]> git.immae.eu Git - github/wallabag/wallabag.git/blame - inc/3rdparty/Readability.php
Merge branch 'Flattr' of git://github.com/tcitworld/poche into tcitworld-Flattr
[github/wallabag/wallabag.git] / inc / 3rdparty / Readability.php
CommitLineData
1a268ba7 1<?php
3db95a85 2/**
1a268ba7
NL
3* Arc90's Readability ported to PHP for FiveFilters.org
4* Based on readability.js version 1.7.1 (without multi-page support)
0ace6cab 5* Updated to allow HTML5 parsing with html5lib
6* Updated with lightClean mode to preserve more images and youtube/vimeo/viddler embeds
1a268ba7
NL
7* ------------------------------------------------------
8* Original URL: http://lab.arc90.com/experiments/readability/js/readability.js
9* Arc90's project URL: http://lab.arc90.com/experiments/readability/
10* JS Source: http://code.google.com/p/arc90labs-readability
11* Ported by: Keyvan Minoukadeh, http://www.keyvan.net
12* More information: http://fivefilters.org/content-only/
13* License: Apache License, Version 2.0
14* Requires: PHP5
0ace6cab 15* Date: 2012-09-19
3db95a85 16*
1a268ba7
NL
17* Differences between the PHP port and the original
18* ------------------------------------------------------
3db95a85
NL
19* Arc90's Readability is designed to run in the browser. It works on the DOM
20* tree (the parsed HTML) after the page's CSS styles have been applied and
21* Javascript code executed. This PHP port does not run inside a browser.
22* We use PHP's ability to parse HTML to build our DOM tree, but we cannot
23* rely on CSS or Javascript support. As such, the results will not always
24* match Arc90's Readability. (For example, if a web page contains CSS style
25* rules or Javascript code which hide certain HTML elements from display,
26* Arc90's Readability will dismiss those from consideration but our PHP port,
1a268ba7 27* unable to understand CSS or Javascript, will not know any better.)
3db95a85
NL
28*
29* Another significant difference is that the aim of Arc90's Readability is
30* to re-present the main content block of a given web page so users can
31* read it more easily in their browsers. Correct identification, clean up,
32* and separation of the content block is only a part of this process.
33* This PHP port is only concerned with this part, it does not include code
34* that relates to presentation in the browser - Arc90 already do
35* that extremely well, and for PDF output there's FiveFilters.org's
1a268ba7 36* PDF Newspaper: http://fivefilters.org/pdf-newspaper/.
3db95a85
NL
37*
38* Finally, this class contains methods that might be useful for developers
39* working on HTML document fragments. So without deviating too much from
40* the original code (which I don't want to do because it makes debugging
41* and updating more difficult), I've tried to make it a little more
42* developer friendly. You should be able to use the methods here on
43* existing DOMElement objects without passing an entire HTML document to
1a268ba7
NL
44* be parsed.
45*/
46
47// This class allows us to do JavaScript like assignements to innerHTML
48require_once(dirname(__FILE__).'/JSLikeHTMLElement.php');
49
50// Alternative usage (for testing only!)
3db95a85 51// uncomment the lines below and call Readability.php in your browser
1a268ba7
NL
52// passing it the URL of the page you'd like content from, e.g.:
53// Readability.php?url=http://medialens.org/alerts/09/090615_the_guardian_climate.php
54
55/*
56if (!isset($_GET['url']) || $_GET['url'] == '') {
0ace6cab 57 die('Please pass a URL to the script. E.g. Readability.php?url=bla.com/story.html');
1a268ba7
NL
58}
59$url = $_GET['url'];
60if (!preg_match('!^https?://!i', $url)) $url = 'http://'.$url;
61$html = file_get_contents($url);
62$r = new Readability($html, $url);
63$r->init();
64echo $r->articleContent->innerHTML;
65*/
66
1a268ba7
NL
67class Readability
68{
0ace6cab 69 public $version = '1.7.1-without-multi-page';
70 public $convertLinksToFootnotes = false;
71 public $revertForcedParagraphElements = true;
72 public $articleTitle;
73 public $articleContent;
74 public $dom;
75 public $url = null; // optional - URL where HTML was retrieved
76 public $debug = false;
77 public $lightClean = true; // preserves more content (experimental) added 2012-09-19
3db95a85 78 protected $body = null; //
0ace6cab 79 protected $bodyCache = null; // Cache the body HTML in case we need to re-use it later
80 protected $flags = 7; // 1 | 2 | 4; // Start with all flags set.
81 protected $success = false; // indicates whether we were able to extract or not
3db95a85 82
0ace6cab 83 /**
84 * All of the regular expressions in use within readability.
85 * Defined up here so we don't instantiate them repeatedly in loops.
