追跡対象動画バルク UPSERT (#22) (#23)

#22 インポート漏れ修正

#22

#22

Co-authored-by: miteruzo <miteruzo@naver.com>
Reviewed-on: #23

db/config.py

@@ -0,0 +1,25 @@
+from __future__ import annotations
+
+import os
+from typing import TypedDict
+
+from eloquent import DatabaseManager, Model  # type: ignore
+
+CONFIG: dict[str, DbConfig] = { 'mysql': { 'driver':    'mysql',
+                                           'host':      'localhost',
+                                           'database':  'nizika_nico',
+                                           'user':      os.environ['MYSQL_USER'],
+                                           'password':  os.environ['MYSQL_PASS'],
+                                           'prefix':    '' } }
+
+DB = DatabaseManager (CONFIG)
+Model.set_connection_resolver (DB)
+
+
+class DbConfig (TypedDict):
+    driver:     str
+    host:       str
+    database:   str
+    user:       str
+    password:   str
+    prefix:     str

db/eloquent.pyi

@@ -115,6 +115,9 @@ class Model:
     @classmethod
     def where_not_null (cls, field: str) -> QueryBuilder[Self]: ...
 
+    @classmethod
+    def max (cls, column: str) -> Any: ...
+
     @classmethod
     def set_connection_resolver (cls, resolver: DatabaseManager) -> None: ...
 
@@ -137,4 +140,6 @@ class QueryBuilder (Generic[_ModelT]):
 
     def where_null (self, field: str) -> QueryBuilder[_ModelT]: ...
 
+    def max (self, column: str) -> Any: ...
+
     def _load_relation (self, relation_name: str) -> QueryBuilder[_ModelT]: ...

db/models.py

@@ -58,6 +58,19 @@ class Tag (Model):
         return self.has_many (VideoTag)
 
 
+class TrackedVideo (Model):
+    id:     int
+    code:   str
+
+    __timestamps__ = False
+
+    def upsert (
+            self,
+            *args:  str,
+    ) -> None:
+        super ().upsert ('code')
+
+
 class User (Model):
     id:     int
     code:   str
@@ -74,6 +87,7 @@ class User (Model):
 class Video (Model):
     id:             int
     code:           str
+    user_id:        int | None
     title:          str
     description:    str
     uploaded_at:    datetime
@@ -81,6 +95,14 @@ class Video (Model):
 
     __timestamps__ = False
 
+    @property
+    def user (
+            self,
+    ) -> User | None:
+        if self.user_id is None:
+            return None
+        return self.belongs_to (User)
+
     @property
     def video_histories (
             self,

db/my_eloquent.py

@@ -19,13 +19,32 @@ class Model (eloquent.Model):
             self,
             *args:  str,
     ) -> None:
-        q = self.query ()
+        row = self._find_upsert_row (*args)
-        for arg in args:
-            q = q.where (arg, getattr (self, arg))
-        row = q.first ()
         if row is not None:
             self.id = row.id
             # pylint: disable = invalid-name
             # pylint: disable = attribute-defined-outside-init
             self._Model__exists = True
             self.save ()
+            return
+
+        try:
+            self.save ()
+        except Exception:
+            row = self._find_upsert_row (*args)
+            if row is None:
+                raise
+            self.id = row.id
+            # pylint: disable = invalid-name
+            # pylint: disable = attribute-defined-outside-init
+            self._Model__exists = True
+            self.save ()
+
+    def _find_upsert_row (
+            self,
+            *args:  str,
+    ):
+        q = self.query ()
+        for arg in args:
+            q = q.where (arg, getattr (self, arg))
+        return q.first ()

