Repository navigation
fix(types,components,plugin-form): declare a form field's step (the spec's number) and className; built-in branches merge the class and derive the number step, ObjectForm stops writing 'any' (objectui#12130) - #12166
Merged
Merged
Conversation
…ives its step; ObjectForm stops writing `step` (objectui#12130) The built-in `input` arm and the unregistered-widget `default` arm draw a native number input under `inputType: 'number'`. Each now derives its step from the metadata carrier by the registered `number` widget's rule: an explicit numeric `step` wins, else `10^-scale`, else `'any'`. So `ObjectForm` no longer writes `step` onto the entries it builds. It used to write `'any'` when no `scale` was declared, a value the spec's numeric `FieldSchema.step` refuses. `objectFormNumericStep-9574.test.tsx` passes unmodified on both of its routes. The premise measurement for round 3, committed before any declaration. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
… into its control's classes (objectui#12130) The built-in `input`, `textarea`, `checkbox` and `switch` branches and the unregistered-widget default spread the entry's `className` over their own classes, so an authored class erased the touch-target height and the read-only tint. Each now writes `cn(own, entry.className)` after the spread, as the built-in `select` already did. The effects pins measure it per element class (a built-in control, a registered widget, a section-divider), and a sweep over every type the form draws pins that the class never lands on the field row, its label or a wrapper the host draws, editable or read-only. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
…assName`; the census has no open row left (objectui#12130) `step` is `FieldSchema.step` by reference on both faces, so HTML's `'any'` is refused, on the tolerant face too, which used to strip it. No producer writes it any more: the controls that draw a number derive `'any'` themselves when a field declares neither `step` nor `scale`. `className` is the classes of the element the field's type renders, measured per type through the real `SchemaRenderer` with and without `@object-ui/fields`: the control, a registered widget's own element, or a section-divider's divider, never the field row or a wrapper the host draws. The census moves both rows into DECLARED_BY_12130; its ledger has no open row left. The face pin, the zod coverage list, form.mdx and one changeset follow. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
…ber input, and that no producer writes `'any'` (objectui#12130) Comment-only. The default arm's note said an object-bound number field reaches it when its widget is unregistered; the entries that draw a number input there are the ones that say `inputType: 'number'` and name no registered widget. The `FormField` note said no producer writes `step`; it is `'any'` that no producer writes. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
…bjectui#12130) Brings in objectui#12155's DOM allowlist for the form's built-in branches. Conflicts in renderFieldComponent resolved by keeping both: the allowlist (`toFormControlDomProps` plus `builtinHostAttributes`) and the explicit react-hook-form binding from objectui#12155, then this branch's merged `className` (`cn(own, entry.className)`) and the number input's derived `step` after the spread. `BUILTIN_HOST_ATTRIBUTES`' note said `ObjectForm` writes `step`; since this branch it writes none, so the note now says `FormField` declares `step` and the number input resolves it from the metadata carrier. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
…lared FormField key; round 1's changeset and the 9574 header follow (objectui#12130) Claim amendment 6114408150, the seat's answers to open questions 1-3: - content/docs/plugins/plugin-form.mdx: `className` moves from the "Not a `FormField` key" table to the declared keys; the two-key sentence becomes one key, keeping `defaultValue`; the clause saying nothing in the contract promises a field-level `className` is deleted; the stale key cardinal is dropped. - scripts/__tests__/check-doc-component-types.test.ts: the `className` assertion flips to `toMatch`, so the pin re-derives the page's now positive claim. The `defaultValue` non-member assertion stays. - .changeset/12130-form-field-channel-keys.md: its "Still undeclared" sentence goes, since no OPEN row remains; the new changeset's "Supersedes" bullet goes with it. Prose only. - objectFormNumericStep-9574.test.tsx: header prose only. Every assertion is byte-identical. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
github-actions
Bot
added
documentation
Contributor
|
changeset-claim-re-read
|
