3 * Arc90's Readability ported to PHP for FiveFilters.org
4 * Based on readability.js version 1.7.1 (without multi-page support)
5 * ------------------------------------------------------
6 * Original URL: http://lab.arc90.com/experiments/readability/js/readability.js
7 * Arc90's project URL: http://lab.arc90.com/experiments/readability/
8 * JS Source: http://code.google.com/p/arc90labs-readability
9 * Ported by: Keyvan Minoukadeh, http://www.keyvan.net
10 * More information: http://fivefilters.org/content-only/
11 * License: Apache License, Version 2.0
15 * Differences between the PHP port and the original
16 * ------------------------------------------------------
17 * Arc90's Readability is designed to run in the browser. It works on the DOM
18 * tree (the parsed HTML) after the page's CSS styles have been applied and
19 * Javascript code executed. This PHP port does not run inside a browser.
20 * We use PHP's ability to parse HTML to build our DOM tree, but we cannot
21 * rely on CSS or Javascript support. As such, the results will not always
22 * match Arc90's Readability. (For example, if a web page contains CSS style
23 * rules or Javascript code which hide certain HTML elements from display,
24 * Arc90's Readability will dismiss those from consideration but our PHP port,
25 * unable to understand CSS or Javascript, will not know any better.)
27 * Another significant difference is that the aim of Arc90's Readability is
28 * to re-present the main content block of a given web page so users can
29 * read it more easily in their browsers. Correct identification, clean up,
30 * and separation of the content block is only a part of this process.
31 * This PHP port is only concerned with this part, it does not include code
32 * that relates to presentation in the browser - Arc90 already do
33 * that extremely well, and for PDF output there's FiveFilters.org's
34 * PDF Newspaper: http://fivefilters.org/pdf-newspaper/.
36 * Finally, this class contains methods that might be useful for developers
37 * working on HTML document fragments. So without deviating too much from
38 * the original code (which I don't want to do because it makes debugging
39 * and updating more difficult), I've tried to make it a little more
40 * developer friendly. You should be able to use the methods here on
41 * existing DOMElement objects without passing an entire HTML document to
45 // This class allows us to do JavaScript like assignements to innerHTML
46 require_once(dirname(__FILE__
).'/JSLikeHTMLElement.php');
48 // Alternative usage (for testing only!)
49 // uncomment the lines below and call Readability.php in your browser
50 // passing it the URL of the page you'd like content from, e.g.:
51 // Readability.php?url=http://medialens.org/alerts/09/090615_the_guardian_climate.php
54 if (!isset($_GET['url']) || $_GET['url'] == '') {
55 die('Please pass a URL to the script. E.g. Readability.php?url=bla.com/story.html');
58 if (!preg_match('!^https?://!i', $url)) $url = 'http://'.$url;
59 $html = file_get_contents($url);
60 $r = new Readability($html, $url);
62 echo $r->articleContent->innerHTML;
69 const FLAG_STRIP_UNLIKELYS
= 1;
70 const FLAG_WEIGHT_CLASSES
= 2;
71 const FLAG_CLEAN_CONDITIONALLY
= 4;
73 public $version = '1.7.1-without-multi-page';
74 public $convertLinksToFootnotes = false;
75 public $revertForcedParagraphElements = true;
77 public $articleContent;
79 public $url = null; // optional - URL where HTML was retrieved
80 public $debug = false;
81 protected $body = null; //
82 protected $bodyCache = null; // Cache the body HTML in case we need to re-use it later
83 protected $flags = Self
::FLAG_CLEAN_CONDITIONALLY
; // 1 | 2 | 4; // Start with all flags set.
84 protected $success = false; // indicates whether we were able to extract or not
87 * All of the regular expressions in use within readability.
88 * Defined up here so we don't instantiate them repeatedly in loops.
