mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-12-25 07:30:22 +00:00
Extractor for websurg.com
This commit is contained in:
parent
f3f34c5b0f
commit
d79a0e233a
@ -109,6 +109,7 @@ from .videofyme import VideofyMeIE
|
|||||||
from .vimeo import VimeoIE, VimeoChannelIE
|
from .vimeo import VimeoIE, VimeoChannelIE
|
||||||
from .vine import VineIE
|
from .vine import VineIE
|
||||||
from .wat import WatIE
|
from .wat import WatIE
|
||||||
|
from .websurg import WeBSurgIE
|
||||||
from .weibo import WeiboIE
|
from .weibo import WeiboIE
|
||||||
from .wimp import WimpIE
|
from .wimp import WimpIE
|
||||||
from .worldstarhiphop import WorldStarHipHopIE
|
from .worldstarhiphop import WorldStarHipHopIE
|
||||||
|
67
youtube_dl/extractor/websurg.py
Normal file
67
youtube_dl/extractor/websurg.py
Normal file
@ -0,0 +1,67 @@
|
|||||||
|
# coding: utf-8
|
||||||
|
|
||||||
|
import re
|
||||||
|
|
||||||
|
from ..utils import (
|
||||||
|
compat_urllib_request,
|
||||||
|
compat_urllib_parse
|
||||||
|
)
|
||||||
|
|
||||||
|
from .common import InfoExtractor
|
||||||
|
|
||||||
|
class WeBSurgIE(InfoExtractor):
|
||||||
|
IE_NAME = u'websurg.com'
|
||||||
|
_VALID_URL = r'http://.*?\.websurg\.com/MEDIA/\?noheader=1&doi=(.*)'
|
||||||
|
|
||||||
|
_TEST = {
|
||||||
|
u'url': u'http://www.websurg.com/MEDIA/?noheader=1&doi=vd01en4012',
|
||||||
|
u'file': u'vd01en4012.mp4',
|
||||||
|
u'params': {
|
||||||
|
u'skip_download': True,
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
_LOGIN_URL = 'http://www.websurg.com/inc/login/login_div.ajax.php?login=1'
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
|
||||||
|
login_form = {
|
||||||
|
'username': self._downloader.params['username'],
|
||||||
|
'password': self._downloader.params['password'],
|
||||||
|
'Submit': 1
|
||||||
|
}
|
||||||
|
|
||||||
|
request = compat_urllib_request.Request(
|
||||||
|
self._LOGIN_URL, compat_urllib_parse.urlencode(login_form))
|
||||||
|
request.add_header(
|
||||||
|
'Content-Type', 'application/x-www-form-urlencoded;charset=utf-8')
|
||||||
|
login_results = compat_urllib_request.urlopen(request).info()
|
||||||
|
|
||||||
|
sessid = re.match(r'PHPSESSID=(.*);',
|
||||||
|
login_results['Set-Cookie']).group(1)
|
||||||
|
request = compat_urllib_request.Request(
|
||||||
|
url, compat_urllib_parse.urlencode(login_form),
|
||||||
|
{'Cookie': 'PHPSESSID=' + sessid + ';'})
|
||||||
|
webpage = compat_urllib_request.urlopen(request).read()
|
||||||
|
|
||||||
|
video_id = re.match(self._VALID_URL, url).group(1)
|
||||||
|
|
||||||
|
url_info = re.search(r'streamer="(.*?)" src="(.*?)"', webpage)
|
||||||
|
|
||||||
|
if url_info is None:
|
||||||
|
self._downloader.report_warning(
|
||||||
|
u'Unable to log in: bad username/password')
|
||||||
|
return
|
||||||
|
|
||||||
|
return {'id': video_id,
|
||||||
|
'title' : re.search(
|
||||||
|
r'property="og:title" content="(.*?)" />'
|
||||||
|
, webpage).group(1),
|
||||||
|
'description': re.search(
|
||||||
|
r'name="description" content="(.*?)" />', webpage).group(1),
|
||||||
|
'ext' : 'mp4',
|
||||||
|
'url' : url_info.group(1) + '/' + url_info.group(2),
|
||||||
|
'thumbnail': re.search(
|
||||||
|
r'property="og:image" content="(.*?)" />', webpage
|
||||||
|
).group(1)
|
||||||
|
}
|
Loading…
Reference in New Issue
Block a user