From 42c80841c846610be280218d53fcde06b0f0063b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicolas=20L=C5=93uillet?= Date: Fri, 6 Dec 2013 09:45:27 +0100 Subject: [change] we now use Full-Text RSS 3.1, thank you so much @fivefilters --- .../libraries/readability/JSLikeHTMLElement.php | 110 ++ inc/3rdparty/libraries/readability/Readability.php | 1138 ++++++++++++++++++++ 2 files changed, 1248 insertions(+) create mode 100644 inc/3rdparty/libraries/readability/JSLikeHTMLElement.php create mode 100644 inc/3rdparty/libraries/readability/Readability.php (limited to 'inc/3rdparty/libraries/readability') diff --git a/inc/3rdparty/libraries/readability/JSLikeHTMLElement.php b/inc/3rdparty/libraries/readability/JSLikeHTMLElement.php new file mode 100644 index 00000000..a8eeccf4 --- /dev/null +++ b/inc/3rdparty/libraries/readability/JSLikeHTMLElement.php @@ -0,0 +1,110 @@ +registerNodeClass('DOMElement', 'JSLikeHTMLElement'); +* $doc->loadHTML('

Para 1

Para 2

'); +* $elem = $doc->getElementsByTagName('div')->item(0); +* +* // print innerHTML +* echo $elem->innerHTML; // prints '

Para 1

Para 2

' +* echo "\n\n"; +* +* // set innerHTML +* $elem->innerHTML = 'FiveFilters.org'; +* echo $elem->innerHTML; // prints 'FiveFilters.org' +* echo "\n\n"; +* +* // print document (with our changes) +* echo $doc->saveXML(); +* @endcode +* +* @author Keyvan Minoukadeh - http://www.keyvan.net - keyvan@keyvan.net +* @see http://fivefilters.org (the project this was written for) +*/ +class JSLikeHTMLElement extends DOMElement +{ + /** + * Used for setting innerHTML like it's done in JavaScript: + * @code + * $div->innerHTML = '

Chapter 2

The story begins...

'; + * @endcode + */ + public function __set($name, $value) { + if ($name == 'innerHTML') { + // first, empty the element + for ($x=$this->childNodes->length-1; $x>=0; $x--) { + $this->removeChild($this->childNodes->item($x)); + } + // $value holds our new inner HTML + if ($value != '') { + $f = $this->ownerDocument->createDocumentFragment(); + // appendXML() expects well-formed markup (XHTML) + $result = @$f->appendXML($value); // @ to suppress PHP warnings + if ($result) { + if ($f->hasChildNodes()) $this->appendChild($f); + } else { + // $value is probably ill-formed + $f = new DOMDocument(); + $value = mb_convert_encoding($value, 'HTML-ENTITIES', 'UTF-8'); + // Using will generate a warning, but so will bad HTML + // (and by this point, bad HTML is what we've got). + // We use it (and suppress the warning) because an HTML fragment will + // be wrapped around tags which we don't really want to keep. + // Note: despite the warning, if loadHTML succeeds it will return true. + $result = @$f->loadHTML(''.$value.''); + if ($result) { + $import = $f->getElementsByTagName('htmlfragment')->item(0); + foreach ($import->childNodes as $child) { + $importedNode = $this->ownerDocument->importNode($child, true); + $this->appendChild($importedNode); + } + } else { + // oh well, we tried, we really did. :( + // this element is now empty + } + } + } + } else { + $trace = debug_backtrace(); + trigger_error('Undefined property via __set(): '.$name.' in '.$trace[0]['file'].' on line '.$trace[0]['line'], E_USER_NOTICE); + } + } + + /** + * Used for getting innerHTML like it's done in JavaScript: + * @code + * $string = $div->innerHTML; + * @endcode + */ + public function __get($name) + { + if ($name == 'innerHTML') { + $inner = ''; + foreach ($this->childNodes as $child) { + $inner .= $this->ownerDocument->saveXML($child); + } + return $inner; + } + + $trace = debug_backtrace(); + trigger_error('Undefined property via __get(): '.$name.' in '.$trace[0]['file'].' on line '.$trace[0]['line'], E_USER_NOTICE); + return null; + } + + public function __toString() + { + return '['.$this->tagName.']'; + } +} +?> \ No newline at end of file diff --git a/inc/3rdparty/libraries/readability/Readability.php b/inc/3rdparty/libraries/readability/Readability.php new file mode 100644 index 00000000..2e8991cc --- /dev/null +++ b/inc/3rdparty/libraries/readability/Readability.php @@ -0,0 +1,1138 @@ +init(); +echo $r->articleContent->innerHTML; +*/ + +class Readability +{ + public $version = '1.7.1-without-multi-page'; + public $convertLinksToFootnotes = false; + public $revertForcedParagraphElements = true; + public $articleTitle; + public $articleContent; + public $dom; + public $url = null; // optional - URL where HTML was retrieved + public $debug = false; + public $lightClean = true; // preserves more content (experimental) added 2012-09-19 + protected $body = null; // + protected $bodyCache = null; // Cache the body HTML in case we need to re-use it later + protected $flags = 7; // 1 | 2 | 4; // Start with all flags set. + protected $success = false; // indicates whether we were able to extract or not + + /** + * All of the regular expressions in use within readability. + * Defined up here so we don't instantiate them repeatedly in loops. + **/ + public $regexps = array( + 'unlikelyCandidates' => '/combx|comment|community|disqus|extra|foot|header|menu|remark|rss|shoutbox|sidebar|sponsor|ad-break|agegate|pagination|pager|popup/i', + 'okMaybeItsACandidate' => '/and|article|body|column|main|shadow/i', + 'positive' => '/article|body|content|entry|hentry|main|page|attachment|pagination|post|text|blog|story/i', + 'negative' => '/combx|comment|com-|contact|foot|footer|_nav|footnote|masthead|media|meta|outbrain|promo|related|scroll|shoutbox|sidebar|sponsor|shopping|tags|tool|widget/i', + 'divToPElements' => '/<(a|blockquote|dl|div|img|ol|p|pre|table|ul)/i', + 'replaceBrs' => '/(]*>[ \n\r\t]*){2,}/i', + 'replaceFonts' => '/<(\/?)font[^>]*>/i', + // 'trimRe' => '/^\s+|\s+$/g', // PHP has trim() + 'normalize' => '/\s{2,}/', + 'killBreaks' => '/((\s| ?)*){1,}/', + 'video' => '!//(player\.|www\.)?(youtube|vimeo|viddler)\.com!i', + 'skipFootnoteLink' => '/^\s*(\[?[a-z0-9]{1,2}\]?|^|edit|citation needed)\s*$/i' + ); + + /* constants */ + const FLAG_STRIP_UNLIKELYS = 1; + const FLAG_WEIGHT_CLASSES = 2; + const FLAG_CLEAN_CONDITIONALLY = 4; + + /** + * Create instance of Readability + * @param string UTF-8 encoded string + * @param string (optional) URL associated with HTML (used for footnotes) + * @param string which parser to use for turning raw HTML into a DOMDocument (either 'libxml' or 'html5lib') + */ + function __construct($html, $url=null, $parser='libxml') + { + $this->url = $url; + /* Turn all double br's into p's */ + $html = preg_replace($this->regexps['replaceBrs'], '