86 **/
87 public $regexps = array(
88 'unlikelyCandidates' => '/combx|comment|community|disqus|extra|foot|header|menu|remark|rss|shoutbox|sidebar|sponsor|ad-break|agegate|pagination|pager|popup/i',
89 'okMaybeItsACandidate' => '/and|article|body|column|main|shadow/i',
90 'positive' => '/article|body|content|entry|hentry|main|page|attachment|pagination|post|text|blog|story/i',
91 'negative' => '/combx|comment|com-|contact|foot|footer|_nav|footnote|masthead|media|meta|outbrain|promo|related|scroll|shoutbox|sidebar|sponsor|shopping|tags|tool|widget/i',
92 'divToPElements' => '/<(a|blockquote|dl|div|img|ol|p|pre|table|ul)/i',
93 'replaceBrs' => '/(<br[^>]*>[ \n\r\t]*){2,}/i',
94 'replaceFonts' => '/<(\/?)font[^>]*>/i',
95 // 'trimRe' => '/^\s+|\s+$/g', // PHP has trim()
96 'normalize' => '/\s{2,}/',
97 'killBreaks' => '/(<br\s*\/?>(\s|&nbsp;?)*){1,}/',
98 'video' => '!//(player\.|www\.)?(youtube|vimeo|viddler)\.com!i',
99 'skipFootnoteLink' => '/^\s*(\[?[a-z0-9]{1,2}\]?|^|edit|citation needed)\s*$/i'
3db95a85
NL
100 );
101
0ace6cab 102 /* constants */
103 const FLAG_STRIP_UNLIKELYS = 1;
104 const FLAG_WEIGHT_CLASSES = 2;
105 const FLAG_CLEAN_CONDITIONALLY = 4;
3db95a85 106
0ace6cab 107 /**
108 * Create instance of Readability
109 * @param string UTF-8 encoded string
110 * @param string (optional) URL associated with HTML (used for footnotes)
111 * @param string which parser to use for turning raw HTML into a DOMDocument (either 'libxml' or 'html5lib')
3db95a85 112 */
0ace6cab 113 function __construct($html, $url=null, $parser='libxml')
114 {
115 $this->url = $url;
116 /* Turn all double br's into p's */
117 $html = preg_replace($this->regexps['replaceBrs'], '</p><p>', $html);
118 $html = preg_replace($this->regexps['replaceFonts'], '<$1span>', $html);
119 $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8");
120 if (trim($html) == '') $html = '<html></html>';
121 if ($parser=='html5lib' && ($this->dom = HTML5_Parser::parse($html))) {
122 // all good
123 } else {
124 $this->dom = new DOMDocument();
125 $this->dom->preserveWhiteSpace = false;
126 @$this->dom->loadHTML($html);
127 }
128 $this->dom->registerNodeClass('DOMElement', 'JSLikeHTMLElement');
129 }
130
131 /**
132 * Get article title element
133 * @return DOMElement
134 */
135 public function getTitle() {
136 return $this->articleTitle;
137 }
3db95a85 138
0ace6cab 139 /**
140 * Get article content element
141 * @return DOMElement
142 */
143 public function getContent() {
144 return $this->articleContent;
3db95a85
NL
145 }
146
0ace6cab 147 /**
148 * Runs readability.
3db95a85 149 *
0ace6cab 150 * Workflow:
151 * 1. Prep the document by removing script tags, css, etc.
152 * 2. Build readability's DOM tree.
153 * 3. Grab the article content from the current dom tree.
154 * 4. Replace the current DOM tree with the new one.
155 * 5. Read peacefully.
156 *
157 * @return boolean true if we found content, false otherwise
158 **/
159 public function init()
160 {
161 if (!isset($this->dom->documentElement)) return false;
162 $this->removeScripts($this->dom);
163 //die($this->getInnerHTML($this->dom->documentElement));
3db95a85 164
0ace6cab 165 // Assume successful outcome
166 $this->success = true;
167
168 $bodyElems = $this->dom->getElementsByTagName('body');
169 if ($bodyElems->length > 0) {
170 if ($this->bodyCache == null) {
171 $this->bodyCache = $bodyElems->item(0)->innerHTML;
172 }
173 if ($this->body == null) {
174 $this->body = $bodyElems->item(0);
175 }
176 }
177
178 $this->prepDocument();
3db95a85 179
0ace6cab 180 //die($this->dom->documentElement->parentNode->nodeType);
181 //$this->setInnerHTML($this->dom->documentElement, $this->getInnerHTML($this->dom->documentElement));
182 //die($this->getInnerHTML($this->dom->documentElement));
183
184 /* Build readability's DOM tree */
185 $overlay = $this->dom->createElement('div');
186 $innerDiv = $this->dom->createElement('div');
187 $articleTitle = $this->getArticleTitle();
188 $articleContent = $this->grabArticle();
189
190 if (!$articleContent) {
191 $this->success = false;
192 $articleContent = $this->dom->createElement('div');
193 $articleContent->setAttribute('id', 'readability-content');
3db95a85 194 $articleContent->innerHTML = '<p>Sorry, Readability was unable to parse this page for content.</p>';
0ace6cab 195 }
3db95a85 196
0ace6cab 197 $overlay->setAttribute('id', 'readOverlay');
198 $innerDiv->setAttribute('id', 'readInner');
199
200 /* Glue the structure of our document together. */
201 $innerDiv->appendChild($articleTitle);
202 $innerDiv->appendChild($articleContent);
203 $overlay->appendChild($innerDiv);
3db95a85 204
0ace6cab 205 /* Clear the old HTML, insert the new content. */
206 $this->body->innerHTML = '';
207 $this->body->appendChild($overlay);
208 //document.body.insertBefore(overlay, document.body.firstChild);
209 $this->body->removeAttribute('style');
210
211 $this->postProcessContent($articleContent);
3db95a85 212
0ace6cab 213 // Set title and content instance variables
214 $this->articleTitle = $articleTitle;
215 $this->articleContent = $articleContent;
3db95a85 216
0ace6cab 217 return $this->success;
218 }
3db95a85 219
0ace6cab 220 /**
221 * Debug
222 */
223 protected function dbg($msg) {
224 if ($this->debug) echo '* ',$msg, "\n";
225 }
3db95a85 226
0ace6cab 227 /**
228 * Run any post-process modifications to article content as necessary.
229 *
230 * @param DOMElement
231 * @return void
232 */
233 public function postProcessContent($articleContent) {
3db95a85 234 if ($this->convertLinksToFootnotes && !preg_match('/wikipedia\.org/', @$this->url)) {
0ace6cab 235 $this->addFootnotes($articleContent);
236 }
237 }
3db95a85 238
0ace6cab 239 /**
240 * Get the article title as an H1.
241 *
242 * @return DOMElement
243 */
244 protected function getArticleTitle() {
245 $curTitle = '';
246 $origTitle = '';
247
248 try {
249 $curTitle = $origTitle = $this->getInnerText($this->dom->getElementsByTagName('title')->item(0));
250 } catch(Exception $e) {}
3db95a85 251
0ace6cab 252 if (preg_match('/ [\|\-] /', $curTitle))
253 {
254 $curTitle = preg_replace('/(.*)[\|\-] .*/i', '$1', $origTitle);
3db95a85 255
0ace6cab 256 if (count(explode(' ', $curTitle)) < 3) {
257 $curTitle = preg_replace('/[^\|\-]*[\|\-](.*)/i', '$1', $origTitle);
258 }
259 }
260 else if (strpos($curTitle, ': ') !== false)
261 {
262 $curTitle = preg_replace('/.*:(.*)/i', '$1', $origTitle);
263
264 if (count(explode(' ', $curTitle)) < 3) {
265 $curTitle = preg_replace('/[^:]*[:](.*)/i','$1', $origTitle);
266 }
267 }
268 else if(strlen($curTitle) > 150 || strlen($curTitle) < 15)
269 {
270 $hOnes = $this->dom->getElementsByTagName('h1');
271 if($hOnes->length == 1)
272 {
273 $curTitle = $this->getInnerText($hOnes->item(0));
274 }
275 }
276
277 $curTitle = trim($curTitle);
278
279 if (count(explode(' ', $curTitle)) <= 4) {
280 $curTitle = $origTitle;
281 }
3db95a85 282
0ace6cab 283 $articleTitle = $this->dom->createElement('h1');
284 $articleTitle->innerHTML = $curTitle;
3db95a85 285
0ace6cab 286 return $articleTitle;
287 }
3db95a85 288
0ace6cab 289 /**
290 * Prepare the HTML document for readability to scrape it.
