yt-dlp/yt_dlp/extractor/viqeo.py

88 lines
3 KiB
Python
Raw Normal View History

2018-07-29 16:05:36 -04:00
from .common import InfoExtractor
from ..utils import (
int_or_none,
str_or_none,
url_or_none,
)
class ViqeoIE(InfoExtractor):
_VALID_URL = r'''(?x)
(?:
viqeo:|
https?://cdn\.viqeo\.tv/embed/*\?.*?\bvid=|
https?://api\.viqeo\.tv/v\d+/data/startup?.*?\bvideo(?:%5B%5D|\[\])=
)
(?P<id>[\da-f]+)
'''
_EMBED_REGEX = [r'<iframe[^>]+\bsrc=(["\'])(?P<url>(?:https?:)?//cdn\.viqeo\.tv/embed/*\?.*?\bvid=[\da-f]+.*?)\1']
2018-07-29 16:05:36 -04:00
_TESTS = [{
'url': 'https://cdn.viqeo.tv/embed/?vid=cde96f09d25f39bee837',
'md5': 'a169dd1a6426b350dca4296226f21e76',
'info_dict': {
'id': 'cde96f09d25f39bee837',
'ext': 'mp4',
'title': 'cde96f09d25f39bee837',
'thumbnail': r're:^https?://.*\.jpg$',
'duration': 76,
},
}, {
'url': 'viqeo:cde96f09d25f39bee837',
'only_matching': True,
}, {
'url': 'https://api.viqeo.tv/v1/data/startup?video%5B%5D=71bbec412ade45c3216c&profile=112',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(
'https://cdn.viqeo.tv/embed/?vid=%s' % video_id, video_id)
data = self._parse_json(
self._search_regex(
r'SLOT_DATA\s*=\s*({.+?})\s*;', webpage, 'slot data'),
video_id)
formats = []
thumbnails = []
for media_file in data['mediaFiles']:
if not isinstance(media_file, dict):
continue
media_url = url_or_none(media_file.get('url'))
if not media_url or not media_url.startswith(('http', '//')):
continue
media_type = str_or_none(media_file.get('type'))
if not media_type:
continue
media_kind = media_type.split('/')[0].lower()
f = {
'url': media_url,
'width': int_or_none(media_file.get('width')),
'height': int_or_none(media_file.get('height')),
}
format_id = str_or_none(media_file.get('quality'))
if media_kind == 'image':
f['id'] = format_id
thumbnails.append(f)
elif media_kind in ('video', 'audio'):
is_audio = media_kind == 'audio'
f.update({
'format_id': 'audio' if is_audio else format_id,
'fps': int_or_none(media_file.get('fps')),
'vcodec': 'none' if is_audio else None,
})
formats.append(f)
self._sort_formats(formats)
duration = int_or_none(data.get('duration'))
return {
'id': video_id,
'title': video_id,
'duration': duration,
'thumbnails': thumbnails,
'formats': formats,
}