', $html); + $html = preg_replace($this->regexps['replaceFonts'], '<$1span>', $html); + $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); + if (trim($html) == '') $html = ''; + if ($parser=='html5lib' && ($this->dom = HTML5_Parser::parse($html))) { + // all good + } else { + $this->dom = new DOMDocument(); + $this->dom->preserveWhiteSpace = false; + @$this->dom->loadHTML($html); + } + $this->dom->registerNodeClass('DOMElement', 'JSLikeHTMLElement'); + } + + /** + * Get article title element + * @return DOMElement + */ + public function getTitle() { + return $this->articleTitle; + } + + /** + * Get article content element + * @return DOMElement + */ + public function getContent() { + return $this->articleContent; + } + + /** + * Runs readability. + * + * Workflow: + * 1. Prep the document by removing script tags, css, etc. + * 2. Build readability's DOM tree. + * 3. Grab the article content from the current dom tree. + * 4. Replace the current DOM tree with the new one. + * 5. Read peacefully. + * + * @return boolean true if we found content, false otherwise + **/ + public function init() + { + if (!isset($this->dom->documentElement)) return false; + $this->removeScripts($this->dom); + //die($this->getInnerHTML($this->dom->documentElement)); + + // Assume successful outcome + $this->success = true; + + $bodyElems = $this->dom->getElementsByTagName('body'); + if ($bodyElems->length > 0) { + if ($this->bodyCache == null) { + $this->bodyCache = $bodyElems->item(0)->innerHTML; + } + if ($this->body == null) { + $this->body = $bodyElems->item(0); + } + } + + $this->prepDocument(); + + //die($this->dom->documentElement->parentNode->nodeType); + //$this->setInnerHTML($this->dom->documentElement, $this->getInnerHTML($this->dom->documentElement)); + //die($this->getInnerHTML($this->dom->documentElement)); + + /* Build readability's DOM tree */ + $overlay = $this->dom->createElement('div'); + $innerDiv = $this->dom->createElement('div'); + $articleTitle = $this->getArticleTitle(); + $articleContent = $this->grabArticle(); + + if (!$articleContent) { + $this->success = false; + $articleContent = $this->dom->createElement('div'); + $articleContent->setAttribute('id', 'readability-content'); + $articleContent->innerHTML = '

Sorry, Readability was unable to parse this page for content.