291 * This includes things like stripping javascript, CSS, and handling terrible markup.
3db95a85 292 *
0ace6cab 293 * @return void
294 **/
295 protected function prepDocument() {
296 /**
297 * In some cases a body element can't be found (if the HTML is totally hosed for example)
298 * so we create a new body node and append it to the document.
299 */
300 if ($this->body == null)
301 {
302 $this->body = $this->dom->createElement('body');
303 $this->dom->documentElement->appendChild($this->body);
304 }
305 $this->body->setAttribute('id', 'readabilityBody');
306
307 /* Remove all style tags in head */
308 $styleTags = $this->dom->getElementsByTagName('style');
309 for ($i = $styleTags->length-1; $i >= 0; $i--)
310 {
311 $styleTags->item($i)->parentNode->removeChild($styleTags->item($i));
312 }
313
314 /* Turn all double br's into p's */
315 /* Note, this is pretty costly as far as processing goes. Maybe optimize later. */
316 //document.body.innerHTML = document.body.innerHTML.replace(readability.regexps.replaceBrs, '</p><p>').replace(readability.regexps.replaceFonts, '<$1span>');
317 // We do this in the constructor for PHP as that's when we have raw HTML - before parsing it into a DOM tree.
318 // Manipulating innerHTML as it's done in JS is not possible in PHP.
319 }
320
321 /**
322 * For easier reading, convert this document to have footnotes at the bottom rather than inline links.
323 * @see http://www.roughtype.com/archives/2010/05/experiments_in.php
324 *
325 * @return void
326 **/
327 public function addFootnotes($articleContent) {
328 $footnotesWrapper = $this->dom->createElement('div');
329 $footnotesWrapper->setAttribute('id', 'readability-footnotes');
330 $footnotesWrapper->innerHTML = '<h3>References</h3>';
3db95a85 331
0ace6cab 332 $articleFootnotes = $this->dom->createElement('ol');
333 $articleFootnotes->setAttribute('id', 'readability-footnotes-list');
334 $footnotesWrapper->appendChild($articleFootnotes);
3db95a85 335
0ace6cab 336 $articleLinks = $articleContent->getElementsByTagName('a');
3db95a85 337
0ace6cab 338 $linkCount = 0;
339 for ($i = 0; $i < $articleLinks->length; $i++)
340 {
341 $articleLink = $articleLinks->item($i);
342 $footnoteLink = $articleLink->cloneNode(true);
343 $refLink = $this->dom->createElement('a');
344 $footnote = $this->dom->createElement('li');
345 $linkDomain = @parse_url($footnoteLink->getAttribute('href'), PHP_URL_HOST);
346 if (!$linkDomain && isset($this->url)) $linkDomain = @parse_url($this->url, PHP_URL_HOST);
347 //linkDomain = footnoteLink.host ? footnoteLink.host : document.location.host,
348 $linkText = $this->getInnerText($articleLink);
3db95a85 349
0ace6cab 350 if ((strpos($articleLink->getAttribute('class'), 'readability-DoNotFootnote') !== false) || preg_match($this->regexps['skipFootnoteLink'], $linkText)) {
351 continue;
352 }
3db95a85 353
0ace6cab 354 $linkCount++;
355
356 /** Add a superscript reference after the article link */
357 $refLink->setAttribute('href', '#readabilityFootnoteLink-' . $linkCount);
358 $refLink->innerHTML = '<small><sup>[' . $linkCount . ']</sup></small>';
359 $refLink->setAttribute('class', 'readability-DoNotFootnote');
360 $refLink->setAttribute('style', 'color: inherit;');
3db95a85 361
0ace6cab 362 //TODO: does this work or should we use DOMNode.isSameNode()?
363 if ($articleLink->parentNode->lastChild == $articleLink) {
364 $articleLink->parentNode->appendChild($refLink);
365 } else {
366 $articleLink->parentNode->insertBefore($refLink, $articleLink->nextSibling);
367 }
368
369 $articleLink->setAttribute('style', 'color: inherit; text-decoration: none;');
370 $articleLink->setAttribute('name', 'readabilityLink-' . $linkCount);
371
372 $footnote->innerHTML = '<small><sup><a href="#readabilityLink-' . $linkCount . '" title="Jump to Link in Article">^</a></sup></small> ';
373
374 $footnoteLink->innerHTML = ($footnoteLink->getAttribute('title') != '' ? $footnoteLink->getAttribute('title') : $linkText);
375 $footnoteLink->setAttribute('name', 'readabilityFootnoteLink-' . $linkCount);
3db95a85 376
0ace6cab 377 $footnote->appendChild($footnoteLink);
378 if ($linkDomain) $footnote->innerHTML = $footnote->innerHTML . '<small> (' . $linkDomain . ')</small>';
3db95a85 379
0ace6cab 380 $articleFootnotes->appendChild($footnote);
381 }
382
383 if ($linkCount > 0) {
3db95a85 384 $articleContent->appendChild($footnotesWrapper);
0ace6cab 385 }
386 }
387
388 /**
389 * Reverts P elements with class 'readability-styled'
390 * to text nodes - which is what they were before.
