mirror of
https://github.com/yt-dlp/yt-dlp
synced 2024-12-26 21:59:08 +01:00
-Updates URL scheme
-Includes suggested changes -Updates test cases
This commit is contained in:
parent
2aa9e75975
commit
38e08a1147
2 changed files with 38 additions and 48 deletions
|
@ -2480,11 +2480,11 @@ from .xinpianchang import XinpianchangIE
|
|||
from .xminus import XMinusIE
|
||||
from .xnxx import XNXXIE
|
||||
from .xstream import XstreamIE
|
||||
from .xumo import XumoIE
|
||||
from .xvideos import (
|
||||
XVideosIE,
|
||||
XVideosQuickiesIE,
|
||||
)
|
||||
from .xumo import XumoIE
|
||||
from .xxxymovies import XXXYMoviesIE
|
||||
from .yahoo import (
|
||||
YahooIE,
|
||||
|
|
|
@ -1,46 +1,39 @@
|
|||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
determine_ext,
|
||||
int_or_none,
|
||||
smuggle_url,
|
||||
traverse_obj,
|
||||
unsmuggle_url,
|
||||
url_or_none
|
||||
)
|
||||
from ..utils import determine_ext, int_or_none, smuggle_url, traverse_obj, unsmuggle_url, url_or_none
|
||||
|
||||
|
||||
class XumoIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?xumo\.tv/[^?#]+/(?P<id>XM[A-Z0-9]{12})'
|
||||
_VALID_URL = r'https?://play\.?xumo\.com/[^?#]+/(?P<id>XM[A-Z0-9]{12})'
|
||||
_TESTS = [{
|
||||
# movie
|
||||
'url': 'https://www.xumo.tv/free-movies/zombie-high/XM01OEG1MIJVJR',
|
||||
'url': 'https://play.xumo.com/free-movies/a-circus-tale-and-a-love-song/XM041I5U497VD3',
|
||||
'params': {
|
||||
'check_formats': True
|
||||
'check_formats': True,
|
||||
},
|
||||
'md5': '2843529c1abf5f86f61165637f89cbae',
|
||||
'md5': 'eaac858a8db4ee5a67d6d16920c24e15',
|
||||
'info_dict': {
|
||||
'id': 'XM01OEG1MIJVJR',
|
||||
'title': 'Zombie High',
|
||||
'id': 'XM041I5U497VD3',
|
||||
'title': 'A Circus Tale & A Love Song',
|
||||
'ext': 'mp4',
|
||||
'description': 'md5:b620cc0afc18a91fb69275047f6c4d22',
|
||||
'duration': 5436,
|
||||
'release_year': 1987,
|
||||
'thumbnail': r're:^https?://.*\.jpg$'
|
||||
}
|
||||
'description': 'md5:aa6372f4785c528ff04c94a275f63446',
|
||||
'duration': 6887,
|
||||
'release_year': 2016,
|
||||
'thumbnail': r're:^https?://.*\.jpg$',
|
||||
},
|
||||
}, {
|
||||
# series
|
||||
'url': 'https://www.xumo.tv/tv-shows/super-mario-world/XM0AN69OG47PRN',
|
||||
# entire series
|
||||
'url': 'https://play.xumo.com/tv-shows/super-mario-world/XM0AN69OG47PRN',
|
||||
'params': {
|
||||
'skip_download': True
|
||||
'skip_download': True,
|
||||
},
|
||||
'playlist_count': 10,
|
||||
'info_dict': {
|
||||
'id': 'XM0AN69OG47PRN',
|
||||
'title': 'Super Mario World'
|
||||
}
|
||||
'title': 'Super Mario World',
|
||||
},
|
||||
}, {
|
||||
# video from network URL
|
||||
'url': 'https://www.xumo.tv/networks/free-movies/99991299/XM02D369HADFRR',
|
||||
# episode of series
|
||||
'url': 'https://play.xumo.com/tv-shows/99991299/XM02D369HADFRR',
|
||||
'md5': 'ed2f396272b39f2e0fe47f02b9ae34cc',
|
||||
'info_dict': {
|
||||
'id': 'XM02D369HADFRR',
|
||||
|
@ -54,10 +47,10 @@ class XumoIE(InfoExtractor):
|
|||
'thumbnail': r're:^https?://.*\.jpg$',
|
||||
'season': 'Season 1',
|
||||
'duration': 1368,
|
||||
}
|
||||
},
|
||||
}, {
|
||||
# episode from generated URL
|
||||
'url': 'https://www.xumo.tv/free-movies/x/XM02D369HADFRR',
|
||||
# video from network-based alternate URL scheme
|
||||
'url': 'https://play.xumo.com/networks/fakenetworkname/99991299/XM02D369HADFRR',
|
||||
