[bilibili] Fix extraction (closes #10375)

Thanks @gdkchan for the algorithm

2 files changed, 37 insertions, 64 deletions

diff --git a/ChangeLog b/ChangeLog
index 5d7a052a5..e055976c5 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,6 +1,7 @@
 version <unreleased>
 
 Extractors
+* [bilibili] Fix extraction (#10375)
 * [openload] Fix extraction (#10408)
 
 
diff --git a/youtube_dl/extractor/bilibili.py b/youtube_dl/extractor/bilibili.py
index d87c38a02..a332fbb69 100644
--- a/youtube_dl/extractor/bilibili.py
+++ b/youtube_dl/extractor/bilibili.py
@@ -1,35 +1,26 @@
 # coding: utf-8
 from __future__ import unicode_literals
 
-import calendar
-import datetime
+import hashlib
 import re
 
 from .common import InfoExtractor
-from ..compat import (
-    compat_etree_fromstring,
-    compat_str,
-    compat_parse_qs,
-    compat_xml_parse_error,
-)
+from ..compat import compat_parse_qs
 from ..utils import (
-    ExtractorError,
     int_or_none,
     float_or_none,
-    xpath_text,
+    unified_timestamp,
 )
 
 
 class BiliBiliIE(InfoExtractor):
-    _WORKING = False
-
     _VALID_URL = r'https?://www\.bilibili\.(?:tv|com)/video/av(?P<id>\d+)'
 
     _TESTS = [{
         'url': 'http://www.bilibili.tv/video/av1074402/',
         'md5': '9fa226fe2b8a9a4d5a69b4c6a183417e',
         'info_dict': {
-            'id': '1554319',
+            'id': '1074402',
             'ext': 'mp4',
             'title': '【金坷垃】金泡沫',
             'description': 'md5:ce18c2a2d2193f0df2917d270f2e5923',
@@ -43,24 +34,28 @@ class BiliBiliIE(InfoExtractor):
     }, {
         'url': 'http://www.bilibili.com/video/av1041170/',
         'info_dict': {
-            'id': '1507019',
+            'id': '1041170',
             'ext': 'mp4',
             'title': '【BD1080P】刀语【诸神&异域】',
             'description': '这是个神奇的故事~每个人不留弹幕不给走哦~切利哦！~',
+            'duration': 3382.259,
             'timestamp': 1396530060,
             'upload_date': '20140403',
+            'thumbnail': 're:^https?://.+\.jpg',
             'uploader': '枫叶逝去',
             'uploader_id': '520116',
         },
     }, {
         'url': 'http://www.bilibili.com/video/av4808130/',
         'info_dict': {
-            'id': '7802182',
+            'id': '4808130',
             'ext': 'mp4',
             'title': '【长篇】哆啦A梦443【钉铛】',
             'description': '(2016.05.27)来组合客人的脸吧&amp;amp;寻母六千里锭 抱歉，又轮到周日上班现在才到家 封面www.pixiv.net/member_illust.php?mode=medium&amp;amp;illust_id=56912929',
+            'duration': 1493.995,
             'timestamp': 1464564180,
             'upload_date': '20160529',
+            'thumbnail': 're:^https?://.+\.jpg',
             'uploader': '喜欢拉面',
             'uploader_id': '151066',
         },
@@ -68,12 +63,14 @@ class BiliBiliIE(InfoExtractor):
         # Missing upload time
         'url': 'http://www.bilibili.com/video/av1867637/',
         'info_dict': {
-            'id': '2880301',
+            'id': '1867637',
             'ext': 'mp4',
             'title': '【HDTV】【喜剧】岳父岳母真难当 （2014）【法国票房冠军】',
             'description': '一个信奉天主教的法国旧式传统资产阶级家庭中有四个女儿。三个女儿却分别找了阿拉伯、犹太、中国丈夫，老夫老妻唯独期盼剩下未嫁的小女儿能找一个信奉天主教的法国白人，结果没想到小女儿找了一位非裔黑人……【这次应该不会跳帧了】',
+            'duration': 5760.0,
             'uploader': '黑夜为猫',
             'uploader_id': '610729',
+            'thumbnail': 're:^https?://.+\.jpg',
         },
         'params': {
             # Just to test metadata extraction
@@ -82,86 +79,61 @@ class BiliBiliIE(InfoExtractor):
         'expected_warnings': ['upload time'],
     }]
 
