[ie/instagram] Rework extractor (#17075)

Closes #7165
Closes #8408
Closes #17074

Authored by: bashonly

README.md

@@ -1916,7 +1916,7 @@ The following extractors use this feature:
 * `dr`: dynamic range to ignore - one or more of `sdr`, `hdr10`, `dv`
 
 #### instagram
-* `app_id`: The value of the `X-IG-App-ID` header used for API requests. Default is the web app ID, `936619743392459`
+* `app_id`: The value of the `X-IG-App-ID` header used for API requests. Can be the actual ID number, `ios`, or `web` (default)
 
 #### niconicochannelplus
 * `max_comments`: Maximum number of comments to extract - default is `120`

yt_dlp/extractor/instagram.py

@@ -1,3 +1,4 @@
+import functools
 import hashlib
 import itertools
 import json
@@ -16,11 +17,10 @@ from ..utils import (
     get_element_by_attribute,
     int_or_none,
     join_nonempty,
-    lowercase_escape,
     str_or_none,
-    str_to_int,
     traverse_obj,
     url_or_none,
+    urlencode_postdata,
 )
 
 _ENCODING_CHARS = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_'
@@ -41,13 +41,31 @@ def _id_to_pk(shortcode):
 
 class InstagramBaseIE(InfoExtractor):
     _API_BASE_URL = 'https://i.instagram.com/api/v1'
+    _BASE_URL = 'https://www.instagram.com/'
     _LOGIN_URL = 'https://www.instagram.com/accounts/login'
+    _APP_IDS = {
+        'ios': '124024574287414',
+        'web': '936619743392459',  # default
+    }
+
+    @property
+    def _is_logged_in(self):
+        return bool(self._get_cookies(self._BASE_URL).get('sessionid'))
+
+    @functools.cached_property
+    def _app_id(self):
+        user_input = self._configuration_arg('app_id', [None], ie_key=InstagramIE)[0]
+        return self._APP_IDS.get(user_input, user_input or self._APP_IDS['web'])
+
+    @property
+    def _is_web_app(self):
+        return self._app_id == self._APP_IDS['web']
 
     @property
     def _api_headers(self):
         return {
-            'X-IG-App-ID': self._configuration_arg('app_id', ['936619743392459'], ie_key=InstagramIE)[0],
+            'X-IG-App-ID': self._app_id,
-            'X-ASBD-ID': '198387',
+            'X-ASBD-ID': '359341',
             'X-IG-WWW-Claim': '0',
             'Origin': 'https://www.instagram.com',
             'Accept': '*/*',
@@ -106,71 +124,83 @@ class InstagramBaseIE(InfoExtractor):
             }
 
     def _extract_product_media(self, product_media):
-        media_id = product_media.get('code') or _pk_to_id(product_media.get('pk'))
+        video_id = traverse_obj(product_media, ('pk', {_pk_to_id}))
-        vcodec = product_media.get('video_codec')
-        dash_manifest_raw = product_media.get('video_dash_manifest')
-        videos_list = product_media.get('video_versions')
-        if not (dash_manifest_raw or videos_list):
-            return {}
 
-        formats = [{
+        formats = traverse_obj(product_media, ('video_versions', lambda _, v: url_or_none(v['url']), {
-            'format_id': fmt.get('id'),
+            'url': 'url',
-            'url': fmt.get('url'),
+            'format_id': (('id', {str}), ('type', {int}, {str_or_none}), any),
-            'width': fmt.get('width'),
+            'width': ('width', {int_or_none}),
-            'height': fmt.get('height'),
+            'height': ('height', {int_or_none}),
-            'vcodec': vcodec,
+        }))
-        } for fmt in videos_list or []]
+        format_info = traverse_obj(product_media, {
-        if dash_manifest_raw:
+            'vcodec': ('video_codec', {str}),
-            formats.extend(self._parse_mpd_formats(self._parse_xml(dash_manifest_raw, media_id), mpd_id='dash'))
+            'acodec': ('has_audio', {bool}, {lambda x: 'none' if x is False else None}),
+        })
+        for f in formats:
+            f.update(format_info)
+
+        dash = traverse_obj(product_media, ('video_dash_manifest', {str}))
+        if dash:
+            formats.extend(self._parse_mpd_formats(self._parse_xml(dash, video_id), mpd_id='dash'))
 
