Skip to content

feat: newcomers found no way in from the flat issues list - #182

Merged
jamesconwaydev merged 2 commits into
mainfrom
dashboards
Nov 22, 2026
Merged

jamesconwaydev merged 2 commits into
mainfrom
dashboards

Conversation

@jamesconwaydev

Copy link
Copy Markdown
Collaborator

Fixes #145.

Problem. Someone new opens the issues page and gets a flat list of 40-odd issues. The six marked good first issue or help wanted are mixed in with milestone tracking issues, and nothing says which are already taken or where each milestone stands.

Change.

  • scripts/dashboards.py builds three dashboards and writes each into the open issue whose body starts with <!-- parsec-dashboard:NAME -->:
    • 🙋 Help wanted across Parsec #179 🙋 Help wanted across Parsec: every open good first issue and help wanted issue, free or taken. An issue counts as taken when it's assigned, has an open linked pull request, or someone outside the core team has said they're on it.
    • 🔀 Community pull requests #180 🔀 Community pull requests: everyone except jamesconwaydev and the bots, sorted into waiting on us, waiting on the author, failing CI, needs a rebase and drafts, longest wait first.
    • 🗺️ Roadmap status #181 🗺️ Roadmap status: each open milestone with its closed and open counts, a progress bar and the next five open issues, then closed milestones one line each. The dashboards sit in M15 themselves, so the script leaves them out of its counts.
  • .github/workflows/dashboards.yml runs it at 23 minutes past each hour, and on demand. The job gets issues: write and nothing else, and checks out without persisting credentials.
  • Titles are escaped so a @name or #12 in one never notifies anyone or cross-links from every snapshot. Bodies shrink their tables to stay under GitHub's 65,536-character limit.

The three issues are created, pinned and filled already, from a local run against this repository.

Tests. scripts/dashboards_test.py, no network:

  • Escaping: fails if a handle or issue number in a title stays bare.
  • Buckets: fails if a maintainer reply doesn't hand the turn to the author, an author reply doesn't hand it back, an outsider's comment counts as ours, or draft, rebase and CI stop taking precedence in that order.
  • Claims: fails if an assignee, a linked pull request or a "I'd like to work on this" comment doesn't mark an issue taken, or a core-team comment does.
  • HelpBoard: fails if an issue with both labels shows up twice, the free count is wrong, or a 400-issue body goes over the limit.
  • Roadmap: fails if a dashboard issue is counted in its milestone, an assignee goes missing, the "more" link is wrong, or closed milestones aren't listed last.

python3 -m unittest dashboards_test passes (14 tests), and actionlint and zizmor report nothing on the workflow.

Three pinned issues now refresh every hour: help wanted, community pull
requests and roadmap status. scripts/dashboards.py finds each one by the
marker in its body and rewrites it.

@foothillsdigital foothillsdigital left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The boards read well, and the 14 tests pass (counted with -v). One thing has to change before this runs every hour with issues: write.

Anyone can take over a dashboard. existing() takes the first open issue whose body starts with the marker, and GET /repos/.../issues lists the newest first. So if a stranger opens an issue whose body starts with <!-- parsec-dashboard:roadmap -->, the next run writes the roadmap into their issue, and the pinned #181 goes stale. I checked it by stubbing gh with two open issues, the stranger's newer #200 and the owner's #181:

{'roadmap': 200}

The fix is a line in existing(): accept an issue only if a maintainer opened it, for example i.get("author_association") in MAINTAINER_ROLES, or i["user"]["login"] in CORE_TEAM plus github-actions[bot] if the job ever creates one. Add a test that a newer issue from someone else with the same marker is ignored.