90 public $regexps = array(
91 'unlikelyCandidates' => '/combx|comment|comments|community|disqus|extra|foot|header|menu|remark|rss|shoutbox|sidebar|sponsor|ad-break|agegate|pagination|pager|popup|tweet|twitter/i',
92 'okMaybeItsACandidate' => '/and|article|body|column|main|shadow/i',
93 'positive' => '/article|body|content|entry|hentry|main|page|pagination|post|text|blog|story|attachment/i',
94 'negative' => '/combx|comment|comments|com-|contact|foot|footer|footnote|masthead|media|meta|outbrain|promo|related|scroll|shoutbox|sidebar|sponsor|shopping|tags|tool|widget/i',
95 'divToPElements' => '/<(a|blockquote|dl|div|ol|p|pre|table|ul)/i',
96 'replaceBrs' => '/(<br[^>]*>[ \n\r\t]*){2,}/i',
97 'replaceFonts' => '/<(\/?)font[^>]*>/i',
98 // 'trimRe' => '/^\s+|\s+$/g', // PHP has trim()
99 'normalize' => '/\s{2,}/',
100 'killBreaks' => '/(<br\s*\/?>(\s| ?)*){1,}/',
101 'video' => '/http:\/\/(www\.)?(youtube|vimeo|dailymotion)\.com/i',
102 'skipFootnoteLink' => '/^\s*(\[?[a-z0-9]{1,2}\]?|^|edit|citation needed)\s*$/i'
106 * Create instance of Readability
107 * @param string UTF-8 encoded string
108 * @param string (optional) URL associated with HTML (used for footnotes)
110 function __construct($html, $url=null)
112 /* Turn all double br's into p's */
113 $html = preg_replace($this->regexps
['replaceBrs'], '</p><p>', $html);
114 $html = preg_replace($this->regexps
['replaceFonts'], '<$1span>', $html);
115 $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8");
116 $this->dom
= new DOMDocument();
117 $this->dom
->preserveWhiteSpace
= false;
118 $this->dom
->registerNodeClass('DOMElement', 'JSLikeHTMLElement');
119 if (trim($html) == '') $html = '<html></html>';
120 @$this->dom
->loadHTML($html);
125 * Get article title element
128 public function getTitle() {
129 return $this->articleTitle
;
133 * Get article content element
136 public function getContent() {
137 return $this->articleContent
;
144 * 1. Prep the document by removing script tags, css, etc.
145 * 2. Build readability's DOM tree.
146 * 3. Grab the article content from the current dom tree.
147 * 4. Replace the current DOM tree with the new one.
148 * 5. Read peacefully.
150 * @return boolean true if we found content, false otherwise
152 public function init()
154 if (!isset($this->dom
->documentElement
)) return false;
155 $this->removeScripts($this->dom
);
156 //die($this->getInnerHTML($this->dom->documentElement));
158 // Assume successful outcome
159 $this->success
= true;
161 $bodyElems = $this->dom
->getElementsByTagName('body');
162 if ($bodyElems->length
> 0) {
163 if ($this->bodyCache
== null) {
164 $this->bodyCache
= $bodyElems->item(0)->innerHTML
;
166 if ($this->body
== null) {
167 $this->body
= $bodyElems->item(0);
171 $this->prepDocument();
173 //die($this->dom->documentElement->parentNode->nodeType);
174 //$this->setInnerHTML($this->dom->documentElement, $this->getInnerHTML($this->dom->documentElement));
175 //die($this->getInnerHTML($this->dom->documentElement));
177 /* Build readability's DOM tree */
178 $overlay = $this->dom
->createElement('div');
179 $innerDiv = $this->dom
->createElement('div');
180 $articleTitle = $this->getArticleTitle();
181 $articleContent = $this->grabArticle();
183 if (!$articleContent) {
184 $this->success
= false;
185 $articleContent = $this->dom
->createElement('div');
186 $articleContent->setAttribute('id', 'readability-content');
187 $articleContent->innerHTML
= '<p>Sorry, Readability was unable to parse this page for content.</p>';
188 return $this->success
;
191 $overlay->setAttribute('id', 'readOverlay');
192 $innerDiv->setAttribute('id', 'readInner');
194 /* Glue the structure of our document together. */
195 $innerDiv->appendChild($articleTitle);
196 $innerDiv->appendChild($articleContent);
197 $overlay->appendChild($innerDiv);
199 /* Clear the old HTML, insert the new content. */
200 $this->body
->innerHTML
= '';
201 $this->body
->appendChild($overlay);
202 //document.body.insertBefore(overlay, document.body.firstChild);
203 $this->body
->removeAttribute('style');
205 $this->postProcessContent($articleContent);
207 // Set title and content instance variables
208 $this->articleTitle
= $articleTitle;
209 $this->articleContent
= $articleContent;
211 return $this->success
;
217 protected function dbg($msg) {
218 if ($this->debug
) echo '* ',$msg, '<br />', "\n";
222 * Run any post-process modifications to article content as necessary.
227 public function postProcessContent($articleContent) {
228 if ($this->convertLinksToFootnotes
&& !preg_match('/wikipedia\.org/', @$this->url
)) {
229 $this->addFootnotes($articleContent);
234 * Get the article title as an H1.