-        thumbnails = [{
-            'url': thumbnail.get('url'),
-            'width': thumbnail.get('width'),
-            'height': thumbnail.get('height'),
-        } for thumbnail in traverse_obj(product_media, ('image_versions2', 'candidates')) or []]
         return {
-            'id': media_id,
+            'id': video_id,
-            'duration': float_or_none(product_media.get('video_duration')),
             'formats': formats,
-            'thumbnails': thumbnails,
+            'duration': traverse_obj(product_media, ('video_duration', {float_or_none})),
+            'thumbnails': list(reversed(traverse_obj(product_media, (
+                'image_versions2', 'candidates',
+                lambda _, v: url_or_none(v['url']), {
+                    'url': 'url',
+                    'width': ('width', {int}),
+                    'height': 'height',
+                },
+            )))),
         }
 
-    def _extract_product(self, product_info):
+    def _extract_product(self, product_info, video_id=None, get_comments=True):
         if isinstance(product_info, list):
             product_info = product_info[0]
 
-        user_info = product_info.get('user') or {}
         info_dict = {
-            'id': _pk_to_id(traverse_obj(product_info, 'pk', 'id', expected_type=str_or_none)[:19]),
+            'id': video_id,
-            'title': product_info.get('title') or f'Video by {user_info.get("username")}',
+            'title': format_field(product_info, [('user', 'username', {str})], 'Video by %s', default=None),
-            'description': traverse_obj(product_info, ('caption', 'text'), expected_type=str_or_none),
+            **traverse_obj(product_info, {
-            'timestamp': int_or_none(product_info.get('taken_at')),
+                'id': ('pk', {_pk_to_id}),
-            'channel': user_info.get('username'),
+                'title': ('title', {str}),
-            'uploader': user_info.get('full_name'),
+                'description': ('caption', 'text', {str}),
-            'uploader_id': str_or_none(user_info.get('pk')),
+                'channel': ('user', 'username', {str}),
-            'view_count': int_or_none(product_info.get('view_count')),
+                'uploader_id': ('user', 'pk', {str_or_none}),
-            'like_count': int_or_none(product_info.get('like_count')),
+                'uploader': ('user', 'full_name', {str}),
-            'comment_count': int_or_none(product_info.get('comment_count')),
+                'timestamp': ('taken_at', {int_or_none}),
-            '__post_extractor': self.extract_comments(_pk_to_id(product_info.get('pk'))),
+                'view_count': ('view_count', {int_or_none}),
+                'like_count': ('like_count', {int_or_none}),
+                'comment_count': ('comment_count', {int_or_none}),
+            }),
             'http_headers': {
                 'Referer': 'https://www.instagram.com/',
             },
         }
-        carousel_media = product_info.get('carousel_media')
+
-        if carousel_media:
+        if carousel_media := traverse_obj(product_info, ('carousel_media', ..., {dict})):
+            comments = None
+            if get_comments and self.get_param('getcomments'):
+                comments = self._get_comments(info_dict.get('id'))
+
             return {
                 '_type': 'playlist',
                 **info_dict,
-                'title': f'Post by {user_info.get("username")}',
+                'title': format_field(info_dict, 'channel', 'Post by %s', default=None),
                 'entries': [{
                     **info_dict,
                     **self._extract_product_media(product_media),
+                    'comments': comments,
                 } for product_media in carousel_media],
             }
 
