1
0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-14 11:07:50 +01:00
youtube-dl/youtube_dl/extractor/xtube.py

55 lines
2.0 KiB
Python
Raw Normal View History

2013-10-28 19:23:48 +01:00
import os
import re
from .common import InfoExtractor
from ..utils import (
compat_urllib_parse_urlparse,
compat_urllib_request,
)
class XTubeIE(InfoExtractor):
_VALID_URL = r'^(?:https?://)?(?:www\.)?(?P<url>xtube\.com/watch\.php\?v=(?P<videoid>[^/?&]+))'
_TEST = {
u'url': u'http://www.xtube.com/watch.php?v=kVTUy_G222_',
u'file': u'kVTUy_G222_.mp4',
u'md5': u'092fbdd3cbe292c920ef6fc6a8a9cdab',
u'info_dict': {
u"title": u"strange erotica",
2013-11-02 22:45:48 +01:00
u"description": u"surreal gay themed erotica...almost an ET kind of thing",
2013-10-28 19:23:48 +01:00
u"uploader": u"greenshowers",
u"age_limit": 18,
}
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('videoid')
url = 'http://www.' + mobj.group('url')
req = compat_urllib_request.Request(url)
req.add_header('Cookie', 'age_verified=1')
webpage = self._download_webpage(req, video_id)
video_title = self._html_search_regex(r'<div class="p_5px[^>]*>([^<]+)', webpage, u'title')
video_uploader = self._html_search_regex(r'so_s\.addVariable\("owner_u", "([^"]+)', webpage, u'uploader', fatal=False)
video_description = self._html_search_regex(r'<p class="video_description">([^<]+)', webpage, u'description', fatal=False)
2013-10-28 19:23:48 +01:00
video_url= self._html_search_regex(r'var videoMp4 = "([^"]+)', webpage, u'video_url').replace('\\/', '/')
path = compat_urllib_parse_urlparse(video_url).path
extension = os.path.splitext(path)[1][1:]
2013-10-28 19:23:48 +01:00
format = path.split('/')[5].split('_')[:2]
format[0] += 'p'
format[1] += 'k'
format = "-".join(format)
2013-10-28 19:23:48 +01:00
return {
'id': video_id,
'title': video_title,
'uploader': video_uploader,
'description': video_description,
'url': video_url,
'ext': extension,
'format': format,
'format_id': format,
'age_limit': 18,
}