1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-16 14:34:31 +01:00
youtube-dl/youtube_dl/extractor/comedycentral.py
2013-10-04 11:14:10 +02:00

188 lines
7.6 KiB
Python

import re
import xml.etree.ElementTree
from .common import InfoExtractor
from ..utils import (
compat_str,
compat_urllib_parse,
ExtractorError,
unified_strdate,
)
class ComedyCentralIE(InfoExtractor):
IE_DESC = u'The Daily Show / Colbert Report'
# urls can be abbreviations like :thedailyshow or :colbert
# urls for episodes like:
# or urls for clips like: http://www.thedailyshow.com/watch/mon-december-10-2012/any-given-gun-day
# or: http://www.colbertnation.com/the-colbert-report-videos/421667/november-29-2012/moon-shattering-news
# or: http://www.colbertnation.com/the-colbert-report-collections/422008/festival-of-lights/79524
_VALID_URL = r"""^(:(?P<shortname>tds|thedailyshow|cr|colbert|colbertnation|colbertreport)
|(https?://)?(www\.)?
(?P<showname>thedailyshow|colbertnation)\.com/
(full-episodes/(?P<episode>.*)|
(?P<clip>
(the-colbert-report-(videos|collections)/(?P<clipID>[0-9]+)/[^/]*/(?P<cntitle>.*?))
|(watch/(?P<date>[^/]*)/(?P<tdstitle>.*)))|
(?P<interview>
extended-interviews/(?P<interID>[0-9]+)/playlist_tds_extended_(?P<interview_title>.*?)/.*?)))
$"""
_TEST = {
u'url': u'http://www.thedailyshow.com/watch/thu-december-13-2012/kristen-stewart',
u'file': u'422212.mp4',
u'md5': u'4e2f5cb088a83cd8cdb7756132f9739d',
u'info_dict': {
u"upload_date": u"20121214",
u"description": u"Kristen Stewart",
u"uploader": u"thedailyshow",
u"title": u"thedailyshow-kristen-stewart part 1"
}
}
_available_formats = ['3500', '2200', '1700', '1200', '750', '400']
_video_extensions = {
'3500': 'mp4',
'2200': 'mp4',
'1700': 'mp4',
'1200': 'mp4',
'750': 'mp4',
'400': 'mp4',
}
_video_dimensions = {
'3500': (1280, 720),
'2200': (960, 540),
'1700': (768, 432),
'1200': (640, 360),
'750': (512, 288),
'400': (384, 216),
}
@classmethod
def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
@staticmethod
def _transform_rtmp_url(rtmp_video_url):
m = re.match(r'^rtmpe?://.*?/(?P<finalid>gsp.comedystor/.*)$', rtmp_video_url)
if not m:
raise ExtractorError(u'Cannot transform RTMP url')
base = 'http://mtvnmobile.vo.llnwd.net/kip0/_pxn=1+_pxI0=Ripod-h264+_pxL0=undefined+_pxM0=+_pxK=18639+_pxE=mp4/44620/mtvnorigin/'
return base + m.group('finalid')
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url, re.VERBOSE)
if mobj is None:
raise ExtractorError(u'Invalid URL: %s' % url)
if mobj.group('shortname'):
if mobj.group('shortname') in ('tds', 'thedailyshow'):
url = u'http://www.thedailyshow.com/full-episodes/'
else:
url = u'http://www.colbertnation.com/full-episodes/'
mobj = re.match(self._VALID_URL, url, re.VERBOSE)
assert mobj is not None
if mobj.group('clip'):
if mobj.group('showname') == 'thedailyshow':
epTitle = mobj.group('tdstitle')
else:
epTitle = mobj.group('cntitle')
dlNewest = False
elif mobj.group('interview'):
epTitle = mobj.group('interview_title')
dlNewest = False
else:
dlNewest = not mobj.group('episode')
if dlNewest:
epTitle = mobj.group('showname')
else:
epTitle = mobj.group('episode')
self.report_extraction(epTitle)
webpage,htmlHandle = self._download_webpage_handle(url, epTitle)
if dlNewest:
url = htmlHandle.geturl()
mobj = re.match(self._VALID_URL, url, re.VERBOSE)
if mobj is None:
raise ExtractorError(u'Invalid redirected URL: ' + url)
if mobj.group('episode') == '':
raise ExtractorError(u'Redirected URL is still not specific: ' + url)
epTitle = mobj.group('episode')
mMovieParams = re.findall('(?:<param name="movie" value="|var url = ")(http://media.mtvnservices.com/([^"]*(?:episode|video).*?:.*?))"', webpage)
if len(mMovieParams) == 0:
# The Colbert Report embeds the information in a without
# a URL prefix; so extract the alternate reference
# and then add the URL prefix manually.
altMovieParams = re.findall('data-mgid="([^"]*(?:episode|video).*?:.*?)"', webpage)
if len(altMovieParams) == 0:
raise ExtractorError(u'unable to find Flash URL in webpage ' + url)
else:
mMovieParams = [("http://media.mtvnservices.com/" + altMovieParams[0], altMovieParams[0])]
uri = mMovieParams[0][1]
indexUrl = 'http://shadow.comedycentral.com/feeds/video_player/mrss/?' + compat_urllib_parse.urlencode({'uri': uri})
indexXml = self._download_webpage(indexUrl, epTitle,
u'Downloading show index',
u'unable to download episode index')
results = []
idoc = xml.etree.ElementTree.fromstring(indexXml)
itemEls = idoc.findall('.//item')
for partNum,itemEl in enumerate(itemEls):
mediaId = itemEl.findall('./guid')[0].text
shortMediaId = mediaId.split(':')[-1]
showId = mediaId.split(':')[-2].replace('.com', '')
officialTitle = itemEl.findall('./title')[0].text
officialDate = unified_strdate(itemEl.findall('./pubDate')[0].text)
configUrl = ('http://www.comedycentral.com/global/feeds/entertainment/media/mediaGenEntertainment.jhtml?' +
compat_urllib_parse.urlencode({'uri': mediaId}))
configXml = self._download_webpage(configUrl, epTitle,
u'Downloading configuration for %s' % shortMediaId)
cdoc = xml.etree.ElementTree.fromstring(configXml)
turls = []
for rendition in cdoc.findall('.//rendition'):
finfo = (rendition.attrib['bitrate'], rendition.findall('./src')[0].text)
turls.append(finfo)
if len(turls) == 0:
self._downloader.report_error(u'unable to download ' + mediaId + ': No videos found')
continue
formats = []
for format, rtmp_video_url in turls:
w, h = self._video_dimensions.get(format, (None, None))
formats.append({
'url': self._transform_rtmp_url(rtmp_video_url),
'ext': self._video_extensions.get(format, 'mp4'),
'format_id': format,
'height': h,
'width': w,
})
effTitle = showId + u'-' + epTitle + u' part ' + compat_str(partNum+1)
info = {
'id': shortMediaId,
'formats': formats,
'uploader': showId,
'upload_date': officialDate,
'title': effTitle,
'thumbnail': None,
'description': compat_str(officialTitle),
}
# TODO: Remove when #980 has been merged
info.update(info['formats'][-1])
results.append(info)
return results