管理者ダッシュボード レポート参照ガイド

[Discourse]

This is a reference guide for describing how the Admin Dashboard Reports function, the data they’re displaying, the corresponding Data Explorer SQL queries, and where to find the Ruby code for each report.

Required user level: Staff

Discourse contains several built-in admin dashboard Reports that can be useful for exploring stats about a community. To access these reports, you can visit discourse.example.com/admin/dashboard/reports on your site ( or click the Reports link at the top of the dashboard). Note that only staff users will have access to these reports.

Data from all users on a site is included in these reports (including staff activity like visiting admin pages). The only condition that is put on users in the reports is that they are ‘real’ users, which is used to exclude the system user from the reports.

Plugins can also add reports to the dashboard with add_report(name, &block).

Ruby models for most reports are located at: discourse/app/models/concerns/reports/. Some reports also reference: discourse/app/models/report.rb

The dashboard-sql topics contain all of the corresponding SQL queries that can be used to generate reports identical to the Admin Dashboard Reports. These queries can be used within the Data Explorer plugin and for Running Data Explorer queries with the Discourse API

Accepted solutions

Displays daily tally of posts marked as solutions.

Ruby code: discourse-solved/plugin.rb at main · discourse/discourse-solved · GitHub

SQL Query: Dashboard Report - Accepted Solutions

Admin Logins

List of admin login times with locations.

Ruby Code: discourse/app/models/concerns/reports/staff_logins.rb

SQL Query: Dashboard Report - Admin Logins

Anonymous

Number of new pageviews by visitors not logged in to an account.

Ruby Code: discourse/app/models/concerns/reports/consolidated_page_views.rb

SQL Query: Dashboard Report - Anonymous

Bookmarks

Number of new topics and posts bookmarked.

Ruby Code: discourse/app/models/concerns/reports/bookmarks.rb

SQL Query: Dashboard Report - Bookmarks

Consolidated API Requests

API usage statistics by date, tracking both regular API requests and user API requests.

Ruby Code: discourse/app/models/concerns/reports/consolidated_api_requests.rb at main · discourse/discourse · GitHub

SQL Query: Dashboard Report - Consolidated API Requests

Consolidated Pageviews

Pageviews for logged in users, anonymous users and crawlers.

Ruby Code: discourse/app/models/concerns/reports/consolidated_page_views.rb

SQL Query: Dashboard Report - Consolidated Pageviews

Consolidated Pageviews with Browser Detection (Experimental)

Pageviews for logged in users, anonymous users, known crawlers and other. This experimental report ensures logged-in/anon requests are coming from real browsers before counting them.

Ruby Code: discourse/app/models/concerns/reports/consolidated_page_views_browser_detection.rb

SQL Query: Dashboard Report - Consolidated Pageviews with Browser Detection

DAU/MAU

Number of members that logged in in the last day divided by number of members that logged in in the last month – returns a % which indicates community ‘stickiness’. Aim for >20%.

Ruby Code: discourse/app/models/concerns/reports/dau_by_mau.rb

SQL Query: Dashboard Report - DAU/MAU

Daily Engaged Users

Number of users that have liked or posted in the last day.

Ruby Code: discourse/app/models/concerns/reports/daily_engaged_users.rb

SQL Query: Dashboard Report - Daily Engaged Users

Emails Sent

Number of new emails sent.

Ruby Code: discourse/app/models/concerns/reports/emails.rb

SQL Query: Dashboard Report - Emails Sent

Flags

Number of new flags.

Ruby Code: discourse/app/models/concerns/reports/flags.rb

SQL Query: Dashboard Report - Flags

Flags Status

List of flags’ statuses including type of flag, poster, flagger, and time to resolution.

Ruby Code: discourse/app/models/concerns/reports/flags_status.rb

SQL Query: Dashboard Report - Flags Status

Likes

Number of new likes.

Ruby Code: discourse/app/models/concerns/reports/likes.rb

SQL Query: Dashboard Report - Likes

Logged In

Number of new pageviews from logged in users.

Ruby Code: discourse/app/controllers/admin/reports_controller.rb#L5

SQL Query: Dashboard Report - Logged In

Moderator Activity

List of moderator activity including flags reviewed, reading time, topics created, posts created, personal messages created, and revisions.

