fix(fields): close the FieldEditWidget delegation road for retired field types - #4942
Merged
Merged
Conversation
…eld types (#4931) `hasFieldEditWidget` answered `resolveInlineEditType(type) in EDIT_WIDGETS`, and `resolveInlineEditType` returns a type unchanged when it is already a key of `EDIT_WIDGETS`. While `owner: UserField` sat in that map the alias table — where objectui#4814's retirement lives — was never consulted, so the retirement never applied: `hasFieldEditWidget('owner')` was `true` and the grid's inline cell editor still rendered a working person picker for a field the record form refuses with a tombstone. Gate on the live `RETIRED_FIELD_TYPES` table so the class closes, not the spelling: `hasFieldEditWidget` -> false, `isInlineExcludedFieldType` -> true, and `FieldEditWidget` renders no control while writing the once-per-spelling prescription. Drop `owner` from `EDIT_WIDGETS` and from the cosmetic `COMPACT_EDIT_TYPES` set (#4914 item 11). The exclusion half is load-bearing: `ObjectGrid` forces `editable: false` on a column exactly when `isFieldInlineEditable` is false, so without it the column stays editable and DataTable falls back to a plain text input that saves a bare string over a stored person reference. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DbRmJD3iPhXjKr6vhd4Qkv
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
…t-widgets-retired-gate
…n refuses it PR #4932 merged after this branch's base and shrank `owner` out of the published field-type unions in `@object-ui/types`, so the four `type: 'owner'` field literals in the new pin no longer compile (TS2322) — the contract twin working exactly as designed. The suite still has to author the refused spelling, because what it pins is the RUNTIME disposition for a field that is ALREADY STORED that way, which no compile-time union can reach. Confine the escape to one helper (`retiredFieldLiteral`) so strictness holds everywhere else in the file, and state there that the refusal is the point rather than an obstacle. Verified the escape is load-bearing, not decorative: removing `as never` reproduces TS2322 at all four call sites. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DbRmJD3iPhXjKr6vhd4Qkv
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
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 #4931
Closes the last measured active-contradiction face of #4814's ruling A′: the
inline-edit delegation road. Its parent #4914 stays open — items 4 and 6–15
remain there as ordinary cleanup, and this PR does not reference that card with
a closing keyword.
Why this seam survived two retirement PRs
hasFieldEditWidgetanswersresolveInlineEditType(type) in EDIT_WIDGETS, andresolveInlineEditTypereturns a type unchanged when it is already a key ofEDIT_WIDGETS:So while
owner: UserFieldsat in that map, the alias table — where #4814'sretirement lives — was never consulted, and the retirement never applied. No
routing table had to list the spelling for it to stay alive, which is why
deleting memberships in #4932 did not reach it.
Re-measured on this branch (not inherited from the card, which measured on
69dbea20a), atorigin/main25c8007d4:hasFieldEditWidget('owner')truefalseisInlineExcludedFieldType('owner')falsetrueFieldEditWidgetcall,type: 'owner'role=buttonreadingSelect…hasFieldEditWidget('user')truetrueThe user-facing consumer, measured before/after
The real direct consumers are the grid's inline cell editor
(
ObjectGrid.tsx:2674), the kanban required-fields dialog(
requiredWhenPrompt.ts:123), andInlineFieldInput's delegation tail. Thegrid is the user-facing one:
isFieldInlineEditable({ type: 'owner' })truefalseObjectGrid.tsx:1989editable: falseownerfieldisFieldInlineEditable({ type: 'user' })truetrueSo the cell degrades to the grid's normal read-only path — the same
disposition
password,formulaandcompositealready get. It does notcrash, and it never saves a string. The read path is unchanged from #4814: the
text cell plus the once-per-spelling console prescription.
What changed, and why the second gate is not optional
The gate consults the live
RETIRED_FIELD_TYPEStable (imported from./field-type-alias, never copied, and not via the./indexbarrel — thatwould cycle), so this closes the class: the next retirement covers this seam
the day it lands in that table, with no edit here.
hasFieldEditWidget()→falsefor any retired spelling, ahead of everytable lookup, so it holds even if a retired spelling is still a key of the
widget map.
isInlineExcludedFieldType()→truefor any retired spelling.FieldEditWidgetrenders no control for a retired spelling and writes theonce-per-spelling prescription.
ownerdropped fromEDIT_WIDGETSand from the cosmeticCOMPACT_EDIT_TYPESsizing set (ownersurvives as a live field-type in ~13 more code faces after #4814's retirement — including a published Zod enum that still accepts it #4914 item 11, same stroke).Gate 2 is load-bearing, and the reverse verification below proves it.
