Merge branch 'fixYTSearch' of https://github.com/xarantolus/youtube-dl into xarantolus-fixYTSearch

youtube_dlc/extractor/youtube.py

@@ -3308,6 +3308,7 @@ class YoutubeSearchURLIE(YoutubeSearchBaseInfoExtractor):
     IE_DESC = 'YouTube.com search URLs'
     IE_NAME = 'youtube:search_url'
     _VALID_URL = r'https?://(?:www\.)?youtube\.com/results\?(.*?&)?(?:search_query|q)=(?P<query>[^&]+)(?:[&]|$)'
+    _SEARCH_DATA = r'(?:window\["ytInitialData"\]|ytInitialData)\W?=\W?({.*?});'
     _TESTS = [{
         'url': 'https://www.youtube.com/results?baz=bar&search_query=youtube-dl+test+video&filters=video&lclk=video',
         'playlist_mincount': 5,
@@ -3319,6 +3320,58 @@ class YoutubeSearchURLIE(YoutubeSearchBaseInfoExtractor):
         'only_matching': True,
     }]
 
+    def _find_videos_in_json(self, extracted):
+        videos = []
+
+        def _real_find(obj):
+            if obj is None or isinstance(obj, str):
+                return
+
+            if type(obj) is list:
+                for elem in obj:
+                    _real_find(elem)
+
+            if type(obj) is dict:
+                if "videoId" in obj:
+                    videos.append(obj)
+                    return
+
+                for _, o in obj.items():
+                    _real_find(o)
+
+        _real_find(extracted)
+
+        return videos
+
+    def extract_videos_from_page_impl(self, page, ids_in_page, titles_in_page):
+        search_response = self._parse_json(self._search_regex(self._SEARCH_DATA, page, 'ytInitialData'), None)
+
+        result_items = self._find_videos_in_json(search_response)
+
+        for renderer in result_items:
+            video_id = try_get(renderer, lambda x: x['videoId'])
+            video_title = try_get(renderer, lambda x: x['title']['runs'][0]['text']) or try_get(renderer, lambda x: x['title']['simpleText'])
+
+            if video_id is None or video_title is None:
+                # we do not have a videoRenderer or title extraction broke
+                continue
+
+            video_title = video_title.strip()
+
+            try:
+                idx = ids_in_page.index(video_id)
+                if video_title and not titles_in_page[idx]:
+                    titles_in_page[idx] = video_title
+            except ValueError:
+                ids_in_page.append(video_id)
+                titles_in_page.append(video_title)
+
+    def extract_videos_from_page(self, page):
+        ids_in_page = []
+        titles_in_page = []
+        self.extract_videos_from_page_impl(page, ids_in_page, titles_in_page)
+        return zip(ids_in_page, titles_in_page)
+
     def _real_extract(self, url):
         mobj = re.match(self._VALID_URL, url)
         query = compat_urllib_parse_unquote_plus(mobj.group('query'))
@@ -3351,6 +3404,8 @@ class YoutubeFeedsInfoExtractor(YoutubeBaseInfoExtractor):
     Subclasses must define the _FEED_NAME and _PLAYLIST_TITLE properties.
     """
     _LOGIN_REQUIRED = True
+    _FEED_DATA = r'(?:window\["ytInitialData"\]|ytInitialData)\W?=\W?({.*?});'
+    _YTCFG_DATA = r"ytcfg.set\(({.*?})\)"
 
     @property
     def IE_NAME(self):
@@ -3359,37 +3414,89 @@ def IE_NAME(self):
     def _real_initialize(self):
         self._login()
 