'; + } + + $overlay->setAttribute('id', 'readOverlay'); + $innerDiv->setAttribute('id', 'readInner'); + + /* Glue the structure of our document together. */ + $innerDiv->appendChild($articleTitle); + $innerDiv->appendChild($articleContent); + $overlay->appendChild($innerDiv); + + /* Clear the old HTML, insert the new content. */ + $this->body->innerHTML = ''; + $this->body->appendChild($overlay); + //document.body.insertBefore(overlay, document.body.firstChild); + $this->body->removeAttribute('style'); + + $this->postProcessContent($articleContent); + + // Set title and content instance variables + $this->articleTitle = $articleTitle; + $this->articleContent = $articleContent; + + return $this->success; + } + + /** + * Debug + */ + protected function dbg($msg) { + if ($this->debug) echo '* ',$msg, "\n"; + } + + /** + * Run any post-process modifications to article content as necessary. + * + * @param DOMElement + * @return void + */ + public function postProcessContent($articleContent) { + if ($this->convertLinksToFootnotes && !preg_match('/wikipedia\.org/', @$this->url)) { + $this->addFootnotes($articleContent); + } + } + + /** + * Get the article title as an H1. + * + * @return DOMElement + */ + protected function getArticleTitle() { + $curTitle = ''; + $origTitle = ''; + + try { + $curTitle = $origTitle = $this->getInnerText($this->dom->getElementsByTagName('title')->item(0)); + } catch(Exception $e) {} + + if (preg_match('/ [\|\-] /', $curTitle)) + { + $curTitle = preg_replace('/(.*)[\|\-] .*/i', '$1', $origTitle); + + if (count(explode(' ', $curTitle)) < 3) { + $curTitle = preg_replace('/[^\|\-]*[\|\-](.*)/i', '$1', $origTitle); + } + } + else if (strpos($curTitle, ': ') !== false) + { + $curTitle = preg_replace('/.*:(.*)/i', '$1', $origTitle); + + if (count(explode(' ', $curTitle)) < 3) { + $curTitle = preg_replace('/[^:]*[:](.*)/i','$1', $origTitle); + } + } + else if(strlen($curTitle) > 150 || strlen($curTitle) < 15) + { + $hOnes = $this->dom->getElementsByTagName('h1'); + if($hOnes->length == 1) + { + $curTitle = $this->getInnerText($hOnes->item(0)); + } + } + + $curTitle = trim($curTitle); + + if (count(explode(' ', $curTitle)) <= 4) { + $curTitle = $origTitle; + } + + $articleTitle = $this->dom->createElement('h1'); + $articleTitle->innerHTML = $curTitle; + + return $articleTitle; + } + + /** + * Prepare the HTML document for readability to scrape it. + * This includes things like stripping javascript, CSS, and handling terrible markup. + * + * @return void + **/ + protected function prepDocument() { + /** + * In some cases a body element can't be found (if the HTML is totally hosed for example) + * so we create a new body node and append it to the document. + */ + if ($this->body == null) + { + $this->body = $this->dom->createElement('body'); + $this->dom->documentElement->appendChild($this->body); + } + $this->body->setAttribute('id', 'readabilityBody'); + + /* Remove all style tags in head */ + $styleTags = $this->dom->getElementsByTagName('style'); + for ($i = $styleTags->length-1; $i >= 0; $i--) + { + $styleTags->item($i)->parentNode->removeChild($styleTags->item($i)); + } + + /* Turn all double br's into p's */ + /* Note, this is pretty costly as far as processing goes. Maybe optimize later. */ + //document.body.innerHTML = document.body.innerHTML.replace(readability.regexps.replaceBrs, '

').replace(readability.regexps.replaceFonts, '<$1span>'); + // We do this in the constructor for PHP as that's when we have raw HTML - before parsing it into a DOM tree. + // Manipulating innerHTML as it's done in JS is not possible in PHP. + } + + /** + * For easier reading, convert this document to have footnotes at the bottom rather than inline links. + * @see http://www.roughtype.com/archives/2010/05/experiments_in.php + * + * @return void + **/ + public function addFootnotes($articleContent) { + $footnotesWrapper = $this->dom->createElement('div'); + $footnotesWrapper->setAttribute('id', 'readability-footnotes'); + $footnotesWrapper->innerHTML = '