SQL Query: Dashboard Report - Moderator Activity

Moderator Warning

Number of warnings sent by personal messages from moderators.

Ruby Code: discourse/app/models/concerns/reports/moderator_warning_private_messages.rb

SQL Query: Dashboard Report - Moderator Warnings

New Contributors

Number of users who made their first post during this period.

Ruby Code: discourse/app/models/concerns/reports/new_contributors.rb

SQL Query: Dashboard Report - New Contributors

Notify Moderators

Number of times moderators have been privately notified by a flag.

Ruby Code: discourse/app/models/concerns/reports/notify_moderators_private_messages.rb

SQL Query: Dashboard Report - Notify Moderators

Notify User

Number of times users have been privately notified by a flag.

Ruby Code: discourse/app/models/concerns/reports/notify_user_private_messages.rb

SQL Query: Dashboard Report - Notify User

Overall Sentiment

Number of posts classified either positive or negative with the “Sentiment” AI, over a specified period.

Ruby Code: discourse-ai/lib/sentiment/entry_point.rb at main · discourse/discourse-ai · GitHub

SQL Query: Dashboard Report - Overall Sentiment

Pageviews

Number of new pageviews from all visitors. Same as the total for Consolidated Pageviews.

Discourse uses the follow query to determine total pageviews:

SQL Query: Dashboard Report - Consolidated Pageviews

Post Edits

Number of new post edits.

Ruby Code: discourse/app/models/concerns/reports/post_edits.rb

SQL Query: Dashboard Report - Post Edits

Posts

New posts created during the selected time period

Ruby Code: discourse/app/models/concerns/reports/posts.rb

SQL Query: Dashboard Report - Posts

Post Emotion

Number of posts classified by AI with one of following emotions: Sadness, Surprise, Fear, Anger, Joy, Disgust - group by poster trust level, over a specified period.

Ruby Code: discourse-ai/lib/sentiment/entry_point.rb at main · discourse/discourse-ai · GitHub

SQL Query: Dashboard Report - Post Emotion

Reactions

List most recent reactions.

Ruby code: discourse-reactions/plugin.rb at main · discourse/discourse-reactions · GitHub

SQL Query: Dashboard Report - Reactions

Signups

New account registrations for this period.

Ruby Code: discourse/app/models/concerns/reports/signups.rb

SQL Query: Dashboard Report - Signups

Site Traffic

Ruby Code: discourse/app/models/concerns/reports/site_traffic.rb

SQL Query: Dashboard Report - Site Traffic

Suspicious Logins

Details of new logins that differ suspiciously from previous logins.

Ruby Code: discourse/app/models/concerns/reports/suspicious_logins.rb

SQL Query: Dashboard Report - Suspicious Logins

System

Number of personal messages sent automatically by the System.

Ruby Code: discourse/app/models/concerns/reports/system_private_messages.rb

SQL Query: Dashboard Report - System

Time to first response

Average time (in hours) of the first response to new topics.

Ruby Code: discourse/app/models/concerns/reports/time_to_first_response.rb + discourse/discourse/blob/main/app/models/topic.rb#L1799-L1844

SQL Query: Dashboard Report - Time to First Response

Top Ignored / Muted Users

Users who have been muted and/or ignored by many other users.

Ruby Code: discourse/app/models/concerns/reports/top_ignored_users.rb

SQL Query: Dashboard Report - Top Ignored / Muted Users

Top Referred Topics

Topics that have received the most clicks from external sources.

Ruby Code: discourse/app/models/concerns/reports/top_referred_topics.rb

SQL Query: Dashboard Report - Top Referred Topics

Top Referrers

Users listed by number of clicks on links they have shared.

Ruby Code: discourse/app/models/concerns/reports/top_referrers.rb

SQL Query: Dashboard Report - Top Referrers

Top Traffic Sources

External sources that have linked to this site the most.

Ruby Code: discourse/app/models/concerns/reports/top_traffic_sources.rb

SQL Query: Dashboard Report - Top Traffic Sources

Top Uploads

List all uploads by extension, filesize and author.

Ruby Code: discourse/app/models/concerns/reports/top_uploads.rb

SQL Query: Dashboard Report - Top Uploads

Top Users by likes received

Top 10 users who have received likes.

Ruby Code: discourse/app/models/concerns/reports/top_users_by_likes_received.rb

SQL Query: Dashboard Report - Top Users by Likes Received

Top Users by likes received from a user with a lower trust level

Top 10 users in a higher trust level being liked by people in a lower trust level.

