diff options
author | Andres Rey <[email protected]> | 2017-12-03 12:24:28 +0000 |
---|---|---|
committer | GitHub <[email protected]> | 2017-12-03 12:24:28 +0000 |
commit | 0d02e2916d1659bb79426969c5d48848ff402598 (patch) | |
tree | 043f85fe97b68d24027bb067029ab6e3b7cb627d /src | |
parent | bbf6068a6440c2390956993a6845952a02219722 (diff) | |
parent | 62deed54da3c5248e5d8ac13805c0d1c0d2bf1d6 (diff) |
Merge pull request #33 from andreskrey/v1.0
v1.0
Diffstat (limited to 'src')
-rw-r--r-- | src/Configuration.php | 289 | ||||
-rw-r--r-- | src/Environment.php | 36 | ||||
-rw-r--r-- | src/HTMLParser.php | 1598 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMAttr.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMCdataSection.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMCharacterData.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMComment.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMDocument.php | 28 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMDocumentFragment.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMDocumentType.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMElement.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMNode.php | 13 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMNotation.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMProcessingInstruction.php | 10 | ||||
-rw-r--r-- | src/Nodes/DOM/DOMText.php | 10 | ||||
-rw-r--r-- | src/Nodes/NodeTrait.php | 430 | ||||
-rw-r--r-- | src/Nodes/NodeUtility.php | 159 | ||||
-rw-r--r-- | src/ParseException.php | 7 | ||||
-rw-r--r-- | src/Readability.php | 1620 | ||||
-rw-r--r-- | src/ReadabilityInterface.php | 92 |
20 files changed, 2298 insertions, 2074 deletions
diff --git a/src/Configuration.php b/src/Configuration.php index d2fa6a7..1a405de 100644 --- a/src/Configuration.php +++ b/src/Configuration.php @@ -2,59 +2,298 @@ namespace andreskrey\Readability; +/** + * Class Configuration. + */ class Configuration { - protected $config; + /** + * @var int + */ + protected $maxTopCandidates = 5; + /** + * @var int + */ + protected $wordThreshold = 500; + /** + * @var bool + */ + protected $articleByLine = false; + /** + * @var bool + */ + protected $stripUnlikelyCandidates = true; + /** + * @var bool + */ + protected $cleanConditionally = true; + /** + * @var bool + */ + protected $weightClasses = true; + /** + * @var bool + */ + protected $removeReadabilityTags = true; + /** + * @var bool + */ + protected $fixRelativeURLs = false; + /** + * @var bool + */ + protected $substituteEntities = false; + /** + * @var bool + */ + protected $normalizeEntities = false; + /** + * @var string + */ + protected $originalURL = 'http://fakehost'; + + /** + * @return int + */ + public function getMaxTopCandidates() + { + return $this->maxTopCandidates; + } /** - * @param array $config + * @param int $maxTopCandidates + * + * @return $this */ - public function __construct(array $config = []) + public function setMaxTopCandidates($maxTopCandidates) { - $this->config = $config; + $this->maxTopCandidates = $maxTopCandidates; + + return $this; } /** - * @param array $config + * @return int */ - public function merge(array $config = []) + public function getWordThreshold() { - $this->config = array_replace_recursive($this->config, $config); + return $this->wordThreshold; } /** - * @param array $config + * @param int $wordThreshold + * + * @return $this */ - public function replace(array $config = []) + public function setWordThreshold($wordThreshold) { - $this->config = $config; + $this->wordThreshold = $wordThreshold; + + return $this; } /** - * @param string $key - * @param mixed $value + * @return bool */ - public function setOption($key, $value) + public function getArticleByLine() { - $this->config[$key] = $value; + return $this->articleByLine; } /** - * @param string|null $key - * @param mixed|null $default + * @param bool $articleByLine * - * @return mixed|null + * @return $this */ - public function getOption($key = null, $default = null) + public function setArticleByLine($articleByLine) { - if ($key === null) { - return $this->config; - } + $this->articleByLine = $articleByLine; - if (!isset($this->config[$key])) { - return $default; - } + return $this; + } - return $this->config[$key]; + /** + * @return bool + */ + public function getStripUnlikelyCandidates() + { + return $this->stripUnlikelyCandidates; + } + + /** + * @param bool $stripUnlikelyCandidates + * + * @return $this + */ + public function setStripUnlikelyCandidates($stripUnlikelyCandidates) + { + $this->stripUnlikelyCandidates = $stripUnlikelyCandidates; + + return $this; + } + + /** + * @return bool + */ + public function getCleanConditionally() + { + return $this->cleanConditionally; + } + + /** + * @param bool $cleanConditionally + * + * @return $this + */ + public function setCleanConditionally($cleanConditionally) + { + $this->cleanConditionally = $cleanConditionally; + + return $this; + } + + /** + * @return bool + */ + public function getWeightClasses() + { + return $this->weightClasses; + } + + /** + * @param bool $weightClasses + * + * @return $this + */ + public function setWeightClasses($weightClasses) + { + $this->weightClasses = $weightClasses; + + return $this; + } + + /** + * @return bool + */ + public function getRemoveReadabilityTags() + { + return $this->removeReadabilityTags; } + + /** + * @param bool $removeReadabilityTags + * + * @return $this + */ + public function setRemoveReadabilityTags($removeReadabilityTags) + { + $this->removeReadabilityTags = $removeReadabilityTags; + + return $this; + } + + /** + * @return bool + */ + public function getFixRelativeURLs() + { + return $this->fixRelativeURLs; + } + + /** + * @param bool $fixRelativeURLs + * + * @return $this + */ + public function setFixRelativeURLs($fixRelativeURLs) + { + $this->fixRelativeURLs = $fixRelativeURLs; + + return $this; + } + + /** + * @return bool + */ + public function getSubstituteEntities() + { + return $this->substituteEntities; + } + + /** + * @param bool $substituteEntities + * + * @return $this + */ + public function setSubstituteEntities($substituteEntities) + { + $this->substituteEntities = $substituteEntities; + + return $this; + } + + /** + * @return bool + */ + public function getNormalizeEntities() + { + return $this->normalizeEntities; + } + + /** + * @param bool $normalizeEntities + * + * @return $this + */ + public function setNormalizeEntities($normalizeEntities) + { + $this->normalizeEntities = $normalizeEntities; + + return $this; + } + + /** + * @return string + */ + public function getOriginalURL() + { + return $this->originalURL; + } + + /** + * @param string $originalURL + * + * @return $this + */ + public function setOriginalURL($originalURL) + { + $this->originalURL = $originalURL; + + return $this; + } + + /** + * @return bool + */ + public function getSummonCthulhu() + { + return $this->summonCthulhu; + } + + /** + * @param bool $summonCthulhu + * + * @return $this + */ + public function setSummonCthulhu($summonCthulhu) + { + $this->summonCthulhu = $summonCthulhu; + + return $this; + } + + /** + * @var bool + */ + protected $summonCthulhu = false; } diff --git a/src/Environment.php b/src/Environment.php deleted file mode 100644 index 6e88783..0000000 --- a/src/Environment.php +++ /dev/null @@ -1,36 +0,0 @@ -<?php - -namespace andreskrey\Readability; - -final class Environment -{ - /** - * @var Configuration - */ - protected $config; - - public function __construct(array $config = []) - { - $this->config = new Configuration($config); - } - - /** - * @return Configuration - */ - public function getConfig() - { - return $this->config; - } - - /** - * @param array $config - * - * @return Environment - */ - public static function createDefaultEnvironment(array $config = []) - { - $environment = new static($config); - - return $environment; - } -} diff --git a/src/HTMLParser.php b/src/HTMLParser.php deleted file mode 100644 index e4ed4bb..0000000 --- a/src/HTMLParser.php +++ /dev/null @@ -1,1598 +0,0 @@ -<?php - -namespace andreskrey\Readability; - -use DOMDocument; - -/** - * Class HTMLParser. - * - * A helper class to parse HTML and get a Readability object. - */ -class HTMLParser -{ - /** - * @var DOMDocument - */ - private $dom = null; - - /** - * TODO Make this an object? Instead of a dumb array. - * - * @var array - */ - private $metadata = []; - - /** - * @var array - */ - private $regexps = [ - 'unlikelyCandidates' => '/banner|breadcrumbs|combx|comment|community|cover-wrap|disqus|extra|foot|header|legends|menu|modal|related|remark|replies|rss|shoutbox|sidebar|skyscraper|social|sponsor|supplemental|ad-break|agegate|pagination|pager|popup|yom-remote/i', - 'okMaybeItsACandidate' => '/and|article|body|column|main|shadow/i', - 'extraneous' => '/print|archive|comment|discuss|e[\-]?mail|share|reply|all|login|sign|single|utility/i', - 'byline' => '/byline|author|dateline|writtenby|p-author/i', - 'replaceFonts' => '/<(\/?)font[^>]*>/gi', - 'normalize' => '/\s{2,}/', - 'videos' => '/\/\/(www\.)?(dailymotion|youtube|youtube-nocookie|player\.vimeo)\.com/i', - 'nextLink' => '/(next|weiter|continue|>([^\|]|$)|»([^\|]|$))/i', - 'prevLink' => '/(prev|earl|old|new|<|«)/i', - 'whitespace' => '/^\s*$/', - 'hasContent' => '/\S$/', - // \x{00A0} is the unicode version of - 'onlyWhitespace' => '/\x{00A0}|\s+/u' - ]; - - private $defaultTagsToScore = [ - 'section', - 'h2', - 'h3', - 'h4', - 'h5', - 'h6', - 'p', - 'td', - 'pre', - ]; - - /** - * @var array - */ - private $alterToDIVExceptions = [ - 'div', - 'article', - 'section', - 'p', - // TODO, check if this is correct, #text elements do not exist in js - '#text', - ]; - - /** - * @var array - */ - private $divToPElements = [ - 'a', - 'blockquote', - 'dl', - 'div', - 'img', - 'ol', - 'p', - 'pre', - 'table', - 'ul', - 'select', - ]; - - /** - * Constructor. - * - * @param array $options Options to override the default ones - */ - public function __construct(array $options = []) - { - $defaults = [ - 'maxTopCandidates' => 5, - 'wordThreshold' => 500, - 'articleByLine' => false, - 'stripUnlikelyCandidates' => true, - 'cleanConditionally' => true, - 'weightClasses' => true, - 'removeReadabilityTags' => true, - 'fixRelativeURLs' => false, - 'substituteEntities' => true, - 'normalizeEntities' => false, - 'summonCthulhu' => false, - 'originalURL' => 'http://fakehost', - ]; - - $this->environment = Environment::createDefaultEnvironment($defaults); - - $this->environment->getConfig()->merge($options); - - // To avoid having a gazillion of errors on malformed HTMLs - libxml_use_internal_errors(true); - } - - /** - * Parse the html. This is the main entry point of the HTMLParser. - * - * @param string $html Full html of the website, page, etc. - * - * #return ? TBD - */ - public function parse($html) - { - $this->dom = $this->loadHTML($html); - - $this->metadata = $this->getMetadata(); - - $this->metadata['image'] = $this->getMainImage(); - - // Checking for minimum HTML to work with. - if (!($root = $this->dom->getElementsByTagName('body')->item(0)) || !$root->firstChild) { - return false; - } - - $parseSuccessful = true; - while (true) { - $root = new Readability($root->firstChild); - - $elementsToScore = $this->getNodes($root); - - $result = $this->rateNodes($elementsToScore); - - /* - * 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. - */ - - // TODO Better way to count resulting text. Textcontent usually has alt titles and that stuff - // that doesn't really count to the quality of the result. - $length = 0; - foreach ($result->getElementsByTagName('p') as $p) { - $length += mb_strlen($p->textContent); - } - if ($result && mb_strlen(preg_replace('/\s/', '', $result->textContent)) < $this->getConfig()->getOption('wordThreshold')) { - $this->dom = $this->loadHTML($html); - $root = $this->dom->getElementsByTagName('body')->item(0); - - if ($this->getConfig()->getOption('stripUnlikelyCandidates')) { - $this->getConfig()->setOption('stripUnlikelyCandidates', false); - } elseif ($this->getConfig()->getOption('weightClasses')) { - $this->getConfig()->setOption('weightClasses', false); - } elseif ($this->getConfig()->getOption('cleanConditionally')) { - $this->getConfig()->setOption('cleanConditionally', false); - } else { - $parseSuccessful = false; - break; - } - } else { - break; - } - } - - if (!$parseSuccessful) { - return false; - } - - $result = $this->postProcessContent($result); - - // Todo, fix return, check for values, maybe create a function to create the return object - return [ - 'title' => isset($this->metadata['title']) ? $this->metadata['title'] : null, - 'author' => isset($this->metadata['author']) ? $this->metadata['author'] : null, - 'image' => isset($this->metadata['image']) ? $this->metadata['image'] : null, - 'images' => $this->getImages(), - 'article' => $result, - 'html' => $result->C14N(), - 'dir' => isset($this->metadata['articleDir']) ? $this->metadata['articleDir'] : null, - ]; - } - - /** - * Creates a DOM Document object and loads the provided HTML on it. - * - * Used for the first load of Readability and subsequent reloads (when disabling flags and rescanning the text) - * Previous versions of Readability used this method one time and cloned the DOM to keep a backup. This caused bugs - * because cloning the DOM object keeps a relation between the clone and the original one, doing changes in both - * objects and ruining the backup. - * - * @param string $html - * - * @return DOMDocument - */ - private function loadHTML($html) - { - $dom = new DOMDocument('1.0', 'utf-8'); - - if (!$this->getConfig()->getOption('substituteEntities')) { - // Keep the original HTML entities - $dom->substituteEntities = false; - } - - if ($this->getConfig()->getOption('normalizeEntities')) { - // Replace UTF-8 characters with the HTML Entity equivalent. Useful to fix html with mixed content - $html = mb_convert_encoding($html, 'HTML-ENTITIES', 'UTF-8'); - } - - if ($this->getConfig()->getOption('summonCthulhu')) { - $html = preg_replace('/<script\b[^>]*>([\s\S]*?)<\/script>/', '', $html); - } - - // Prepend the XML tag to avoid having issues with special characters. Should be harmless. - $dom->loadHTML('<?xml encoding="UTF-8">' . $html); - $dom->encoding = 'UTF-8'; - - $this->removeScripts($dom); - - $this->prepDocument($dom); - - return $dom; - } - - /** - * @return Configuration - */ - public function getConfig() - { - return $this->environment->getConfig(); - } - - /** - * Removes all the scripts of the html. - * - * @param DOMDocument $dom - */ - private function removeScripts(DOMDocument $dom) - { - $toRemove = ['script', 'noscript']; - - foreach ($toRemove as $tag) { - while ($script = $dom->getElementsByTagName($tag)) { - if ($script->item(0)) { - $script->item(0)->parentNode->removeChild($script->item(0)); - } else { - break; - } - } - } - } - - /** - * Prepares the document for parsing. - * - * @param DOMDocument $dom - */ - private function prepDocument(DOMDocument $dom) - { - /* - * DOMNodeList must be converted to an array before looping over it. - * This is done to avoid node shifting when removing nodes. - * - * Reverse traversing cannot be done here because we need to find brs that are right next to other brs. - * (If we go the other way around we need to search for previous nodes forcing the creation of new functions - * that will be used only here) - */ - foreach (iterator_to_array($dom->getElementsByTagName('br')) as $br) { - $next = $br->nextSibling; - - /* - * Whether 2 or more <br> elements have been found and replaced with a - * <p> block. - */ - $replaced = false; - - /* - * If we find a <br> chain, remove the <br>s until we hit another element - * or non-whitespace. This leaves behind the first <br> in the chain - * (which will be replaced with a <p> later). - */ - while (($next = $this->nextElement($next)) && ($next->nodeName === 'br')) { - $replaced = true; - $brSibling = $next->nextSibling; - $next->parentNode->removeChild($next); - $next = $brSibling; - } - - /* - * If we removed a <br> chain, replace the remaining <br> with a <p>. Add - * all sibling nodes as children of the <p> until we hit another <br> - * chain. - */ - - if ($replaced) { - $p = $dom->createElement('p'); - $br->parentNode->replaceChild($p, $br); - - $next = $p->nextSibling; - while ($next) { - // If we've hit another <br><br>, we're done adding children to this <p>. - if ($next->nodeName === 'br') { - $nextElem = $this->nextElement($next); - if ($nextElem && $nextElem->nodeName === 'br') { - break; - } - } - - // Otherwise, make this node a child of the new <p>. - $sibling = $next->nextSibling; - $p->appendChild($next); - $next = $sibling; - } - } - } - - // Replace font tags with span - $fonts = $dom->getElementsByTagName('font'); - $length = $fonts->length; - for ($i = 0; $i < $length; $i++) { - $font = $fonts->item($length - 1 - $i); - $span = new Readability($font); - $span->setNodeTag('span', true); - } - } - - public function postProcessContent(DOMDocument $article) - { - // Readability cannot open relative uris so we convert them to absolute uris. - if ($this->getConfig()->getOption('fixRelativeURLs')) { - foreach (iterator_to_array($article->getElementsByTagName('a')) as $link) { - /** @var \DOMElement $link */ - $href = $link->getAttribute('href'); - if ($href) { - // Replace links with javascript: URIs with text content, since - // they won't work after scripts have been removed from the page. - if (strpos($href, 'javascript:') === 0) { - $text = $article->createTextNode($link->textContent); - $link->parentNode->replaceChild($text, $link); - } else { - $link->setAttribute('href', $this->toAbsoluteURI($href)); - } - } - } - - foreach ($article->getElementsByTagName('img') as $img) { - /** @var \DOMElement $img */ - $src = $img->getAttribute('src'); - if ($src) { - $img->setAttribute('src', $this->toAbsoluteURI($src)); - } - } - } - - return $article; - } - - private function toAbsoluteURI($uri) - { - list($pathBase, $scheme, $prePath) = $this->getPathInfo($this->getConfig()->getOption('originalURL')); - - // If this is already an absolute URI, return it. - if (preg_match('/^[a-zA-Z][a-zA-Z0-9\+\-\.]