mirror of
				https://github.com/yt-dlp/yt-dlp.git
				synced 2025-10-25 03:31:03 +00:00 
			
		
		
		
	
		
			
				
	
	
		
			128 lines
		
	
	
		
			3.8 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
			
		
		
	
	
			128 lines
		
	
	
		
			3.8 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
| # coding: utf-8
 | |
| from __future__ import unicode_literals
 | |
| 
 | |
| from .common import InfoExtractor
 | |
| from ..utils import (
 | |
|     ExtractorError,
 | |
|     js_to_json,
 | |
| )
 | |
| 
 | |
| 
 | |
| class PicartoIE(InfoExtractor):
 | |
|     _VALID_URL = r'https?://(?:www.)?picarto\.tv/(?P<id>[a-zA-Z0-9]+)'
 | |
|     _TEST = {
 | |
|         'url': 'https://picarto.tv/Setz',
 | |
|         'info_dict': {
 | |
|             'id': 'Setz',
 | |
|             'ext': 'mp4',
 | |
|             'title': 're:^Setz [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$',
 | |
|             'timestamp': int,
 | |
|             'is_live': True
 | |
|         },
 | |
|         'skip': 'Stream is offline',
 | |
|     }
 | |
| 
 | |
|     @classmethod
 | |
|     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)
 | |
| 
 | |
|         data = self._download_json(
 | |
|             'https://ptvintern.picarto.tv/ptvapi', channel_id, query={
 | |
|                 'query': '''{
 | |
|   channel(name: "%s") {
 | |
|     adult
 | |
|     id
 | |
|     online
 | |
|     stream_name
 | |
|     title
 | |
|   }
 | |
|   getLoadBalancerUrl(channel_name: "%s") {
 | |
|     url
 | |
|   }
 | |
| }''' % (channel_id, channel_id),
 | |
|             })['data']
 | |
|         metadata = data['channel']
 | |
| 
 | |
|         if metadata.get('online') == 0:
 | |
|             raise ExtractorError('Stream is offline', expected=True)
 | |
|         title = metadata['title']
 | |
| 
 | |
|         cdn_data = self._download_json(
 | |
|             data['getLoadBalancerUrl']['url'] + '/stream/json_' + metadata['stream_name'] + '.js',
 | |
|             channel_id, 'Downloading load balancing info')
 | |
| 
 | |
|         formats = []
 | |
|         for source in (cdn_data.get('source') or []):
 | |
|             source_url = source.get('url')
 | |
|             if not source_url:
 | |
|                 continue
 | |
|             source_type = source.get('type')
 | |
|             if source_type == 'html5/application/vnd.apple.mpegurl':
 | |
|                 formats.extend(self._extract_m3u8_formats(
 | |
|                     source_url, channel_id, 'mp4', m3u8_id='hls', fatal=False))
 | |
|             elif source_type == 'html5/video/mp4':
 | |
|                 formats.append({
 | |
|                     'url': source_url,
 | |
|                 })
 | |
|         self._sort_formats(formats)
 | |
| 
 | |
|         mature = metadata.get('adult')
 | |
|         if mature is None:
 | |
|             age_limit = None
 | |
|         else:
 | |
|             age_limit = 18 if mature is True else 0
 | |
| 
 | |
|         return {
 | |
|             'id': channel_id,
 | |
|             'title': title.strip(),
 | |
|             'is_live': True,
 | |
|             'channel': channel_id,
 | |
|             'channel_id': metadata.get('id'),
 | |
|             'channel_url': 'https://picarto.tv/%s' % channel_id,
 | |
|             'age_limit': age_limit,
 | |
|             'formats': formats,
 | |
|         }
 | |
| 
 | |
| 
 | |
| class PicartoVodIE(InfoExtractor):
 | |
|     _VALID_URL = r'https?://(?:www.)?picarto\.tv/videopopout/(?P<id>[^/?#&]+)'
 | |
|     _TESTS = [{
 | |
|         'url': 'https://picarto.tv/videopopout/ArtofZod_2017.12.12.00.13.23.flv',
 | |
|         'md5': '3ab45ba4352c52ee841a28fb73f2d9ca',
 | |
|         'info_dict': {
 | |
|             'id': 'ArtofZod_2017.12.12.00.13.23.flv',
 | |
|             'ext': 'mp4',
 | |
|             'title': 'ArtofZod_2017.12.12.00.13.23.flv',
 | |
|             'thumbnail': r're:^https?://.*\.jpg'
 | |
|         },
 | |
|     }, {
 | |
|         'url': 'https://picarto.tv/videopopout/Plague',
 | |
|         'only_matching': True,
 | |
|     }]
 | |
| 
 | |
|     def _real_extract(self, url):
 | |
|         video_id = self._match_id(url)
 | |
| 
 | |
|         webpage = self._download_webpage(url, video_id)
 | |
| 
 | |
|         vod_info = self._parse_json(
 | |
|             self._search_regex(
 | |
|                 r'(?s)#vod-player["\']\s*,\s*(\{.+?\})\s*\)', webpage,
 | |
|                 'vod player'),
 | |
|             video_id, transform_source=js_to_json)
 | |
| 
 | |
|         formats = self._extract_m3u8_formats(
 | |
|             vod_info['vod'], video_id, 'mp4', entry_protocol='m3u8_native',
 | |
|             m3u8_id='hls')
 | |
|         self._sort_formats(formats)
 | |
| 
 | |
|         return {
 | |
|             'id': video_id,
 | |
|             'title': video_id,
 | |
|             'thumbnail': vod_info.get('vodThumb'),
 | |
|             'formats': formats,
 | |
|         }
 | 