References

'; + + $articleFootnotes = $this->dom->createElement('ol'); + $articleFootnotes->setAttribute('id', 'readability-footnotes-list'); + $footnotesWrapper->appendChild($articleFootnotes); + + $articleLinks = $articleContent->getElementsByTagName('a'); + + $linkCount = 0; + for ($i = 0; $i < $articleLinks->length; $i++) + { + $articleLink = $articleLinks->item($i); + $footnoteLink = $articleLink->cloneNode(true); + $refLink = $this->dom->createElement('a'); + $footnote = $this->dom->createElement('li'); + $linkDomain = @parse_url($footnoteLink->getAttribute('href'), PHP_URL_HOST); + if (!$linkDomain && isset($this->url)) $linkDomain = @parse_url($this->url, PHP_URL_HOST); + //linkDomain = footnoteLink.host ? footnoteLink.host : document.location.host, + $linkText = $this->getInnerText($articleLink); + + if ((strpos($articleLink->getAttribute('class'), 'readability-DoNotFootnote') !== false) || preg_match($this->regexps['skipFootnoteLink'], $linkText)) { + continue; + } + + $linkCount++; + + /** Add a superscript reference after the article link */ + $refLink->setAttribute('href', '#readabilityFootnoteLink-' . $linkCount); + $refLink->innerHTML = '^{[' . $linkCount . ']}'; + $refLink->setAttribute('class', 'readability-DoNotFootnote'); + $refLink->setAttribute('style', 'color: inherit;'); + + //TODO: does this work or should we use DOMNode.isSameNode()? + if ($articleLink->parentNode->lastChild == $articleLink) { + $articleLink->parentNode->appendChild($refLink); + } else { + $articleLink->parentNode->insertBefore($refLink, $articleLink->nextSibling); + } + + $articleLink->setAttribute('style', 'color: inherit; text-decoration: none;'); + $articleLink->setAttribute('name', 'readabilityLink-' . $linkCount); + + $footnote->innerHTML = '^{^} '; + + $footnoteLink->innerHTML = ($footnoteLink->getAttribute('title') != '' ? $footnoteLink->getAttribute('title') : $linkText); + $footnoteLink->setAttribute('name', 'readabilityFootnoteLink-' . $linkCount); + + $footnote->appendChild($footnoteLink); + if ($linkDomain) $footnote->innerHTML = $footnote->innerHTML . ' (' . $linkDomain . ')'; + + $articleFootnotes->appendChild($footnote); + } + + if ($linkCount > 0) { + $articleContent->appendChild($footnotesWrapper); + } + } + + /** + * Reverts P elements with class 'readability-styled' + * to text nodes - which is what they were before. + * + * @param DOMElement + * @return void + */ + function revertReadabilityStyledElements($articleContent) { + $xpath = new DOMXPath($articleContent->ownerDocument); + $elems = $xpath->query('.//p[@class="readability-styled"]', $articleContent); + //$elems = $articleContent->getElementsByTagName('p'); + for ($i = $elems->length-1; $i >= 0; $i--) { + $e = $elems->item($i); + $e->parentNode->replaceChild($articleContent->ownerDocument->createTextNode($e->textContent), $e); + //if ($e->hasAttribute('class') && $e->getAttribute('class') == 'readability-styled') { + // $e->parentNode->replaceChild($this->dom->createTextNode($e->textContent), $e); + //} + } + } + + /** + * Prepare the article node for display. Clean out any inline styles, + * iframes, forms, strip extraneous

tags, etc. + * + * @param DOMElement + * @return void + */ + function prepArticle($articleContent) { + $this->cleanStyles($articleContent); + $this->killBreaks($articleContent); + if ($this->revertForcedParagraphElements) { + $this->revertReadabilityStyledElements($articleContent); + } + + /* Clean out junk from the article content */ + $this->cleanConditionally($articleContent, 'form'); + $this->clean($articleContent, 'object'); + $this->clean($articleContent, 'h1'); + + /** + * If there is only one h2, they are probably using it + * as a header and not a subheader, so remove it since we already have a header. + ***/ + if (!$this->lightClean && ($articleContent->getElementsByTagName('h2')->length == 1)) { + $this->clean($articleContent, 'h2'); + } + $this->clean($articleContent, 'iframe'); + + $this->cleanHeaders($articleContent); + + /* Do these last as the previous stuff may have removed junk that will affect these */ + $this->cleanConditionally($articleContent, 'table'); + $this->cleanConditionally($articleContent, 'ul'); + $this->cleanConditionally($articleContent, 'div'); + + /* Remove extra paragraphs */ + $articleParagraphs = $articleContent->getElementsByTagName('p'); + for ($i = $articleParagraphs->length-1; $i >= 0; $i--) + { + $imgCount = $articleParagraphs->item($i)->getElementsByTagName('img')->length; + $embedCount = $articleParagraphs->item($i)->getElementsByTagName('embed')->length; + $objectCount = $articleParagraphs->item($i)->getElementsByTagName('object')->length; + $iframeCount = $articleParagraphs->item($i)->getElementsByTagName('iframe')->length; + + if ($imgCount === 0 && $embedCount === 0 && $objectCount === 0 && $iframeCount === 0 && $this->getInnerText($articleParagraphs->item($i), false) == '') + { + $articleParagraphs->item($i)->parentNode->removeChild($articleParagraphs->item($i)); + } + } + + try { + $articleContent->innerHTML = preg_replace('/]*>\s*

innerHTML); + //articleContent.innerHTML = articleContent.innerHTML.replace(/]*>\s*

