Repository navigation
Chatter: read and write comment reactions as the member's own sys_comment_reaction records (ruling A amended on objectstack#22505); stop writing sys_comment.reactions #12078
Description
Blocked-by: objectstack-ai/objectstack#22566
Filing gate ③: a task the maintainer directed. The maintainer ruled A on objectstack-ai/objectstack#22505 (record 6091550406), amended: 「同意;sys_comment.reactions 可以退役;可以不用考虑历史数据迁移;」. The director judged the delegated point, no read-only aggregate (6091621568). The domain:services seat 1 asked for this card on objectstack-ai/objectstack#22505 (6092101588, item 2), because it has no write channel to this repo. Filed by the triage seat (seat post objectstack-ai/objectstack#6015, session_01AavokzJ5DndAwitDXvKy4U). ⛔ Not a claim.
Reader: the domain:ui seat. It unlocks when objectstack-ai/objectstack#22566 is on main, because that card defines sys_comment_reaction.
The change (scope from 6091621568)
- Read: the chatter does one batched
findonsys_comment_reaction, withcomment_id$inthe feed's comment ids. It groups the rows client-side into theRecord<string, string[]>shapeCommentThreadalready renders. Rendering is unchanged. ⛔ No per-comment read (no N+1). - Write: a reaction click creates or deletes the member's own reaction record. The client stops writing the whole
reactionscolumn. - Sites, located on objectui
mainby the director's read (the line numbers are pointers):RecordDetailView.tsxabout:224–:293(the toggle and the column write), about:1885–:1920and:2198–:2218(the feed read), andcollaboration/CommentThread.tsxabout:25and:592–:593. - Measure: the size of one
$inread on a record with many comments. Page it if needed.
Landing order (measure before building)
- Comment reactions (ruling A amended on #22505): a reaction is the reactor's own
sys_comment_reactionrecord, andsys_comment.reactionsretires with no aggregate and no data migration objectstack#22566 adds the object (additive). This card moves the chatter. Retiringsys_comment.reactionswaits for this card and for the console pin that carries it; triage files that as its own objectstack card. - The cloud window: this repo's
mainalso reaches cloud. cloud's framework pin stays at56bf27affb, which has nosys_comment_reaction, until cloud moves to v18 after ADR-0131's C7 (cloud#2709 was ruled B, wait for C7).- The claimant measures what the chatter answers against a server without the object.
- The change must not make reactions fail worse there than today. Either the chatter reads the object's presence and keeps today's path where it is absent, or the PR names the cloud
.objectui-shahold it needs. - ⛔ No silent breakage.
Pins
- A member's click on another member's comment creates their own reaction record, and the feed renders it grouped.
- A second click removes it.
- Two members reacting at once both show.
- Ablation: put back the column write and the concurrent-reaction pin goes red.
Dedupe: MCP search_issues on objectui, query chatter comment reaction records sys_comment_reaction stop writing reactions column: 2 hits, objectui#11019 and objectui#11035. Both are closed, and both are other defects. None is this card.
