fix(app-shell,plugin-detail): one sys_activity row to FeedItem constructor; stop dropping author-extended types on the console record page

.changeset/5896-feeditem-single-constructor.md

@@ -0,0 +1,48 @@
+---
+'@object-ui/plugin-detail': minor
+'@object-ui/app-shell': minor
+---
+
+One `sys_activity` row → `FeedItem` constructor, and the console record page
+stops dropping author-extended activity types in silence (objectui#5896).
+
+**The defect.** `RecordDetailView`'s `sys_activity` merge read the shared type
+table (objectui#5878) and then built the `FeedItem` itself, ending in
+`if (!feedType) continue;`. That one `continue` collapsed two different
+situations: a type the table maps to `undefined` **on purpose** (`commented` /
+`mentioned` / `login` / `logout`), and a type the table has never heard of. The
+second is an **author-extended** value — `sys_activity.type` is
+author-extensible (objectstack#11507 direction 4, ruled 2026-08-24), every
+column on that table is `readonly` so objectql never validates a write, and
+ADR-0052 §5b.2 forwards an author's `activityMilestones[].type` into it
+verbatim. So an activity that happened, was written and is queryable had no row
+on the console record page: no placeholder, no empty state, no console message.
+Stored, queryable, invisible — objectui#5840's failure mode reached by another
+route, and one objectui#5969 (PR #6112) had already removed from the block
+side, leaving the two surfaces disagreeing about the same row of the same table.
+
+**The fix is convergence, not a second decision.** `@object-ui/plugin-detail`
+now exports the whole reading — `activityRowToFeedItem`,
+`UNMAPPED_ACTIVITY_FEED_TYPE` and the `resetUnknownActivityTypeWarnings` test
+seam alongside `ACTIVITY_TYPE_TO_FEED_TYPE` — and `RecordDetailView` calls the
+constructor instead of paraphrasing it. Publishing the table alone had left the
+mirror one level up, and it had already drifted three ways: the silent drop, a
+timestamp fallback that could leave `createdAt` `undefined` where the helper
+yields `''`, and a second hand-written system-actor lookup.
+
+**Behaviour change on the console record page** (breaking in the objectui sense,
+shipped `minor` — objectui's `major` tracks `@objectstack`):
+
+- an unmapped `sys_activity.type` now **renders** through the generic
+  `UNMAPPED_ACTIVITY_FEED_TYPE` (`'system'`) presentation instead of vanishing,
+  and is announced once per distinct value on `console.warn` — a **missing
+  decision**, not lost data: the row is visible, what it lacks is its own icon
+  and colour. `FeedItemType` is a closed spec enum, so minting a kind for "we
+  don't know" would be a platform change, not this surface's.
+- `createdAt` is always a string for a row with neither a usable `timestamp`
+  nor a `created_at`.
+
+**Unchanged, deliberately:** the four exclusions still produce no row and no
+warning. They are decisions — comment content lives in `sys_comment`, and
+login/logout are account events rather than record activity — and a warning
+about a decision teaches authors to ignore the channel.

packages/app-shell/src/views/RecordDetailView.activityMapIdentity-5878.test.tsx

@@ -54,13 +54,18 @@
  * stopped delivering rows fails as a broken probe rather than reporting the
  * consumer forked.
  *
- * ## Out of scope, deliberately
+ * ## What used to be out of scope here — closed by objectui#5896
  *
- * Only the TABLE is shared. The row -> `FeedItem` construction around it is
- * still written twice (this merge builds the item inline; the block calls
- * `activityRowToFeedItem`), so the console surface still drops an unmapped
- * type SILENTLY where the block warns once. That is the same class of mirror
- * one level up, and it is not this card's fence.
+ * This card shared only the TABLE; the row -> `FeedItem` construction around
+ * it stayed written twice, so the console surface still dropped an unmapped
+ * type SILENTLY where the block warned once. objectui#5896 converged that too:
+ * the merge now calls `activityRowToFeedItem`, and an unmapped type RENDERS
+ * through `UNMAPPED_ACTIVITY_FEED_TYPE` with one diagnostic. The pins for that
+ * behaviour and for the single constructor live in
+ * `RecordDetailView.activityUnmappedType-5896.test.tsx`; this file keeps
+ * pinning the TABLE's identity, which the convergence preserves — the shared
+ * constructor reads the same shared object, so both legs below still measure a
+ * real read.
  */
 
 import * as React from 'react';
@@ -308,19 +313,32 @@ describe('the previously mapped kinds are unchanged — regression controls', ()
     }
   });
 
-  it('the console surface still shows a mapped row and still drops the rest', async () => {
-    renderWith([
-      activityRow('updated', 0),
-      activityRow('commented', 1),
-      activityRow('zzz_not_an_activity_type', 2),
-    ]);
+  it('a mapped row shows, a DELIBERATE exclusion does not', async () => {
+    // Updated by objectui#5896. This leg used to also assert that
+    // `zzz_not_an_activity_type` produced no row — it pinned the silent drop,
+    // which was the defect, not the contract. Under the objectstack#11507
+    // direction-4 ruling an unmapped type is an author-extended value and
+    // renders through the fallback; that is asserted in
+    // `RecordDetailView.activityUnmappedType-5896.test.tsx`, together with the
+    // diagnostic that accompanies it.
+    //
+    // What survives here is the distinction the table draws and this card is
+    // about: `commented` maps to `undefined` ON PURPOSE (its content lives in
+    // `sys_comment`), and a decision stays a decision.
+    const warn = vi.spyOn(console, 'warn').mockImplementation(() => {});
+    try {
+      renderWith([activityRow('updated', 0), activityRow('commented', 1)]);
 
-    expect(await screen.findByText(summaryFor('updated'))).toBeTruthy();
-    // `commented` is a deliberate exclusion (its content comes from
-    // `sys_comment`); the third is a type nothing maps.
-    await waitFor(() => {
-      expect(screen.queryByText(summaryFor('commented'))).toBeNull();
-      expect(screen.queryByText(summaryFor('zzz_not_an_activity_type'))).toBeNull();
-    });
+      expect(await screen.findByText(summaryFor('updated'))).toBeTruthy();
+      await waitFor(() => {
+        expect(screen.queryByText(summaryFor('commented'))).toBeNull();
+      });
+      // ...and silently: no `[record:activity]` diagnostic for an exclusion.
+      expect(
+        warn.mock.calls.map((c) => String(c[0])).filter((m) => m.includes('[record:activity]')),
+      ).toHaveLength(0);
+    } finally {
+      warn.mockRestore();
+    }
   });
 });