DOMLex.php 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339
  1. <?php
  2. /**
  3. * Parser that uses PHP 5's DOM extension (part of the core).
  4. *
  5. * In PHP 5, the DOM XML extension was revamped into DOM and added to the core.
  6. * It gives us a forgiving HTML parser, which we use to transform the HTML
  7. * into a DOM, and then into the tokens. It is blazingly fast (for large
  8. * documents, it performs twenty times faster than
  9. * HTMLPurifier_Lexer_DirectLex,and is the default choice for PHP 5.
  10. *
  11. * @note Any empty elements will have empty tokens associated with them, even if
  12. * this is prohibited by the spec. This is cannot be fixed until the spec
  13. * comes into play.
  14. *
  15. * @note PHP's DOM extension does not actually parse any entities, we use
  16. * our own function to do that.
  17. *
  18. * @warning DOM tends to drop whitespace, which may wreak havoc on indenting.
  19. * If this is a huge problem, due to the fact that HTML is hand
  20. * edited and you are unable to get a parser cache that caches the
  21. * the output of HTML Purifier while keeping the original HTML lying
  22. * around, you may want to run Tidy on the resulting output or use
  23. * HTMLPurifier_DirectLex
  24. */
  25. class HTMLPurifier_Lexer_DOMLex extends HTMLPurifier_Lexer
  26. {
  27. /**
  28. * @type HTMLPurifier_TokenFactory
  29. */
  30. private $factory;
  31. public function __construct()
  32. {
  33. // setup the factory
  34. parent::__construct();
  35. $this->factory = new HTMLPurifier_TokenFactory();
  36. }
  37. /**
  38. * @param string $html
  39. * @param HTMLPurifier_Config $config
  40. * @param HTMLPurifier_Context $context
  41. * @return HTMLPurifier_Token[]
  42. */
  43. public function tokenizeHTML($html, $config, $context)
  44. {
  45. $html = $this->normalize($html, $config, $context);
  46. // attempt to armor stray angled brackets that cannot possibly
  47. // form tags and thus are probably being used as emoticons
  48. if ($config->get('Core.AggressivelyFixLt')) {
  49. $char = '[^a-z!\/]';
  50. $comment = "/<!--(.*?)(-->|\z)/is";
  51. $html = preg_replace_callback($comment, array($this, 'callbackArmorCommentEntities'), $html);
  52. do {
  53. $old = $html;
  54. $html = preg_replace("/<($char)/i", '&lt;\\1', $html);
  55. } while ($html !== $old);
  56. $html = preg_replace_callback($comment, array($this, 'callbackUndoCommentSubst'), $html); // fix comments
  57. }
  58. // preprocess html, essential for UTF-8
  59. $html = $this->wrapHTML($html, $config, $context);
  60. $doc = new DOMDocument();
  61. $doc->encoding = 'UTF-8'; // theoretically, the above has this covered
  62. $options = 0;
  63. if ($config->get('Core.AllowParseManyTags') && defined('LIBXML_PARSEHUGE')) {
  64. $options |= LIBXML_PARSEHUGE;
  65. }
  66. set_error_handler(array($this, 'muteErrorHandler'));
  67. // loadHTML() fails on PHP 5.3 when second parameter is given
  68. if ($options) {
  69. $doc->loadHTML($html, $options);
  70. } else {
  71. $doc->loadHTML($html);
  72. }
  73. restore_error_handler();
  74. $body = $doc->getElementsByTagName('html')->item(0)-> // <html>
  75. getElementsByTagName('body')->item(0); // <body>
  76. $div = $body->getElementsByTagName('div')->item(0); // <div>
  77. $tokens = array();
  78. $this->tokenizeDOM($div, $tokens, $config);
  79. // If the div has a sibling, that means we tripped across
  80. // a premature </div> tag. So remove the div we parsed,
  81. // and then tokenize the rest of body. We can't tokenize
  82. // the sibling directly as we'll lose the tags in that case.
  83. if ($div->nextSibling) {
  84. $body->removeChild($div);
  85. $this->tokenizeDOM($body, $tokens, $config);
  86. }
  87. return $tokens;
  88. }
  89. /**
  90. * Iterative function that tokenizes a node, putting it into an accumulator.
