[ie/sohu] Fix extractor (#7628)

Closes #1667, Closes #7463
Authored by: c-basalt, bashonly

yt_dlp/extractor/_extractors.py

@@ -1795,7 +1795,10 @@
 from .slutload import SlutloadIE
 from .smotrim import SmotrimIE
 from .snotr import SnotrIE
-from .sohu import SohuIE
+from .sohu import (
+    SohuIE,
+    SohuVIE,
+)
 from .sonyliv import (
     SonyLIVIE,
     SonyLIVSeriesIE,

yt_dlp/extractor/sohu.py

@@ -1,3 +1,4 @@
+import base64
 import re
 
 from .common import InfoExtractor
@@ -8,7 +9,12 @@
 from ..utils import (
     ExtractorError,
     int_or_none,
+    float_or_none,
+    url_or_none,
+    unified_timestamp,
     try_get,
+    urljoin,
+    traverse_obj,
 )
 
 
@@ -31,13 +37,20 @@ class SohuIE(InfoExtractor):
             'id': '409385080',
             'ext': 'mp4',
             'title': '《2015湖南卫视羊年元宵晚会》唐嫣《花好月圆》',
-        }
+        },
+        'skip': 'no longer available',
     }, {
         'url': 'http://my.tv.sohu.com/us/232799889/78693464.shtml',
         'info_dict': {
             'id': '78693464',
             'ext': 'mp4',
             'title': '【爱范品】第31期：MWC见不到的奇葩手机',
+            'uploader': '爱范儿视频',
+            'duration': 213,
+            'timestamp': 1425519600,
+            'upload_date': '20150305',
+            'thumbnail': 'http://e3f49eaa46b57.cdn.sohucs.com//group1/M10/83/FA/MTAuMTAuODguODA=/6_14cbccdde5eg104SysCutcloud_78693464_7_0b.jpg',
+            'tags': ['爱范儿', '爱范品', 'MWC', '手机'],
         }
     }, {
         'note': 'Multipart video',
@@ -45,6 +58,12 @@ class SohuIE(InfoExtractor):
         'info_dict': {
             'id': '78910339',
             'title': '【神探苍实战秘籍】第13期 战争之影 赫卡里姆',
+            'uploader': '小苍cany',
+            'duration': 744.0,
+            'timestamp': 1426269360,
+            'upload_date': '20150313',
+            'thumbnail': 'http://e3f49eaa46b57.cdn.sohucs.com//group1/M11/89/57/MTAuMTAuODguODA=/6_14cea022a1dg102SysCutcloud_78910339_8_0b.jpg',
+            'tags': ['小苍MM', '英雄联盟', '实战秘籍'],
         },
         'playlist': [{
             'info_dict': {
@@ -75,6 +94,11 @@ class SohuIE(InfoExtractor):
             'id': '78932792',
             'ext': 'mp4',
             'title': 'youtube-dl testing video',
+            'duration': 360,
+            'timestamp': 1426348620,
+            'upload_date': '20150314',
+            'thumbnail': 'http://e3f49eaa46b57.cdn.sohucs.com//group1/M02/8A/00/MTAuMTAuODguNzk=/6_14cee1be192g102SysCutcloud_78932792_7_7b.jpg',
+            'tags': [],
         },
         'params': {
             'skip_download': True
@@ -100,7 +124,7 @@ def _fetch_data(vid_id, mytv=False):
 
         webpage = self._download_webpage(url, video_id)
 
-        title = re.sub(r' - 搜狐视频$', '', self._og_search_title(webpage))
+        title = re.sub(r'( - 高清正版在线观看)? - 搜狐视频$', '', self._og_search_title(webpage))
 
         vid = self._html_search_regex(
             r'var vid ?= ?["\'](\d+)["\']',
@@ -132,7 +156,9 @@ def _fetch_data(vid_id, mytv=False):
                 allot = format_data['allot']
 
                 data = format_data['data']
-                clips_url = data['clipsURL']
+                clip_url = traverse_obj(data, (('clipsURL', 'mp4PlayUrl'), i, {url_or_none}), get_all=False)
+                if not clip_url:
+                    raise ExtractorError(f'Unable to extract url for clip {i}')
                 su = data['su']
 
                 video_url = 'newflv.sohu.ccgslb.net'
@@ -142,9 +168,9 @@ def _fetch_data(vid_id, mytv=False):
                 while 'newflv.sohu.ccgslb.net' in video_url:
                     params = {
                         'prot': 9,
-                        'file': clips_url[i],
+                        'file': clip_url,
                         'new': su[i],
-                        'prod': 'flash',
+                        'prod': 'h5n',
                         'rb': 1,
                     }
 
