mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-05 07:53:18 +00:00
86e5f3ed2e
Using https://github.com/asottile/pyupgrade 1. `__future__` imports and `coding: utf-8` were removed 2. Files were rewritten with `pyupgrade --py36-plus --keep-percent-format` 3. f-strings were cherry-picked from `pyupgrade --py36-plus` Extractors are left untouched (except removing header) to avoid unnecessary merge conflicts
132 lines
5 KiB
Python
132 lines
5 KiB
Python
from .common import InfoExtractor
|
|
from ..utils import (
|
|
float_or_none,
|
|
int_or_none,
|
|
parse_iso8601,
|
|
try_get,
|
|
)
|
|
|
|
|
|
class NineCNineMediaIE(InfoExtractor):
|
|
IE_NAME = '9c9media'
|
|
_GEO_COUNTRIES = ['CA']
|
|
_VALID_URL = r'9c9media:(?P<destination_code>[^:]+):(?P<id>\d+)'
|
|
_API_BASE_TEMPLATE = 'http://capi.9c9media.com/destinations/%s/platforms/desktop/contents/%s/'
|
|
|
|
def _real_extract(self, url):
|
|
destination_code, content_id = self._match_valid_url(url).groups()
|
|
api_base_url = self._API_BASE_TEMPLATE % (destination_code, content_id)
|
|
content = self._download_json(api_base_url, content_id, query={
|
|
'$include': '[Media.Name,Season,ContentPackages.Duration,ContentPackages.Id]',
|
|
})
|
|
title = content['Name']
|
|
content_package = content['ContentPackages'][0]
|
|
package_id = content_package['Id']
|
|
content_package_url = api_base_url + 'contentpackages/%s/' % package_id
|
|
content_package = self._download_json(
|
|
content_package_url, content_id, query={
|
|
'$include': '[HasClosedCaptions]',
|
|
})
|
|
|
|
if (not self.get_param('allow_unplayable_formats')
|
|
and try_get(content_package, lambda x: x['Constraints']['Security']['Type'])):
|
|
self.report_drm(content_id)
|
|
|
|
manifest_base_url = content_package_url + 'manifest.'
|
|
formats = []
|
|
formats.extend(self._extract_m3u8_formats(
|
|
manifest_base_url + 'm3u8', content_id, 'mp4',
|
|
'm3u8_native', m3u8_id='hls', fatal=False))
|
|
formats.extend(self._extract_f4m_formats(
|
|
manifest_base_url + 'f4m', content_id,
|
|
f4m_id='hds', fatal=False))
|
|
formats.extend(self._extract_mpd_formats(
|
|
manifest_base_url + 'mpd', content_id,
|
|
mpd_id='dash', fatal=False))
|
|
self._sort_formats(formats)
|
|
|
|
thumbnails = []
|
|
for image in (content.get('Images') or []):
|
|
image_url = image.get('Url')
|
|
if not image_url:
|
|
continue
|
|
thumbnails.append({
|
|
'url': image_url,
|
|
'width': int_or_none(image.get('Width')),
|
|
'height': int_or_none(image.get('Height')),
|
|
})
|
|
|
|
tags, categories = [], []
|
|
for source_name, container in (('Tags', tags), ('Genres', categories)):
|
|
for e in content.get(source_name, []):
|
|
e_name = e.get('Name')
|
|
if not e_name:
|
|
continue
|
|
container.append(e_name)
|
|
|
|
season = content.get('Season') or {}
|
|
|
|
info = {
|
|
'id': content_id,
|
|
'title': title,
|
|
'description': content.get('Desc') or content.get('ShortDesc'),
|
|
'timestamp': parse_iso8601(content.get('BroadcastDateTime')),
|
|
'episode_number': int_or_none(content.get('Episode')),
|
|
'season': season.get('Name'),
|
|
'season_number': int_or_none(season.get('Number')),
|
|
'season_id': season.get('Id'),
|
|
'series': try_get(content, lambda x: x['Media']['Name']),
|
|
'tags': tags,
|
|
'categories': categories,
|
|
'duration': float_or_none(content_package.get('Duration')),
|
|
'formats': formats,
|
|
'thumbnails': thumbnails,
|
|
}
|
|
|
|
if content_package.get('HasClosedCaptions'):
|
|
info['subtitles'] = {
|
|
'en': [{
|
|
'url': manifest_base_url + 'vtt',
|
|
'ext': 'vtt',
|
|
}, {
|
|
'url': manifest_base_url + 'srt',
|
|
'ext': 'srt',
|
|
}]
|
|
}
|
|
|
|
return info
|
|
|
|
|
|
class CPTwentyFourIE(InfoExtractor):
|
|
IE_NAME = 'cp24'
|
|
_GEO_COUNTRIES = ['CA']
|
|
_VALID_URL = r'https?://(?:www\.)?cp24\.com/news/(?P<id>[^?#]+)'
|
|
|
|
_TESTS = [{
|
|
'url': 'https://www.cp24.com/news/video-shows-atm-being-ripped-out-of-business-by-pickup-truck-driver-in-mississauga-1.5676877',
|
|
'info_dict': {
|
|
'id': '2328005',
|
|
'ext': 'mp4',
|
|
'title': 'WATCH: Truck rips ATM from Mississauga business',
|
|
'description': 'md5:cf7498480885f080a754389a2b2f7073',
|
|
'timestamp': 1637618377,
|
|
'episode_number': None,
|
|
'season': 'Season 0',
|
|
'season_number': 0,
|
|
'season_id': 57974,
|
|
'series': 'CTV News Toronto',
|
|
'duration': 26.86,
|
|
'thumbnail': 'http://images2.9c9media.com/image_asset/2014_11_5_2eb609a0-475b-0132-fbd6-34b52f6f1279_jpg_2000x1125.jpg',
|
|
'upload_date': '20211122',
|
|
},
|
|
'params': {'skip_download': True, 'format': 'bv'}
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
display_id = self._match_id(url)
|
|
webpage = self._download_webpage(url, display_id)
|
|
id, destination = self._search_regex(
|
|
r'getAuthStates\("(?P<id>[^"]+)",\s?"(?P<destination>[^"]+)"\);',
|
|
webpage, 'video id and destination', group=('id', 'destination'))
|
|
return self.url_result(f'9c9media:{destination}:{id}', ie=NineCNineMediaIE.ie_key(), video_id=id)
|