library/HTMLPurifier/Lexer/DOMLex.php

   1 <?php
   2
   3 /**
   4  * Parser that uses PHP 5's DOM extension (part of the core).
   5  *
   6  * In PHP 5, the DOM XML extension was revamped into DOM and added to the core.
   7  * It gives us a forgiving HTML parser, which we use to transform the HTML
   8  * into a DOM, and then into the tokens.  It is blazingly fast (for large
   9  * documents, it performs twenty times faster than
  10  * HTMLPurifier_Lexer_DirectLex,and is the default choice for PHP 5.
  11  *
  12  * @note Any empty elements will have empty tokens associated with them, even if
  13  * this is prohibited by the spec. This is cannot be fixed until the spec
  14  * comes into play.
  15  *
  16  * @note PHP's DOM extension does not actually parse any entities, we use
  17  *       our own function to do that.
  18  *
  19  * @warning DOM tends to drop whitespace, which may wreak havoc on indenting.
  20  *          If this is a huge problem, due to the fact that HTML is hand
  21  *          edited and you are unable to get a parser cache that caches the
  22  *          the output of HTML Purifier while keeping the original HTML lying
  23  *          around, you may want to run Tidy on the resulting output or use
  24  *          HTMLPurifier_DirectLex
  25  */
  26
  27 class HTMLPurifier_Lexer_DOMLex extends HTMLPurifier_Lexer
  28 {
  29
  30     /**
  31      * @type HTMLPurifier_TokenFactory
  32      */
  33     private $factory;
  34
  35     public function __construct()
  36     {
  37         // setup the factory
  38         parent::__construct();
  39         $this->factory = new HTMLPurifier_TokenFactory();
  40     }
  41
  42     /**
  43      * @param string $html
  44      * @param HTMLPurifier_Config $config
  45      * @param HTMLPurifier_Context $context
  46      * @return HTMLPurifier_Token[]
  47      */
  48     public function tokenizeHTML($html, $config, $context)
  49     {
  50         $html = $this->normalize($html, $config, $context);
  51
  52         // attempt to armor stray angled brackets that cannot possibly
  53         // form tags and thus are probably being used as emoticons
  54         if ($config->get('Core.AggressivelyFixLt')) {
  55             $char = '[^a-z!\/]';
  56             $comment = "/<!--(.*?)(-->|\z)/is";
  57             $html = preg_replace_callback($comment, array($this, 'callbackArmorCommentEntities'), $html);
  58             do {
  59                 $old = $html;
  60                 $html = preg_replace("/<($char)/i", '&lt;\\1', $html);
  61             } while ($html !== $old);
  62             $html = preg_replace_callback($comment, array($this, 'callbackUndoCommentSubst'), $html); // fix comments
  63         }
  64
  65         // preprocess html, essential for UTF-8
  66         $html = $this->wrapHTML($html, $config, $context);
  67
  68         $doc = new DOMDocument();
  69         $doc->encoding = 'UTF-8'; // theoretically, the above has this covered
  70
  71         set_error_handler(array($this, 'muteErrorHandler'));
  72         $doc->loadHTML($html);
  73         restore_error_handler();
  74
  75         $tokens = array();
  76         $this->tokenizeDOM(
  77             $doc->getElementsByTagName('html')->item(0)-> // <html>
  78             getElementsByTagName('body')->item(0)-> //   <body>
  79             getElementsByTagName('div')->item(0), //     <div>
  80             $tokens
  81         );
  82         return $tokens;
  83     }
  84
  85     /**
  86      * Iterative function that tokenizes a node, putting it into an accumulator.
  87      * To iterate is human, to recurse divine - L. Peter Deutsch
  88      * @param DOMNode $node DOMNode to be tokenized.
  89      * @param HTMLPurifier_Token[] $tokens   Array-list of already tokenized tokens.
  90      * @return HTMLPurifier_Token of node appended to previously passed tokens.
  91      */
  92     protected function tokenizeDOM($node, &$tokens)
  93     {
  94         $level = 0;
  95         $nodes = array($level => array($node));
  96         $closingNodes = array();
  97         do {
  98             while (!empty($nodes[$level])) {
  99                 $node = array_shift($nodes[$level]); // FIFO
 100                 $collect = $level > 0 ? true : false;
 101                 $needEndingTag = $this->createStartNode($node, $tokens, $collect);
 102                 if ($needEndingTag) {
 103                     $closingNodes[$level][] = $node;
 104                 }
 105                 if ($node->childNodes && $node->childNodes->length) {
 106                     $level++;
 107                     $nodes[$level] = array();
 108                     foreach ($node->childNodes as $childNode) {
 109                         array_push($nodes[$level], $childNode);
 110                     }
 111                 }
 112             }
 113             $level--;
 114             if ($level && isset($closingNodes[$level])) {
 115                 while ($node = array_pop($closingNodes[$level])) {
 116                     $this->createEndNode($node, $tokens);
 117                 }
 118             }
 119         } while ($level > 0);
 120     }
 121
 122     /**
 123      * @param DOMNode $node DOMNode to be tokenized.
 124      * @param HTMLPurifier_Token[] $tokens   Array-list of already tokenized tokens.
 125      * @param bool $collect  Says whether or start and close are collected, set to
 126      *                    false at first recursion because it's the implicit DIV
 127      *                    tag you're dealing with.