*:/', $uri)) { - return $uri; - } - - // Scheme-rooted relative URI. - if (substr($uri, 0, 2) === '//') { - return $scheme . '://' . substr($uri, 2); - } - - // Prepath-rooted relative URI. - if (substr($uri, 0, 1) === '/') { - return $prePath . $uri; - } - - // Dotslash relative URI. - if (strpos($uri, './') === 0) { - return $pathBase . substr($uri, 2); - } - // Ignore hash URIs: - if (substr($uri, 0, 1) === '#') { - return $uri; - } - - // Standard relative URI; add entire path. pathBase already includes a - // trailing "/". - return $pathBase . $uri; - } - - /** - * @param string $url - * - * @return array [$pathBase, $scheme, $prePath] - */ - public function getPathInfo($url) - { - $pathBase = parse_url($url, PHP_URL_SCHEME) . '://' . parse_url($url, PHP_URL_HOST) . dirname(parse_url($url, PHP_URL_PATH)) . '/'; - $scheme = parse_url($pathBase, PHP_URL_SCHEME); - $prePath = $scheme . '://' . parse_url($pathBase, PHP_URL_HOST); - - return [$pathBase, $scheme, $prePath]; - } - - private function nextElement($node) - { - $next = $node; - while ($next - && $next->nodeName !== '#text' - && trim($next->textContent)) { - $next = $next->nextSibling; - } - - return $next; - } - - /** - * Tries to guess relevant info from metadata of the html. - * - * @return array Metadata info. May have title, excerpt and or byline. - */ - private function getMetadata() - { - $metadata = $values = []; - // Match "description", or Twitter's "twitter:description" (Cards) - // in name attribute. - $namePattern = '/^\s*((twitter)\s*:\s*)?(description|title|image)\s*$/i'; - - // Match Facebook's Open Graph title & description properties. - $propertyPattern = '/^\s*og\s*:\s*(description|title|image)\s*$/i'; - - foreach ($this->dom->getElementsByTagName('meta') as $meta) { - /* @var Readability $meta */ - $elementName = $meta->getAttribute('name'); - $elementProperty = $meta->getAttribute('property'); - - if (in_array('author', [$elementName, $elementProperty])) { - $metadata['byline'] = $meta->getAttribute('content'); - continue; - } - - $name = null; - if (preg_match($namePattern, $elementName)) { - $name = $elementName; - } elseif (preg_match($propertyPattern, $elementProperty)) { - $name = $elementProperty; - } - - if ($name) { - $content = $meta->getAttribute('content'); - if ($content) { - // Convert to lowercase and remove any whitespace - // so we can match below. - $name = preg_replace('/\s/', '', strtolower($name)); - $values[$name] = trim($content); - } - } - } - if (array_key_exists('description', $values)) { - $metadata['excerpt'] = $values['description']; - } elseif (array_key_exists('og:description', $values)) { - // Use facebook open graph description. - $metadata['excerpt'] = $values['og:description']; - } elseif (array_key_exists('twitter:description', $values)) { - // Use twitter cards description. - $metadata['excerpt'] = $values['twitter:description']; - } - - $metadata['title'] = $this->getTitle(); - - if (!$metadata['title']) { - if (array_key_exists('og:title', $values)) { - // Use facebook open graph title. - $metadata['title'] = $values['og:title']; - } elseif (array_key_exists('twitter:title', $values)) { - // Use twitter cards title. - $metadata['title'] = $values['twitter:title']; - } - } - - if (array_key_exists('og:image', $values) || array_key_exists('twitter:image', $values)) { - $metadata['image'] = array_key_exists('og:image', $values) ? $values['og:image'] : $values['twitter:image']; - } else { - $metadata['image'] = null; - } - - return $metadata; - } - - /** - * Tries to get the main article image. Will only update the metadata if the getMetadata function couldn't - * find a correct image. - * - * @return bool|string URL of the top image or false if unsuccessful. - */ - public function getMainImage() - { - $imgUrl = false; - - if ($this->metadata['image'] !== null) { - $imgUrl = $this->metadata['image']; - } - - if (!$imgUrl) { - foreach ($this->dom->getElementsByTagName('link') as $link) { - /** @var \DOMElement $link */ - /* - * Check for the rel attribute, then check if the rel attribute is either img_src or image_src, and - * finally check for the existence of the href attribute, which should hold the image url. - */ - if ($link->hasAttribute('rel') && ($link->getAttribute('rel') === 'img_src' || $link->getAttribute('rel') === 'image_src') && $link->hasAttribute('href')) { - $imgUrl = $link->getAttribute('href'); - break; - } - } - } - - if (!empty($imgUrl) && $this->getConfig()->getOption('fixRelativeURLs')) { - $imgUrl = $this->toAbsoluteURI($imgUrl); - } - - return $imgUrl; - } - - /** - * @return array - */ - public function getImages() - { - $result = []; - if (!empty($this->metadata['image'])) { - $result[] = $this->metadata['image']; - } - if (null == $this->dom) { - return $result; - } - - foreach ($this->dom->getElementsByTagName('img') as $img) { - if ($src = $img->getAttribute('src')) { - $result[] = $src; - } - } - - if ($this->getConfig()->getOption('fixRelativeURLs')) { - foreach ($result as &$imgSrc) { - $imgSrc = $this->toAbsoluteURI($imgSrc); - } - } - - $result = array_unique(array_filter($result)); - - return $result; - } - - /** - * 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 Readability $readability - * - * @return int - */ - public function getLinkDensity($readability) - { - $linkLength = 0; - $textLength = mb_strlen($readability->getTextContent(true)); - - if (!$textLength) { - return 0; - } - - $links = $readability->getAllLinks(); - - if ($links) { - /** @var Readability $link */ - foreach ($links as $link) { - $linkLength += mb_strlen($link->getTextContent(true)); - } - } - - return $linkLength / $textLength; - } - - /** - * Returns the title of the html. Prioritizes the title from the metadata against the title tag. - * - * @return string|null - */ - private function getTitle() - { - $originalTitle = null; - - if (isset($this->metadata['title'])) { - $originalTitle = $this->metadata['title']; - } else { - $titleTag = $this->dom->getElementsByTagName('title'); - if ($titleTag->length > 0) { - $originalTitle = $titleTag->item(0)->nodeValue; - } - } - - if ($originalTitle === null) { - return null; - } - - $curTitle = $originalTitle; - $titleHadHierarchicalSeparators = false; - - /* - * If there's a separator in the title, first remove the final part - * - * Sanity warning: if you eval this match in PHPStorm's "Evaluate expression" box, it will return false - * I can assure you it works properly if you let the code run. - */ - if (preg_match('/ [\|\-\\\\\/>»] /i', $curTitle)) { - $titleHadHierarchicalSeparators = (bool)preg_match('/ [\\\\\/>»] /', $curTitle); - $curTitle = preg_replace('/(.*)[\|\-\\\\\/>»] .*/i', '$1', $originalTitle); - - // If the resulting title is too short (3 words or fewer), remove - // the first part instead: - if (count(preg_split('/\s+/', $curTitle)) < 3) { - $curTitle = preg_replace('/[^\|\-\\\\\/>»]*[\|\-\\\\\/>»](.*)/i', '$1', $originalTitle); - } - } elseif (strpos($curTitle, ': ') !== false) { - // Check if we have an heading containing this exact string, so we - // could assume it's the full title. - $match = false; - for ($i = 1; $i <= 2; $i++) { - foreach ($this->dom->getElementsByTagName('h' . $i) as $hTag) { - // Trim texts to avoid having false negatives when the title is surrounded by spaces or tabs - if (trim($hTag->nodeValue) === trim($curTitle)) { - $match = true; - } - } - } - - // If we don't, let's extract the title out of the original title string. - if (!$match) { - $curTitle = substr($originalTitle, strrpos($originalTitle, ':') + 1); - - // If the title is now too short, try the first colon instead: - if (count(preg_split('/\s+/', $curTitle)) < 3) { - $curTitle = substr($originalTitle, strpos($originalTitle, ':') + 1); - } - } - } elseif (mb_strlen($curTitle) > 150 || mb_strlen($curTitle) < 15) { - $hOnes = $this->dom->getElementsByTagName('h1'); - - if ($hOnes->length === 1) { - $curTitle = $hOnes->item(0)->nodeValue; - } - } - - $curTitle = trim($curTitle); - - /* - * If we now have 4 words or fewer as our title, and either no - * 'hierarchical' separators (\, /, > or ») were found in the original - * title or we decreased the number of words by more than 1 word, use - * the original title. - */ - $curTitleWordCount = count(preg_split('/\s+/', $curTitle)); - $originalTitleWordCount = count(preg_split('/\s+/', preg_replace('/[\|\-\\\\\/>»]+/', '', $originalTitle))) - 1; - - if ($curTitleWordCount <= 4 && - (!$titleHadHierarchicalSeparators || $curTitleWordCount !== $originalTitleWordCount)) { - $curTitle = $originalTitle; - } - - return $curTitle; - } - - /** - * Gets nodes from the root element. - * - * @param $node Readability - * - * @return array - */ - private function getNodes(Readability $node) - { - $stripUnlikelyCandidates = $this->getConfig()->getOption('stripUnlikelyCandidates'); - - $elementsToScore = []; - - /* - * First, node prepping. Trash nodes that look cruddy (like ones with the - * class name "comment", etc), and turn divs into P tags where they have been - * used inappropriately (as in, where they contain no other block level elements.) - */ - - while ($node) { - $matchString = $node->getAttribute('class') . ' ' . $node->getAttribute('id'); - - // Remove DOMComments nodes as we don't need them and mess up children counting - if ($node->nodeTypeEqualsTo(XML_COMMENT_NODE)) { - $node = $node->removeAndGetNext($node); - continue; - } - - // Check to see if this node is a byline, and remove it if it is. - if ($this->checkByline($node, $matchString)) { - $node = $node->removeAndGetNext($node); - continue; - } - - // Remove unlikely candidates - if ($stripUnlikelyCandidates) { - if ( - preg_match($this->regexps['unlikelyCandidates'], $matchString) && - !preg_match($this->regexps['okMaybeItsACandidate'], $matchString) && - !$node->tagNameEqualsTo('body') && - !$node->tagNameEqualsTo('a') - ) { - $node = $node->removeAndGetNext($node); - continue; - } - } - - // Remove DIV, SECTION, and HEADER nodes without any content(e.g. text, image, video, or iframe). - if (($node->tagNameEqualsTo('div') || $node->tagNameEqualsTo('section') || $node->tagNameEqualsTo('header') || - $node->tagNameEqualsTo('h1') || $node->tagNameEqualsTo('h2') || $node->tagNameEqualsTo('h3') || - $node->tagNameEqualsTo('h4') || $node->tagNameEqualsTo('h5') || $node->tagNameEqualsTo('h6') || - $node->tagNameEqualsTo('p')) && - $node->isElementWithoutContent()) { - $node = $node->removeAndGetNext($node); - continue; - } - - if (in_array(strtolower($node->getTagName()), $this->defaultTagsToScore)) { - $elementsToScore[] = $node; - } - - // Turn all divs that don't have children block level elements into p's - if ($node->tagNameEqualsTo('div')) { - /* - * Sites like http://mobile.slate.com encloses each paragraph with a DIV - * element. DIVs with only a P element inside and no text content can be - * safely converted into plain P elements to avoid confusing the scoring - * algorithm with DIVs with are, in practice, paragraphs. - */ - if ($this->hasSinglePNode($node)) { - $pNode = $node->getChildren(true)[0]; - $node->replaceChild($pNode); - $node = $pNode; - $elementsToScore[] = $node; - } elseif (!$this->hasSingleChildBlockElement($node)) { - $node->setNodeTag('p'); - $elementsToScore[] = $node; - } else { - // EXPERIMENTAL - foreach ($node->getChildren() as $child) { - /** @var Readability $child */ - if ($child->isText() && mb_strlen(trim($child->getTextContent())) > 0) { - $newNode = $node->createNode($child, 'p'); - $child->replaceChild($newNode); - } - } - } - } - - $node = $node->getNextNode($node); - } - - return $elementsToScore; - } - - /** - * Assign scores to each node. This function will rate each node and return a Readability object for each one. - * - * @param array $nodes - * - * @return DOMDocument|bool - */ - private function rateNodes($nodes) - { - $candidates = []; - - /** @var Readability $node */ - foreach ($nodes as $node) { - if (!$node->getParent()) { - continue; - } - // Discard nodes with less than 25 characters, without blank space - if (mb_strlen($node->getTextContent(true)) < 25) { - continue; - } - - $ancestors = $node->getNodeAncestors(); - - // Exclude nodes with no ancestor - if (count($ancestors) === 0) { - continue; - } - - // Start with a point for the paragraph itself as a base. - $contentScore = 1; - - // Add points for any commas within this paragraph. - $contentScore += count(explode(',', $node->getTextContent(true))); - - // For every 100 characters in this paragraph, add another point. Up to 3 points. - $contentScore += min(floor(mb_strlen($node->getTextContent(true)) / 100), 3); - - // Initialize and score ancestors. - /** @var Readability $ancestor */ - foreach ($ancestors as $level => $ancestor) { - if (!$ancestor->isInitialized()) { - $ancestor->initializeNode(); - $candidates[] = $ancestor; - } - - /* - * Node score divider: - * - parent: 1 (no division) - * - grandparent: 2 - * - great grandparent+: ancestor level * 3 - */ - - if ($level === 0) { - $scoreDivider = 1; - } elseif ($level === 1) { - $scoreDivider = 2; - } else { - $scoreDivider = $level * 3; - } - - $currentScore = $ancestor->getContentScore(); - $ancestor->setContentScore($currentScore + ($contentScore / $scoreDivider)); - } - } - - /* - * TODO This is an horrible hack because I don't know how to properly pass by reference. - * When candidates are added to the $candidates array, they lose the reference to the original object - * and on each loop, the object inside $candidates doesn't get updated. This function restores the score - * by getting it of the data-readability tag. This should be fixed using proper references and good coding - * practices (which I lack) - */ - - foreach ($candidates as $candidate) { - $candidate->reloadScore(); - } - - /* - * After we've calculated scores, loop through all of the possible - * candidate nodes we found and find the one with the highest score. - */ - - $topCandidates = []; - foreach ($candidates as $candidate) { - - /* - * 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. - */ - - $candidate->setContentScore($candidate->getContentScore() * (1 - $this->getLinkDensity($candidate))); - - for ($i = 0; $i < $this->getConfig()->getOption('maxTopCandidates'); $i++) { - $aTopCandidate = isset($topCandidates[$i]) ? $topCandidates[$i] : null; - - if (!