1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-24 11:11:54 +00:00
youtube-dl/youtube_dl/extractor/discovery.py
2016-05-02 07:07:35 +06:00

97 lines
3.5 KiB
Python

from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import (
parse_duration,
parse_iso8601,
)
from ..compat import compat_str
class DiscoveryIE(InfoExtractor):
_VALID_URL = r'''(?x)https?://(?:www\.)?(?:
discovery|
investigationdiscovery|
discoverylife|
animalplanet|
ahctv|
destinationamerica|
sciencechannel|
tlc|
velocity
)\.com/(?:[^/]+/)*(?P<id>[^./?#]+)'''
_TESTS = [{
'url': 'http://www.discovery.com/tv-shows/mythbusters/videos/mission-impossible-outtakes.htm',
'info_dict': {
'id': '20769',
'ext': 'mp4',
'title': 'Mission Impossible Outtakes',
'description': ('Watch Jamie Hyneman and Adam Savage practice being'
' each other -- to the point of confusing Jamie\'s dog -- and '
'don\'t miss Adam moon-walking as Jamie ... behind Jamie\'s'
' back.'),
'duration': 156,
'timestamp': 1302032462,
'upload_date': '20110405',
'uploader_id': '103207',
},
'params': {
'skip_download': True, # requires ffmpeg
}
}, {
'url': 'http://www.discovery.com/tv-shows/mythbusters/videos/mythbusters-the-simpsons',
'info_dict': {
'id': 'mythbusters-the-simpsons',
'title': 'MythBusters: The Simpsons',
},
'playlist_mincount': 10,
}, {
'url': 'http://www.animalplanet.com/longfin-eels-maneaters/',
'info_dict': {
'id': '78326',
'ext': 'mp4',
'title': 'Longfin Eels: Maneaters?',
'description': 'Jeremy Wade tests whether or not New Zealand\'s longfin eels are man-eaters by covering himself in fish guts and getting in the water with them.',
'upload_date': '20140725',
'timestamp': 1406246400,
'duration': 116,
'uploader_id': '103207',
},
'params': {
'skip_download': True, # requires ffmpeg
}
}]
def _real_extract(self, url):
display_id = self._match_id(url)
info = self._download_json(url + '?flat=1', display_id)
video_title = info.get('playlist_title') or info.get('video_title')
entries = []
for idx, video_info in enumerate(info['playlist']):
subtitles = {}
caption_url = video_info.get('captionsUrl')
if caption_url:
subtitles = {
'en': [{
'url': caption_url,
}]
}
entries.append({
'_type': 'url_transparent',
'url': 'http://players.brightcove.net/103207/default_default/index.html?videoId=ref:%s' % video_info['referenceId'],
'id': compat_str(video_info['id']),
'title': video_info['title'],
'description': video_info.get('description'),
'duration': parse_duration(video_info.get('video_length')),
'webpage_url': video_info.get('href') or video_info.get('url'),
'thumbnail': video_info.get('thumbnailURL'),
'alt_title': video_info.get('secondary_title'),
'timestamp': parse_iso8601(video_info.get('publishedDate')),
'subtitles': subtitles,
})
return self.playlist_result(entries, display_id, video_title)