Ruby Code: discourse/app/models/concerns/reports/top_users_by_likes_received_from_inferior_trust_level.rb

SQL Query: Dashboard Report - Top Users by Likes Received from a User with a Lower Trust Level

Top Users by likes received from a variety of people

Top 10 users who have had likes from a wide range of people.

Ruby Code: discourse/app/models/concerns/reports/top_users_by_likes_received_from_a_variety_of_people.rb

SQL Query: Dashboard Report - Top Users by Likes Received From a Variety of People

Topics

New topics created during this period.

Ruby Code: discourse/app/models/concerns/reports/topics.rb

SQL Query: Dashboard Report - Topics

Topics with no response

Number of new topics created that did not receive a response.

Ruby Code: discourse/app/models/concerns/reports/topics_with_no_response.rb

SQL Query: Dashboard Report - Topics with No Response

Topic View Stats

Ruby Code: discourse/app/models/concerns/reports/topic_view_stats.rb

SQL Query: Dashboard Report - Topic View Stats

Trending Search Terms

Most popular search terms with their click-through rates.

Ruby Code: discourse/app/models/concerns/reports/trending_search.rb

SQL Query: Dashboard Report - Trending Search Terms

Trust Level growth

Number of users who increased their Trust Level during this period.

The Trust Level Growth report is pulling data from the user_histories table in the Discourse database. Specifically, this report is counting the number of times a user_histories.action is recorded for an increase in a user trust level.

Ruby Code: discourse/app/models/concerns/reports/trust_level_growth.rb

SQL Query: Dashboard Report - Trust Level Growth

Unaccepted policies

This dashboard report identifies topics with policies that have not been accepted by certain users.

Ruby code: discourse-policy/plugin.rb at main · discourse/discourse-policy · GitHub

SQL Query: Dashboard Report - Unaccepted Policies

User Flagging Ratio

List of users ordered by ratio of staff response to their flags (disagreed to agreed).

Ruby Code: discourse/app/models/concerns/reports/user_flagging_ratio.rb

SQL Query: Dashboard Report - User Flagging Ratio

User notes

List most recent user notes.

Ruby code: discourse-user-notes/plugin.rb at main · discourse/discourse-user-notes · GitHub

SQL Query: Dashboard Report - User Notes

User Profile Views

Total new views of user profiles.

Ruby Code: discourse/app/models/concerns/reports/profile_views.rb

SQL Query: Dashboard Report - User Profile Views

User Visits

The total number of logged-in user visits in the forum for the selected time period (today, yesterday, last 7 days, etc).

A User Visit is counted anytime a unique logged in user visits the site, up to once per day. For example, if a user visited a site every day within a week, Discourse would count that as 7 user visits.

Ruby Code: discourse/app/models/concerns/reports/visits.rb

SQL Query: Dashboard Report - User Visits

User Visits (mobile)

Number of unique logged-in users who visited using a mobile device.

Ruby Code: discourse/app/models/concerns/reports/mobile_visits.rb

SQL Query: Dashboard Report - User Visits

User-to-User (excluding replies)

Number of newly initiated personal messages.

Ruby Code: discourse/app/models/concerns/reports/user_to_user_private_messages.rb

SQL Query: Dashboard Report - User-to-User

User-to-User (with replies)

Number of all new personal messages and responses.

Ruby Code: discourse/app/models/concerns/reports/user_to_user_private_messages_with_replies.rb

SQL Query: Dashboard Report - User-to-User

Users per Trust Level

Number of users grouped by trust level.

Ruby Code: discourse/app/models/concerns/reports/users_by_trust_level.rb

SQL Query: Dashboard Report - Users Per Trust Level

Users per Type

Number of users grouped by admin, moderator, suspended, and silenced.

Ruby Code: discourse/app/models/concerns/reports/users_by_type.rb

SQL Query: Dashboard Report - Users Per Type

Web Crawler Pageviews

Total pageviews from web crawlers over time.

Ruby Code: discourse/app/models/report.rb

SQL Query: Dashboard Report - Web Crawler Pageviews

Web Crawler User Agents

List of web crawler user agents, sorted by pageviews.