$aTopCandidate || $candidate->getContentScore() > $aTopCandidate->getContentScore()) { - array_splice($topCandidates, $i, 0, [$candidate]); - if (count($topCandidates) > $this->getConfig()->getOption('maxTopCandidates')) { - array_pop($topCandidates); - } - break; - } - } - } - - $topCandidate = isset($topCandidates[0]) ? $topCandidates[0] : null; - $neededToCreateTopCandidate = false; - $parentOfTopCandidate = null; - - /* - * If we still have no top candidate, just use the body as a last resort. - * We also have to copy the body node so it is something we can modify. - */ - - if ($topCandidate === null || $topCandidate->tagNameEqualsTo('body')) { - // Move all of the page's children into topCandidate - $topCandidate = new DOMDocument('1.0', 'utf-8'); - $topCandidate->encoding = 'UTF-8'; - $topCandidate->appendChild($topCandidate->createElement('div', '')); - $kids = $this->dom->getElementsByTagName('body')->item(0)->childNodes; - - // Cannot be foreached, don't ask me why. - for ($i = 0; $i < $kids->length; $i++) { - $import = $topCandidate->importNode($kids->item($i), true); - $topCandidate->firstChild->appendChild($import); - } - - // Readability must be created using firstChild to grab the DOMElement instead of the DOMDocument. - $topCandidate = new Readability($topCandidate->firstChild); - $topCandidate->initializeNode(); - - //TODO on the original code, $topCandidate is added to the page variable, which holds the whole HTML - // Should be done this here also? (line 823 in readability.js) - } elseif ($topCandidate) { - // Find a better top candidate node if it contains (at least three) nodes which belong to `topCandidates` array - // and whose scores are quite closed with current `topCandidate` node. - $alternativeCandidateAncestors = []; - for ($i = 1; $i < count($topCandidates); $i++) { - if ($topCandidates[$i]->getContentScore() / $topCandidate->getContentScore() >= 0.75) { - array_push($alternativeCandidateAncestors, $topCandidates[$i]->getNodeAncestors(false)); - } - } - - $MINIMUM_TOPCANDIDATES = 3; - if (count($alternativeCandidateAncestors) >= $MINIMUM_TOPCANDIDATES) { - $parentOfTopCandidate = $topCandidate->getParent(); - while (!$parentOfTopCandidate->tagNameEqualsTo('body')) { - $listsContainingThisAncestor = 0; - for ($ancestorIndex = 0; $ancestorIndex < count($alternativeCandidateAncestors) && $listsContainingThisAncestor < $MINIMUM_TOPCANDIDATES; $ancestorIndex++) { - $listsContainingThisAncestor += (int)in_array($parentOfTopCandidate, $alternativeCandidateAncestors[$ancestorIndex]); - } - if ($listsContainingThisAncestor >= $MINIMUM_TOPCANDIDATES) { - $topCandidate = $parentOfTopCandidate; - break; - } - $parentOfTopCandidate = $parentOfTopCandidate->getParent(); - } - } - - /* - * Because of our bonus system, parents of candidates might have scores - * themselves. They get half of the node. There won't be nodes with higher - * scores than our topCandidate, but if we see the score going *up* in the first - * few steps up the tree, that's a decent sign that there might be more content - * lurking in other places that we want to unify in. The sibling stuff - * below does some of that - but only if we've looked high enough up the DOM - * tree. - */ - - $parentOfTopCandidate = $topCandidate->getParent(); - $lastScore = $topCandidate->getContentScore(); - - // The scores shouldn't get too low. - $scoreThreshold = $lastScore / 3; - - /* @var Readability $parentOfTopCandidate */ - while (!$parentOfTopCandidate->tagNameEqualsTo('body')) { - $parentScore = $parentOfTopCandidate->getContentScore(); - if ($parentScore < $scoreThreshold) { - break; - } - - if ($parentScore > $lastScore) { - // Alright! We found a better parent to use. - $topCandidate = $parentOfTopCandidate; - break; - } - $lastScore = $parentOfTopCandidate->getContentScore(); - $parentOfTopCandidate = $parentOfTopCandidate->getParent(); - } - - // If the top candidate is the only child, use parent instead. This will help sibling - // joining logic when adjacent content is actually located in parent's sibling node. - $parentOfTopCandidate = $topCandidate->getParent(); - while (!$parentOfTopCandidate->tagNameEqualsTo('body') && count($parentOfTopCandidate->getChildren(true)) === 1) { - $topCandidate = $parentOfTopCandidate; - $parentOfTopCandidate = $topCandidate->getParent(); - } - } - - /* - * 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 = new DOMDocument('1.0', 'utf-8'); - $articleContent->createElement('div'); - - $siblingScoreThreshold = max(10, $topCandidate->getContentScore() * 0.2); - // Keep potential top candidate's parent node to try to get text direction of it later. - $parentOfTopCandidate = $topCandidate->getParent(); - $siblings = $parentOfTopCandidate->getChildren(); - - $hasContent = false; - - /** @var Readability $sibling */ - foreach ($siblings as $sibling) { - $append = false; - - if ($sibling->compareNodes($sibling, $topCandidate)) { - $append = true; - } else { - $contentBonus = 0; - - // Give a bonus if sibling nodes and top candidates have the example same classname - if ($sibling->getAttribute('class') === $topCandidate->getAttribute('class') && $topCandidate->getAttribute('class') !== '') { - $contentBonus += $topCandidate->getContentScore() * 0.2; - } - if ($sibling->getContentScore() + $contentBonus >= $siblingScoreThreshold) { - $append = true; - } elseif ($sibling->tagNameEqualsTo('p')) { - $linkDensity = $this->getLinkDensity($sibling); - $nodeContent = $sibling->getTextContent(true); - - if (mb_strlen($nodeContent) > 80 && $linkDensity < 0.25) { - $append = true; - } elseif ($nodeContent && mb_strlen($nodeContent) < 80 && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent)) { - $append = true; - } - } - } - - if ($append) { - $hasContent = true; - - if (!in_array(strtolower($sibling->getTagName()), $this->alterToDIVExceptions)) { - /* - * 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. - */ - - $sibling->setNodeTag('div'); - } - - $import = $articleContent->importNode($sibling->getDOMNode(), true); - $articleContent->appendChild($import); - - /* - * No node shifting needs to be check because when calling getChildren, an array is made with the - * children of the parent node, instead of using the DOMElement childNodes function, which, when used - * along with appendChild, would shift the nodes position and the current foreach will behave in - * unpredictable ways. - */ - } - } - - $articleContent = $this->prepArticle($articleContent); - - if ($hasContent) { - // Find out text direction from ancestors of final top candidate. - $ancestors = array_merge([$parentOfTopCandidate, $topCandidate], $parentOfTopCandidate->getNodeAncestors()); - foreach ($ancestors as $ancestor) { - $articleDir = $ancestor->getAttribute('dir'); - if ($articleDir) { - $this->metadata['articleDir'] = $articleDir; - break; - } - } - - return $articleContent; - } else { - return false; - } - } - - /** - * TODO To be moved to Readability. - * - * @param DOMDocument $article - * - * @return DOMDocument - */ - public function prepArticle(DOMDocument $article) - { - $this->_cleanStyles($article); - $this->_clean($article, 'style'); - - // Check for data tables before we continue, to avoid removing items in - // those tables, which will often be isolated even though they're - // visually linked to other content-ful elements (text, images, etc.). - $this->_markDataTables($article); - - // Clean out junk from the article content - $this->_cleanConditionally($article, 'form'); - $this->_cleanConditionally($article, 'fieldset'); - $this->_clean($article, 'object'); - $this->_clean($article, 'embed'); - $this->_clean($article, 'h1'); - $this->_clean($article, 'footer'); - - // Clean out elements have "share" in their id/class combinations from final top candidates, - // which means we don't remove the top candidates even they have "share". - foreach ($article->childNodes as $child) { - $this->_cleanMatchedNodes($child, '/share/i'); - } - - /* - * If there is only one h2 and its text content substantially equals article title, - * they are probably using it as a header and not a subheader, - * so remove it since we already extract the title separately. - */ - $h2 = $article->getElementsByTagName('h2'); - if ($h2->length === 1) { - $lengthSimilarRate = (mb_strlen($h2->item(0)->textContent) - mb_strlen($this->metadata['title'])) / max(mb_strlen($this->metadata['title']), 1); - - if (abs($lengthSimilarRate) < 0.5) { - if ($lengthSimilarRate > 0) { - $titlesMatch = strpos($h2->item(0)->textContent, $this->metadata['title']) !== false; - } else { - $titlesMatch = strpos($this->metadata['title'], $h2->item(0)->textContent) !== false; - } - if ($titlesMatch) { - $this->_clean($article, 'h2'); - } - } - } - - $this->_clean($article, 'iframe'); - $this->_clean($article, 'input'); - $this->_clean($article, 'textarea'); - $this->_clean($article, 'select'); - $this->_clean($article, 'button'); - $this->_cleanHeaders($article); - - // Do these last as the previous stuff may have removed junk - // that will affect these - $this->_cleanConditionally($article, 'table'); - $this->_cleanConditionally($article, 'ul'); - $this->_cleanConditionally($article, 'div'); - - $this->_cleanExtraParagraphs($article); - - $this->_cleanReadabilityTags($article); - - foreach (iterator_to_array($article->getElementsByTagName('br')) as $br) { - $next = $br->nextSibling; - if ($next && $next->nodeName === 'p') { - $br->parentNode->removeChild($br); - } - } - - return $article; - } - - /** - * Look for 'data' (as opposed to 'layout') tables, for which we use - * similar checks as - * https://dxr.mozilla.org/mozilla-central/rev/71224049c0b52ab190564d3ea0eab089a159a4cf/accessible/html/HTMLTableAccessible.cpp#920. - * - * TODO To be moved to Readability. WARNING: check if we actually keep the "readabilityDataTable" param and - * maybe switch to a readability data-tag? - * - * @param DOMDocument $article - * - * @return void - */ - public function _markDataTables(DOMDocument $article) - { - $tables = $article->getElementsByTagName('table'); - foreach ($tables as $table) { - /** @var \DOMElement $table */ - $role = $table->getAttribute('role'); - if ($role === 'presentation') { - $table->readabilityDataTable = false; - continue; - } - $datatable = $table->getAttribute('datatable'); - if ($datatable == '0') { - $table->readabilityDataTable = false; - continue; - } - $summary = $table->getAttribute('summary'); - if ($summary) { - $table->readabilityDataTable = true; - continue; - } - - $caption = $table->getElementsByTagName('caption'); - if ($caption->length > 0 && $caption->item(0)->childNodes->length > 0) { - $table->readabilityDataTable = true; - continue; - } - - // If the table has a descendant with any of these tags, consider a data table: - foreach (['col', 'colgroup', 'tfoot', 'thead', 'th'] as $dataTableDescendants) { - if ($table->getElementsByTagName($dataTableDescendants)->length > 0) { - $table->readabilityDataTable = true; - continue 2; - } - } - - // Nested tables indicate a layout table: - if ($table->getElementsByTagName('table')->length > 0) { - $table->readabilityDataTable = false; - continue; - } - - $sizeInfo = $this->_getRowAndColumnCount($table); - if ($sizeInfo['rows'] >= 10 || $sizeInfo['columns'] > 4) { - $table->readabilityDataTable = true; - continue; - } - // Now just go by size entirely: - $table->readabilityDataTable = $sizeInfo['rows'] * $sizeInfo['columns'] > 10; - } - } - - /** - * Return an array indicating how many rows and columns this table has. - * - * @param \DOMElement $table - * - * @return array - */ - public function _getRowAndColumnCount(\DOMElement $table) - { - $rows = $columns = 0; - $trs = $table->getElementsByTagName('tr'); - foreach ($trs as $tr) { - /** @var \DOMElement $tr */ - $rowspan = $tr->getAttribute('rowspan'); - $rows += ($rowspan || 1); - - // Now look for column-related info - $columnsInThisRow = 0; - $cells = $tr->getElementsByTagName('td'); - foreach ($cells as $cell) { - /** @var \DOMElement $cell */ - $colspan = $cell->getAttribute('colspan'); - $columnsInThisRow += ($colspan || 1); - } - $columns = max($columns, $columnsInThisRow); - } - - return ['rows' => $rows, 'columns' => $columns]; - } - - /** - * TODO To be moved to Readability. - * - * @param DOMDocument $article - * - * @return void - */ - public function _cleanReadabilityTags(DOMDocument $article) - { - if ($this->getConfig()->getOption('removeReadabilityTags')) { - foreach ($article->getElementsByTagName('*') as $tag) { - if ($tag->hasAttribute('data-readability')) { - $tag->removeAttribute('data-readability'); - } - } - } - } - - /** - * Remove the style attribute on every e and under. - * TODO: To be moved to Readability. - * - * @param $node \DOMDocument|\DOMNode - **/ - public function _cleanStyles($node) - { - if (property_exists($node, 'tagName') && $node->tagName === 'svg') { - return; - } - - // Do not bother if there's no method to remove an attribute - if (method_exists($node, 'removeAttribute')) { - $presentational_attributes = ['align', 'background', 'bgcolor', 'border', 'cellpadding', 'cellspacing', 'frame', 'hspace', 'rules', 'style', 'valign', 'vspace']; - // Remove `style` and deprecated presentational attributes - foreach ($presentational_attributes as $presentational_attribute) { - $node->removeAttribute($presentational_attribute); - } - - $deprecated_size_attribute_elems = ['table', 'th', 'td', 'hr', 'pre']; - if (property_exists($node, 'tagName') && in_array($node->tagName, $deprecated_size_attribute_elems)) { - $node->removeAttribute('width'); - $node->removeAttribute('height'); - } - } - - $cur = $node->firstChild; - while ($cur !== null) { - $this->_cleanStyles($cur); - $cur = $cur->nextSibling; - } - } - - /** - * Clean out elements whose id/class combinations match specific string. - * - * TODO To be moved to readability - * - * @param string $regex Match id/class combination. - * - * @return void - **/ - public function _cleanMatchedNodes($node, $regex) - { - $node = new Readability($node); - $endOfSearchMarkerNode = $node->getNextNode($node, true); - $next = $node->getNextNode($node); - while ($next && $next !== $endOfSearchMarkerNode) { - if (preg_match($regex, sprintf('%s %s', $next->getAttribute('class'), $next->getAttribute('id')))) { - $next = $next->removeAndGetNext($next); - } else { - $next = $next->getNextNode($next); - } - } - } - - /** - * TODO To be moved to Readability. - * - * @param DOMDocument $article - * - * @return void - */ - public function _cleanExtraParagraphs(DOMDocument $article) - { - $paragraphs = $article->getElementsByTagName('p'); - $length = $paragraphs->length; - - for ($i = 0; $i < $length; $i++) { - $paragraph = $paragraphs->item($length - 1 - $i); - - $imgCount = $paragraph->getElementsByTagName('img')->length; - $embedCount = $paragraph->getElementsByTagName('embed')->length; - $objectCount = $paragraph->getElementsByTagName('object')->length; - // At this point, nasty iframes have been removed, only remain embedded video ones. - $iframeCount = $paragraph->getElementsByTagName('iframe')->length; - $totalCount = $imgCount + $embedCount + $objectCount + $iframeCount; - - if ($totalCount === 0 && !preg_replace($this->regexps['onlyWhitespace'], '', $paragraph->textContent)) { - // TODO must be done via readability - $paragraph->parentNode->removeChild($paragraph); - } - } - } - - /** - * TODO To be moved to Readability. - * - * @param DOMDocument $article - * - * @return void - */ - public function _cleanConditionally(DOMDocument $article, $tag) - { - if (!