Whoops, forgot to edit WHATSNEW
[htmlpurifier.git] / library / HTMLPurifier / Lexer / DOMLex.php
blob6238a99e34e59cb38d4e61eebdc3517dae101066
1 <?php
3 /**
4 * Parser that uses PHP 5's DOM extension (part of the core).
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.
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.
16 * @note PHP's DOM extension does not actually parse any entities, we use
17 * our own function to do that.
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
27 class HTMLPurifier_Lexer_DOMLex extends HTMLPurifier_Lexer
30 /**
31 * @type HTMLPurifier_TokenFactory
33 private $factory;
35 public function __construct()
37 // setup the factory
38 parent::__construct();
39 $this->factory = new HTMLPurifier_TokenFactory();
42 /**
43 * @param string $html
44 * @param HTMLPurifier_Config $config
45 * @param HTMLPurifier_Context $context
46 * @return HTMLPurifier_Token[]
48 public function tokenizeHTML($html, $config, $context)
50 $html = $this->normalize($html, $config, $context);
52 // attempt to armor stray angled brackets that cannot possibly
53 // form tags and thus are probably being used as emoticons
54 if ($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
65 // preprocess html, essential for UTF-8
66 $html = $this->wrapHTML($html, $config, $context);
68 $doc = new DOMDocument();
69 $doc->encoding = 'UTF-8'; // theoretically, the above has this covered
71 set_error_handler(array($this, 'muteErrorHandler'));
72 $doc->loadHTML($html);
73 restore_error_handler();
75 $body = $doc->getElementsByTagName('html')->item(0)-> // <html>
76 getElementsByTagName('body')->item(0); // <body>
78 $div = $body->getElementsByTagName('div')->item(0); // <div>
79 $tokens = array();
80 $this->tokenizeDOM($div, $tokens, $config);
81 // If the div has a sibling, that means we tripped across
82 // a premature </div> tag. So remove the div we parsed,
83 // and then tokenize the rest of body. We can't tokenize
84 // the sibling directly as we'll lose the tags in that case.
85 if ($div->nextSibling) {
86 $body->removeChild($div);
87 $this->tokenizeDOM($body, $tokens, $config);
89 return $tokens;
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[] $tokens Array-list of already tokenized tokens.
97 * @return HTMLPurifier_Token of node appended to previously passed tokens.
99 protected function tokenizeDOM($node, &$tokens, $config)
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, $config);
109 if ($needEndingTag) {
110 $closingNodes[$level][] = $node;
112 if ($node->childNodes && $node->childNodes->length) {
113 $level++;
114 $nodes[$level] = new HTMLPurifier_Queue();
115 foreach ($node->childNodes as $childNode) {
116 $nodes[$level]->push($childNode);
120 $level--;
121 if ($level && isset($closingNodes[$level])) {
122 while ($node = array_pop($closingNodes[$level])) {
123 $this->createEndNode($node, $tokens);
126 } while ($level > 0);
130 * Portably retrieve the tag name of a node; deals with older versions
131 * of libxml like 2.7.6
132 * @param DOMNode $node
134 protected function getTagName($node)
136 if (property_exists($node, 'tagName')) {
137 return $node->tagName;
138 } else if (property_exists($node, 'nodeName')) {
139 return $node->nodeName;
140 } else if (property_exists($node, 'localName')) {
141 return $node->localName;
143 return null;
147 * Portably retrieve the data of a node; deals with older versions
148 * of libxml like 2.7.6
149 * @param DOMNode $node
151 protected function getData($node)
153 if (property_exists($node, 'data')) {
154 return $node->data;
155 } else if (property_exists($node, 'nodeValue')) {
156 return $node->nodeValue;
157 } else if (property_exists($node, 'textContent')) {
158 return $node->textContent;
160 return null;
165 * @param DOMNode $node DOMNode to be tokenized.
166 * @param HTMLPurifier_Token[] $tokens Array-list of already tokenized tokens.
167 * @param bool $collect Says whether or start and close are collected, set to
168 * false at first recursion because it's the implicit DIV
169 * tag you're dealing with.
170 * @return bool if the token needs an endtoken
171 * @todo data and tagName properties don't seem to exist in DOMNode?
