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         $body = $doc->getElementsByTagName('html')->item(0)-> // <html>
  76                       getElementsByTagName('body')->item(0);  // <body>
  77
  78         $div = $body->getElementsByTagName('div')->item(0); // <div>
  79         $tokens = array();
  80         $this->tokenizeDOM($div, $tokens);
  81         // If the div has a sibling, that means we tripped across
  82         // a premature </div> tag.  So remove the div we parsed,
  83         // and then tokenize the rest of body.  We can't tokenize
  84         // the sibling directly as we'll lose the tags in that case.
  85         if ($div->nextSibling) {
  86             $body->removeChild($div);
  87             $this->tokenizeDOM($body, $tokens);
  88         }
  89         return $tokens;
  90     }
  91
  92     /**
  93      * Iterative function that tokenizes a node, putting it into an accumulator.
  94      * To iterate is human, to recurse divine - L. Peter Deutsch
  95      * @param DOMNode $node DOMNode to be tokenized.
  96      * @param HTMLPurifier_Token[] $tokens   Array-list of already tokenized tokens.
  97      * @return HTMLPurifier_Token of node appended to previously passed tokens.
  98      */
  99     protected function tokenizeDOM($node, &$tokens)
 100     {
 101         $level = 0;
 102         $nodes = array($level => new HTMLPurifier_Queue(array($node)));
 103         $closingNodes = array();
 104         do {
 105             while (!$nodes[$level]->isEmpty()) {
 106                 $node = $nodes[$level]->shift(); // FIFO
 107                 $collect = $level > 0 ? true : false;
 108                 $needEndingTag = $this->createStartNode($node, $tokens, $collect);
 109                 if ($needEndingTag) {
 110                     $closingNodes[$level][] = $node;
 111                 }
 112                 if ($node->childNodes && $node->childNodes->length) {
 113                     $level++;
 114                     $nodes[$level] = new HTMLPurifier_Queue();
 115                     foreach ($node->childNodes as $childNode) {
 116                         $nodes[$level]->push($childNode);
 117                     }
 118                 }
 119             }
 120             $level--;
 121             if ($level && isset($closingNodes[$level])) {
 122                 while ($node = array_pop($closingNodes[$level])) {
 123                     $this->createEndNode($node, $tokens);
 124                 }
 125             }
 126         } while ($level > 0);
 127     }
 128
 129     /**
 130      * @param DOMNode $node DOMNode to be tokenized.
 131      * @param HTMLPurifier_Token[] $tokens   Array-list of already tokenized tokens.
 132      * @param bool $collect  Says whether or start and close are collected, set to
 133      *                    false at first recursion because it's the implicit DIV
 134      *                    tag you're dealing with.
 135      * @return bool if the token needs an endtoken
 136      * @todo data and tagName properties don't seem to exist in DOMNode?
 137      */
 138     protected function createStartNode($node, &$tokens, $collect)
 139     {
 140         // intercept non element nodes. WE MUST catch all of them,
 141         // but we're not getting the character reference nodes because
 142         // those should have been preprocessed
 143         if ($node->nodeType === XML_TEXT_NODE) {
 144             $tokens[] = $this->factory->createText($node->data);
 145             return false;
 146         } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
 147             // undo libxml's special treatment of <script> and <style> tags
 148             $last = end($tokens);
 149             $data = $node->data;
 150             // (note $node->tagname is already normalized)
 151             if ($last instanceof HTMLPurifier_Token_Start && ($last->name == 'script' || $last->name == 'style')) {
 152                 $new_data = trim($data);
 153                 if (substr($new_data, 0, 4) === '<!--') {
 154                     $data = substr($new_data, 4);
 155                     if (substr($data, -3) === '-->') {
 156                         $data = substr($data, 0, -3);
 157                     } else {
 158                         // Highly suspicious! Not sure what to do...
