X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/6d835d4d6903cf5e18ad01844736929e06d16004..233624c1db781ee7dabbaf88453cf18e248dd20d:/youtube_dl/extractor/nuvid.py diff --git a/youtube_dl/extractor/nuvid.py b/youtube_dl/extractor/nuvid.py index 58ec81f..ab6bfcd 100644 --- a/youtube_dl/extractor/nuvid.py +++ b/youtube_dl/extractor/nuvid.py @@ -5,13 +5,11 @@ import re from .common import InfoExtractor from ..utils import ( parse_duration, - unified_strdate, - compat_urllib_request, ) class NuvidIE(InfoExtractor): - _VALID_URL = r'^https?://(?:www|m)\.nuvid\.com/video/(?P[0-9]+)' + _VALID_URL = r'https?://(?:www|m)\.nuvid\.com/video/(?P[0-9]+)' _TEST = { 'url': 'http://m.nuvid.com/video/1310741/', 'md5': 'eab207b7ac4fccfb4e23c86201f11277', @@ -20,37 +18,38 @@ class NuvidIE(InfoExtractor): 'ext': 'mp4', 'title': 'Horny babes show their awesome bodeis and', 'duration': 129, - 'upload_date': '20140508', 'age_limit': 18, } } def _real_extract(self, url): - mobj = re.match(self._VALID_URL, url) - video_id = mobj.group('id') + video_id = self._match_id(url) - formats = [] + page_url = 'http://m.nuvid.com/video/%s' % video_id + webpage = self._download_webpage( + page_url, video_id, 'Downloading video page') + # When dwnld_speed exists and has a value larger than the MP4 file's + # bitrate, Nuvid returns the MP4 URL + # It's unit is 100bytes/millisecond, see mobile-nuvid-min.js for the algorithm + self._set_cookie('nuvid.com', 'dwnld_speed', '10.0') + mp4_webpage = self._download_webpage( + page_url, video_id, 'Downloading video page for MP4 format') - for dwnld_speed, format_id in [(0, '3gp'), (5, 'mp4')]: - request = compat_urllib_request.Request( - 'http://m.nuvid.com/play/%s' % video_id) - request.add_header('Cookie', 'skip_download_page=1; dwnld_speed=%d; adv_show=1' % dwnld_speed) - webpage = self._download_webpage( - request, video_id, 'Downloading %s page' % format_id) - video_url = self._html_search_regex( - r'', webpage, '%s video URL' % format_id, fatal=False) - if not video_url: - continue + html5_video_re = r'(?s)<(?:video|audio)[^<]*(?:>.*?]*)?\s+src=["\'](.*?)["\']', + video_url = self._html_search_regex(html5_video_re, webpage, video_id) + mp4_video_url = self._html_search_regex(html5_video_re, mp4_webpage, video_id) + formats = [{ + 'url': video_url, + }] + if mp4_video_url != video_url: formats.append({ - 'url': video_url, - 'format_id': format_id, + 'url': mp4_video_url, }) - webpage = self._download_webpage( - 'http://m.nuvid.com/video/%s' % video_id, video_id, 'Downloading video page') title = self._html_search_regex( [r'', - r'
\s*]*>([^<]+)'], webpage, 'title').strip() + r'
\s*]*>([^<]+)', + r']+class="title_thumb">([^<]+)'], webpage, 'title').strip() thumbnails = [ { 'url': thumb_url, @@ -58,9 +57,8 @@ class NuvidIE(InfoExtractor): ] thumbnail = thumbnails[0]['url'] if thumbnails else None duration = parse_duration(self._html_search_regex( - r'\s*(\d{2}:\d{2})', webpage, 'duration', fatal=False)) - upload_date = unified_strdate(self._html_search_regex( - r'\s*(\d{4}-\d{2}-\d{2})', webpage, 'upload date', fatal=False)) + [r'\s*(\d{2}:\d{2})', + r']+class="view_time">([^<]+)'], webpage, 'duration', fatal=False)) return { 'id': video_id, @@ -68,7 +66,6 @@ class NuvidIE(InfoExtractor): 'thumbnails': thumbnails, 'thumbnail': thumbnail, 'duration': duration, - 'upload_date': upload_date, 'age_limit': 18, 'formats': formats, - } \ No newline at end of file + }