DOMLex.php 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291
  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. $doc->encoding = 'UTF-8'; // theoretically, the above has this covered
  62. set_error_handler(array($this, 'muteErrorHandler'));
  63. $doc->loadHTML($html);
  64. restore_error_handler();
  65. $body = $doc->getElementsByTagName('html')->item(0)-> // <html>
  66. getElementsByTagName('body')->item(0); // <body>
  67. $div = $body->getElementsByTagName('div')->item(0); // <div>
  68. $tokens = array();
  69. $this->tokenizeDOM($div, $tokens, $config);
  70. // If the div has a sibling, that means we tripped across
  71. // a premature </div> tag. So remove the div we parsed,
  72. // and then tokenize the rest of body. We can't tokenize
  73. // the sibling directly as we'll lose the tags in that case.
  74. if ($div->nextSibling) {
  75. $body->removeChild($div);
  76. $this->tokenizeDOM($body, $tokens, $config);
  77. }
  78. return $tokens;
  79. }
  80. /**
  81. * Iterative function that tokenizes a node, putting it into an accumulator.
  82. * To iterate is human, to recurse divine - L. Peter Deutsch
  83. * @param DOMNode $node DOMNode to be tokenized.
  84. * @param HTMLPurifier_Token[] $tokens Array-list of already tokenized tokens.
  85. * @return HTMLPurifier_Token of node appended to previously passed tokens.
  86. */
  87. protected function tokenizeDOM($node, &$tokens, $config)
  88. {
  89. $level = 0;
  90. $nodes = array($level => new HTMLPurifier_Queue(array($node)));
  91. $closingNodes = array();
  92. do {
  93. while (!$nodes[$level]->isEmpty()) {
  94. $node = $nodes[$level]->shift(); // FIFO
  95. $collect = $level > 0 ? true : false;
  96. $needEndingTag = $this->createStartNode($node, $tokens, $collect, $config);
  97. if ($needEndingTag) {
  98. $closingNodes[$level][] = $node;
  99. }
  100. if ($node->childNodes && $node->childNodes->length) {
  101. $level++;
  102. $nodes[$level] = new HTMLPurifier_Queue();
  103. foreach ($node->childNodes as $childNode) {
  104. $nodes[$level]->push($childNode);
  105. }
  106. }
  107. }
  108. $level--;
  109. if ($level && isset($closingNodes[$level])) {
  110. while ($node = array_pop($closingNodes[$level])) {
  111. $this->createEndNode($node, $tokens);
  112. }
  113. }
  114. } while ($level > 0);
  115. }
  116. /**
  117. * @param DOMNode $node DOMNode to be tokenized.
  118. * @param HTMLPurifier_Token[] $tokens Array-list of already tokenized tokens.
  119. * @param bool $collect Says whether or start and close are collected, set to
  120. * false at first recursion because it's the implicit DIV
  121. * tag you're dealing with.
  122. * @return bool if the token needs an endtoken
  123. * @todo data and tagName properties don't seem to exist in DOMNode?
  124. */
  125. protected function createStartNode($node, &$tokens, $collect, $config)
  126. {
  127. // intercept non element nodes. WE MUST catch all of them,
  128. // but we're not getting the character reference nodes because
  129. // those should have been preprocessed
  130. if ($node->nodeType === XML_TEXT_NODE) {
  131. $tokens[] = $this->factory->createText($node->data);
  132. return false;
  133. } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
  134. // undo libxml's special treatment of <script> and <style> tags
  135. $last = end($tokens);
  136. $data = $node->data;
  137. // (note $node->tagname is already normalized)
  138. if ($last instanceof HTMLPurifier_Token_Start && ($last->name == 'script' || $last->name == 'style')) {
  139. $new_data = trim($data);
  140. if (substr($new_data, 0, 4) === '<!--') {
  141. $data = substr($new_data, 4);
  142. if (substr($data, -3) === '-->') {
  143. $data = substr($data, 0, -3);
  144. } else {
  145. // Highly suspicious! Not sure what to do...
