finding(fields): LocationField's THIRD refusal sentence (the #6715 residue arm) is still hard-coded English after the #6755 ruling keyed the other two

[os-sam]

Filed unassigned by the dev seat implementing objectui#6755, as an out-of-scope finding. Recording, not choosing.

Dedup before filing: repo-scoped REST list of all 252 open issues (3 pages, state=open), title+body grepped for 6715 / residue / refusedResidue / is not a number / plain decimals — no match. Control: grepping the same corpus for LocationField returns objectui#6755, so the read was live rather than empty.

What

objectui#6755 ruled (2026-08-29, maintainer) that a widget's own refusal sentence goes through useFieldTranslation + FIELD_DEFAULTS, and locked scope to three sentences: ObjectField's Invalid JSON and LocationField's two refusal sentences.

LocationField now has three. The card was measured on 2026-08-29 06:24 and the ruling written at 14:53; objectui#6715 added a third refusal arm — residue, for a half that is only partly a number (12abc, 34) — whose sentence is built by refusedResidueMessage() in packages/fields/src/widgets/LocationField.tsx:

`Not saved: ${named} ${verb}. Enter plain decimals (example: 30.2741, 120.1551).`

with verb being is not a number / are not numbers and named being latitude "12abc"-style. It is a string literal in the source, exactly like the two the ruling keyed.

Why it is filed rather than folded into the #6755 PR

The ruling says 「范围锁定这 3 句」 and names one same-class instance (objectui#4752) that must NOT ride along. This is a fourth sentence the ruling could not have seen — but reading a scope lock as "except when I find one more" is the dev seat's call to make, not mine. So the #6755 PR keys the three ruled sentences and leaves this one, and no test in that PR pins the residue arm's English text (pinning it would read as endorsement).

The measured consequence of leaving it

All three arms render through the SAME <p> element and the same refusalError state. After #6755 lands, that one line speaks the reader's language when the format or range arm fired and English when the residue arm did — the same shape objectui#4028 was reported for ("four Chinese labels around one English one"), inside a single widget.

Cost if it is ruled in

Mechanically identical to what the #6755 PR already does, one arm wider: one key (fields.location.refusedResidue — an interpolated one: the half names and the quoted text are data), ten pack entries, and it becomes bound by check:i18n-drift like the other three. The English default stays byte-identical, so objectui#6715's own pins and plugin-form's ObjectForm.locationResidue.test.tsx are unaffected.

Not measured: whether the interpolation should carry named/verb as holes or be keyed per arity (_one / _other). That is a real authoring choice — verb is grammar, and a pack whose plural rules differ cannot use an English verb form as a hole — and it is why this is not a two-line copy of the other two keys.

Related

objectui#6755 (the ruling and the PR that keys the other two) · objectui#6715 (added this arm) · objectui#6716 (added the first two) · objectui#4752 (the same class in @object-ui/components, still its own card)

[os-sam]

PM note for triage — I am NOT grading this, but three facts bear on how you grade it

domain:ui execution seat, session session_013hfmP9hoMd3dJwTh85J4yB. ⛔ No label touched, card
left unassigned. Grading is yours; this is the context I hold that you would otherwise have to
re-derive.

1. The scope lock this card respects did not exclude this sentence — it predates it

The ruling says 「范围锁定这 3 句」, and the dev seat was right not to fold a fourth in. But the
lock and the exclusion are not the same thing:

• The ruling explicitly excluded objectui#4752 — a different package, a different widget.
  That is a decision.
• The count three was a measurement of the tree on 2026-08-29 06:24, and the ruling was
  written at 14:53. objectui#6715 added the residue arm after both. Nobody declined to key
  it; nobody could see it.

⇒ this is new terrain, not excluded terrain. That distinction is the whole grading question,
and I would rather state it than have the next reader infer "the maintainer said three, so three
is the answer" from a number that was never a choice.

2. The consequence is measured, and it is worse than "one more English string"

All three arms render through the same <p> element and the same refusalError state. After
PR #6890 lands, that single line speaks the reader's language when the format or range arm fired
and English when the residue arm did — a bilingual element, not a bilingual screen. The filer
names the precedent exactly: this is the shape objectui#4028 was reported for ("four Chinese
labels around one English one"), except compressed into one widget's one line.

Partial translation of a single sentence position reads to a user as a bug, not as a missing
translation.

3. It is genuinely NOT a two-line copy — there is a real authoring choice

The card is right to flag this and I want it weighted, because it is the reason this is not
"just do it": verb is grammar (is not a number / are not numbers). Passing an English
verb form through a {{hole}} hands translators a fragment they cannot inflect around — packs
with different plural rules (ar has dual) cannot produce a correct sentence from it. The
alternatives are per-arity keys or a re-shaped sentence, and named (latitude "12abc") carries
a translatable noun too.

So whoever takes it should be told to measure what this channel already does — do any of the
11 widgets already on useFieldTranslation handle arity, and how — and to stop and report if
the answer turns out to be a design choice rather than a precedent to follow. That is a real
round, not a rider.

What I am doing meanwhile, and why

I am landing PR #6890 rather than holding it. Holding a green, correctly-scoped, ruled PR to
wait on an unruled sibling would invert the lesson this lane keeps relearning — an unlanded PR
blocks everything behind it, and #6890's own scope is not in question. The bilingual window is
real but short at this repo's landing cadence, and it is visible only on a malformed-coordinate
path.

⛔ I am not dispatching this card: it carries no pm:* label, so it is not in my queue, and
taking it would be jumping your grading. If you grade it into pm:queue + domain:ui I will
pick it up on the next pass.

---

Generated by Claude Code