-    # BiliBili blocks keys from time to time. The current key is extracted from
-    # the Android client
-    # TODO: find the sign algorithm used in the flash player
-    _APP_KEY = '86385cdc024c0f6c'
+    _APP_KEY = '6f90a59ac58a4123'
+    _BILIBILI_KEY = '0bfd84cc3940035173f35e6777508326'
 
     def _real_extract(self, url):
-        mobj = re.match(self._VALID_URL, url)
-        video_id = mobj.group('id')
+        video_id = self._match_id(url)
 
         webpage = self._download_webpage(url, video_id)
 
-        params = compat_parse_qs(self._search_regex(
+        cid = compat_parse_qs(self._search_regex(
             [r'EmbedPlayer\([^)]+,\s*"([^"]+)"\)',
              r'<iframe[^>]+src="https://secure\.bilibili\.com/secure,([^"]+)"'],
-            webpage, 'player parameters'))
-        cid = params['cid'][0]
-
-        info_xml_str = self._download_webpage(
-            'http://interface.bilibili.com/v_cdn_play',
-            cid, query={'appkey': self._APP_KEY, 'cid': cid},
-            note='Downloading video info page')
-
-        err_msg = None
-        durls = None
-        info_xml = None
-        try:
-            info_xml = compat_etree_fromstring(info_xml_str.encode('utf-8'))
-        except compat_xml_parse_error:
-            info_json = self._parse_json(info_xml_str, video_id, fatal=False)
-            err_msg = (info_json or {}).get('error_text')
-        else:
-            err_msg = xpath_text(info_xml, './message')
+            webpage, 'player parameters'))['cid'][0]
+
+        payload = 'appkey=%s&cid=%s&otype=json&quality=2&type=mp4' % (self._APP_KEY, cid)
+        sign = hashlib.md5((payload + self._BILIBILI_KEY).encode('utf-8')).hexdigest()
 
-        if info_xml is not None:
-            durls = info_xml.findall('./durl')
-        if not durls:
-            if err_msg:
-                raise ExtractorError('%s said: %s' % (self.IE_NAME, err_msg), expected=True)
-            else:
-                raise ExtractorError('No videos found!')
+        video_info = self._download_json(
+            'http://interface.bilibili.com/playurl?%s&sign=%s' % (payload, sign),
+            video_id, note='Downloading video info page')
 
         entries = []
 
-        for durl in durls:
-            size = xpath_text(durl, ['./filesize', './size'])
+        for idx, durl in enumerate(video_info['durl']):
             formats = [{
-                'url': durl.find('./url').text,
-                'filesize': int_or_none(size),
+                'url': durl['url'],
+                'filesize': int_or_none(durl['size']),
             }]
-            for backup_url in durl.findall('./backup_url/url'):
+            for backup_url in durl['backup_url']:
                 formats.append({
-                    'url': backup_url.text,
+                    'url': backup_url,
                     # backup URLs have lower priorities
-                    'preference': -2 if 'hd.mp4' in backup_url.text else -3,
+                    'preference': -2 if 'hd.mp4' in backup_url else -3,
                 })
 
             self._sort_formats(formats)
 
             entries.append({
-                'id': '%s_part%s' % (cid, xpath_text(durl, './order')),
-                'duration': int_or_none(xpath_text(durl, './length'), 1000),
+                'id': '%s_part%s' % (video_id, idx),
+                'duration': float_or_none(durl.get('length'), 1000),
                 'formats': formats,
             })
 
         title = self._html_search_regex('<h1[^>]+title="([^"]+)">', webpage, 'title')
         description = self._html_search_meta('description', webpage)
-        datetime_str = self._html_search_regex(
-            r'<time[^>]+datetime="([^"]+)"', webpage, 'upload time', fatal=False)
-        timestamp = None
-        if datetime_str:
-            timestamp = calendar.timegm(datetime.datetime.strptime(datetime_str, '%Y-%m-%dT%H:%M').timetuple())
+        timestamp = unified_timestamp(self._html_search_regex(
+            r'<time[^>]+datetime="([^"]+)"', webpage, 'upload time', fatal=False))
 
         # TODO 'view_count' requires deobfuscating Javascript
         info = {
-            'id': compat_str(cid),
+            'id': video_id,
             'title': title,
             'description': description,
             'timestamp': timestamp,
             'thumbnail': self._html_search_meta('thumbnailUrl', webpage),
-            'duration': float_or_none(xpath_text(info_xml, './timelength'), scale=1000),
+            'duration': float_or_none(video_info.get('timelength'), scale=1000),
         }
 
         uploader_mobj = re.search(