mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-23 09:01:43 +00:00
[picarto] Extract more metadata (closes #16518)
This commit is contained in:
parent
f17a24a6df
commit
730c0d12a0
@ -1,6 +1,7 @@
|
||||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
import time
|
||||
|
||||
from .common import InfoExtractor
|
||||
@ -8,6 +9,7 @@
|
||||
from ..utils import (
|
||||
ExtractorError,
|
||||
js_to_json,
|
||||
try_get,
|
||||
update_url_query,
|
||||
urlencode_postdata,
|
||||
)
|
||||
@ -32,7 +34,9 @@ def suitable(cls, url):
|
||||
return False if PicartoVodIE.suitable(url) else super(PicartoIE, cls).suitable(url)
|
||||
|
||||
def _real_extract(self, url):
|
||||
channel_id = self._match_id(url)
|
||||
mobj = re.match(self._VALID_URL, url)
|
||||
channel_id = mobj.group('id')
|
||||
|
||||
metadata = self._download_json(
|
||||
'https://api.picarto.tv/v1/channel/name/' + channel_id,
|
||||
channel_id)
|
||||
@ -45,7 +49,7 @@ def _real_extract(self, url):
|
||||
data=urlencode_postdata({'loadbalancinginfo': channel_id}),
|
||||
note='Downloading load balancing info')
|
||||
|
||||
token = self._VALID_URL_RE.match(url).group('token') or 'public'
|
||||
token = mobj.group('token') or 'public'
|
||||
params = {
|
||||
'con': int(time.time() * 1000),
|
||||
'token': token,
|
||||
@ -99,9 +103,11 @@ def _real_extract(self, url):
|
||||
|
||||
return {
|
||||
'id': channel_id,
|
||||
'title': self._live_title(channel_id),
|
||||
'title': self._live_title(metadata.get('title') or channel_id),
|
||||
'is_live': True,
|
||||
'thumbnail': metadata.get('thumbnails', {}).get('web'),
|
||||
'thumbnail': try_get(metadata, lambda x: x['thumbnails']['web']),
|
||||
'channel': channel_id,
|
||||
'channel_url': 'https://picarto.tv/%s' % channel_id,
|
||||
'age_limit': age_limit,
|
||||
'formats': formats,
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user