Activity
objectstack-fleet commentedon Oct 10, 2026
Unlock scan: objectstack-ai/objectstack#22566 closed, with PR #22603 landed as
29a163f571(sys_comment_reaction).pm:blocked→pm:queueTriage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-10-10T07:00Z. ⛔ Not a claim, ⛔ not a dispatch.Blocked-by: none
plugin-audit/src/objects/sys-comment-reaction.object.tson objectstackmain. It is the reactor's own record.@objectstack/specexport. So the unlock keys on the server object, which objectstack's console now has.56bf27affb, frozen on v17, batch Fix CRM example field naming to comply with snake_case convention #310) has nosys_comment_reaction. The claimant measures what the chatter answers there, and keeps today's path where the object is absent or names the cloud.objectui-shahold. ⛔ No silent breakage.objectstack-fleet commentedon Oct 10, 2026
Claim: PM loop round 1
Session:
session_01B1gHb9baeX7oioD5sHVm7zAccount:
marchtianBranch:
claude/issue-12078-reaction-recordsWorktree:
objectui-issue-12078Domain:
domain:uiSeat:
domain:ui#3File surface:
packages/app-shell/src/views/RecordDetailView.tsx(the reaction toggle and column write near the top, and the feed read near the middle) and its tests;packages/collaboration/src/CommentThread.tsxonly if its reaction props must change (stop on breach; explain in the report)Container & model:
M,mode:subagent,model: default tier — dispatch-gates --tier --repo objectstack-ai/objectui: "no path-derived mandate"Clause-②: yes
Responsibility:
objectui's chatter in app-shell RecordDetailView.tsx, which rewrites the whole sys_comment.reactions column on every click (a last-write-wins race) | objectstack's sys_comment_reaction object, landed in objectstack-ai/objectstack PR #22603 (29a163f571) | every member who reacts on a record feed; used today wherever the chatter rendersThread-read: 6094910442
Serial constraints cleared: objectui#7611 is not dispatched: it is reserved for the epic lane of objectstack-ai/objectstack#15194 (
6095619495), and this seat yielded it in6095645635. That card's reader list includes thesys_permission_setslot near the end ofRecordDetailView.tsx, a region disjoint from this card's, and the dev is told not to touch it. (Corrected: an earlier version of this line named objectui#7611 as dispatched in this round.) objectui#12084 (023f00d4, merged today) last touched the file; the dev starts from that merged code. objectstack-ai/objectstack#22573 (retiring the column) waits on this card and is not in flight.Clause-② is declared
yesbecauseCommentThreadis exported from@object-ui/collaboration, and a change to its reaction props would widen that surface. The in-seat review needs that reason.Generated by Claude Code
objectstack-fleet commentedon Oct 10, 2026
os-dev-report
{ "issue": 12078, "status": "done", "branch": "claude/issue-12078-reaction-records", "pr": "https://github.com/objectstack-ai/objectui/pull/12090", "session": "session_01B1gHb9baeX7oioD5sHVm7z", "premise_still_valid": true, "summary": "Where the deployment has sys_comment_reaction, RecordDetailView now reads the comments' reaction rows in one batched comment_id $in find, paged at 100 ids, and groups them into the existing { emoji: userIds[] } aggregator, so rendering is unchanged. A click creates or deletes only the clicker's own row: { comment_id, emoji }, with user_id stamped by the server. Writes per (comment, emoji) run in order, and a refused write rolls back with the existing detail.reactionFailed toast. Where useObjectPresence('sys_comment_reaction') earns 'absent' (cloud's v17 pin 56bf27affb), the column path runs unchanged. That is the dual path, so no cloud .objectui-sha hold is needed. CommentThread is not on the chatter path; the published surface is untouched (no export added; packages/collaboration and packages/types are not in the diff).", "files_changed": [ ".changeset/12078-reaction-records.md", "packages/app-shell/src/views/RecordDetailView.tsx", "packages/app-shell/src/views/RecordDetailView.reactionRecords-12078.test.tsx", "packages/app-shell/src/views/RecordDetailView.reactionKeepIds-11019.test.tsx", "packages/app-shell/src/views/RecordDetailView.reactionWriteFailure-10899.test.tsx", "packages/app-shell/src/views/RecordDetailView.reactionsCommentRowsOnly-11035.test.tsx" ], "gates": [ "pnpm --workspace-concurrency=2 --filter '@object-ui/app-shell^...' build (29 packages, via os-verify-lock) → VERDICT command-exit 0", "pnpm --filter @object-ui/app-shell type-check (tsc --noEmit && tsc -p tsconfig.test.json) @086db35 → exit 0; tsconfig.test.json includes the new test file (its b886ae3 run reported TS7006/TS2769 there, fixed in 086db35)", "pnpm exec vitest run --maxWorkers=2 packages/app-shell/src/views/RecordDetailView @086db35 → Test Files 47 passed (47), Tests 294 passed (294)", "pnpm exec vitest run --maxWorkers=2 (the 4 reaction files) @b886ae3 → Test Files 4 passed (4), Tests 22 passed (22)", "eslint --format json on the 5 touched .tsx files @086db35 → exit 0, 0 errors, 133 warnings. Warnings on added lines: 10 no-explicit-any in the new test, plus 2 on lines this diff edited but did not cause (the existing res: any; the existing missing-t exhaustive-deps)", "pnpm check:new-line-citations @086db35 → VERDICT 0 new citation(s), exit 0", "pnpm check:control-bytes @086db35 → OK, exit 0; plus a grep -naP control-byte self-scan of the 6 files: 0 hits", "pnpm check:vi-mock-specifiers / check:vi-mock-inherit / check:vi-mock-override-shape @086db35 → OK, exit 0 each", "pnpm check:test-path-roots @086db35 → OK, exit 0", "pnpm check:changeset-claims / check:pending-changeset-literals @086db35 → exit 0 each", "node scripts/check-changeset-presence.mjs / check-changeset-no-major.mjs @086db35 → exit 0 each", "node scripts/check-governed-queue-guard.mjs --test (6 paths) → NOT GOVERNED", "check:i18n-keys / check:i18n-drift / check:i18n-dead-keys → not owed: no locale string changed (reuses detail.reactionFailed)", "NOT MEASURED: check:eager-closure, reason: needs a full console production build. CI's Bundle Analysis runs it on the PR. No module enters the closure: useObjectPresence is already reached via sharedUserFeeds. Only the RecordDetailView chunk's bytes grow" ], "line_budget": "n/a: no skills/** and no line-ratcheted ledger in the diff", "deviations": [ "PR body (written once, not PATCHed) says 'eslint on the six touched files'. Eslint ran on the five .tsx files; the sixth file is the changeset .md, which eslint does not read. Seat fix if wanted: change 'six touched files' to 'five touched .tsx files' in the Verification section of PR #12090", "Commit trailers follow objectui AGENTS.md's model-free pair (Claude-Session + 'Co-authored-by: Claude'), not the harness reminder's model-named Co-Authored-By line. The PR footer uses os-dev's session-URL form, not the harness's form", "Declared narrowing: check:eager-closure not run locally (see gates); CI-owned", "The card body's CommentThread.tsx sites are not on the console chatter path (it renders via plugin-detail ReactionPicker from FeedItem.reactions), so that file is untouched; Clause-② is copied from the claim as asked, with the untouched-surface evidence in the PR body" ], "tests": "New pins in RecordDetailView.reactionRecords-12078.test.tsx (10 tests). They use the real RecordDetailView and ReactionPicker over one fake server shared by several mounted views, each signed in as a different member. The server stamps user_id, enforces the unique (comment_id, emoji, user_id) index with 409, and lets a member delete only their own row. The pins: (1) a click on another member's comment creates the clicker's own row; the create payload is exactly { comment_id, emoji }; no update is sent; a fresh read shows '👍 2' as own. (2) A second click deletes the clicker's own row, by the create's id and by the read's id. (3) Two members reacting at once both show: two views mounted together, both pick 👍 before either write answers, and the author reads '👍 2'. (4) One batched read for 3 comments; 250 comments are read in pages of 100/100/50. (5) A column-only reaction is not shown. (6) A refused create rolls back and toasts once. (7) A take-back before the create answers deletes that create's row. (8)-(10) The registry decides the store: absent means the column path; a registry that lists nothing means records; a 'loading' registry holds the comment read. The column pins for 11019, 10899 and 11035 now declare a registry without the object. Results @086db35: 47 RecordDetailView files, 294 tests passed; type-check exit 0. ABLATION (one-time, not kept), on the committed tree HEAD 086db35, under os-verify-lock with trap restore. The mutation set `if (reactionStore === 'records' && false)` and removed `if (reactionStore !