[myvideo] Modernize

This commit is contained in:
Jaime Marquínez Ferrándiz 2014-03-09 18:58:34 +01:00
parent 9d9d70c462
commit 439a1fffcb

View File

@ -1,3 +1,5 @@
from __future__ import unicode_literals
import binascii
import base64
import hashlib
@ -14,18 +16,16 @@ from ..utils import (
)
class MyVideoIE(InfoExtractor):
"""Information Extractor for myvideo.de."""
_VALID_URL = r'(?:http://)?(?:www\.)?myvideo\.de/(?:[^/]+/)?watch/([0-9]+)/([^?/]+).*'
IE_NAME = u'myvideo'
_VALID_URL = r'http://(?:www\.)?myvideo\.de/(?:[^/]+/)?watch/(?P<id>[0-9]+)/[^?/]+.*'
IE_NAME = 'myvideo'
_TEST = {
u'url': u'http://www.myvideo.de/watch/8229274/bowling_fail_or_win',
u'file': u'8229274.flv',
u'md5': u'2d2753e8130479ba2cb7e0a37002053e',
u'info_dict': {
u"title": u"bowling-fail-or-win"
'url': 'http://www.myvideo.de/watch/8229274/bowling_fail_or_win',
'md5': '2d2753e8130479ba2cb7e0a37002053e',
'info_dict': {
'id': '8229274',
'ext': 'flv',
'title': 'bowling-fail-or-win',
}
}
@ -53,10 +53,7 @@ class MyVideoIE(InfoExtractor):
def _real_extract(self,url):
mobj = re.match(self._VALID_URL, url)
if mobj is None:
raise ExtractorError(u'invalid URL: %s' % url)
video_id = mobj.group(1)
video_id = mobj.group('id')
GK = (
b'WXpnME1EZGhNRGhpTTJNM01XVmhOREU0WldNNVpHTTJOakpt'
@ -74,37 +71,33 @@ class MyVideoIE(InfoExtractor):
video_url = mobj.group(1) + '.flv'
video_title = self._html_search_regex('<title>([^<]+)</title>',
webpage, u'title')
webpage, 'title')
video_ext = self._search_regex('[.](.+?)$', video_url, u'extension')
return [{
'id': video_id,
'url': video_url,
'uploader': None,
'upload_date': None,
'title': video_title,
'ext': video_ext,
}]
return {
'id': video_id,
'url': video_url,
'title': video_title,
}
mobj = re.search(r'data-video-service="/service/data/video/%s/config' % video_id, webpage)
if mobj is not None:
request = compat_urllib_request.Request('http://www.myvideo.de/service/data/video/%s/config' % video_id, '')
response = self._download_webpage(request, video_id,
u'Downloading video info')
'Downloading video info')
info = json.loads(base64.b64decode(response).decode('utf-8'))
return {'id': video_id,
'title': info['title'],
'url': info['streaming_url'].replace('rtmpe', 'rtmpt'),
'play_path': info['filename'],
'ext': 'flv',
'thumbnail': info['thumbnail'][0]['url'],
}
return {
'id': video_id,
'title': info['title'],
'url': info['streaming_url'].replace('rtmpe', 'rtmpt'),
'play_path': info['filename'],
'ext': 'flv',
'thumbnail': info['thumbnail'][0]['url'],
}
# try encxml
mobj = re.search('var flashvars={(.+?)}', webpage)
if mobj is None:
raise ExtractorError(u'Unable to extract video')
raise ExtractorError('Unable to extract video')
params = {}
encxml = ''
@ -118,7 +111,7 @@ class MyVideoIE(InfoExtractor):
params['domain'] = 'www.myvideo.de'
xmldata_url = '%s?%s' % (encxml, compat_urllib_parse.urlencode(params))
if 'flash_playertype=MTV' in xmldata_url:
self._downloader.report_warning(u'avoiding MTV player')
self._downloader.report_warning('avoiding MTV player')
xmldata_url = (
'http://www.myvideo.de/dynamic/get_player_video_xml.php'
'?flash_playertype=D&ID=%s&_countlimit=4&autorun=yes'
@ -144,7 +137,7 @@ class MyVideoIE(InfoExtractor):
video_url = compat_urllib_parse.unquote(mobj.group(1))
if 'myvideo2flash' in video_url:
self.report_warning(
u'Rewriting URL to use unencrypted rtmp:// ...',
'Rewriting URL to use unencrypted rtmp:// ...',
video_id)
video_url = video_url.replace('rtmpe://', 'rtmp://')
@ -152,39 +145,31 @@ class MyVideoIE(InfoExtractor):
# extract non rtmp videos
mobj = re.search('path=\'(http.*?)\' source=\'(.*?)\'', dec_data)
if mobj is None:
raise ExtractorError(u'unable to extract url')
raise ExtractorError('unable to extract url')
video_url = compat_urllib_parse.unquote(mobj.group(1)) + compat_urllib_parse.unquote(mobj.group(2))
video_file = self._search_regex('source=\'(.*?)\'', dec_data, u'video file')
video_file = self._search_regex('source=\'(.*?)\'', dec_data, 'video file')
video_file = compat_urllib_parse.unquote(video_file)
if not video_file.endswith('f4m'):
ppath, prefix = video_file.split('.')
video_playpath = '%s:%s' % (prefix, ppath)
video_hls_playlist = ''
else:
video_playpath = ''
video_hls_playlist = (
video_file
).replace('.f4m', '.m3u8')
video_swfobj = self._search_regex('swfobject.embedSWF\(\'(.+?)\'', webpage, u'swfobj')
video_swfobj = self._search_regex('swfobject.embedSWF\(\'(.+?)\'', webpage, 'swfobj')
video_swfobj = compat_urllib_parse.unquote(video_swfobj)
video_title = self._html_search_regex("<h1(?: class='globalHd')?>(.*?)</h1>",
webpage, u'title')
webpage, 'title')
return [{
'id': video_id,
'url': video_url,
'tc_url': video_url,
'uploader': None,
'upload_date': None,
'title': video_title,
'ext': u'flv',
'play_path': video_playpath,
'video_file': video_file,
'video_hls_playlist': video_hls_playlist,
'player_url': video_swfobj,
}]
return {
'id': video_id,
'url': video_url,
'tc_url': video_url,
'title': video_title,
'ext': 'flv',
'play_path': video_playpath,
'player_url': video_swfobj,
}