from .common import InfoExtractor from ..utils import ( parse_duration, parse_filesize, str_to_int, ) class SnotrIE(InfoExtractor): _VALID_URL = r'http?://(?:www\.)?snotr\.com/video/(?P\d+)/([\w]+)' _TESTS = [{ 'url': 'http://www.snotr.com/video/13708/Drone_flying_through_fireworks', 'info_dict': { 'id': '13708', 'ext': 'mp4', 'title': 'Drone flying through fireworks!', 'duration': 248, 'filesize_approx': 40700000, 'description': 'A drone flying through Fourth of July Fireworks', 'thumbnail': r're:^https?://.*\.jpg$', }, 'expected_warnings': ['description'], }, { 'url': 'http://www.snotr.com/video/530/David_Letteman_-_George_W_Bush_Top_10', 'info_dict': { 'id': '530', 'ext': 'mp4', 'title': 'David Letteman - George W. Bush Top 10', 'duration': 126, 'filesize_approx': 8500000, 'description': 'The top 10 George W. Bush moments, brought to you by David Letterman!', 'thumbnail': r're:^https?://.*\.jpg$', } }] def _real_extract(self, url): mobj = self._match_valid_url(url) video_id = mobj.group('id') webpage = self._download_webpage(url, video_id) title = self._og_search_title(webpage) description = self._og_search_description(webpage) info_dict = self._parse_html5_media_entries( url, webpage, video_id, m3u8_entry_protocol='m3u8_native')[0] view_count = str_to_int(self._html_search_regex( r']*>\s*]*>Views:\s*]*>([\d,\.]+)', webpage, 'view count', fatal=False)) duration = parse_duration(self._html_search_regex( r']*>\s*]*>Length:\s*]*>([\d:]+)', webpage, 'duration', fatal=False)) filesize_approx = parse_filesize(self._html_search_regex( r']*>\s*]*>Filesize:\s*]*>([^<]+)', webpage, 'filesize', fatal=False)) info_dict.update({ 'id': video_id, 'description': description, 'title': title, 'view_count': view_count, 'duration': duration, 'filesize_approx': filesize_approx, }) return info_dict