238 protected function getArticleTitle() {
243 $curTitle = $origTitle = $this->getInnerText($this->dom
->getElementsByTagName('title')->item(0));
244 } catch(Exception
$e) {}
246 if (preg_match('/ [\|\-] /', $curTitle))
248 $curTitle = preg_replace('/(.*)[\|\-] .*/i', '$1', $origTitle);
250 if (count(explode(' ', $curTitle)) < 3) {
251 $curTitle = preg_replace('/[^\
|\
-]*[\
|\
-](.*)/i
', '$1', $origTitle);
254 else if (strpos($curTitle, ': ') !== false)
256 $curTitle = preg_replace('/.*:(.*)/i', '$1', $origTitle);
258 if (count(explode(' ', $curTitle)) < 3) {
259 $curTitle = preg_replace('/[^
:]*[:](.*)/i
','$1', $origTitle);
262 else if(strlen($curTitle) > 150 || strlen($curTitle) < 15)
264 $hOnes = $this->dom
->getElementsByTagName('h1');
265 if($hOnes->length
== 1)
267 $curTitle = $this->getInnerText($hOnes->item(0));
271 $curTitle = trim($curTitle);
273 if (count(explode(' ', $curTitle)) <= 4) {
274 $curTitle = $origTitle;
277 $articleTitle = $this->dom
->createElement('h1');
278 $articleTitle->innerHTML
= $curTitle;
280 return $articleTitle;
284 * Prepare the HTML document for readability to scrape it.
285 * This includes things like stripping javascript, CSS, and handling terrible markup.
289 protected function prepDocument() {
291 * In some cases a body element can't be found (if the HTML is totally hosed for example)
292 * so we create a new body node and append it to the document.
294 if ($this->body
== null)
296 $this->body
= $this->dom
->createElement('body');
297 $this->dom
->documentElement
->appendChild($this->body
);
299 $this->body
->setAttribute('id', 'readabilityBody');
301 /* Remove all style tags in head */
302 $styleTags = $this->dom
->getElementsByTagName('style');
303 for ($i = $styleTags->length
-1; $i >= 0; $i--)
305 $styleTags->item($i)->parentNode
->removeChild($styleTags->item($i));
308 /* Turn all double br's into p's */
309 /* Note, this is pretty costly as far as processing goes. Maybe optimize later. */
310 //document.body.innerHTML = document.body.innerHTML.replace(readability.regexps.replaceBrs, '</p><p>').replace(readability.regexps.replaceFonts, '<$1span>');
311 // We do this in the constructor for PHP as that's when we have raw HTML - before parsing it into a DOM tree.
312 // Manipulating innerHTML as it's done in JS is not possible in PHP.
316 * For easier reading, convert this document to have footnotes at the bottom rather than inline links.
317 * @see http://www.roughtype.com/archives/2010/05/experiments_in.php
321 public function addFootnotes($articleContent) {
322 $footnotesWrapper = $this->dom
->createElement('div');
323 $footnotesWrapper->setAttribute('id', 'readability-footnotes');
324 $footnotesWrapper->innerHTML
= '<h3>References</h3>';
326 $articleFootnotes = $this->dom
->createElement('ol');
327 $articleFootnotes->setAttribute('id', 'readability-footnotes-list');
328 $footnotesWrapper->appendChild($articleFootnotes);
330 $articleLinks = $articleContent->getElementsByTagName('a');
333 for ($i = 0; $i < $articleLinks->length
; $i++
)
335 $articleLink = $articleLinks->item($i);
336 $footnoteLink = $articleLink->cloneNode(true);
337 $refLink = $this->dom
->createElement('a');
338 $footnote = $this->dom
->createElement('li');
339 $linkDomain = @parse_url($footnoteLink->getAttribute('href'), PHP_URL_HOST
);
340 if (!$linkDomain && isset($this->url
)) $linkDomain = @parse_url($this->url
, PHP_URL_HOST
);
341 //linkDomain = footnoteLink.host ? footnoteLink.host : document.location.host,
342 $linkText = $this->getInnerText($articleLink);
344 if ((strpos($articleLink->getAttribute('class'), 'readability-DoNotFootnote') !== false) || preg_match($this->regexps
['skipFootnoteLink'], $linkText)) {
350 /** Add a superscript reference after the article link */
351 $refLink->setAttribute('href', '#readabilityFootnoteLink-' . $linkCount);
352 $refLink->innerHTML
= '<small><sup>[' . $linkCount . ']</sup></small>';
353 $refLink->setAttribute('class', 'readability-DoNotFootnote');
354 $refLink->setAttribute('style', 'color: inherit;');
356 //TODO: does this work or should we use DOMNode.isSameNode()?
357 if ($articleLink->parentNode
->lastChild
== $articleLink) {
358 $articleLink->parentNode
->appendChild($refLink);
360 $articleLink->parentNode
->insertBefore($refLink, $articleLink->nextSibling
);
363 $articleLink->setAttribute('style', 'color: inherit; text-decoration: none;');
364 $articleLink->setAttribute('name', 'readabilityLink-' . $linkCount);
366 $footnote->innerHTML
= '<small><sup><a href="#readabilityLink-' . $linkCount . '" title="Jump to Link in Article">^</a></sup></small> ';
368 $footnoteLink->innerHTML
= ($footnoteLink->getAttribute('title') != '' ? $footnoteLink->getAttribute('title') : $linkText);
369 $footnoteLink->setAttribute('name', 'readabilityFootnoteLink-' . $linkCount);
371 $footnote->appendChild($footnoteLink);
372 if ($linkDomain) $footnote->innerHTML
= $footnote->innerHTML
. '<small> (' . $linkDomain . ')</small>';
374 $articleFootnotes->appendChild($footnote);
377 if ($linkCount > 0) {
378 $articleContent->appendChild($footnotesWrapper);
383 * Reverts P elements with class 'readability-styled'
384 * to text nodes - which is what they were before.