         return {
             **info_dict,
             **self._extract_product_media(product_info),
+            '__post_extractor': self.extract_comments(info_dict.get('id')) if get_comments else None,
         }
 
     def _get_comments(self, video_id):
@@ -182,9 +212,9 @@ class InstagramBaseIE(InfoExtractor):
         for comment_dict in comment_data or []:
             yield {
                 'author': traverse_obj(comment_dict, ('node', 'owner', 'username'), ('user', 'username')),
-                'author_id': traverse_obj(comment_dict, ('node', 'owner', 'id'), ('user', 'pk')),
+                'author_id': traverse_obj(comment_dict, ('node', 'owner', 'id'), ('user', 'pk'), expected_type=str_or_none),
                 'author_thumbnail': traverse_obj(comment_dict, ('node', 'owner', 'profile_pic_url'), ('user', 'profile_pic_url'), expected_type=url_or_none),
-                'id': traverse_obj(comment_dict, ('node', 'id'), 'pk'),
+                'id': traverse_obj(comment_dict, ('node', 'id'), 'pk', expected_type=str_or_none),
                 'text': traverse_obj(comment_dict, ('node', 'text'), 'text'),
                 'like_count': traverse_obj(comment_dict, ('node', 'edge_liked_by', 'count'), 'comment_like_count', expected_type=int_or_none),
                 'timestamp': traverse_obj(comment_dict, ('node', 'created_at'), 'created_at', expected_type=int_or_none),
@@ -350,6 +380,8 @@ class InstagramIE(InstagramBaseIE):
         'only_matching': True,
     }]
 
+    _lsd_token = None
+
     @classmethod
     def _extract_embed_urls(cls, url, webpage):
         res = tuple(super()._extract_embed_urls(url, webpage))
@@ -361,165 +393,107 @@ class InstagramIE(InstagramBaseIE):
         if mobj:
             return [mobj.group('link')]
 
+    def _real_initialize(self):
+        if self._is_logged_in:
+            return
+        if not self._lsd_token:
+            webpage = self._download_webpage(
+                self._BASE_URL, None, 'Setting up session',
+                impersonate=True, require_impersonation=True)
+            eqmc = self._search_json(
+                r'<script\b[^>]* id="__eqmc"[^>]*>', webpage, 'eqmc JSON', None, default={})
+            self._lsd_token = (
+                traverse_obj(eqmc, ('l', {str}))
+                or self._search_regex(r'\["LSD",\[\],\{"token":"([^"]+)"', webpage, 'LSD token'))
+
     def _real_extract(self, url):
         video_id, url = self._match_valid_url(url).group('id', 'url')
-        media, webpage = {}, ''
+        media_id = str(_id_to_pk(video_id))
 
-        if self._get_cookies(url).get('sessionid'):
+        if self._is_logged_in:
-            info = traverse_obj(self._download_json(
+            return self._extract_product(self._download_json(
-                f'{self._API_BASE_URL}/media/{_id_to_pk(video_id)}/info/', video_id,
+                f'{self._API_BASE_URL}/media/{media_id}/info/', video_id,
-                fatal=False, errnote='Video info extraction failed',
+                'Downloading video info', 'Video info extraction failed',
-                note='Downloading video info', headers=self._api_headers), ('items', 0))
+                impersonate=self._is_web_app, headers=self._api_headers)['items'][0])
-            if info:
-                media.update(info)
-                return self._extract_product(media)
 
         api_check = self._download_json(
-            f'{self._API_BASE_URL}/web/get_ruling_for_content/?content_type=MEDIA&target_id={_id_to_pk(video_id)}',
+            f'{self._API_BASE_URL}/web/get_ruling_for_content/', video_id,
-            video_id, headers=self._api_headers, fatal=False, note='Setting up session', errnote=False) or {}
+            'Checking post accessibility', errnote=False, fatal=False,
-        csrf_token = self._get_cookies('https://www.instagram.com').get('csrftoken')
+            impersonate=True, require_impersonation=True, headers=self._api_headers,
+            query={'content_type': 'MEDIA', 'target_id': media_id}) or {}
 