391 *
392 * @param DOMElement
393 * @return void
394 */
395 function revertReadabilityStyledElements($articleContent) {
396 $xpath = new DOMXPath($articleContent->ownerDocument);
397 $elems = $xpath->query('.//p[@class="readability-styled"]', $articleContent);
398 //$elems = $articleContent->getElementsByTagName('p');
399 for ($i = $elems->length-1; $i >= 0; $i--) {
400 $e = $elems->item($i);
401 $e->parentNode->replaceChild($articleContent->ownerDocument->createTextNode($e->textContent), $e);
402 //if ($e->hasAttribute('class') && $e->getAttribute('class') == 'readability-styled') {
403 // $e->parentNode->replaceChild($this->dom->createTextNode($e->textContent), $e);
404 //}
405 }
406 }
3db95a85 407
0ace6cab 408 /**
409 * Prepare the article node for display. Clean out any inline styles,
410 * iframes, forms, strip extraneous <p> tags, etc.
411 *
412 * @param DOMElement
413 * @return void
414 */
415 function prepArticle($articleContent) {
416 $this->cleanStyles($articleContent);
417 $this->killBreaks($articleContent);
418 if ($this->revertForcedParagraphElements) {
419 $this->revertReadabilityStyledElements($articleContent);
420 }
421
422 /* Clean out junk from the article content */
423 $this->cleanConditionally($articleContent, 'form');
424 $this->clean($articleContent, 'object');
425 $this->clean($articleContent, 'h1');
426
427 /**
428 * If there is only one h2, they are probably using it
429 * as a header and not a subheader, so remove it since we already have a header.
430 ***/
431 if (!$this->lightClean && ($articleContent->getElementsByTagName('h2')->length == 1)) {
3db95a85 432 $this->clean($articleContent, 'h2');
0ace6cab 433 }
434 $this->clean($articleContent, 'iframe');
435
436 $this->cleanHeaders($articleContent);
437
438 /* Do these last as the previous stuff may have removed junk that will affect these */
439 $this->cleanConditionally($articleContent, 'table');
440 $this->cleanConditionally($articleContent, 'ul');
441 $this->cleanConditionally($articleContent, 'div');
442
443 /* Remove extra paragraphs */
444 $articleParagraphs = $articleContent->getElementsByTagName('p');
445 for ($i = $articleParagraphs->length-1; $i >= 0; $i--)
446 {
447 $imgCount = $articleParagraphs->item($i)->getElementsByTagName('img')->length;
448 $embedCount = $articleParagraphs->item($i)->getElementsByTagName('embed')->length;
449 $objectCount = $articleParagraphs->item($i)->getElementsByTagName('object')->length;
450 $iframeCount = $articleParagraphs->item($i)->getElementsByTagName('iframe')->length;
3db95a85 451
0ace6cab 452 if ($imgCount === 0 && $embedCount === 0 && $objectCount === 0 && $iframeCount === 0 && $this->getInnerText($articleParagraphs->item($i), false) == '')
453 {
454 $articleParagraphs->item($i)->parentNode->removeChild($articleParagraphs->item($i));
455 }
456 }
457
458 try {
459 $articleContent->innerHTML = preg_replace('/<br[^>]*>\s*<p/i', '<p', $articleContent->innerHTML);
3db95a85 460 //articleContent.innerHTML = articleContent.innerHTML.replace(/<br[^>]*>\s*<p/gi, '<p');
0ace6cab 461 }
462 catch (Exception $e) {
463 $this->dbg("Cleaning innerHTML of breaks failed. This is an IE strict-block-elements bug. Ignoring.: " . $e);
464 }
465 }
3db95a85 466
0ace6cab 467 /**
468 * Initialize a node with the readability object. Also checks the
469 * className/id for special names to add to its score.
470 *
471 * @param Element
472 * @return void
473 **/
474 protected function initializeNode($node) {
475 $readability = $this->dom->createAttribute('readability');
476 $readability->value = 0; // this is our contentScore
3db95a85 477 $node->setAttributeNode($readability);
0ace6cab 478
479 switch (strtoupper($node->tagName)) { // unsure if strtoupper is needed, but using it just in case
480 case 'DIV':
481 $readability->value += 5;
482 break;
483
484 case 'PRE':
485 case 'TD':
486 case 'BLOCKQUOTE':
487 $readability->value += 3;
488 break;
3db95a85 489
0ace6cab 490 case 'ADDRESS':
491 case 'OL':
492 case 'UL':
493 case 'DL':
494 case 'DD':
495 case 'DT':
496 case 'LI':
497 case 'FORM':
498 $readability->value -= 3;
499 break;
500
501 case 'H1':
502 case 'H2':
503 case 'H3':
504 case 'H4':
505 case 'H5':
506 case 'H6':
507 case 'TH':
508 $readability->value -= 5;
509 break;
510 }
511 $readability->value += $this->getClassWeight($node);
512 }
3db95a85 513
0ace6cab 514 /***
515 * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is
516 * most likely to be the stuff a user wants to read. Then return it wrapped up in a div.
517 *
518 * @return DOMElement
519 **/
520 protected function grabArticle($page=null) {
521 $stripUnlikelyCandidates = $this->flagIsActive(self::FLAG_STRIP_UNLIKELYS);
522 if (!$page) $page = $this->dom;
523 $allElements = $page->getElementsByTagName('*');
524 /**
525 * First, node prepping. Trash nodes that look cruddy (like ones with the class name "comment", etc), and turn divs
526 * into P tags where they have been used inappropriately (as in, where they contain no other block level elements.)
527 *
528 * Note: Assignment from index for performance. See http://www.peachpit.com/articles/article.aspx?p=31567&seqNum=5
529 * TODO: Shouldn't this be a reverse traversal?