 128      * @return bool if the token needs an endtoken
 129      * @todo data and tagName properties don't seem to exist in DOMNode?
 130      */
 131     protected function createStartNode($node, &$tokens, $collect)
 132     {
 133         // intercept non element nodes. WE MUST catch all of them,
 134         // but we're not getting the character reference nodes because
 135         // those should have been preprocessed
 136         if ($node->nodeType === XML_TEXT_NODE) {
 137             $tokens[] = $this->factory->createText($node->data);
 138             return false;
 139         } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
 140             // undo libxml's special treatment of <script> and <style> tags
 141             $last = end($tokens);
 142             $data = $node->data;
 143             // (note $node->tagname is already normalized)
 144             if ($last instanceof HTMLPurifier_Token_Start && ($last->name == 'script' || $last->name == 'style')) {
 145                 $new_data = trim($data);
 146                 if (substr($new_data, 0, 4) === '<!--') {
 147                     $data = substr($new_data, 4);
 148                     if (substr($data, -3) === '-->') {
 149                         $data = substr($data, 0, -3);
 150                     } else {
 151                         // Highly suspicious! Not sure what to do...
 152                     }
 153                 }
 154             }
 155             $tokens[] = $this->factory->createText($this->parseData($data));
 156             return false;
 157         } elseif ($node->nodeType === XML_COMMENT_NODE) {
 158             // this is code is only invoked for comments in script/style in versions
 159             // of libxml pre-2.6.28 (regular comments, of course, are still
 160             // handled regularly)
 161             $tokens[] = $this->factory->createComment($node->data);
 162             return false;
 163         } elseif ($node->nodeType !== XML_ELEMENT_NODE) {
 164             // not-well tested: there may be other nodes we have to grab
 165             return false;
 166         }
 167
 168         $attr = $node->hasAttributes() ? $this->transformAttrToAssoc($node->attributes) : array();
 169
 170         // We still have to make sure that the element actually IS empty
 171         if (!$node->childNodes->length) {
 172             if ($collect) {
 173                 $tokens[] = $this->factory->createEmpty($node->tagName, $attr);
 174             }
 175             return false;
 176         } else {
 177             if ($collect) {
 178                 $tokens[] = $this->factory->createStart(
 179                     $tag_name = $node->tagName, // somehow, it get's dropped
 180                     $attr
 181                 );
 182             }
 183             return true;
 184         }
 185     }
 186
 187     /**
 188      * @param DOMNode $node
 189      * @param HTMLPurifier_Token[] $tokens
 190      */
 191     protected function createEndNode($node, &$tokens)
 192     {
 193         $tokens[] = $this->factory->createEnd($node->tagName);
 194     }
 195
 196
 197     /**
 198      * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
 199      *
 200      * @param DOMNamedNodeMap $node_map DOMNamedNodeMap of DOMAttr objects.
 201      * @return array Associative array of attributes.
 202      */
 203     protected function transformAttrToAssoc($node_map)
 204     {
 205         // NamedNodeMap is documented very well, so we're using undocumented
 206         // features, namely, the fact that it implements Iterator and
 207         // has a ->length attribute
 208         if ($node_map->length === 0) {
 209             return array();
 210         }
 211         $array = array();
 212         foreach ($node_map as $attr) {
 213             $array[$attr->name] = $attr->value;
 214         }
 215         return $array;
 216     }
 217
 218     /**
 219      * An error handler that mutes all errors
 220      * @param int $errno
 221      * @param string $errstr
 222      */
 223     public function muteErrorHandler($errno, $errstr)
 224     {
 225     }
 226
 227     /**
 228      * Callback function for undoing escaping of stray angled brackets
 229      * in comments
 230      * @param array $matches
 231      * @return string
 232      */
 233     public function callbackUndoCommentSubst($matches)
 234     {
 235         return '<!--' . strtr($matches[1], array('&amp;' => '&', '&lt;' => '<')) . $matches[2];
 236     }
 237
 238     /**
 239      * Callback function that entity-izes ampersands in comments so that
 240      * callbackUndoCommentSubst doesn't clobber them
 241      * @param array $matches
 242      * @return string
 243      */
 244     public function callbackArmorCommentEntities($matches)
 245     {
 246         return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
 247     }
 248
 249     /**
 250      * Wraps an HTML fragment in the necessary HTML
 251      * @param string $html
 252      * @param HTMLPurifier_Config $config
 253      * @param HTMLPurifier_Context $context
 254      * @return string
 255      */
 256     protected function wrapHTML($html, $config, $context)
 257     {
 258         $def = $config->getDefinition('HTML');
 259         $ret = '';
 260
 261         if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
 262             $ret .= '<!DOCTYPE html ';
 263             if (!empty($def->doctype->dtdPublic)) {
 264                 $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
 265             }
 266             if (!empty($def->doctype->dtdSystem)) {
 267                 $ret .= '"' . $def->doctype->dtdSystem . '" ';
 268             }
 269             $ret .= '>';
 270         }
 271
 272         $ret .= '<html><head>';
 273         $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
 274         // No protection if $html contains a stray </div>!
 275         $ret .= '</head><body><div>' . $html . '</div></body></html>';
 276         return $ret;
 277     }
 278 }
 279
 280 // vim: et sw=4 sts=4