X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/ced7488f6d3a519b2c1b1cbd31048743fb8285bd..fe6cc1b2aebf52f3fcdcc22ca8e846b67a406cc0:/youtube_dl/extractor/miomio.py?ds=sidebyside diff --git a/youtube_dl/extractor/miomio.py b/youtube_dl/extractor/miomio.py index 937ba0f..40f72d6 100644 --- a/youtube_dl/extractor/miomio.py +++ b/youtube_dl/extractor/miomio.py @@ -25,10 +25,7 @@ class MioMioIE(InfoExtractor): 'title': '【SKY】字幕 铠武昭和VS平成 假面骑士大战FEAT战队 魔星字幕组 字幕', 'duration': 5923, }, - 'params': { - # The server provides broken file - 'skip_download': True, - } + 'skip': 'Unable to load videos', }, { 'url': 'http://www.miomio.tv/watch/cc184024/', 'info_dict': { @@ -47,17 +44,14 @@ class MioMioIE(InfoExtractor): 'skip': 'Unable to load videos', }, { # new 'h5' player - 'url': 'http://www.miomio.tv/watch/cc273295/', - 'md5': '', + 'url': 'http://www.miomio.tv/watch/cc273997/', + 'md5': '0b27a4b4495055d826813f8c3a6b2070', 'info_dict': { - 'id': '273295', + 'id': '273997', 'ext': 'mp4', - 'title': 'アウト×デラックス 20160526', - }, - 'params': { - # intermittent HTTP 500 - 'skip_download': True, + 'title': 'マツコの知らない世界【劇的進化SP!ビニール傘&冷凍食品2016】 1_2 - 16 05 31', }, + 'skip': 'Unable to load videos', }] def _extract_mioplayer(self, webpage, video_id, title, http_headers): @@ -101,9 +95,18 @@ class MioMioIE(InfoExtractor): return entries + def _download_chinese_webpage(self, *args, **kwargs): + # Requests with English locales return garbage + headers = { + 'Accept-Language': 'zh-TW,en-US;q=0.7,en;q=0.3', + } + kwargs.setdefault('headers', {}).update(headers) + return self._download_webpage(*args, **kwargs) + def _real_extract(self, url): video_id = self._match_id(url) - webpage = self._download_webpage(url, video_id) + webpage = self._download_chinese_webpage( + url, video_id) title = self._html_search_meta( 'description', webpage, 'title', fatal=True) @@ -113,10 +116,10 @@ class MioMioIE(InfoExtractor): if '_h5' in mioplayer_path: player_url = compat_urlparse.urljoin(url, mioplayer_path) - player_webpage = self._download_webpage( + player_webpage = self._download_chinese_webpage( player_url, video_id, note='Downloading player webpage', headers={'Referer': url}) - entries = self._parse_html5_media_entries(player_url, player_webpage) + entries = self._parse_html5_media_entries(player_url, player_webpage, video_id) http_headers = {'Referer': player_url} else: http_headers = {'Referer': 'http://www.miomio.tv%s' % mioplayer_path}