+    def _find_videos_in_json(self, extracted):
+        videos = []
+        c = {}
+
+        def _real_find(obj):
+            if obj is None or isinstance(obj, str):
+                return
+
+            if type(obj) is list:
+                for elem in obj:
+                    _real_find(elem)
+
+            if type(obj) is dict:
+                if "videoId" in obj:
+                    videos.append(obj)
+                    return
+
+                if "nextContinuationData" in obj:
+                    c["continuation"] = obj["nextContinuationData"]
+                    return
+
+                for _, o in obj.items():
+                    _real_find(o)
+
+        _real_find(extracted)
+
+        return videos, try_get(c, lambda x: x["continuation"])
+
     def _entries(self, page):
-        # The extraction process is the same as for playlists, but the regex
+        info = []
-        # for the video ids doesn't contain an index
+
-        ids = []
+        yt_conf = self._parse_json(self._search_regex(self._YTCFG_DATA, page, 'ytcfg.set', default="null"), None, fatal=False)
-        more_widget_html = content_html = page
+
+        search_response = self._parse_json(self._search_regex(self._FEED_DATA, page, 'ytInitialData'), None)
+
         for page_num in itertools.count(1):
-            matches = re.findall(r'href="\s*/watch\?v=([0-9A-Za-z_-]{11})', content_html)
+            video_info, continuation = self._find_videos_in_json(search_response)
 
-            # 'recommended' feed has infinite 'load more' and each new portion spins
+            new_info = []
-            # the same videos in (sometimes) slightly different order, so we'll check
+
-            # for unicity and break when portion has no new videos
+            for v in video_info:
-            new_ids = list(filter(lambda video_id: video_id not in ids, orderedSet(matches)))
+                v_id = try_get(v, lambda x: x['videoId'])
-            if not new_ids:
+                if not v_id:
+                    continue
+
+                have_video = False
+                for old in info:
+                    if old['videoId'] == v_id:
+                        have_video = True
+                        break
+
+                if not have_video:
+                    new_info.append(v)
+
+            if not new_info:
                 break
 
-            ids.extend(new_ids)
+            info.extend(new_info)
 
-            for entry in self._ids_to_results(new_ids):
+            for video in new_info:
-                yield entry
+                yield self.url_result(try_get(video, lambda x: x['videoId']), YoutubeIE.ie_key(), video_title=try_get(video, lambda x: x['title']['runs'][0]['text']) or try_get(video, lambda x: x['title']['simpleText']))
 
-            mobj = re.search(r'data-uix-load-more-href="/?(?P<more>[^"]+)"', more_widget_html)
+            if not continuation or not yt_conf:
-            if not mobj:
                 break
 
-            more = self._download_json(
+            search_response = self._download_json(
-                'https://www.youtube.com/%s' % mobj.group('more'), self._PLAYLIST_TITLE,
+                'https://www.youtube.com/browse_ajax', self._PLAYLIST_TITLE,
                 'Downloading page #%s' % page_num,
                 transform_source=uppercase_escape,
-                headers=self._YOUTUBE_CLIENT_HEADERS)
+                query={
-            content_html = more['content_html']
+                    "ctoken": try_get(continuation, lambda x: x["continuation"]),
-            more_widget_html = more['load_more_widget_html']
+                    "continuation": try_get(continuation, lambda x: x["continuation"]),
+                    "itct": try_get(continuation, lambda x: x["clickTrackingParams"])
+                },
+                headers={
+                    "X-YouTube-Client-Name": try_get(yt_conf, lambda x: x["INNERTUBE_CONTEXT_CLIENT_NAME"]),
+                    "X-YouTube-Client-Version": try_get(yt_conf, lambda x: x["INNERTUBE_CONTEXT_CLIENT_VERSION"]),
+                    "X-Youtube-Identity-Token": try_get(yt_conf, lambda x: x["ID_TOKEN"]),
+                    "X-YouTube-Device": try_get(yt_conf, lambda x: x["DEVICE"]),
+                    "X-YouTube-Page-CL": try_get(yt_conf, lambda x: x["PAGE_CL"]),
+                    "X-YouTube-Page-Label": try_get(yt_conf, lambda x: x["PAGE_BUILD_LABEL"]),
+                    "X-YouTube-Variants-Checksum": try_get(yt_conf, lambda x: x["VARIANTS_CHECKSUM"]),
+                })
 
     def _real_extract(self, url):
         page = self._download_webpage(