  91. * To iterate is human, to recurse divine - L. Peter Deutsch
  92. * @param DOMNode $node DOMNode to be tokenized.
  93. * @param HTMLPurifier_Token[] $tokens Array-list of already tokenized tokens.
  94. * @return HTMLPurifier_Token of node appended to previously passed tokens.
  95. */
  96. protected function tokenizeDOM($node, &$tokens, $config)
  97. {
  98. $level = 0;
  99. $nodes = array($level => new HTMLPurifier_Queue(array($node)));
  100. $closingNodes = array();
  101. do {
  102. while (!$nodes[$level]->isEmpty()) {
  103. $node = $nodes[$level]->shift(); // FIFO
  104. $collect = $level > 0 ? true : false;
  105. $needEndingTag = $this->createStartNode($node, $tokens, $collect, $config);
  106. if ($needEndingTag) {
  107. $closingNodes[$level][] = $node;
  108. }
  109. if ($node->childNodes && $node->childNodes->length) {
  110. $level++;
  111. $nodes[$level] = new HTMLPurifier_Queue();
  112. foreach ($node->childNodes as $childNode) {
  113. $nodes[$level]->push($childNode);
  114. }
  115. }
  116. }
  117. $level--;
  118. if ($level && isset($closingNodes[$level])) {
  119. while ($node = array_pop($closingNodes[$level])) {
  120. $this->createEndNode($node, $tokens);
  121. }
  122. }
  123. } while ($level > 0);
  124. }
  125. /**
  126. * Portably retrieve the tag name of a node; deals with older versions
  127. * of libxml like 2.7.6
  128. * @param DOMNode $node
  129. */
  130. protected function getTagName($node)
  131. {
  132. if (isset($node->tagName)) {
  133. return $node->tagName;
  134. } else if (isset($node->nodeName)) {
  135. return $node->nodeName;
  136. } else if (isset($node->localName)) {
  137. return $node->localName;
  138. }
  139. return null;
  140. }
  141. /**
  142. * Portably retrieve the data of a node; deals with older versions
  143. * of libxml like 2.7.6
  144. * @param DOMNode $node
  145. */
  146. protected function getData($node)
  147. {
  148. if (isset($node->data)) {
  149. return $node->data;
  150. } else if (isset($node->nodeValue)) {
  151. return $node->nodeValue;
  152. } else if (isset($node->textContent)) {
  153. return $node->textContent;
  154. }
  155. return null;
  156. }
  157. /**
  158. * @param DOMNode $node DOMNode to be tokenized.
  159. * @param HTMLPurifier_Token[] $tokens Array-list of already tokenized tokens.
  160. * @param bool $collect Says whether or start and close are collected, set to
  161. * false at first recursion because it's the implicit DIV
  162. * tag you're dealing with.
  163. * @return bool if the token needs an endtoken
  164. * @todo data and tagName properties don't seem to exist in DOMNode?
  165. */
  166. protected function createStartNode($node, &$tokens, $collect, $config)
  167. {
  168. // intercept non element nodes. WE MUST catch all of them,
  169. // but we're not getting the character reference nodes because
  170. // those should have been preprocessed
  171. if ($node->nodeType === XML_TEXT_NODE) {
  172. $data = $this->getData($node); // Handle variable data property
  173. if ($data !== null) {
  174. $tokens[] = $this->factory->createText($data);
  175. }
  176. return false;
  177. } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
  178. // undo libxml's special treatment of <script> and <style> tags
  179. $last = end($tokens);
  180. $data = $node->data;
  181. // (note $node->tagname is already normalized)
  182. if ($last instanceof HTMLPurifier_Token_Start && ($last->name == 'script' || $last->name == 'style')) {
  183. $new_data = trim($data);
  184. if (substr($new_data, 0, 4) === '<!--') {
  185. $data = substr($new_data, 4);
  186. if (substr($data, -3) === '-->') {
  187. $data = substr($data, 0, -3);
  188. } else {
  189. // Highly suspicious! Not sure what to do...