389 function revertReadabilityStyledElements($articleContent) {
390 $xpath = new DOMXPath($articleContent->ownerDocument
);
391 $elems = $xpath->query('.//p[@class="readability-styled"]', $articleContent);
392 //$elems = $articleContent->getElementsByTagName('p');
393 for ($i = $elems->length
-1; $i >= 0; $i--) {
394 $e = $elems->item($i);
395 $e->parentNode
->replaceChild($articleContent->ownerDocument
->createTextNode($e->textContent
), $e);
396 //if ($e->hasAttribute('class') && $e->getAttribute('class') == 'readability-styled') {
397 // $e->parentNode->replaceChild($this->dom->createTextNode($e->textContent), $e);
403 * Prepare the article node for display. Clean out any inline styles,
404 * iframes, forms, strip extraneous <p> tags, etc.
409 function prepArticle($articleContent) {
410 $this->cleanStyles($articleContent);
411 $this->killBreaks($articleContent);
412 if ($this->revertForcedParagraphElements
) {
413 $this->revertReadabilityStyledElements($articleContent);
416 /* Clean out junk from the article content */
417 $this->cleanConditionally($articleContent, 'form');
418 $this->clean($articleContent, 'object');
419 $this->clean($articleContent, 'h1');
422 * If there is only one h2, they are probably using it
423 * as a header and not a subheader, so remove it since we already have a header.
425 if ($articleContent->getElementsByTagName('h2')->length
== 1) {
426 $this->clean($articleContent, 'h2');
428 $this->clean($articleContent, 'iframe');
430 $this->cleanHeaders($articleContent);
432 /* Do these last as the previous stuff may have removed junk that will affect these */
433 $this->cleanConditionally($articleContent, 'table');
434 $this->cleanConditionally($articleContent, 'ul');
435 $this->cleanConditionally($articleContent, 'div');
437 /* Remove extra paragraphs */
438 $articleParagraphs = $articleContent->getElementsByTagName('p');
439 for ($i = $articleParagraphs->length
-1; $i >= 0; $i--)
441 $imgCount = $articleParagraphs->item($i)->getElementsByTagName('img')->length
;
442 $embedCount = $articleParagraphs->item($i)->getElementsByTagName('embed')->length
;
443 $objectCount = $articleParagraphs->item($i)->getElementsByTagName('object')->length
;
445 if ($imgCount === 0 && $embedCount === 0 && $objectCount === 0 && $this->getInnerText($articleParagraphs->item($i), false) == '')
447 $articleParagraphs->item($i)->parentNode
->removeChild($articleParagraphs->item($i));
452 $articleContent->innerHTML
= preg_replace('/<br[^>]*>\s*<p/i', '<p', $articleContent->innerHTML
);
453 //articleContent.innerHTML = articleContent.innerHTML.replace(/<br[^>]*>\s*<p/gi, '<p');
455 catch (Exception
$e) {
456 $this->dbg("Cleaning innerHTML of breaks failed. This is an IE strict-block-elements bug. Ignoring.: " . $e);
461 * Initialize a node with the readability object. Also checks the
462 * className/id for special names to add to its score.
467 protected function initializeNode($node) {
468 $readability = $this->dom
->createAttribute('readability');
469 $readability->value
= 0; // this is our contentScore
470 $node->setAttributeNode($readability);
472 switch (strtoupper($node->tagName
)) { // unsure if strtoupper is needed, but using it just in case
474 $readability->value +
= 5;
480 $readability->value +
= 3;
491 $readability->value
-= 3;
501 $readability->value
-= 5;
504 $readability->value +
= $this->getClassWeight($node);
508 * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is
509 * most likely to be the stuff a user wants to read. Then return it wrapped up in a div.
513 protected function grabArticle($page=null) {
514 $stripUnlikelyCandidates = $this->flagIsActive(self
::FLAG_STRIP_UNLIKELYS
);
515 if (!$page) $page = $this->dom
;
516 $allElements = $page->getElementsByTagName('*');
518 * First, node prepping. Trash nodes that look cruddy (like ones with the class name "comment", etc), and turn divs
519 * into P tags where they have been used inappropriately (as in, where they contain no other block level elements.)