…ld key (objectui#12130) Claim amendment 6114675131: the same correction as plugin-form.mdx. `className` moves from the "Not a `FormField` key" list into the declared keys, and the stale key cardinal is dropped. Claude-Session: https://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73 Co-authored-by: Claude <noreply@anthropic.com>
Contributor
✅ Console Performance Budget
The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it. 📦 Bundle Size Report
Size Limits
|
Contributor
Author
Contract reviewServed-tier: Inputs: card #12130 (body, 19 comments), PR #12166 (body, 16 files, net diff ① Derived judgments
② Semver level
③ Boundary flags
Implemented-by: VERDICT: PASS Generated by Claude Code |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #12130
Clause-②: yes (narrowing): declaring
stepandclassNamewidens the strict authoring face by two keys. Typingstepas the spec's number narrows the tolerant face, from acceptingstep: 'any'(stripped) to refusing it atfields[].step.Round 3 of objectui#12130 (
step,className), the census's last two OPEN rows. After it the census ledger has no OPEN row. Implemented by the os-dev for thedomain:spec@ objectui seat, sessionhttps://claude.ai/code/session_01V8qYZseSLfKH3V8tPu2f73, claim 6113481475 as amended by 6114408150. Basec0c2f79, synced by mergingmainf3d5408(objectui#12155's DOM allowlist; conflicts kept both sides). Head9c93615.Premises (measured before any declaration): both held
step. The built-in number input can derive the step from the metadata carrier (field.field || field; on theObjectFormroute that is the raw object field withscale), byNumberField's rule: an explicitstep, else10^-scale, elseany. SoObjectFormnow writes nostep.objectFormNumericStep-9574.test.tsxpasses on both routes with every assertion byte-identical tomain: its 7expect(lines diff empty, and only header prose changed. Ablating the derivation turns its two route-1 tests red. A hand-authoredinputwithinputType: 'number'steps byany/0.01(scale 2) /5(explicit), where it had no step attribute / none /5before.className. Measured per type through the realSchemaRenderer, with components only and with@object-ui/fields, in three states. Every reader puts it on an element the field's own widget or branch renders: the control in most types, the widget's own box fortags,file,imageandslider, the first sub-input foraddressandgeolocation, a display element, or asection-divider's divider. None puts it on the field row, the label, the host's read-only group or outside the row.lookup,master_detail,userandsignaturedraw no class from it.What changed
packages/components/src/renderers/form/form.tsx: theinputanddefaultarms derive a number input's step (nativeNumberStep), and the built-in branches merge an entry'sclassNamewith their own classes (cn(own, className)), asselectalready did. Before, the entry's class replaced the touch-target and read-only classes.packages/plugin-form/src/ObjectForm.tsx: theformField.step = … : 'any'write is removed.FormField/FormFieldSchema:stepdeclared by reference to the spec'sFieldSchema.step;classNamedeclared asstring.DECLARED_BY_12130), the effects and components-only pins (with/without cases, one per element class, and a never-the-row sweep), the face pin and the zod coverage list.form.mdxandplugin-form.mdx(classNameis now a declared key). Its pin incheck-doc-component-types.test.tsflips totoMatch..changeset/12130-step-classname.md(types / componentsminor, plugin-formpatch, breaking semantics stated). Round 1's pending changeset loses its 「Still undeclared」 sentence.Gates run (code at
215da7e;9c93615adds theplugin-form/README.mdcorrection only)os-verify-lock:scripts/__tests__/「Test Files 180 passed | 2 skipped (182)」; the 9574 pin 「Tests 3 passed (3)」;examples/schema-catalog/「Test Files 46 passed (46)」. Atb1fd4c4:packages/types/,packages/plugin-form/,packages/components/: all exit 0.pnpm --filtertype-checkfor types, components, plugin-form and example-schema-catalog: exit 0.pnpm check: 「0 did not validate」.check-doc-component-types,-doc-snippet-types,-doc-json-fences,check:doc-fences,-doc-example-ids: all exit 0.check-changeset-presence「7 source file(s) of 3 released package(s) changed, and this change declares 1 changeset(s)」.-claims,-no-major,-overwriteand-fixedexit 0 (-overwritereports the round-1 changeset edit, report-only).check:pending-changeset-literals,check:control-bytesandcheck:new-line-citations(「0 new citation(s)」): exit 0.Acceptance notes
packages/plugin-form/README.mdlistedclassNameas 「Not aFormFieldkey」. It ships with the package, so it is corrected in this PR (9c93615, claim amendment6114675131):classNamemoves to the declared keys and the stale 「23 keys」 count is dropped.lookup,master_detail,userandsignatureacceptclassNameand draw nothing from it. That is noted, not filed.Generated by Claude Code