530 **/
531 $node = null;
532 $nodesToScore = array();
533 for ($nodeIndex = 0; ($node = $allElements->item($nodeIndex)); $nodeIndex++) {
534 //for ($nodeIndex=$targetList->length-1; $nodeIndex >= 0; $nodeIndex--) {
535 //$node = $targetList->item($nodeIndex);
536 $tagName = strtoupper($node->tagName);
537 /* Remove unlikely candidates */
538 if ($stripUnlikelyCandidates) {
539 $unlikelyMatchString = $node->getAttribute('class') . $node->getAttribute('id');
540 if (
541 preg_match($this->regexps['unlikelyCandidates'], $unlikelyMatchString) &&
542 !preg_match($this->regexps['okMaybeItsACandidate'], $unlikelyMatchString) &&
543 $tagName != 'BODY'
544 )
545 {
546 $this->dbg('Removing unlikely candidate - ' . $unlikelyMatchString);
547 //$nodesToRemove[] = $node;
548 $node->parentNode->removeChild($node);
549 $nodeIndex--;
550 continue;
3db95a85 551 }
0ace6cab 552 }
553
554 if ($tagName == 'P' || $tagName == 'TD' || $tagName == 'PRE') {
555 $nodesToScore[] = $node;
556 }
557
558 /* Turn all divs that don't have children block level elements into p's */
559 if ($tagName == 'DIV') {
560 if (!preg_match($this->regexps['divToPElements'], $node->innerHTML)) {
561 //$this->dbg('Altering div to p');
562 $newNode = $this->dom->createElement('p');
563 try {
564 $newNode->innerHTML = $node->innerHTML;
565 //$nodesToReplace[] = array('new'=>$newNode, 'old'=>$node);
566 $node->parentNode->replaceChild($newNode, $node);
567 $nodeIndex--;
568 $nodesToScore[] = $node; // or $newNode?
569 }
570 catch(Exception $e) {
571 $this->dbg('Could not alter div to p, reverting back to div.: ' . $e);
572 }
573 }
574 else
575 {
576 /* EXPERIMENTAL */
577 // TODO: change these p elements back to text nodes after processing
578 for ($i = 0, $il = $node->childNodes->length; $i < $il; $i++) {
579 $childNode = $node->childNodes->item($i);
580 if ($childNode->nodeType == 3) { // XML_TEXT_NODE
581 //$this->dbg('replacing text node with a p tag with the same content.');
582 $p = $this->dom->createElement('p');
583 $p->innerHTML = $childNode->nodeValue;
584 $p->setAttribute('style', 'display: inline;');
585 $p->setAttribute('class', 'readability-styled');
586 $childNode->parentNode->replaceChild($p, $childNode);
587 }
588 }
589 }
590 }
591 }
3db95a85 592
0ace6cab 593 /**
594 * Loop through all paragraphs, and assign a score to them based on how content-y they look.
595 * Then add their score to their parent node.
596 *
597 * A score is determined by things like number of commas, class names, etc. Maybe eventually link density.
598 **/
599 $candidates = array();
600 for ($pt=0; $pt < count($nodesToScore); $pt++) {
601 $parentNode = $nodesToScore[$pt]->parentNode;
602 // $grandParentNode = $parentNode ? $parentNode->parentNode : null;
603 $grandParentNode = !$parentNode ? null : (($parentNode->parentNode instanceof DOMElement) ? $parentNode->parentNode : null);
604 $innerText = $this->getInnerText($nodesToScore[$pt]);
605
606 if (!$parentNode || !isset($parentNode->tagName)) {
607 continue;
608 }
609
610 /* If this paragraph is less than 25 characters, don't even count it. */
611 if(strlen($innerText) < 25) {
612 continue;
613 }
614
615 /* Initialize readability data for the parent. */
3db95a85 616 if (!$parentNode->hasAttribute('readability'))
0ace6cab 617 {
618 $this->initializeNode($parentNode);
619 $candidates[] = $parentNode;
620 }
621
622 /* Initialize readability data for the grandparent. */
623 if ($grandParentNode && !$grandParentNode->hasAttribute('readability') && isset($grandParentNode->tagName))
624 {
625 $this->initializeNode($grandParentNode);
626 $candidates[] = $grandParentNode;
627 }
628
629 $contentScore = 0;
630
631 /* Add a point for the paragraph itself as a base. */
632 $contentScore++;
633
634 /* Add points for any commas within this paragraph */
635 $contentScore += count(explode(',', $innerText));
3db95a85 636
0ace6cab 637 /* For every 100 characters in this paragraph, add another point. Up to 3 points. */
638 $contentScore += min(floor(strlen($innerText) / 100), 3);
3db95a85 639
0ace6cab 640 /* Add the score to the parent. The grandparent gets half. */
641 $parentNode->getAttributeNode('readability')->value += $contentScore;
642
643 if ($grandParentNode) {
3db95a85 644 $grandParentNode->getAttributeNode('readability')->value += $contentScore/2;
0ace6cab 645 }
646 }
647
648 /**
649 * After we've calculated scores, loop through all of the possible candidate nodes we found
650 * and find the one with the highest score.
651 **/
652 $topCandidate = null;
653 for ($c=0, $cl=count($candidates); $c < $cl; $c++)
654 {
655 /**
656 * Scale the final candidates score based on link density. Good content should have a
657 * relatively small link density (5% or less) and be mostly unaffected by this operation.
658 **/
659 $readability = $candidates[$c]->getAttributeNode('readability');
660 $readability->value = $readability->value * (1-$this->getLinkDensity($candidates[$c]));
661
662 $this->dbg('Candidate: ' . $candidates[$c]->tagName . ' (' . $candidates[$c]->getAttribute('class') . ':' . $candidates[$c]->getAttribute('id') . ') with score ' . $readability->value);
663
664 if (!$topCandidate || $readability->value > (int)$topCandidate->getAttribute('readability')) {
665 $topCandidate = $candidates[$c];
666 }
667 }
668
669 /**
670 * If we still have no top candidate, just use the body as a last resort.
671 * We also have to copy the body node so it is something we can modify.
672 **/
673 if ($topCandidate === null || strtoupper($topCandidate->tagName) == 'BODY')
674 {
675 $topCandidate = $this->dom->createElement('div');
676 if ($page instanceof DOMDocument) {
677 if (!isset($page->documentElement)) {
678 // we don't have a body either? what a mess! :)
679 } else {
680 $topCandidate->innerHTML = $page->documentElement->innerHTML;
681 $page->documentElement->innerHTML = '';
682 $page->documentElement->appendChild($topCandidate);
683 }
684 } else {
685 $topCandidate->innerHTML = $page->innerHTML;
686 $page->innerHTML = '';
687 $page->appendChild($topCandidate);
688 }
689 $this->initializeNode($topCandidate);
690 }
691
692 /**
693 * Now that we have the top candidate, look through its siblings for content that might also be related.