@@ -193,6 +219,75 @@ def _fetch_data(vid_id, mytv=False):
                 'entries': playlist,
                 'id': video_id,
                 'title': title,
+                'duration': traverse_obj(vid_data, ('data', 'totalDuration', {float_or_none})),
             }
 
-        return info
+        if mytv:
+            publish_time = unified_timestamp(self._search_regex(
+                r'publishTime:\s*["\'](\d+-\d+-\d+ \d+:\d+)["\']', webpage, 'publish time', fatal=False))
+        else:
+            publish_time = traverse_obj(vid_data, ('tv_application_time', {unified_timestamp}))
+
+        return {
+            'timestamp': publish_time - 8 * 3600 if publish_time else None,
+            **traverse_obj(vid_data, {
+                'alt_title': ('data', 'subName', {str}),
+                'uploader': ('wm_data', 'wm_username', {str}),
+                'thumbnail': ('data', 'coverImg', {url_or_none}),
+                'tags': ('data', 'tag', {str.split}),
+            }),
+            **info,
+        }
+
+
+class SohuVIE(InfoExtractor):
+    _VALID_URL = r'https?://tv\.sohu\.com/v/(?P<id>[\w=-]+)\.html(?:$|[#?])'
+
+    _TESTS = [{
+        'note': 'Multipart video',
+        'url': 'https://tv.sohu.com/v/MjAyMzA2MTQvbjYwMTMxNTE5Mi5zaHRtbA==.html',
+        'info_dict': {
+            'id': '601315192',
+            'title': '《淬火丹心》第1集',
+            'alt_title': '“点天灯”发生事故',
+            'duration': 2701.692,
+            'timestamp': 1686758040,
+            'upload_date': '20230614',
+            'thumbnail': 'http://photocdn.tv.sohu.com/img/20230614/vrsa_hor_1686738763256_454010551.jpg',
+        },
+        'playlist_mincount': 9,
+        'skip': 'Only available in China',
+    }, {
+        'url': 'https://tv.sohu.com/v/dXMvMjMyNzk5ODg5Lzc4NjkzNDY0LnNodG1s.html',
+        'info_dict': {
+            'id': '78693464',
+            'ext': 'mp4',
+            'title': '【爱范品】第31期：MWC见不到的奇葩手机',
+            'uploader': '爱范儿视频',
+            'duration': 213,
+            'timestamp': 1425519600,
+            'upload_date': '20150305',
+            'thumbnail': 'http://e3f49eaa46b57.cdn.sohucs.com//group1/M10/83/FA/MTAuMTAuODguODA=/6_14cbccdde5eg104SysCutcloud_78693464_7_0b.jpg',
+            'tags': ['爱范儿', '爱范品', 'MWC', '手机'],
+        }
+    }, {
+        'note': 'Multipart video',
+        'url': 'https://tv.sohu.com/v/dXMvMjQyNTYyMTYzLzc4OTEwMzM5LnNodG1s.html?src=pl',
+        'info_dict': {
+            'id': '78910339',
+            'title': '【神探苍实战秘籍】第13期 战争之影 赫卡里姆',
+            'uploader': '小苍cany',
+            'duration': 744.0,
+            'timestamp': 1426269360,
+            'upload_date': '20150313',
+            'thumbnail': 'http://e3f49eaa46b57.cdn.sohucs.com//group1/M11/89/57/MTAuMTAuODguODA=/6_14cea022a1dg102SysCutcloud_78910339_8_0b.jpg',
+            'tags': ['小苍MM', '英雄联盟', '实战秘籍'],
+        },
+        'playlist_mincount': 3,
+    }]
+
+    def _real_extract(self, url):
+        encoded_id = self._match_id(url)
+        path = base64.urlsafe_b64decode(encoded_id).decode()
+        subdomain = 'tv' if re.match(r'\d+/n\d+\.shtml', path) else 'my.tv'
+        return self.url_result(urljoin(f'http://{subdomain}.sohu.com/', path), SohuIE)