mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-14 19:17:53 +01:00
63 lines
2.0 KiB
Python
63 lines
2.0 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import json
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
js_to_json,
|
|
qualities,
|
|
)
|
|
|
|
|
|
class TassIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:tass\.ru|itar-tass\.com)/[^/]+/(?P<id>\d+)'
|
|
_TESTS = [
|
|
{
|
|
'url': 'http://tass.ru/obschestvo/1586870',
|
|
'md5': '3b4cdd011bc59174596b6145cda474a4',
|
|
'info_dict': {
|
|
'id': '1586870',
|
|
'ext': 'mp4',
|
|
'title': 'Посетителям московского зоопарка показали красную панду',
|
|
'description': 'Приехавшую из Дублина Зейну можно увидеть в павильоне "Кошки тропиков"',
|
|
'thumbnail': r're:^https?://.*\.jpg$',
|
|
},
|
|
},
|
|
{
|
|
'url': 'http://itar-tass.com/obschestvo/1600009',
|
|
'only_matching': True,
|
|
},
|
|
]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
sources = json.loads(js_to_json(self._search_regex(
|
|
r'(?s)sources\s*:\s*(\[.+?\])', webpage, 'sources')))
|
|
|
|
quality = qualities(['sd', 'hd'])
|
|
|
|
formats = []
|
|
for source in sources:
|
|
video_url = source.get('file')
|
|
if not video_url or not video_url.startswith('http') or not video_url.endswith('.mp4'):
|
|
continue
|
|
label = source.get('label')
|
|
formats.append({
|
|
'url': video_url,
|
|
'format_id': label,
|
|
'quality': quality(label),
|
|
})
|
|
self._sort_formats(formats)
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': self._og_search_title(webpage),
|
|
'description': self._og_search_description(webpage),
|
|
'thumbnail': self._og_search_thumbnail(webpage),
|
|
'formats': formats,
|
|
}
|