feat(types): name InputShorthandSchema and UiCalendarSchema on the root barrel (objectui#9406) (#9525)

Fixes #9406

Implements decision batch #133 item 2, letter (a) (maintainer
2026-09-14, recorded on the card). Nothing here is a judgement call of
this branch's: the ruling names the file, the act, the bump level and
the pin.

## The exact published delta

Two names join **one** list: the root barrel
`packages/types/src/index.ts`, on its existing `export type { ... } from
'./form.js';` clause, from the module `packages/types/src/form.ts`.

| name | declared in | was on `@object-ui/types/form` | was on
`@object-ui/types/zod` | root `@object-ui/types` |
|---|---|---|---|---|
| `InputShorthandSchema` | `form.ts` | yes | yes (objectui#9067 / PR
#9405) | **added here** |
| `UiCalendarSchema` | `form.ts` | yes | yes (objectui#9067 / PR #9405)
| **added here** |

**No other export moves.** Measured on the published artifact rather
than asserted: `packages/types` was built at the branch point and at
this head, and `dist/index.d.ts` differs by exactly one line, which
differs by exactly these two names inserted between `CalendarSchema,`
and `FieldValidationRules,`. Nothing else in that file changed, in
either direction.

The dispatch cautioned that this package's emit is not byte-reproducible
(measured on PR objectui#9517), so the comparison carries a
**same-source control**: two consecutive clean builds of byte-identical
source, compared before any normalisation. For `dist/index.d.ts` on this
tree they came back **byte-identical** (raw differing lines 0, sorted
differing lines 0), so the delta above is a reading and not emit noise.
Reported as measured: the instability that card saw on another emitted
file did not appear on this one, which the control establishes rather
than assumes.

The source diff is **one additive hunk**, 16 insertions and 0 deletions
on `index.ts` — two names and the comment that says why they are there.
The comment does not reach the published artifact (the emitted clause is
one comma-joined line), so the published surface change is exactly the
two names.

## Why it mattered more than a missing name

From the root entry point the compiler did not merely fail, it
**suggested a different component**. Reproduced live on this branch by
removing the two lines again and running the package's own `tsc -p
tsconfig.test.json`:

```
error TS2724: '"../index"' has no exported member named 'InputShorthandSchema'. Did you mean 'InputOTPSchema'?
error TS2724: '"../index"' has no exported member named 'UiCalendarSchema'. Did you mean 'CalendarSchema'?
```

`ui:calendar` and `calendar` resolve to different renderers, and
`InputOTPSchema` is not an input at all. An author who accepts the
suggestion writes an import that compiles and means something else.

## The pin derives the gap; it does not assert a pair

`packages/types/src/__tests__/form-barrel-mirror-9406.test.ts`. On every
run it reads `form.ts`'s own export list and the root barrel's
`./form.js` re-export list, and the verdict is the **set difference** —
so the declaration that lands in `form.ts` next month and is forgotten
fails it. A pin asserting "these two names are present" would have
passed that day, which is this class reopening a third time.

- Names deliberately left off carry a **ledger row with the reason**
(`ABSENT_BY_DECISION`), names nobody has ruled on carry one saying so
(`ABSENT_PENDING_DECISION`) — the two-ledger shape
`arm-named-export-8784.test.ts` settled for the zod barrel.
- **Rows cannot rot:** a row whose name `form.ts` no longer declares, or
which has since reached the barrel, fails the file and asks to be
deleted.
- Both populations carry controls, so a dark instrument fails instead of
passing: exactly one `./form.js` clause, the barrel list is a subset of
what the module declares, the four names the card measured all read
non-zero, and a star re-export from either module under the pin fails
rather than silently shortening the denominator.
- `SemanticElementSchema` / `HtmlElementSchema` are recorded as the
ruling asked, with the reason and the ruling that made them so
(objectui#9067, decision batch #121 item 5), and **checked** — they are
declared in `layout.ts`, not `form.ts`, so they sit outside this pin's
population by construction, and the block fails the day either reaches
the root barrel. They are not added here.

## One finding, ledgered and reported rather than repaired

Running the derived pin over this list surfaced two more names in the
same state: **`CommandItem` and `CommandGroup`** (the element types of
`CommandSchema.groups` and of its `items`), reachable only through the
`/form` subpath while `CommandSchema` sits on the root list. That is
objectui#7697's shape exactly. The ruling authorises exactly two names
and says "additive, no other line moves", so they are ledgered as
**undecided** with the reason and handed to the dispatching seat for
their own card. This PR does not move them.