$this->getConfig()->getOption('cleanConditionally')) { - return; - } - - $isList = in_array($tag, ['ul', 'ol']); - - /* - * Gather counts for other typical elements embedded within. - * Traverse backwards so we can remove nodes at the same time - * without effecting the traversal. - */ - - $DOMNodeList = $article->getElementsByTagName($tag); - $length = $DOMNodeList->length; - for ($i = 0; $i < $length; $i++) { - $node = $DOMNodeList->item($length - 1 - $i); - - $node = new Readability($node); - - // First check if we're in a data table, in which case don't remove us. - if ($node->hasAncestorTag($node, 'table', -1) && isset($node->readabilityDataTable)) { - continue; - } - - $weight = $node->getClassWeight(); - - if ($weight < 0) { - $this->removeNode($node->getDOMNode()); - continue; - } - - if (substr_count($node->getTextContent(), ',') < 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. - */ - - // TODO Horrible hack, must be removed once this function is inside Readability - $p = $node->getDOMNode()->getElementsByTagName('p')->length; - $img = $node->getDOMNode()->getElementsByTagName('img')->length; - $li = $node->getDOMNode()->getElementsByTagName('li')->length - 100; - $input = $node->getDOMNode()->getElementsByTagName('input')->length; - - $embedCount = 0; - $embeds = $node->getDOMNode()->getElementsByTagName('embed'); - - foreach ($embeds as $embedNode) { - if (preg_match($this->regexps['videos'], $embedNode->C14N())) { - $embedCount++; - } - } - - $linkDensity = $this->getLinkDensity($node); - $contentLength = mb_strlen($node->getTextContent(true)); - - $haveToRemove = - ($img > 1 && $p / $img < 0.5 && !$node->hasAncestorTag($node, 'figure')) || - (!$isList && $li > $p) || - ($input > floor($p / 3)) || - (!$isList && $contentLength < 25 && ($img === 0 || $img > 2) && !$node->hasAncestorTag($node, 'figure')) || - (!$isList && $weight < 25 && $linkDensity > 0.2) || - ($weight >= 25 && $linkDensity > 0.5) || - (($embedCount === 1 && $contentLength < 75) || $embedCount > 1); - - if ($haveToRemove) { - $this->removeNode($node->getDOMNode()); - } - } - } - } - - /** - * Clean a node of all elements of type "tag". - * (Unless it's a youtube/vimeo video. People love movies.). - * - * TODO To be moved to Readability - * - * @param $article DOMDocument - * @param $tag string tag to clean - * - * @return void - **/ - public function _clean(DOMDocument $article, $tag) - { - $isEmbed = in_array($tag, ['object', 'embed', 'iframe']); - - $DOMNodeList = $article->getElementsByTagName($tag); - $length = $DOMNodeList->length; - for ($i = 0; $i < $length; $i++) { - $item = $DOMNodeList->item($length - 1 - $i); - - // Allow youtube and vimeo videos through as people usually want to see those. - if ($isEmbed) { - $attributeValues = []; - foreach ($item->attributes as $name => $value) { - $attributeValues[] = $value->nodeValue; - } - $attributeValues = implode('|', $attributeValues); - - // First, check the elements attributes to see if any of them contain youtube or vimeo - if (preg_match($this->regexps['videos'], $attributeValues)) { - continue; - } - - // Then check the elements inside this element for the same. - if (preg_match($this->regexps['videos'], $item->C14N())) { - continue; - } - } - $this->removeNode($item); - } - } - - /** - * Clean out spurious headers from an Element. Checks things like classnames and link density. - * - * TODO To be moved to Readability - * - * @param DOMDocument $article - * - * @return void - **/ - public function _cleanHeaders(DOMDocument $article) - { - for ($headerIndex = 1; $headerIndex < 3; $headerIndex++) { - $headers = $article->getElementsByTagName('h' . $headerIndex); - foreach ($headers as $header) { - $header = new Readability($header); - if ($header->getClassWeight() < 0) { - $this->removeNode($header->getDOMNode()); - } - } - } - } - - /** - * Remove the passed node. - * - * TODO To be moved to Readability - * - * @param \DOMNode $node - * - * @return void - **/ - public function removeNode(\DOMNode $node) - { - $parent = $node->parentNode; - if ($parent) { - $parent->removeChild($node); - } - } - - /** - * Checks if the node is a byline. - * - * @param Readability $node - * @param string $matchString - * - * @return bool - */ - private function checkByline($node, $matchString) - { - if (!$this->getConfig()->getOption('articleByLine')) { - return false; - } - - /* - * Check if the byline is already set - */ - if (isset($this->metadata['byline'])) { - return false; - } - - $rel = $node->getAttribute('rel'); - - if ($rel === 'author' || preg_match($this->regexps['byline'], $matchString) && $this->isValidByline($node->getTextContent())) { - $this->metadata['byline'] = trim($node->getTextContent()); - - return true; - } - - return false; - } - - /** - * Checks the validity of a byLine. Based on string length. - * - * @param string $text - * - * @return bool - */ - private function isValidByline($text) - { - if (gettype($text) == 'string') { - $byline = trim($text); - - return (mb_strlen($byline) > 0) && (mb_strlen($text) < 100); - } - - return false; - } - - /** - * Checks if the current node has a single child and if that child is a P node. - * Useful to convert <div><p> nodes to a single <p> node and avoid confusing the scoring system since div with p - * tags are, in practice, paragraphs. - * - * @param Readability $node - * - * @return bool - */ - private function hasSinglePNode(Readability $node) - { - // There should be exactly 1 element child which is a P: - if (count($children = $node->getChildren(true)) !== 1 || !$children[0]->tagNameEqualsTo('p')) { - return false; - } - - // And there should be no text nodes with real content (param true on ->getChildren) - foreach ($children as $child) { - /** @var $child Readability */ - if ($child->nodeTypeEqualsTo(XML_TEXT_NODE) && !preg_match('/\S$/', $child->getTextContent())) { - return false; - } - } - - return true; - } - - private function hasSingleChildBlockElement(Readability $node) - { - $result = false; - if ($node->hasChildren()) { - /** @var Readability $child */ - foreach ($node->getChildren() as $child) { - if (in_array($child->getTagName(), $this->divToPElements)) { - $result = true; - } else { - // If any of the hasSingleChildBlockElement calls return true, return true then. - $result = ($result || $this->hasSingleChildBlockElement($child)); - } - } - } - - return $result; - } -} diff --git a/src/Nodes/DOM/DOMAttr.php b/src/Nodes/DOM/DOMAttr.php new file mode 100644 index 0000000..91729f3 --- /dev/null +++ b/src/Nodes/DOM/DOMAttr.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMAttr extends \DOMAttr +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMCdataSection.php b/src/Nodes/DOM/DOMCdataSection.php new file mode 100644 index 0000000..4535e4e --- /dev/null +++ b/src/Nodes/DOM/DOMCdataSection.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMCdataSection extends \DOMCdataSection +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMCharacterData.php b/src/Nodes/DOM/DOMCharacterData.php new file mode 100644 index 0000000..e4db11a --- /dev/null +++ b/src/Nodes/DOM/DOMCharacterData.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMCharacterData extends \DOMCharacterData +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMComment.php b/src/Nodes/DOM/DOMComment.php new file mode 100644 index 0000000..13b004a --- /dev/null +++ b/src/Nodes/DOM/DOMComment.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMComment extends \DOMComment +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMDocument.php b/src/Nodes/DOM/DOMDocument.php new file mode 100644 index 0000000..b99d464 --- /dev/null +++ b/src/Nodes/DOM/DOMDocument.php @@ -0,0 +1,28 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMDocument extends \DOMDocument +{ + use NodeTrait; + + public function __construct($version, $encoding) + { + parent::__construct($version, $encoding); + + $this->registerNodeClass('DOMAttr', DOMAttr::class); + $this->registerNodeClass('DOMCdataSection', DOMCdataSection::class); + $this->registerNodeClass('DOMCharacterData', DOMCharacterData::class); + $this->registerNodeClass('DOMComment', DOMComment::class); + $this->registerNodeClass('DOMDocument', self::class); + $this->registerNodeClass('DOMDocumentFragment', DOMDocumentFragment::class); + $this->registerNodeClass('DOMDocumentType', DOMDocumentType::class); + $this->registerNodeClass('DOMElement', DOMElement::class); + $this->registerNodeClass('DOMNode', DOMNode::class); + $this->registerNodeClass('DOMNotation', DOMNotation::class); + $this->registerNodeClass('DOMProcessingInstruction', DOMProcessingInstruction::class); + $this->registerNodeClass('DOMText', DOMText::class); + } +} diff --git a/src/Nodes/DOM/DOMDocumentFragment.php b/src/Nodes/DOM/DOMDocumentFragment.php new file mode 100644 index 0000000..bcb8946 --- /dev/null +++ b/src/Nodes/DOM/DOMDocumentFragment.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMDocumentFragment extends \DOMDocumentFragment +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMDocumentType.php b/src/Nodes/DOM/DOMDocumentType.php new file mode 100644 index 0000000..18705a7 --- /dev/null +++ b/src/Nodes/DOM/DOMDocumentType.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMDocumentType extends \DOMDocumentType +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMElement.php b/src/Nodes/DOM/DOMElement.php new file mode 100644 index 0000000..c07670b --- /dev/null +++ b/src/Nodes/DOM/DOMElement.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMElement extends \DOMElement +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMNode.php b/src/Nodes/DOM/DOMNode.php new file mode 100644 index 0000000..f1944c4 --- /dev/null +++ b/src/Nodes/DOM/DOMNode.php @@ -0,0 +1,13 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +/** + * @method getAttribute($attribute) + */ +class DOMNode extends \DOMNode +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMNotation.php b/src/Nodes/DOM/DOMNotation.php new file mode 100644 index 0000000..2e888ce --- /dev/null +++ b/src/Nodes/DOM/DOMNotation.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMNotation extends \DOMNotation +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMProcessingInstruction.php b/src/Nodes/DOM/DOMProcessingInstruction.php new file mode 100644 index 0000000..9dd4c5c --- /dev/null +++ b/src/Nodes/DOM/DOMProcessingInstruction.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMProcessingInstruction extends \DOMProcessingInstruction +{ + use NodeTrait; +} diff --git a/src/Nodes/DOM/DOMText.php b/src/Nodes/DOM/DOMText.php new file mode 100644 index 0000000..42c575b --- /dev/null +++ b/src/Nodes/DOM/DOMText.php @@ -0,0 +1,10 @@ +<?php + +namespace andreskrey\Readability\Nodes\DOM; + +use andreskrey\Readability\Nodes\NodeTrait; + +class DOMText extends \DOMText +{ + use NodeTrait; +} diff --git a/src/Nodes/NodeTrait.php b/src/Nodes/NodeTrait.php new file mode 100644 index 0000000..f9bf8f9 --- /dev/null +++ b/src/Nodes/NodeTrait.php @@ -0,0 +1,430 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +use andreskrey\Readability\Nodes\DOM\DOMElement; +use andreskrey\Readability\Nodes\DOM\DOMNode; +use andreskrey\Readability\Nodes\DOM\DOMText; + +trait NodeTrait +{ + /** + * Content score of the node. Used to determine the value of the content. + * + * @var int + */ + public $contentScore = 0; + + /** + * Flag for initialized status. + * + * @var bool + */ + private $initialized = false; + + /** + * Flag data tables. + * + * @var bool + */ + private $readabilityDataTable = false; + + /** + * @var array + */ + private $divToPElements = [ + 'a', + 'blockquote', + 'dl', + 'div', + 'img', + 'ol', + 'p', + 'pre', + 'table', + 'ul', + 'select', + ]; + + /** + * initialized getter. + * + * @return bool + */ + public function isInitialized() + { + return $this->initialized; + } + + /** + * @return bool + */ + public function isReadabilityDataTable() + { + return $this->readabilityDataTable; + } + + /** + * @param bool $param + */ + public function setReadabilityDataTable($param) + { + $this->readabilityDataTable = $param; + } + + /** + * Initializer. Calculates the current score of the node and returns a full Readability object. + * + * @ TODO: I don't like the weightClasses param. How can we get the config here? + * + * @param $weightClasses bool Weight classes? + * + * @return static + */ + public function initializeNode($weightClasses) + { + if (!$this->isInitialized()) { + $contentScore = 0; + + switch ($this->nodeName) { + case 'div': + $contentScore += 5; + break; + + case 'pre': + case 'td': + case 'blockquote': + $contentScore += 3; + break; + + case 'address': + case 'ol': + case 'ul': + case 'dl': + case 'dd': + case 'dt': + case 'li': + case 'form': + $contentScore -= 3; + break; + + case 'h1': + case 'h2': + case 'h3': + case 'h4': + case 'h5': + case 'h6': + case 'th': + $contentScore -= 5; + break; + } + + $this->contentScore = $contentScore + ($weightClasses ? $this->getClassWeight() : 0); + + $this->initialized = true; + } + + return $this; + } + + /** + * Override for native getAttribute method. Some nodes have the getAttribute method, some don't, so we need + * to check first the existence of the attributes property. + * + * @param $attributeName string Attribute to retrieve + * + * @return string + */ + public function getAttribute($attributeName) + { + if (!is_null($this->attributes)) { + return parent::getAttribute($attributeName); + } + + return ''; + } + + /** + * Get the ancestors of the current node. + * + * @param int|bool $maxLevel Max amount of ancestors to get. False for all of them + * + * @return array + */ + public function getNodeAncestors($maxLevel = 3) + { + $ancestors = []; + $level = 0; + + $node = $this->parentNode; + + while ($node) { + $ancestors[] = $node; + $level++; + if ($level === $maxLevel) { + break; + } + $node = $node->parentNode; + } + + return $ancestors; + } + + /** + * Returns all links from the current element. + * + * @return array + */ + public function getAllLinks() + { + return iterator_to_array($this->getElementsByTagName('a')); + } + + /** + * 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. + * + * @return int + */ + public function getLinkDensity() + { + $linkLength = 0; + $textLength = mb_strlen($this->getTextContent(true)); + + if (!$textLength) { + return 0; + } + + $links = $this->getAllLinks(); + + if ($links) { + /** @var DOMElement $link */ + foreach ($links as $link) { + $linkLength += mb_strlen($link->getTextContent(true)); + } + } + + return $linkLength / $textLength; + } + + /** + * Calculates the weight of the class/id of the current element. + * + * @return int + */ + public function getClassWeight() + { + $weight = 0; + + // Look for a special classname + $class = $this->getAttribute('class'); + if (trim($class)) { + if (preg_match(NodeUtility::$regexps['negative'], $class)) { + $weight -= 25; + } + + if (preg_match(NodeUtility::$regexps['positive'], $class)) { + $weight += 25; + } + } + + // Look for a special ID + $id = $this->getAttribute('id'); + if (trim($id)) { + if (preg_match(NodeUtility::$regexps['negative'], $id)) { + $weight -= 25; + } + + if (preg_match(NodeUtility::$regexps['positive'], $id)) { + $weight += 25; + } + } + + return $weight; + } + + /** + * Returns the full text of the node. + * + * @param bool $normalize Normalize white space? + * + * @return string + */ + public function getTextContent($normalize = false) + { + $nodeValue = $this->nodeValue; + if ($normalize) { + $nodeValue = trim(preg_replace('/\s{2,}/', ' ', $nodeValue)); + } + + return $nodeValue; + } + + /** + * Returns the children of the current node. + * + * @param bool $filterEmptyDOMText Filter empty DOMText nodes? + * + * @return array + */ + public function getChildren($filterEmptyDOMText = false) + { + $ret = iterator_to_array($this->childNodes); + if ($filterEmptyDOMText) { + // Array values is used to discard the key order. Needs to be 0 to whatever without skipping any number + $ret = array_values(array_filter($ret, function ($node) { + return $node->nodeName !