feat: add community experience dashboard

[hezyin]

Add a new dashboard that computes 8 key metrics for dev experience in the community. Here's the metric list:

Issue Metrics:

1. Time to Initial Issue Response
2. Issue Resolution Time
3. Issue Response Rate within SLA (user can customize their SLA in the dashboard)
4. Number of Good First Issues (user can customize the issue label that represents good first issues in their repo)

PR Metrics:

1. Time to Initial PR Review
2. PR Resolution Time
3. PR Resolution Rate within SLA (user can customize their SLA in the dashboard)
4. PR Closed W/O Merging Rate

Future work includes differentiating between community issues/PRs vs core team issues/PRs.

Example Screehshot:

[Startrekzky]

Add or comment_rank is null in the last line? Otherwise, the non-responded issue won't be calculated.

with issue_comment_list as(
  select
    i.id as issue_id,
    i.url,
    i.title,
    i.created_date as issue_created_date,
    ic.id as comment_id,
    ic.created_date as comment_date,
    ic.body,
    case when ic.id is not null then rank() over (partition by i.id order by ic.created_date asc) else null end as comment_rank
  from
    lake.issues i
    join lake.board_issues bi on i.id = bi.issue_id
    join lake.boards b on bi.board_id = b.id
    left join lake.issue_comments ic on i.id = ic.issue_id
  where
    date(i.created_date) BETWEEN
      curdate() - INTERVAL DAYOFMONTH(curdate())-1 DAY - INTERVAL 1 month and
      curdate() - INTERVAL DAYOFMONTH(curdate()) DAY
    and b.id in ($repo_id)
)

select
  100 * sum(case when (TIMESTAMPDIFF(MINUTE, issue_created_date,comment_date))/60 < $iir_sla then 1 else null end) / count(*)
from issue_comment_list
where comment_rank = 1 or comment_rank is null

[hezyin]

Updated

[Startrekzky]

Similar to Issue response time, shall we remove and status = 'closed? Otherwise, the opened PR will not be calculated.

select
  100 * sum(case when TIMESTAMPDIFF(Minute, created_date, closed_date) / 1440 < $prrt_sla then 1 else 0 end) / count(*)
from 
	pull_requests pr
where 
  date(created_date) BETWEEN
    curdate() - INTERVAL DAYOFMONTH(curdate())-1 DAY - INTERVAL 1 month and
    curdate() - INTERVAL DAYOFMONTH(curdate()) DAY
	and pr.base_repo_id in ($repo_id)

[hezyin]

Updated. Even though it's not necessary, I added closed_date to the condition to make the query easier to understand.

select
  100 * sum(case when closed_date and TIMESTAMPDIFF(Minute, created_date, closed_date) / 1440 < $prrt_sla then 1 else 0 end) / count(*)
from 
	pull_requests pr
where 
  date(created_date) BETWEEN
    curdate() - INTERVAL DAYOFMONTH(curdate())-1 DAY - INTERVAL 1 month and
    curdate() - INTERVAL DAYOFMONTH(curdate()) DAY
	and pr.base_repo_id in ($repo_id)

[Startrekzky]

Adding closed_date to the condition will affect the result of count(*) and the final result

[Startrekzky]

it's safer to count(distinct i.id) then count(*)

[hezyin]

count(distinct.id) is indeed more error tolerant, but it's also more prone to hiding data duplication bugs. Maybe we should go with count(*) here?

[Startrekzky]

LGTM