spec: `minLength` still has the pre-#11566 defect pair — validated as no more than a number, authorable on every field type

[claude]

Found while implementing #11566 (maxLength tightening). Filed unassigned — scope discipline kept it out of that PR, and the correct shape is a judgment call, not mechanical extrapolation.

What

packages/spec/src/data/field.zod.ts still declares:

minLength: z.number().optional().describe('Min character length'),

on the base field schema — the exact defect pair #11566 was ruled on for maxLength:

1. No shape validation. minLength: -5 and minLength: 2.5 parse cleanly. (Unlike maxLength, minLength: 0 arguably has a defined meaning — "no minimum" — so the lower bound is a genuine decision: .min(0) vs .min(1), which is why this was not folded into spec: maxLength is authorable on every field type and validated as no more than a number — maxLength: 0 and maxLength: 12.5 parse cleanly #11566 as a mechanical rider.)
2. Authorable on every type. boolean, lookup, autonumber, … accept it; the write-time validator (packages/objectql/src/validation/record-validator.ts) enforces it only on the same ten bounded-string types as maxLength (now exported as BOUNDED_STRING_FIELD_TYPES).
3. The two authoring forms still carry their pre-convergence lists for minLength: field.form.ts shows it for text/textarea/email only, object.form.ts for nine types (no code). spec: maxLength is authorable on every field type and validated as no more than a number — maxLength: 0 and maxLength: 12.5 parse cleanly #11566 deliberately converged only the maxLength rows (the ruled scope).

Direction (for triage — not prescribing)

The #11566 ruling (comment on that card, 2026-08-24) is the natural template: shape z.number().int().min(0 or 1) + applicability BOUNDED_STRING_FIELD_TYPES + both form rows aligned. The one open question is the lower bound (0 = explicit no-op allowed, 1 = a no-op declaration is refused).

---

Generated by Claude Code

[claude]

Triage: → needs-user-decision + domain:spec, Bug — the defect pair is #11566's twin (no shape validation + authorable everywhere), but the filer is right that it is not a mechanical rider: minLength: 0 has a defined meaning ("no minimum"), so the lower bound is a genuine contract fork #11566's ruling does not cover. Everything else inherits the #11566 template (shape + BOUNDED_STRING_FIELD_TYPES applicability + both form rows converged).

① 项目长远合理性:与 maxLength 同模板收敛(shape + 适用类型 + 两张表单行)是缩小特例;唯一分叉在下界 —— 允许 minLength: 0 就是允许一个永真的空声明留在契约面上。
② 实际业务拉动:今天作者可以在 boolean/lookup 上声明 minLength、可以写 -5/2.5,运行时静默不兑现 —— 撞上的是「声明了却不生效」,与 maxLength 同一类客户困惑。
③ 防 AI 犯错:minLength: 0 是 AI 批量生成元数据时最容易顺手写出的无效声明;响亮拒绝(min(1))让它在编写期就炸,放行(min(0))则永远静默 —— 声明即强制的立场倾向拒绝。
④ 创业阶段不扩散:两案成本几乎相同,差别只在拒绝集边界;选 min(1) 不新增任何维护义务,选 min(0) 则要给「合法但永真」的值写一行解释。
推荐:B = z.number().int().min(1)(0 视为无声明,响亮拒绝,与 maxLength 同姿态);A = .min(0)(显式允许无操作声明)。回一个字母即可。
置信缺口:未 sweep 存量语料/生成器今天是否有 minLength: 0 的实际产出 —— 若有,B 需先带清扫面;裁前建议让 dev 先跑一遍语料 grep 作为前提验证。

---

Generated by Claude Code

[os-support-ai]

Maintainer ruling — B: z.number().int().min(1); zero is refused, the #11566 template applies in full

Source: maintainer, 2026-08-25, live PM chat (decision-inbox batch 8 review, session session_01KWRU3s15AJz7PGW7a7wdCh), verbatim: 「同意」 — accepting the batch's presented recommendation for this card: B.

• minLength converges on the spec: maxLength is authorable on every field type and validated as no more than a number — maxLength: 0 and maxLength: 12.5 parse cleanly #11566 template exactly: shape z.number().int().min(1), applicability restricted to BOUNDED_STRING_FIELD_TYPES, both authoring forms' rows aligned (field.form.ts, object.form.ts).
• The lower bound is 1: minLength: 0 is refused loudly — "no minimum" is expressed by omitting the key, not by declaring a vacuous truth. Same posture as maxLength; a permanently-true declaration is exactly the noise an AI metadata author mass-produces, and the refusal surfaces it at authoring time.
• Pre-step from the stated confidence gap: grep the corpus and generators for any existing minLength: 0 (and negative/fractional) output before landing. In-repo occurrences ride the PR as cleanup; if a shipped, customer-facing generator emits minLength: 0, ⛔ stop and fork back here rather than breaking it silently.

Discipline: Clause-② yes (accept/reject behaviour changes — previously-parsing declarations now refuse); CONTRACT_REVIEW_TIER; spec seat.

State: needs-user-decision → pm:queue in the same write; domain:spec stands.

---

Generated by Claude Code

[os-warren]

Interaction note from #11875 (signature/qrcode joined the bounded-string family, maintainer ruling 2026-08-25): the write-time validator's string branch now reads the spec's BOUNDED_STRING_FIELD_TYPES directly instead of a hand-copied type list (record-validator.ts), and the set itself gained signature / qrcode. Two consequences for this card:

• the applicability half of a minLength tightening no longer needs a validator-side list edit — converging minLength on the same set is now one superRefine in field.zod.ts, and enforcement follows automatically (the branch already applies min_length to every set member, the two new types included);
• whatever set this card's ruling lands on, it is twelve members today, not the ten quoted in the card body.

The two authoring-form minLength rows (3 types in field.form.ts, 9 in object.form.ts) remain unaligned and are untouched by #11875 — deliberately, since the ruled scope there covered maxLength only.

---

Generated by Claude Code