mirror of
https://codeberg.org/polarisfm/youtube-dl
synced 2024-11-30 12:17:54 +01:00
96 lines
3.0 KiB
Python
96 lines
3.0 KiB
Python
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
determine_ext,
|
|
int_or_none,
|
|
parse_duration,
|
|
url_or_none,
|
|
)
|
|
|
|
|
|
class YouJizzIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:\w+\.)?youjizz\.com/videos/(?:[^/#?]*-(?P<id>\d+)\.html|embed/(?P<embed_id>\d+))'
|
|
_TESTS = [{
|
|
'url': 'http://www.youjizz.com/videos/zeichentrick-1-2189178.html',
|
|
'md5': 'b1e1dfaa8bb9537d8b84eeda9cf4acf4',
|
|
'info_dict': {
|
|
'id': '2189178',
|
|
'ext': 'mp4',
|
|
'title': 'Zeichentrick 1',
|
|
'age_limit': 18,
|
|
'duration': 2874,
|
|
}
|
|
}, {
|
|
'url': 'http://www.youjizz.com/videos/-2189178.html',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'https://www.youjizz.com/videos/embed/31991001',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('id') or mobj.group('embed_id')
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
title = self._html_search_regex(
|
|
r'<title>(.+?)</title>', webpage, 'title')
|
|
|
|
formats = []
|
|
|
|
encodings = self._parse_json(
|
|
self._search_regex(
|
|
r'encodings\s*=\s*(\[.+?\]);\n', webpage, 'encodings',
|
|
default='[]'),
|
|
video_id, fatal=False)
|
|
for encoding in encodings:
|
|
if not isinstance(encoding, dict):
|
|
continue
|
|
format_url = url_or_none(encoding.get('filename'))
|
|
if not format_url:
|
|
continue
|
|
if determine_ext(format_url) == 'm3u8':
|
|
formats.extend(self._extract_m3u8_formats(
|
|
format_url, video_id, 'mp4', entry_protocol='m3u8_native',
|
|
m3u8_id='hls', fatal=False))
|
|
else:
|
|
format_id = encoding.get('name') or encoding.get('quality')
|
|
height = int_or_none(self._search_regex(
|
|
r'^(\d+)[pP]', format_id, 'height', default=None))
|
|
formats.append({
|
|
'url': format_url,
|
|
'format_id': format_id,
|
|
'height': height,
|
|
})
|
|
|
|
if formats:
|
|
info_dict = {
|
|
'formats': formats,
|
|
}
|
|
else:
|
|
# YouJizz's HTML5 player has invalid HTML
|
|
webpage = webpage.replace('"controls', '" controls')
|
|
info_dict = self._parse_html5_media_entries(
|
|
url, webpage, video_id)[0]
|
|
|
|
duration = parse_duration(self._search_regex(
|
|
r'<strong>Runtime:</strong>([^<]+)', webpage, 'duration',
|
|
default=None))
|
|
uploader = self._search_regex(
|
|
r'<strong>Uploaded By:.*?<a[^>]*>([^<]+)', webpage, 'uploader',
|
|
default=None)
|
|
|
|
info_dict.update({
|
|
'id': video_id,
|
|
'title': title,
|
|
'age_limit': self._rta_search(webpage),
|
|
'duration': duration,
|
|
'uploader': uploader,
|
|
})
|
|
|
|
return info_dict
|