yt-dlp/yt_dlp/extractor/arkena.py
2022-11-17 11:40:17 +05:30

151 lines
7.0 KiB
Python

from .common import InfoExtractor
from ..utils import (
ExtractorError,
float_or_none,
int_or_none,
parse_iso8601,
parse_qs,
try_get,
)
class ArkenaIE(InfoExtractor):
_VALID_URL = r'''(?x)
https?://
(?:
video\.(?:arkena|qbrick)\.com/play2/embed/player\?|
play\.arkena\.com/(?:config|embed)/avp/v\d/player/media/(?P<id>[^/]+)/[^/]+/(?P<account_id>\d+)
)
'''
# See https://support.arkena.com/display/PLAY/Ways+to+embed+your+video
_EMBED_REGEX = [r'<iframe[^>]+src=(["\'])(?P<url>(?:https?:)?//play\.arkena\.com/embed/avp/.+?)\1']
_TESTS = [{
'url': 'https://video.qbrick.com/play2/embed/player?accountId=1034090&mediaId=d8ab4607-00090107-aab86310',
'md5': '97f117754e5f3c020f5f26da4a44ebaf',
'info_dict': {
'id': 'd8ab4607-00090107-aab86310',
'ext': 'mp4',
'title': 'EM_HT20_117_roslund_v2.mp4',
'timestamp': 1608285912,
'upload_date': '20201218',
'duration': 1429.162667,
'subtitles': {
'sv': 'count:3',
},
},
}, {
'url': 'https://play.arkena.com/embed/avp/v2/player/media/b41dda37-d8e7-4d3f-b1b5-9a9db578bdfe/1/129411',
'only_matching': True,
}, {
'url': 'https://play.arkena.com/config/avp/v2/player/media/b41dda37-d8e7-4d3f-b1b5-9a9db578bdfe/1/129411/?callbackMethod=jQuery1111023664739129262213_1469227693893',
'only_matching': True,
}, {
'url': 'http://play.arkena.com/config/avp/v1/player/media/327336/darkmatter/131064/?callbackMethod=jQuery1111002221189684892677_1469227595972',
'only_matching': True,
}, {
'url': 'http://play.arkena.com/embed/avp/v1/player/media/327336/darkmatter/131064/',
'only_matching': True,
}, {
'url': 'http://video.arkena.com/play2/embed/player?accountId=472718&mediaId=35763b3b-00090078-bf604299&pageStyling=styled',
'only_matching': True,
}]
def _real_extract(self, url):
mobj = self._match_valid_url(url)
video_id = mobj.group('id')
account_id = mobj.group('account_id')
# Handle http://video.arkena.com/play2/embed/player URL
if not video_id:
qs = parse_qs(url)
video_id = qs.get('mediaId', [None])[0]
account_id = qs.get('accountId', [None])[0]
if not video_id or not account_id:
raise ExtractorError('Invalid URL', expected=True)
media = self._download_json(
'https://video.qbrick.com/api/v1/public/accounts/%s/medias/%s' % (account_id, video_id),
video_id, query={
# https://video.qbrick.com/docs/api/examples/library-api.html
'fields': 'asset/resources/*/renditions/*(height,id,language,links/*(href,mimeType),type,size,videos/*(audios/*(codec,sampleRate),bitrate,codec,duration,height,width),width),created,metadata/*(title,description),tags',
})
metadata = media.get('metadata') or {}
title = metadata['title']
duration = None
formats = []
thumbnails = []
subtitles = {}
for resource in media['asset']['resources']:
for rendition in (resource.get('renditions') or []):
rendition_type = rendition.get('type')
for i, link in enumerate(rendition.get('links') or []):
href = link.get('href')
if not href:
continue
if rendition_type == 'image':
thumbnails.append({
'filesize': int_or_none(rendition.get('size')),
'height': int_or_none(rendition.get('height')),
'id': rendition.get('id'),
'url': href,
'width': int_or_none(rendition.get('width')),
})
elif rendition_type == 'subtitle':
subtitles.setdefault(rendition.get('language') or 'en', []).append({
'url': href,
})
elif rendition_type == 'video':
f = {
'filesize': int_or_none(rendition.get('size')),
'format_id': rendition.get('id'),
'url': href,
}
video = try_get(rendition, lambda x: x['videos'][i], dict)
if video:
if not duration:
duration = float_or_none(video.get('duration'))
f.update({
'height': int_or_none(video.get('height')),
'tbr': int_or_none(video.get('bitrate'), 1000),
'vcodec': video.get('codec'),
'width': int_or_none(video.get('width')),
})
audio = try_get(video, lambda x: x['audios'][0], dict)
if audio:
f.update({
'acodec': audio.get('codec'),
'asr': int_or_none(audio.get('sampleRate')),
})
formats.append(f)
elif rendition_type == 'index':
mime_type = link.get('mimeType')
if mime_type == 'application/smil+xml':
formats.extend(self._extract_smil_formats(
href, video_id, fatal=False))
elif mime_type == 'application/x-mpegURL':
formats.extend(self._extract_m3u8_formats(
href, video_id, 'mp4', 'm3u8_native',
m3u8_id='hls', fatal=False))
elif mime_type == 'application/hds+xml':
formats.extend(self._extract_f4m_formats(
href, video_id, f4m_id='hds', fatal=False))
elif mime_type == 'application/dash+xml':
formats.extend(self._extract_f4m_formats(
href, video_id, f4m_id='hds', fatal=False))
elif mime_type == 'application/vnd.ms-sstr+xml':
formats.extend(self._extract_ism_formats(
href, video_id, ism_id='mss', fatal=False))
return {
'id': video_id,
'title': title,
'description': metadata.get('description'),
'timestamp': parse_iso8601(media.get('created')),
'thumbnails': thumbnails,
'subtitles': subtitles,
'duration': duration,
'tags': media.get('tags'),
'formats': formats,
}