projects / github / wallabag / wallabag.git / blame
| Commit | Line | Data |
|---|---|---|
| eb1af592 NL |
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 = $url; | |
| 26 | } | |
| 27 | ||
| 28 | public function isCorrect() | |
| 29 | { | |
| 149df445 | 30 | return filter_var($this->url, FILTER_VALIDATE_URL) !== FALSE; |
| eb1af592 NL |
31 | } |
| 32 | ||
| 33 | public function clean() | |
| 34 | { | |
| 35 | $url = html_entity_decode(trim($this->url)); | |
| 36 | ||
| 37 | $stuff = strpos($url,'&utm_source='); | |
| 38 | if ($stuff !== FALSE) | |
| 39 | $url = substr($url, 0, $stuff); | |
| 40 | $stuff = strpos($url,'?utm_source='); | |
| 41 | if ($stuff !== FALSE) | |
| 42 | $url = substr($url, 0, $stuff); | |
| 43 | $stuff = strpos($url,'#xtor=RSS-'); | |
| 44 | if ($stuff !== FALSE) | |
| 45 | $url = substr($url, 0, $stuff); | |
| 46 | ||
| 47 | $this->url = $url; | |
| 48 | } | |
| 49 | ||
| 50 | public function fetchContent() | |
| 51 | { | |
| 52 | if ($this->isCorrect()) { | |
| 53 | $this->clean(); | |
| 54 | $html = Encoding::toUTF8(Tools::getFile($this->getUrl())); | |
| 55 | ||
| 56 | # if Tools::getFile() if not able to retrieve HTTPS content, try the same URL with HTTP protocol | |
| 57 | if (!preg_match('!^https?://!i', $this->getUrl()) && (!isset($html) || strlen($html) <= 0)) { | |
| 58 | $this->setUrl('http://' . $this->getUrl()); | |
| 59 | $html = Encoding::toUTF8(Tools::getFile($this->getUrl())); | |
| 60 | } | |
| 61 | ||
| 62 | if (function_exists('tidy_parse_string')) { | |
| 63 | $tidy = tidy_parse_string($html, array(), 'UTF8'); | |
| 64 | $tidy->cleanRepair(); | |
| 2bf93dc0 | 65 | |
| 70c39d16 NL |
66 | //Warning: tidy might fail so, ensure there is still a content |
| 67 | $body = $tidy->body(); | |
| 2bf93dc0 | 68 | |
| 70c39d16 NL |
69 | //hasChildren does not seem to work, just check the string |
| 70 | //returned (and do not forget to clean the white spaces) | |
| 71 | if (preg_replace('/\s+/', '', $body->value) !== "<body></body>") { | |
| 72 | $html = $tidy->value; | |
| 73 | } | |
| 7374ff30 | 74 | } |
| eb1af592 NL |
75 | |
| 76 | $parameters = array(); | |
| 77 | if (isset($html) and strlen($html) > 0) | |
| 78 | { | |
| 79 | $readability = new Readability($html, $this->getUrl()); | |
| 80 | $readability->convertLinksToFootnotes = CONVERT_LINKS_FOOTNOTES; | |
| 81 | $readability->revertForcedParagraphElements = REVERT_FORCED_PARAGRAPH_ELEMENTS; | |
| 82 | ||
| 83 | if($readability->init()) | |
| 84 | { | |
| 85 | $content = $readability->articleContent->innerHTML; | |
| 8340fedd | 86 | $parameters['title'] = ($readability->articleTitle->innerHTML != '' ? $readability->articleTitle->innerHTML : _('Untitled')); |
| eb1af592 NL |
87 | $parameters['content'] = $content; |
| 88 | ||
| 89 | return $parameters; | |
| 90 | } | |
| 91 | } | |
| 92 | } | |
| 93 | else { | |
| 94 | #$msg->add('e', _('error during url preparation : the link is not valid')); | |
| 95 | Tools::logm($this->getUrl() . ' is not a valid url'); | |
| 96 | } | |
| 97 | ||
| 98 | return FALSE; | |
| 99 | } | |
| 100 | } |