache that caches the
  25  *          the output of HTML Purifier while keeping the original HTML lying
  26  *          around, you may want to run Tidy on the resulting output or use
  27  *          HTMLPurifier_DirectLex
  28  */
  29
  30 class HTMLPurifier_Lexer_DOMLex extends HTMLPurifier_Lexer
  31 {
  32
  33     private $factory;
  34
  35     public function __construct() {
  36         // setup the factory
  37         parent::HTMLPurifier_Lexer();
  38         $this->factory = new HTMLPurifier_TokenFactory();
  39     }
  40
  41     public function tokenizeHTML($html, $config, &$context) {
  42
  43         $html = $this->normalize($html, $config, $context);
  44
  45         // attempt to armor stray angled brackets that cannot possibly
  46         // form tags and thus are probably being used as emoticons
  47         if ($config->get('Core', 'AggressivelyFixLt')) {
  48             $char = '[^a-z!\/]';
  49             $comment = "/<!--(.*?)(-->|\z)/is";
  50             $html = preg_replace_callback($comment, array('HTMLPurifier_Lexer_DOMLex', 'callbackArmorCommentEntities'), $html);
  51             $html = preg_replace("/<($char)/i", '&lt;\\1', $html);
  52             $html = preg_replace_callback($comment, array('HTMLPurifier_Lexer_DOMLex', 'callbackUndoCommentSubst'), $html); // fix comments
  53         }
  54
  55         // preprocess html, essential for UTF-8
  56         $html = $this->wrapHTML($html, $config, $context);
  57
  58         $doc = new DOMDocument();
  59         $doc->encoding = 'UTF-8'; // theoretically, the above has this covered
  60
  61         set_error_handler(array($this, 'muteErrorHandler'));
  62         $doc->loadHTML($html);
  63         restore_error_handler();
  64
  65         $tokens = array();
  66         $this->tokenizeDOM(
  67             $doc->getElementsByTagName('html')->item(0)-> // <html>
  68                   getElementsByTagName('body')->item(0)-> //   <body>
  69                   getElementsByTagName('div')->item(0)    //     <div>
  70             , $tokens);
  71         return $tokens;
  72     }
  73
  74     /**
  75      * Recursive function that tokenizes a node, putting it into an accumulator.
  76      *
  77      * @param $node     DOMNode to be tokenized.
  78      * @param $tokens   Array-list of already tokenized tokens.
  79      * @param $collect  Says whether or start and close are collected, set to
  80      *                  false at first recursion because it's the implicit DIV
  81      *                  tag you're dealing with.
  82      * @returns Tokens of node appended to previously passed tokens.
  83      */
  84     protected function tokenizeDOM($node, &$tokens, $collect = false) {
  85
  86         // intercept non element nodes. WE MUST catch all of them,
  87         // but we're not getting the character reference nodes because
  88         // those should have been preprocessed
  89         if ($node->nodeType === XML_TEXT_NODE) {
  90             $tokens[] = $this->factory->createText($node->data);
  91             return;
  92         } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
  93             // undo libxml's special treatment of <script> and <style> tags
  94             $last = end($tokens);
  95             $data = $node->data;
  96             // (note $node->tagname is already normalized)
  97             if ($last instanceof HTMLPurifier_Token_Start && $last->name == 'script') {
  98                 $new_data = trim($data);
  99                 if (substr($new_data, 0, 4) === '<!--') {
 100                     $data = substr($new_data, 4);
 101                     if (substr($data, -3) === '-->') {
 102                         $data = substr($data, 0, -3);
 103                     } else {
 104                         // Highly suspicious! Not sure what to do...
 105                     }
 106                 }
 107             }
 108             $tokens[] = $this->factory->createText($this->parseData($data));
 109             return;
 110         } elseif ($node->nodeType === XML_COMMENT_NODE) {
 111             // this is code is only invoked for comments in script/style in versions
 112             // of libxml pre-2.6.28 (regular comments, of course, are still
 113             // handled regularly)
 114             $tokens[] = $this->factory->createComment($node->data);
 115             return;
 116         } elseif (
 117             // not-well tested: there may be other nodes we have to grab
 118             $node->nodeType !== XML_ELEMENT_NODE
 119         ) {
 120             return;
 121         }
 122
 123         $attr = $node->hasAttributes() ?
 124             $this->transformAttrToAssoc($node->attributes) :
 125             array();
 126
 127         // We still have to make sure that the element actually IS empty
 128         if (!$node->childNodes->length) {
 129             if ($collect) {
 130                 $tokens[] = $this->factory->createEmpty($node->tagName, $attr);
 131             }
 132         } else {
 133             if ($collect) { // don't wrap on first iteration
 134                 $tokens[] = $this->factory->createStart(
 135                     $tag_name = $node->tagName, // somehow, it get's dropped
 136                     $attr
 137                 );
 138             }
 139             foreach ($node->childNodes as $node) {
 140                 // remember, it's an accumulator. Otherwise, we'd have
 141                 // to use array_merge
 142                 $this->tokenizeDOM($node, $tokens, true);
 143             }
 144             if ($collect) {
 145                 $tokens[] = $this->factory->createEnd($tag_name);
 146             }
 147         }
 148
 149     }
 150
 151     /**
 152      * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
 153      *
 154      * @param $attribute_list DOMNamedNodeMap of DOMAttr objects.
 155      * @returns Associative array of attributes.
 156      */
 157     protected function transformAttrToAssoc($node_map) {
 158         // NamedNodeMap is documented very well, so we're using undocumented
 159         // features, namely, the fact that it implements Iterator and
 160         // has a ->length attribute
 161         if ($node_map->length === 0) return array();
 162         $array = array();
 163         foreach ($node_map as $attr) {
 164             $array[$attr->name] = $attr->value;
 165         }
 166         return $array;
 167     }
 168
 169     /**
 170      * An error handler that mutes all errors
 171      */
 172     public function muteErrorHandler($errno, $errstr) {}
 173
 174     /**
 175      * Callback function for undoing escaping of stray angled brackets
 176      * in comments
 177      */
 178     function callbackUndoCommentSubst($matches) {
 179         return '<!--' . strtr($matches[1], array('&amp;'=>'&','&lt;'=>'<')) . $matches[2];
 180     }
 181
 182     /**
 183      * Callback function that entity-izes ampersands in comments so that
 184      * callbackUndoCommentSubst doesn't clobber them
 185      */
 186     function callbackArmorCommentEntities($matches) {
 187         return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
 188     }
 189
 190     /**
 191      * Wraps an HTML fragment in the necessary HTML
 192      */
 193     function wrapHTML($html, $config, &$context) {
 194         $def = $config->getDefinition('HTML');
 195         $ret = '';
 196
 197         if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
 198             $ret .= '<!DOCTYPE html ';
 199             if (!empty($def->doctype->dtdPublic)) $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
 200             if (!empty($def->doctype->dtdSystem)) $ret .= '"' . $def->doctype->dtdSystem . '" ';
 201             $ret .= '>';
 202         }
 203
 204         $ret .= '<html><head>';
 205         $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
 206         $ret .= '</head><body><div>'.$html.'</div></body></html>';
 207         return $ret;
 208     }
 209
 210 }
 211