mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-10-31 07:50:11 +01:00
34df1c1f60
Authored by: seproDev
84 lines
3.0 KiB
Python
84 lines
3.0 KiB
Python
from .common import InfoExtractor
|
|
from ..utils import (
|
|
ExtractorError,
|
|
mimetype2ext,
|
|
url_or_none,
|
|
)
|
|
from ..utils.traversal import traverse_obj
|
|
|
|
|
|
class VidlyIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:vid\.ly/|(?:s\.)?vid\.ly/embeded\.html\?(?:[^#]+&)?link=)(?P<id>\w+)'
|
|
_EMBED_REGEX = [r'<script[^>]+\bsrc=[\'"](?P<url>(?:https?:)?//vid\.ly/\w+/embed[^\'"]+)',
|
|
r'<iframe[^>]+\bsrc=[\'"](?P<url>(?:https?:)?//(?:s\.)?vid\.ly/embeded\.html\?(?:[^#\'"]+&)?link=\w+[^\'"]+)']
|
|
_TESTS = [{
|
|
# JWPlayer 7, Embeds forbidden
|
|
'url': 'https://vid.ly/2i3o9j/embed',
|
|
'info_dict': {
|
|
'id': '2i3o9j',
|
|
'ext': 'mp4',
|
|
'title': '2i3o9j',
|
|
'thumbnail': r're:https://\w+\.cloudfront\.net/',
|
|
},
|
|
}, {
|
|
# JWPlayer 6
|
|
'url': 'http://s.vid.ly/embeded.html?link=jw_test&new=1&autoplay=true&controls=true',
|
|
'info_dict': {
|
|
'id': 'jw_test',
|
|
'ext': 'mp4',
|
|
'title': '2x8m8t',
|
|
'thumbnail': r're:https://\w+\.cloudfront\.net/',
|
|
},
|
|
}, {
|
|
# Vidlyplayer
|
|
'url': 'https://vid.ly/7x0e6l',
|
|
'info_dict': {
|
|
'id': '7x0e6l',
|
|
'ext': 'mp4',
|
|
'title': '7x0e6l',
|
|
},
|
|
}]
|
|
_WEBPAGE_TESTS = [{
|
|
'url': 'https://www.petfinder.com/dog/gus-57378930/tn/ooltewah/furever-furkids-rescue-tn592/',
|
|
'info_dict': {
|
|
'id': 'w8p5b0',
|
|
'ext': 'mp4',
|
|
'title': 'w8p5b0',
|
|
'thumbnail': r're:https://\w+\.cloudfront\.net/',
|
|
}
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
embed_script = self._download_webpage(
|
|
f'https://vid.ly/{video_id}/embed', video_id, headers={'Referer': 'https://vid.ly/'})
|
|
player = self._search_json(r'initCallback\(', embed_script, 'player', video_id)
|
|
|
|
player_type = player.get('player') or ''
|
|
if player_type.startswith('jwplayer'):
|
|
return self._parse_jwplayer_data(player['config'], video_id)
|
|
elif not player_type.startswith('vidly'):
|
|
raise ExtractorError(f'Unknown player type {player_type!r}')
|
|
|
|
formats = []
|
|
ext = mimetype2ext(traverse_obj(player, ('config', 'type')))
|
|
for source, fid in [('source', 'sd'), ('source_hd', 'hd')]:
|
|
if traverse_obj(player, ('config', source, {url_or_none})):
|
|
formats.append({
|
|
'url': player['config'][source],
|
|
'format_id': f'http-{fid}',
|
|
'ext': ext,
|
|
})
|
|
# Has higher quality formats
|
|
formats.extend(self._extract_m3u8_formats(
|
|
f'https://d3fenhwk93s16g.cloudfront.net/{video_id}/hls.m3u8', video_id,
|
|
fatal=False, note='Requesting higher quality m3u8 formats',
|
|
errnote='No higher quality m3u8 formats found') or [])
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': video_id,
|
|
'formats': formats,
|
|
}
|