  190. }
  191. }
  192. }
  193. $tokens[] = $this->factory->createText($this->parseText($data, $config));
  194. return false;
  195. } elseif ($node->nodeType === XML_COMMENT_NODE) {
  196. // this is code is only invoked for comments in script/style in versions
  197. // of libxml pre-2.6.28 (regular comments, of course, are still
  198. // handled regularly)
  199. $tokens[] = $this->factory->createComment($node->data);
  200. return false;
  201. } elseif ($node->nodeType !== XML_ELEMENT_NODE) {
  202. // not-well tested: there may be other nodes we have to grab
  203. return false;
  204. }
  205. $attr = $node->hasAttributes() ? $this->transformAttrToAssoc($node->attributes) : array();
  206. $tag_name = $this->getTagName($node); // Handle variable tagName property
  207. if (empty($tag_name)) {
  208. return (bool) $node->childNodes->length;
  209. }
  210. // We still have to make sure that the element actually IS empty
  211. if (!$node->childNodes->length) {
  212. if ($collect) {
  213. $tokens[] = $this->factory->createEmpty($tag_name, $attr);
  214. }
  215. return false;
  216. } else {
  217. if ($collect) {
  218. $tokens[] = $this->factory->createStart($tag_name, $attr);
  219. }
  220. return true;
  221. }
  222. }
  223. /**
  224. * @param DOMNode $node
  225. * @param HTMLPurifier_Token[] $tokens
  226. */
  227. protected function createEndNode($node, &$tokens)
  228. {
  229. $tag_name = $this->getTagName($node); // Handle variable tagName property
  230. $tokens[] = $this->factory->createEnd($tag_name);
  231. }
  232. /**
  233. * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
  234. *
  235. * @param DOMNamedNodeMap $node_map DOMNamedNodeMap of DOMAttr objects.
  236. * @return array Associative array of attributes.
  237. */
  238. protected function transformAttrToAssoc($node_map)
  239. {
  240. // NamedNodeMap is documented very well, so we're using undocumented
  241. // features, namely, the fact that it implements Iterator and
  242. // has a ->length attribute
  243. if ($node_map->length === 0) {
  244. return array();
  245. }
  246. $array = array();
  247. foreach ($node_map as $attr) {
  248. $array[$attr->name] = $attr->value;
  249. }
  250. return $array;
  251. }
  252. /**
  253. * An error handler that mutes all errors
  254. * @param int $errno
  255. * @param string $errstr
  256. */
  257. public function muteErrorHandler($errno, $errstr)
  258. {
  259. }
  260. /**
  261. * Callback function for undoing escaping of stray angled brackets
  262. * in comments
  263. * @param array $matches
  264. * @return string
  265. */
  266. public function callbackUndoCommentSubst($matches)
  267. {
  268. return '<!--' . strtr($matches[1], array('&amp;' => '&', '&lt;' => '<')) . $matches[2];
  269. }
  270. /**
  271. * Callback function that entity-izes ampersands in comments so that
  272. * callbackUndoCommentSubst doesn't clobber them
  273. * @param array $matches
  274. * @return string
  275. */
  276. public function callbackArmorCommentEntities($matches)
  277. {
  278. return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
  279. }
  280. /**
  281. * Wraps an HTML fragment in the necessary HTML
  282. * @param string $html
  283. * @param HTMLPurifier_Config $config
  284. * @param HTMLPurifier_Context $context
  285. * @return string
  286. */
  287. protected function wrapHTML($html, $config, $context, $use_div = true)
  288. {
  289. $def = $config->getDefinition('HTML');
  290. $ret = '';
  291. if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
  292. $ret .= '<!DOCTYPE html ';
  293. if (!empty($def->doctype->dtdPublic)) {
  294. $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
  295. }
  296. if (!empty($def->doctype->dtdSystem)) {
  297. $ret .= '"' . $def->doctype->dtdSystem . '" ';
  298. }
  299. $ret .= '>';
  300. }
  301. $ret .= '<html><head>';
  302. $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
  303. // No protection if $html contains a stray </div>!
  304. $ret .= '</head><body>';
  305. if ($use_div) $ret .= '<div>';
  306. $ret .= $html;
  307. if ($use_div) $ret .= '</div>';
  308. $ret .= '</body></html>';
  309. return $ret;
  310. }
  311. }
  312. // vim: et sw=4 sts=4