mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-15 05:33:05 +00:00
92 lines
4.3 KiB
Python
92 lines
4.3 KiB
Python
from .common import InfoExtractor
|
|
from ..utils import extract_attributes, multipart_encode, url_or_none
|
|
from ..utils.traversal import traverse_obj
|
|
|
|
|
|
class PiaLiveIE(InfoExtractor):
|
|
PLAYER_ROOT_URL = 'https://player.pia-live.jp/'
|
|
PIA_LIVE_API_URL = 'https://api.pia-live.jp'
|
|
API_KEY = 'kfds)FKFps-dms9e'
|
|
_VALID_URL = r'https?://player\.pia-live\.jp/stream/(?P<id>[\w-]+)'
|
|
|
|
_TESTS = [
|
|
{
|
|
'url': 'https://player.pia-live.jp/stream/4JagFBEIM14s_hK9aXHKf3k3F3bY5eoHFQxu68TC6krUDqGOwN4d61dCWQYOd6CTxl4hjya9dsfEZGsM4uGOUdax60lEI4twsXGXf7crmz8Gk__GhupTrWxA7RFRVt76',
|
|
'info_dict': {
|
|
'id': '88f3109a-f503-4d0f-a9f7-9f39ac745d84',
|
|
'display_id': '2431867_001',
|
|
'title': 'こながめでたい日2024の視聴ページ | PIA LIVE STREAM(ぴあライブストリーム)',
|
|
'live_status': 'was_live',
|
|
'comment_count': int,
|
|
},
|
|
'params': {
|
|
'getcomments': True,
|
|
'skip_download': True,
|
|
'ignore_no_formats_error': True,
|
|
},
|
|
},
|
|
{
|
|
'url': 'https://player.pia-live.jp/stream/4JagFBEIM14s_hK9aXHKf3k3F3bY5eoHFQxu68TC6krJdu0GVBVbVy01IwpJ6J3qBEm3d9TCTt1d0eWpsZGj7DrOjVOmS7GAWGwyscMgiThopJvzgWC4H5b-7XQjAfRZ',
|
|
'info_dict': {
|
|
'id': '9ce8b8ba-f6d1-4d1f-83a0-18c3148ded93',
|
|
'display_id': '2431867_002',
|
|
'title': 'こながめでたい日2024の視聴ページ | PIA LIVE STREAM(ぴあライブストリーム)',
|
|
'live_status': 'was_live',
|
|
'comment_count': int,
|
|
},
|
|
'params': {
|
|
'getcomments': True,
|
|
'skip_download': True,
|
|
'ignore_no_formats_error': True,
|
|
},
|
|
},
|
|
]
|
|
|
|
def _extract_vars(self, variable, html):
|
|
return self._search_regex(
|
|
rf'(?:var|const|let)\s+{variable}\s*=\s*(["\'])(?P<value>(?:(?!\1).)+)\1',
|
|
html, f'variable {variable}', group='value')
|
|
|
|
def _real_extract(self, url):
|
|
video_key = self._match_id(url)
|
|
webpage = self._download_webpage(url, video_key)
|
|
|
|
program_code = self._extract_vars('programCode', webpage)
|
|
article_code = self._extract_vars('articleCode', webpage)
|
|
|
|
payload, content_type = multipart_encode({
|
|
'play_url': video_key,
|
|
'api_key': self.API_KEY})
|
|
|
|
player_tag_list = self._download_json(
|
|
f'{self.PIA_LIVE_API_URL}/perf/player-tag-list/{program_code}', program_code,
|
|
data=payload, headers={'Content-Type': content_type, 'Referer': self.PLAYER_ROOT_URL},
|
|
note='Fetching player tag list', errnote='Unable to fetch player tag list')
|
|
if self.get_param('getcomments'):
|
|
chat_room_url = traverse_obj(self._download_json(
|
|
f'{self.PIA_LIVE_API_URL}/perf/chat-tag-list/{program_code}/{article_code}', program_code,
|
|
data=payload, headers={'Content-Type': content_type, 'Referer': self.PLAYER_ROOT_URL},
|
|
note='Fetching chat info', errnote='Unable to fetch chat info', fatal=False),
|
|
('data', 'chat_one_tag', {extract_attributes}, 'src', {url_or_none}))
|
|
|
|
return self.url_result(
|
|
extract_attributes(player_tag_list['data']['movie_one_tag'])['src'], url_transparent=True,
|
|
video_title=self._html_extract_title(webpage), display_id=program_code,
|
|
__post_extractor=self.extract_comments(program_code, chat_room_url))
|
|
|
|
def _get_comments(self, video_id, chat_room_url):
|
|
if not chat_room_url:
|
|
return
|
|
if comment_page := self._download_webpage(
|
|
chat_room_url, video_id, headers={'Referer': f'{self.PLAYER_ROOT_URL}'},
|
|
note='Fetching comment page', errnote='Unable to fetch comment page', fatal=False):
|
|
yield from traverse_obj(self._search_json(
|
|
r'var\s+_history\s*=', comment_page, 'comment list',
|
|
video_id, contains_pattern=r'\[(?s:.+)\]', fatal=False), (..., {
|
|
'timestamp': 0,
|
|
'author_is_uploader': (1, {lambda x: x == 2}),
|
|
'author': 2,
|
|
'text': 3,
|
|
'id': 4,
|
|
}))
|