 159                     }
 160                 }
 161             }
 162             $tokens[] = $this->factory->createText($this->parseData($data));
 163             return false;
 164         } elseif ($node->nodeType === XML_COMMENT_NODE) {
 165             // this is code is only invoked for comments in script/style in versions
 166             // of libxml pre-2.6.28 (regular comments, of course, are still
 167             // handled regularly)
 168             $tokens[] = $this->factory->createComment($node->data);
 169             return false;
 170         } elseif ($node->nodeType !== XML_ELEMENT_NODE) {
 171             // not-well tested: there may be other nodes we have to grab
 172             return false;
 173         }
 174
 175         $attr = $node->hasAttributes() ? $this->transformAttrToAssoc($node->attributes) : array();
 176
 177         // We still have to make sure that the element actually IS empty
 178         if (!$node->childNodes->length) {
 179             if ($collect) {
 180                 $tokens[] = $this->factory->createEmpty($node->tagName, $attr);
 181             }
 182             return false;
 183         } else {
 184             if ($collect) {
 185                 $tokens[] = $this->factory->createStart(
 186                     $tag_name = $node->tagName, // somehow, it get's dropped
 187                     $attr
 188                 );
 189             }
 190             return true;
 191         }
 192     }
 193
 194     /**
 195      * @param DOMNode $node
 196      * @param HTMLPurifier_Token[] $tokens
 197      */
 198     protected function createEndNode($node, &$tokens)
 199     {
 200         $tokens[] = $this->factory->createEnd($node->tagName);
 201     }
 202
 203
 204     /**
 205      * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
 206      *
 207      * @param DOMNamedNodeMap $node_map DOMNamedNodeMap of DOMAttr objects.
 208      * @return array Associative array of attributes.
 209      */
 210     protected function transformAttrToAssoc($node_map)
 211     {
 212         // NamedNodeMap is documented very well, so we're using undocumented
 213         // features, namely, the fact that it implements Iterator and
 214         // has a ->length attribute
 215         if ($node_map->length === 0) {
 216             return array();
 217         }
 218         $array = array();
 219         foreach ($node_map as $attr) {
 220             $array[$attr->name] = $attr->value;
 221         }
 222         return $array;
 223     }
 224
 225     /**
 226      * An error handler that mutes all errors
 227      * @param int $errno
 228      * @param string $errstr
 229      */
 230     public function muteErrorHandler($errno, $errstr)
 231     {
 232     }
 233
 234     /**
 235      * Callback function for undoing escaping of stray angled brackets
 236      * in comments
 237      * @param array $matches
 238      * @return string
 239      */
 240     public function callbackUndoCommentSubst($matches)
 241     {
 242         return '<!--' . strtr($matches[1], array('&amp;' => '&', '&lt;' => '<')) . $matches[2];
 243     }
 244
 245     /**
 246      * Callback function that entity-izes ampersands in comments so that
 247      * callbackUndoCommentSubst doesn't clobber them
 248      * @param array $matches
 249      * @return string
 250      */
 251     public function callbackArmorCommentEntities($matches)
 252     {
 253         return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
 254     }
 255
 256     /**
 257      * Wraps an HTML fragment in the necessary HTML
 258      * @param string $html
 259      * @param HTMLPurifier_Config $config
 260      * @param HTMLPurifier_Context $context
 261      * @return string
 262      */
 263     protected function wrapHTML($html, $config, $context, $use_div = true)
 264     {
 265         $def = $config->getDefinition('HTML');
 266         $ret = '';
 267
 268         if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
 269             $ret .= '<!DOCTYPE html ';
 270             if (!empty($def->doctype->dtdPublic)) {
 271                 $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
 272             }
 273             if (!empty($def->doctype->dtdSystem)) {
 274                 $ret .= '"' . $def->doctype->dtdSystem . '" ';
 275             }
 276             $ret .= '>';
 277         }
 278
 279         $ret .= '<html><head>';
 280         $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
 281         // No protection if $html contains a stray </div>!
 282         $ret .= '</head><body>';
 283         if ($use_div) $ret .= '<div>';
 284         $ret .= $html;
 285         if ($use_div) $ret .= '</div>';
 286         $ret .= '</body></html>';
 287         return $ret;
 288     }
 289 }
 290
 291 // vim: et sw=4 sts=4