boyska/rss-bridge
1
0
Fork 0
forked from blallo/rss-bridge
Code Releases Activity

Merge pull request #332 from LogMANOriginal/WordPressBridge

Browse source 
Improve WordPress bridge and remove LeMotDuJour + Raymond (test before merge!)
This commit is contained in:
Mitsu 2016-08-05 01:59:16 +02:00 committed by GitHub
commit da1e32267b
3 changed files with 94 additions and 143 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

55
bridges/LeMotDuJourBridge.php
View file

@ -1,55 +0,0 @@
<?php
class LeMotDuJourBridge extends BridgeAbstract{
public function loadMetadatas() {
$this->maintainer = "qwertygc";
$this->name = "LeMotDuJour Bridge";
$this->uri = "http://www.lemotdujour.com/";
$this->description = "Returns the newest articles.";
$this->update = "2014-05-25";
}
public function collectData(array $param){
function StripCDATA($string) {
$string = str_replace('<![CDATA[', '', $string);
$string = str_replace(']]>', '', $string);
return $string;
}
function ExtractContent($url) {
$html2 = $this->file_get_html($url);
$text = $html2->find('div.single-contenu', 0)->innertext;
return $text;
}
$html = $this->file_get_html('http://feeds2.feedburner.com/lemotdujour/lemotdujour') or $this->returnError('Could not request LeMotDuJour.', 404);
$limit = 0;
foreach($html->find('item') as $element) {
if($limit < 10) {
$item = new \Item();
$item->title = StripCDATA($element->find('title', 0)->innertext);
$item->uri = StripCDATA($element->find('guid', 0)->plaintext);
$item->timestamp = strtotime($element->find('pubDate', 0)->plaintext);
$item->content = ExtractContent($item->uri);
$this->items[] = $item;
$limit++;
}
}
}
public function getName(){
return 'LeMotDuJour Bridge';
}
public function getURI(){
return 'http://lemotdujour.com/';
}
public function getCacheDuration(){
return 3600*2; // 2 hours
// return 0; // 2 hours
}
}

53
bridges/RaymondBridge.php
View file

@ -1,53 +0,0 @@
<?php
class RaymondBridge extends BridgeAbstract{
public function loadMetadatas() {
$this->maintainer = "pit-fgfjiudghdf";
$this->name = "Raymond";
$this->uri = "http://www.raymond.cc";
$this->description = "Returns the 3 newest posts from Raymond.cc (full text)";
$this->update = "2014-05-26";
}
public function collectData(array $param){
function raymondStripCDATA($string) {
$string = str_replace('<![CDATA[', '', $string);
$string = str_replace(']]>', '', $string);
return $string;
}
function raymondExtractContent($url) {
$html2 = $this->file_get_html($url);
$text = $html2->find('div.entry-content', 0)->innertext;
$text = preg_replace('/class="ad".*/', '', $text);
$text = strip_tags($text, '<p><a><i><strong><em><img>');
$text = str_replace('(adsbygoogle = window.adsbygoogle || []).push({});', '', $text);
return $text;
}
$html = $this->file_get_html('http://www.raymond.cc/blog/feed') or $this->returnError('Could not request raymond.', 404);
$limit = 0;
foreach($html->find('item') as $element) {
if($limit < 3) {
$item = new \Item();
$item->title = raymondStripCDATA($element->find('title', 0)->innertext);
$item->uri = raymondStripCDATA($element->find('guid', 0)->plaintext);
$item->timestamp = strtotime($element->find('pubDate', 0)->plaintext);
$item->content = raymondExtractContent($item->uri);
$this->items[] = $item;
$limit++;
}
}
}
public function getName(){
return 'raymond';
}
public function getURI(){
return 'http://www.raymond.cc/blog';
}
public function getCacheDuration(){
return 3600*12; // 12 hour
}
}

