[chilloutzone] Added support for chilloutzone.net

Added support for chilloutzone.net videos including embedded youtube
and vimeo movies. In case you find a not working movie, drop me an
email.
This commit is contained in:
Andreas Schmitz 2014-02-06 11:44:44 +01:00
parent df872ec4e7
commit 46a073bfac
2 changed files with 101 additions and 0 deletions

View File

@ -25,6 +25,7 @@ from .canalplus import CanalplusIE
from .canalc2 import Canalc2IE
from .cbs import CBSIE
from .channel9 import Channel9IE
from .chilloutzone import ChilloutzoneIE
from .cinemassacre import CinemassacreIE
from .clipfish import ClipfishIE
from .cliphunter import CliphunterIE

View File

@ -0,0 +1,100 @@
import re
import base64
import urllib
import json
from .common import InfoExtractor
video_container = ('.mp4', '.mkv', '.flv')
class ChilloutzoneIE(InfoExtractor):
_VALID_URL = r'(?:https?://)?(?:www\.)?chilloutzone\.net/video/(?P<id>[\w|-]+).html'
_TEST = {
u'url': u'http://www.chilloutzone.net/video/enemene-meck-alle-katzen-weg.html',
u'file': u'18088-enemene-meck-alle-katzen-weg.mp4',
u'md5': u'a76f3457e813ea0037e5244f509e66d1',
u'info_dict': {
u"id": u"18088",
u"ext": u"mp4",
u"title": u"Enemene Meck - Alle Katzen weg"
}
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
webpage_url = 'http://www.chilloutzone.net/video/' + video_id + '.html'
# Log that we are starting to download the page
self.report_download_webpage(webpage_url)
webpage = self._download_webpage(webpage_url, video_id)
# Log that we are starting to parse the page
self.report_extraction(video_id)
# Find base64 decoded file info
base64_video_info = self._html_search_regex(r'var cozVidData = "(.+?)";', webpage, u'video Data')
# decode string and find video file
decoded_video_info = base64.b64decode(base64_video_info)
video_info_dict = json.loads(decoded_video_info)
# get video information from dict
media_url = video_info_dict['mediaUrl']
description = video_info_dict['description']
title = video_info_dict['title']
native_platform = video_info_dict['nativePlatform']
native_video_id = video_info_dict['nativeVideoId']
source_priority = video_info_dict['sourcePriority']
# Start video extraction
video_url = ''
# If nativePlatform is None a fallback mechanism is used (i.e. youtube embed)
if native_platform == None:
# Look for other video urls
video_url = self._html_search_regex(r'<iframe.* src="(.+?)".*', webpage, u'fallback Video URL')
if 'youtube' in video_url:
self.to_screen(u'Youtube video detected:')
print video_url
return self.url_result(video_url, ie='Youtube')
# For debugging purposes
#print video_info_dict
#print native_platform
#print native_video_id
#print source_priority
#print media_url
# Non Fallback: Decide to use native source (e.g. youtube or vimeo) or
# the own CDN
if source_priority == 'native':
if native_platform == 'youtube':
self.to_screen(u'Youtube video detected:')
video_url = 'https://www.youtube.com/watch?v=' + native_video_id
print video_url
return self.url_result(video_url, ie='Youtube')
if native_platform == 'vimeo':
self.to_screen(u'Vimeo video detected:')
video_url = 'http://vimeo.com/' + native_video_id
print video_url
return self.url_result(video_url, ie='Vimeo')
# No redirect, use coz media url
video_url = media_url
if video_url.endswith('.mp4') == False:
self.report_warning(u'Url does not contain a video container')
return []
return [{
'id': video_id,
'url': video_url,
'ext': 'mp4',
'title': title,
'description': description
}]