'md5': 'ed2f396272b39f2e0fe47f02b9ae34cc',
|
||||
'info_dict': {
|
||||
'id': 'XM02D369HADFRR',
|
||||
|
@ -71,7 +64,7 @@ class XumoIE(InfoExtractor):
|
|||
'thumbnail': r're:^https?://.*\.jpg$',
|
||||
'season': 'Season 1',
|
||||
'duration': 1368,
|
||||
}
|
||||
},
|
||||
}]
|
||||
|
||||
_INFO_URL = 'https://valencia-app-mds.xumo.com/v2/assets/asset/'
|
||||
|
@ -97,7 +90,7 @@ class XumoIE(InfoExtractor):
|
|||
'episodes.descriptions',
|
||||
'episodes.hasCaptions',
|
||||
'episodes.ratings',
|
||||
]
|
||||
],
|
||||
}
|
||||
|
||||
def _get_video_links(self, video_id, info_json):
|
||||
|
@ -127,14 +120,10 @@ class XumoIE(InfoExtractor):
|
|||
formats.extend(fmts)
|
||||
subtitles = self._merge_subtitles(subtitles, subs)
|
||||
|
||||
for caption in traverse_obj(info_json, ('providers', ..., 'captions', ...)) or []:
|
||||
sub_url = caption.get('url')
|
||||
ext = determine_ext(sub_url)
|
||||
if ext not in ('vtt', 'srt'):
|
||||
continue
|
||||
lang = caption.get('lang') or 'und'
|
||||
subtitles.setdefault(lang, []).append({'ext': ext, 'url': caption.get('url')})
|
||||
|
||||
for caption in traverse_obj(info_json, ('providers', ..., 'captions', ...)):
|
||||
subtitles.setdefault(caption.get('lang') or 'und', []).append({
|
||||
'url': caption.get('url'),
|
||||
})
|
||||
return formats, subtitles
|
||||
|
||||
def _real_extract(self, url):
|
||||
|
@ -148,12 +137,13 @@ class XumoIE(InfoExtractor):
|
|||
content_type = media_metadata['contentType']
|
||||
|
||||
if content_type == 'SERIES':
|
||||
# series => return set of URLs pointing to episodes
|
||||
video_links = []
|
||||
for episode in traverse_obj(media_metadata, ('seasons', ..., 'episodes', ...)) or []:
|
||||
smuggled_url = smuggle_url('https://www.xumo.tv/free-movies/x/' + episode['id'], {'series_title': traverse_obj(media_metadata, 'title')})
|
||||
video_links.append(self.url_result(smuggled_url, video_title=episode.get('episodeTitle')))
|
||||
return self.playlist_result(video_links, playlist_id=media_id, playlist_title=title)
|
||||
# series => return set of URLs pointing to individual episodes and smuggle series title to avoid extra API call for each episode
|
||||
return self.playlist_result([
|
||||
self.url_result(
|
||||
smuggle_url(f'https://play.xumo.com/tv-shows/x/{episode["id"]}', {'series': title}),
|
||||
XumoIE, episode['id'], episode.get('episodeTitle'))
|
||||
for episode in traverse_obj(media_metadata, ('seasons', ..., 'episodes', ...))
|
||||
], media_id, title)
|
||||
|
||||
# video => return video info
|
||||
|
||||
|
@ -165,7 +155,7 @@ class XumoIE(InfoExtractor):
|
|||
if is_episode:
|
||||
season_number = int_or_none(media_metadata.get('season'))
|
||||
if smuggled_data:
|
||||
series_title = traverse_obj(smuggled_data, 'series_title')
|
||||
series_title = traverse_obj(smuggled_data, 'series')
|
||||
else:
|
||||
series_data = self._download_json(f'{self._INFO_URL}{media_metadata["connectorId"]}.json', media_id, query=self._INFO_QUERY_PARAMS)
|
||||
series_title = traverse_obj(series_data, 'title')
|
||||
|
@ -184,5 +174,5 @@ class XumoIE(InfoExtractor):
|
|||
'episode_number': media_metadata.get('episode'),
|
||||
'season_number': season_number,
|
||||
'episode': title if is_episode else None,
|
||||
'series': series_title
|
||||
'series': series_title,
|
||||
}
|
||||
|
|
Loading…
Reference in a new issue