694 * Things like preambles, content split by ads that we removed, etc.
695 **/
696 $articleContent = $this->dom->createElement('div');
697 $articleContent->setAttribute('id', 'readability-content');
698 $siblingScoreThreshold = max(10, ((int)$topCandidate->getAttribute('readability')) * 0.2);
699 $siblingNodes = $topCandidate->parentNode->childNodes;
700 if (!isset($siblingNodes)) {
701 $siblingNodes = new stdClass;
702 $siblingNodes->length = 0;
703 }
704
705 for ($s=0, $sl=$siblingNodes->length; $s < $sl; $s++)
706 {
707 $siblingNode = $siblingNodes->item($s);
708 $append = false;
709
710 $this->dbg('Looking at sibling node: ' . $siblingNode->nodeName . (($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->hasAttribute('readability')) ? (' with score ' . $siblingNode->getAttribute('readability')) : ''));
711
712 //dbg('Sibling has score ' . ($siblingNode->readability ? siblingNode.readability.contentScore : 'Unknown'));
713
714 if ($siblingNode === $topCandidate)
715 // or if ($siblingNode->isSameNode($topCandidate))
716 {
717 $append = true;
718 }
719
720 $contentBonus = 0;
721 /* Give a bonus if sibling nodes and top candidates have the example same classname */
722 if ($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->getAttribute('class') == $topCandidate->getAttribute('class') && $topCandidate->getAttribute('class') != '') {
723 $contentBonus += ((int)$topCandidate->getAttribute('readability')) * 0.2;
724 }
725
726 if ($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->hasAttribute('readability') && (((int)$siblingNode->getAttribute('readability')) + $contentBonus) >= $siblingScoreThreshold)
727 {
728 $append = true;
729 }
3db95a85 730
0ace6cab 731 if (strtoupper($siblingNode->nodeName) == 'P') {
732 $linkDensity = $this->getLinkDensity($siblingNode);
733 $nodeContent = $this->getInnerText($siblingNode);
734 $nodeLength = strlen($nodeContent);
3db95a85 735
0ace6cab 736 if ($nodeLength > 80 && $linkDensity < 0.25)
737 {
738 $append = true;
739 }
740 else if ($nodeLength < 80 && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent))
741 {
742 $append = true;
743 }
744 }
745
746 if ($append)
747 {
748 $this->dbg('Appending node: ' . $siblingNode->nodeName);
749
750 $nodeToAppend = null;
751 $sibNodeName = strtoupper($siblingNode->nodeName);
752 if ($sibNodeName != 'DIV' && $sibNodeName != 'P') {
753 /* We have a node that isn't a common block level element, like a form or td tag. Turn it into a div so it doesn't get filtered out later by accident. */
3db95a85 754
0ace6cab 755 $this->dbg('Altering siblingNode of ' . $sibNodeName . ' to div.');
756 $nodeToAppend = $this->dom->createElement('div');
757 try {
758 $nodeToAppend->setAttribute('id', $siblingNode->getAttribute('id'));
759 $nodeToAppend->innerHTML = $siblingNode->innerHTML;
760 }
761 catch(Exception $e)
762 {
763 $this->dbg('Could not alter siblingNode to div, reverting back to original.');
764 $nodeToAppend = $siblingNode;
765 $s--;
766 $sl--;
767 }
768 } else {
769 $nodeToAppend = $siblingNode;
770 $s--;
771 $sl--;
772 }
3db95a85 773
0ace6cab 774 /* To ensure a node does not interfere with readability styles, remove its classnames */
775 $nodeToAppend->removeAttribute('class');
776
777 /* Append sibling and subtract from our list because it removes the node when you append to another node */
778 $articleContent->appendChild($nodeToAppend);
779 }
780 }
781
782 /**
783 * So we have all of the content that we need. Now we clean it up for presentation.
784 **/
785 $this->prepArticle($articleContent);
786
787 /**
788 * Now that we've gone through the full algorithm, check to see if we got any meaningful content.
789 * If we didn't, we may need to re-run grabArticle with different flags set. This gives us a higher
790 * likelihood of finding the content, and the sieve approach gives us a higher likelihood of
791 * finding the -right- content.
792 **/
793 if (strlen($this->getInnerText($articleContent, false)) < 250)
794 {
795 // TODO: find out why element disappears sometimes, e.g. for this URL http://www.businessinsider.com/6-hedge-fund-etfs-for-average-investors-2011-7
796 // in the meantime, we check and create an empty element if it's not there.
797 if (!isset($this->body->childNodes)) $this->body = $this->dom->createElement('body');
798 $this->body->innerHTML = $this->bodyCache;
3db95a85 799
0ace6cab 800 if ($this->flagIsActive(self::FLAG_STRIP_UNLIKELYS)) {
801 $this->removeFlag(self::FLAG_STRIP_UNLIKELYS);
802 return $this->grabArticle($this->body);
803 }
804 else if ($this->flagIsActive(self::FLAG_WEIGHT_CLASSES)) {
805 $this->removeFlag(self::FLAG_WEIGHT_CLASSES);
3db95a85 806 return $this->grabArticle($this->body);
0ace6cab 807 }
808 else if ($this->flagIsActive(self::FLAG_CLEAN_CONDITIONALLY)) {
809 $this->removeFlag(self::FLAG_CLEAN_CONDITIONALLY);
810 return $this->grabArticle($this->body);
811 }
812 else {
5f9bff0f
NL
813 # this line was commented by Nicolas Lœuillet 8/8/13 due to some urls not parsed
814 // return false;
0ace6cab 815 }
816 }
817 return $articleContent;
818 }
3db95a85 819
0ace6cab 820 /**
821 * Remove script tags from document
822 *
823 * @param DOMElement
824 * @return void
825 */
826 public function removeScripts($doc) {
827 $scripts = $doc->getElementsByTagName('script');
828 for($i = $scripts->length-1; $i >= 0; $i--)
829 {
830 $scripts->item($i)->parentNode->removeChild($scripts->item($i));
831 }
832 }
3db95a85 833
0ace6cab 834 /**
835 * Get the inner text of a node.
836 * This also strips out any excess whitespace to be found.
