inc/poche/Url.class.php

   1 <?php
   2 /**
   3  * poche, a read it later open source system
   4  *
   5  * @category   poche
   6  * @author     Nicolas Lœuillet <support@inthepoche.com>
   7  * @copyright  2013
   8  * @license    http://www.wtfpl.net/ see COPYING file
   9  */
  10
  11 class Url
  12 {
  13     public $url;
  14
  15     function __construct($url)
  16     {
  17         $this->url = base64_decode($url);
  18     }
  19
  20     public function getUrl() {
  21         return $this->url;
  22     }
  23
  24     public function setUrl($url) {
  25         $this->url = $url;
  26     }
  27
  28     public function isCorrect()
  29     {
  30         $pattern = '|^(.*:)//([a-z\-.]+)(:[0-9]+)?(.*)$|i';
  31
  32         return preg_match($pattern, $this->url);
  33     }
  34
  35     public function clean()
  36     {
  37         $url = html_entity_decode(trim($this->url));
  38
  39         $stuff = strpos($url,'&utm_source=');
  40         if ($stuff !== FALSE)
  41             $url = substr($url, 0, $stuff);
  42         $stuff = strpos($url,'?utm_source=');
  43         if ($stuff !== FALSE)
  44             $url = substr($url, 0, $stuff);
  45         $stuff = strpos($url,'#xtor=RSS-');
  46         if ($stuff !== FALSE)
  47             $url = substr($url, 0, $stuff);
  48
  49         $this->url = $url;
  50     }
  51
  52     public function fetchContent()
  53     {
  54         if ($this->isCorrect()) {
  55             $this->clean();
  56             $html = Encoding::toUTF8(Tools::getFile($this->getUrl()));
  57
  58             # if Tools::getFile() if not able to retrieve HTTPS content, try the same URL with HTTP protocol
  59             if (!preg_match('!^https?://!i', $this->getUrl()) && (!isset($html) || strlen($html) <= 0)) {
  60                 $this->setUrl('http://' . $this->getUrl());
  61                 $html = Encoding::toUTF8(Tools::getFile($this->getUrl()));
  62             }
  63
  64             if (function_exists('tidy_parse_string')) {
  65                 $tidy = tidy_parse_string($html, array(), 'UTF8');
  66                 $tidy->cleanRepair();
  67
  68                 //Warning: tidy might fail so, ensure there is still a content
  69                 $body = $tidy->body();
  70
  71                 //hasChildren does not seem to work, just check the string
  72                 //returned (and do not forget to clean the white spaces)
  73                 if (preg_replace('/\s+/', '', $body->value) !== "<body></body>") {
  74                     $html = $tidy->value;
  75                 }
  76             }
  77
  78             $parameters = array();
  79             if (isset($html) and strlen($html) > 0)
  80             {
  81                 $readability = new Readability($html, $this->getUrl());
  82                 $readability->convertLinksToFootnotes = CONVERT_LINKS_FOOTNOTES;
  83                 $readability->revertForcedParagraphElements = REVERT_FORCED_PARAGRAPH_ELEMENTS;
  84
  85                 if($readability->init())
  86                 {
  87                     $content = $readability->articleContent->innerHTML;
  88                     $parameters['title'] = ($readability->articleTitle->innerHTML != '' ? $readability->articleTitle->innerHTML : _('Untitled'));
  89                     $parameters['content'] = $content;
  90
  91                     return $parameters;
  92                 }
  93             }
  94         }
  95         else {
  96             #$msg->add('e', _('error during url preparation : the link is not valid'));
  97             Tools::logm($this->getUrl() . ' is not a valid url');
  98         }
  99
 100         return FALSE;
 101     }
 102 }