X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/099764c0c91f4ad7db03d9347798f8619383ea7e..eeb287ed58128659e5de200bba1e0762fb972560:/youtube_dl/extractor/addanime.py?ds=sidebyside diff --git a/youtube_dl/extractor/addanime.py b/youtube_dl/extractor/addanime.py index 11f149f..203936e 100644 --- a/youtube_dl/extractor/addanime.py +++ b/youtube_dl/extractor/addanime.py @@ -15,8 +15,7 @@ from ..utils import ( class AddAnimeIE(InfoExtractor): - - _VALID_URL = r'^http://(?:\w+\.)?add-anime\.net/watch_video\.php\?(?:.*?)v=(?P[\w_]+)(?:.*)' + _VALID_URL = r'^http://(?:\w+\.)?add-anime\.net/watch_video\.php\?(?:.*?)v=(?P[\w_]+)(?:.*)' _TEST = { 'url': 'http://www.add-anime.net/watch_video.php?v=24MR3YO5SAS9', 'md5': '72954ea10bc979ab5e2eb288b21425a0', @@ -29,9 +28,9 @@ class AddAnimeIE(InfoExtractor): } def _real_extract(self, url): + video_id = self._match_id(url) + try: - mobj = re.match(self._VALID_URL, url) - video_id = mobj.group('video_id') webpage = self._download_webpage(url, video_id) except ExtractorError as ee: if not isinstance(ee.cause, compat_HTTPError) or \ @@ -49,7 +48,7 @@ class AddAnimeIE(InfoExtractor): r'a\.value = ([0-9]+)[+]([0-9]+)[*]([0-9]+);', redir_webpage) if av is None: - raise ExtractorError(u'Cannot find redirect math task') + raise ExtractorError('Cannot find redirect math task') av_res = int(av.group(1)) + int(av.group(2)) * int(av.group(3)) parsed_url = compat_urllib_parse_urlparse(url)