521 * Note: Assignment from index for performance. See http://www.peachpit.com/articles/article.aspx?p=31567&seqNum=5
522 * TODO: Shouldn't this be a reverse traversal?
525 $nodesToScore = array();
526 for ($nodeIndex = 0; ($node = $allElements->item($nodeIndex)); $nodeIndex++
) {
527 //for ($nodeIndex=$targetList->length-1; $nodeIndex >= 0; $nodeIndex--) {
528 //$node = $targetList->item($nodeIndex);
529 $tagName = strtoupper($node->tagName
);
530 /* Remove unlikely candidates */
531 if ($stripUnlikelyCandidates) {
532 $unlikelyMatchString = $node->getAttribute('class') . $node->getAttribute('id');
534 preg_match($this->regexps
['unlikelyCandidates'], $unlikelyMatchString) &&
535 !preg_match($this->regexps
['okMaybeItsACandidate'], $unlikelyMatchString) &&
539 $this->dbg('Removing unlikely candidate - ' . $unlikelyMatchString);
540 //$nodesToRemove[] = $node;
541 $node->parentNode
->removeChild($node);
547 if ($tagName == 'P' || $tagName == 'TD' || $tagName == 'PRE') {
548 $nodesToScore[] = $node;
551 /* Turn all divs that don't have children block level elements into p's */
552 if ($tagName == 'DIV') {
553 if (!preg_match($this->regexps
['divToPElements'], $node->innerHTML
)) {
554 //$this->dbg('Altering div to p');
555 $newNode = $this->dom
->createElement('p');
557 $newNode->innerHTML
= $node->innerHTML
;
558 //$nodesToReplace[] = array('new'=>$newNode, 'old'=>$node);
559 $node->parentNode
->replaceChild($newNode, $node);
561 $nodesToScore[] = $node; // or $newNode?
563 catch(Exception
$e) {
564 $this->dbg('Could not alter div to p, reverting back to div.: ' . $e);
570 // TODO: change these p elements back to text nodes after processing
571 for ($i = 0, $il = $node->childNodes
->length
; $i < $il; $i++
) {
572 $childNode = $node->childNodes
->item($i);
573 if ($childNode->nodeType
== 3) { // XML_TEXT_NODE
574 //$this->dbg('replacing text node with a p tag with the same content.');
575 $p = $this->dom
->createElement('p');
576 $p->innerHTML
= $childNode->nodeValue
;
577 $p->setAttribute('style', 'display: inline;');
578 $p->setAttribute('class', 'readability-styled');
579 $childNode->parentNode
->replaceChild($p, $childNode);
587 * Loop through all paragraphs, and assign a score to them based on how content-y they look.
588 * Then add their score to their parent node.
590 * A score is determined by things like number of commas, class names, etc. Maybe eventually link density.
592 $candidates = array();
593 for ($pt=0; $pt < count($nodesToScore); $pt++
) {
594 $parentNode = $nodesToScore[$pt]->parentNode
;
595 // $grandParentNode = $parentNode ? $parentNode->parentNode : null;
596 $grandParentNode = !$parentNode ? null : (($parentNode->parentNode
instanceof DOMElement
) ? $parentNode->parentNode
: null);
597 $innerText = $this->getInnerText($nodesToScore[$pt]);
599 if (!$parentNode || !isset($parentNode->tagName
)) {
603 /* If this paragraph is less than 25 characters, don't even count it. */
604 if(strlen($innerText) < 25) {
608 /* Initialize readability data for the parent. */
609 if (!$parentNode->hasAttribute('readability'))
611 $this->initializeNode($parentNode);
612 $candidates[] = $parentNode;
615 /* Initialize readability data for the grandparent. */
616 if ($grandParentNode && !$grandParentNode->hasAttribute('readability') && isset($grandParentNode->tagName
))
618 $this->initializeNode($grandParentNode);
619 $candidates[] = $grandParentNode;
624 /* Add a point for the paragraph itself as a base. */
627 /* Add points for any commas within this paragraph */
628 $contentScore +
= count(explode(',', $innerText));
630 /* For every 100 characters in this paragraph, add another point. Up to 3 points. */
631 $contentScore +
= min(floor(strlen($innerText) / 100), 3);
633 /* Add the score to the parent. The grandparent gets half. */
634 $parentNode->getAttributeNode('readability')->value +
= $contentScore;
636 if ($grandParentNode) {
637 $grandParentNode->getAttributeNode('readability')->value +
= $contentScore/2;
642 * After we've calculated scores, loop through all of the possible candidate nodes we found
643 * and find the one with the highest score.
645 $topCandidate = null;
646 for ($c=0, $cl=count($candidates); $c < $cl; $c++
)
649 * Scale the final candidates score based on link density. Good content should have a
650 * relatively small link density (5% or less) and be mostly unaffected by this operation.
