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/saitosan.py

75 lines
2.9 KiB
Python
Raw Normal View History

from .common import InfoExtractor
from ..utils import ExtractorError, try_get
class SaitosanIE(InfoExtractor):
IE_NAME = 'Saitosan'
_VALID_URL = r'https?://(?:www\.)?saitosan\.net/bview.html\?id=(?P<id>[0-9]+)'
_TESTS = [{
'url': 'http://www.saitosan.net/bview.html?id=10031846',
'info_dict': {
'id': '10031846',
'ext': 'mp4',
'title': '井下原 和弥',
'uploader': '井下原 和弥',
'thumbnail': 'http://111.171.196.85:8088/921f916f-7f55-4c97-b92e-5d9d0fef8f5f/thumb',
'is_live': True,
},
'params': {
# m3u8 download
'skip_download': True,
},
'skip': 'Broadcasts are ephemeral',
},
{
'url': 'http://www.saitosan.net/bview.html?id=10031795',
'info_dict': {
'id': '10031795',
'ext': 'mp4',
'title': '橋本',
'uploader': '橋本',
'thumbnail': 'http://111.171.196.85:8088/1a3933e1-a01a-483b-8931-af15f37f8082/thumb',
'is_live': True,
},
'params': {
# m3u8 download
'skip_download': True,
},
'skip': 'Broadcasts are ephemeral',
}]
def _real_extract(self, url):
b_id = self._match_id(url)
base = 'http://hankachi.saitosan-api.net:8002/socket.io/?transport=polling&EIO=3'
sid = self._download_socket_json(base, b_id, note='Opening socket').get('sid')
base += '&sid=' + sid
self._download_webpage(base, b_id, note='Polling socket')
payload = '420["room_start_join",{"room_id":"%s"}]' % b_id
payload = '%s:%s' % (len(payload), payload)
self._download_webpage(base, b_id, data=payload, note='Polling socket with payload')
response = self._download_socket_json(base, b_id, note='Polling socket')
if not response.get('ok'):
err = response.get('error') or {}
raise ExtractorError(
'%s said: %s - %s' % (self.IE_NAME, err.get('code', '?'), err.get('msg', 'Unknown')) if err
else 'The socket reported that the broadcast could not be joined. Maybe it\'s offline or the URL is incorrect',
expected=True, video_id=b_id)
self._download_webpage(base, b_id, data='26:421["room_finish_join",{}]', note='Polling socket')
b_data = self._download_socket_json(base, b_id, note='Getting broadcast metadata from socket')
m3u8_url = b_data.get('url')
self._download_webpage(base, b_id, data='1:1', note='Closing socket', fatal=False)
return {
'id': b_id,
'title': b_data.get('name'),
'formats': self._extract_m3u8_formats(m3u8_url, b_id, 'mp4', live=True),
'thumbnail': m3u8_url.replace('av.m3u8', 'thumb'),
'uploader': try_get(b_data, lambda x: x['broadcast_user']['name']), # same as title
'is_live': True
}