This commit is contained in:
rzhxeo 2013-10-27 10:10:28 +01:00
parent 49a25557b0
commit 5b11143d05
2 changed files with 59 additions and 0 deletions

View File

@ -72,6 +72,7 @@ from .jeuxvideo import JeuxVideoIE
from .jukebox import JukeboxIE from .jukebox import JukeboxIE
from .justintv import JustinTVIE from .justintv import JustinTVIE
from .kankan import KankanIE from .kankan import KankanIE
from .keezmovies import KeezMoviesIE
from .kickstarter import KickStarterIE from .kickstarter import KickStarterIE
from .keek import KeekIE from .keek import KeekIE
from .liveleak import LiveLeakIE from .liveleak import LiveLeakIE

View File

@ -0,0 +1,58 @@
import os
import re
from .common import InfoExtractor
from ..utils import (
compat_urllib_parse_urlparse,
compat_urllib_request,
compat_urllib_parse,
unescapeHTML,
)
from ..aes import (
aes_decrypt_text
)
class KeezMoviesIE(InfoExtractor):
_VALID_URL = r'^(?:https?://)?(?:www\.)?(?P<url>keezmovies\.com/video/.+?(?P<videoid>[0-9]+))'
_TEST = {
u'url': u'http://www.keezmovies.com/video/petite-asian-lady-mai-playing-in-bathtub-1214711',
u'file': u'1214711.mp4',
u'md5': u'6e297b7e789329923fcf83abb67c9289',
u'info_dict': {
u"title": u"Petite Asian Lady Mai Playing In Bathtub",
}
}
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)
# embedded video
mobj = re.search(r'href="([^"]+)"></iframe>', webpage)
if mobj:
embedded_url = mobj.group(1)
return self.playlist_result([self.url_result(embedded_url)], playlist_id=video_id)
video_title = self._html_search_regex(r'<h1 [^>]*>([^<]+)', webpage, u'title')
video_url = compat_urllib_parse.unquote(self._html_search_regex(r'video_url=(.+?)&amp;', webpage, u'video_url'))
if webpage.find('encrypted=true')!=-1:
password = self._html_search_regex(r'video_title=(.+?)&amp;', webpage, u'password')
video_url = aes_decrypt_text(video_url, password, 32).decode('utf-8')
path = compat_urllib_parse_urlparse( video_url ).path
extension = os.path.splitext( path )[1][1:]
format = path.split('/')[4].split('_')[:2]
format = "-".join( format )
return {
'id': video_id,
'title': video_title,
'url': video_url,
'ext': extension,
'format': format,
'format_id': format,
}