1
0
mirror of https://github.com/l1ving/youtube-dl synced 2024-11-26 02:13:22 +08:00
youtube-dl/youtube_dl/extractor/theplatform.py
2013-12-26 21:18:18 +01:00

87 lines
2.9 KiB
Python

import re
import json
from .common import InfoExtractor
from ..utils import (
ExtractorError,
xpath_with_ns,
)
_x = lambda p: xpath_with_ns(p, {'smil': 'http://www.w3.org/2005/SMIL21/Language'})
class ThePlatformIE(InfoExtractor):
_VALID_URL = r'(?:https?://link\.theplatform\.com/s/[^/]+/|theplatform:)(?P<id>[^/\?]+)'
_TEST = {
# from http://www.metacafe.com/watch/cb-e9I_cZgTgIPd/blackberrys_big_bold_z30/
u'url': u'http://link.theplatform.com/s/dJ5BDC/e9I_cZgTgIPd/meta.smil?format=smil&Tracking=true&mbr=true',
u'info_dict': {
u'id': u'e9I_cZgTgIPd',
u'ext': u'flv',
u'title': u'Blackberry\'s big, bold Z30',
u'description': u'The Z30 is Blackberry\'s biggest, baddest mobile messaging device yet.',
u'duration': 247,
},
u'params': {
# rtmp download
u'skip_download': True,
},
}
def _get_info(self, video_id):
smil_url = ('http://link.theplatform.com/s/dJ5BDC/{0}/meta.smil?'
'format=smil&mbr=true'.format(video_id))
meta = self._download_xml(smil_url, video_id)
try:
error_msg = next(
n.attrib['abstract']
for n in meta.findall(_x('.//smil:ref'))
if n.attrib.get('title') == u'Geographic Restriction')
except StopIteration:
pass
else:
raise ExtractorError(error_msg, expected=True)
info_url = 'http://link.theplatform.com/s/dJ5BDC/{0}?format=preview'.format(video_id)
info_json = self._download_webpage(info_url, video_id)
info = json.loads(info_json)
head = meta.find(_x('smil:head'))
body = meta.find(_x('smil:body'))
base_url = head.find(_x('smil:meta')).attrib['base']
switch = body.find(_x('smil:switch'))
formats = []
for f in switch.findall(_x('smil:video')):
attr = f.attrib
width = int(attr['width'])
height = int(attr['height'])
vbr = int(attr['system-bitrate']) // 1000
format_id = '%dx%d_%dk' % (width, height, vbr)
formats.append({
'format_id': format_id,
'url': base_url,
'play_path': 'mp4:' + attr['src'],
'ext': 'flv',
'width': width,
'height': height,
'vbr': vbr,
})
self._sort_formats(formats)
return {
'id': video_id,
'title': info['title'],
'formats': formats,
'description': info['description'],
'thumbnail': info['defaultThumbnailUrl'],
'duration': info['duration']//1000,
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
return self._get_info(video_id)