0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-23 02:25:11 +00:00
yt-dlp/yt_dlp/extractor/mixch.py

119 lines
4 KiB
Python
Raw Normal View History

from .common import InfoExtractor
from ..networking.exceptions import HTTPError
from ..utils import (
ExtractorError,
UserNotLive,
int_or_none,
str_or_none,
url_or_none,
)
from ..utils.traversal import traverse_obj
class MixchIE(InfoExtractor):
IE_NAME = 'mixch'
_VALID_URL = r'https?://(?:www\.)?mixch\.tv/u/(?P<id>\d+)'
_TESTS = [{
'url': 'https://mixch.tv/u/16943797/live',
'skip': 'don\'t know if this live persists',
'info_dict': {
'id': '16943797',
'ext': 'mp4',
'title': '#EntView #カリナ #セブチ 2024-05-05 06:58',
'comment_count': int,
'view_count': int,
'timestamp': 1714726805,
'uploader': 'Ent.View K-news🎶💕',
'uploader_id': '16943797',
'live_status': 'is_live',
'upload_date': '20240503',
},
}, {
'url': 'https://mixch.tv/u/16137876/live',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
data = self._download_json(f'https://mixch.tv/api-web/users/{video_id}/live', video_id)
if not traverse_obj(data, ('liveInfo', {dict})):
raise UserNotLive(video_id=video_id)
return {
'id': video_id,
'uploader_id': video_id,
**traverse_obj(data, {
'title': ('liveInfo', 'title', {str}),
'comment_count': ('liveInfo', 'comments', {int_or_none}),
'view_count': ('liveInfo', 'visitor', {int_or_none}),
'timestamp': ('liveInfo', 'created', {int_or_none}),
'uploader': ('broadcasterInfo', 'name', {str}),
}),
'formats': [{
'format_id': 'hls',
'url': data['liveInfo']['hls'],
'ext': 'mp4',
'protocol': 'm3u8',
}],
'is_live': True,
'__post_extractor': self.extract_comments(video_id),
}
def _get_comments(self, video_id):
yield from traverse_obj(self._download_json(
f'https://mixch.tv/api-web/lives/{video_id}/messages', video_id,
note='Downloading comments', errnote='Failed to download comments'), (..., {
'author': ('name', {str}),
'author_id': ('user_id', {str_or_none}),
'id': ('message_id', {str}, {lambda x: x or None}),
'text': ('body', {str}),
'timestamp': ('created', {int}),
}))
class MixchArchiveIE(InfoExtractor):
IE_NAME = 'mixch:archive'
_VALID_URL = r'https?://(?:www\.)?mixch\.tv/archive/(?P<id>\d+)'
_TESTS = [{
'url': 'https://mixch.tv/archive/421',
'skip': 'paid video, no DRM. expires at Jan 23',
'info_dict': {
'id': '421',
'ext': 'mp4',
'title': '96NEKO SHOW TIME',
},
}, {
'url': 'https://mixch.tv/archive/1213',
'skip': 'paid video, no DRM. expires at Dec 31, 2023',
'info_dict': {
'id': '1213',
'ext': 'mp4',
'title': '【特別トーク番組アーカイブス】Merm4id×燐舞曲 2nd LIVE「VERSUS」',
'release_date': '20231201',
'thumbnail': str,
},
}, {
'url': 'https://mixch.tv/archive/1214',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
try:
info_json = self._download_json(
f'https://mixch.tv/api-web/archive/{video_id}', video_id)['archive']
except ExtractorError as e:
if isinstance(e.cause, HTTPError) and e.cause.status == 401:
self.raise_login_required()
raise
return {
'id': video_id,
'title': traverse_obj(info_json, ('title', {str})),
'formats': self._extract_m3u8_formats(info_json['archiveURL'], video_id),
'thumbnail': traverse_obj(info_json, ('thumbnailURL', {url_or_none})),
}