diff options
Diffstat (limited to 'src')
25 files changed, 151 insertions, 140 deletions
diff --git a/src/NodeClass/DOMAttr.php b/src/NodeClass/DOMAttr.php deleted file mode 100644 index ea8672d..0000000 --- a/src/NodeClass/DOMAttr.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMAttr extends \DOMAttr -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMCdataSection.php b/src/NodeClass/DOMCdataSection.php deleted file mode 100644 index 438ac99..0000000 --- a/src/NodeClass/DOMCdataSection.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMCdataSection extends \DOMCdataSection -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMCharacterData.php b/src/NodeClass/DOMCharacterData.php deleted file mode 100644 index 480980e..0000000 --- a/src/NodeClass/DOMCharacterData.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMCharacterData extends \DOMCharacterData -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMComment.php b/src/NodeClass/DOMComment.php deleted file mode 100644 index 416460b..0000000 --- a/src/NodeClass/DOMComment.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMComment extends \DOMComment -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMDocumentFragment.php b/src/NodeClass/DOMDocumentFragment.php deleted file mode 100644 index cc8b753..0000000 --- a/src/NodeClass/DOMDocumentFragment.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMDocumentFragment extends \DOMDocumentFragment -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMDocumentType.php b/src/NodeClass/DOMDocumentType.php deleted file mode 100644 index 13f7829..0000000 --- a/src/NodeClass/DOMDocumentType.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMDocumentType extends \DOMDocumentType -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMElement.php b/src/NodeClass/DOMElement.php deleted file mode 100644 index a7dc36a..0000000 --- a/src/NodeClass/DOMElement.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMElement extends \DOMElement -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMNotation.php b/src/NodeClass/DOMNotation.php deleted file mode 100644 index 3e09bbc..0000000 --- a/src/NodeClass/DOMNotation.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMNotation extends \DOMNotation -{ - use NodeClassTrait; -} diff --git a/src/NodeClass/DOMText.php b/src/NodeClass/DOMText.php deleted file mode 100644 index 80ef6c8..0000000 --- a/src/NodeClass/DOMText.php +++ /dev/null @@ -1,8 +0,0 @@ -<?php - -namespace andreskrey\Readability\NodeClass; - -class DOMText extends \DOMText -{ - use NodeClassTrait; -} diff --git a/src/Nodes/DOMAttr.php b/src/Nodes/DOMAttr.php new file mode 100644 index 0000000..c31517a --- /dev/null +++ b/src/Nodes/DOMAttr.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMAttr extends \DOMAttr +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMCdataSection.php b/src/Nodes/DOMCdataSection.php new file mode 100644 index 0000000..f3a56f0 --- /dev/null +++ b/src/Nodes/DOMCdataSection.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMCdataSection extends \DOMCdataSection +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMCharacterData.php b/src/Nodes/DOMCharacterData.php new file mode 100644 index 0000000..e5087d9 --- /dev/null +++ b/src/Nodes/DOMCharacterData.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMCharacterData extends \DOMCharacterData +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMComment.php b/src/Nodes/DOMComment.php new file mode 100644 index 0000000..fd2b8b5 --- /dev/null +++ b/src/Nodes/DOMComment.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMComment extends \DOMComment +{ + use NodeTrait; +} diff --git a/src/NodeClass/DOMDocument.php b/src/Nodes/DOMDocument.php index 98b1215..510cdf2 100644 --- a/src/NodeClass/DOMDocument.php +++ b/src/Nodes/DOMDocument.php @@ -1,10 +1,10 @@ <?php -namespace andreskrey\Readability\NodeClass; +namespace andreskrey\Readability\Nodes; class DOMDocument extends \DOMDocument { - use NodeClassTrait; + use NodeTrait; public function __construct($version, $encoding) { diff --git a/src/Nodes/DOMDocumentFragment.php b/src/Nodes/DOMDocumentFragment.php new file mode 100644 index 0000000..d5f013e --- /dev/null +++ b/src/Nodes/DOMDocumentFragment.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMDocumentFragment extends \DOMDocumentFragment +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMDocumentType.php b/src/Nodes/DOMDocumentType.php new file mode 100644 index 0000000..81e426b --- /dev/null +++ b/src/Nodes/DOMDocumentType.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMDocumentType extends \DOMDocumentType +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMElement.php b/src/Nodes/DOMElement.php new file mode 100644 index 0000000..6ca0a29 --- /dev/null +++ b/src/Nodes/DOMElement.