+        csrf_token = self._get_cookies('https://www.instagram.com').get('csrftoken')
         if not csrf_token:
-            self.report_warning('No csrf token set by Instagram API', video_id)
+            self.report_warning('No CSRF token set by Instagram API', video_id)
         else:
             csrf_token = csrf_token.value if api_check.get('status') == 'ok' else None
             if not csrf_token:
                 self.report_warning('Instagram API is not granting access', video_id)
 
-        variables = {
+        response = self._download_json(
-            'shortcode': video_id,
+            'https://www.instagram.com/api/graphql', video_id,
-            'child_comment_count': 3,
+            impersonate=True, require_impersonation=True,
-            'fetch_comment_count': 40,
+            headers=filter_dict({
-            'parent_comment_count': 24,
-            'has_threaded_comments': True,
-        }
-        general_info = self._download_json(
-            'https://www.instagram.com/graphql/query/', video_id, fatal=False, errnote=False,
-            headers={
                 **self._api_headers,
-                'X-CSRFToken': csrf_token or '',
+                'X-FB-Friendly-Name': 'PolarisLoggedOutDesktopWWWPostRootContentQuery',
+                'X-CSRFToken': csrf_token,
+                'X-FB-LSD': self._lsd_token,
                 'X-Requested-With': 'XMLHttpRequest',
                 'Referer': url,
-            }, query={
+            }), data=urlencode_postdata({
-                'doc_id': '8845758582119845',
+                'av': '0',
-                'variables': json.dumps(variables, separators=(',', ':')),
+                '__d': 'www',
-            })
+                '__user': '0',
+                'dpr': '1',
+                'lsd': self._lsd_token,
+                'fb_api_caller_class': 'RelayModern',
+                'fb_api_req_friendly_name': 'PolarisLoggedOutDesktopWWWPostRootContentQuery',
+                'server_timestamps': 'true',
+                'variables': json.dumps({'media_id': media_id}, separators=(',', ':')),
+                'doc_id': '27130156389949648',
+            }))
 
-        if not general_info:
+        media = traverse_obj(response, ('data', 'xig_polaris_media', {dict}))
-            self.report_warning('General metadata extraction failed (some metadata might be missing).', video_id)
+        product_info = traverse_obj(media, ('if_not_gated_logged_out', {dict}))
-            webpage, urlh = self._download_webpage_handle(url, video_id)
+        if not product_info:
-            shared_data = self._search_json(
+            error = join_nonempty('title', 'description', delim=': ', from_dict=api_check)
-                r'window\._sharedData\s*=', webpage, 'shared data', video_id, fatal=False) or {}
+            if 'Restricted Video' in error:
+                self.raise_login_required(error)
+            elif error:
+                raise ExtractorError(error, expected=True)
+            elif len(video_id) > 28:
+                # It's a private post (video_id == shortcode + 28 extra characters)
+                # Only raise after getting empty response; sometimes "long"-shortcode posts are public
+                self.raise_login_required(
+                    'This content is only available for registered users who follow this account')
+            raise ExtractorError(
+                'Instagram sent an empty media response. Check if this post is accessible in your '
+                f'browser without being logged-in. If it is not, then u{self._login_hint()[1:]}. '
+                'Otherwise, if the post is accessible in browser without being logged-in'
+                f'{bug_reports_message(before=",")}', expected=True)
 