837 *
838 * @param DOMElement $
839 * @param boolean $normalizeSpaces (default: true)
840 * @return string
841 **/
842 public function getInnerText($e, $normalizeSpaces=true) {
843 $textContent = '';
844
845 if (!isset($e->textContent) || $e->textContent == '') {
846 return '';
847 }
848
849 $textContent = trim($e->textContent);
850
851 if ($normalizeSpaces) {
852 return preg_replace($this->regexps['normalize'], ' ', $textContent);
853 } else {
854 return $textContent;
855 }
856 }
857
858 /**
859 * Get the number of times a string $s appears in the node $e.
860 *
861 * @param DOMElement $e
862 * @param string - what to count. Default is ","
863 * @return number (integer)
864 **/
865 public function getCharCount($e, $s=',') {
866 return substr_count($this->getInnerText($e), $s);
867 }
868
869 /**
870 * Remove the style attribute on every $e and under.
871 *
872 * @param DOMElement $e
873 * @return void
874 */
875 public function cleanStyles($e) {
876 if (!is_object($e)) return;
877 $elems = $e->getElementsByTagName('*');
878 foreach ($elems as $elem) {
879 $elem->removeAttribute('style');
880 }
881 }
3db95a85 882
0ace6cab 883 /**
884 * Get the density of links as a percentage of the content
885 * This is the amount of text that is inside a link divided by the total text in the node.
3db95a85 886 *
0ace6cab 887 * @param DOMElement $e
888 * @return number (float)
889 */
890 public function getLinkDensity($e) {
891 $links = $e->getElementsByTagName('a');
892 $textLength = strlen($this->getInnerText($e));
893 $linkLength = 0;
894 for ($i=0, $il=$links->length; $i < $il; $i++)
895 {
896 $linkLength += strlen($this->getInnerText($links->item($i)));
897 }
898 if ($textLength > 0) {
899 return $linkLength / $textLength;
900 } else {
901 return 0;
902 }
903 }
3db95a85 904
0ace6cab 905 /**
3db95a85 906 * Get an elements class/id weight. Uses regular expressions to tell if this
0ace6cab 907 * element looks good or bad.
908 *
909 * @param DOMElement $e
910 * @return number (Integer)
911 */
912 public function getClassWeight($e) {
913 if(!$this->flagIsActive(self::FLAG_WEIGHT_CLASSES)) {
914 return 0;
915 }
916
917 $weight = 0;
918
919 /* Look for a special classname */
920 if ($e->hasAttribute('class') && $e->getAttribute('class') != '')
921 {
922 if (preg_match($this->regexps['negative'], $e->getAttribute('class'))) {
923 $weight -= 25;
924 }
925 if (preg_match($this->regexps['positive'], $e->getAttribute('class'))) {
926 $weight += 25;
927 }
928 }
929
930 /* Look for a special ID */
931 if ($e->hasAttribute('id') && $e->getAttribute('id') != '')
932 {
933 if (preg_match($this->regexps['negative'], $e->getAttribute('id'))) {
934 $weight -= 25;
935 }
936 if (preg_match($this->regexps['positive'], $e->getAttribute('id'))) {
937 $weight += 25;
938 }
939 }
940 return $weight;
941 }
942
943 /**
944 * Remove extraneous break tags from a node.
945 *
946 * @param DOMElement $node
947 * @return void
948 */
949 public function killBreaks($node) {
950 $html = $node->innerHTML;
951 $html = preg_replace($this->regexps['killBreaks'], '<br />', $html);
952 $node->innerHTML = $html;
953 }
954
955 /**
956 * Clean a node of all elements of type "tag".
957 * (Unless it's a youtube/vimeo video. People love movies.)
958 *
959 * Updated 2012-09-18 to preserve youtube/vimeo iframes
960 *
961 * @param DOMElement $e
962 * @param string $tag
963 * @return void
964 */
965 public function clean($e, $tag) {
966 $targetList = $e->getElementsByTagName($tag);
967 $isEmbed = ($tag == 'iframe' || $tag == 'object' || $tag == 'embed');
3db95a85 968
0ace6cab 969 for ($y=$targetList->length-1; $y >= 0; $y--) {
970 /* Allow youtube and vimeo videos through as people usually want to see those. */
971 if ($isEmbed) {
972 $attributeValues = '';
973 for ($i=0, $il=$targetList->item($y)->attributes->length; $i < $il; $i++) {
974 $attributeValues .= $targetList->item($y)->attributes->item($i)->value . '|'; // DOMAttr? (TODO: test)
975 }
3db95a85 976
0ace6cab 977 /* First, check the elements attributes to see if any of them contain youtube or vimeo */
978 if (preg_match($this->regexps['video'], $attributeValues)) {
979 continue;
980 }
981
982 /* Then check the elements inside this element for the same. */
983 if (preg_match($this->regexps['video'], $targetList->item($y)->innerHTML)) {
984 continue;
985 }
986 }
987 $targetList->item($y)->parentNode->removeChild($targetList->item($y));
988 }
989 }
3db95a85 990
0ace6cab 991 /**
992 * Clean an element of all tags of type "tag" if they look fishy.
3db95a85 993 * "Fishy" is an algorithm based on content length, classnames,
0ace6cab 994 * link density, number of images & embeds, etc.
995 *
996 * @param DOMElement $e
997 * @param string $tag
998 * @return void
999 */
1000 public function cleanConditionally($e, $tag) {
1001 if (!$this->flagIsActive(self::FLAG_CLEAN_CONDITIONALLY)) {
1002 return;
1003 }
1004
1005 $tagsList = $e->getElementsByTagName($tag);
1006 $curTagsLength = $tagsList->length;
1007
1008 /**
1009 * Gather counts for other typical elements embedded within.
1010 * Traverse backwards so we can remove nodes at the same time without effecting the traversal.
1011 *
1012 * TODO: Consider taking into account original contentScore here.