Ruby Code: discourse/app/models/concerns/reports/web_crawlers.rb

SQL Query: Dashboard Report - Web Crawler User Agents

Last edited by @SaraDev 2025-10-15T20:49:23Z

Last checked by @SaraDev 2025-01-28T21:42:37Z

Check document

Perform check on document:

[pfaffman (Jay Pfaffman)]

@SaraDev

Discourse には、コミュニティに関する統計情報を探索するのに役立つ、いくつかの組み込みの管理者ダッシュボードレポートが含まれています。これらのレポートにアクセスするには、サイトの /admin/dashboard/reports にアクセスしてください。

/admin にリンクが見つかりません。読み方が足りないのでしょうか？これらのレポートはもっと見つけやすい場所にあるべきだと思います。これらのレポートがあることは知っていたと思いますが、探しても見つかりませんでした。

数分で見つけることができましたが、次のようなものを追加すると良いかもしれません。

プラグインは add_report(name, &block) を使用してレポートをダッシュボードに追加できます。

[maiki]

@pfaffman

/admin にはこのリンクが表示されません。

@pfaffman

これはもっと見つけやすいはずです。これらのレポートがここにあることは知っていたと思いますが、探しても見つかりませんでした。

ええ、サイトが最初に作成されたときに、スタッフへのPMで言及すると良いかもしれませんね。

[pfaffman (Jay Pfaffman)]

すみません。どこかで見たことがあると思ったのですが。

人々は物事を読もうとしませんね…しかし、プラグインでそれを行う方法を見つけるためにソースコードを読むことはできますか？

しかし、上記を次のように更新しますか？

@SaraDev

これらのレポートにアクセスするには、サイトの /admin/dashboard/reports にアクセスするか、ダッシュボードの レポート リンクをクリックしてください。

それが本当に私を混乱させた原因だと思います。(しかし、いいえ、言い訳はありません。)

[maiki]

@pfaffman

しかし、上記を更新するかもしれません

トピックをウィキにしました。どうぞ！

[hello-smile6 (9pfs)]

@SaraDev

30%以上にすることを目指してください。

UI（UIは20%を使用しています）と一致しませんが、どちらを使用すべきでしょうか？

[JammyDodger]

よく見つけました。最近20%に更新されました。OPで変更します。

Is there a difference in target retention rate between a support-first forum and discussion-first forum?

Here’s a PR with that change: Longer term, I think a more nuanced approach could be powerful. I’d love for communities to be able to set their own goals (with some guidance like what you shared above about why they might choose one vs. the other) and use the dashboard or other related features to help them track their results vs. what they are aiming for.

[Yotam_Hagay]

@SaraDev

トップトラフィックソース

このサイトに最も多くリンクしている外部リンク。

Rubyコード: discourse/app/models/concerns/reports/top_traffic_sources.rb

@SaraDev さん、このレポートの出力はSQLクエリで取得できますか？共有していただけますか？
ありがとうございます。

[SaraDev (Sara Devlaeminck)]

@Yotam_Hagay

こんにちは @SaraDev このレポートの出力はSQLクエリで取得できますか？共有していただけますか？
よろしくお願いします

はい、トップトラフィックソースの以下のSQLレポートを使用できます。

-- [params]
-- date :start_date = 01/05/2023
-- date :end_date = 03/06/2023

WITH count_links AS (

SELECT COUNT(*) AS clicks,
       ind.name AS domain
FROM incoming_links il
  INNER JOIN posts p ON p.deleted_at ISNULL AND p.id = il.post_id
  INNER JOIN topics t ON t.deleted_at ISNULL AND t.id = p.topic_id
  INNER JOIN incoming_referers ir ON ir.id = il.incoming_referer_id
  INNER JOIN incoming_domains ind ON ind.id = ir.incoming_domain_id
WHERE t.archetype = 'regular'
  AND il.created_at::date BETWEEN :start_date AND :end_date
GROUP BY ind.name
ORDER BY clicks DESC
),

count_topics AS (

SELECT COUNT(DISTINCT p.topic_id) AS topics,
       ind.name AS domain
FROM incoming_links il
INNER JOIN posts p ON p.deleted_at ISNULL AND p.id = il.post_id
INNER JOIN topics t ON t.deleted_at ISNULL AND t.id = p.topic_id
INNER JOIN incoming_referers ir ON ir.id = il.incoming_referer_id
INNER JOIN incoming_domains ind ON ind.id = ir.incoming_domain_id
WHERE t.archetype = 'regular'
  AND il.created_at > (CURRENT_TIMESTAMP - INTERVAL '30 DAYS')
GROUP BY ind.name
)

