# coding: utf-8 from __future__ import unicode_literals from .common import InfoExtractor from ..compat import compat_str from ..utils import ( ExtractorError, int_or_none, InAdvancePagedList, float_or_none, unescapeHTML, ) class TudouIE(InfoExtractor): IE_NAME = 'tudou' _VALID_URL = r'https?://(?:www\.)?tudou\.com/(?:(?:programs|wlplay)/view|(?:listplay|albumplay)/[\w-]{11})/(?P<id>[\w-]{11})' _TESTS = [{ 'url': 'http://www.tudou.com/listplay/zzdE77v6Mmo/2xN2duXMxmw.html', 'md5': '140a49ed444bd22f93330985d8475fcb', 'info_dict': { 'id': '159448201', 'ext': 'f4v', 'title': '卡马乔国足开大脚长传冲吊集锦', 'thumbnail': 're:^https?://.*\.jpg$', 'timestamp': 1372113489000, 'description': '卡马乔卡家军,开大脚先进战术不完全集锦!', 'duration': 289.04, 'view_count': int, 'filesize': int, } }, { 'url': 'http://www.tudou.com/programs/view/ajX3gyhL0pc/', 'info_dict': { 'id': '117049447', 'ext': 'f4v', 'title': 'La Sylphide-Bolshoi-Ekaterina Krysanova & Vyacheslav Lopatin 2012', 'thumbnail': 're:^https?://.*\.jpg$', 'timestamp': 1349207518000, 'description': 'md5:294612423894260f2dcd5c6c04fe248b', 'duration': 5478.33, 'view_count': int, 'filesize': int, } }] _PLAYER_URL = 'http://js.tudouui.com/bin/lingtong/PortalPlayer_177.swf' # Translated from tudou/tools/TVCHelper.as in PortalPlayer_193.swf # 0001, 0002 and 4001 are not included as they indicate temporary issues TVC_ERRORS = { '0003': 'The video is deleted or does not exist', '1001': 'This video is unavailable due to licensing issues', '1002': 'This video is unavailable as it\'s under review', '1003': 'This video is unavailable as it\'s under review', '3001': 'Password required', '5001': 'This video is available in Mainland China only due to licensing issues', '7001': 'This video is unavailable', '8001': 'This video is unavailable due to licensing issues', } def _url_for_id(self, video_id, quality=None): info_url = 'http://v2.tudou.com/f?id=' + compat_str(video_id) if quality: info_url += '&hd' + quality xml_data = self._download_xml(info_url, video_id, 'Opening the info XML page') error = xml_data.attrib.get('error') if error is not None: raise ExtractorError('Tudou said: %s' % error, expected=True) final_url = xml_data.text return final_url def _real_extract(self, url): video_id = self._match_id(url) item_data = self._download_json( 'http://www.tudou.com/tvp/getItemInfo.action?ic=%s' % video_id, video_id) youku_vcode = item_data.get('vcode') if youku_vcode: return self.url_result('youku:' + youku_vcode, ie='Youku') if not item_data.get('itemSegs'): tvc_code = item_data.get('tvcCode') if tvc_code: err_msg = self.TVC_ERRORS.get(tvc_code) if err_msg: raise ExtractorError('Tudou said: %s' % err_msg, expected=True) raise ExtractorError('Unexpected error %s returned from Tudou' % tvc_code) raise ExtractorError('Unxpected error returned from Tudou') title = unescapeHTML(item_data['kw']) description = item_data.get('desc') thumbnail_url = item_data.get('pic') view_count = int_or_none(item_data.get('playTimes')) timestamp = int_or_none(item_data.get('pt')) segments = self._parse_json(item_data['itemSegs'], video_id) # It looks like the keys are the arguments that have to be passed as # the hd field in the request url, we pick the higher # Also, filter non-number qualities (see issue #3643). quality = sorted(filter(lambda k: k.isdigit(), segments.keys()), key=lambda k: int(k))[-1] parts = segments[quality] len_parts = len(parts) if len_parts > 1: self.to_screen('%s: found %s parts' % (video_id, len_parts)) def part_func(partnum): part = parts[partnum] part_id = part['k'] final_url = self._url_for_id(part_id, quality) ext = (final_url.split('?')[0]).split('.')[-1] return [{ 'id': '%s' % part_id, 'url': final_url, 'ext': ext, 'title': title, 'thumbnail': thumbnail_url, 'description': description, 'view_count': view_count, 'timestamp': timestamp, 'duration': float_or_none(part.get('seconds'), 1000), 'filesize': int_or_none(part.get('size')), 'http_headers': { 'Referer': self._PLAYER_URL, }, }] entries = InAdvancePagedList(part_func, len_parts, 1) return { '_type': 'multi_video', 'entries': entries, 'id': video_id, 'title': title, } class TudouPlaylistIE(InfoExtractor): IE_NAME = 'tudou:playlist' _VALID_URL = r'https?://(?:www\.)?tudou\.com/listplay/(?P<id>[\w-]{11})\.html' _TESTS = [{ 'url': 'http://www.tudou.com/listplay/zzdE77v6Mmo.html', 'info_dict': { 'id': 'zzdE77v6Mmo', }, 'playlist_mincount': 209, }] def _real_extract(self, url): playlist_id = self._match_id(url) playlist_data = self._download_json( 'http://www.tudou.com/tvp/plist.action?lcode=%s' % playlist_id, playlist_id) entries = [self.url_result( 'http://www.tudou.com/programs/view/%s' % item['icode'], 'Tudou', item['icode'], item['kw']) for item in playlist_data['items']] return self.playlist_result(entries, playlist_id) class TudouAlbumIE(InfoExtractor): IE_NAME = 'tudou:album' _VALID_URL = r'https?://(?:www\.)?tudou\.com/album(?:cover|play)/(?P<id>[\w-]{11})' _TESTS = [{ 'url': 'http://www.tudou.com/albumplay/v5qckFJvNJg.html', 'info_dict': { 'id': 'v5qckFJvNJg', }, 'playlist_mincount': 45, }] def _real_extract(self, url): album_id = self._match_id(url) album_data = self._download_json( 'http://www.tudou.com/tvp/alist.action?acode=%s' % album_id, album_id) entries = [self.url_result( 'http://www.tudou.com/programs/view/%s' % item['icode'], 'Tudou', item['icode'], item['kw']) for item in album_data['items']] return self.playlist_result(entries, album_id)