yt-dlp/yt_dlp/extractor/rule34video.py

66 lines
2.4 KiB
Python
Raw Normal View History

import re
from ..utils import parse_duration, unescapeHTML
from .common import InfoExtractor
class Rule34VideoIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?rule34video\.com/videos/(?P<id>\d+)'
_TESTS = [
{
'url': 'https://rule34video.com/videos/3065157/shot-it-mmd-hmv/',
'md5': 'ffccac2c23799dabbd192621ae4d04f3',
'info_dict': {
'id': '3065157',
'ext': 'mp4',
'title': 'Shot It-(mmd hmv)',
'thumbnail': 'https://rule34video.com/contents/videos_screenshots/3065000/3065157/preview.jpg',
'duration': 347.0,
'age_limit': 18,
'tags': 'count:14'
}
},
{
'url': 'https://rule34video.com/videos/3065296/lara-in-trouble-ep-7-wildeerstudio/',
'md5': '6bb5169f9f6b38cd70882bf2e64f6b86',
'info_dict': {
'id': '3065296',
'ext': 'mp4',
'title': 'Lara in Trouble Ep. 7 [WildeerStudio]',
'thumbnail': 'https://rule34video.com/contents/videos_screenshots/3065000/3065296/preview.jpg',
'duration': 938.0,
'age_limit': 18,
'tags': 'count:50'
}
},
]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
formats = []
for mobj in re.finditer(r'<a[^>]+href="(?P<video_url>[^"]+download=true[^"]+)".*>(?P<ext>[^\s]+) (?P<quality>[^<]+)p</a>', webpage):
url, ext, quality = mobj.groups()
formats.append({
'url': url,
'ext': ext.lower(),
'quality': quality,
})
2022-04-04 08:27:35 +00:00
title = self._html_extract_title(webpage)
thumbnail = self._html_search_regex(r'preview_url:\s+\'([^\']+)\'', webpage, 'thumbnail', default=None)
duration = self._html_search_regex(r'"icon-clock"></i>\s+<span>((?:\d+:?)+)', webpage, 'duration', default=None)
return {
'id': video_id,
'formats': formats,
'title': title,
'thumbnail': thumbnail,
'duration': parse_duration(duration),
'age_limit': 18,
'tags': list(map(unescapeHTML, re.findall(
r'<a class="tag_item"[^>]+\bhref="https://rule34video\.com/tags/\d+/"[^>]*>(?P<tag>[^>]*)</a>', webpage))),
}