get_comments_by_video_code.py

@@ -0,0 +1,55 @@
+# pylint: disable = missing-class-docstring
+# pylint: disable = missing-function-docstring
+
+"""
+動画コードからコメントのリストを取得し，JSON 形式で出力する．
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import datetime
+from typing import TypedDict
+
+from db.config import DB
+from db.models import Video
+
+DB
+
+
+def main (
+        video_code: str,
+) -> None:
+    video = Video.where ('code', video_code).first ()
+    if video:
+        comments: list[CommentDict] = []
+        for row in video.comments:
+            comment: CommentDict = {
+                    'id':           row.id,
+                    'video_id':     row.video_id,
+                    'comment_no':   row.comment_no,
+                    'user_id':      row.user_id,
+                    'content':      row.content,
+                    'posted_at':    row.posted_at,
+                    'nico_count':   row.nico_count,
+                    'vpos_ms':      row.vpos_ms }
+            comments.append (comment)
+        print (json.dumps (comments, default = str))
+    else:
+        print ('[]')
+
+
+class CommentDict (TypedDict):
+    id:         int
+    video_id:   int
+    comment_no: int
+    user_id:    int
+    content:    str
+    posted_at:  datetime
+    nico_count: int
+    vpos_ms:    int
+
+
+if __name__ == '__main__':
+    main (sys.argv[1])

get_kiriban_list.py

@@ -0,0 +1,65 @@
+# pylint: disable = missing-class-docstring
+# pylint: disable = missing-function-docstring
+
+"""
+動画履歴の情報を取得し，JSON 形式で出力する．
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import date, datetime
+from typing import cast
+
+from db.config import DB
+from db.models import Video, VideoHistory
+
+DB
+
+
+def main (
+        views_counts:   list[int],
+        base_date:      date,
+) -> None:
+    kiriban_list: list[tuple[int, str, str]] = []
+
+    latest_fetched_at = cast (date | None,
+                              (VideoHistory
+                               .where ('fetched_at', '<=', base_date)
+                               .max ('fetched_at')))
+    if latest_fetched_at is None:
+        print ('[]')
+        return
+
+    for views_count in views_counts:
+        targets = { vh.video.code for vh in (
+                VideoHistory
+                .where ('fetched_at', latest_fetched_at)
+                .where ('views_count', '>=', views_count)
+                .get ()) }
+
+        for code in targets:
+            if code in [kiriban[1] for kiriban in kiriban_list]:
+                continue
+
+            previous_views_count: int | None = (
+                    VideoHistory
+                    .where_has ('video', lambda q, code = code: q.where ('code', code))
+                    .where ('fetched_at', '<', latest_fetched_at)
+                    .max ('views_count'))
+            if previous_views_count is None:
+                previous_views_count = 0
+            if previous_views_count >= views_count:
+                continue
+
+            kiriban_list.append ((views_count, code,
+                                  (cast (Video, Video.where ('code', code).first ())
+                                   .uploaded_at)))
+
+    print (json.dumps (kiriban_list, default = str))
+
+
+if __name__ == '__main__':
+    main (list (map (int, sys.argv[2:])),
+          datetime.strptime (sys.argv[1], '%Y-%m-%d').date ())

get_videos.py

@@ -0,0 +1,52 @@
+# pylint: disable = missing-class-docstring
+# pylint: disable = missing-function-docstring
+
+"""
+全動画の情報を取得し，JSON 形式で出力する．
+"""
+
+from __future__ import annotations
+
+import json
+from datetime import date, datetime
+from typing import TypedDict
+
+from db.config import DB
+from db.models import Video
+
+DB
+
+
+def main (
+) -> None:
+    videos: list[VideoDict] = []
+    for row in Video.all ():
+        deleted_at = row.deleted_at.date () if row.deleted_at else None
+        video: VideoDict = { 'id':          row.id,
+                             'code':        row.code,
+                             'user':        getattr (row.user, 'code', None),
+                             'title':       row.title,
+                             'description': row.description,
+                             'tags':        [],
+                             'uploaded_at': row.uploaded_at,
+                             'deleted_at':  deleted_at }
+        for video_tag in row.video_tags:
+            if video_tag.untagged_at is None:
+                video['tags'].append (video_tag.tag.name)
+        videos.append (video)
+
+    print (json.dumps (videos, default = str))
+
+class VideoDict (TypedDict):
+    id:             int
+    code:           str
+    user:           str | None
+    title:          str
+    description:    str
+    tags:           list[str]
+    uploaded_at:    datetime
+    deleted_at:     date | None
+
+
+if __name__ == '__main__':
+    main ()

