mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-04 23:35:04 +00:00
53 lines
1.9 KiB
Python
53 lines
1.9 KiB
Python
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
float_or_none,
|
|
int_or_none,
|
|
)
|
|
|
|
|
|
class DotsubIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?dotsub\.com/view/(?P<id>[^/]+)'
|
|
_TEST = {
|
|
'url': 'https://dotsub.com/view/9c63db2a-fa95-4838-8e6e-13deafe47f09',
|
|
'md5': '21c7ff600f545358134fea762a6d42b6',
|
|
'info_dict': {
|
|
'id': '9c63db2a-fa95-4838-8e6e-13deafe47f09',
|
|
'ext': 'flv',
|
|
'title': 'MOTIVATION - "It\'s Possible" Best Inspirational Video Ever',
|
|
'description': 'md5:41af1e273edbbdfe4e216a78b9d34ac6',
|
|
'thumbnail': 're:^https?://dotsub.com/media/9c63db2a-fa95-4838-8e6e-13deafe47f09/p',
|
|
'duration': 198,
|
|
'uploader': 'liuxt',
|
|
'timestamp': 1385778501.104,
|
|
'upload_date': '20131130',
|
|
'view_count': int,
|
|
}
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
info = self._download_json(
|
|
'https://dotsub.com/api/media/%s/metadata' % video_id, video_id)
|
|
video_url = info.get('mediaURI')
|
|
|
|
if not video_url:
|
|
webpage = self._download_webpage(url, video_id)
|
|
video_url = self._search_regex(
|
|
[r'<source[^>]+src="([^"]+)"', r'"file"\s*:\s*\'([^\']+)'],
|
|
webpage, 'video url')
|
|
|
|
return {
|
|
'id': video_id,
|
|
'url': video_url,
|
|
'ext': 'flv',
|
|
'title': info['title'],
|
|
'description': info.get('description'),
|
|
'thumbnail': info.get('screenshotURI'),
|
|
'duration': int_or_none(info.get('duration'), 1000),
|
|
'uploader': info.get('user'),
|
|
'timestamp': float_or_none(info.get('dateCreated'), 1000),
|
|
'view_count': int_or_none(info.get('numberOfViews')),
|
|
}
|