[videobam] Add support for videobam.com (Closes #2411)

pull/8/head
Sergey M. 11 years ago
parent f38da66731
commit 3e123c1e28

@ -241,6 +241,7 @@ from .vesti import VestiIE
from .vevo import VevoIE
from .vice import ViceIE
from .viddler import ViddlerIE
from .videobam import VideoBamIE
from .videodetective import VideoDetectiveIE
from .videofyme import VideofyMeIE
from .videopremium import VideoPremiumIE

@ -0,0 +1,79 @@
from __future__ import unicode_literals
import re
import json
from .common import InfoExtractor
from ..utils import int_or_none
class VideoBamIE(InfoExtractor):
_VALID_URL = r'http://(?:www\.)?videobam\.com/(?:videos/download/)?(?P<id>[a-zA-Z]+)'
_TESTS = [
{
'url': 'http://videobam.com/OiJQM',
'md5': 'db471f27763a531f10416a0c58b5a1e0',
'info_dict': {
'id': 'OiJQM',
'ext': 'mp4',
'title': 'Is Alcohol Worse Than Ecstasy?',
'description': 'md5:d25b96151515c91debc42bfbb3eb2683',
'uploader': 'frihetsvinge',
},
},
{
'url': 'http://videobam.com/pqLvq',
'md5': 'd9a565b5379a99126ef94e1d7f9a383e',
'note': 'HD video',
'info_dict': {
'id': 'pqLvq',
'ext': 'mp4',
}
},
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
page = self._download_webpage('http://videobam.com/%s' % video_id, video_id, 'Downloading page')
formats = []
for preference, format_id in enumerate(['low', 'high']):
mobj = re.search(r"%s: '(?P<url>[^']+)'" % format_id, page)
if not mobj:
continue
formats.append({
'url': mobj.group('url'),
'ext': 'mp4',
'format_id': format_id,
'preference': preference,
})
if not formats:
player_config = json.loads(self._html_search_regex(r'var player_config = ({.+?});', page, 'player config'))
formats = [{
'url': item['url'],
'ext': 'mp4',
} for item in player_config['playlist'] if 'autoPlay' in item]
self._sort_formats(formats)
title = self._og_search_title(page, default='VideoBam', fatal=False)
description = self._og_search_description(page, default=None)
thumbnail = self._og_search_thumbnail(page)
uploader = self._html_search_regex(r'Upload by ([^<]+)</a>', page, 'uploader', fatal=False, default=None)
view_count = int_or_none(
self._html_search_regex(r'<strong>Views:</strong> (\d+) ', page, 'view count', fatal=False))
return {
'id': video_id,
'title': title,
'description': description,
'thumbnail': thumbnail,
'uploader': uploader,
'view_count': view_count,
'formats': formats,
}
Loading…
Cancel
Save