1013 */
1014 for ($i=$curTagsLength-1; $i >= 0; $i--) {
1015 $weight = $this->getClassWeight($tagsList->item($i));
1016 $contentScore = ($tagsList->item($i)->hasAttribute('readability')) ? (int)$tagsList->item($i)->getAttribute('readability') : 0;
3db95a85 1017
0ace6cab 1018 $this->dbg('Cleaning Conditionally ' . $tagsList->item($i)->tagName . ' (' . $tagsList->item($i)->getAttribute('class') . ':' . $tagsList->item($i)->getAttribute('id') . ')' . (($tagsList->item($i)->hasAttribute('readability')) ? (' with score ' . $tagsList->item($i)->getAttribute('readability')) : ''));
1019
1020 if ($weight + $contentScore < 0) {
1021 $tagsList->item($i)->parentNode->removeChild($tagsList->item($i));
1022 }
1023 else if ( $this->getCharCount($tagsList->item($i), ',') < 10) {
1024 /**
1025 * If there are not very many commas, and the number of
1026 * non-paragraph elements is more than paragraphs or other ominous signs, remove the element.
1027 **/
1028 $p = $tagsList->item($i)->getElementsByTagName('p')->length;
1029 $img = $tagsList->item($i)->getElementsByTagName('img')->length;
1030 $li = $tagsList->item($i)->getElementsByTagName('li')->length-100;
1031 $input = $tagsList->item($i)->getElementsByTagName('input')->length;
1032 $a = $tagsList->item($i)->getElementsByTagName('a')->length;
1033
1034 $embedCount = 0;
1035 $embeds = $tagsList->item($i)->getElementsByTagName('embed');
1036 for ($ei=0, $il=$embeds->length; $ei < $il; $ei++) {
1037 if (preg_match($this->regexps['video'], $embeds->item($ei)->getAttribute('src'))) {
3db95a85 1038 $embedCount++;
0ace6cab 1039 }
1040 }
1041 $embeds = $tagsList->item($i)->getElementsByTagName('iframe');
1042 for ($ei=0, $il=$embeds->length; $ei < $il; $ei++) {
1043 if (preg_match($this->regexps['video'], $embeds->item($ei)->getAttribute('src'))) {
3db95a85 1044 $embedCount++;
0ace6cab 1045 }
1046 }
1047
1048 $linkDensity = $this->getLinkDensity($tagsList->item($i));
1049 $contentLength = strlen($this->getInnerText($tagsList->item($i)));
1050 $toRemove = false;
1051
1052 if ($this->lightClean) {
1053 $this->dbg('Light clean...');
1054 if ( ($img > $p) && ($img > 4) ) {
1055 $this->dbg(' more than 4 images and more image elements than paragraph elements');
1056 $toRemove = true;
1057 } else if ($li > $p && $tag != 'ul' && $tag != 'ol') {
1058 $this->dbg(' too many <li> elements, and parent is not <ul> or <ol>');
1059 $toRemove = true;
1060 } else if ( $input > floor($p/3) ) {
1061 $this->dbg(' too many <input> elements');
3db95a85 1062 $toRemove = true;
0ace6cab 1063 } else if ($contentLength < 25 && ($embedCount === 0 && ($img === 0 || $img > 2))) {
1064 $this->dbg(' content length less than 25 chars, 0 embeds and either 0 images or more than 2 images');
1065 $toRemove = true;
1066 } else if($weight < 25 && $linkDensity > 0.2) {
1067 $this->dbg(' weight smaller than 25 and link density above 0.2');
1068 $toRemove = true;
1069 } else if($a > 2 && ($weight >= 25 && $linkDensity > 0.5)) {
1070 $this->dbg(' more than 2 links and weight above 25 but link density greater than 0.5');
1071 $toRemove = true;
1072 } else if($embedCount > 3) {
1073 $this->dbg(' more than 3 embeds');
1074 $toRemove = true;
1075 }
1076 } else {
1077 $this->dbg('Standard clean...');
1078 if ( $img > $p ) {
1079 $this->dbg(' more image elements than paragraph elements');
1080 $toRemove = true;
1081 } else if ($li > $p && $tag != 'ul' && $tag != 'ol') {
1082 $this->dbg(' too many <li> elements, and parent is not <ul> or <ol>');
1083 $toRemove = true;
1084 } else if ( $input > floor($p/3) ) {
1085 $this->dbg(' too many <input> elements');
3db95a85 1086 $toRemove = true;
0ace6cab 1087 } else if ($contentLength < 25 && ($img === 0 || $img > 2) ) {
1088 $this->dbg(' content length less than 25 chars and 0 images, or more than 2 images');
1089 $toRemove = true;
1090 } else if($weight < 25 && $linkDensity > 0.2) {
1091 $this->dbg(' weight smaller than 25 and link density above 0.2');
1092 $toRemove = true;
1093 } else if($weight >= 25 && $linkDensity > 0.5) {
1094 $this->dbg(' weight above 25 but link density greater than 0.5');
1095 $toRemove = true;
1096 } else if(($embedCount == 1 && $contentLength < 75) || $embedCount > 1) {
1097 $this->dbg(' 1 embed and content length smaller than 75 chars, or more than one embed');
1098 $toRemove = true;
1099 }
1100 }
1101
1102 if ($toRemove) {
1103 //$this->dbg('Removing: '.$tagsList->item($i)->innerHTML);
1104 $tagsList->item($i)->parentNode->removeChild($tagsList->item($i));
1105 }
1106 }
1107 }
1108 }
1109
1110 /**
1111 * Clean out spurious headers from an Element. Checks things like classnames and link density.
1112 *
1113 * @param DOMElement $e
1114 * @return void
1115 */
1116 public function cleanHeaders($e) {
1117 for ($headerIndex = 1; $headerIndex < 3; $headerIndex++) {
1118 $headers = $e->getElementsByTagName('h' . $headerIndex);
1119 for ($i=$headers->length-1; $i >=0; $i--) {
1120 if ($this->getClassWeight($headers->item($i)) < 0 || $this->getLinkDensity($headers->item($i)) > 0.33) {
1121 $headers->item($i)->parentNode->removeChild($headers->item($i));
1122 }
1123 }
1124 }
1125 }
1126
1127 public function flagIsActive($flag) {
1128 return ($this->flags & $flag) > 0;
1129 }
3db95a85 1130
0ace6cab 1131 public function addFlag($flag) {
1132 $this->flags = $this->flags | $flag;
1133 }
3db95a85 1134
0ace6cab 1135 public function removeFlag($flag) {
1136 $this->flags = $this->flags & ~$flag;
1137 }
1138}