class KeezMoviesIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?keezmovies\.com/video/(?:(?P<display_id>[^/]+)-)?(?P<id>\d+)'
_TESTS = [{
- 'url': 'http://www.keezmovies.com/video/petite-asian-lady-mai-playing-in-bathtub-1214711',
- 'md5': '1c1e75d22ffa53320f45eeb07bc4cdc0',
+ 'url': 'https://www.keezmovies.com/video/arab-wife-want-it-so-bad-i-see-she-thirsty-and-has-tiny-money-18070681',
+ 'md5': '2ac69cdb882055f71d82db4311732a1a',
'info_dict': {
- 'id': '1214711',
- 'display_id': 'petite-asian-lady-mai-playing-in-bathtub',
+ 'id': '18070681',
+ 'display_id': 'arab-wife-want-it-so-bad-i-see-she-thirsty-and-has-tiny-money',
'ext': 'mp4',
- 'title': 'Petite Asian Lady Mai Playing In Bathtub',
- 'thumbnail': 're:^https?://.*\.jpg$',
+ 'title': 'Arab wife want it so bad I see she thirsty and has tiny money.',
+ 'thumbnail': None,
'view_count': int,
'age_limit': 18,
}
}, {
- 'url': 'http://www.keezmovies.com/video/1214711',
+ 'url': 'http://www.keezmovies.com/video/18070681',
'only_matching': True,
}]
- def _extract_info(self, url):
+ def _extract_info(self, url, fatal=True):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
- display_id = mobj.group('display_id') or video_id
+ display_id = (mobj.group('display_id')
+ if 'display_id' in mobj.groupdict()
+ else None) or mobj.group('id')
webpage = self._download_webpage(
url, display_id, headers={'Cookie': 'age_verified=1'})
encrypted = False
def extract_format(format_url, height=None):
- if not isinstance(format_url, compat_str) or not format_url.startswith('http'):
+ if not isinstance(format_url, compat_str) or not format_url.startswith(('http', '//')):
return
if format_url in format_urls:
return
raise ExtractorError(
'Video %s is no longer available' % video_id, expected=True)
- self._sort_formats(formats)
+ try:
+ self._sort_formats(formats)
+ except ExtractorError:
+ if fatal:
+ raise
if not title:
title = self._html_search_regex(
}
def _real_extract(self, url):
- webpage, info = self._extract_info(url)
+ webpage, info = self._extract_info(url, fatal=False)
+ if not info['formats']:
+ return self.url_result(url, 'Generic')
info['view_count'] = str_to_int(self._search_regex(
r'<b>([\d,.]+)</b> Views?', webpage, 'view count', fatal=False))
return info