_ Git - youtube-dl/blob - youtube_dl/extractor/xuite.py

   1 # -*- coding: utf-8 -*-
   2 from __future__ import unicode_literals
   3
   4 import base64
   5 from .common import InfoExtractor
   6 from ..compat import (
   7     compat_urlparse,
   8     compat_urllib_parse_unquote,
   9     compat_parse_qs
  10 )
  11 from ..utils import (
  12     ExtractorError,
  13     parse_iso8601,
  14     parse_duration
  15 )
  16
  17 # ref: http://stackoverflow.com/questions/475074/regex-to-parse-or-validate-base64-data
  18 REGEX_BASE64 = r'(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?'
  19
  20 CST_ZONE = +8  # China Standard Time
  21
  22
  23 class XuiteIE(InfoExtractor):
  24     _VALID_URL = r'http://vlog.xuite.net/play/(?P<id>%s)(/.*)?' % REGEX_BASE64
  25     _TESTS = [{
  26         # Audio
  27         'url': 'http://vlog.xuite.net/play/RGkzc1ZULTM4NjA5MTQuZmx2',
  28         'md5': '63a42c705772aa53fd4c1a0027f86adf',
  29         'info_dict': {
  30             'id': 'RGkzc1ZULTM4NjA5MTQuZmx2',
  31             'ext': 'mp3',
  32             'upload_date': '20110902',
  33             'uploader_id': '15973816',
  34             'uploader': '阿能',
  35             'timestamp': 1314932940,
  36             'title': '孤單南半球-歐德陽',
  37             'thumbnail': 're:^https?://.*\.jpg$',
  38             'categories': ['個人短片'],
  39             'duration': 247.246
  40         }
  41     }, {
  42         # Audio with alternative form of url
  43         'url': 'http://vlog.xuite.net/play/S1dDUjdyLTMyOTc3NjcuZmx2/%E5%AD%AB%E7%87%95%E5%A7%BF-%E7%9C%BC%E6%B7%9A%E6%88%90%E8%A9%A9',
  44         'md5': 'c91645f324de53d82ebb80930e1a73d2',
  45         'info_dict': {
  46             'id': 'S1dDUjdyLTMyOTc3NjcuZmx2',
  47             'ext': 'mp3',
  48             'upload_date': '20101226',
  49             'uploader_id': '10102699',
  50             'uploader': '蠍',
  51             'timestamp': 1293367080,
  52             'title': '孫燕姿-眼淚成詩',
  53             'thumbnail': 're:^https?://.*\.jpg$',
  54             'categories': ['個人短片'],
  55             'duration': 223.19
  56         }
  57     }, {
  58         # Video with only one format
  59         'url': 'http://vlog.xuite.net/play/TkRZNjhULTM0NDE2MjkuZmx2',
  60         'md5': 'c45737fc8ac5dc8ac2f92ecbcecf505e',
  61         'info_dict': {
  62             'id': 'TkRZNjhULTM0NDE2MjkuZmx2',
  63             'ext': 'mp4',
  64             'upload_date': '20110306',
  65             'uploader_id': '10400126',
  66             'uploader': 'Valen',
  67             'timestamp': 1299383640,
  68             'title': '孫燕姿 - 眼淚成詩',
  69             'thumbnail': 're:^https?://.*\.jpg$',
  70             'categories': ['影視娛樂'],
  71             'duration': 217.399
  72         }
  73     }, {
  74         # Video with two formats
  75         'url': 'http://vlog.xuite.net/play/bWo1N1pLLTIxMzAxMTcwLmZsdg==',
  76         'md5': '1166e0f461efe55b62e26a2d2a68e6de',
  77         'info_dict': {
  78             'id': 'bWo1N1pLLTIxMzAxMTcwLmZsdg==',
  79             'ext': 'mp4',
  80             'upload_date': '20150117',
  81             'uploader_id': '242127761',
  82             'uploader': '我只是想認真點',
  83             'timestamp': 1421481240,
  84             'title': '暗殺教室 02',
  85             'thumbnail': 're:^https?://.*\.jpg$',
  86             'categories': ['電玩動漫'],
  87             'duration': 1384.907
  88         }
  89     }]
  90
  91     def _flv_config(self, media_id):
  92         base64_media_id = base64.b64encode(media_id.encode('utf-8')).decode('utf-8')
  93         flv_config_url = 'http://vlog.xuite.net/flash/player?media=' + base64_media_id
  94         flv_config = self._download_xml(flv_config_url, 'flv config')
  95
  96         prop_dict = {}
  97         for prop in flv_config.findall('./property'):
  98             prop_id = base64.b64decode(prop.attrib['id']).decode('utf-8')
  99
 100             if not prop.text:
 101                 continue  # CDATA may be empty in flv config
 102
 103             encoded_content = base64.b64decode(prop.text).decode('utf-8')
 104             prop_dict[prop_id] = compat_urllib_parse_unquote(encoded_content)
 105
 106         return prop_dict
 107
 108     def _type_string(self, media_url):
 109         query_string = compat_urlparse.urlparse(media_url).query
 110         type_string = compat_parse_qs(query_string)['q'][0]
 111         return type_string
 112
 113     def _guess_ext(self, media_url):
 114         type_string = self._type_string(media_url)
 115         if type_string == 'mp3':
 116             return 'mp3'
 117         elif type_string == '360' or type_string == '720':
 118             return 'mp4'
 119         else:
 120             raise ExtractorError('Unknown type string %s' % type_string)
 121
 122     def _real_extract(self, url):
 123         video_id = self._match_id(url)
 124
 125         page = self._download_webpage(url, video_id)
 126         media_id = self._html_search_regex(r'data-mediaid="(\d+)"', page, 'media id')
 127         flv_config = self._flv_config(media_id)
 128
 129         timestamp_local = parse_iso8601(flv_config['publish_datetime'], ' ')
 130         timestamp_gmt = timestamp_local - CST_ZONE * 3600
 131
 132         ret_attrs = {
 133             'id': video_id,
 134             'title': flv_config['title'],
 135             'thumbnail': flv_config['thumb'],
 136             'uploader': flv_config['author_name'],
 137             'timestamp': timestamp_gmt,
 138             'uploader_id': flv_config['author_id'],
 139             'duration': parse_duration(flv_config['duration']),
 140             'categories': [flv_config['category']]
 141         }
 142
 143         if 'hq_src' in flv_config:
 144             urls = [flv_config['src'], flv_config['hq_src']]
 145
 146             ret_attrs['formats'] = []
 147
 148             for url in urls:
 149                 ret_attrs['formats'].append({
 150                     'url': url,
 151                     'ext': self._guess_ext(url),
 152                     'format_id': self._type_string(url),
 153                     'height': int(self._type_string(url))
 154                 })
 155         else:
 156             ret_attrs['url'] = flv_config['src']
 157             ret_attrs['ext'] = self._guess_ext(flv_config['src'])
 158
 159         return ret_attrs