X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/9dc487f48b50767cf540fa36c3de2c386fd74c04..f8df414a4abcde0ddd39325dac26ca071d2d15c6:/youtube_dl/extractor/leeco.py?ds=sidebyside diff --git a/youtube_dl/extractor/leeco.py b/youtube_dl/extractor/leeco.py index 63f581c..0a07c13 100644 --- a/youtube_dl/extractor/leeco.py +++ b/youtube_dl/extractor/leeco.py @@ -20,16 +20,16 @@ from ..utils import ( int_or_none, orderedSet, parse_iso8601, - sanitized_Request, str_or_none, url_basename, + urshift, ) class LeIE(InfoExtractor): IE_DESC = '乐视网' - _VALID_URL = r'https?://(?:www\.le\.com/ptv/vplay|sports\.le\.com/video)/(?P\d+)\.html' - + _VALID_URL = r'https?://(?:www\.le\.com/ptv/vplay|(?:sports\.le|(?:www\.)?lesports)\.com/(?:match|video))/(?P\d+)\.html' + _GEO_COUNTRIES = ['CN'] _URL_TEMPLATE = 'http://www.le.com/ptv/vplay/%s.html' _TESTS = [{ @@ -50,7 +50,7 @@ class LeIE(InfoExtractor): 'id': '1415246', 'ext': 'mp4', 'title': '美人天下01', - 'description': 'md5:f88573d9d7225ada1359eaf0dbf8bcda', + 'description': 'md5:28942e650e82ed4fcc8e4de919ee854d', }, 'params': { 'hls_prefer_native': True, @@ -68,30 +68,28 @@ class LeIE(InfoExtractor): 'params': { 'hls_prefer_native': True, }, - 'skip': 'Only available in China', }, { 'url': 'http://sports.le.com/video/25737697.html', 'only_matching': True, + }, { + 'url': 'http://www.lesports.com/match/1023203003.html', + 'only_matching': True, + }, { + 'url': 'http://sports.le.com/match/1023203003.html', + 'only_matching': True, }] - @staticmethod - def urshift(val, n): - return val >> n if val >= 0 else (val + 0x100000000) >> n - - # ror() and calc_time_key() are reversed from a embedded swf file in KLetvPlayer.swf + # ror() and calc_time_key() are reversed from a embedded swf file in LetvPlayer.swf def ror(self, param1, param2): _loc3_ = 0 while _loc3_ < param2: - param1 = self.urshift(param1, 1) + ((param1 & 1) << 31) + param1 = urshift(param1, 1) + ((param1 & 1) << 31) _loc3_ += 1 return param1 def calc_time_key(self, param1): - _loc2_ = 773625421 - _loc3_ = self.ror(param1, _loc2_ % 13) - _loc3_ = _loc3_ ^ _loc2_ - _loc3_ = self.ror(_loc3_, _loc2_ % 17) - return _loc3_ + _loc2_ = 185025305 + return self.ror(param1, _loc2_ % 17) ^ _loc2_ # see M3U8Encryption class in KLetvPlayer.swf @staticmethod @@ -113,75 +111,81 @@ class LeIE(InfoExtractor): return bytes(_loc7_) - def _real_extract(self, url): - media_id = self._match_id(url) - page = self._download_webpage(url, media_id) - params = { - 'id': media_id, - 'platid': 1, - 'splatid': 101, - 'format': 1, - 'tkey': self.calc_time_key(int(time.time())), - 'domain': 'www.le.com' - } - play_json_req = sanitized_Request( - 'http://api.le.com/mms/out/video/playJson?' + compat_urllib_parse_urlencode(params) - ) - cn_verification_proxy = self._downloader.params.get('cn_verification_proxy') - if cn_verification_proxy: - play_json_req.add_header('Ytdl-request-proxy', cn_verification_proxy) - - play_json = self._download_json( - play_json_req, - media_id, 'Downloading playJson data') - + def _check_errors(self, play_json): # Check for errors - playstatus = play_json['playstatus'] + playstatus = play_json['msgs']['playstatus'] if playstatus['status'] == 0: flag = playstatus['flag'] if flag == 1: - msg = 'Country %s auth error' % playstatus['country'] + self.raise_geo_restricted() else: - msg = 'Generic error. flag = %d' % flag - raise ExtractorError(msg, expected=True) + raise