⚠️ A related correction to the filing card's framing, from the disk at
this branch point: the card reads "re-exports 28 of the 30 interfaces
`src/form.ts` exports", which conflates two populations. `form.ts`
exports 30 interfaces plus 2 type aliases; the barrel's list carried 28
names, of which 26 are interfaces. So the gap was **four** names, not
two — the two this ruling names, plus the pair above. The card's own
two-zero reading is confirmed unchanged: both names read 0 on the barrel
while `CalendarSchema` and `InputSchema` read non-zero in the same pass.

## Checks run before pushing

| check | how | result |
|---|---|---|
| the new pin | `pnpm exec vitest run
packages/types/src/__tests__/form-barrel-mirror-9406.test.ts` | 16
passed |
| the pin, with the change removed | same command, `index.ts` taken back
to the branch point | 1 failed, naming both missing names |
| the type-level leg, with the change removed | `tsc -p
tsconfig.test.json` | exit 2, the two TS2724 lines quoted above |
| package type-check | `pnpm --filter @object-ui/types run type-check`
(all three `tsc` legs) | exit 0 |
| package suite | `pnpm exec vitest run packages/types/` | 190 files,
4345 tests passed |
| package lint | `pnpm --filter @object-ui/types run lint` | exit 0, 0
errors |
| changeset declaration | `check-changeset-presence.mjs`, `pnpm
changeset:check`, `pnpm check:changeset-claims` | exit 0 |
| line citations | `pnpm check:new-line-citations` | exit 0 |
| control bytes | `pnpm check:control-bytes` and a direct sweep of the
diff | exit 0, no hits |

Both ablation legs restored from `HEAD` and verified by blob hash
against the `HEAD` blob, with `git diff HEAD` empty afterwards.
Repository-scale runs (`Lint`, `Type Check`, the test shards, `Build &
E2E`, `Build Docs`) are CI's.

## Changeset

`.changeset/9406-types-root-barrel-two-names.md`, **`minor`** on
`@object-ui/types` — the level the ruling sets, on the ground that this
is a published surface gaining two names, not on the size of the diff.

## For the Clause-2 contract review

The reviewed surface is one clause in one file. The published delta is
the two names in the table above and nothing else, established against
the built `dist/index.d.ts` with a same-source control. Both names were
already published on two other entry points of this same package, so the
third is aligned rather than widened; both subpaths are untouched and
every previously published name is still published. Draft, not enqueued,
no auto-merge, no labels written by this branch.

---
_Generated by [Claude
Code](https://claude.ai/code/session_011QreXiyMEqKLN4U5daMPVa)_

Co-authored-by: Claude <noreply@anthropic.com>

Author: os-tesla

.changeset/9406-types-root-barrel-two-names.md

@@ -0,0 +1,34 @@
+---
+"@object-ui/types": minor
+---
+
+`InputShorthandSchema` and `UiCalendarSchema` are now named exports of `@object-ui/types`
+itself, not only of `@object-ui/types/form` and `@object-ui/types/zod` (objectui#9406).
+
+Before this release, `import type { UiCalendarSchema } from '@object-ui/types'` read
+`TS2724` while `CalendarSchema` — its neighbour on the very same re-export list — resolved
+in the same compilation pass. The failure was not merely "the name is not there": the
+compiler answered with a suggestion naming a DIFFERENT component (`CalendarSchema` for
+`UiCalendarSchema`, `InputOTPSchema` for `InputShorthandSchema`), and `ui:calendar` and
+`calendar` resolve to different renderers. An author who takes the suggestion writes an
+import that compiles and means something else, which is worse than one that fails.
+
+The change is purely additive — two names join the root barrel's existing named re-export
+list from `./form.js`, nothing is removed, renamed or narrowed, and both subpaths keep
+working. Both names were already published on two other entry points of this same package,
+so this aligns the third rather than widening the surface; it is the route objectui#7697
+took for `ComboboxOption`, on the same list, for the same shape.
+
+`form-barrel-mirror-9406.test.ts` keeps it closed, and it is DERIVED rather than a pair of
+presence assertions: it reads `form.ts`'s export list and the root barrel's `./form.js`
+re-export list on every run and names whatever is in the first and not the second. A pin
+asserting "these two names are present" would pass on the day the next declaration lands in
+`form.ts` and is forgotten, which is this class reopening a third time. Names deliberately
+left off the list get a ledger row carrying the reason instead, and a row goes red once its
+name reaches the barrel or stops being declared.
+
+Running that pin against this list surfaced two more names in the same state,
+`CommandItem` and `CommandGroup` — the element types of `CommandSchema.groups` and of its
+`items`, reachable only through the `/form` subpath while `CommandSchema` sits on the root
+list. objectui#9406's ruling authorises exactly two names, so they are ledgered as
+undecided rather than moved, and reported for their own card.