SELECT cl.domain,
       cl.clicks AS "Clicks",
       ct.topics AS "Topics"
FROM count_links cl
JOIN count_topics ct ON cl.domain = ct.domain
LIMIT 10

このクエリでは、日付パラメータは 日/月/年 の形式の日付を受け入れることに注意してください。

[Yotam_Hagay]

@SaraDev

はい、トップトラフィックソースには次のSQLレポートを使用できます。

@SaraDevさん、クエリの共有ありがとうございます。
このレポートとincoming_linksテーブルに関する、より一般的な質問ですが、これは投稿ページへのトラフィックのみを表しており、フォーラムの全ページへのトラフィックではないですよね？

コンテキスト：フォーラム全体のトラフィックの傾向を分析したいと考えており、トップトラフィックソースレポートからソースごとの全体的なトラフィックを取得できることを期待していました。
しかし、全体的なトラフィックが過去1か月で約272K（ユーザーと匿名）であるのに対し、同じ期間のトラフィックソースレポートの総クリック数は59Kにすぎません。
また、topicsテーブルとpostsテーブルとの内部結合を使用しているため、クリックにpost_idが添付されていない場合はカウントしないことになります。

私の結論を確認していただけますでしょうか。また、incoming_linksテーブルの背後にあるロジックについて少し説明していただけると幸いです。

[Yotam_Hagay]

@SaraDev

投稿

この期間中に作成された新しい投稿

@SaraDev 様
このクエリを実行しましたが、結果が一般タブの「投稿」レポートと正確に一致しません。
例：11月30日：
クエリ = 112 件の投稿
レポート = 120 件の投稿
この差について確認していただけますでしょうか？
よろしくお願いします。

[JammyDodger]

念のためお知らせします @Yotam_Hagay - サラさんがスレッドの作成者ですが、ガイドは全員の責任です 各投稿でメンションする必要はありません。

[Yotam_Hagay]

明確にしてくれてありがとう、@JammyDodger。

他にタグ付けしたり、回答を得るために頼ったりできる人はいますか？

[Yotam_Hagay]

@SaraDev

SQL Time to first response

--[params]
-- date :date_start
-- date :date_end

WITH first_reply AS (
    SELECT 
        p.topic_id, 
        MIN(post_number) post_number, 
        t.created_at
    FROM posts p
    INNER JOIN topics t ON (p.topic_id = t.id)
    WHERE p.deleted_at IS NULL
        AND p.post_number != 1
        AND p.post_type = 1
        AND p.user_id > 0
        AND t.user_id > 0
        AND t.deleted_at IS NULL
        AND t.archetype = 'regular'
        AND t.created_at::date BETWEEN :date_start AND :date_end
    GROUP BY p.topic_id, t.created_at
    ORDER BY 2 DESC
)

SELECT 
    p.topic_id, 
    fr.created_at::date dt_topic_created,
    (p.created_at - fr.created_at) response_time
FROM posts p
INNER JOIN first_reply fr 
    ON fr.topic_id = p.topic_id 
    AND fr.post_number = p.post_number
    AND p.created_at > fr.created_at
ORDER BY response_time

このクエリの結果は、「Time to first response」レポートとはわずかに異なります。
例えば11月8日：
クエリ：93時間
レポート：116時間
どなたかアドバイスをいただけますか？

[JammyDodger]

これらの中には調査に時間がかかるものがあるかもしれませんが、私自身も何が起こっているのか調べてみます（SQLスキルとRubyスキルの間には大きな隔たりがありますが ）。

しかし、この情報をすべて確定させるために、引き続き調査結果を教えてください。

[JammyDodger]

@Yotam_Hagay

@SaraDev さん、このクエリを実行しましたが、結果が一般タブの投稿レポートと正確に一致しません。
例：11月30日
クエリ = 112 件の投稿
レポート = 120 件の投稿
この差を確認していただけますか？
よろしくお願いします。

投稿については、在庫レポートがトピックの投稿とシステムユーザーからの投稿もカウントしていると思いますが、post_typeが1のもの（つまり、ウィスパー、小さなアクション投稿、またはモデレーターアクションではないもの）のみを対象としている可能性があります。