ExtractorError('Generic error. flag = %d' % flag, expected=True) - playurl = play_json['playurl'] - - formats = ['350', '1000', '1300', '720p', '1080p'] - dispatch = playurl['dispatch'] + def _real_extract(self, url): + media_id = self._match_id(url) + page = self._download_webpage(url, media_id) - urls = [] - for format_id in formats: - if format_id in dispatch: - media_url = playurl['domain'][0] + dispatch[format_id][0] - media_url += '&' + compat_urllib_parse_urlencode({ + play_json_flash = self._download_json( + 'http://player-pc.le.com/mms/out/video/playJson', + media_id, 'Downloading flash playJson data', query={ + 'id': media_id, + 'platid': 1, + 'splatid': 101, + 'format': 1, + 'source': 1000, + 'tkey': self.calc_time_key(int(time.time())), + 'domain': 'www.le.com', + 'region': 'cn', + }, + headers=self.geo_verification_headers()) + self._check_errors(play_json_flash) + + def get_flash_urls(media_url, format_id): + nodes_data = self._download_json( + media_url, media_id, + 'Download JSON metadata for format %s' % format_id, + query={ 'm3v': 1, 'format': 1, 'expect': 3, - 'rateid': format_id, + 'tss': 'ios', }) - nodes_data = self._download_json( - media_url, media_id, - 'Download JSON metadata for format %s' % format_id) + req = self._request_webpage( + nodes_data['nodelist'][0]['location'], media_id, + note='Downloading m3u8 information for format %s' % format_id) - req = self._request_webpage( - nodes_data['nodelist'][0]['location'], media_id, - note='Downloading m3u8 information for format %s' % format_id) + m3u8_data = self.decrypt_m3u8(req.read()) - m3u8_data = self.decrypt_m3u8(req.read()) + return { + 'hls': encode_data_uri(m3u8_data, 'application/vnd.apple.mpegurl'), + } - url_info_dict = { - 'url': encode_data_uri(m3u8_data, 'application/vnd.apple.mpegurl'), - 'ext': determine_ext(dispatch[format_id][1]), - 'format_id': format_id, - 'protocol': 'm3u8', + extracted_formats = [] + formats = [] + playurl = play_json_flash['msgs']['playurl'] + play_domain = playurl['domain'][0] + + for format_id, format_data in playurl.get('dispatch', []).items(): + if format_id in extracted_formats: + continue + extracted_formats.append(format_id) + + media_url = play_domain + format_data[0] + for protocol, format_url in get_flash_urls(media_url, format_id).items(): + f = { + 'url': format_url, + 'ext': determine_ext(format_data[1]), + 'format_id': '%s-%s' % (protocol, format_id), + 'protocol': 'm3u8_native' if protocol == 'hls' else 'http', + 'quality': int_or_none(format_id), } if format_id[-1:] == 'p': - url_info_dict['height'] = int_or_none(format_id[:-1]) + f['height'] = int_or_none(format_id[:-1]) - urls.append(url_info_dict) + formats.append(f) + self._sort_formats(formats, ('height', 'quality', 'format_id')) publish_time = parse_iso8601(self._html_search_regex( r'发布时间 ([^<>]+) ', page, 'publish time', default=None), @@ -190,7 +194,7 @@ class LeIE(InfoExtractor): return { 'id': media_id, - 'formats': urls, + 'formats': formats, 'title': playurl['title'], 'thumbnail': playurl['pic'], 'description': description, @@ -344,8 +348,8 @@ class LetvCloudIE(InfoExtractor): return formats def _real_extract(self, url): - uu_mobj = re.search('uu=([\w]+)', url) - vu_mobj = re.search('vu=([\w]+)', url) + uu_mobj = re.search(r'uu=([\w]+)', url) + vu_mobj = re.search(r'vu=([\w]+)', url) if not uu_mobj or not vu_mobj: raise ExtractorError('Invalid URL: %s' % url, expected=True)