== '#text' || mb_strlen(trim($node->nodeValue)); + })); + } + + return $ret; + } + + /** + * Return an array indicating how many rows and columns this table has. + * + * @return array + */ + public function getRowAndColumnCount() + { + $rows = $columns = 0; + $trs = $this->getElementsByTagName('tr'); + foreach ($trs as $tr) { + /** @var \DOMElement $tr */ + $rowspan = $tr->getAttribute('rowspan'); + $rows += ($rowspan || 1); + + // Now look for column-related info + $columnsInThisRow = 0; + $cells = $tr->getElementsByTagName('td'); + foreach ($cells as $cell) { + /** @var \DOMElement $cell */ + $colspan = $cell->getAttribute('colspan'); + $columnsInThisRow += ($colspan || 1); + } + $columns = max($columns, $columnsInThisRow); + } + + return ['rows' => $rows, 'columns' => $columns]; + } + + /** + * Creates a new node based on the text content of the original node. + * + * @param $originalNode DOMNode + * @param $tagName string + * + * @return DOMElement + */ + public function createNode($originalNode, $tagName) + { + $text = $originalNode->getTextContent(); + $newNode = $originalNode->ownerDocument->createElement($tagName, $text); + + return $newNode; + } + + /** + * Check if a given node has one of its ancestor tag name matching the + * provided one. + * + * @param DOMElement $node + * @param string $tagName + * @param int $maxDepth + * + * @return bool + */ + public function hasAncestorTag($node, $tagName, $maxDepth = 3) + { + $depth = 0; + while ($node->parentNode) { + if ($maxDepth > 0 && $depth > $maxDepth) { + return false; + } + if ($node->parentNode->nodeName === $tagName) { + return true; + } + $node = $node->parentNode; + $depth++; + } + + return false; + } + + /** + * Checks if the current node has a single child and if that child is a P node. + * Useful to convert <div><p> nodes to a single <p> node and avoid confusing the scoring system since div with p + * tags are, in practice, paragraphs. + * + * @param DOMNode $node + * + * @return bool + */ + public function hasSinglePNode() + { + // There should be exactly 1 element child which is a P: + if (count($children = $this->getChildren(true)) !== 1 || $children[0]->nodeName !== 'p') { + return false; + } + + // And there should be no text nodes with real content (param true on ->getChildren) + foreach ($children as $child) { + /** @var $child DOMNode */ + if ($child->nodeType === XML_TEXT_NODE && !preg_match('/\S$/', $child->getTextContent())) { + return false; + } + } + + return true; + } + + /** + * Check if the current element has a single child block element. + * Block elements are the ones defined in the divToPElements array. + * + * @return bool + */ + public function hasSingleChildBlockElement() + { + $result = false; + if ($this->hasChildNodes()) { + foreach ($this->getChildren() as $child) { + if (in_array($child->nodeName, $this->divToPElements)) { + $result = true; + } else { + // If any of the hasSingleChildBlockElement calls return true, return true then. + /** @var $child DOMElement */ + $result = ($result || $child->hasSingleChildBlockElement()); + } + } + } + + return $result; + } + + /** + * Determines if a node has no content or it is just a bunch of dividing lines and/or whitespace. + * + * @return bool + */ + public function isElementWithoutContent() + { + return $this instanceof DOMElement && + mb_strlen(preg_replace(NodeUtility::$regexps['onlyWhitespace'], '', $this->textContent)) === 0 && + ($this->childNodes->length === 0 || + $this->childNodes->length === $this->getElementsByTagName('br')->length + $this->getElementsByTagName('hr')->length + /* + * Special PHP DOMDocument case: We also need to count how many DOMText we have inside the node. + * If there's an empty tag with an space inside and a BR (for example "<p> <br/></p>) counting only BRs and + * HRs will will say that the example has 2 nodes, instead of one. This happens because in DOMDocument, + * DOMTexts are also nodes (which doesn't happen in JS). So we need to also count how many DOMText we + * are dealing with (And at this point we know they are empty or are just whitespace, because of the + * mb_strlen in this chain of checks). + */ + + count(array_filter(iterator_to_array($this->childNodes), function ($child) { + return $child instanceof DOMText; + })) + + ); + } +} diff --git a/src/Nodes/NodeUtility.php b/src/Nodes/NodeUtility.php new file mode 100644 index 0000000..0e1072f --- /dev/null +++ b/src/Nodes/NodeUtility.php @@ -0,0 +1,159 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +use andreskrey\Readability\Nodes\DOM\DOMDocument; +use andreskrey\Readability\Nodes\DOM\DOMElement; +use andreskrey\Readability\Nodes\DOM\DOMNode; + +/** + * Class NodeUtility. + */ +class NodeUtility +{ + /** + * Collection of regexps to check the node usability. + * + * @var array + */ + public static $regexps = [ + 'unlikelyCandidates' => '/banner|breadcrumbs|combx|comment|community|cover-wrap|disqus|extra|foot|header|legends|menu|modal|related|remark|replies|rss|shoutbox|sidebar|skyscraper|social|sponsor|supplemental|ad-break|agegate|pagination|pager|popup|yom-remote/i', + 'okMaybeItsACandidate' => '/and|article|body|column|main|shadow/i', + 'extraneous' => '/print|archive|comment|discuss|e[\-]?mail|share|reply|all|login|sign|single|utility/i', + 'byline' => '/byline|author|dateline|writtenby|p-author/i', + 'replaceFonts' => '/<(\/?)font[^>]*>/gi', + 'normalize' => '/\s{2,}/', + 'videos' => '/\/\/(www\.)?(dailymotion|youtube|youtube-nocookie|player\.vimeo)\.com/i', + 'nextLink' => '/(next|weiter|continue|>([^\|]|$)|»([^\|]|$))/i', + 'prevLink' => '/(prev|earl|old|new|<|«)/i', + 'whitespace' => '/^\s*$/', + 'hasContent' => '/\S$/', + 'positive' => '/article|body|content|entry|hentry|h-entry|main|page|pagination|post|text|blog|story/i', + 'negative' => '/hidden|^hid$| hid$| hid |^hid |banner|combx|comment|com-|contact|foot|footer|footnote|masthead|media|meta|modal|outbrain|promo|related|scroll|share|shoutbox|sidebar|skyscraper|sponsor|shopping|tags|tool|widget/i', + // \x{00A0} is the unicode version of + 'onlyWhitespace' => '/\x{00A0}|\s+/u' + ]; + + /** + * Imported from the Element class on league\html-to-markdown. + * + * @param $node + * + * @return DOMElement + */ + public static function nextElement($node) + { + $next = $node; + while ($next + && $next->nodeName !== '#text' + && trim($next->textContent)) { + $next = $next->nextSibling; + } + + return $next; + } + + /** + * Changes the node tag name. Since tagName on DOMElement is a read only value, this must be done creating a new + * element with the new tag name and importing it to the main DOMDocument. + * + * @param string $value + * @param bool $importAttributes + * + * @return DOMNode + */ + public static function setNodeTag($node, $value, $importAttributes = false) + { + $new = new DOMDocument('1.0', 'utf-8'); + $new->appendChild($new->createElement($value)); + + $children = $node->childNodes; + /** @var $children \DOMNodeList $i */ + for ($i = 0; $i < $children->length; $i++) { + $import = $new->importNode($children->item($i), true); + $new->firstChild->appendChild($import); + } + + if ($importAttributes) { + // Import attributes from the original node. + foreach ($node->attributes as $attribute) { + $new->firstChild->setAttribute($attribute->nodeName, $attribute->nodeValue); + } + } + + // The import must be done on the firstChild of $new, since $new is a DOMDocument and not a DOMElement. + $import = $node->ownerDocument->importNode($new->firstChild, true); + $node->parentNode->replaceChild($import, $node); + + return $import; + } + + /** + * Removes the current node and returns the next node to be parsed (child, sibling or parent). + * + * @param DOMNode $node + * + * @return DOMNode + */ + public static function removeAndGetNext($node) + { + $nextNode = self::getNextNode($node, true); + $node->parentNode->removeChild($node); + + return $nextNode; + } + + /** + * Remove the selected node. + * + * @param $node DOMElement + * + * @return void + **/ + public static function removeNode($node) + { + $parent = $node->parentNode; + if ($parent) { + $parent->removeChild($node); + } + } + + /** + * Returns the next node. First checks for children (if the flag allows it), then for siblings, and finally + * for parents. + * + * @param DOMNode $originalNode + * @param bool $ignoreSelfAndKids + * + * @return DOMNode + */ + public static function getNextNode($originalNode, $ignoreSelfAndKids = false) + { + /* + * Traverse the DOM from node to node, starting at the node passed in. + * Pass true for the second parameter to indicate this node itself + * (and its kids) are going away, and we want the next node over. + * + * Calling this in a loop will traverse the DOM depth-first. + */ + + // First check for kids if those aren't being ignored + if (!$ignoreSelfAndKids && $originalNode->firstChild) { + return $originalNode->firstChild; + } + + // Then for siblings... + if ($originalNode->nextSibling) { + return $originalNode->nextSibling; + } + + // And finally, move up the parent chain *and* find a sibling + // (because this is depth-first traversal, we will have already + // seen the parent nodes themselves). + do { + $originalNode = $originalNode->parentNode; + } while ($originalNode && !$originalNode->nextSibling); + + return ($originalNode) ? $originalNode->nextSibling : $originalNode; + } +} diff --git a/src/ParseException.php b/src/ParseException.php new file mode 100644 index 0000000..646c04a --- /dev/null +++ b/src/ParseException.php @@ -0,0 +1,7 @@ +<?php + +namespace andreskrey\Readability; + +class ParseException extends \Exception +{ +} diff --git a/src/Readability.php b/src/Readability.php index bf8cdf8..19ccd5d 100644 --- a/src/Readability.php +++ b/src/Readability.php @@ -2,547 +2,1521 @@ namespace andreskrey\Readability; -use League\HTMLToMarkdown\Element; +use andreskrey\Readability\Nodes\DOM\DOMDocument; +use andreskrey\Readability\Nodes\DOM\DOMElement; +use andreskrey\Readability\Nodes\DOM\DOMNode; +use andreskrey\Readability\Nodes\DOM\DOMText; +use andreskrey\Readability\Nodes\NodeUtility; /** - * Class DOMElement. - * - * This is a extension of the original Element class from League\HTMLToMarkdown\Element. - * This class adds functions specific to Readability.php and overloads some of them to fit the purpose of this project. + * Class Readability. */ -class Readability extends Element implements ReadabilityInterface +class Readability { /** - * @var \DOMNode|\DOMElement + * Main DOMDocument where all the magic happens. + * + * @var DOMDocument + */ + protected $dom; + + /** + * Title of the article. + * + * @var string|null + */ + protected $title = null; + + /** + * HTML content article. + * + * @var string|null + */ + protected $content = null; + + /** + * Excerpt of the article. + * + * @var string|null */ - protected $node; + protected $excerpt = null; /** - * @var int + * Main image of the article. + * + * @var string|null + */ + protected $image = null; + + /** + * Author of the article. Extracted from the byline tags and other social media properties. + * + * @var string|null + */ + protected $author = null; + + /** + * Direction of the text. + * + * @var string|null + */ + protected $direction = null; + + /** + * Configuration object. + * + * @var Configuration */ - protected $contentScore = 0; + private $configuration; /** - * @var int + * @var array */ - protected $initialized = false; + private $defaultTagsToScore = [ + 'section', + 'h2', + 'h3', + 'h4', + 'h5', + 'h6', + 'p', + 'td', + 'pre', + ]; /** * @var array */ - private $regexps = [ - 'positive' => '/article|body|content|entry|hentry|h-entry|main|page|pagination|post|text|blog|story/i', - 'negative' => '/hidden|^hid$| hid$| hid |^hid |banner|combx|comment|com-|contact|foot|footer|footnote|masthead|media|meta|modal|outbrain|promo|related|scroll|share|shoutbox|sidebar|skyscraper|sponsor|shopping|tags|tool|widget/i', + private $alterToDIVExceptions = [ + 'div', + 'article', + 'section', + 'p', ]; /** - * Constructor. + * Readability constructor. * - * @param \DOMNode $node Selected element from DOMDocument + * @param Configuration $configuration */ - public function __construct(\DOMNode $node) + public function __construct(Configuration $configuration) { - parent::__construct($node); - - /* - * Restore the score if the object has been already scored. - * - * An if must be added before calling the getAttribute function, because if we reach the DOMDocument - * by getting the node parents we'll get a undefined function fatal error - */ - if (method_exists($node, 'getAttribute')) { - if ($node->hasAttribute('data-readability')) { - // Node was initialized previously. Restoring score and setting flag. - $this->initialized = true; - $score = $node->getAttribute('data-readability'); - $this->setContentScore($score); - } - } + $this->configuration = $configuration; } /** - * Checks for the tag name. Case insensitive. + * Main parse function. * - * @param string $value Name to compare to the current tag + * @param $html * - * @return bool + * @throws ParseException + * + * @return array|bool */ - public function tagNameEqualsTo($value) + public function parse($html) { - $tagName = $this->getTagName(); - if (strtolower($value) === strtolower($tagName)) { - return true; + $this->dom = $this->loadHTML($html); + + $this->getMetadata(); + + $this->getMainImage(); + + // Checking for minimum HTML to work with. + if (!($root = $this->dom->getElementsByTagName('body')->item(0)) || !$root->firstChild) { + throw new ParseException('Invalid or incomplete HTML.'); } - return false; + while (true) { + $root = $root->firstChild; + + $elementsToScore = $this->getNodes($root); + + $result = $this->rateNodes($elementsToScore); + + /* + * 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. + */ + + $length = 0; + foreach ($result->getElementsByTagName('p') as $p) { + $length += mb_strlen($p->textContent); + } + if ($result && mb_strlen(preg_replace('/\s/', '', $result->textContent)) < $this->configuration->getWordThreshold()) { + $this->dom = $this->loadHTML($html); + $root = $this->dom->getElementsByTagName('body')->item(0); + + if ($this->configuration->getStripUnlikelyCandidates()) { + $this->configuration->setStripUnlikelyCandidates(false); + } elseif ($this->configuration->getWeightClasses()) { + $this->configuration->setWeightClasses(false); + } elseif ($this->configuration->getCleanConditionally()) { + $this->configuration->setCleanConditionally(false); + } else { + throw new ParseException('Could not parse text.'); + } + } else { + break; + } + } + + $result = $this->postProcessContent($result); + + // If we haven't found an excerpt in the article's metadata, use the article's + // first paragraph as the excerpt. This can be used for displaying a preview of + // the article's content. + if (!$this->getExcerpt()) { + $paragraphs = $result->getElementsByTagName('p'); + if ($paragraphs->length > 0) { + $this->setExcerpt(trim($paragraphs->item(0)->textContent)); + } + } + + $this->setContent($result->C14N()); + + return true; } /** - * Checks for the node type. + * Creates a DOM Document object and loads the provided HTML on it. * - * @param string $value Type of node to compare to + * Used for the first load of Readability and subsequent reloads (when disabling flags and rescanning the text) + * Previous versions of Readability used this method one time and cloned the DOM to keep a backup. This caused bugs + * because cloning the DOM object keeps a relation between the clone and the original one, doing changes in both + * objects and ruining the backup. * - * @return bool + * @param string $html + * + * @return DOMDocument */ - public function nodeTypeEqualsTo($value) + private function loadHTML($html) { - return $this->node->nodeType === $value; + // To avoid throwing a gazillion of errors on malformed HTMLs + libxml_use_internal_errors(true); + + $dom = new DOMDocument('1.0', 'utf-8'); + + if (!$this->configuration->getSubstituteEntities()) { + // Keep the original HTML entities + $dom->substituteEntities = false; + } + + if ($this->configuration->getNormalizeEntities()) { + // Replace UTF-8 characters with the HTML Entity equivalent. Useful to fix html with mixed content + $html = mb_convert_encoding($html, 'HTML-ENTITIES', 'UTF-8'); + } + + if ($this->configuration->getSummonCthulhu()) { + $html = preg_replace('/<script\b[^>]*>([\s\S]*?)