dbg("Cleaning innerHTML of breaks failed. This is an IE strict-block-elements bug. Ignoring.: " . $e); + } + } + + /** + * Initialize a node with the readability object. Also checks the + * className/id for special names to add to its score. + * + * @param Element + * @return void + **/ + protected function initializeNode($node) { + $readability = $this->dom->createAttribute('readability'); + $readability->value = 0; // this is our contentScore + $node->setAttributeNode($readability); + + switch (strtoupper($node->tagName)) { // unsure if strtoupper is needed, but using it just in case + case 'DIV': + $readability->value += 5; + break; + + case 'PRE': + case 'TD': + case 'BLOCKQUOTE': + $readability->value += 3; + break; + + case 'ADDRESS': + case 'OL': + case 'UL': + case 'DL': + case 'DD': + case 'DT': + case 'LI': + case 'FORM': + $readability->value -= 3; + break; + + case 'H1': + case 'H2': + case 'H3': + case 'H4': + case 'H5': + case 'H6': + case 'TH': + $readability->value -= 5; + break; + } + $readability->value += $this->getClassWeight($node); + } + + /*** + * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is + * most likely to be the stuff a user wants to read. Then return it wrapped up in a div. + * + * @return DOMElement + **/ + protected function grabArticle($page=null) { + $stripUnlikelyCandidates = $this->flagIsActive(self::FLAG_STRIP_UNLIKELYS); + if (!$page) $page = $this->dom; + $allElements = $page->getElementsByTagName('*'); + /** + * First, node prepping. Trash nodes that look cruddy (like ones with the class name "comment", etc), and turn divs + * into P tags where they have been used inappropriately (as in, where they contain no other block level elements.) + * + * Note: Assignment from index for performance. See http://www.peachpit.com/articles/article.aspx?p=31567&seqNum=5 + * TODO: Shouldn't this be a reverse traversal? + **/ + $node = null; + $nodesToScore = array(); + for ($nodeIndex = 0; ($node = $allElements->item($nodeIndex)); $nodeIndex++) { + //for ($nodeIndex=$targetList->length-1; $nodeIndex >= 0; $nodeIndex--) { + //$node = $targetList->item($nodeIndex); + $tagName = strtoupper($node->tagName); + /* Remove unlikely candidates */ + if ($stripUnlikelyCandidates) { + $unlikelyMatchString = $node->getAttribute('class') . $node->getAttribute('id'); + if ( + preg_match($this->regexps['unlikelyCandidates'], $unlikelyMatchString) && + !preg_match($this->regexps['okMaybeItsACandidate'], $unlikelyMatchString) && + $tagName != 'BODY' + ) + { + $this->dbg('Removing unlikely candidate - ' . $unlikelyMatchString); + //$nodesToRemove[] = $node; + $node->parentNode->removeChild($node); + $nodeIndex--; + continue; + } + } + + if ($tagName == 'P' || $tagName == 'TD' || $tagName == 'PRE') { + $nodesToScore[] = $node; + } + + /* Turn all divs that don't have children block level elements into p's */ + if ($tagName == 'DIV') { + if (!preg_match($this->regexps['divToPElements'], $node->innerHTML)) { + //$this->dbg('Altering div to p'); + $newNode = $this->dom->createElement('p'); + try { + $newNode->innerHTML = $node->innerHTML; + //$nodesToReplace[] = array('new'=>$newNode, 'old'=>$node); + $node->parentNode->replaceChild($newNode, $node); + $nodeIndex--; + $nodesToScore[] = $node; // or $newNode? + } + catch(Exception $e) { + $this->dbg('Could not alter div to p, reverting back to div.: ' . $e); + } + } + else + { + /* EXPERIMENTAL */ + // TODO: change these p elements back to text nodes after processing + for ($i = 0, $il = $node->childNodes->length; $i < $il; $i++) { + $childNode = $node->childNodes->item($i); + if ($childNode->nodeType == 3) { // XML_TEXT_NODE + //$this->dbg('replacing text node with a p tag with the same content.'); + $p = $this->dom->createElement('p'); + $p->innerHTML = $childNode->nodeValue; + $p->setAttribute('style', 'display: inline;'); + $p->setAttribute('class', 'readability-styled'); + $childNode->parentNode->replaceChild($p, $childNode); + } + } + } + } + } + + /** + * Loop through all paragraphs, and assign a score to them based on how content-y they look. + * Then add their score to their parent node. + * + * A score is determined by things like number of commas, class names, etc. Maybe eventually link density. + **/ + $candidates = array(); + for ($pt=0; $pt < count($nodesToScore); $pt++) { + $parentNode = $nodesToScore[$pt]->parentNode; + // $grandParentNode = $parentNode ? $parentNode->parentNode : null; + $grandParentNode = !$parentNode ? null : (($parentNode->parentNode instanceof DOMElement) ? $parentNode->parentNode : null); + $innerText = $this->getInnerText($nodesToScore[$pt]); + + if (!$parentNode || !isset($parentNode->tagName)) { + continue; + } + + /* If this paragraph is less than 25 characters, don't even count it. */ + if(strlen($innerText) < 25) { + continue; + } + + /* Initialize readability data for the parent. */ + if (!