652 $readability = $candidates[$c]->getAttributeNode('readability');
653 $readability->value
= $readability->value
* (1-$this->getLinkDensity($candidates[$c]));
655 $this->dbg('Candidate: ' . $candidates[$c]->tagName
. ' (' . $candidates[$c]->getAttribute('class') . ':' . $candidates[$c]->getAttribute('id') . ') with score ' . $readability->value
);
657 if (!$topCandidate || $readability->value
> (int)$topCandidate->getAttribute('readability')) {
658 $topCandidate = $candidates[$c];
663 * If we still have no top candidate, just use the body as a last resort.
664 * We also have to copy the body node so it is something we can modify.
666 if ($topCandidate === null || strtoupper($topCandidate->tagName
) == 'BODY')
668 $topCandidate = $this->dom
->createElement('div');
669 if ($page instanceof DOMDocument
) {
670 if (!isset($page->documentElement
)) {
671 // we don't have a body either? what a mess! :)
673 $topCandidate->innerHTML
= $page->documentElement
->innerHTML
;
674 $page->documentElement
->innerHTML
= '';
675 $page->documentElement
->appendChild($topCandidate);
678 $topCandidate->innerHTML
= $page->innerHTML
;
679 $page->innerHTML
= '';
680 $page->appendChild($topCandidate);
682 $this->initializeNode($topCandidate);
686 * Now that we have the top candidate, look through its siblings for content that might also be related.
687 * Things like preambles, content split by ads that we removed, etc.
689 $articleContent = $this->dom
->createElement('div');
690 $articleContent->setAttribute('id', 'readability-content');
691 $siblingScoreThreshold = max(10, ((int)$topCandidate->getAttribute('readability')) * 0.2);
692 $siblingNodes = $topCandidate->parentNode
->childNodes
;
693 if (!isset($siblingNodes)) {
694 $siblingNodes = new stdClass
;
695 $siblingNodes->length
= 0;
698 for ($s=0, $sl=$siblingNodes->length
; $s < $sl; $s++
)
700 $siblingNode = $siblingNodes->item($s);
703 $this->dbg('Looking at sibling node: ' . $siblingNode->nodeName
. (($siblingNode->nodeType
=== XML_ELEMENT_NODE
&& $siblingNode->hasAttribute('readability')) ? (' with score ' . $siblingNode->getAttribute('readability')) : ''));
705 //dbg('Sibling has score ' . ($siblingNode->readability ? siblingNode.readability.contentScore : 'Unknown'));
707 if ($siblingNode === $topCandidate)
708 // or if ($siblingNode->isSameNode($topCandidate))
714 /* Give a bonus if sibling nodes and top candidates have the example same classname */
715 if ($siblingNode->nodeType
=== XML_ELEMENT_NODE
&& $siblingNode->getAttribute('class') == $topCandidate->getAttribute('class') && $topCandidate->getAttribute('class') != '') {
716 $contentBonus +
= ((int)$topCandidate->getAttribute('readability')) * 0.2;
719 if ($siblingNode->nodeType
=== XML_ELEMENT_NODE
&& $siblingNode->hasAttribute('readability') && (((int)$siblingNode->getAttribute('readability')) +
$contentBonus) >= $siblingScoreThreshold)
724 if (strtoupper($siblingNode->nodeName
) == 'P') {
725 $linkDensity = $this->getLinkDensity($siblingNode);
726 $nodeContent = $this->getInnerText($siblingNode);
727 $nodeLength = strlen($nodeContent);
729 if ($nodeLength > 80 && $linkDensity < 0.25)
733 else if ($nodeLength < 80 && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent))
739 /* Look for a special classname */
740 if ($siblingNode->nodeType
=== XML_ELEMENT_NODE
&& $siblingNode->hasAttribute('class') && $siblingNode->getAttribute('class') != '')
742 if (preg_match($this->regexps
['okMaybeItsACandidate'], $siblingNode->getAttribute('class'))) {
747 /* Look for a special classname */
748 if ($siblingNode->nodeType
=== XML_ELEMENT_NODE
&& $siblingNode->hasAttribute('id') && $siblingNode->getAttribute('id') != '')
750 if (preg_match($this->regexps
['okMaybeItsACandidate'], $siblingNode->getAttribute('id'))) {
758 $this->dbg('Appending node: ' . $siblingNode->nodeName
);
760 $nodeToAppend = null;
761 $sibNodeName = strtoupper($siblingNode->nodeName
);
762 if ($sibNodeName != 'DIV' && $sibNodeName != 'P') {
763 /* 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. */
765 $this->dbg('Altering siblingNode of ' . $sibNodeName . ' to div.');
766 $nodeToAppend = $this->dom
->createElement('div');
768 $nodeToAppend->setAttribute('id', $siblingNode->getAttribute('id'));
769 $nodeToAppend->innerHTML
= $siblingNode->innerHTML
;
773 $this->dbg('Could not alter siblingNode to div, reverting back to original.');
774 $nodeToAppend = $siblingNode;
779 $nodeToAppend = $siblingNode;
784 /* To ensure a node does not interfere with readability styles, remove its classnames */
785 $nodeToAppend->removeAttribute('class');
787 /* Append sibling and subtract from our list because it removes the node when you append to another node */
788 $articleContent->appendChild($nodeToAppend);
793 * So we have all of the content that we need. Now we clean it up for presentation.
