mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2025-06-27 17:08:32 +00:00
79 lines
2.3 KiB
Python
79 lines
2.3 KiB
Python
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
smuggle_url,
|
|
unsmuggle_url,
|
|
)
|
|
from ..utils.traversal import traverse_obj
|
|
|
|
|
|
class DzsecurityLiveIE(InfoExtractor):
|
|
_VALID_URL = r'https?://live\.dzsecurity\.net/live/player/(?P<id>[\w-]+)'
|
|
_EMBED_REGEX = [rf'<iframe[^>]+\bsrc\s*=\s*["\'](?P<url>{_VALID_URL})']
|
|
|
|
_WEBPAGE_TESTS = [{
|
|
'url': 'https://www.echoroukonline.com/live',
|
|
'info_dict': {
|
|
'id': 'echorouktv',
|
|
'title': str,
|
|
'ext': 'mp4',
|
|
'live_status': 'is_live',
|
|
},
|
|
}, {
|
|
'url': 'https://www.echoroukonline.com/live-news',
|
|
'info_dict': {
|
|
'id': 'echorouknews',
|
|
'title': str,
|
|
'ext': 'mp4',
|
|
'live_status': 'is_live',
|
|
},
|
|
}, {
|
|
'url': 'https://elhayat.dz/%D8%A7%D9%84%D8%A8%D8%AB-%D8%A7%D9%84%D8%AD%D9%8A/',
|
|
'info_dict': {
|
|
'id': 'elhayattv',
|
|
'title': str,
|
|
'ext': 'mp4',
|
|
'live_status': 'is_live',
|
|
},
|
|
}, {
|
|
'url': 'https://www.ennaharonline.com/live',
|
|
'info_dict': {
|
|
'id': 'ennahartv',
|
|
'title': str,
|
|
'ext': 'mp4',
|
|
'live_status': 'is_live',
|
|
},
|
|
'skip': 'Geo-restricted to Algeria',
|
|
}]
|
|
|
|
@classmethod
|
|
def _extract_embed_urls(cls, url, webpage):
|
|
for embed_url in super()._extract_embed_urls(url, webpage):
|
|
yield smuggle_url(embed_url, {'referer': url})
|
|
|
|
def _real_extract(self, url):
|
|
url, smuggled_data = unsmuggle_url(url, {})
|
|
|
|
stream_id = self._match_id(url)
|
|
|
|
title = stream_id
|
|
referer = smuggled_data['referer']
|
|
if referer:
|
|
webpage = self._download_webpage(referer, referer)
|
|
title = self._html_extract_title(webpage, default=title)
|
|
|
|
player_page = self._download_webpage(url, stream_id, headers=traverse_obj(smuggled_data, {'Referer': 'referer'}))
|
|
|
|
m3u8_url = 'https:' + self._search_regex(
|
|
r'src:\s*location\.protocol\s*\+\s*"(//[^"]+\.m3u8\?[^"]+)"',
|
|
player_page,
|
|
'm3u8 URL',
|
|
)
|
|
|
|
return {
|
|
'id': stream_id,
|
|
'title': title,
|
|
'formats': self._extract_m3u8_formats(m3u8_url, stream_id),
|
|
'is_live': True,
|
|
}
|