0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-22 02:15:12 +00:00
yt-dlp/yt_dlp/extractor/xanimu.py
2023-01-07 02:48:34 +05:30

51 lines
2.1 KiB
Python

import re
from .common import InfoExtractor
from ..utils import int_or_none
class XanimuIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?xanimu\.com/(?P<id>[^/]+)/?'
_TESTS = [{
'url': 'https://xanimu.com/51944-the-princess-the-frog-hentai/',
'md5': '899b88091d753d92dad4cb63bbf357a7',
'info_dict': {
'id': '51944-the-princess-the-frog-hentai',
'ext': 'mp4',
'title': 'The Princess + The Frog Hentai',
'thumbnail': 'https://xanimu.com/storage/2020/09/the-princess-and-the-frog-hentai.jpg',
'description': r're:^Enjoy The Princess \+ The Frog Hentai',
'duration': 207.0,
'age_limit': 18
}
}, {
'url': 'https://xanimu.com/huge-expansion/',
'only_matching': True
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
formats = []
for format in ['videoHigh', 'videoLow']:
format_url = self._search_json(r'var\s+%s\s*=' % re.escape(format), webpage, format,
video_id, default=None, contains_pattern=r'[\'"]([^\'"]+)[\'"]')
if format_url:
formats.append({
'url': format_url,
'format_id': format,
'quality': -2 if format.endswith('Low') else None,
})
return {
'id': video_id,
'formats': formats,
'title': self._search_regex(r'[\'"]headline[\'"]:\s*[\'"]([^"]+)[\'"]', webpage,
'title', default=None) or self._html_extract_title(webpage),
'thumbnail': self._html_search_meta('thumbnailUrl', webpage, default=None),
'description': self._html_search_meta('description', webpage, default=None),
'duration': int_or_none(self._search_regex(r'duration:\s*[\'"]([^\'"]+?)[\'"]',
webpage, 'duration', fatal=False)),
'age_limit': 18
}