[telecinco] fix extraction
This commit is contained in:
parent
e474996541
commit
f86b299d0e
|
@ -5,14 +5,11 @@ import json
|
|||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from .ooyala import OoyalaIE
|
||||
from ..utils import (
|
||||
clean_html,
|
||||
determine_ext,
|
||||
int_or_none,
|
||||
str_or_none,
|
||||
try_get,
|
||||
urljoin,
|
||||
)
|
||||
|
||||
|
||||
|
@ -28,7 +25,7 @@ class TelecincoIE(InfoExtractor):
|
|||
'description': 'md5:716caf5601e25c3c5ab6605b1ae71529',
|
||||
},
|
||||
'playlist': [{
|
||||
'md5': 'adb28c37238b675dad0f042292f209a7',
|
||||
'md5': '7ee56d665cfd241c0e6d80fd175068b0',
|
||||
'info_dict': {
|
||||
'id': 'JEA5ijCnF6p5W08A1rNKn7',
|
||||
'ext': 'mp4',
|
||||
|
@ -38,7 +35,7 @@ class TelecincoIE(InfoExtractor):
|
|||
}]
|
||||
}, {
|
||||
'url': 'http://www.cuatro.com/deportes/futbol/barcelona/Leo_Messi-Champions-Roma_2_2052780128.html',
|
||||
'md5': '9468140ebc300fbb8b9d65dc6e5c4b43',
|
||||
'md5': 'c86fe0d99e3bdb46b7950d38bf6ef12a',
|
||||
'info_dict': {
|
||||
'id': 'jn24Od1zGLG4XUZcnUnZB6',
|
||||
'ext': 'mp4',
|
||||
|
@ -48,7 +45,7 @@ class TelecincoIE(InfoExtractor):
|
|||
},
|
||||
}, {
|
||||
'url': 'http://www.mediaset.es/12meses/campanas/doylacara/conlatratanohaytrato/Ayudame-dar-cara-trata-trato_2_1986630220.html',
|
||||
'md5': 'ae2dc6b7b50b2392076a51c0f70e01f6',
|
||||
'md5': 'eddb50291df704ce23c74821b995bcac',
|
||||
'info_dict': {
|
||||
'id': 'aywerkD2Sv1vGNqq9b85Q2',
|
||||
'ext': 'mp4',
|
||||
|
@ -90,58 +87,24 @@ class TelecincoIE(InfoExtractor):
|
|||
|
||||
def _parse_content(self, content, url):
|
||||
video_id = content['dataMediaId']
|
||||
if content.get('dataCmsId') == 'ooyala':
|
||||
return self.url_result(
|
||||
'ooyala:%s' % video_id, OoyalaIE.ie_key(), video_id)
|
||||
config_url = urljoin(url, content['dataConfig'])
|
||||
config = self._download_json(
|
||||
config_url, video_id, 'Downloading config JSON')
|
||||
content['dataConfig'], video_id, 'Downloading config JSON')
|
||||
title = config['info']['title']
|
||||
|
||||
def mmc_url(mmc_type):
|
||||
return re.sub(
|
||||
r'/(?:flash|html5)\.json', '/%s.json' % mmc_type,
|
||||
config['services']['mmc'])
|
||||
|
||||
duration = None
|
||||
formats = []
|
||||
for mmc_type in ('flash', 'html5'):
|
||||
mmc = self._download_json(
|
||||
mmc_url(mmc_type), video_id,
|
||||
'Downloading %s mmc JSON' % mmc_type, fatal=False)
|
||||
if not mmc:
|
||||
continue
|
||||
if not duration:
|
||||
duration = int_or_none(mmc.get('duration'))
|
||||
for location in mmc['locations']:
|
||||
gat = self._proto_relative_url(location.get('gat'), 'http:')
|
||||
gcp = location.get('gcp')
|
||||
ogn = location.get('ogn')
|
||||
if None in (gat, gcp, ogn):
|
||||
continue
|
||||
token_data = {
|
||||
'gcp': gcp,
|
||||
'ogn': ogn,
|
||||
'sta': 0,
|
||||
}
|
||||
media = self._download_json(
|
||||
gat, video_id, data=json.dumps(token_data).encode('utf-8'),
|
||||
headers={
|
||||
'Content-Type': 'application/json;charset=utf-8',
|
||||
'Referer': url,
|
||||
}, fatal=False) or {}
|
||||
stream = media.get('stream') or media.get('file')
|
||||
if not stream:
|
||||
continue
|
||||
ext = determine_ext(stream)
|
||||
if ext == 'f4m':
|
||||
formats.extend(self._extract_f4m_formats(
|
||||
stream + '&hdcore=3.2.0&plugin=aasp-3.2.0.77.18',
|
||||
video_id, f4m_id='hds', fatal=False))
|
||||
elif ext == 'm3u8':
|
||||
formats.extend(self._extract_m3u8_formats(
|
||||
stream, video_id, 'mp4', 'm3u8_native',
|
||||
m3u8_id='hls', fatal=False))
|
||||
services = config['services']
|
||||
caronte = self._download_json(services['caronte'], video_id)
|
||||
stream = caronte['dls'][0]['stream']
|
||||
headers = self.geo_verification_headers()
|
||||
headers.update({
|
||||
'Content-Type': 'application/json;charset=UTF-8',
|
||||
'Origin': re.match(r'https?://[^/]+', url).group(0),
|
||||
})
|
||||
cdn = self._download_json(
|
||||
caronte['cerbero'], video_id, data=json.dumps({
|
||||
'bbx': caronte['bbx'],
|
||||
'gbx': self._download_json(services['gbx'], video_id)['gbx'],
|
||||
}).encode(), headers=headers)['tokens']['1']['cdn']
|
||||
formats = self._extract_m3u8_formats(
|
||||
stream + '?' + cdn, video_id, 'mp4', 'm3u8_native', m3u8_id='hls')
|
||||
self._sort_formats(formats)
|
||||
|
||||
return {
|
||||
|
@ -149,7 +112,7 @@ class TelecincoIE(InfoExtractor):
|
|||
'title': title,
|
||||
'formats': formats,
|
||||
'thumbnail': content.get('dataPoster') or config.get('poster', {}).get('imageUrl'),
|
||||
'duration': duration,
|
||||
'duration': int_or_none(content.get('dataDuration')),
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
|
|
Loading…
Reference in New Issue