Handle group messages

[nonumpa]

Fixes #13

Description

Not ready

• (done)Change validCategories from categoryName to categoryId
• (done)Should filter out the category which negative vote > positive vote
• Self-introduction wording
• (done)Should disable job queue rate limit and add concurrency to env variable
• Maybe add a leave group command, it seems people can't kick others in the room

Flow

webhook/index (get group message event) -> GroupHandler(jobQueue) -> processGroupEvent -> groupMessage

Response rule

• Chat-bot self-introduction : Input Hi cofacts or Hi confacts .
• Ignore message : Input length less than 10 won't reply and post to cofacts-api.
• Reply :
  • The article found should be in validCategories and be Identical enough.
    • validCategories: 疾病、醫藥🆕, COVID-19 疫情🆕, 科技、資安、隱私, 保健秘訣、食品安全, 免費訊息詐騙
  • The reply of the article should:
    1. Candidate's positiveFeedbackCount > candidate's negativeFeedbackCount.
    2. Candidate's positiveFeedbackCount > non-rumors' positiveFeedbackCount.
    3. If replyCount > 2, check if rumor count >> non-rumor count (2:1).

Job queue

• jobQueue will process events which replytoken are not expired.
• If the event which jobQueue is going to process is expired, it will be skipped and added to expiredJobQueue.
• expiredJobQueue start only when there's no jobs in jobQueue(drained).
• expiredJobQueue pause when there's jobs adding to jobQueue.

GA

Add two user scope custom dimension

1. Message Source: user | group | room will set every time sends the event.
2. NumberOfGroupMembers: It will only set when bot join group.

Article event in group

Category / Action / Label

• If user trigger chat-bot to introduce self:
  • UserInput / Intro /
• If we found a articles in database that matches the message:
  • UserInput / ArticleSearch / ArticleFound
  • Article / Search / <article id> for each article found
• We don't send event If nothing found in database

To calculate total group the bot joined, we set ga event value 1 as join, -1 as leave

Category / Action / Value

• Join Event
  • 'Group' / 'Join' / 1
• Leave Event
  • 'Group' / 'Leave' / -1

Snapshot shows where to create custom dimension in ga.

[MrOrz]

I'm still reviewing, just send out 2 comments here

[MrOrz]

Thanks for the implementation!

I have added some comments and some hints on GA. Happy Lunar New Year!

[MrOrz]

I am confused by the 4 jobs and the changing processGroupEvent mock. May I know what would happen to the 4 jobs here?

[nonumpa]

有點難描述，我用中文好了 哈哈

這裡要測試 expireQueue 執行時，如果有新的 event 進來，expireQueue 要被暫停，並且執行完 jobQueue 後再繼續，利用監聽 expiredJobQueue 的 active event 加上 concurrency 設成 1，我們可以確保在 expiredJobQueue 執行期間，插入一個沒過期的 job。

1. 因為這裡把 GroupHandler 的 concurrency 設成 1，所以四個 expiredJob 會在 jobQueue 的 process function processJob 一個一個(不保證順序)處理並加進 expiredJobQueue，這裡並不會執行到 processJob 裡面的 processGroupEvent。

2. 當所有(四個) jobQueue 的 job 被執行完之後(onDrained)，expiredJobQueue 會開始(resume)執行。

3. 當第一個 expired job 開始執行時，會同時送出一個 expiredJobQueue 的 job active event。

4. 在收到 expired job active event 的時候加入一個沒過期的 job (到 jobQueue)，並且為了讓它成功，把 processGroupEvent 改成 resolve，在 jobQueue on('completed' 改回來。
   其實這邊我期望是可以 mockImplementationOnce(expired x 4 + success x 1)，但是 processGroupEvent 被執行的順序可能是

   • testExpiredId1 -> (expiredJobQueue active event) -> successJobId -> testExpiredId2 -> ...
   • testExpiredId1 -> (expiredJobQueue active event) testExpiredId2 （幾乎同時）-> successJobId -> ...

   因為我們不能保證 active event 收到的時機，不知道 success mock 要放在第幾個順位才能正確讓 successJobId 是 resolve 的，所以就在收到第一個 expiredJobQueue (job) 的 active event 時把它改成 resolve，在它完成時再改回 reject，這樣就會造成一個問題是：把 processGroupEvent 改成 resolve 時剛好 testExpiredId2 開始了 expiredJobQueue 還沒 pause，所以 testExpiredId2 的結果就是 complete 的(理論上 expiredJobQueue 的 job 要在 processGroupEvent 裡面全被 reject)
   這裡只 expect jobQueue 的 success/fail count，因為 expiredJobQueue 的結果可能每次都不一樣

[MrOrz]

感謝感謝
我在想如果「這裡要測試 expireQueue 執行時，如果有新的 event 進來，expireQueue 要被暫停，並且執行完 jobQueue 後再繼續」

若把下面這兩塊拆成兩個分開的 test case，會不會比較簡單

1. expireQueue 執行時，如果有新的 event 進來，expireQueue 要被暫停 --> 先往 expireQueue 塞東西，再 gh.addJob，然後 expect expireQueue 的 status
2. 執行完 jobQueue 後再繼續 --> 先準備一個暫停的 expireQueue，然後 gh.addJob，然後在該 job 完成之後 expect expireQueue 的 status 是否不是 pause

[coveralls]

Pull Request Test Coverage Report for Build 1270

• 116 of 132 (87.88%) changed or added relevant lines in 8 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.3%) to 86.709%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/webhook/handlers/groupHandler.js	28	32	87.5%
src/webhook/index.js	5	10	50.0%
src/webhook/lineClient.js	0	7	0.0%

Totals
Change from base Build 1240:	0.3%
Covered Lines:	918
Relevant Lines:	1045

---

💛 - Coveralls

[MrOrz]

Seems that most issues are resolved. Things left are

1. documentation regarding the custom dimensions; and if we should use custom metric instead of custom dimension for user count in chat groups
2. excessive comment
3. simplify test case

I think we can discuss on 1 because setting custom metric / dimension cannot be undone in the future. Documentations, comments & tests can wait.

[MrOrz]

We will merge first and just setup 1st custom dimension. Leave 2nd dimension / metric for the future.