mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-24 21:56:53 +01:00
115 lines
3.8 KiB
Python
115 lines
3.8 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
from .adobepass import AdobePassIE
|
|
from .uplynk import UplynkPreplayIE
|
|
from ..compat import compat_str
|
|
from ..utils import (
|
|
HEADRequest,
|
|
int_or_none,
|
|
parse_age_limit,
|
|
parse_duration,
|
|
try_get,
|
|
unified_timestamp,
|
|
)
|
|
|
|
|
|
class FOXIE(AdobePassIE):
|
|
_VALID_URL = r'https?://(?:www\.)?fox\.com/watch/(?P<id>[\da-fA-F]+)'
|
|
_TESTS = [{
|
|
# clip
|
|
'url': 'https://www.fox.com/watch/4b765a60490325103ea69888fb2bd4e8/',
|
|
'md5': 'ebd296fcc41dd4b19f8115d8461a3165',
|
|
'info_dict': {
|
|
'id': '4b765a60490325103ea69888fb2bd4e8',
|
|
'ext': 'mp4',
|
|
'title': 'Aftermath: Bruce Wayne Develops Into The Dark Knight',
|
|
'description': 'md5:549cd9c70d413adb32ce2a779b53b486',
|
|
'duration': 102,
|
|
'timestamp': 1504291893,
|
|
'upload_date': '20170901',
|
|
'creator': 'FOX',
|
|
'series': 'Gotham',
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
# episode, geo-restricted
|
|
'url': 'https://www.fox.com/watch/087036ca7f33c8eb79b08152b4dd75c1/',
|
|
'only_matching': True,
|
|
}, {
|
|
# episode, geo-restricted, tv provided required
|
|
'url': 'https://www.fox.com/watch/30056b295fb57f7452aeeb4920bc3024/',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
video = self._download_json(
|
|
'https://api.fox.com/fbc-content/v1_4/video/%s' % video_id,
|
|
video_id, headers={
|
|
'apikey': 'abdcbed02c124d393b39e818a4312055',
|
|
'Content-Type': 'application/json',
|
|
'Referer': url,
|
|
})
|
|
|
|
title = video['name']
|
|
release_url = video['videoRelease']['url']
|
|
|
|
description = video.get('description')
|
|
duration = int_or_none(video.get('durationInSeconds')) or int_or_none(
|
|
video.get('duration')) or parse_duration(video.get('duration'))
|
|
timestamp = unified_timestamp(video.get('datePublished'))
|
|
age_limit = parse_age_limit(video.get('contentRating'))
|
|
|
|
data = try_get(
|
|
video, lambda x: x['trackingData']['properties'], dict) or {}
|
|
|
|
creator = data.get('brand') or data.get('network') or video.get('network')
|
|
|
|
series = video.get('seriesName') or data.get(
|
|
'seriesName') or data.get('show')
|
|
season_number = int_or_none(video.get('seasonNumber'))
|
|
episode = video.get('name')
|
|
episode_number = int_or_none(video.get('episodeNumber'))
|
|
release_year = int_or_none(video.get('releaseYear'))
|
|
|
|
if data.get('authRequired'):
|
|
# TODO: AP
|
|
pass
|
|
|
|
info = {
|
|
'id': video_id,
|
|
'title': title,
|
|
'description': description,
|
|
'duration': duration,
|
|
'timestamp': timestamp,
|
|
'age_limit': age_limit,
|
|
'creator': creator,
|
|
'series': series,
|
|
'season_number': season_number,
|
|
'episode': episode,
|
|
'episode_number': episode_number,
|
|
'release_year': release_year,
|
|
}
|
|
|
|
urlh = self._request_webpage(HEADRequest(release_url), video_id)
|
|
video_url = compat_str(urlh.geturl())
|
|
|
|
if UplynkPreplayIE.suitable(video_url):
|
|
info.update({
|
|
'_type': 'url_transparent',
|
|
'url': video_url,
|
|
'ie_key': UplynkPreplayIE.ie_key(),
|
|
})
|
|
else:
|
|
m3u8_url = self._download_json(release_url, video_id)['playURL']
|
|
formats = self._extract_m3u8_formats(
|
|
m3u8_url, video_id, 'mp4',
|
|
entry_protocol='m3u8_native', m3u8_id='hls')
|
|
self._sort_formats(formats)
|
|
info['formats'] = formats
|
|
return info
|