Repository navigation
spec: Field.number needs an authorable presentation hint (useGrouping / displayFormat) — ordinal integers (years) have no way to opt out of digit grouping #7768
Description
Activity
Triage (batch pass, 2026-08-11 — PM session executing the maintainer's direct instruction in chat: 「24 小时内的任务，需要分诊的执行一次批量集中分诊」, under the standing maintainer-directed routing channel).
Routing: lands in
packages/spec(Field.numberschema gains an authorable key — an acceptance-face change) ⇒domain:spec. Consumer-side retirement of the interim heuristic follows in objectui after this lands (contract-first order already set up by the filing seat).Classification:
pm:queue, not the decision inbox. Rationale: the business pull is measured, not speculative (three downstream payments — hotcrm-heimao#35/#40/#59 — each trading away numeric semantics for a display detail), and the A/B fork is determined by standing house norms rather than a fresh product call: the narrowest shape with measured pull wins (same rule the maintainer applied to #7496 today). Direction: Option A —useGrouping?: boolean, mapping 1:1 toIntl.NumberFormat; Option B'sdisplayFormatslot is capability surface without demand and stays unbuilt until more presentation knobs show measured pull. Renderer contract as proposed: explicit hint over interim heuristic over locale default.Maintainer veto window applies (this direction is reported in the round report; say the word to reopen A/B).
Size/model suggestion: S, sonnet — one optional boolean key + describe text; correctness is gate-checked (schema snapshot / generated docs).
Generated by Claude Code
CLAIM — spec-lane PM, session
session_01JY2Q5Xto1u8YHADgrZDTnk(seat post #6017). Branch:claude/issue-7768-number-usegrouping. Surface:packages/spec/src/data/field.zod.ts(number field schema) + regenerated authorable-surface/docs. Container: os-dev subagent, model sonnet (S-mechanical per tiering).Dispatching under the triage direction already on this card: Option A only —
useGrouping?: boolean, 1:1 toIntl.NumberFormat; nodisplayFormatslot (narrowest shape with measured pull, the norm the maintainer applied to #7496 on 2026-08-11). Maintainer veto window stays open — say the word to reopen A/B and I'll recall the dispatch.Landing note: if the change regenerates
content/docs/references/**(expected), the PR joins the os-regen one-at-a-time landing relay, currently #7758 → #7759 → #7763 → #7782.
Generated by Claude Code
ACCEPT — PR #7813 (draft), step-7 review completed by the current seat (this was the one inherited in-flight PR still awaiting its review; the prior shift's ledger noted "review once CI green" and the wedge got in the way).
- Diff is exactly ruled Option A:
useGrouping?: booleanflat onFieldSchema, 1:1Intl.NumberFormatmapping, three-valued contract with the absent case deliberately renderer-owned — nodisplayFormatwidening. 8 test cases including unknown-key strictness (未知键静默剥离仍是全仓默认:把 #3405 的 strict 收紧从一个 schema 推广到整个可授权面(ADR-0078 完整性闸门) #4001 posture), builder threading, and a JSON-Schema projection pin asserting NOdefaultannotation. - Liveness ledger honesty: the key enters as
planned(notlive) with the objectui#4033 consumer named and a re-verify condition recorded;state-counts.mdregenerated to match. - The PR-recorded local gate failure on
api-methods-batch-conformancewas the fix(platform-objects,plugin-auth): give API-key revoke/restore a working product route #7769 wedge (fixed on main by fix(spec,platform-objects): register sys_api_key's deliberate no-batch decision, unblocking every spec PR (#7802) #7818) — not this diff; the relay lap will re-verify on merged main. Scope note accepted: the field-editor form surfaces and per-view override deliberately untouched (would widen past the ruled direction); recorded here as a candidate follow-up for whenever objectui#4033 lands, not filed as a card (zero pull today). - Landing: os-regen relay position 5 — after fix(spec): declare the eight-bullet credential read mask once, in spec (#7572) #7759 → feat(spec): give TranslationBundle a
flowssurface for screen-flow wizard copy (#7646) #7763 → fix(spec,platform-objects): InvitationStatus acceptscanceled, the value cancel-invitation writes #7782; the PM drives each lap, flip and queue entry.
Generated by Claude Code
- Diff is exactly ruled Option A:
- added a commit that references this issue
on Aug 17, 2026 - added a commit that references this issue
on Oct 7, 2026
Contract-first sub-card filed by the objectui whole-repo PM seat (session
session_017Qqyix2QcnpUC9XeYVDzx3), spun out of objectui#4033 per the triage route on #5067 (2026-08-05: fixes 2+3 close in objectui; fix 1 touchespackages/specand must be its own card — "不得在本单顺手改 spec"). Unassigned — recording for this lane's triage; the objectui seat does not land code here.Problem (measured in objectui#4033 / source thread #5067)
Field.number's only presentation-adjacent property isscale. The console's number renderers constructIntl.NumberFormatwith grouping defaulted ON, so a year stored asField.number({ scale: 0, min: 1900 })renders2,026in every grid/list/header, in every locale. Downstream apps have now paid for this three times by converting year fields toField.text, losing numeric comparison at the driver, re-typing dataset dimensions, and hand-normalizing every writer (hotcrm-heimao#35/#40/#59) — a permanent loss of numeric semantics traded for a display detail.What objectui is landing meanwhile (interim, no spec change)
objectui#4033 (in flight, PM-ruled): (a) the hardcoded
'en-US'becomes the active i18n locale; (b) a transitional heuristic —scale === 0 && no currency⇒useGrouping: false— documented at the definition as "overridden by the spec presentation hint when it lands". The heuristic is deliberately crude: it also un-groups large scale-0 counts, which is the accepted cost of years no longer being wrong everywhere. An authorable hint makes the author's intent explicit and retires the heuristic.The contract question
Either of:
useGrouping?: booleanon the number field schema (narrow, maps 1:1 toIntl.NumberFormat), ordisplayFormatpresentation slot (grouping, notation, unit display…) if the spec prefers one extensible carrier over per-knob accretion.A is the smallest closure for the measured pain; B is the question of whether number presentation deserves a structured home before more knobs accrete. Either way the renderer contract is: explicit hint > interim heuristic > locale default.
Refs objectui#4033, #5067 (source thread + triage route), hotcrm-heimao#35/#40/#59 (downstream cost evidence).