129
bridges/WordPressBridge.php
View file

@ -1,7 +1,10 @@
<?php <?php
define('WORDPRESS_TYPE_ATOM', 1); // Content is of type ATOM
define('WORDPRESS_TYPE_RSS', 2); // Content is of type RSS
class WordPressBridge extends BridgeAbstract { class WordPressBridge extends BridgeAbstract {
private $url; private $url;
public $sitename; // Name of the site
public function loadMetadatas() { public function loadMetadatas() {
@ -9,7 +12,7 @@ class WordPressBridge extends BridgeAbstract {
$this->name = "Wordpress Bridge"; $this->name = "Wordpress Bridge";
$this->uri = "https://wordpress.org/"; $this->uri = "https://wordpress.org/";
$this->description = "Returns the 3 newest full posts of a Wordpress blog"; $this->description = "Returns the 3 newest full posts of a Wordpress blog";
$this->update = "2016-08-02"; $this->update = "2016-08-04";
$this->parameters[] = $this->parameters[] =
'[ '[
@ -19,60 +22,118 @@ class WordPressBridge extends BridgeAbstract {
"identifier" : "url" "identifier" : "url"
} }
]'; ]';
}
// Returns the content type for a given html dom
function DetectContentType($html){
if($html->find('entry'))
return WORDPRESS_TYPE_ATOM;
if($html->find('item'))
return WORDPRESS_TYPE_RSS;
return WORDPRESS_TYPE_ATOM; // Make ATOM default
}
// Replaces all 'link' tags with 'url' for simplehtmldom to actually find 'links' ('url')
function ReplaceLinkTagsWithUrlTags($element){
// We need to fix the 'link' tag as simplehtmldom cannot parse it (just rename it and load back as dom)
$element_text = $element->outertext;
$element_text = str_replace('<link>', '<url>', $element_text);
$element_text = str_replace('</link>', '</url>', $element_text);
$element_text = str_replace('<link ', '<url ', $element_text);
return str_get_html($element_text);
}
function StripCDATA($string) {
$string = str_replace('<![CDATA[', '', $string);
$string = str_replace(']]>', '', $string);
return $string;
}
function ClearContent($content) {
$content = preg_replace('/<script[^>]*>[^<]*<\/script>/', '', $content);
$content = preg_replace('/<div class="wpa".*/', '', $content);
$content = preg_replace('/<form.*\/form>/', '', $content);
return $content;
} }
public function collectData(array $param) { public function collectData(array $param) {
function StripCDATA($string) {
$string = str_replace('<![CDATA[', '', $string);
$string = str_replace(']]>', '', $string);
return $string;
}
function clearContent($content) {
$content = preg_replace('/<script.*\/script>/', '', $content);
$content = preg_replace('/<div class="wpa".*/', '', $content);
return $content;
}
$this->processParams($param); $this->processParams($param);
if (!$this->hasUrl()) { if (!$this->hasUrl()) {
$this->returnError('You must specify a URL', 400); $this->returnError('You must specify a URL', 400);
} }
$this->url = $this->url.'/feed/atom'; $this->url = $this->url.'/feed/atom';
$html = $this->file_get_html($this->url) or $this->returnError("Could not request {$this->url}.", 404); $html = $this->file_get_html($this->url) or $this->returnError("Could not request {$this->url}.", 404);
$posts = $html->find('entry');
if(!empty($posts) ) { // Notice: We requested an ATOM feed, however some sites return RSS feeds instead!
$this->name = $html->find('title', 0)->plaintext; $type = $this->DetectContentType($html);
$i=0;
foreach ($html->find('entry') as $article) { if($type === WORDPRESS_TYPE_RSS)
$posts = $html->find('item');
else
$posts = $html->find('entry');
if(!empty($posts) ) {
$this->sitename = $html->find('title', 0)->plaintext;
$i=0;
foreach ($posts as $article) {
if($i < 3) { if($i < 3) {
$this->items[$i]->uri = $article->find('link', 0)->getAttribute('href');
$this->items[$i]->title = StripCDATA($article->find('title', 0)->plaintext);
$this->items[$i]->author = trim($article->find('author', 0)->innertext);
$this->items[$i]->timestamp = strtotime($article->find('updated', 0)->innertext);
$article_html = $this->file_get_html($this->items[$i]->uri); $item = new \Item();
$this->items[$i]->content = clearContent($article_html->find('article', 0)->innertext);
if(empty($this->items[$i]->content))
$this->items[$i]->content = clearContent($article_html->find('.single-content', 0)->innertext); // another common content div
if(empty($this->items[$i]->content))
$this->items[$i]->content = clearContent($article_html->find('.post', 0)->innertext); // for old WordPress themes without HTML5
$article = $this->ReplaceLinkTagsWithUrlTags($article);
if($type === WORDPRESS_TYPE_RSS){
$item->uri = $article->find('url', 0)->innertext; // 'link' => 'url'!
$item->title = $article->find('title', 0)->plaintext;
$item->author = trim($this->StripCDATA($article->find('dc:creator', 0)->innertext));
$item->timestamp = strtotime($article->find('pubDate', 0)->innertext);
} else {
$item->uri = $article->find('url', 0)->getAttribute('href'); // 'link' => 'url'!
$item->title = $this->StripCDATA($article->find('title', 0)->plaintext);
$item->author = trim($article->find('author', 0)->innertext);
$item->timestamp = strtotime($article->find('updated', 0)->innertext);
}
$article_html = $this->file_get_html($item->uri);
// Attempt to find most common content div
if(empty($item->content)){
$article = $article_html->find('article', 0);
if(!empty($article)){
$item->content = $this->ClearContent($article->innertext);
}
}
// another common content div
if(empty($item->content)){
$article = $article_html->find('.single-content', 0);
if(!empty($article)){
$item->content = $this->ClearContent($article->innertext);
}
}
// for old WordPress themes without HTML5
if(empty($item->content)){
$article = $article_html->find('.post', 0);
if(!empty($article)){
$item->content = $this->ClearContent($article->innertext);
}
}
$this->items[] = $item;
$i++; $i++;
} }
} }
} } else {
else {
$this->returnError("Sorry, {$this->url} doesn't seem to be a Wordpress blog.", 404); $this->returnError("Sorry, {$this->url} doesn't seem to be a Wordpress blog.", 404);
} }
} }
public function getName() { public function getName() {
return "{$this->name} - Wordpress Bridge"; return "{$this->sitename} - Wordpress Bridge";
} }
public function getURI() { public function getURI() {
@ -93,6 +154,4 @@ class WordPressBridge extends BridgeAbstract {
private function processParams($param) { private function processParams($param) {
$this->url = $param['url']; $this->url = $param['url'];
} }
} }