infong/php-readability
1
0
Fork 0
mirror of https://github.com/j0k3r/php-readability.git synced 2026-09-27 06:26:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Cleanup

Browse Source
This commit is contained in:
Jeremy Benoist
2016-02-29 10:22:07 +01:00
parent dc590542f0
commit 850ade16b6
1 changed files with 17 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/Readability.php
+17 -17
View File
@@ -69,7 +69,7 @@ class Readability
* Defined up here so we don't instantiate them repeatedly in loops. * Defined up here so we don't instantiate them repeatedly in loops.
*/ */
public $regexps = array( public $regexps = array(
'unlikelyCandidates' => '/display\s*:\s*none|ignore|\binfo|annoy|clock|date|time|author|intro|links|hidd?e|about|archive|\bprint|bookmark|tags|share|search|social|robot|published|combx|comment|mast(?:head)|subscri|community|category|disqus|extra|head|head(?:er|note)|floor|foot(?:er|note)|menu|tool|function|nav|remark|rss|shoutbox|tool|widget|meta|banner|sponsor|adsense|inner-?ad|ad-|sponsor|\badv\b|\bads\b|agr?egate?|pager|sidebar|popup|tweet|twitter/i', 'unlikelyCandidates' => '/display\s*:\s*none|ignore|\binfo|annoy|clock|date|time|author|intro|links|hidd?e|about|archive|\bprint|bookmark|tags|tag-list|share|search|social|robot|published|combx|comment|mast(?:head)|subscri|community|category|disqus|extra|head|head(?:er|note)|floor|foot(?:er|note)|menu|tool|function|nav|remark|rss|shoutbox|tool|widget|meta|banner|sponsor|adsense|inner-?ad|ad-|sponsor|\badv\b|\bads\b|agr?egate?|pager|sidebar|popup|tweet|twitter/i',
'okMaybeItsACandidate' => '/article\b|contain|\bcontent|column|general|detail|shadow|lightbox|blog|body|entry|main|page/i', 'okMaybeItsACandidate' => '/article\b|contain|\bcontent|column|general|detail|shadow|lightbox|blog|body|entry|main|page/i',
'positive' => '/read|full|article|body|\bcontent|contain|entry|main|markdown|page|attach|pagination|post|text|blog|story/i', 'positive' => '/read|full|article|body|\bcontent|contain|entry|main|markdown|page|attach|pagination|post|text|blog|story/i',
'negative' => '/bottom|stat|info|discuss|e[\-]?mail|comment|reply|log.{2}(n|ed)|sign|single|combx|com-|contact|_nav|link|media|\bout|promo|\bad-|related|scroll|shoutbox|sidebar|sponsor|shopping|teaser|recommend/i', 'negative' => '/bottom|stat|info|discuss|e[\-]?mail|comment|reply|log.{2}(n|ed)|sign|single|combx|com-|contact|_nav|link|media|\bout|promo|\bad-|related|scroll|shoutbox|sidebar|sponsor|shopping|teaser|recommend/i',
@@ -789,6 +789,7 @@ class Readability
$grandParentNode->getAttributeNode('readability')->value += $contentScore / self::GRANDPARENT_SCORE_DIVISOR; $grandParentNode->getAttributeNode('readability')->value += $contentScore / self::GRANDPARENT_SCORE_DIVISOR;
} }
} }
/* /*
* Node prepping: trash nodes that look cruddy (like ones with the class name "comment", etc). * Node prepping: trash nodes that look cruddy (like ones with the class name "comment", etc).
* This is faster to do before scoring but safer after. * This is faster to do before scoring but safer after.
@@ -800,7 +801,7 @@ class Readability
$node = $candidates->item($c); $node = $candidates->item($c);
// node should be readable but not inside of an article otherwise it's probably non-readable block // node should be readable but not inside of an article otherwise it's probably non-readable block
if ($node->hasAttribute('readability') && (int) $node->getAttributeNode('readability')->value < 40 && ($node->parentNode ? strcasecmp($node->parentNode->tagName, 'article') !== 0 : true)) { if ($node->hasAttribute('readability') && (int) $node->getAttributeNode('readability')->value < 40 && ($node->parentNode ? strcasecmp($node->parentNode->tagName, 'article') !== 0 : true)) {
$this->dbg('Removing unlikely candidate '.$node->getNodePath().' by "'.$node->tagName.'" with readability '.($node->hasAttribute('readability') ? (int) $node->getAttributeNode('readability')->value : 0)); $this->dbg('Removing unlikely candidate (using note) '.$node->getNodePath().' by "'.$node->tagName.'" with readability '.($node->hasAttribute('readability') ? (int) $node->getAttributeNode('readability')->value : 0));
$node->parentNode->removeChild($node); $node->parentNode->removeChild($node);
} }
} }
@@ -809,15 +810,15 @@ class Readability
for ($node = null, $c = $candidates->length - 1; $c >= 0; --$c) { for ($node = null, $c = $candidates->length - 1; $c >= 0; --$c) {
$node = $candidates->item($c); $node = $candidates->item($c);
$tagName = $node->tagName;
/* Remove unlikely candidates */ // Remove unlikely candidates
$unlikelyMatchString = $node->getAttribute('class').' '.$node->getAttribute('id').' '.$node->getAttribute('style'); $unlikelyMatchString = $node->getAttribute('class').' '.$node->getAttribute('id').' '.$node->getAttribute('style');
//$this->dbg('Processing '.$node->getNodePath().' by "'. $unlikelyMatchString.'" with readability '.($node->hasAttribute('readability') ? (int)$node->getAttributeNode('readability')->value : 0));
