|
|
|
@ -5,6 +5,7 @@ from __future__ import absolute_import
|
|
|
|
|
|
|
|
|
|
import base64
|
|
|
|
|
import datetime
|
|
|
|
|
import itertools
|
|
|
|
|
import netrc
|
|
|
|
|
import os
|
|
|
|
|
import re
|
|
|
|
@ -3812,8 +3813,6 @@ class PornotubeIE(InfoExtractor):
|
|
|
|
|
|
|
|
|
|
return [info]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class YouJizzIE(InfoExtractor):
|
|
|
|
|
"""Information extractor for youjizz.com."""
|
|
|
|
|
_VALID_URL = r'^(?:https?://)?(?:\w+\.)?youjizz\.com/videos/(?P<videoid>[^.]+).html$'
|
|
|
|
@ -3860,6 +3859,47 @@ class YouJizzIE(InfoExtractor):
|
|
|
|
|
|
|
|
|
|
return [info]
|
|
|
|
|
|
|
|
|
|
class EightTracksIE(InfoExtractor):
|
|
|
|
|
IE_NAME = '8tracks'
|
|
|
|
|
_VALID_URL = r'https?://8tracks.com/(?P<user>[^/]+)/(?P<id>[^/]+)'
|
|
|
|
|
|
|
|
|
|
def _real_extract(self, url):
|
|
|
|
|
mobj = re.match(self._VALID_URL, url)
|
|
|
|
|
if mobj is None:
|
|
|
|
|
raise ExtractorError(u'Invalid URL: %s' % url)
|
|
|
|
|
playlist_id = mobj.group('id')
|
|
|
|
|
|
|
|
|
|
webpage = self._download_webpage(url, playlist_id)
|
|
|
|
|
|
|
|
|
|
m = re.search(r"new TRAX.Mix\((.*?)\);\n*\s*TRAX.initSearchAutocomplete\('#search'\);", webpage, flags=re.DOTALL)
|
|
|
|
|
if not m:
|
|
|
|
|
raise ExtractorError(u'Cannot find trax information')
|
|
|
|
|
json_like = m.group(1)
|
|
|
|
|
data = json.loads(json_like)
|
|
|
|
|
|
|
|
|
|
session = str(random.randint(0, 1000000000))
|
|
|
|
|
mix_id = data['id']
|
|
|
|
|
track_count = data['tracks_count']
|
|
|
|
|
first_url = 'http://8tracks.com/sets/%s/play?player=sm&mix_id=%s&format=jsonh' % (session, mix_id)
|
|
|
|
|
next_url = first_url
|
|
|
|
|
res = []
|
|
|
|
|
for i in itertools.count():
|
|
|
|
|
api_json = self._download_webpage(next_url, playlist_id,
|
|
|
|
|
note=u'Downloading song information %s/%s' % (str(i+1), track_count),
|
|
|
|
|
errnote=u'Failed to download song information')
|
|
|
|
|
api_data = json.loads(api_json)
|
|
|
|
|
track_data = api_data[u'set']['track']
|
|
|
|
|
info = {
|
|
|
|
|
'id': track_data['id'],
|
|
|
|
|
'url': track_data['track_file_stream_url'],
|
|
|
|
|
'title': track_data['name'],
|
|
|
|
|
'ext': 'm4a',
|
|
|
|
|
}
|
|
|
|
|
res.append(info)
|
|
|
|
|
if api_data['set']['at_last_track']:
|
|
|
|
|
break
|
|
|
|
|
next_url = 'http://8tracks.com/sets/%s/next?player=sm&mix_id=%s&format=jsonh&track_id=%s' % (session, mix_id, track_data['id'])
|
|
|
|
|
return res
|
|
|
|
|
|
|
|
|
|
def gen_extractors():
|
|
|
|
|
""" Return a list of an instance of every supported extractor.
|
|
|
|
@ -3906,6 +3946,7 @@ def gen_extractors():
|
|
|
|
|
SteamIE(),
|
|
|
|
|
UstreamIE(),
|
|
|
|
|
RBMARadioIE(),
|
|
|
|
|
EightTracksIE(),
|
|
|
|
|
GenericIE()
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|