795 $this->prepArticle($articleContent);
798 * Now that we've gone through the full algorithm, check to see if we got any meaningful content.
799 * If we didn't, we may need to re-run grabArticle with different flags set. This gives us a higher
800 * likelihood of finding the content, and the sieve approach gives us a higher likelihood of
801 * finding the -right- content.
803 if (strlen($this->getInnerText($articleContent, false)) < 250)
805 // 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
806 // in the meantime, we check and create an empty element if it's not there.
807 if (!isset($this->body
->childNodes
)) $this->body
= $this->dom
->createElement('body');
808 $this->body
->innerHTML
= $this->bodyCache
;
810 if ($this->flagIsActive(self
::FLAG_STRIP_UNLIKELYS
)) {
811 $this->removeFlag(self
::FLAG_STRIP_UNLIKELYS
);
812 return $this->grabArticle($this->body
);
814 else if ($this->flagIsActive(self
::FLAG_WEIGHT_CLASSES
)) {
815 $this->removeFlag(self
::FLAG_WEIGHT_CLASSES
);
816 return $this->grabArticle($this->body
);
818 else if ($this->flagIsActive(self
::FLAG_CLEAN_CONDITIONALLY
)) {
819 $this->removeFlag(self
::FLAG_CLEAN_CONDITIONALLY
);
820 return $this->grabArticle($this->body
);
826 return $articleContent;
830 * Remove script tags from document
835 public function removeScripts($doc) {
836 $scripts = $doc->getElementsByTagName('script');
837 for($i = $scripts->length
-1; $i >= 0; $i--)
839 $scripts->item($i)->parentNode
->removeChild($scripts->item($i));
844 * Get the inner text of a node.
845 * This also strips out any excess whitespace to be found.
847 * @param DOMElement $
848 * @param boolean $normalizeSpaces (default: true)
851 public function getInnerText($e, $normalizeSpaces=true) {
854 if (!isset($e->textContent
) || $e->textContent
== '') {
858 $textContent = trim($e->textContent
);
860 if ($normalizeSpaces) {
861 return preg_replace($this->regexps
['normalize'], ' ', $textContent);
868 * Get the number of times a string $s appears in the node $e.
870 * @param DOMElement $e
871 * @param string - what to count. Default is ","
872 * @return number (integer)
874 public function getCharCount($e, $s=',') {
875 return substr_count($this->getInnerText($e), $s);
879 * Remove the style attribute on every $e and under.
881 * @param DOMElement $e
884 public function cleanStyles($e) {
885 if (!is_object($e)) return;
886 $elems = $e->getElementsByTagName('*');
887 foreach ($elems as $elem) {
888 $elem->removeAttribute('style');
893 * Get the density of links as a percentage of the content
894 * This is the amount of text that is inside a link divided by the total text in the node.
896 * @param DOMElement $e
897 * @return number (float)
899 public function getLinkDensity($e) {
900 $links = $e->getElementsByTagName('a');
901 $textLength = strlen($this->getInnerText($e));
903 for ($i=0, $il=$links->length
; $i < $il; $i++
)
905 $linkLength +
= strlen($this->getInnerText($links->item($i)));
907 if ($textLength > 0) {
908 return $linkLength / $textLength;
915 * Get an elements class/id weight. Uses regular expressions to tell if this
916 * element looks good or bad.
918 * @param DOMElement $e
919 * @return number (Integer)
921 public function getClassWeight($e) {
922 if(!$this->flagIsActive(self
::FLAG_WEIGHT_CLASSES
)) {
928 /* Look for a special classname */
929 if ($e->hasAttribute('class') && $e->getAttribute('class') != '')
931 if (preg_match($this->regexps
['negative'], $e->getAttribute('class'))) {
934 if (preg_match($this->regexps
['positive'], $e->getAttribute('class'))) {
939 /* Look for a special ID */
940 if ($e->hasAttribute('id') && $e->getAttribute('id') != '')
942 if (preg_match($this->regexps
['negative'], $e->getAttribute('id'))) {
945 if (preg_match($this->regexps
['positive'], $e->getAttribute('id'))) {
953 * Remove extraneous break tags from a node.
955 * @param DOMElement $node
958 public function killBreaks($node) {
959 $html = $node->innerHTML
;
960 $html = preg_replace($this->regexps
['killBreaks'], '<br />', $html);
961 $node->innerHTML
= $html;
965 * Clean a node of all elements of type "tag".
