0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-14 01:05:06 +00:00
yt-dlp/yt_dlp/extractor/tenplay.py

171 lines
6.5 KiB
Python
Raw Normal View History

import base64
import datetime as dt
import functools
import itertools
from .common import InfoExtractor
from ..networking import HEADRequest
from ..utils import int_or_none, traverse_obj, urlencode_postdata, urljoin
class TenPlayIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?10play\.com\.au/(?:[^/]+/)+(?P<id>tpv\d{6}[a-z]{5})'
_NETRC_MACHINE = '10play'
_TESTS = [{
'url': 'https://10play.com.au/neighbours/web-extras/season-39/nathan-borg-is-the-first-aussie-actor-with-a-cochlear-implant-to-join-neighbours/tpv210128qupwd',
'info_dict': {
'id': '6226844312001',
'ext': 'mp4',
'title': 'Nathan Borg Is The First Aussie Actor With A Cochlear Implant To Join Neighbours',
'alt_title': 'Nathan Borg Is The First Aussie Actor With A Cochlear Implant To Join Neighbours',
'description': 'md5:a02d0199c901c2dd4c796f1e7dd0de43',
'duration': 186,
'season': 'Season 39',
'season_number': 39,
'series': 'Neighbours',
'thumbnail': r're:https://.*\.jpg',
'uploader': 'Channel 10',
'age_limit': 15,
'timestamp': 1611810000,
'upload_date': '20210128',
'uploader_id': '2199827728001',
},
'params': {
'skip_download': True,
},
'skip': 'Only available in Australia',
}, {
'url': 'https://10play.com.au/todd-sampsons-body-hack/episodes/season-4/episode-7/tpv200921kvngh',
'info_dict': {
'id': '6192880312001',
'ext': 'mp4',
'title': "Todd Sampson's Body Hack - S4 Ep. 2",
'description': 'md5:fa278820ad90f08ea187f9458316ac74',
'age_limit': 15,
'timestamp': 1600770600,
'upload_date': '20200922',
'uploader': 'Channel 10',
'uploader_id': '2199827728001',
},
'params': {
'skip_download': True,
},
}, {
'url': 'https://10play.com.au/how-to-stay-married/web-extras/season-1/terrys-talks-ep-1-embracing-change/tpv190915ylupc',
'only_matching': True,
}]
2021-01-01 12:26:37 +00:00
_GEO_BYPASS = False
_AUS_AGES = {
'G': 0,
'PG': 15,
'M': 15,
'MA': 15,
'MA15+': 15,
'R': 18,
'X': 18,
}
def _get_bearer_token(self, video_id):
username, password = self._get_login_info()
if username is None or password is None:
self.raise_login_required('Your 10play account\'s details must be provided with --username and --password.')
_timestamp = dt.datetime.now().strftime('%Y%m%d000000')
_auth_header = base64.b64encode(_timestamp.encode('ascii')).decode('ascii')
data = self._download_json('https://10play.com.au/api/user/auth', video_id, 'Getting bearer token', headers={
'X-Network-Ten-Auth': _auth_header,
}, data=urlencode_postdata({
'email': username,
'password': password,
}))
return 'Bearer ' + data['jwt']['accessToken']
def _real_extract(self, url):
content_id = self._match_id(url)
data = self._download_json(
'https://10play.com.au/api/v1/videos/' + content_id, content_id)
headers = {}
if data.get('memberGated') is True:
_token = self._get_bearer_token(content_id)
headers = {'Authorization': _token}
_video_url = self._download_json(
data.get('playbackApiEndpoint'), content_id, 'Downloading video JSON',
headers=headers).get('source')
2021-01-01 12:26:37 +00:00
m3u8_url = self._request_webpage(HEADRequest(
_video_url), content_id).url
2021-01-01 12:26:37 +00:00
if '10play-not-in-oz' in m3u8_url:
self.raise_geo_restricted(countries=['AU'])
formats = self._extract_m3u8_formats(m3u8_url, content_id, 'mp4')
return {
2021-01-01 12:26:37 +00:00
'formats': formats,
'subtitles': {'en': [{'url': data.get('captionUrl')}]} if data.get('captionUrl') else None,
'id': data.get('altId') or content_id,
'duration': data.get('duration'),
'title': data.get('subtitle'),
'alt_title': data.get('title'),
'description': data.get('description'),
'age_limit': self._AUS_AGES.get(data.get('classification')),
'series': data.get('tvShow'),
'season_number': int_or_none(data.get('season')),
'episode_number': int_or_none(data.get('episode')),
'timestamp': data.get('published'),
'thumbnail': data.get('imageUrl'),
'uploader': 'Channel 10',
2021-01-01 12:26:37 +00:00
'uploader_id': '2199827728001',
}
class TenPlaySeasonIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?10play\.com\.au/(?P<show>[^/?#]+)/episodes/(?P<season>[^/?#]+)/?(?:$|[?#])'
_TESTS = [{
'url': 'https://10play.com.au/masterchef/episodes/season-14',
'info_dict': {
'title': 'Season 14',
'id': 'MjMyOTIy',
},
'playlist_mincount': 64,
}, {
'url': 'https://10play.com.au/the-bold-and-the-beautiful-fast-tracked/episodes/season-2022',
'info_dict': {
'title': 'Season 2022',
'id': 'Mjc0OTIw',
},
'playlist_mincount': 256,
}]
def _entries(self, load_more_url, display_id=None):
skip_ids = []
for page in itertools.count(1):
episodes_carousel = self._download_json(
load_more_url, display_id, query={'skipIds[]': skip_ids},
note=f'Fetching episodes page {page}')
episodes_chunk = episodes_carousel['items']
skip_ids.extend(ep['id'] for ep in episodes_chunk)
for ep in episodes_chunk:
yield ep['cardLink']
if not episodes_carousel['hasMore']:
break
def _real_extract(self, url):
show, season = self._match_valid_url(url).group('show', 'season')
season_info = self._download_json(
f'https://10play.com.au/api/shows/{show}/episodes/{season}', f'{show}/{season}')
episodes_carousel = traverse_obj(season_info, (
'content', 0, 'components', (
lambda _, v: v['title'].lower() == 'episodes',
(..., {dict}),
)), get_all=False) or {}
playlist_id = episodes_carousel['tpId']
return self.playlist_from_matches(
self._entries(urljoin(url, episodes_carousel['loadMoreUrl']), playlist_id),
playlist_id, traverse_obj(season_info, ('content', 0, 'title', {str})),
getter=functools.partial(urljoin, url))