$parentNode->hasAttribute('readability')) + { + $this->initializeNode($parentNode); + $candidates[] = $parentNode; + } + + /* Initialize readability data for the grandparent. */ + if ($grandParentNode && !$grandParentNode->hasAttribute('readability') && isset($grandParentNode->tagName)) + { + $this->initializeNode($grandParentNode); + $candidates[] = $grandParentNode; + } + + $contentScore = 0; + + /* Add a point for the paragraph itself as a base. */ + $contentScore++; + + /* Add points for any commas within this paragraph */ + $contentScore += count(explode(',', $innerText)); + + /* For every 100 characters in this paragraph, add another point. Up to 3 points. */ + $contentScore += min(floor(strlen($innerText) / 100), 3); + + /* Add the score to the parent. The grandparent gets half. */ + $parentNode->getAttributeNode('readability')->value += $contentScore; + + if ($grandParentNode) { + $grandParentNode->getAttributeNode('readability')->value += $contentScore/2; + } + } + + /** + * After we've calculated scores, loop through all of the possible candidate nodes we found + * and find the one with the highest score. + **/ + $topCandidate = null; + for ($c=0, $cl=count($candidates); $c < $cl; $c++) + { + /** + * Scale the final candidates score based on link density. Good content should have a + * relatively small link density (5% or less) and be mostly unaffected by this operation. + **/ + $readability = $candidates[$c]->getAttributeNode('readability'); + $readability->value = $readability->value * (1-$this->getLinkDensity($candidates[$c])); + + $this->dbg('Candidate: ' . $candidates[$c]->tagName . ' (' . $candidates[$c]->getAttribute('class') . ':' . $candidates[$c]->getAttribute('id') . ') with score ' . $readability->value); + + if (!$topCandidate || $readability->value > (int)$topCandidate->getAttribute('readability')) { + $topCandidate = $candidates[$c]; + } + } + + /** + * If we still have no top candidate, just use the body as a last resort. + * We also have to copy the body node so it is something we can modify. + **/ + if ($topCandidate === null || strtoupper($topCandidate->tagName) == 'BODY') + { + $topCandidate = $this->dom->createElement('div'); + if ($page instanceof DOMDocument) { + if (!isset($page->documentElement)) { + // we don't have a body either? what a mess! :) + } else { + $topCandidate->innerHTML = $page->documentElement->innerHTML; + $page->documentElement->innerHTML = ''; + $page->documentElement->appendChild($topCandidate); + } + } else { + $topCandidate->innerHTML = $page->innerHTML; + $page->innerHTML = ''; + $page->appendChild($topCandidate); + } + $this->initializeNode($topCandidate); + } + + /** + * Now that we have the top candidate, look through its siblings for content that might also be related. + * Things like preambles, content split by ads that we removed, etc. + **/ + $articleContent = $this->dom->createElement('div'); + $articleContent->setAttribute('id', 'readability-content'); + $siblingScoreThreshold = max(10, ((int)$topCandidate->getAttribute('readability')) * 0.2); + $siblingNodes = $topCandidate->parentNode->childNodes; + if (!isset($siblingNodes)) { + $siblingNodes = new stdClass; + $siblingNodes->length = 0; + } + + for ($s=0, $sl=$siblingNodes->length; $s < $sl; $s++) + { + $siblingNode = $siblingNodes->item($s); + $append = false; + + $this->dbg('Looking at sibling node: ' . $siblingNode->nodeName . (($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->hasAttribute('readability')) ? (' with score ' . $siblingNode->getAttribute('readability')) : '')); + + //dbg('Sibling has score ' . ($siblingNode->readability ? siblingNode.readability.contentScore : 'Unknown')); + + if ($siblingNode === $topCandidate) + // or if ($siblingNode->isSameNode($topCandidate)) + { + $append = true; + } + + $contentBonus = 0; + /* Give a bonus if sibling nodes and top candidates have the example same classname */ + if ($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->getAttribute('class') == $topCandidate->getAttribute('class') && $topCandidate->getAttribute('class') != '') { + $contentBonus += ((int)$topCandidate->getAttribute('readability')) * 0.2; + } + + if ($siblingNode->nodeType === XML_ELEMENT_NODE && $siblingNode->hasAttribute('readability') && (((int)$siblingNode->getAttribute('readability')) + $contentBonus) >= $siblingScoreThreshold) + { + $append = true; + } + + if (strtoupper($siblingNode->nodeName) == 'P') { + $linkDensity = $this->getLinkDensity($siblingNode); + $nodeContent = $this->getInnerText($siblingNode); + $nodeLength = strlen($nodeContent); + + if ($nodeLength > 80 && $linkDensity < 0.25) + { + $append = true; + } + else if ($nodeLength < 80 && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent)) + { + $append = true; + } + } + + if ($append) + { + $this->dbg('Appending node: ' . $siblingNode->nodeName); + + $nodeToAppend = null; + $sibNodeName = strtoupper($siblingNode->nodeName); + if ($sibNodeName != 'DIV' && $sibNodeName != 'P') { + /* 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. */ + + $this->dbg('Altering siblingNode of ' . $sibNodeName . ' to div.'); + $nodeToAppend = $this->dom->createElement('div'); + try { + $nodeToAppend->setAttribute('id', $siblingNode->getAttribute('id')); + $nodeToAppend->innerHTML = $siblingNode->innerHTML; + } + catch(Exception $e) + { + $this->dbg('Could not alter siblingNode to div, reverting back to original.'); + $nodeToAppend = $siblingNode; + $s--; + $sl--; + } + } else { + $nodeToAppend = $siblingNode; + $s--; + $sl--; + } + + /* To ensure a node does not interfere with readability styles, remove its classnames */ + $nodeToAppend->removeAttribute('class'); + + /* Append sibling and subtract from our list because it removes the node when you append to another node */ + $articleContent->appendChild($nodeToAppend); + } + } + + /** + * So we have all of the content that we need. Now we clean it up for presentation. + **/ + $this->prepArticle($articleContent); + + /** + * Now that we've gone through the full algorithm, check to see if we got any meaningful content. + * If we didn't, we may need to re-run grabArticle with different flags set. This gives us a higher + * likelihood of finding the content, and the sieve approach gives us a higher likelihood of + * finding the -right- content. + **/ + if (strlen($this->getInnerText($articleContent, false)) < 250) + { + // 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 + // in the meantime, we check and create an empty element if it's not there. + if (!isset($this->body->childNodes)) $this->body = $this->dom->createElement('body'); + $this->body->innerHTML = $this->bodyCache; + + if ($this->flagIsActive(self::FLAG_STRIP_UNLIKELYS)) { + $this->removeFlag(self::FLAG_STRIP_UNLIKELYS); + return $this->grabArticle($this->body); + } + else if ($this->flagIsActive(self::FLAG_WEIGHT_CLASSES)) { + $this->removeFlag(self::FLAG_WEIGHT_CLASSES); + return $this->grabArticle($this->body); + } + else if ($this->flagIsActive(self::FLAG_CLEAN_CONDITIONALLY)) { + $this->removeFlag(self::FLAG_CLEAN_CONDITIONALLY); + return $this->grabArticle($this->body); + } + else { + return false; + } + } + return $articleContent; + } + + /** + * Remove script tags from document + * + * @param DOMElement + * @return void + */ + public function removeScripts($doc) { + $scripts = $doc->getElementsByTagName('script'); + for($i = $scripts->length-1; $i >= 0; $i--) + { + $scripts->item($i)->parentNode->removeChild($scripts->item($i)); + } + } + + /** + * Get the inner text of a node. + * This also strips out any excess whitespace to be found. + * + * @param DOMElement $ + * @param boolean $normalizeSpaces (default: true) + * @return string + **/ + public function getInnerText($e, $normalizeSpaces=true) { + $textContent = ''; + + if (!isset($e->textContent) || $e->textContent == '') { + return ''; + } + + $textContent = trim($e->textContent); + + if ($normalizeSpaces) { + return preg_replace($this->regexps['normalize'], ' ', $textContent); + } else { + return $textContent; + } + } + + /** + * Get the number of times a string $s appears in the node $e. + * + * @param DOMElement $e + * @param string - what to count. Default is "," + * @return number (integer) + **/ + public function getCharCount($e, $s=',') { + return substr_count($this->getInnerText($e), $s); + } + + /** + * Remove the style attribute on every $e and under. + * + * @param DOMElement $e + * @return void + */ + public function cleanStyles($e) { + if (!is_object($e)) return; + $elems = $e->getElementsByTagName('*'); + foreach ($elems as $elem) { + $elem->removeAttribute('style'); + } + } + + /** + * Get the density of links as a percentage of the content + * This is the amount of text that is inside a link divided by the total text in the node. + * + * @param DOMElement $e + * @return number (float) + */ + public function getLinkDensity($e) { + $links = $e->getElementsByTagName('a'); + $textLength = strlen($this->getInnerText($e)); + $linkLength = 0; + for ($i=0, $il=$links->length; $i < $il; $i++) + { + $linkLength += strlen($this->getInnerText($links->item($i))); + } + if ($textLength > 0) { + return $linkLength / $textLength; + } else { + return 0; + } + } + + /** + * Get an elements class/id weight. Uses regular expressions to tell if this + * element looks good or bad. + * + * @param DOMElement $e + * @return number (Integer) + */ + public function getClassWeight($e) { + if(!