|
|
from __future__ import unicode_literals
import re import json
from .common import InfoExtractor from ..utils import find_xpath_attr, compat_str
class NBCIE(InfoExtractor): _VALID_URL = r'http://www\.nbc\.com/[^/]+/video/[^/]+/(?P<id>n?\d+)'
_TEST = { 'url': 'http://www.nbc.com/chicago-fire/video/i-am-a-firefighter/2734188', 'md5': '54d0fbc33e0b853a65d7b4de5c06d64e', 'info_dict': { 'id': 'u1RInQZRN7QJ', 'ext': 'flv', 'title': 'I Am a Firefighter', 'description': 'An emergency puts Dawson\'sf irefighter skills to the ultimate test in this four-part digital series.', }, }
def _real_extract(self, url): mobj = re.match(self._VALID_URL, url) video_id = mobj.group('id') webpage = self._download_webpage(url, video_id) theplatform_url = self._search_regex('class="video-player video-player-full" data-mpx-url="(.*?)"', webpage, 'theplatform url') if theplatform_url.startswith('//'): theplatform_url = 'http:' + theplatform_url return self.url_result(theplatform_url)
class NBCNewsIE(InfoExtractor): _VALID_URL = r'''(?x)https?://www\.nbcnews\.com/
((video/.+?/(?P<id>\d+))| (feature/[^/]+/(?P<title>.+))) '''
_TESTS = [ { 'url': 'http://www.nbcnews.com/video/nbc-news/52753292', 'md5': '47abaac93c6eaf9ad37ee6c4463a5179', 'info_dict': { 'id': '52753292', 'ext': 'flv', 'title': 'Crew emerges after four-month Mars food study', 'description': 'md5:24e632ffac72b35f8b67a12d1b6ddfc1', }, }, { 'url': 'http://www.nbcnews.com/feature/edward-snowden-interview/how-twitter-reacted-snowden-interview-n117236', 'md5': 'b2421750c9f260783721d898f4c42063', 'info_dict': { 'id': 'I1wpAI_zmhsQ', 'ext': 'flv', 'title': 'How Twitter Reacted To The Snowden Interview', 'description': 'md5:65a0bd5d76fe114f3c2727aa3a81fe64', }, 'add_ie': ['ThePlatform'], }, ]
def _real_extract(self, url): mobj = re.match(self._VALID_URL, url) video_id = mobj.group('id') if video_id is not None: all_info = self._download_xml('http://www.nbcnews.com/id/%s/displaymode/1219' % video_id, video_id) info = all_info.find('video')
return { 'id': video_id, 'title': info.find('headline').text, 'ext': 'flv', 'url': find_xpath_attr(info, 'media', 'type', 'flashVideo').text, 'description': compat_str(info.find('caption').text), 'thumbnail': find_xpath_attr(info, 'media', 'type', 'thumbnail').text, } else: # "feature" pages use theplatform.com title = mobj.group('title') webpage = self._download_webpage(url, title) bootstrap_json = self._search_regex( r'var bootstrapJson = ({.+})\s*$', webpage, 'bootstrap json', flags=re.MULTILINE) bootstrap = json.loads(bootstrap_json) info = bootstrap['results'][0]['video'] playlist_url = info['fallbackPlaylistUrl'] + '?form=MPXNBCNewsAPI' mpxid = info['mpxId'] all_videos = self._download_json(playlist_url, title)['videos'] # The response contains additional videos info = next(v for v in all_videos if v['mpxId'] == mpxid)
return { '_type': 'url', # We get the best quality video 'url': info['videoAssets'][-1]['publicUrl'], 'ie_key': 'ThePlatform', }
|