if (mb_strlen($unlikelyMatchString) > 3 && // don't process "empty" strings if (mb_strlen($unlikelyMatchString) > 3 && // don't process "empty" strings
preg_match($this->regexps['unlikelyCandidates'], $unlikelyMatchString) && preg_match($this->regexps['unlikelyCandidates'], $unlikelyMatchString) &&
!preg_match($this->regexps['okMaybeItsACandidate'], $unlikelyMatchString) !preg_match($this->regexps['okMaybeItsACandidate'], $unlikelyMatchString)
) { ) {
$this->dbg('Removing unlikely candidate '.$node->getNodePath().' by "'.$unlikelyMatchString.'" with readability '.($node->hasAttribute('readability') ? (int) $node->getAttributeNode('readability')->value : 0)); $this->dbg('Removing unlikely candidate (using conf) '.$node->getNodePath().' by "'.$unlikelyMatchString.'" with readability '.($node->hasAttribute('readability') ? (int) $node->getAttributeNode('readability')->value : 0));
$node->parentNode->removeChild($node); $node->parentNode->removeChild($node);
--$nodeIndex; --$nodeIndex;
} }
@@ -934,9 +935,8 @@ class Readability
$nodeContent = $this->getInnerText($siblingNode, true, true); $nodeContent = $this->getInnerText($siblingNode, true, true);
$nodeLength = mb_strlen($nodeContent); $nodeLength = mb_strlen($nodeContent);
if ($nodeLength > self::MIN_NODE_LENGTH && $linkDensity < self::MAX_LINK_DENSITY) { if (($nodeLength > self::MIN_NODE_LENGTH && $linkDensity < self::MAX_LINK_DENSITY)
$append = true; || ($nodeLength < self::MIN_NODE_LENGTH && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent))) {
} elseif ($nodeLength < self::MIN_NODE_LENGTH && $linkDensity === 0 && preg_match('/\.( |$)/', $nodeContent)) {
$append = true; $append = true;
} }
} }
@@ -946,15 +946,15 @@ class Readability
$nodeToAppend = null; $nodeToAppend = null;
if (strcasecmp($siblingNodeName, 'div') !== 0 && strcasecmp($siblingNodeName, 'p') !== 0) { if (strcasecmp($siblingNodeName, 'div') !== 0 && strcasecmp($siblingNodeName, 'p') !== 0) {
/* 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. */ // We have a node that isn't a common block level element, like a form or td tag. Turn it into a div so it doesn't get filtered out later by accident.
$this->dbg('Altering siblingNode '.$siblingNodeName.' to div.'); $this->dbg('Altering siblingNode "'.$siblingNodeName.'" to "div".');
$nodeToAppend = $this->dom->createElement('div'); $nodeToAppend = $this->dom->createElement('div');
try { try {
$nodeToAppend->setAttribute('alt', $siblingNodeName); $nodeToAppend->setAttribute('alt', $siblingNodeName);
$nodeToAppend->innerHTML = $siblingNode->innerHTML; $nodeToAppend->innerHTML = $siblingNode->innerHTML;
} catch (Exception $e) { } catch (Exception $e) {
$this->dbg('Could not alter siblingNode '.$siblingNodeName.' to div, reverting to original.'); $this->dbg('Could not alter siblingNode "'.$siblingNodeName.'" to "div", reverting to original.');
$nodeToAppend = $siblingNode; $nodeToAppend = $siblingNode;
--$s; --$s;
--$sl; --$sl;
@@ -1161,9 +1161,9 @@ class Readability
} }
$weight = 0; $weight = 0;
/* Look for a special classname */ // Look for a special classname
$weight += $this->weightAttribute($e, 'class'); $weight += $this->weightAttribute($e, 'class');
/* Look for a special ID */ // Look for a special ID
$weight += $this->weightAttribute($e, 'id'); $weight += $this->weightAttribute($e, 'id');
return $weight; return $weight;
@@ -1196,18 +1196,18 @@ class Readability
$isEmbed = ($tag === 'audio' || $tag === 'video' || $tag === 'iframe' || $tag === 'object' || $tag === 'embed'); $isEmbed = ($tag === 'audio' || $tag === 'video' || $tag === 'iframe' || $tag === 'object' || $tag === 'embed');
for ($cur_item = null, $y = $targetList->length - 1; $y >= 0; --$y) { for ($cur_item = null, $y = $targetList->length - 1; $y >= 0; --$y) {
/* Allow youtube and vimeo videos through as people usually want to see those. */ // Allow youtube and vimeo videos through as people usually want to see those.
$cur_item = $targetList->item($y); $cur_item = $targetList->item($y);
if ($isEmbed) { if ($isEmbed) {
$attributeValues = $cur_item->getAttribute('src').' '.$cur_item->getAttribute('href'); $attributeValues = $cur_item->getAttribute('src').' '.$cur_item->getAttribute('href');
/* First, check the elements attributes to see if any of them contain known media hosts */ // First, check the elements attributes to see if any of them contain known media hosts
if (preg_match($this->regexps['media'], $attributeValues)) { if (preg_match($this->regexps['media'], $attributeValues)) {
continue; continue;
} }
/* Then check the elements inside this element for the same. */ // Then check the elements inside this element for the same.
if (preg_match($this->regexps['media'], $targetList->item($y)->innerHTML)) { if (preg_match($this->regexps['media'], $targetList->item($y)->innerHTML)) {
continue; continue;
} }