Adding a meta-bridge for all Gawker publications (lifehacker, kotaku, you name it, ...)

This commit is contained in:
Nicolas Delsaux 2014-04-02 10:55:42 +01:00
parent 1a12f48e2e
commit fd71ceae82
2 changed files with 123 additions and 13 deletions

89
bridges/Gawker.php Normal file
View file

@ -0,0 +1,89 @@
<?php
/**
*
* @name Gawker media
* @description A bridge allowing access to any of the numerous Gawker media blogs (Lifehacker, deadspin, Kotaku, Jezebel, and so on
* @update 27/03/2014
* @use1(site="site")
*/
class Gawker extends HttpCachingBridgeAbstract{
private $uri;
private $name;
public function collectData(array $param){
if (empty($param['site'])) {
trigger_error("If no site is provided, nothing is gonna happen", E_USER_ERROR);
} else {
$this->uri = $param['site'];
}
$html = file_get_html($this->getURI()) or $this->returnError('Could not request '.$this->getURI(), 404);
$this->message("loaded HTML from ".$this->getURI());
// customize name
$this->name = $html->find('title', 0)->innertext;
foreach($html->find('.main-column') as $content) {
$this->parseContent($content);
}
}
public function parseContent($content) {
foreach($content->find('.headline') as $headline) {
foreach($headline->find('a') as $articleLink) {
// notice we only use article from this gawker site (as gawker like to see us visit other sites)
if(strpos($articleLink->href, $this->getURI())>=0) {
$this->parseLink($articleLink);
}
}
}
}
public function parseLink($infoLink) {
$item = new Item();
$item->uri = $infoLink->href;
$item->title = $infoLink->innertext;
try {
// now load that uri from cache
// $this->message("loading page ".$item->uri);
$articlePage = str_get_html($this->get_cached($item->uri));
if(is_object($articlePage)) {
$content = $articlePage->find('.post-content', 0);
$this->defaultImageSrcTo($content, $this->getURI());
$item->content = $content->innertext;
// http://stackoverflow.com/q/22715928/15619
$publishtime = $articlePage->find('.publish-time', 0)->getAttribute("data-publishtime");
// don't know what I'm doing there, but http://www.epochconverter.com/programming/functions-php.php#epoch2date recommends it
$item->timestamp = $this->js_to_unix_timestamp($publishtime);
$vcard = $articlePage->find('.vcard', 0);
if(is_object($vcard)) {
$item->name = $vcard->find('a', 0)->innertext;
}
} else {
throw new Exception("cache content for ".$item->uri." is NOT a Simple DOM parser object !");
}
} catch(Exception $e) {
$this->message("obtaining ".$item->uri." resulted in exception ".$e->getMessage().". Deleting cached page ...");
// maybe file is incorrect. it should be discarded from cache
$this->remove_from_cache($item->url);
$item->content = $e->getMessage();
}
$this->items[] = $item;
}
function js_to_unix_timestamp($jsTimestamp){
return $jsTimestamp/1000;
}
public function getName(){
return $this->name;
}
public function getURI(){
return $this->uri;
}
public function getCacheDuration(){
return 3600; // 1h
}
public function getDescription(){
return "Gawker press blog content.";
}
}

View file

@ -109,33 +109,54 @@ abstract class HttpCachingBridgeAbstract extends BridgeAbstract {
$filename = $filename."index.html";
}
if(file_exists($filename)) {
// $this->message("loading cached file from ".$filename." for page at url ".$url);
// $this->message("loading cached file from ".$filename." for page at url ".$url);
// TODO touch file and its parent, and try to do neighbour deletion
$currentPath = $filename;
while(!$pageCacheDir==$currentPath) {
touch($currentPath);
$currentPath = dirname($currentPath);
}
$this->refresh_in_cache($pageCacheDir, $filename);
} else {
// $this->message("we have no local copy of ".$url." Downloading !");
// $this->message("we have no local copy of ".$url." Downloading to ".$filename);
$dir = substr($filename, 0, strrpos($filename, '/'));
if(!is_dir($dir)) {
// $this->message("creating directories for ".$dir);
mkdir($dir, 0777, true);
}
$this->download_remote($url, $filename);
}
return file_get_contents($filename);
}
private function refresh_in_cache($pageCacheDir, $filename) {
$currentPath = $filename;
while(!$pageCacheDir==$currentPath) {
touch($currentPath);
$currentPath = dirname($currentPath);
}
}
public function download_remote($url , $save_path) {
$f = fopen( $save_path , 'w+');
$handle = fopen($url , "rb");
while (!feof($handle)) {
$contents = fread($handle, 8192);
fwrite($f , $contents);
if($f) {
$handle = fopen($url , "rb");
if($handle) {
while (!feof($handle)) {
$contents = fread($handle, 8192);
if($contents) {
fwrite($f , $contents);
}
}
fclose($handle);
}
fclose($f);
}
fclose($handle);
fclose($f);
}
public function remove_from_cache($url) {
$simplified_url = str_replace(["http://", "https://", "?", "&", "="], ["", "", "/", "/", "/"], $url);
// TODO build this from the variable given to Cache
$pageCacheDir = __DIR__ . '/../cache/'."pages/";
$filename = realpath($pageCacheDir.$simplified_url);
$this->message("removing from cache \"".$filename."\" WELL, NOT REALLY");
// filename is NO GOOD
// unlink($filename);
}
public function message($text) {