X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/474aa0db26d02763aedd55afc30b4897534576e8..refs/heads/master:/youtube_dl/extractor/redtube.py?ds=inline diff --git a/youtube_dl/extractor/redtube.py b/youtube_dl/extractor/redtube.py index b1bde1e..2d2f6a9 100644 --- a/youtube_dl/extractor/redtube.py +++ b/youtube_dl/extractor/redtube.py @@ -4,6 +4,7 @@ import re from .common import InfoExtractor from ..utils import ( + determine_ext, ExtractorError, int_or_none, merge_dicts, @@ -57,7 +58,7 @@ class RedTubeIE(InfoExtractor): if not info.get('title'): info['title'] = self._html_search_regex( - (r']+class="(?:video_title_text|videoTitle)[^"]*">(?P(?:(?!\1).)+)</h\1>', + (r'<h(\d)[^>]+class="(?:video_title_text|videoTitle|video_title)[^"]*">(?P<title>(?:(?!\1).)+)</h\1>', r'(?:videoTitle|title)\s*:\s*(["\'])(?P<title>(?:(?!\1).)+)\1',), webpage, 'title', group='title', default=None) or self._og_search_title(webpage) @@ -77,7 +78,7 @@ class RedTubeIE(InfoExtractor): }) medias = self._parse_json( self._search_regex( - r'mediaDefinition\s*:\s*(\[.+?\])', webpage, + r'mediaDefinition["\']?\s*:\s*(\[.+?}\s*\])', webpage, 'media definitions', default='{}'), video_id, fatal=False) if medias and isinstance(medias, list): @@ -85,6 +86,12 @@ class RedTubeIE(InfoExtractor): format_url = url_or_none(media.get('videoUrl')) if not format_url: continue + if media.get('format') == 'hls' or 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)) + continue format_id = media.get('quality') formats.append({ 'url': format_url,