<\/script>/', '', $html); + } + + // Prepend the XML tag to avoid having issues with special characters. Should be harmless. + $dom->loadHTML('<?xml encoding="UTF-8">' . $html); + $dom->encoding = 'UTF-8'; + + $this->removeScripts($dom); + + $this->prepDocument($dom); + + return $dom; } /** - * Get the ancestors of the current node. - * - * @param int|bool $maxLevel Max amount of ancestors to get. False for all of them - * - * @return array + * Tries to guess relevant info from metadata of the html. Sets the results in the Readability properties. */ - public function getNodeAncestors($maxLevel = 3) + private function getMetadata() { - $ancestors = []; - $level = 0; + $values = []; + // Match "description", or Twitter's "twitter:description" (Cards) + // in name attribute. + $namePattern = '/^\s*((twitter)\s*:\s*)?(description|title|image)\s*$/i'; - $node = $this->getParent(); + // Match Facebook's Open Graph title & description properties. + $propertyPattern = '/^\s*og\s*:\s*(description|title|image)\s*$/i'; - while ($node) { - $ancestors[] = $node; - $level++; - if ($level === $maxLevel) { - break; + foreach ($this->dom->getElementsByTagName('meta') as $meta) { + /* @var DOMNode $meta */ + $elementName = $meta->getAttribute('name'); + $elementProperty = $meta->getAttribute('property'); + + if (in_array('author', [$elementName, $elementProperty])) { + $this->setAuthor($meta->getAttribute('content')); + continue; + } + + $name = null; + if (preg_match($namePattern, $elementName)) { + $name = $elementName; + } elseif (preg_match($propertyPattern, $elementProperty)) { + $name = $elementProperty; + } + + if ($name) { + $content = $meta->getAttribute('content'); + if ($content) { + // Convert to lowercase and remove any whitespace + // so we can match below. + $name = preg_replace('/\s/', '', strtolower($name)); + $values[$name] = trim($content); + } + } + } + if (array_key_exists('description', $values)) { + $this->setExcerpt($values['description']); + } elseif (array_key_exists('og:description', $values)) { + // Use facebook open graph description. + $this->setExcerpt($values['og:description']); + } elseif (array_key_exists('twitter:description', $values)) { + // Use twitter cards description. + $this->setExcerpt($values['twitter:description']); + } + + $this->setTitle($this->getArticleTitle()); + + if (!$this->getTitle()) { + if (array_key_exists('og:title', $values)) { + // Use facebook open graph title. + $this->setTitle($values['og:title']); + } elseif (array_key_exists('twitter:title', $values)) { + // Use twitter cards title. + $this->setTitle($values['twitter:title']); } - $node = $node->getParent(); } - return $ancestors; + if (array_key_exists('og:image', $values) || array_key_exists('twitter:image', $values)) { + $this->setImage(array_key_exists('og:image', $values) ? $values['og:image'] : $values['twitter:image']); + } } /** - * Overloading the getParent function from League\HTMLToMarkdown\Element due to a bug when there are no more parents - * on the selected element. + * Returns all the images of the parsed article. * - * @return Readability|null + * @return array */ - public function getParent() + public function getImages() { - $node = $this->node->parentNode; + $result = []; + if ($this->getImage()) { + $result[] = $this->getImage(); + } + + if (null == $this->dom) { + return $result; + } + + foreach ($this->dom->getElementsByTagName('img') as $img) { + if ($src = $img->getAttribute('src')) { + $result[] = $src; + } + } - return ($node) ? new self($node) : null; + if ($this->configuration->getFixRelativeURLs()) { + foreach ($result as &$imgSrc) { + $imgSrc = $this->toAbsoluteURI($imgSrc); + } + } + + $result = array_unique(array_filter($result)); + + return $result; } /** - * Returns all links from the current element. - * - * @return array|null + * Tries to get the main article image. Will only update the metadata if the getMetadata function couldn't + * find a correct image. */ - public function getAllLinks() + public function getMainImage() { - if (($this->isText())) { - return null; - } else { - $links = []; - foreach ($this->node->getElementsByTagName('a') as $link) { - $links[] = new self($link); + $imgUrl = false; + + if ($this->getImage() !== null) { + $imgUrl = $this->getImage(); + } + + if (!$imgUrl) { + foreach ($this->dom->getElementsByTagName('link') as $link) { + /** @var \DOMElement $link */ + /* + * Check for the rel attribute, then check if the rel attribute is either img_src or image_src, and + * finally check for the existence of the href attribute, which should hold the image url. + */ + if ($link->hasAttribute('rel') && ($link->getAttribute('rel') === 'img_src' || $link->getAttribute('rel') === 'image_src') && $link->hasAttribute('href')) { + $imgUrl = $link->getAttribute('href'); + break; + } } + } - return $links; + if (!empty($imgUrl) && $this->configuration->getFixRelativeURLs()) { + $this->setImage($this->toAbsoluteURI($imgUrl)); } } /** - * Initializer. Calculates the current score of the node and returns a full Readability object. + * Returns the title of the html. Prioritizes the title from the metadata against the title tag. * - * @return Readability + * @return string|null */ - public function initializeNode() + private function getArticleTitle() { - if (!$this->initialized) { - $contentScore = 0; + $originalTitle = null; - switch ($this->getTagName()) { - case 'div': - $contentScore += 5; - break; + if ($this->getTitle()) { + $originalTitle = $this->getTitle(); + } else { + $titleTag = $this->dom->getElementsByTagName('title'); + if ($titleTag->length > 0) { + $originalTitle = $titleTag->item(0)->nodeValue; + } + } - case 'pre': - case 'td': - case 'blockquote': - $contentScore += 3; - break; + if ($originalTitle === null) { + return null; + } - case 'address': - case 'ol': - case 'ul': - case 'dl': - case 'dd': - case 'dt': - case 'li': - case 'form': - $contentScore -= 3; - break; + $curTitle = $originalTitle; + $titleHadHierarchicalSeparators = false; - case 'h1': - case 'h2': - case 'h3': - case 'h4': - case 'h5': - case 'h6': - case 'th': - $contentScore -= 5; - break; + /* + * If there's a separator in the title, first remove the final part + * + * Sanity warning: if you eval this match in PHPStorm's "Evaluate expression" box, it will return false + * I can assure you it works properly if you let the code run. + */ + if (preg_match('/ [\|\-\\\\\/>»] /i', $curTitle)) { + $titleHadHierarchicalSeparators = (bool)preg_match('/ [\\\\\/>»] /', $curTitle); + $curTitle = preg_replace('/(.*)[\|\-\\\\\/>»] .*/i', '$1', $originalTitle); + + // If the resulting title is too short (3 words or fewer), remove + // the first part instead: + if (count(preg_split('/\s+/', $curTitle)) < 3) { + $curTitle = preg_replace('/[^\|\-\\\\\/>»]*[\|\-\\\\\/>»](.*)/i', '$1', $originalTitle); + } + } elseif (strpos($curTitle, ': ') !== false) { + // Check if we have an heading containing this exact string, so we + // could assume it's the full title. + $match = false; + for ($i = 1; $i <= 2; $i++) { + foreach ($this->dom->getElementsByTagName('h' . $i) as $hTag) { + // Trim texts to avoid having false negatives when the title is surrounded by spaces or tabs + if (trim($hTag->nodeValue) === trim($curTitle)) { + $match = true; + } + } } - $this->setContentScore($contentScore + $this->getClassWeight()); + // If we don't, let's extract the title out of the original title string. + if (!$match) { + $curTitle = substr($originalTitle, strrpos($originalTitle, ':') + 1); - $this->initialized = true; + // If the title is now too short, try the first colon instead: + if (count(preg_split('/\s+/', $curTitle)) < 3) { + $curTitle = substr($originalTitle, strpos($originalTitle, ':') + 1); + } + } + } elseif (mb_strlen($curTitle) > 150 || mb_strlen($curTitle) < 15) { + $hOnes = $this->dom->getElementsByTagName('h1'); + + if ($hOnes->length === 1) { + $curTitle = $hOnes->item(0)->nodeValue; + } } - return $this; + $curTitle = trim($curTitle); + + /* + * If we now have 4 words or fewer as our title, and either no + * 'hierarchical' separators (\, /, > or ») were found in the original + * title or we decreased the number of words by more than 1 word, use + * the original title. + */ + $curTitleWordCount = count(preg_split('/\s+/', $curTitle)); + $originalTitleWordCount = count(preg_split('/\s+/', preg_replace('/[\|\-\\\\\/>»]+/', '', $originalTitle))) - 1; + + if ($curTitleWordCount <= 4 && + (!$titleHadHierarchicalSeparators || $curTitleWordCount !== $originalTitleWordCount)) { + $curTitle = $originalTitle; + } + + return $curTitle; } /** - * Calculates the weight of the class/id of the current element. + * Convert URI to an absolute URI. * - * @todo check for flag that lets this function run or not + * @param $uri string URI to convert * - * @return int + * @return string */ - public function getClassWeight() + private function toAbsoluteURI($uri) { - // TODO To implement. How to get config from html parser from readability -// if ($this->getConfig()->getOption('weightClasses')) { -// return 0; -// } -// - $weight = 0; + list($pathBase, $scheme, $prePath) = $this->getPathInfo($this->configuration->getOriginalURL()); - // Look for a special classname - $class = $this->getAttribute('class'); - if (trim($class)) { - if (preg_match($this->regexps['negative'], $class)) { - $weight -= 25; - } + // If this is already an absolute URI, return it. + if (preg_match('/^[a-zA-Z][a-zA-Z0-9\+\-\.]*:/', $uri)) { + return $uri; + } - if (preg_match($this->regexps['positive'], $class)) { - $weight += 25; - } + // Scheme-rooted relative URI. + if (substr($uri, 0, 2) === '//') { + return $scheme . '://' . substr($uri, 2); } - // Look for a special ID - $id = $this->getAttribute('id'); - if (trim($id)) { - if (preg_match($this->regexps['negative'], $id)) { - $weight -= 25; - } + // Prepath-rooted relative URI. + if (substr($uri, 0, 1) === '/') { + return $prePath . $uri; + } - if (preg_match($this->regexps['positive'], $id)) { - $weight += 25; - } + // Dotslash relative URI. + if (strpos($uri, './') === 0) { + return $pathBase . substr($uri, 2); + } + // Ignore hash URIs: + if (substr($uri, 0, 1) === '#') { + return $uri; } - return $weight; + // Standard relative URI; add entire path. pathBase already includes a + // trailing "/". + return $pathBase . $uri; } /** - * Returns the current score of the Readability object. + * Returns full path info of an URL. * - * @return int + * @param string $url + * + * @return array [$pathBase, $scheme, $prePath] */ - public function getContentScore() + public function getPathInfo($url) { - return $this->contentScore; + $pathBase = parse_url($url, PHP_URL_SCHEME) . '://' . parse_url($url, PHP_URL_HOST) . dirname(parse_url($url, PHP_URL_PATH)) . '/'; + $scheme = parse_url($pathBase, PHP_URL_SCHEME); + $prePath = $scheme . '://' . parse_url($pathBase, PHP_URL_HOST); + + return [$pathBase, $scheme, $prePath]; } /** - * Returns the current score of the Readability object. + * Gets nodes from the root element. * - * @param int $score + * @param $node DOMNode|DOMText * - * @return int + * @return array */ - public function setContentScore($score) + private function getNodes($node) { - // Check if the setAttribute method exists, as some elements lack of it (and calling it anyway throws an exception) - if (method_exists($this->node, 'setAttribute')) { - $this->contentScore = (float)$score; + $stripUnlikelyCandidates = $this->configuration->getStripUnlikelyCandidates(); + + $elementsToScore = []; + + /* + * First, node prepping. Trash nodes that look cruddy (like ones with the + * class name "comment", etc), and turn divs into P tags where they have been + * used inappropriately (as in, where they contain no other block level elements.) + */ + + while ($node) { + $matchString = $node->getAttribute('class') . ' ' . $node->getAttribute('id'); - // Set score in an attribute of the tag to prevent losing it while creating new Readability objects. - $this->node->setAttribute('data-readability', $this->contentScore); + // Remove DOMComments nodes as we don't need them and mess up children counting + if ($node->nodeType === XML_COMMENT_NODE) { + $node = NodeUtility::removeAndGetNext($node); + continue; + } + + // Check to see if this node is a byline, and remove it if it is. + if ($this->checkByline($node, $matchString)) { + $node = NodeUtility::removeAndGetNext($node); + continue; + } + + // Remove unlikely candidates + if ($stripUnlikelyCandidates) { + if ( + preg_match(NodeUtility::$regexps['unlikelyCandidates'], $matchString) && + !preg_match(NodeUtility::$regexps['okMaybeItsACandidate'], $matchString) && + $node->nodeName !== 'body' && + $node->nodeName !== 'a' + ) { + $node = NodeUtility::removeAndGetNext($node); + continue; + } + } + + // Remove DIV, SECTION, and HEADER nodes without any content(e.g. text, image, video, or iframe). + if (($node->nodeName === 'div' || $node->nodeName === 'section' || $node->nodeName === 'header' || + $node->nodeName === 'h1' || $node->nodeName === 'h2' || $node->nodeName === 'h3' || + $node->nodeName === 'h4' || $node->nodeName === 'h5' || $node->nodeName === 'h6' || + $node->nodeName === 'p') && + $node->isElementWithoutContent()) { + $node = NodeUtility::removeAndGetNext($node); + continue; + } - return $this->contentScore; + if (in_array(strtolower($node->nodeName), $this->defaultTagsToScore)) { + $elementsToScore[] = $node; + } + + // Turn all divs that don't have children block level elements into p's + if ($node->nodeName === 'div') { + /* + * Sites like http://mobile.slate.com encloses each paragraph with a DIV + * element. DIVs with only a P element inside and no text content can be + * safely converted into plain P elements to avoid confusing the scoring + * algorithm with DIVs with are, in practice, paragraphs. + */ + if ($node->hasSinglePNode()) { + $pNode = $node->getChildren(true)[0]; + $node->parentNode->replaceChild($pNode, $node); + $node = $pNode; + $elementsToScore[] = $node; + } elseif (!$node->hasSingleChildBlockElement()) { + $node = NodeUtility::setNodeTag($node, 'p'); + $elementsToScore[] = $node; + } else { + // EXPERIMENTAL + foreach ($node->getChildren() as $child) { + /** @var $child DOMNode */ + if ($child->nodeType === XML_TEXT_NODE && mb_strlen(trim($child->getTextContent())) > 0) { + $newNode = $node->createNode($child, 'p'); + $child->parentNode->replaceChild($newNode, $child); + } + } + } + } + + $node = NodeUtility::getNextNode($node); } - return 0; + return $elementsToScore; } /** - * Returns the full text of the node. + * Checks if the node is a byline. * - * @param bool $normalize Normalize white space? + * @param DOMNode $node + * @param string $matchString * - * @return string + * @return bool */ - public function getTextContent($normalize = false) + private function checkByline($node, $matchString) { - $nodeValue = $this->node->nodeValue; - if ($normalize) { - $nodeValue = trim(preg_replace('/\s{2,}/', ' ', $nodeValue)); + if (!