1
0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-15 03:27:46 +01:00
youtube-dl/youtube_dl/extractor/space.py
Jaime Marquínez Ferrándiz 065472936a Add an extractor for space.com (fixes #1718)
It uses Brightcove, but requires some special process for getting a url with the playerKey field in some videos
2013-11-06 17:37:39 +01:00

36 lines
1.4 KiB
Python

import re
from .common import InfoExtractor
from .brightcove import BrightcoveIE
from ..utils import RegexNotFoundError, ExtractorError
class SpaceIE(InfoExtractor):
_VALID_URL = r'https?://www\.space\.com/\d+-(?P<title>[^/\.\?]*?)-video.html'
_TEST = {
u'add_ie': ['Brightcove'],
u'url': u'http://www.space.com/23373-huge-martian-landforms-detail-revealed-by-european-probe-video.html',
u'info_dict': {
u'id': u'2780937028001',
u'ext': u'mp4',
u'title': u'Huge Martian Landforms\' Detail Revealed By European Probe | Video',
u'description': u'md5:db81cf7f3122f95ed234b631a6ea1e61',
u'uploader': u'TechMedia Networks',
},
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
title = mobj.group('title')
webpage = self._download_webpage(url, title)
try:
# Some videos require the playerKey field, which isn't define in
# the BrightcoveExperience object
brightcove_url = self._og_search_video_url(webpage)
except RegexNotFoundError:
# Other videos works fine with the info from the object
brightcove_url = BrightcoveIE._extract_brightcove_url(webpage)
if brightcove_url is None:
raise ExtractorError(u'The webpage does not contain a video', expected=True)
return self.url_result(brightcove_url, BrightcoveIE.ie_key())