-            if shared_data and self._LOGIN_URL not in urlh.url:
+        info_dict = self._extract_product(product_info, video_id=video_id, get_comments=False)
-                media.update(traverse_obj(
+        is_playlist = info_dict.get('_type') == 'playlist'
-                    shared_data, ('entry_data', 'PostPage', 0, 'graphql', 'shortcode_media'),
+        if not is_playlist and not info_dict.get('formats'):
-                    ('entry_data', 'PostPage', 0, 'media'), expected_type=dict) or {})
+            self.raise_no_formats('There is no video in this post', expected=True)
-            else:
-                self.report_warning('Main webpage is locked behind the login page. Retrying with embed webpage (some metadata might be missing).')
-                webpage = self._download_webpage(
-                    f'{url}/embed/', video_id, note='Downloading embed webpage', fatal=False) or ''
-                additional_data = self._search_json(
-                    r'window\.__additionalDataLoaded\s*\(\s*[^,]+,', webpage, 'additional data', video_id, fatal=False)
-                if not additional_data and not media:
-                    self.raise_login_required('Requested content is not available, rate-limit reached or login required')
 
-                product_item = traverse_obj(additional_data, ('items', 0), expected_type=dict)
+        comments = traverse_obj(media, (
-                if product_item:
+            'comments_connection', 'edges', lambda _, v: v['node']['text'], 'node', {
-                    media.update(product_item)
+                'author': ('user', 'username', {str}),
-                    return self._extract_product(media)
+                'author_id': ('user', 'pk', {str_or_none}),
-
+                'id': ('pk', {str_or_none}),
-                media.update(traverse_obj(
+                'text': ('text', {str}),
-                    additional_data, ('graphql', 'shortcode_media'), 'shortcode_media', expected_type=dict) or {})
+                'timestamp': ('created_at', {int_or_none}),
+                'like_count': ('comment_like_count', {int_or_none}),
+            }))
 
+        if is_playlist:
+            for entry in info_dict['entries']:
+                entry['comments'] = comments
         else:
-            xdt_shortcode_media = traverse_obj(general_info, ('data', 'xdt_shortcode_media', {dict})) or {}
+            info_dict['comments'] = comments
-            if not xdt_shortcode_media:
-                error = join_nonempty('title', 'description', delim=': ', from_dict=api_check)
-                if 'Restricted Video' in error:
-                    self.raise_login_required(error)
-                elif error:
-                    raise ExtractorError(error, expected=True)
-                elif len(video_id) > 28:
-                    # It's a private post (video_id == shortcode + 28 extra characters)
-                    # Only raise after getting empty response; sometimes "long"-shortcode posts are public
-                    self.raise_login_required(
-                        'This content is only available for registered users who follow this account')
-                raise ExtractorError(
-                    'Instagram sent an empty media response. Check if this post is accessible in your '
-                    f'browser without being logged-in. If it is not, then u{self._login_hint()[1:]}. '
-                    'Otherwise, if the post is accessible in browser without being logged-in'
-                    f'{bug_reports_message(before=",")}', expected=True)
-            media.update(xdt_shortcode_media)
 