migrations/2026_01_01_041600_create_tracked_videos.sql

@@ -0,0 +1,2 @@
+CREATE TABLE `nizika_nico`.`tracked_videos` (`id` BIGINT NOT NULL AUTO_INCREMENT , `code` VARCHAR(16) NOT NULL COMMENT '動画コード' , PRIMARY KEY (`id`)) ENGINE = InnoDB COMMENT = '追跡対象動画';
+ALTER TABLE `tracked_videos` ADD UNIQUE(`code`);

migrations/2026_03_05_122300_add_user_id_to_videos.sql

@@ -0,0 +1,3 @@
+ALTER TABLE `videos` ADD `user_id` BIGINT NULL DEFAULT NULL COMMENT 'ユーザ Id.' AFTER `code`;
+ALTER TABLE `videos` ADD INDEX(`user_id`);
+ALTER TABLE `videos` ADD FOREIGN KEY (`user_id`) REFERENCES `users`(`id`) ON DELETE RESTRICT ON UPDATE CASCADE;

tracked_videos/put_bulk_upsert.py

@@ -0,0 +1,19 @@
+import sys
+
+from db.config import DB
+from db.models import TrackedVideo
+
+DB
+
+
+def main (
+        video_codes:    list[str],
+) -> None:
+    for code in video_codes:
+        tv = TrackedVideo ()
+        tv.code = code
+        tv.upsert ()
+
+
+if __name__ == '__main__':
+    main (sys.argv[1:])

update_db.py

@@ -8,115 +8,157 @@
 from __future__ import annotations
 
 import json
-import os
+import logging
 import random
 import string
 import time
 import unicodedata
-from datetime import datetime, timedelta
+from datetime import date, datetime, timedelta
 from typing import Any, TypedDict, cast
 
 import jaconv
 import requests
-from eloquent import DatabaseManager, Model
 
-from db.models import Comment, Tag, User, Video, VideoHistory, VideoTag
+from db.config import DB
+from db.models import (Comment,
+                       Tag,
+                       TrackedVideo,
+                       User,
+                       Video,
+                       VideoHistory,
+                       VideoTag)
+
+logger = logging.getLogger (__name__)
+logging.basicConfig (
+        level = logging.INFO,
+        format = '%(asctime)s %(levelname)s %(message)s')
 
 
 def main (
 ) -> None:
-    config: dict[str, DbConfig] = { 'mysql': { 'driver':    'mysql',
-                                               'host':      'localhost',
-                                               'database':  'nizika_nico',
-                                               'user':      os.environ['MYSQL_USER'],
-                                               'password':  os.environ['MYSQL_PASS'],
-                                               'prefix':    '' } }
-    db = DatabaseManager (config)
-    Model.set_connection_resolver (db)
-
     now = datetime.now ()
+    today = now.date ()
 
-    api_data = search_nico_by_tags (['伊地知ニジカ', 'ぼざろクリーチャーシリーズ'])
+    search_result = search_nico_by_tags (['伊地知ニジカ',
-    update_tables (api_data, now)
+                                          'ぼざろクリーチャーシリーズ',
+                                          'ぼざろクリーチャーシリーズ外伝'])
+    comments_by_video_code = fetch_comments_by_video_code (search_result['videos'])
+
+    context: UpdateContext = { 'api_data':               search_result['videos'],
+                               'comments_by_video_code': comments_by_video_code,
+                               'deletable':              search_result['is_complete'] }
+
+    connection = DB.connection ()
+    connection.begin_transaction ()
+    try:
+        update_tables (context, now, today)
+        connection.commit ()
+    except Exception:
+        connection.rollback ()
+        raise
 
 
 def update_tables (
-        api_data:   list[VideoResult],
+        context:  UpdateContext,
         now:      datetime,
+        today:    date,
 ) -> None:
     alive_video_codes: list[str] = []
 