  146. }
  147. }
  148. }
  149. $tokens[] = $this->factory->createText($this->parseText($data, $config));
  150. return false;
  151. } elseif ($node->nodeType === XML_COMMENT_NODE) {
  152. // this is code is only invoked for comments in script/style in versions
  153. // of libxml pre-2.6.28 (regular comments, of course, are still
  154. // handled regularly)
  155. $tokens[] = $this->factory->createComment($node->data);
  156. return false;
  157. } elseif ($node->nodeType !== XML_ELEMENT_NODE) {
  158. // not-well tested: there may be other nodes we have to grab
  159. return false;
  160. }
  161. $attr = $node->hasAttributes() ? $this->transformAttrToAssoc($node->attributes) : array();
  162. // We still have to make sure that the element actually IS empty
  163. if (!$node->childNodes->length) {
  164. if ($collect) {
  165. $tokens[] = $this->factory->createEmpty($node->tagName, $attr);
  166. }
  167. return false;
  168. } else {
  169. if ($collect) {
  170. $tokens[] = $this->factory->createStart(
  171. $tag_name = $node->tagName, // somehow, it get's dropped
  172. $attr
  173. );
  174. }
  175. return true;
  176. }
  177. }
  178. /**
  179. * @param DOMNode $node
  180. * @param HTMLPurifier_Token[] $tokens
  181. */
  182. protected function createEndNode($node, &$tokens)
  183. {
  184. $tokens[] = $this->factory->createEnd($node->tagName);
  185. }
  186. /**
  187. * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
  188. *
  189. * @param DOMNamedNodeMap $node_map DOMNamedNodeMap of DOMAttr objects.
  190. * @return array Associative array of attributes.
  191. */
  192. protected function transformAttrToAssoc($node_map)
  193. {
  194. // NamedNodeMap is documented very well, so we're using undocumented
  195. // features, namely, the fact that it implements Iterator and
  196. // has a ->length attribute
  197. if ($node_map->length === 0) {
  198. return array();
  199. }
  200. $array = array();
  201. foreach ($node_map as $attr) {
  202. $array[$attr->name] = $attr->value;
  203. }
  204. return $array;
  205. }
  206. /**
  207. * An error handler that mutes all errors
  208. * @param int $errno
  209. * @param string $errstr
  210. */
  211. public function muteErrorHandler($errno, $errstr)
  212. {
  213. }
  214. /**
  215. * Callback function for undoing escaping of stray angled brackets
  216. * in comments
  217. * @param array $matches
  218. * @return string
  219. */
  220. public function callbackUndoCommentSubst($matches)
  221. {
  222. return '<!--' . strtr($matches[1], array('&amp;' => '&', '&lt;' => '<')) . $matches[2];
  223. }
  224. /**
  225. * Callback function that entity-izes ampersands in comments so that
  226. * callbackUndoCommentSubst doesn't clobber them
  227. * @param array $matches
  228. * @return string
  229. */
  230. public function callbackArmorCommentEntities($matches)
  231. {
  232. return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
  233. }
  234. /**
  235. * Wraps an HTML fragment in the necessary HTML
  236. * @param string $html
  237. * @param HTMLPurifier_Config $config
  238. * @param HTMLPurifier_Context $context
  239. * @return string
  240. */
  241. protected function wrapHTML($html, $config, $context, $use_div = true)
  242. {
  243. $def = $config->getDefinition('HTML');
  244. $ret = '';
  245. if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
  246. $ret .= '<!DOCTYPE html ';
  247. if (!empty($def->doctype->dtdPublic)) {
  248. $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
  249. }
  250. if (!empty($def->doctype->dtdSystem)) {
  251. $ret .= '"' . $def->doctype->dtdSystem . '" ';
  252. }
  253. $ret .= '>';
  254. }
  255. $ret .= '<html><head>';
  256. $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
  257. // No protection if $html contains a stray </div>!
  258. $ret .= '</head><body>';
  259. if ($use_div) $ret .= '<div>';
  260. $ret .= $html;
  261. if ($use_div) $ret .= '</div>';
  262. $ret .= '</body></html>';
  263. return $ret;
  264. }
  265. }
  266. // vim: et sw=4 sts=4