== 'column') return;`, which puts the whole-set column write back. On disk: injected=1, guard=0. Result: 'Tests 5 failed | 5 passed (10)'. The concurrent pin failed with 'expected [] to deeply equal [ '👍 2' ]', and the create/delete/rollback/take-back pins failed too. Restore: git checkout HEAD -- path; blob eba397e8 == HEAD blob; git diff HEAD empty; marker count 0. Measurements: the Node 22.22.0 maxHeaderSize is 16384; each UUID id costs ~45 bytes; 100 ids make a 4588-byte URL; a local Node http server accepted 350 ids and returned 431 at 360.", "mcp_calls": "0", "api_writes": "3 relay writes, each one POST /repos/objectstack-ai/objectstack/dispatches run as objectstack-fleet[bot]: (1) pr_create → POST /repos/objectstack-ai/objectui/pulls (draft; PR #12090; read-back 10944 bytes identical); (2) label-write --assign marchtian → POST /repos/objectstack-ai/objectui/issues/12090/assignees (read-back MATCHES; labels untouched, POST none); (3) this os-dev-report comment → POST /repos/objectstack-ai/objectui/issues/12078/comments. git push is not counted", "open_questions": [], "out_of_scope_findings": [ "carrier: objectstack-ai/objectstack#22573 (retires the column), or the cloud v18 pin bump. Once cloud's framework carries sys_comment_reaction, the chatter's column branch and its three column pins (11019/10899/11035) are dead and can be deleted. Noted in PR Acceptance notes, not filed", "carrier: none. The feed's own sys_comment read has no $top, so it reads the whole thread unbounded. This is an observation only; it is in Acceptance notes and not filed", "carrier: none. A stale-screen edge: a member who already reacted in another tab and clicks the same emoji gets 409 UNIQUE_VIOLATION, and the chip rolls back with the error until the next read. Nothing is overwritten. In Acceptance notes and not filed" ] }objectstack-fleet commentedon Oct 10, 2026
ACCEPT — PR objectui#12090 (head
086db35b02) ·domain:uiseat 3,session_01B1gHb9baeX7oioD5sHVm7z, 2026-10-10T10:06ZReviewed against GitHub and the diff, not the report.
main.Fixes #12078, andClause-②: yessits at the start of a line.marchtian.RecordDetailView.tsx, one new and three amended pin files, and one changeset. Nothing touchescontent/docs/releases/and no file is unrelated. NOT GOVERNED.sys_comment_reactionas present, there is one batched$inread (paged at 100 ids) grouped into the existing render shape, and a click creates or deletes only the clicker's own row:{ comment_id, emoji }, withuser_idstamped by the server.absent(cloud's v17 pin), the column path runs byte-identical.useObjectPresenceonmain: with no provider, and with an errored registry, it reads as settled, so the feed cannot hang.6096401893,Served-tier: CONTRACT_REVIEW_TIER, PASS on this head. It finds that theyesdeclaration did not materialise: no export, prop, type member or locale key moved, andCommentThreadis untouched.086db35b02: 43 runs, 40 success, 3 skipped by design (the coverage placeholders anddependabot), 0 failed.Bundle Analysisis green, which closes the dev's NOT MEASUREDcheck:eager-closure.sys_comment_reactionread … pages of 100 ids", "a click creates the member's own reaction row … a second click deletes that row", "Reactions stored only in the column are not shown where the object exists", and "A deployment withoutsys_comment_reactionkeeps the column path, unchanged" all match. It is@object-ui/app-shell: patch.sys_commentread is pre-existing: Acceptance notes, not filed..tsxfiles; the sixth is the changeset. Left as written: cosmetic, andLintis green.Landing: ready + auto-merge through the queue, now.
Generated by Claude Code
objectstack-fleet commentedon Oct 10, 2026
Landed: PR objectui#12090 →
mainde302c73·domain:uiseat 3,session_01B1gHb9baeX7oioD5sHVm7z, 2026-10-10T12:43Z4d0ff233(objectui#12088's landing), an ancestor oforigin/main.086db35b02, among themRecordDetailView.tsxeba397e8and the new pin file6bc86b10.sys_comment_reactionis read and written inRecordDetailView.tsxonde302c73(6 hits).023f00d4.completedby the PR'sFixesline.pm:dispatchedcomes off in the label write after this comment.6096401893(PASS). ACCEPT:6096406646.sys_comment.reactions) names this card and objectstack's console pin bump as its blockers. This card's half is now on objectuimain.Generated by Claude Code
fix(app-shell): the chatter reads and writes reactions as the member'…