-    for datum in api_data:
+    for datum in context['api_data']:
-        tag_names: list[str] = datum['tags'].split ()
+        tag_names = datum['tags'].split ()
+        normalised_tag_names = {normalise (tag_name) for tag_name in tag_names}
+
+        user: User | None = None
+        if datum['userId'] is not None:
+            user = User.where ('code', str (datum['userId'])).first ()
+            if user is None:
+                user = User ()
+                user.code = str (datum['userId'])
+                user.save ()
+
         video = Video ()
         video.code = datum['contentId']
+        video.user_id = user.id if user else None
         video.title = datum['title']
         video.description = datum['description'] or ''
         video.uploaded_at = datetime.fromisoformat (datum['startTime'])
         video.deleted_at = None
         video.upsert ()
         alive_video_codes.append (video.code)
+
         video_history = VideoHistory ()
         video_history.video_id = video.id
-        video_history.fetched_at = now
+        video_history.fetched_at = today
         video_history.views_count = datum['viewCounter']
         video_history.upsert ()
+
         video_tags = [video_tag for video_tag in video.video_tags
-                                if video_tag.untagged_at is not None]
+                                if video_tag.untagged_at is None]
-        tag: Tag | None
-        video_tag: VideoTag | None
         for video_tag in video_tags:
             tag = video_tag.tag
-            if (tag is not None
+            if tag is None:
-                    and (normalise (tag.name) not in map (normalise, tag_names))):
+                continue
-                video_tag.untagged_at = now
+            if normalise (tag.name) in normalised_tag_names:
+                continue
+            video_tag.untagged_at = today
             video_tag.save ()
+
         for tag_name in tag_names:
             tag = Tag.where ('name', tag_name).first ()
             if tag is None:
                 tag = Tag ()
                 tag.name = tag_name
                 tag.save ()
+
             video_tag = (VideoTag.where ('video_id', video.id)
                                  .where ('tag_id', tag.id)
-                                 .where_null ('untagged_at')
                                  .first ())
             if video_tag is None:
                 video_tag = VideoTag ()
                 video_tag.video_id = video.id
                 video_tag.tag_id = tag.id
-                video_tag.tagged_at = now
+
+            video_tag.tagged_at = getattr (video_tag, 'tagged_at', None) or today
             video_tag.untagged_at = None
-                video_tag.save ()
+            video_tag.upsert ()
-        for com in fetch_comments (video.code):
+
+        for com in context['comments_by_video_code'].get (video.code, []):
             user = User.where ('code', com['userId']).first ()
             if user is None:
                 user = User ()
                 user.code = com['userId']
                 user.save ()
+
             comment = Comment ()
             comment.video_id = video.id
             comment.comment_no = com['no']
             comment.user_id = user.id
             comment.content = com['body']
             comment.posted_at = datetime.fromisoformat (com['postedAt'])
-            comment.nico_count = com['nicoruCount']
+            comment.nico_count = com.get ('nicoruCount', 0)
-            comment.vpos_ms = com['vposMs']
+            comment.vpos_ms = com.get ('vposMs', 0)
             comment.upsert ()
 
-    # 削除動画
+    if not context['deletable']:
+        logger.warning ('skip soft-delete because the latest fetch was incomplete')
+        return
+
+    if not alive_video_codes:
+        logger.warning ('skip soft-delete because no alive videos were fetched')
+        return
+
     videos = (Video.where_not_in ('code', alive_video_codes)
                    .where_null ('deleted_at')
                    .get ())
     for video in videos:
-        if video.code not in alive_video_codes:
         video.deleted_at = now
         video.save ()
 
 
-def fetch_comments (
+def fetch_video_data (
         video_code: str,
-) -> list[CommentResult]:
+) -> dict[str, Any]:
     time.sleep (1.2)
 
     headers = { 'X-Frontend-Id':      '6',
@@ -128,15 +170,38 @@ def fetch_comments (
             + '_'
             + str (random.randrange (10 ** 12, 10 ** 13)))
 