Not blocking, but worth a look while you're in there:

  • The claim regex matches more than claims. "I would like to know why this fails" and "the claim in the README is wrong" both match, so a question would mark an issue as taken. Requiring a verb after i'd like to / i would like to (work|take|try|help|pick|fix) and dropping the bare claim would cut most of that.
  • CORE_TEAM only lists you, so my PRs (and any other maintainer's) will show up on the community board, and a maintainer's "I'll take this" counts as a community claim. If that's what you want, fine. Otherwise use authorAssociation there too, the way bucket already does.

existing() took the newest open issue carrying a marker, so a stranger's
issue with the roadmap marker would have been written to instead of the
pinned one. Only issues a maintainer or this workflow opened count now.

Also: a question like "I would like to know why" no longer reads as a
claim, and maintainers other than the core team are left off the
community boards, judged by authorAssociation as bucket() already did.
@jamesconwaydev

Copy link
Copy Markdown
Collaborator Author

Good catch, thank you. That one would have been easy to miss for a long time.

  • The takeover. existing() now only counts an issue opened by a maintainer (author_association in MAINTAINER_ROLES, or the core team) or by github-actions[bot], for the day the workflow makes one. Existing.test_a_strangers_issue_with_the_marker_is_ignored is your case: a newer feat(bench): parsec-xml had no numbers against libxml2 or anyone #200 from a stranger and our 🗺️ Roadmap status #181, newest first. It finds 181.
  • Claims. i'd like to / i would like to / i'd love to now need one of work|take|try|help|pick|fix after them, and the bare claim is gone. test_questions_are_not_claims has your two sentences plus "I'd love to see this land".
  • Maintainers. There's a maintainer(login, association) helper now, used by outside(), bucket() and claim(), so your PRs stay off the community board and your "I'll take this" isn't counted as a community claim.

18 tests pass, and a dry run against the live repo still finds #179, #180 and #181.

@foothillsdigital foothillsdigital left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved. My stub from the first review (a stranger's newer #200 with the marker, then #181) now finds 181, and both of my sentences no longer count as claims. 18 tests pass, counted with -v, and CI is green.

@jamesconwaydev
jamesconwaydev added this pull request to the merge queue Nov 22, 2026
Merged via the queue into main with commit 3f3dfd4 Nov 22, 2026
33 checks passed
@jamesconwaydev
jamesconwaydev deleted the dashboards branch November 22, 2026 01:30
jamesconwaydev added a commit to conn3xorg/conn3x that referenced this pull request Nov 22, 2026
* fix(dashboards): anyone could take over a dashboard issue

publish() mapped each marker to the first open issue whose body starts
with <!-- conn3x-dashboard:<marker> -->. GitHub lists issues newest
first and anyone can open one, so an issue a stranger opened with the
help-wanted marker would get the next hourly write, from a job with
issues: write, while the pinned #383 went stale. They'd control the
title and labels of an issue the project keeps updating.

Only an issue a maintainer opened, or github-actions[bot], which is who
publish() creates missing boards as, now counts. The lookup is its own
function, existing_boards(), so it can be tested.

From the same review:
- CLAIM matched questions: "I would like to know why", "the claim in
  the README", "I'd love to see this". "I'd like to" and "I'd love to"
  now need a verb of taking the work, and the bare "claim" is gone.
- Other maintainers counted as community: claim(), outside() and
  bucket() now use maintainer(), with authorAssociation asked for in
  the help-wanted and PR queries.

Found in review of Parsec's copy of this script (parsec-html/parsec#182).

Signed-off-by: James Conway <148910373+jamesconwaydev@users.noreply.github.com>

* test(dashboards): the takeover tests sat below unittest.main() and never ran in CI

CI runs python3 scripts/dashboards_test.py, so unittest.main() ran and
exited before Boards and Outside were defined: 14 tests ran, not 18
(Foothills' review on #1250). Move both classes above it.

Signed-off-by: James Conway <148910373+jamesconwaydev@users.noreply.github.com>

---------

Signed-off-by: James Conway <148910373+jamesconwaydev@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

community For people using or helping with the project

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Pinned dashboards on the issues page

2 participants