173 protected function createStartNode($node, &$tokens, $collect, $config)
175 // intercept non element nodes. WE MUST catch all of them,
176 // but we're not getting the character reference nodes because
177 // those should have been preprocessed
178 if ($node->nodeType === XML_TEXT_NODE) {
179 $data = $this->getData($node); // Handle variable data property
180 if ($data !== null) {
181 $tokens[] = $this->factory->createText($data);
183 return false;
184 } elseif ($node->nodeType === XML_CDATA_SECTION_NODE) {
185 // undo libxml's special treatment of <script> and <style> tags
186 $last = end($tokens);
187 $data = $node->data;
188 // (note $node->tagname is already normalized)
189 if ($last instanceof HTMLPurifier_Token_Start && ($last->name == 'script' || $last->name == 'style')) {
190 $new_data = trim($data);
191 if (substr($new_data, 0, 4) === '<!--') {
192 $data = substr($new_data, 4);
193 if (substr($data, -3) === '-->') {
194 $data = substr($data, 0, -3);
195 } else {
196 // Highly suspicious! Not sure what to do...
200 $tokens[] = $this->factory->createText($this->parseText($data, $config));
201 return false;
202 } elseif ($node->nodeType === XML_COMMENT_NODE) {
203 // this is code is only invoked for comments in script/style in versions
204 // of libxml pre-2.6.28 (regular comments, of course, are still
205 // handled regularly)
206 $tokens[] = $this->factory->createComment($node->data);
207 return false;
208 } elseif ($node->nodeType !== XML_ELEMENT_NODE) {
209 // not-well tested: there may be other nodes we have to grab
210 return false;
212 $attr = $node->hasAttributes() ? $this->transformAttrToAssoc($node->attributes) : array();
213 $tag_name = $this->getTagName($node); // Handle variable tagName property
214 if (empty($tag_name)) {
215 return (bool) $node->childNodes->length;
217 // We still have to make sure that the element actually IS empty
218 if (!$node->childNodes->length) {
219 if ($collect) {
220 $tokens[] = $this->factory->createEmpty($tag_name, $attr);
222 return false;
223 } else {
224 if ($collect) {
225 $tokens[] = $this->factory->createStart($tag_name, $attr);
227 return true;
232 * @param DOMNode $node
233 * @param HTMLPurifier_Token[] $tokens
235 protected function createEndNode($node, &$tokens)
237 $tag_name = $this->getTagName($node); // Handle variable tagName property
238 $tokens[] = $this->factory->createEnd($tag_name);
242 * Converts a DOMNamedNodeMap of DOMAttr objects into an assoc array.
244 * @param DOMNamedNodeMap $node_map DOMNamedNodeMap of DOMAttr objects.
245 * @return array Associative array of attributes.
247 protected function transformAttrToAssoc($node_map)
249 // NamedNodeMap is documented very well, so we're using undocumented
250 // features, namely, the fact that it implements Iterator and
251 // has a ->length attribute
252 if ($node_map->length === 0) {
253 return array();
255 $array = array();
256 foreach ($node_map as $attr) {
257 $array[$attr->name] = $attr->value;
259 return $array;
263 * An error handler that mutes all errors
264 * @param int $errno
265 * @param string $errstr
267 public function muteErrorHandler($errno, $errstr)
272 * Callback function for undoing escaping of stray angled brackets
273 * in comments
274 * @param array $matches
275 * @return string
277 public function callbackUndoCommentSubst($matches)
279 return '<!--' . strtr($matches[1], array('&amp;' => '&', '&lt;' => '<')) . $matches[2];
283 * Callback function that entity-izes ampersands in comments so that
284 * callbackUndoCommentSubst doesn't clobber them
285 * @param array $matches
286 * @return string
288 public function callbackArmorCommentEntities($matches)
290 return '<!--' . str_replace('&', '&amp;', $matches[1]) . $matches[2];
294 * Wraps an HTML fragment in the necessary HTML
295 * @param string $html
296 * @param HTMLPurifier_Config $config
297 * @param HTMLPurifier_Context $context
298 * @return string
300 protected function wrapHTML($html, $config, $context, $use_div = true)
302 $def = $config->getDefinition('HTML');
303 $ret = '';
305 if (!empty($def->doctype->dtdPublic) || !empty($def->doctype->dtdSystem)) {
306 $ret .= '<!DOCTYPE html ';
307 if (!empty($def->doctype->dtdPublic)) {
308 $ret .= 'PUBLIC "' . $def->doctype->dtdPublic . '" ';
310 if (!empty($def->doctype->dtdSystem)) {
311 $ret .= '"' . $def->doctype->dtdSystem . '" ';
313 $ret .= '>';
316 $ret .= '<html><head>';
317 $ret .= '<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />';
318 // No protection if $html contains a stray </div>!
319 $ret .= '</head><body>';
320 if ($use_div) $ret .= '<div>';
321 $ret .= $html;
322 if ($use_div) $ret .= '</div>';
323 $ret .= '</body></html>';
324 return $ret;
328 // vim: et sw=4 sts=4