SQLは以下のようになると考えられます。

--[params]
-- date :start_date
-- date :end_date

SELECT
    p.created_at::date AS "Day",
    COUNT(p.id) AS "Count"
FROM posts p
INNER JOIN topics t ON t.id = p.topic_id AND t.deleted_at ISNULL
WHERE p.created_at::date BETWEEN :start_date AND :end_date
    AND p.deleted_at ISNULL
    AND t.archetype = 'regular'
    AND p.post_type = 1
GROUP BY p.created_at::date
ORDER BY 1

これをサイトで実行して、一致するかどうか確認していただけますか？

[Yotam_Hagay]

Jammyさん、ありがとうございます。確認します！
現在、初回応答時間（Time to first response）の分析に取り組んでいるため、こちらの件も確認していただけると幸いです。

[JammyDodger]

SQLバージョンを確認したところ、OPからの返信を除外するために AND p.user_id <> t.user_id が欠けていると思います。それを追加すると、OPと他の誰かからの最初の応答との間の正確な時間が出力されます。

--[params]
-- date :date_start
-- date :date_end

WITH first_reply AS (
    SELECT
        p.topic_id,
        MIN(post_number) post_number,
        t.created_at
    FROM posts p
    INNER JOIN topics t ON (p.topic_id = t.id)
    WHERE p.deleted_at IS NULL
        AND p.user_id <> t.user_id
        AND p.post_number != 1
        AND p.post_type = 1
        AND p.user_id > 0
        AND t.user_id > 0
        AND t.deleted_at IS NULL
        AND t.archetype = 'regular'
        AND t.created_at::date BETWEEN :date_start AND :date_end
    GROUP BY p.topic_id, t.created_at
    ORDER BY 2 DESC
)

SELECT
    p.topic_id,
    fr.created_at::date dt_topic_created,
    (p.created_at - fr.created_at) response_time
FROM posts p
INNER JOIN first_reply fr
    ON fr.topic_id = p.topic_id
    AND fr.post_number = p.post_number
    AND p.created_at > fr.created_at
ORDER BY response_time

また、ストックレポートはSQLの時間の代わりに小数で表示されていると思います。それに合わせるように再度試してみます。

---

AVG を含めるように少し更新しました。これにより、ストックレポートの出力とより似たものになります。

--[params]
-- date :date_start
-- date :date_end

WITH first_reply AS (
    SELECT
        p.topic_id,
        MIN(post_number) post_number,
        t.created_at
    FROM posts p
    INNER JOIN topics t ON p.topic_id = t.id
    WHERE p.deleted_at IS NULL
        AND p.user_id <> t.user_id
        AND p.post_type = 1
        AND p.user_id > 0
        AND t.user_id > 0
        AND t.deleted_at IS NULL
        AND t.archetype = 'regular'
        AND t.created_at::date BETWEEN :date_start AND :date_end
    GROUP BY p.topic_id, t.created_at
)

SELECT
    fr.created_at::date dt_topic_created,
    AVG(p.created_at - fr.created_at) response_time
FROM posts p
INNER JOIN first_reply fr
    ON fr.topic_id = p.topic_id
    AND fr.post_number = p.post_number
    AND p.created_at > fr.created_at
GROUP BY fr.created_at::date
ORDER BY response_time

これで、一方は小数で、もう一方はHH:MMであるということを考慮すれば、ストックレポートと一致するようです。SQLの response_time を小数に変換する方法があると思いますが、HH:MMの方が直感的な方法だと思います。（また、必要ないかもしれない余分な基準がいくつかあると思いますが、安全策として、どちらか確信が持てるまでそのままにしておきました ）

これを実行して、どのように一致するか確認していただけますか？

[Yotam_Hagay]

はい、これで株主報告書の数値と一致するようになりました。ありがとうございます。
1点だけコメントがあります。
以下のAVG関数は、時間が24時間以上の場合（日数部分が欠けていると思われます）、結果が欠落することがわかりました。

AVG(p.created_at - fr.created_at)::time response_time

[JammyDodger]

time へのキャストは悪い選択でしたね。:)\n\n::time を削除すると、より正確な（しかし、見るのが少し大変な）バージョンに戻ります。\n\n上のものも編集しますね。