$this->configuration->getArticleByLine()) { + return false; + } + + /* + * Check if the byline is already set + */ + if ($this->getAuthor()) { + return false; } - return $nodeValue; + $rel = $node->getAttribute('rel'); + + if ($rel === 'author' || preg_match(NodeUtility::$regexps['byline'], $matchString) && $this->isValidByline($node->getTextContent())) { + $this->setAuthor(trim($node->getTextContent())); + + return true; + } + + return false; } /** - * Changes the node tag name. Since tagName on DOMElement is a read only value, this must be done creating a new - * element with the new tag name and importing it to the main DOMDocument. + * Checks the validity of a byLine. Based on string length. * - * @param string $value - * @param bool $importAttributes + * @param string $text + * + * @return bool */ - public function setNodeTag($value, $importAttributes = false) + private function isValidByline($text) { - $new = new \DOMDocument(); - $new->appendChild($new->createElement($value)); - - $childs = $this->node->childNodes; - for ($i = 0; $i < $childs->length; $i++) { - $import = $new->importNode($childs->item($i), true); - $new->firstChild->appendChild($import); - } + if (gettype($text) == 'string') { + $byline = trim($text); - if ($importAttributes) { - // Import attributes from the original node. - foreach ($this->node->attributes as $attribute) { - $new->firstChild->setAttribute($attribute->nodeName, $attribute->nodeValue); - } + return (mb_strlen($byline) > 0) && (mb_strlen($text) < 100); } - // The import must be done on the firstChild of $new, since $new is a DOMDocument and not a DOMElement. - $import = $this->node->ownerDocument->importNode($new->firstChild, true); - $this->node->parentNode->replaceChild($import, $this->node); - - $this->node = $import; + return false; } /** - * Returns the current DOMNode. + * Removes all the scripts of the html. * - * @return \DOMNode + * @param DOMDocument $dom */ - public function getDOMNode() + private function removeScripts(DOMDocument $dom) { - return $this->node; + $toRemove = ['script', 'noscript']; + + foreach ($toRemove as $tag) { + while ($script = $dom->getElementsByTagName($tag)) { + if ($script->item(0)) { + $script->item(0)->parentNode->removeChild($script->item(0)); + } else { + break; + } + } + } } /** - * Removes the current node and returns the next node to be parsed (child, sibling or parent). - * - * @param Readability $node + * Prepares the document for parsing. * - * @return Readability + * @param DOMDocument $dom */ - public function removeAndGetNext($node) + private function prepDocument(DOMDocument $dom) { - $nextNode = $this->getNextNode($node, true); - $node->node->parentNode->removeChild($node->node); + /* + * DOMNodeList must be converted to an array before looping over it. + * This is done to avoid node shifting when removing nodes. + * + * Reverse traversing cannot be done here because we need to find brs that are right next to other brs. + * (If we go the other way around we need to search for previous nodes forcing the creation of new functions + * that will be used only here) + */ + foreach (iterator_to_array($dom->getElementsByTagName('br')) as $br) { + $next = $br->nextSibling; + + /* + * Whether 2 or more <br> elements have been found and replaced with a + * <p> block. + */ + $replaced = false; + + /* + * If we find a <br> chain, remove the <br>s until we hit another element + * or non-whitespace. This leaves behind the first <br> in the chain + * (which will be replaced with a <p> later). + */ + while (($next = NodeUtility::nextElement($next)) && ($next->nodeName === 'br')) { + $replaced = true; + $brSibling = $next->nextSibling; + $next->parentNode->removeChild($next); + $next = $brSibling; + } + + /* + * If we removed a <br> chain, replace the remaining <br> with a <p>. Add + * all sibling nodes as children of the <p> until we hit another <br> + * chain. + */ + + if ($replaced) { + $p = $dom->createElement('p'); + $br->parentNode->replaceChild($p, $br); + + $next = $p->nextSibling; + while ($next) { + // If we've hit another <br><br>, we're done adding children to this <p>. + if ($next->nodeName === 'br') { + $nextElem = NodeUtility::nextElement($next); + if ($nextElem && $nextElem->nodeName === 'br') { + break; + } + } + + // Otherwise, make this node a child of the new <p>. + $sibling = $next->nextSibling; + $p->appendChild($next); + $next = $sibling; + } + } + } - return $nextNode; + // Replace font tags with span + $fonts = $dom->getElementsByTagName('font'); + $length = $fonts->length; + for ($i = 0; $i < $length; $i++) { + $font = $fonts->item($length - 1 - $i); + NodeUtility::setNodeTag($font, 'span', true); + } } /** - * Returns the next node. First checks for childs (if the flag allows it), then for siblings, and finally - * for parents. + * Assign scores to each node. Returns full article parsed or false on error. * - * @param Readability $originalNode - * @param bool $ignoreSelfAndKids + * @param array $nodes * - * @return Readability + * @return DOMDocument|bool */ - public function getNextNode($originalNode, $ignoreSelfAndKids = false) + private function rateNodes($nodes) { + $candidates = []; + + /** @var DOMElement $node */ + foreach ($nodes as $node) { + if (is_null($node->parentNode)) { + continue; + } + + // Discard nodes with less than 25 characters, without blank space + if (mb_strlen($node->getTextContent(true)) < 25) { + continue; + } + + $ancestors = $node->getNodeAncestors(); + + // Exclude nodes with no ancestor + if (count($ancestors) === 0) { + continue; + } + + // Start with a point for the paragraph itself as a base. + $contentScore = 1; + + // Add points for any commas within this paragraph. + $contentScore += count(explode(',', $node->getTextContent(true))); + + // For every 100 characters in this paragraph, add another point. Up to 3 points. + $contentScore += min(floor(mb_strlen($node->getTextContent(true)) / 100), 3); + + /** @var $ancestor DOMElement */ + foreach ($ancestors as $level => $ancestor) { + if (!$ancestor->isInitialized()) { + $ancestor->initializeNode($this->configuration->getWeightClasses()); + $candidates[] = $ancestor; + } + + /* + * Node score divider: + * - parent: 1 (no division) + * - grandparent: 2 + * - great grandparent+: ancestor level * 3 + */ + + if ($level === 0) { + $scoreDivider = 1; + } elseif ($level === 1) { + $scoreDivider = 2; + } else { + $scoreDivider = $level * 3; + } + + $currentScore = $ancestor->contentScore; + $ancestor->contentScore = $currentScore + ($contentScore / $scoreDivider); + } + } + /* - * Traverse the DOM from node to node, starting at the node passed in. - * Pass true for the second parameter to indicate this node itself - * (and its kids) are going away, and we want the next node over. - * - * Calling this in a loop will traverse the DOM depth-first. + * After we've calculated scores, loop through all of the possible + * candidate nodes we found and find the one with the highest score. + */ + + $topCandidates = []; + foreach ($candidates as $candidate) { + + /* + * 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. + */ + + $candidate->contentScore = $candidate->contentScore * (1 - $candidate->getLinkDensity()); + + for ($i = 0; $i < $this->configuration->getMaxTopCandidates(); $i++) { + $aTopCandidate = isset($topCandidates[$i]) ? $topCandidates[$i] : null; + + if (!$aTopCandidate || $candidate->contentScore > $aTopCandidate->contentScore) { + array_splice($topCandidates, $i, 0, [$candidate]); + if (count($topCandidates) > $this->configuration->getMaxTopCandidates()) { + array_pop($topCandidates); + } + break; + } + } + } + + $topCandidate = isset($topCandidates[0]) ? $topCandidates[0] : null; + $parentOfTopCandidate = null; + + /* + * If we still have no top candidate, just use the body as a last resort. + * We also have to copy the body node so it is something we can modify. */ - // First check for kids if those aren't being ignored - if (!$ignoreSelfAndKids && $originalNode->node->firstChild) { - return new self($originalNode->node->firstChild); + if ($topCandidate === null || $topCandidate->nodeName === 'body') { + // Move all of the page's children into topCandidate + $topCandidate = new DOMDocument('1.0', 'utf-8'); + $topCandidate->encoding = 'UTF-8'; + $topCandidate->appendChild($topCandidate->createElement('div', '')); + $kids = $this->dom->getElementsByTagName('body')->item(0)->childNodes; + + // Cannot be foreached, don't ask me why. + for ($i = 0; $i < $kids->length; $i++) { + $import = $topCandidate->importNode($kids->item($i), true); + $topCandidate->firstChild->appendChild($import); + } + + // Candidate must be created using firstChild to grab the DOMElement instead of the DOMDocument. + $topCandidate = $topCandidate->firstChild; + } elseif ($topCandidate) { + // Find a better top candidate node if it contains (at least three) nodes which belong to `topCandidates` array + // and whose scores are quite closed with current `topCandidate` node. + $alternativeCandidateAncestors = []; + for ($i = 1; $i < count($topCandidates); $i++) { + if ($topCandidates[$i]->contentScore / $topCandidate->contentScore >= 0.75) { + array_push($alternativeCandidateAncestors, $topCandidates[$i]->getNodeAncestors(false)); + } + } + + $MINIMUM_TOPCANDIDATES = 3; + if (count($alternativeCandidateAncestors) >= $MINIMUM_TOPCANDIDATES) { + $parentOfTopCandidate = $topCandidate->parentNode; + while ($parentOfTopCandidate->nodeName !== 'body') { + $listsContainingThisAncestor = 0; + for ($ancestorIndex = 0; $ancestorIndex < count($alternativeCandidateAncestors) && $listsContainingThisAncestor < $MINIMUM_TOPCANDIDATES; $ancestorIndex++) { + $listsContainingThisAncestor += (int)in_array($parentOfTopCandidate, $alternativeCandidateAncestors[$ancestorIndex]); + } + if ($listsContainingThisAncestor >= $MINIMUM_TOPCANDIDATES) { + $topCandidate = $parentOfTopCandidate; + break; + } + $parentOfTopCandidate = $parentOfTopCandidate->parentNode; + } + } + + /* + * Because of our bonus system, parents of candidates might have scores + * themselves. They get half of the node. There won't be nodes with higher + * scores than our topCandidate, but if we see the score going *up* in the first + * few steps up the tree, that's a decent sign that there might be more content + * lurking in other places that we want to unify in. The sibling stuff + * below does some of that - but only if we've looked high enough up the DOM + * tree. + */ + + $parentOfTopCandidate = $topCandidate->parentNode; + $lastScore = $topCandidate->contentScore; + + // The scores shouldn't get too low. + $scoreThreshold = $lastScore / 3; + + /* @var DOMElement $parentOfTopCandidate */ + while ($parentOfTopCandidate->nodeName !== 'body') { + $parentScore = $parentOfTopCandidate->contentScore; + if ($parentScore < $scoreThreshold) { + break; + } + + if ($parentScore > $lastScore) { + // Alright! We found a better parent to use. + $topCandidate = $parentOfTopCandidate; + break; + } + $lastScore = $parentOfTopCandidate->contentScore; + $parentOfTopCandidate = $parentOfTopCandidate->parentNode; + } + + // If the top candidate is the only child, use parent instead. This will help sibling + // joining logic when adjacent content is actually located in parent's sibling node. + $parentOfTopCandidate = $topCandidate->parentNode; + while ($parentOfTopCandidate->nodeName !== 'body' && count($parentOfTopCandidate->getChildren(true)) === 1) { + $topCandidate = $parentOfTopCandidate; + $parentOfTopCandidate = $topCandidate->parentNode; + } } - // Then for siblings... - if ($originalNode->node->nextSibling) { - return new self($originalNode->node->nextSibling); + /* + * 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 = new DOMDocument('1.0', 'utf-8'); + $articleContent->createElement('div'); + + $siblingScoreThreshold = max(10, $topCandidate->contentScore * 0.2); + // Keep potential top candidate's parent node to try to get text direction of it later. + $parentOfTopCandidate = $topCandidate->parentNode; + $siblings = $parentOfTopCandidate->getChildren(); + + $hasContent = false; + + /** @var DOMElement $sibling */ + foreach ($siblings as $sibling) { + $append = false; + + if ($sibling === $topCandidate) { + $append = true; + } else { + $contentBonus = 0; + + // Give a bonus if sibling nodes and top candidates have the example same classname + if ($sibling->getAttribute('class') === $topCandidate->getAttribute('class') && $topCandidate->getAttribute('class') !== '') { + $contentBonus += $topCandidate->contentScore * 0.2; + } + if ($sibling->contentScore + $contentBonus >= $siblingScoreThreshold) { + $append = true; + } elseif ($sibling->nodeName === 'p') { + $linkDensity = $sibling->getLinkDensity(); + $nodeContent = $sibling->getTextContent(true); + + if (mb_strlen($nodeContent) > 80 && $linkDensity < 0.25) { + $append = true; + } elseif ($nodeContent && mb_strlen($nodeContent) < 80 && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent)) { + $append = true; + } + } + } + + if ($append) { + $hasContent = true; + + if (!in_array(strtolower($sibling->nodeName), $this->alterToDIVExceptions)) { + /* + * 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. + */ + + $sibling = NodeUtility::setNodeTag($sibling, 'div'); + } + + $import = $articleContent->importNode($sibling, true); + $articleContent->appendChild($import); + + /* + * No node shifting needs to be check because when calling getChildren, an array is made with the + * children of the parent node, instead of using the DOMElement childNodes function, which, when used + * along with appendChild, would shift the nodes position and the current foreach will behave in + * unpredictable ways. + */ + } } - // And finally, move up the parent chain *and* find a sibling - // (because this is depth-first traversal, we will have already - // seen the parent nodes themselves). - do { - $originalNode = $originalNode->getParent(); - } while ($originalNode && !$originalNode->node->nextSibling); + $articleContent = $this->prepArticle($articleContent); + + if ($hasContent) { + // Find out text direction from ancestors of final top candidate. + $ancestors = array_merge([$parentOfTopCandidate, $topCandidate], $parentOfTopCandidate->getNodeAncestors()); + foreach ($ancestors as $ancestor) { + $articleDir = $ancestor->getAttribute('dir'); + if ($articleDir) { + $this->setDirection($articleDir); + break; + } + } - return ($originalNode) ? new self($originalNode->node->nextSibling) : $originalNode; + return $articleContent; + } else { + return false; + } } /** - * Compares nodes. Checks for tag name and text content. - * - * It's a replacement of the original JS code, which looked like this: - * - * $node1 == $node2 + * Cleans up the final article. * - * I'm not sure this works the same in PHP, so I created a mock function to check the actual content of the node. - * Should serve the same porpuse as the original comparison. + * @param DOMDocument $article * - * @param Readability $node1 - * @param Readability $node2 - * - * @return bool + * @return DOMDocument */ - public function compareNodes($node1, $node2) + public function prepArticle(DOMDocument $article) { - if ($node1->getTagName() !== $node2->getTagName()) { - return false; + $this->_cleanStyles($article); + $this->_clean($article, 'style'); + + // Check for data tables before we continue, to avoid removing items in + // those tables, which will often be isolated even though they're + // visually linked to other content-ful elements (text, images, etc.). + $this->_markDataTables($article); + + // Clean out junk from the article content + $this->_cleanConditionally($article, 'form'); + $this->_cleanConditionally($article, 'fieldset'); + $this->_clean($article, 'object'); + $this->_clean($article, 'embed'); + $this->_clean($article, 'h1'); + $this->_clean($article, 'footer'); + + // Clean out elements have "share" in their id/class combinations from final top candidates, + // which means we don't remove the top candidates even they have "share". + foreach ($article->childNodes as $child) { + $this->_cleanMatchedNodes($child, '/share/i'); } - if ($node1->getTextContent(true) !== $node2->getTextContent(true)) { - return false; + /* + * If there is only one h2 and its text content substantially equals article title, + * they are probably using it as a header and not a subheader, + * so remove it since we already extract the title separately. + */ + $h2 = $article->getElementsByTagName('h2'); + if ($h2->length === 1) { + $lengthSimilarRate = (mb_strlen($h2->item(0)->textContent) - mb_strlen($this->getTitle())) / max(mb_strlen($this->getTitle()), 1); + + if (abs($lengthSimilarRate) < 0.5) { + if ($lengthSimilarRate > 0) { + $titlesMatch = strpos($h2->item(0)->textContent, $this->getTitle()) !== false; + } else { + $titlesMatch = strpos($this->getTitle(), $h2->item(0)->textContent) !