ObjectGridforces
editable: falseon a column exactly whenisFieldInlineEditableisfalse, and that predicate consults
isInlineExcludedFieldType. Closing gate 1alone would have left the column editable, handed the cell editor a
nullwidget, and let DataTable fall back to its built-in plain text input — which
for a stored person reference displays a coerced value and writes a bare string
straight back over it. That would have traded a visible wrong answer for a
silent one.
Gate 3 answers with the console prescription rather than a rendered
RetiredFieldTombstonedeliberately. This component's hosts are cell-sized, and#4814 settled the same question the same way for the read path — "there is no
visible alert a table CELL can carry without wrecking the row". The visible
refusal remains the record form's job. Returning
nullis this function'sdocumented contract for "no dedicated widget", not a new fallback, and every
in-repo host is routed away from it by gates 1–2.
Reverse verification — direction predicted first, both legs from the committed state
Leg A — restore
owner: UserFieldtoEDIT_WIDGETS, keep the gates.Predicted: all pins stay green, because the gates key on the retirement
table rather than on map absence. That is the class-closure property, so a red
here would have disproved it. Observed:
Test Files 2 passed (2) / Tests 21 passed (21). Note this is a reversal of the naive "restoring the key reddensthe pins" expectation — reported as observed.
Leg B — strip all three gates, keep the
EDIT_WIDGETSdeletion (i.e. theweaker spelling-only fix). Predicted: road 1 and road 3 stay green
(
hasFieldEditWidgetis alreadyfalsevia map absence; the alias table stillfires the prescription as a side effect), and exactly the two exclusion pins
redden. Observed exactly that —
Tests 2 failed | 10 passed (12):This is the empirical argument for gate 2: a delegation-road-only fix passes
every other pin in the file while leaving the grid's text-box path open.
Both legs restored with
git checkout e255f9123 --+ the edited path;git diff HEADempty afterwards.
Tests
New pin
packages/fields/src/__tests__/FieldEditWidget.retiredFieldType.test.tsxis quantified over
RETIRED_FIELD_TYPESthroughout (the structural shape#4932 used), with a non-vacuity guard on the table and on the live relational
types. It pins both wrong outcomes in both directions — a working picker and a
quiet text box — plus the once-per-spelling dedupe, the
$expand-ed value case,and the
usersurvivor.Two existing fixtures were triaged rather than re-spelled, because each pinned
the exact branch this PR removes:
FieldEditWidget.test.ts—hasFieldEditWidget('owner') === truein therelational-picker list;
ownerdropped, the three survivors kept.complex-widgets.test.tsx— "owner field resolves an expanded-object valueinline" removed, not re-spelled: its value-shape contract is unchanged and
still pinned by the identical
usercase directly above it.One of my own assertions failed on first run and was corrected to measured
reality rather than forced:
hasFieldEditWidget('not-a-real-type')istrue,not false — an unknown type resolves through the alias table's
|| 'field:text'tail onto the
texteditor. That sharpens the contrast the card is about, andthe test now says so: the retired spelling is the one type the text tail does
not catch, and had
ownermerely been dropped from the widget map it wouldhave landed on exactly that fallback.
hasFieldEditWidget('owner')is deliberately still not pinned by#4932's
InlineFieldInput.retiredFieldType.test.tsx, so this fix does not reddenit — confirmed green in the run below.
Evidence
Union run at
e255f9123(= final commit = pushed remote head), heavy runs underflock -E 99 -w 540 /tmp/os-heavy-verify.lock, dependency closures built beforeevery type-check verdict.
packages/fields/→Test Files 103 passed (103) / Tests 1705 passed (1705)packages/plugin-grid/ packages/plugin-kanban/ packages/plugin-detail/(thethree real consumers) →
Test Files 167 passed (167) / Tests 1514 passed (1514)type-checkfor@object-ui/fields,plugin-grid,plugin-kanban,plugin-detail→ allDone, script echoed astsc --noEmit && tsc -p tsconfig.test.json(no zero-match trap). First attemptfailed on unbuilt
@object-ui/mobile/@object-ui/permissions— the closuretrap, not this change; green after building the closures.
check:control-bytesOK (4419 files),check:self-importOK,check:phantom-depsOK,check:published-distOK (39packages),
changeset:checkOK,check:doc-typesOK. Plus a manualcontrol-byte scan of the five changed files: clean.
No exported signature changed, so this is a behavior narrowing rather than a type
narrowing; a real patch changeset for
@object-ui/fieldsrecords it. Noskip-changesetlabel was created or applied (phantom in this repo, #4912).useris untouched throughout — it keeps the picker, its compact sizing and itsexpanded-object resolution. The record-owner idiom survives verbatim as
{ type: 'user', name: 'owner' }.Generated by Claude Code