-    url = (f"https://www.nicovideo.jp/api/watch/v3_guest/{ video_code }"
+    url = (f'https://www.nicovideo.jp/api/watch/v3_guest/{ video_code }'
-            + f"?actionTrackId={ action_track_id }")
+           + f'?actionTrackId={ action_track_id }')
 
-    res = requests.post (url, headers = headers, timeout = 60).json ()
+    return requests.post (url, headers = headers, timeout = 60).json ()
 
+
+def fetch_comments_by_video_code (
+        videos: list[VideoResult],
+) -> dict[str, list[CommentResult]]:
+    comments_by_video_code: dict[str, list[CommentResult]] = {}
+
+    for video in videos:
+        video_code = video['contentId']
         try:
-        nv_comment = res['data']['comment']['nvComment']
+            comments_by_video_code[video_code] = fetch_comments (video_code)
-    except KeyError:
+        except (KeyError,
-        return []
+                TypeError,
+                ValueError,
+                requests.RequestException) as exc:
+            logger.warning ('failed to fetch comments: %s (%s)', video_code, exc)
+            comments_by_video_code[video_code] = []
+
+    return comments_by_video_code
+
+
+def fetch_comments (
+        video_code: str,
+) -> list[CommentResult]:
+    video_data = fetch_video_data (video_code)
+    nv_comment = (video_data.get ('data', {})
+                           .get ('comment', {})
+                           .get ('nvComment'))
     if nv_comment is None:
         return []
 
@@ -150,32 +215,63 @@ def fetch_comments (
 
     url = nv_comment['server'] + '/v1/threads'
 
-    res = (requests.post (url, json.dumps (params),
+    response = requests.post (url,
+                              json = params,
                               headers = headers,
                               timeout = 60)
-            .json ())
+    response.raise_for_status ()
+    res = response.json ()
 
-    try:
+    return select_comments_from_threads (res)
-        return res['data']['threads'][1]['comments']
+
-    except (IndexError, KeyError):
+
+def select_comments_from_threads (
+        response: dict[str, Any],
+) -> list[CommentResult]:
+    threads = response.get ('data', {}).get ('threads', [])
+    if not isinstance (threads, list):
         return []
 
+    main_comments: list[CommentResult] = []
+    fallback_comments: list[CommentResult] = []
 
-def search_nico_by_tag (
+    for thread in threads:
-        tag:    str,
+        comments = thread.get ('comments') if isinstance (thread, dict) else None
-) -> list[VideoResult]:
+        if not isinstance (comments, list):
-    return search_nico_by_tags ([tag])
+            continue
+
+        casted_comments = cast (list[CommentResult], comments)
+        if len (casted_comments) > len (fallback_comments):
+            fallback_comments = casted_comments
+
+        fork = str (thread.get ('fork', '')).lower ()
+        label = str (thread.get ('label', '')).lower ()
+        thread_id = str (thread.get ('id', '')).lower ()
+        if fork == 'main' or 'main' in label or 'main' in thread_id:
+            main_comments = casted_comments
+
+    selected_comments = main_comments or fallback_comments
+    deduped_comments: dict[int, CommentResult] = {}
+    for comment in selected_comments:
+        comment_no = comment.get ('no')
+        if not isinstance (comment_no, int):
+            continue
+        deduped_comments[comment_no] = comment
+
+    return [deduped_comments[comment_no]
+            for comment_no in sorted (deduped_comments)]
 
 
 def search_nico_by_tags (
         tags:   list[str],
-) -> list[VideoResult]:
+) -> SearchNicoResult:
     today = datetime.now ()
 
     url = ('https://snapshot.search.nicovideo.jp'
            + '/api/v2/snapshot/video/contents/search')
 
