mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-27 19:36:50 +01:00
[vvvvid] Add new extractor(closes #5915)
This commit is contained in:
parent
38be3bc568
commit
7b1e80792b
@ -1110,6 +1110,7 @@ from .vporn import VpornIE
|
||||
from .vrt import VRTIE
|
||||
from .vube import VubeIE
|
||||
from .vuclip import VuClipIE
|
||||
from .vvvvid import VVVVIDIE
|
||||
from .vyborymos import VyboryMosIE
|
||||
from .vzaar import VzaarIE
|
||||
from .walla import WallaIE
|
||||
|
148
youtube_dl/extractor/vvvvid.py
Normal file
148
youtube_dl/extractor/vvvvid.py
Normal file
@ -0,0 +1,148 @@
|
||||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
ExtractorError,
|
||||
int_or_none,
|
||||
str_or_none,
|
||||
)
|
||||
|
||||
|
||||
class VVVVIDIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?vvvvid\.it/#!(?:show|anime|film|series)/(?P<show_id>\d+)/[^/]+/(?P<season_id>\d+)/(?P<id>[0-9]+)'
|
||||
_TESTS = [{
|
||||
# video_type == 'video/vvvvid'
|
||||
'url': 'https://www.vvvvid.it/#!show/434/perche-dovrei-guardarlo-di-dario-moccia/437/489048/ping-pong',
|
||||
'md5': 'b8d3cecc2e981adc3835adf07f6df91b',
|
||||
'info_dict': {
|
||||
'id': '489048',
|
||||
'ext': 'mp4',
|
||||
'title': 'Ping Pong',
|
||||
},
|
||||
}, {
|
||||
# video_type == 'video/rcs'
|
||||
'url': 'https://www.vvvvid.it/#!show/376/death-note-live-action/377/482493/episodio-01',
|
||||
'md5': '33e0edfba720ad73a8782157fdebc648',
|
||||
'info_dict': {
|
||||
'id': '482493',
|
||||
'ext': 'mp4',
|
||||
'title': 'Episodio 01',
|
||||
},
|
||||
}]
|
||||
_conn_id = None
|
||||
|
||||
def _real_initialize(self):
|
||||
if not self._conn_id:
|
||||
user = self._downloader.cache.load('vvvvid', 'user') or {}
|
||||
self._conn_id = user.get('conn_id')
|
||||
if not self._conn_id:
|
||||
self._conn_id = self._download_json(
|
||||
'https://www.vvvvid.it/user/login',
|
||||
None, headers=self.geo_verification_headers())['data']['conn_id']
|
||||
self._downloader.cache.store(
|
||||
'vvvvid', 'user', {
|
||||
'conn_id': self._conn_id,
|
||||
})
|
||||
|
||||
def _real_extract(self, url):
|
||||
show_id, season_id, video_id = re.match(self._VALID_URL, url).groups()
|
||||
response = self._download_json(
|
||||
'https://www.vvvvid.it/vvvvid/ondemand/%s/season/%s' % (show_id, season_id),
|
||||
video_id, headers=self.geo_verification_headers(), query={
|
||||
'conn_id': self._conn_id,
|
||||
})
|
||||
if response['result'] == 'error':
|
||||
raise ExtractorError('%s said: %s' % (
|
||||
self.IE_NAME, response['message']), expected=True)
|
||||
|
||||
vid = int(video_id)
|
||||
video_data = list(filter(
|
||||
lambda episode: episode.get('video_id') == vid, response['data']))[0]
|
||||
formats = []
|
||||
|
||||
# vvvvid embed_info decryption algorithm is reverse engineered from function $ds(h) at vvvvid.js
|
||||
def ds(h):
|
||||
g = "MNOPIJKL89+/4567UVWXQRSTEFGHABCDcdefYZabstuvopqr0123wxyzklmnghij"
|
||||
|
||||
def f(m):
|
||||
l = []
|
||||
o = 0
|
||||
b = False
|
||||
m_len = len(m)
|
||||
while ((not b) and o < m_len):
|
||||
n = m[o] << 2
|
||||
o += 1
|
||||
k = -1
|
||||
j = -1
|
||||
if o < m_len:
|
||||
n += m[o] >> 4
|
||||
o += 1
|
||||
if o < m_len:
|
||||
k = (m[o - 1] << 4) & 255
|
||||
k += m[o] >> 2
|
||||
o += 1
|
||||
if o < m_len:
|
||||
j = (m[o - 1] << 6) & 255
|
||||
j += m[o]
|
||||
o += 1
|
||||
else:
|
||||
b = True
|
||||
else:
|
||||
b = True
|
||||
else:
|
||||
b = True
|
||||
l.append(n)
|
||||
if k != -1:
|
||||
l.append(k)
|
||||
if j != -1:
|
||||
l.append(j)
|
||||
return l
|
||||
|
||||
c = []
|
||||
for e in h:
|
||||
c.append(g.index(e))
|
||||
|
||||
c_len = len(c)
|
||||
for e in range(c_len * 2 - 1, -1, -1):
|
||||
a = c[e % c_len] ^ c[(e + 1) % c_len]
|
||||
c[e % c_len] = a
|
||||
|
||||
c = f(c)
|
||||
d = ''
|
||||
for e in c:
|
||||
d += chr(e)
|
||||
|
||||
return d
|
||||
|
||||
for quality in ('_sd', ''):
|
||||
embed_code = video_data.get('embed_info' + quality)
|
||||
if not embed_code:
|
||||
continue
|
||||
embed_code = ds(embed_code)
|
||||
video_type = video_data.get('video_type')
|
||||
if video_type in ('video/rcs', 'video/kenc'):
|
||||
formats.extend(self._extract_akamai_formats(
|
||||
embed_code, video_id))
|
||||
else:
|
||||
formats.extend(self._extract_wowza_formats(
|
||||
'http://sb.top-ix.org/videomg/_definst_/mp4:%s/playlist.m3u8' % embed_code, video_id))
|
||||
self._sort_formats(formats)
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': video_data['title'],
|
||||
'formats': formats,
|
||||
'thumbnail': video_data.get('thumbnail'),
|
||||
'duration': int_or_none(video_data.get('length')),
|
||||
'series': video_data.get('show_title'),
|
||||
'season_id': season_id,
|
||||
'season_number': video_data.get('season_number'),
|
||||
'episode_id': str_or_none(video_data.get('id')),
|
||||
'epidode_number': int_or_none(video_data.get('number')),
|
||||
'episode_title': video_data['title'],
|
||||
'view_count': int_or_none(video_data.get('views')),
|
||||
'like_count': int_or_none(video_data.get('video_likes')),
|
||||
}
|
Loading…
Reference in New Issue
Block a user