fix(spec): state the row-cap guard ElementDataSourceGate implements, and record where the per-kind view limit actually lands (#19228)

[os-warren]

Part of #19228

Clause-②: no

⚠️ Read this first — the card's causal claim did not reproduce

#19228 says an APPLIED .default(100) makes the react tier's 「fills it only when unset」 arm
structurally unreachable. Measured end to end at the objectui pin this repo builds against
(.objectui-sha = 87af769e9, 2026-09-21T06:30–06:40Z), it does not:

• The arm's guard reads the element-face key — readLimit(base, 'limit') where base is the
  object-kanban / object-timeline node. That key is .optional() with no applied default,
  so an author's silence is still silence at parse time.
• The applied default #19226 added sits on the view-face per-kind blocks
  (GalleryConfigSchema / KanbanConfigSchema / TimelineConfigSchema). Different schemas.
• Nothing carries the view-face default across: composed.limit = config.limit ?? savedViewLimit(view),
  and savedViewLimit reads view.pagination.pageSize, else a flat limit on the runtime saved-view
  RECORD — ElementSavedView is an untyped string-keyed record from the adapter's own listViews()
  (core/src/data-scope/element-data-source.ts:96, fallback at :237-241), ⛔ NOT a document
  ListViewSchema accepts: a flat limit on a view document is refused unrecognized_keys. A third
  face, named here so it is not read as the author face — and never a per-kind block. ListView's
  baseProps (plugin-list/src/ListView.tsx:2840-2865) carries no limit on any branch.

⛔ So no .default() moves in this PR, and no precedence is picked — both fences the dispatch set
hold. What DID measure as a real 说明书脱节 is a different, smaller thing, and that is what is repaired.

What is repaired

1. The describe's 「only when unset」 is right for its face — and now names the face.
The branch is if (!fromView || !isUsableRowLimit(authored))
(react/src/element-data-source/ElementDataSourceGate.tsx:316-331).
⚠️ RETRACTED — rounds 1 and 3 both got this backwards, and round 4 measured it. On THIS face the
displacement arm is unreachable: the key is z.number().int().positive().optional() and the gate's
isUsableRowLimit admits integers above zero, so every value the contract accepts is usable and the
「set but not usable」 cell is EMPTY (measured per value: 1 / 25 / 100 / 5000 accepted and usable;
0 / -1 / 2.5 / '100' / null refused by both). ⇒ 「fills this key only when it is unset」 — which round 1
retired as too narrow — was never too narrow; it was right for its face, and is restored. The view half
is pagination.pageSize ALONE on this face, per the bullet above. The describe states the rule that is
reachable from the accept set; the unreachable arm is recorded in the docblock, ⛔ not in author prose.

2. The per-kind view limit is recorded as WHERE IT LANDS. ⚠️ An earlier revision of this section said 「read by nobody」; that was measured with an instrument blind to spreads and is retracted — see the CORRECTION sections below. Kanban and timeline DO read it; gallery alone does not.
Two instruments, because the first one's answer was wrong. PROPERTY-ACCESS over all 8,228 files tracked at the pin — probe \.(kanban|gallery|timeline)(\?)?\.limit\b → 0 lines; control \.(kanban|gallery|timeline)(\?)?\.(groupByField|scale|coverField)\b → 13 lines across 6 files. ⚠️ That zero is WRONG as a claim about readers: a spread carries the key without spelling it. SPREADS, by the predicate 「a spread whose target is the object literal an adapter returns as the node」, return four, and they overturn it. Recorded in two places, on two different faces: on rowLimitKey in view.zod.ts (the VIEW face — flattened onto the node and read, on both routes for kanban, on the plugin-view route for timeline, and read by nobody for gallery), and on the object-timeline door in component.zod.ts (the ELEMENT face — a node's own nested timeline.limit, which is the one read on no route at all). ⛔ Recorded, not repaired: what should read it is the card's open half.

3. ObjectTimelinePropsSchema carries two authorable row caps on one strictObject.
Taking TimelineConfigSchema by reference imported #19226's limit beside the flat limit. The
published member list for that door omitted the new key; it now names it and says it is inert here.
The stale limit read anchor (:234, :254, at the old pin 53ded82b) is corrected to :407 at
the current pin; ⛔ the other anchors in that list are NOT swept — they remain the 53ded82b
readings the docblock header names, and the note says so.