== false; + } + if ($titlesMatch) { + $this->_clean($article, 'h2'); + } + } } - return true; + $this->_clean($article, 'iframe'); + $this->_clean($article, 'input'); + $this->_clean($article, 'textarea'); + $this->_clean($article, 'select'); + $this->_clean($article, 'button'); + $this->_cleanHeaders($article); + + // Do these last as the previous stuff may have removed junk + // that will affect these + $this->_cleanConditionally($article, 'table'); + $this->_cleanConditionally($article, 'ul'); + $this->_cleanConditionally($article, 'div'); + + $this->_cleanExtraParagraphs($article); + + foreach (iterator_to_array($article->getElementsByTagName('br')) as $br) { + $next = $br->nextSibling; + if ($next && $next->nodeName === 'p') { + $br->parentNode->removeChild($br); + } + } + + return $article; } /** - * Replaces child node with a new one. + * Look for 'data' (as opposed to 'layout') tables, for which we use + * similar checks as + * https://dxr.mozilla.org/mozilla-central/rev/71224049c0b52ab190564d3ea0eab089a159a4cf/accessible/html/HTMLTableAccessible.cpp#920. * - * @param Readability $newNode + * @param DOMDocument $article + * + * @return void */ - public function replaceChild(self $newNode) + public function _markDataTables(DOMDocument $article) { - $this->node->parentNode->replaceChild($newNode->node, $this->node); + $tables = $article->getElementsByTagName('table'); + foreach ($tables as $table) { + /** @var DOMElement $table */ + $role = $table->getAttribute('role'); + if ($role === 'presentation') { + $table->setReadabilityDataTable(false); + continue; + } + $datatable = $table->getAttribute('datatable'); + if ($datatable == '0') { + $table->setReadabilityDataTable(false); + continue; + } + $summary = $table->getAttribute('summary'); + if ($summary) { + $table->setReadabilityDataTable(true); + continue; + } + + $caption = $table->getElementsByTagName('caption'); + if ($caption->length > 0 && $caption->item(0)->childNodes->length > 0) { + $table->setReadabilityDataTable(true); + continue; + } + + // If the table has a descendant with any of these tags, consider a data table: + foreach (['col', 'colgroup', 'tfoot', 'thead', 'th'] as $dataTableDescendants) { + if ($table->getElementsByTagName($dataTableDescendants)->length > 0) { + $table->setReadabilityDataTable(true); + continue 2; + } + } + + // Nested tables indicate a layout table: + if ($table->getElementsByTagName('table')->length > 0) { + $table->setReadabilityDataTable(false); + continue; + } + + $sizeInfo = $table->getRowAndColumnCount(); + if ($sizeInfo['rows'] >= 10 || $sizeInfo['columns'] > 4) { + $table->setReadabilityDataTable(true); + continue; + } + // Now just go by size entirely: + $table->setReadabilityDataTable($sizeInfo['rows'] * $sizeInfo['columns'] > 10); + } } /** - * Creates a new node based on the text content of the original node. - * - * @param Readability $originalNode - * @param string $tagName + * Remove the style attribute on every e and under. * - * @return Readability - */ - public function createNode(self $originalNode, $tagName) + * @param $node DOMDocument|DOMNode + **/ + public function _cleanStyles($node) { - $text = $originalNode->getTextContent(); - $newNode = $originalNode->node->ownerDocument->createElement($tagName, $text); + if (property_exists($node, 'tagName') && $node->tagName === 'svg') { + return; + } + + // Do not bother if there's no method to remove an attribute + if (method_exists($node, 'removeAttribute')) { + $presentational_attributes = ['align', 'background', 'bgcolor', 'border', 'cellpadding', 'cellspacing', 'frame', 'hspace', 'rules', 'style', 'valign', 'vspace']; + // Remove `style` and deprecated presentational attributes + foreach ($presentational_attributes as $presentational_attribute) { + $node->removeAttribute($presentational_attribute); + } - return new static($newNode); + $deprecated_size_attribute_elems = ['table', 'th', 'td', 'hr', 'pre']; + if (property_exists($node, 'tagName') && in_array($node->tagName, $deprecated_size_attribute_elems)) { + $node->removeAttribute('width'); + $node->removeAttribute('height'); + } + } + + $cur = $node->firstChild; + while ($cur !== null) { + $this->_cleanStyles($cur); + $cur = $cur->nextSibling; + } } /** - * Checks if the object is initialized. + * Clean out elements whose id/class combinations match specific string. * - * @return bool - */ - public function isInitialized() + * @param $node DOMElement Node to clean + * @param $regex string Match id/class combination. + * + * @return void + **/ + public function _cleanMatchedNodes($node, $regex) { - return $this->initialized; + $endOfSearchMarkerNode = NodeUtility::getNextNode($node, true); + $next = NodeUtility::getNextNode($node); + while ($next && $next !== $endOfSearchMarkerNode) { + if (preg_match($regex, sprintf('%s %s', $next->getAttribute('class'), $next->getAttribute('id')))) { + $next = NodeUtility::removeAndGetNext($next); + } else { + $next = NodeUtility::getNextNode($next); + } + } } /** - * Reloads the score stores in the data-readability tag. + * @param DOMDocument $article * - * @return int|bool + * @return void */ - public function reloadScore() + public function _cleanExtraParagraphs(DOMDocument $article) { - if (method_exists($this->node, 'getAttribute')) { - if ($this->node->hasAttribute('data-readability')) { - $this->initialized = true; - $score = $this->node->getAttribute('data-readability'); - $this->setContentScore($score); + $paragraphs = $article->getElementsByTagName('p'); + $length = $paragraphs->length; - return $score; + for ($i = 0; $i < $length; $i++) { + $paragraph = $paragraphs->item($length - 1 - $i); + + $imgCount = $paragraph->getElementsByTagName('img')->length; + $embedCount = $paragraph->getElementsByTagName('embed')->length; + $objectCount = $paragraph->getElementsByTagName('object')->length; + // At this point, nasty iframes have been removed, only remain embedded video ones. + $iframeCount = $paragraph->getElementsByTagName('iframe')->length; + $totalCount = $imgCount + $embedCount + $objectCount + $iframeCount; + + if ($totalCount === 0 && !preg_replace(NodeUtility::$regexps['onlyWhitespace'], '', $paragraph->textContent)) { + $paragraph->parentNode->removeChild($paragraph); } } - - return false; } /** - * Check if a given node has one of its ancestor tag name matching the - * provided one. - * - * @param Readability $node - * @param string $tagName - * @param int $maxDepth + * @param DOMDocument $article * - * @return bool + * @return void */ - public function hasAncestorTag(self $node, $tagName, $maxDepth = 3) + public function _cleanConditionally(DOMDocument $article, $tag) { - $depth = 0; - while ($node->getParent()) { - if ($maxDepth > 0 && $depth > $maxDepth) { - return false; + if (!$this->configuration->getCleanConditionally()) { + return; + } + + $isList = in_array($tag, ['ul', 'ol']); + + /* + * Gather counts for other typical elements embedded within. + * Traverse backwards so we can remove nodes at the same time + * without effecting the traversal. + */ + + $DOMNodeList = $article->getElementsByTagName($tag); + $length = $DOMNodeList->length; + for ($i = 0; $i < $length; $i++) { + /** @var $node DOMElement */ + $node = $DOMNodeList->item($length - 1 - $i); + + // First check if we're in a data table, in which case don't remove us. + if ($node->hasAncestorTag($node, 'table', -1) && $node->isReadabilityDataTable()) { + continue; } - if ($node->getParent()->tagNameEqualsTo($tagName)) { - return true; + + $weight = 0; + if ($this->configuration->getWeightClasses()) { + $weight = $node->getClassWeight(); } - $node = $node->getParent(); - $depth++; - } - return false; + if ($weight < 0) { + NodeUtility::removeNode($node); + continue; + } + + if (substr_count($node->getTextContent(), ',') < 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 = $node->getElementsByTagName('p')->length; + $img = $node->getElementsByTagName('img')->length; + $li = $node->getElementsByTagName('li')->length - 100; + $input = $node->getElementsByTagName('input')->length; + + $embedCount = 0; + $embeds = $node->getElementsByTagName('embed'); + + foreach ($embeds as $embedNode) { + if (preg_match(NodeUtility::$regexps['videos'], $embedNode->C14N())) { + $embedCount++; + } + } + + $linkDensity = $node->getLinkDensity(); + $contentLength = mb_strlen($node->getTextContent(true)); + + $haveToRemove = + ($img > 1 && $p / $img < 0.5 && !$node->hasAncestorTag($node, 'figure')) || + (!$isList && $li > $p) || + ($input > floor($p / 3)) || + (!$isList && $contentLength < 25 && ($img === 0 || $img > 2) && !$node->hasAncestorTag($node, 'figure')) || + (!$isList && $weight < 25 && $linkDensity > 0.2) || + ($weight >= 25 && $linkDensity > 0.5) || + (($embedCount === 1 && $contentLength < 75) || $embedCount > 1); + + if ($haveToRemove) { + NodeUtility::removeNode($node); + } + } + } } /** - * Returns the children of the current node. + * Clean a node of all elements of type "tag". + * (Unless it's a youtube/vimeo video. People love movies.). * - * @param bool $filterEmptyDOMText Filter empty DOMText nodes? + * @param $article DOMDocument + * @param $tag string tag to clean * - * @return array - */ - public function getChildren($filterEmptyDOMText = false) + * @return void + **/ + public function _clean(DOMDocument $article, $tag) { - $ret = []; - /** @var \DOMNode $node */ - foreach ($this->node->childNodes as $node) { - if ($filterEmptyDOMText && $node->nodeName === '#text' && !trim($node->nodeValue)) { - continue; + $isEmbed = in_array($tag, ['object', 'embed', 'iframe']); + + $DOMNodeList = $article->getElementsByTagName($tag); + $length = $DOMNodeList->length; + for ($i = 0; $i < $length; $i++) { + $item = $DOMNodeList->item($length - 1 - $i); + + // Allow youtube and vimeo videos through as people usually want to see those. + if ($isEmbed) { + $attributeValues = []; + foreach ($item->attributes as $name => $value) { + $attributeValues[] = $value->nodeValue; + } + $attributeValues = implode('|', $attributeValues); + + // First, check the elements attributes to see if any of them contain youtube or vimeo + if (preg_match(NodeUtility::$regexps['videos'], $attributeValues)) { + continue; + } + + // Then check the elements inside this element for the same. + if (preg_match(NodeUtility::$regexps['videos'], $item->C14N())) { + continue; + } } - - $ret[] = new static($node); + NodeUtility::removeNode($item); } + } - return $ret; + /** + * Clean out spurious headers from an Element. Checks things like classnames and link density. + * + * @param DOMDocument $article + * + * @return void + **/ + public function _cleanHeaders(DOMDocument $article) + { + for ($headerIndex = 1; $headerIndex < 3; $headerIndex++) { + $headers = $article->getElementsByTagName('h' . $headerIndex); + /** @var $header DOMElement */ + foreach ($headers as $header) { + $weight = 0; + if ($this->configuration->getWeightClasses()) { + $weight = $header->getClassWeight(); + } + + if ($weight < 0) { + NodeUtility::removeNode($header); + } + } + } } /** - * Determines if a node has no content or it is just a bunch of dividing lines and/or whitespace. + * @param DOMDocument $article * - * @return bool + * @return DOMDocument */ - public function isElementWithoutContent() + public function postProcessContent(DOMDocument $article) { - return $this->node instanceof \DOMElement && - // /\x{00A0}|\s+/u TODO to be replaced with regexps array - mb_strlen(preg_replace('/\x{00A0}|\s+/u', '', $this->node->textContent)) === 0 && - ($this->node->childNodes->length === 0 || - $this->node->childNodes->length === $this->node->getElementsByTagName('br')->length + $this->node->getElementsByTagName('hr')->length + // Readability cannot open relative uris so we convert them to absolute uris. + if ($this->configuration->getFixRelativeURLs()) { + foreach (iterator_to_array($article->getElementsByTagName('a')) as $link) { + /** @var DOMElement $link */ + $href = $link->getAttribute('href'); + if ($href) { + // Replace links with javascript: URIs with text content, since + // they won't work after scripts have been removed from the page. + if (strpos($href, 'javascript:') === 0) { + $text = $article->createTextNode($link->textContent); + $link->parentNode->replaceChild($text, $link); + } else { + $link->setAttribute('href', $this->toAbsoluteURI($href)); + } + } + } + + foreach ($article->getElementsByTagName('img') as $img) { + /** @var DOMElement $img */ /* - * Special DOMDocument case: We also need to count how many DOMText we have inside the node. - * If there's an empty tag with an space inside and a BR (for example "<p> <br/></p>) counting only BRs and - * HRs will will say that the example has 2 nodes, instead of one. This happens because in DOMDocument, - * DOMTexts are also nodes (which doesn't happen in JS). So we need to also count how many DOMText we - * are dealing with (And at this point we know they are empty or are just whitespace, because of the - * mb_strlen in this chain of checks). + * Extract all possible sources of img url and select the first one on the list. */ - + count(array_filter(iterator_to_array($this->node->childNodes), function ($child) { - return $child instanceof \DOMText; - })) + $url = [ + $img->getAttribute('src'), + $img->getAttribute('data-original'), + $img->getAttribute('data-url') + ]; + + $src = array_filter($url); + $src = reset($src); + if ($src) { + $img->setAttribute('src', $this->toAbsoluteURI($src)); + } + } + } + + return $article; + } - ); + /** + * @return null|string + */ + public function __toString() + { + return sprintf('<h1>%s</h1>%s', $this->getTitle(), $this->getContent()); + } + + /** + * @return string|null + */ + public function getTitle() + { + return $this->title; + } + + /** + * @param string $title + */ + protected function setTitle($title) + { + $this->title = $title; + } + + /** + * @return string|null + */ + public function getContent() + { + return $this->content; + } + + /** + * @param string $content + */ + protected function setContent($content) + { + $this->content = $content; + } + + /** + * @return null|string + */ + public function getExcerpt() + { + return $this->excerpt; + } + + /** + * @param null|string $excerpt + */ + public function setExcerpt($excerpt) + { + $this->excerpt = $excerpt; + } + + /** + * @return string|null + */ + public function getImage() + { + return $this->image; + } + + /** + * @param string $image + */ + protected function setImage($image) + { + $this->image = $image; + } + + /** + * @return string|null + */ + public function getAuthor() + { + return $this->author; + } + + /** + * @param string $author + */ + protected function setAuthor($author) + { + $this->author = $author; + } + + /** + * @return null|string + */ + public function getDirection() + { + return $this->direction; + } + + /** + * @param null|string $direction + */ + public function setDirection($direction) + { + $this->direction = $direction; } } diff --git a/src/ReadabilityInterface.php b/src/ReadabilityInterface.php deleted file mode 100644 index 0dee01b..0000000 --- a/src/ReadabilityInterface.php +++ /dev/null @@ -1,92 +0,0 @@ -<?php - -namespace andreskrey\Readability; - -use League\HTMLToMarkdown\ElementInterface; - -interface ReadabilityInterface extends ElementInterface -{ - /** - * @param string $value - * - * @return bool - */ - public function tagNameEqualsTo($value); - - /** - * @return int - */ - public function getNodeAncestors(); - - /** - * @return Readability|null - */ - public function getAllLinks(); - - /** - * @return int - */ - public function getContentScore(); - - /** - * @return Readability - */ - public function initializeNode(); - - /** - * @return int - */ - public function getClassWeight(); - - /** - * @param int $score - * - * @return int - */ - public function setContentScore($score); - - /** - * @param bool $normalize Normalize white space? - * - * @return string - */ - public function getTextContent($normalize); - - /** - * @param string $value - */ - public function setNodeTag($value); - - /** - * @return \DOMNode - */ - public function getDOMNode(); - - /** - * @param Readability $node - * - * @return Readability - */ - public function removeAndGetNext($node); - - /** - * @param Readability $originalNode - * @param bool $ignoreSelfAndKids - * - * @return Readability - */ - public function getNextNode($originalNode, $ignoreSelfAndKids = false); - - /** - * @param Readability $node1 - * @param Readability $node2 - * - * @return bool - */ - public function compareNodes($node1, $node2); - - /** - * @param Readability $newNode - */ - public function replaceChild(Readability $newNode); -} |