$this->flagIsActive(self::FLAG_WEIGHT_CLASSES)) { + return 0; + } + + $weight = 0; + + /* Look for a special classname */ + if ($e->hasAttribute('class') && $e->getAttribute('class') != '') + { + if (preg_match($this->regexps['negative'], $e->getAttribute('class'))) { + $weight -= 25; + } + if (preg_match($this->regexps['positive'], $e->getAttribute('class'))) { + $weight += 25; + } + } + + /* Look for a special ID */ + if ($e->hasAttribute('id') && $e->getAttribute('id') != '') + { + if (preg_match($this->regexps['negative'], $e->getAttribute('id'))) { + $weight -= 25; + } + if (preg_match($this->regexps['positive'], $e->getAttribute('id'))) { + $weight += 25; + } + } + return $weight; + } + + /** + * Remove extraneous break tags from a node. + * + * @param DOMElement $node + * @return void + */ + public function killBreaks($node) { + $html = $node->innerHTML; + $html = preg_replace($this->regexps['killBreaks'], '
', $html); + $node->innerHTML = $html; + } + + /** + * Clean a node of all elements of type "tag". + * (Unless it's a youtube/vimeo video. People love movies.) + * + * Updated 2012-09-18 to preserve youtube/vimeo iframes + * + * @param DOMElement $e + * @param string $tag + * @return void + */ + public function clean($e, $tag) { + $targetList = $e->getElementsByTagName($tag); + $isEmbed = ($tag == 'iframe' || $tag == 'object' || $tag == 'embed'); + + for ($y=$targetList->length-1; $y >= 0; $y--) { + /* Allow youtube and vimeo videos through as people usually want to see those. */ + if ($isEmbed) { + $attributeValues = ''; + for ($i=0, $il=$targetList->item($y)->attributes->length; $i < $il; $i++) { + $attributeValues .= $targetList->item($y)->attributes->item($i)->value . '|'; // DOMAttr? (TODO: test) + } + + /* First, check the elements attributes to see if any of them contain youtube or vimeo */ + if (preg_match($this->regexps['video'], $attributeValues)) { + continue; + } + + /* Then check the elements inside this element for the same. */ + if (preg_match($this->regexps['video'], $targetList->item($y)->innerHTML)) { + continue; + } + } + $targetList->item($y)->parentNode->removeChild($targetList->item($y)); + } + } + + /** + * Clean an element of all tags of type "tag" if they look fishy. + * "Fishy" is an algorithm based on content length, classnames, + * link density, number of images & embeds, etc. + * + * @param DOMElement $e + * @param string $tag + * @return void + */ + public function cleanConditionally($e, $tag) { + if (!$this->flagIsActive(self::FLAG_CLEAN_CONDITIONALLY)) { + return; + } + + $tagsList = $e->getElementsByTagName($tag); + $curTagsLength = $tagsList->length; + + /** + * Gather counts for other typical elements embedded within. + * Traverse backwards so we can remove nodes at the same time without effecting the traversal. + * + * TODO: Consider taking into account original contentScore here. + */ + for ($i=$curTagsLength-1; $i >= 0; $i--) { + $weight = $this->getClassWeight($tagsList->item($i)); + $contentScore = ($tagsList->item($i)->hasAttribute('readability')) ? (int)$tagsList->item($i)->getAttribute('readability') : 0; + + $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')) : '')); + + if ($weight + $contentScore < 0) { + $tagsList->item($i)->parentNode->removeChild($tagsList->item($i)); + } + else if ( $this->getCharCount($tagsList->item($i), ',') < 10) { + /** + * If there are not very many commas, and the number of + * non-paragraph elements is more than paragraphs or other ominous signs, remove the element. + **/ + $p = $tagsList->item($i)->getElementsByTagName('p')->length; + $img = $tagsList->item($i)->getElementsByTagName('img')->length; + $li = $tagsList->item($i)->getElementsByTagName('li')->length-100; + $input = $tagsList->item($i)->getElementsByTagName('input')->length; + $a = $tagsList->item($i)->getElementsByTagName('a')->length; + + $embedCount = 0; + $embeds = $tagsList->item($i)->getElementsByTagName('embed'); + for ($ei=0, $il=$embeds->length; $ei < $il; $ei++) { + if (preg_match($this->regexps['video'], $embeds->item($ei)->getAttribute('src'))) { + $embedCount++; + } + } + $embeds = $tagsList->item($i)->getElementsByTagName('iframe'); + for ($ei=0, $il=$embeds->length; $ei < $il; $ei++) { + if (preg_match($this->regexps['video'], $embeds->item($ei)->getAttribute('src'))) { + $embedCount++; + } + } + + $linkDensity = $this->getLinkDensity($tagsList->item($i)); + $contentLength = strlen($this->getInnerText($tagsList->item($i))); + $toRemove = false; + + if ($this->lightClean) { + $this->dbg('Light clean...'); + if ( ($img > $p) && ($img > 4) ) { + $this->dbg(' more than 4 images and more image elements than paragraph elements'); + $toRemove = true; + } else if ($li > $p && $tag != 'ul' && $tag != 'ol') { + $this->dbg(' too many

elements, and parent is not

elements, and parent is not