Repository navigation
feat: prw dashboard, a live table of your PRs grouped by task - #25
Conversation
`prw dashboard` (alias `prw top`) shows your pull requests across every repository on GitHub and Bitbucket Cloud as a table refreshed every minute, like top for PRs. It finds them itself: GitHub's search (yours, to review, recently merged), and on Bitbucket one request per workspace for yours and a reviewer query per recently active repository (every 10 minutes) for the ones to review. - internal/dashboard: rows are tasks — the issue key in the branch or title (PROJ-123, overridable with issue_pattern), else the source branch without cherry-pick / backport decorations — so a change and its cherry-picks share a row, across repositories. Columns are target branches in promotion order (dev → staging/qa → rc/release → main/master), merged by tier so naming differences line up; stacked PRs' targets share an "other" column. Cells show the PR (an OSC 8 hyperlink) and its most pressing state. group_by and branch_order override the heuristics. - Request budget: a PR's CI is only re-read when its head changes, while it runs, or every 15 minutes; Bitbucket merge checks are reused for 30 minutes while the PR is unchanged; enrichment runs 4 at a time. Clients count their requests and the footer shows them per hour. - Bitbucket client: CurrentUser, Workspaces, UserPullRequests (per workspace, partial responses incl. participants), RecentRepos, ReviewPullRequests, MergeStateWithin. PRs carry source/target branches on both services. - Flags: --review, --merged --since 24h, --mine=false, --interval, --once, --json, --group-by, --width, --no-links. Config: group_by, issue_pattern, branch_order, bitbucket_workspaces, bitbucket_user. - Terminal width and ANSI escapes via the standard library only (Windows console API, Unix ioctl). Verified live: the owner's GitHub PRs (open, merged, stacked, a fresh CI failure) and public Bitbucket PRs of an Atlassian user across a workspace. The Bitbucket /user and /user/workspaces calls need a real token to verify. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9296510518
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| for { | ||
| items, warnings := collector.Refresh(ctx, filter) |
There was a problem hiding this comment.
When the dashboard runs live with --merged, MergedSince is calculated only at startup and the same filter is reused for every refresh. PRs therefore never age out of the selected window: a dashboard left running for several days still includes merges from its first day while labeling them as merged in the last 24 hours. Recompute the cutoff from the selected duration before each refresh and update the view's filter accordingly.
Useful? React with 👍 / 👎.
| if q.state == StateMerged && it.UpdatedAt.Before(f.MergedSince) { | ||
| continue |
There was a problem hiding this comment.
For a cutoff partway through a UTC day, the date-only search also retrieves PRs merged earlier that day. Comparing UpdatedAt does not exclude those PRs if they received a comment or another update after the cutoff, so --merged --since can show PRs merged outside the requested window even on a single refresh. Read the PR's merged_at value and use that for the exact cutoff comparison instead.
Useful? React with 👍 / 👎.
| func (c *Collector) storeCI(it Item) { | ||
| c.mu.Lock() | ||
| defer c.mu.Unlock() | ||
| c.ci[it.Key()] = ciSnapshot{sha: it.HeadSHA, item: it, readAt: c.Now(), running: it.CI == "pending" || it.CI == ""} |
There was a problem hiding this comment.
When one check has failed but others are still running, github.SummarizeCI returns failure or error because those states take precedence over pending. This snapshot consequently marks the entire suite as finished and suppresses CI reads for 15 minutes, even though CIDone < CITotal. Completion and successful retries on the same head can remain invisible throughout that interval. Include unfinished check counts in the running predicate so partially running suites continue to refresh.
Useful? React with 👍 / 👎.
prw dashboard(aliasprw top): your PRs across every repository on GitHub and Bitbucket Cloud, as a live table refreshed every minute — liketopfor PRs. No need to watch them first: prw finds them.Built for workflows where a change is opened on
devand cherry-picked torcandmasteracross many repositories — but generic:PROJ-123, configurableissue_pattern), else the source branch without cherry-pick/backport decorations;--group-by issue|branch|title|none.dev|develop,rc|release|release/*,main|master) so different naming across repos lines up; stacked PRs' targets share anothercolumn;branch_orderoverrides.✓Napprovals).--review,--merged --since 24h,--mine=false;--once,--json.Discovery & request budget
author:@me/workspaces/{ws}/pullrequests/{user}, partial response incl. participants)is:merged merged:>=state=MERGED,q=updated_on > …review-requested:@meCI is re-read only when a PR's head changes, while it runs, or every 15 min; Bitbucket merge checks are reused 30 min while a PR is unchanged; 4 requests in flight. Clients now count their requests; the footer shows them per hour.
Atlassian removed the cross-workspace
/pullrequests/{user}endpoint (verified: 404), hence per-workspace discovery (bitbucket_workspaces, default: all of the token owner's).Test plan
go test ./... -race,go vet ./..., cross-compiles for linux/darwin/windows/freebsd (CGO_ENABLED=0)other), rendering (states, emoji/hyperlink width, alignment), collector CI policy (finished cached, running/new head/15 min re-read, failed read keeps last state), sources against fakes, command view / JSON / flags/userand/user/workspaceswith a real token (owner to try)🤖 Generated with Claude Code