The three relayed consumer readings, reproduced first-hand

Comment 5747444798 marked these 「复核的读数,⛔ 本席未复现」. All three reproduce by shape; two
of the three file:line pairs are wrong at this pin, which is why shape was the instrument.

relayed	reproduced at pin 87af769e9
ListView.tsx:2493 forwards the timeline block nested	✓ shape — the nested forward is :3084 inside case 'timeline': (:3062-3117). :2493 is a useMemo dependency array here (:2496). Only startDateField / endDateField / titleField / groupByField / colorField / scale are hoisted flat; limit is not
ObjectTimeline.tsx:234 reads only the flat schema.limit	✓ shape — the one $top is :407, resolveRowLimit(schema.limit, DEFAULT_TIMELINE_LIMIT); also :279 / :281 / :442. No nested read anywhere
nothing reads timeline.limit	⚠️ RETRACTED — ✓ 0 on that instrument, but see the CORRECTION sections: four flat spreads land the key where kanban/timeline read it

The zero's instrument and its reach radius. git grep over the 8,228 files tracked at commit
87af769e9 — a literal text search of property-access spellings. Lit control on the identical
receiver alternation (schema.timeline / timelineConfig / resolvedTimeline / mergedTimeline /
tCfg): .startDateField → 7 hits, .limit → 0. ⚠️ A first attempt at this probe returned
0 for the control too — a broken bracket expression — and is recorded here because a dead instrument
and a clean result are indistinguishable without one.

One known target deliberately outside the radius: a computed read (cfg[key] with key from a
variable) or a wholesale spread of the config into a query builder is invisible to a text search.
⚠️ CORRECTION (seat, after the at-tier review at head 85ad2898). That hole was named and
then closed by ASSERTION rather than by an instrument, and the assertion was false. A spread carries
a key without ever spelling it, so this radius owed a SECOND instrument with its own lit control; it
was not built. Built afterwards, it returns four flat spreads, not one: ListView.tsx:2979 and
ObjectView.tsx:1638 (...restKanban), ObjectView.tsx:1697 (...(viewOptions.gallery || {}))
and ObjectView.tsx:1725 (...(viewOptions.timeline || {})). Neither restKanban destructure
strips limit. ⭐ Naming a limitation is not discharging it. Also outside the radius: untracked
build output, and objectui HEAD rather than the pin.

Verification

• pnpm --filter @objectstack/spec build && typecheck && test — 509 files / 14,890 tests pass
  (post-merge run on this head).
• Consumer package: @objectstack/lint — the only package outside packages/spec naming any of
  the touched symbols (in a comment, not a runtime read; lit control: ListViewSchema /
  PageComponentSchema fire across 8+ files, so the instrument discriminates). Its first run was a
  Failed to resolve entry cascade from unbuilt workspace deps; after
  pnpm --filter '@objectstack/lint^...' build it reads 106 files / 4,034 tests pass, 5 skipped.
• pnpm --filter @objectstack/spec check:generated — all 15 generated artifacts up to date
  (content/docs/references/ui/component.mdx regenerated via --fix; three table rows changed in component.mdx and nine in view.mdx,
  nothing else).
• pnpm lint (repo-wide eslint . --no-inline-config) — exit 0, the whole union, no narrowing.
• check:react-declaration-parity run exactly as lint.yml runs it
  (MANIFEST="$PWD/sdui.manifest.json" … --baseline react-declaration-parity.baseline.json --strict)
  — exit 0, no new declaration divergence. ⚠️ Run WITHOUT the baseline it exits 1 on 126
  pre-existing divergences; that invocation is not the gate.
• Derived gate families (scripts/pm/dispatch-gates.mjs --commands, reconciled with --ran):
  108 derived, 100 measured green, 8 NOT MEASURED — every one of the 8 a PREREQUISITE NOT MET
  exit 3 (unbuilt sibling packages; check-plugin-teardown-shape --self-test needs an unshallow
  clone). ⛔ Reported as what they are, never as a pass. None of the 8 reads packages/spec/src/ui/.
• origin/main merged through scripts/pm/os-regen-merge.sh; no regeneration debt, and main's
  newest entry (filter-between-field-reference-endpoint-refused) survives at the same 2-file count
  on both sides.