-    result_data: list[VideoResult] = []
+    result_by_video_code: dict[str, VideoResult] = {}
+    is_complete = True
     to = datetime (2022, 12, 3)
     while to <= today:
         time.sleep (1.2)
@@ -194,6 +290,7 @@ def search_nico_by_tags (
                                      'targets':    'tagsExact',
                                      '_sort':      '-viewCounter',
                                      'fields':     ('contentId,'
+                                                    'userId,'
                                                     'title,'
                                                     'tags,'
                                                     'description,'
@@ -201,23 +298,67 @@ def search_nico_by_tags (
                                                     'startTime'),
                                      '_limit':     100,
                                      'jsonFilter': query_filter }
-        res = requests.get (url, params = cast (dict[str, int | str], params), timeout = 60).json ()
         try:
-            result_data += res['data']
+            response = requests.get (
-        except KeyError:
+                    url,
-            pass
+                    params = cast (dict[str, int | str], params),
+                    timeout = 60)
+            response.raise_for_status ()
+            res = response.json ()
+            for datum in cast (list[VideoResult], res.get ('data', [])):
+                result_by_video_code[datum['contentId']] = datum
+        except (ValueError, requests.RequestException) as exc:
+            logger.warning ('snapshot fetch failed: %s - %s (%s)',
+                            to.date (),
+                            until.date (),
+                            exc)
+            is_complete = False
         to = until + timedelta (days = 1)
 
-    return result_data
+    for video in TrackedVideo.get ():
+        if video.code in result_by_video_code:
+            continue
+        try:
+            tracked_video = video
+            video_data = fetch_video_data (tracked_video.code)['data']
+            owner = video_data.get ('owner') or {}
+            video_info = video_data['video']
+            result_by_video_code[tracked_video.code] = {
+                    'contentId':    tracked_video.code,
+                    'userId':       owner.get ('id'),
+                    'title':        video_info['title'],
+                    'tags':         ' '.join (map (lambda t: t['name'],
+                                                   video_data['tag']['items'])),
+                    'description':  video_info['description'],
+                    'viewCounter':  video_info['count']['view'],
+                    'startTime':    video_info['registeredAt'] }
+        except (KeyError,
+                TypeError,
+                ValueError,
+                requests.RequestException) as exc:
+            logger.warning ('tracked video fetch failed: %s (%s)', video.code, exc)
+            is_complete = False
+
+    return { 'videos':      list (result_by_video_code.values ()),
+             'is_complete': is_complete }
 
 
-class DbConfig (TypedDict):
+def normalise (
-    driver:     str
+        text: str,
-    host:       str
+) -> str:
-    database:   str
+    return jaconv.hira2kata (
-    user:       str
+            unicodedata.normalize ('NFKC', text.strip ())).lower ()
-    password:   str
+
-    prefix:     str
+
+class SearchNicoResult (TypedDict):
+    videos:      list['VideoResult']
+    is_complete: bool
+
+
+class UpdateContext (TypedDict):
+    api_data:                list['VideoResult']
+    comments_by_video_code:  dict[str, list['CommentResult']]
+    deletable:               bool
 
 
 class VideoSearchParam (TypedDict):
@@ -231,6 +372,7 @@ class VideoSearchParam (TypedDict):
 
 class VideoResult (TypedDict):
     contentId:    str
+    userId:       int | None
     title:        str
     tags:         str
     description:  str | None
@@ -239,25 +381,12 @@ class VideoResult (TypedDict):
 
 
 class CommentResult (TypedDict):
-    id:             str
     no:           int
-    vposMs:         int
-    body:           str
-    commands:       list[str]
     userId:       str
-    isPremium:      bool
+    body:         str
-    score:          int
     postedAt:     str
     nicoruCount:  int
-    nicoruId:       Any
+    vposMs:       int
-    source:         str
-    isMyPost:       bool
-
-
-def normalise (
-        s:  str,
-) -> str:
-    return jaconv.hira2kata (unicodedata.normalize ('NFKC', s)).lower ()
 
 
 if __name__ == '__main__':