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMElement extends \DOMElement +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMNode.php b/src/Nodes/DOMNode.php new file mode 100644 index 0000000..79a352b --- /dev/null +++ b/src/Nodes/DOMNode.php @@ -0,0 +1,14 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +/** + * Class DOMNode + * + * @method getAttribute($attribute) + * @package andreskrey\Readability\Nodes + */ +class DOMNode extends \DOMNode +{ + use NodeTrait; +} diff --git a/src/Nodes/DOMNotation.php b/src/Nodes/DOMNotation.php new file mode 100644 index 0000000..a4802e0 --- /dev/null +++ b/src/Nodes/DOMNotation.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMNotation extends \DOMNotation +{ + use NodeTrait; +} diff --git a/src/NodeClass/DOMProcessingInstruction.php b/src/Nodes/DOMProcessingInstruction.php index 0c615c6..bd80997 100644 --- a/src/NodeClass/DOMProcessingInstruction.php +++ b/src/Nodes/DOMProcessingInstruction.php @@ -1,8 +1,8 @@ <?php -namespace andreskrey\Readability\NodeClass; +namespace andreskrey\Readability\Nodes; class DOMProcessingInstruction extends \DOMProcessingInstruction { - use NodeClassTrait; + use NodeTrait; } diff --git a/src/Nodes/DOMText.php b/src/Nodes/DOMText.php new file mode 100644 index 0000000..43d2ba9 --- /dev/null +++ b/src/Nodes/DOMText.php @@ -0,0 +1,8 @@ +<?php + +namespace andreskrey\Readability\Nodes; + +class DOMText extends \DOMText +{ + use NodeTrait; +} diff --git a/src/NodeClass/NodeClassTrait.php b/src/Nodes/NodeTrait.php index ae3eeb1..5847178 100644 --- a/src/NodeClass/NodeClassTrait.php +++ b/src/Nodes/NodeTrait.php @@ -1,10 +1,8 @@ <?php -namespace andreskrey\Readability\NodeClass; +namespace andreskrey\Readability\Nodes; -use andreskrey\Readability\NodeUtility; - -trait NodeClassTrait +trait NodeTrait { /** * Content score of the node. Used to determine the value of the content. diff --git a/src/NodeUtility.php b/src/Nodes/NodeUtility.php index 7fbdd45..8938a49 100644 --- a/src/NodeUtility.php +++ b/src/Nodes/NodeUtility.php @@ -1,10 +1,6 @@ <?php -namespace andreskrey\Readability; - -use andreskrey\Readability\NodeClass\DOMDocument; -use andreskrey\Readability\NodeClass\DOMElement; -use andreskrey\Readability\NodeClass\DOMNode; +namespace andreskrey\Readability\Nodes; /** * Class NodeUtility. diff --git a/src/ParseException.php b/src/ParseException.php new file mode 100644 index 0000000..335851f --- /dev/null +++ b/src/ParseException.php @@ -0,0 +1,9 @@ +<?php + +namespace andreskrey\Readability; + + +class ParseException extends \Exception +{ + +} diff --git a/src/Readability.php b/src/Readability.php index db0774a..60a314a 100644 --- a/src/Readability.php +++ b/src/Readability.php @@ -2,10 +2,11 @@ namespace andreskrey\Readability; -use andreskrey\Readability\NodeClass\DOMDocument; -use andreskrey\Readability\NodeClass\DOMElement; -use andreskrey\Readability\NodeClass\DOMNode; -use andreskrey\Readability\NodeClass\DOMText; +use andreskrey\Readability\Nodes\DOMDocument; +use andreskrey\Readability\Nodes\DOMElement; +use andreskrey\Readability\Nodes\DOMNode; +use andreskrey\Readability\Nodes\DOMText; +use andreskrey\Readability\Nodes\NodeUtility; /** * Class Readability. @@ -101,15 +102,57 @@ class Readability public function __construct(Configuration $configuration) { $this->configuration = $configuration; + } + /** + * 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) + { // To avoid having 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; } /** * Main parse function. * * @param $html + * @throws ParseException * * @return array|bool */ @@ -123,10 +166,9 @@ class Readability // Checking for minimum HTML to work with. if (!($root = $this->dom->getElementsByTagName('body')->item(0)) || !$root->firstChild) { - return false; + throw new ParseException('Invalid or incomplete HTML.'); } - $parseSuccessful = true; while (true) { $root = $root->firstChild; @@ -142,8 +184,6 @@ class Readability * 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); @@ -159,18 +199,13 @@ class Readability } elseif ($this->configuration->getCleanConditionally()) { $this->configuration->setCleanConditionally(false); } else { - $parseSuccessful = false; - break; + throw new ParseException('Could not parse text.'); } } else { break; } } - if (!$parseSuccessful) { - return false; - } - $result = $this->postProcessContent($result); $this->setContent($result->C14N()); @@ -179,47 +214,6 @@ class Readability } /** - * 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->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; - } - - /** * Tries to guess relevant info from metadata of the html. Sets the results in the Readability properties. */ private function getMetadata() |