⛔ Refused / not done

• ⛔ No precedence picked. Which of the per-kind view limit, a view's pagination.pageSize and
  a component's flat limit wins is untouched, in prose and in code.
• ⛔ .default(100) not moved, demoted or removed on any of the three view configs.
• ⛔ Nothing written in objectui — read-only at the pin, as dispatched.
• ⛔ The 14 other read anchors in the object-timeline docblock were NOT re-swept at the current pin.

Acceptance notes

• ⚠️ File-surface deviation, declared. The claim's declared surface named
  packages/spec/src/ui/component.zod.ts. Two files fall outside it: packages/spec/src/ui/view.zod.ts
  (the applied default the card is about lives in rowLimitKey there, not in component.zod.ts —
  the claim and the dispatch were both written against a stale location) and
  content/docs/references/ui/component.mdx (the mandatory regeneration from a .describe() change;
  check:docs reds without it — the claim anticipated regeneration but named
  packages/spec/json-schema/, which this diff leaves untouched). Both are inside the seat's measured
  serial-constraint clearance: zero open PRs hold any path under packages/spec/src/ui/.
• ⚠️ Part of, not Fixes. This PR delivers the 说明书 half triage queued; the card's own two contract questions stay open, so line 1 is deliberately Part of #19228 and merging this does not close the card.
• noted, not filed: the object-timeline docblock header claims every read anchor was taken at pin
  53ded82b while .objectui-sha is 87af769e9; only the limit anchor is re-read here. Carrier: a
  future object-timeline card, or the seat that next re-pins that block.
• noted, not filed: spec: declare an author-settable row ceiling for gallery (and kanban) view configs — the protocol lacks the knob objectui#7390 was ruled to read (principle: 协议不正确的先改协议) #17393's own pin text argues an authorable ceiling on the non-grid four 「would be
  surface no renderer reads」 — the measurement above says the ceiling it DID add is exactly that.
  The pin is still correct about the four; the irony is a reading, not a defect.

⚠️ CORRECTION — the per-kind view limit is NOT inert

Seat, after the at-tier review at head 85ad2898, re-measured first-hand at the pin. This PR's
earlier claim of zero read points was taken with an instrument blind to spreads. Four flat spreads
carry the key onto the generated node (ListView.tsx:2979, ObjectView.tsx:1638 / :1697 /
:1725), and ObjectKanban.tsx / ObjectTimeline.tsx read the resulting flat limit.

⛔ ADR-0049 enforce-or-remove is therefore NOT a live option for kanban or timeline — retiring
that key would break two adapters that actively forward it. It IS the honest reading for gallery
alone, whose renderer contains no limit at all (0 occurrences, against a lit control that fires
on the same file).

⚠️ One distinction this correction itself owes: the $top those keys would govern is not issued on
either route today, because both hosts hand the child its rows as a data prop and the child
short-circuits its own fetch. So 「governs no query on these two routes」 is true; 「read by nobody」
is not.

⭐ For the decision box: through those same spreads an APPLIED default puts an authored-LOOKING flat
limit: 100 on every node built from a spec-parsed view — a value no author wrote.

Cross-repo, ⛔ not acted on

objectui#7390's acceptance face is where this gets consumed. ⚠️ Corrected: kanban and timeline have
a read point today; gallery is the one that owes one. Stated here for the seat to file in
objectui's queue with a Blocked-by: line; ⛔ nothing was written there.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

3 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• the SDK route bridge reached 60 of 215 client-bound route-ledger rows — the other 155 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 155: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 100 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 136 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 37ed9ae04b7b656114009378fe62cd186ab3b168 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 304229705818df682866ab8d570fca5a1b152cda — the merge of head b922105e1629344f85c6f03519b9ff8bb64cbe86 into base 37ed9ae04b7b656114009378fe62cd186ab3b168, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 304229705818df682866ab8d570fca5a1b152cda && git checkout 304229705818df682866ab8d570fca5a1b152cda
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 37ed9ae04b7b656114009378fe62cd186ab3b168 b922105e1629344f85c6f03519b9ff8bb64cbe86 && git checkout -B drift-repro 37ed9ae04b7b656114009378fe62cd186ab3b168 && git merge --no-ff b922105e1629344f85c6f03519b9ff8bb64cbe86

node scripts/docs-audit/affected-docs.mjs --json 37ed9ae04b7b656114009378fe62cd186ab3b168

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.