-        username = traverse_obj(media, ('owner', 'username')) or self._search_regex(
+        return info_dict
-            r'"owner"\s*:\s*{\s*"username"\s*:\s*"(.+?)"', webpage, 'username', fatal=False)
-
-        description = (
-            traverse_obj(media, ('edge_media_to_caption', 'edges', 0, 'node', 'text'), expected_type=str)
-            or media.get('caption'))
-        if not description:
-            description = self._search_regex(
-                r'"caption"\s*:\s*"(.+?)"', webpage, 'description', default=None)
-            if description is not None:
-                description = lowercase_escape(description)
-
-        video_url = media.get('video_url')
-        if not video_url:
-            nodes = traverse_obj(media, ('edge_sidecar_to_children', 'edges', ..., 'node'), expected_type=dict) or []
-            if nodes:
-                return self.playlist_result(
-                    self._extract_nodes(nodes, True), video_id,
-                    format_field(username, None, 'Post by %s'), description)
-            raise ExtractorError('There is no video in this post', expected=True)
-
-        formats = [{
-            'url': video_url,
-            'width': self._get_dimension('width', media, webpage),
-            'height': self._get_dimension('height', media, webpage),
-        }]
-        dash = traverse_obj(media, ('dash_info', 'video_dash_manifest'))
-        if dash:
-            formats.extend(self._parse_mpd_formats(self._parse_xml(dash, video_id), mpd_id='dash'))
-
-        comment_data = traverse_obj(media, ('edge_media_to_parent_comment', 'edges'))
-        comments = [{
-            'author': traverse_obj(comment_dict, ('node', 'owner', 'username')),
-            'author_id': traverse_obj(comment_dict, ('node', 'owner', 'id')),
-            'id': traverse_obj(comment_dict, ('node', 'id')),
-            'text': traverse_obj(comment_dict, ('node', 'text')),
-            'timestamp': traverse_obj(comment_dict, ('node', 'created_at'), expected_type=int_or_none),
-        } for comment_dict in comment_data] if comment_data else None
-
-        display_resources = (
-            media.get('display_resources')
-            or [{'src': media.get(key)} for key in ('display_src', 'display_url')]
-            or [{'src': self._og_search_thumbnail(webpage)}])
-        thumbnails = [{
-            'url': thumbnail['src'],
-            'width': thumbnail.get('config_width'),
-            'height': thumbnail.get('config_height'),
-        } for thumbnail in display_resources if thumbnail.get('src')]
-
-        return {
-            'id': video_id,
-            'formats': formats,
-            'title': media.get('title') or f'Video by {username}',
-            'description': description,
-            'duration': float_or_none(media.get('video_duration')),
-            'timestamp': traverse_obj(media, 'taken_at_timestamp', 'date', expected_type=int_or_none),
-            'uploader_id': traverse_obj(media, ('owner', 'id')),
-            'uploader': traverse_obj(media, ('owner', 'full_name')),
-            'channel': username,
-            'like_count': self._get_count(media, 'likes', 'preview_like') or str_to_int(self._search_regex(
-                r'data-log-event="likeCountClick"[^>]*>[^\d]*([\d,\.]+)', webpage, 'like count', fatal=False)),
-            'comment_count': self._get_count(media, 'comments', 'preview_comment', 'to_comment', 'to_parent_comment'),
-            'comments': comments,
-            'thumbnails': thumbnails,
-            'http_headers': {
-                'Referer': 'https://www.instagram.com/',
-            },
-        }
 
 
 class InstagramPlaylistBaseIE(InstagramBaseIE):
@@ -697,12 +671,12 @@ class InstagramStoryIE(InstagramBaseIE):
         if username == 'highlights' and not story_id:  # story id is only mandatory for highlights
             raise ExtractorError('Input URL is missing a highlight ID', expected=True)
         display_id = story_id or username
-        story_info = self._download_webpage(url, display_id)
+        story_info = self._download_webpage(url, display_id, impersonate=self._is_web_app)
         user_info = self._search_json(r'"user":', story_info, 'user info', display_id, fatal=False)
         if not user_info:
             self.raise_login_required('This content is unreachable')
 
-        user_id = traverse_obj(user_info, 'pk', 'id', expected_type=str)
+        user_id = traverse_obj(user_info, 'pk', 'id', expected_type=str_or_none)
         if username == 'highlights':
             story_info_url = f'highlight:{story_id}'
         else:
@@ -712,7 +686,8 @@ class InstagramStoryIE(InstagramBaseIE):
 
         videos = traverse_obj(self._download_json(
             f'{self._API_BASE_URL}/feed/reels_media/?reel_ids={story_info_url}',
-            display_id, errnote=False, fatal=False, headers=self._api_headers), 'reels')
+            display_id, errnote=False, fatal=False, impersonate=self._is_web_app,
+            headers=self._api_headers), 'reels')
         if not videos:
             self.raise_login_required('You need to log in to access this content')
         user_info = traverse_obj(videos, (user_id, 'user', {dict})) or {}