966 * (Unless it's a youtube/vimeo video. People love movies.)
968 * @param DOMElement $e
972 public function clean($e, $tag) {
973 $targetList = $e->getElementsByTagName($tag);
974 $isEmbed = ($tag == 'object' || $tag == 'embed');
976 for ($y=$targetList->length
-1; $y >= 0; $y--) {
977 /* Allow youtube and vimeo videos through as people usually want to see those. */
979 $attributeValues = '';
980 for ($i=0, $il=$targetList->item($y)->attributes
->length
; $i < $il; $i++
) {
981 $attributeValues .= $targetList->item($y)->attributes
->item($i)->value
. '|'; // DOMAttr? (TODO: test)
984 /* First, check the elements attributes to see if any of them contain youtube or vimeo */
985 if (preg_match($this->regexps
['video'], $attributeValues)) {
989 /* Then check the elements inside this element for the same. */
990 if (preg_match($this->regexps
['video'], $targetList->item($y)->innerHTML
)) {
994 $targetList->item($y)->parentNode
->removeChild($targetList->item($y));
999 * Clean an element of all tags of type "tag" if they look fishy.
1000 * "Fishy" is an algorithm based on content length, classnames,
1001 * link density, number of images & embeds, etc.
1003 * @param DOMElement $e
1004 * @param string $tag
1007 public function cleanConditionally($e, $tag) {
1008 if (!$this->flagIsActive(self
::FLAG_CLEAN_CONDITIONALLY
)) {
1012 $tagsList = $e->getElementsByTagName($tag);
1013 $curTagsLength = $tagsList->length
;
1016 * Gather counts for other typical elements embedded within.
1017 * Traverse backwards so we can remove nodes at the same time without effecting the traversal.
1019 * TODO: Consider taking into account original contentScore here.
1021 for ($i=$curTagsLength-1; $i >= 0; $i--) {
1022 $weight = $this->getClassWeight($tagsList->item($i));
1023 $contentScore = ($tagsList->item($i)->hasAttribute('readability')) ? (int)$tagsList->item($i)->getAttribute('readability') : 0;
1025 $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')) : ''));
1027 if ($weight +
$contentScore < 0) {
1028 $tagsList->item($i)->parentNode
->removeChild($tagsList->item($i));
1030 else if ( $this->getCharCount($tagsList->item($i), ',') < 10) {
1032 * If there are not very many commas, and the number of
1033 * non-paragraph elements is more than paragraphs or other ominous signs, remove the element.
1035 $p = $tagsList->item($i)->getElementsByTagName('p')->length
;
1036 $img = $tagsList->item($i)->getElementsByTagName('img')->length
;
1037 $li = $tagsList->item($i)->getElementsByTagName('li')->length
-100;
1038 $input = $tagsList->item($i)->getElementsByTagName('input')->length
;
1041 $embeds = $tagsList->item($i)->getElementsByTagName('embed');
1042 for ($ei=0, $il=$embeds->length
; $ei < $il; $ei++
) {
1043 if (preg_match($this->regexps
['video'], $embeds->item($ei)->getAttribute('src'))) {
1048 $linkDensity = $this->getLinkDensity($tagsList->item($i));
1049 $contentLength = strlen($this->getInnerText($tagsList->item($i)));
1054 } else if ($li > $p && $tag != 'ul' && $tag != 'ol') {
1056 } else if ( $input > floor($p/3) ) {
1058 } else if ($contentLength < 25 && ($img === 0 || $img > 2) ) {
1060 } else if($weight < 25 && $linkDensity > 0.2) {
1062 } else if($weight >= 25 && $linkDensity > 0.5) {
1064 } else if(($embedCount == 1 && $contentLength < 75) || $embedCount > 1) {
1069 $tagsList->item($i)->parentNode
->removeChild($tagsList->item($i));
1076 * Clean out spurious headers from an Element. Checks things like classnames and link density.
1078 * @param DOMElement $e
1081 public function cleanHeaders($e) {
1082 for ($headerIndex = 1; $headerIndex < 3; $headerIndex++
) {
1083 $headers = $e->getElementsByTagName('h' . $headerIndex);
1084 for ($i=$headers->length
-1; $i >=0; $i--) {
1085 if ($this->getClassWeight($headers->item($i)) < 0 || $this->getLinkDensity($headers->item($i)) > 0.33) {
1086 $headers->item($i)->parentNode
->removeChild($headers->item($i));
1092 public function flagIsActive($flag) {
1093 return ($this->flags
& $flag) > 0;
1096 public function addFlag($flag) {
1097 $this->flags
= $this->flags
| $flag;
1100 public function removeFlag($flag) {
1101 $this->flags
= $this->flags
& ~
$flag;