[ted] Style fixes

This commit is contained in:
Jaime Marquínez Ferrándiz 2014-03-05 13:27:26 +01:00
parent ca1fee34f2
commit bacac173a9

View File

@ -12,22 +12,25 @@
class TEDIE(SubtitlesInfoExtractor):
_VALID_URL=r'''(?x)http://www\.ted\.com/
(
(?P<type_playlist>playlists(?:/\d+)?) # We have a playlist
|
((?P<type_talk>talks)) # We have a simple talk
)
(/lang/(.*?))? # The url may contain the language
/(?P<name>\w+) # Here goes the name and then ".html"
'''
_VALID_URL = r'''(?x)http://www\.ted\.com/
(
(?P<type_playlist>playlists(?:/\d+)?) # We have a playlist
|
((?P<type_talk>talks)) # We have a simple talk
)
(/lang/(.*?))? # The url may contain the language
/(?P<name>\w+) # Here goes the name and then ".html"
'''
_TEST = {
'url': 'http://www.ted.com/talks/dan_dennett_on_our_consciousness.html',
'file': '102.mp4',
'md5': '4ea1dada91e4174b53dac2bb8ace429d',
'info_dict': {
'title': 'The illusion of consciousness',
'description': 'Philosopher Dan Dennett makes a compelling argument that not only don\'t we understand our own consciousness, but that half the time our brains are actively fooling us.',
'description': ('Philosopher Dan Dennett makes a compelling '
'argument that not only don\'t we understand our own '
'consciousness, but that half the time our brains are '
'actively fooling us.'),
'uploader': 'Dan Dennett',
}
}
@ -39,18 +42,18 @@ class TEDIE(SubtitlesInfoExtractor):
}
def _extract_info(self, webpage):
info_json = self._search_regex(r'q\("\w+.init",({.+})\)</script>', webpage, 'info json')
info_json = self._search_regex(r'q\("\w+.init",({.+})\)</script>',
webpage, 'info json')
return json.loads(info_json)
def _real_extract(self, url):
m=re.match(self._VALID_URL, url, re.VERBOSE)
m = re.match(self._VALID_URL, url, re.VERBOSE)
name = m.group('name')
if m.group('type_talk'):
return self._talk_info(url)
else :
name=m.group('name')
return self._talk_info(url, name)
else:
return self._playlist_videos_info(url, name)
def _playlist_videos_info(self, url, name):
'''Returns the videos of the playlist'''
@ -68,11 +71,8 @@ def _playlist_videos_info(self, url, name):
playlist_id=compat_str(playlist_info['id']),
playlist_title=playlist_info['title'])
def _talk_info(self, url, video_id=0):
"""Return the video for the talk in the url"""
m = re.match(self._VALID_URL, url)
video_name = m.group('name')
webpage = self._download_webpage(url, video_id, 'Downloading \"%s\" page' % video_name)
def _talk_info(self, url, video_name):
webpage = self._download_webpage(url, video_name)
self.report_extraction(video_name)
talk_info = self._extract_info(webpage)['talks'][0]