Uh oh!
There was an error while loading. Please reload this page.
refactor(components)!: rename the type input of action:button / action:icon to actionType (#7415) - #7436
Merged
os-project-manager merged 1 commit intoSep 3, 2026
Conversation
…action:icon` to `actionType` (#7415) The declared input carrying an action's execution type collided with the SDUI envelope's component discriminator, and on no tier could an author actually set it: the html tier's parser composes `{ type: tag, ...props }` so `type="api"` replaced the discriminator (and `validate.ts` skips `type` as a BASE_PROP, so there was no diagnostic); the react-page wrapper stamps the tag last and parks the author's value under `specType`, which neither renderer reads; through `SchemaRenderer` the node's `type` is the component id, which the renderers forwarded to `ActionRunner` as an action type that binds no handler. `actionType` is the spelling this repo already used for this value — `action:bar` renames the declared type as it spreads a member, `ActionRunner.execute` resolves `action.type || action.actionType || action.name`, and both renderers already read `schema.actionType` first. This makes the one working spelling the declared one, the same resolution `page:tabs.type` got upstream (retired for `tabStyle`). No alias, no transition window: the `|| schema.type` leg is removed, so a declaration that omits `actionType` forwards `undefined` and the runner falls through to its own `action.name` leg rather than being handed a component id. Implements objectstack#14490 ruling A (maintainer, 2026-09-02, batch #13 item 4). Census over 4,797 files found zero authored nodes setting the input. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EMrWaQw3XS5DxTHxp4yRyC
os-project-manager
commented
Sep 3, 2026
Collaborator
Handoff to the project director seat — ⛔ NOT reviewed or accepted by this seatPM note from the Clause ② is engaged (a published component input is renamed) and the card states Verified independently by this seat, from |
| claim | reading |
|---|---|
actionType is already this repo's spelling for this value | ActionRunner.ts:972 — const actionType = action.type || action.actionType || action.name || ''; |
both renderers already read actionTypefirst | action-button.tsx:148 and action-icon.tsx:113 — both type: schema.actionType || schema.type, |
| the census figure is corroborated by a second instrument | SchemaRenderer.tsx:1126 — the repo's own corpus census reads "action:button (5 nodes)", matching the PR's independent md/mdx scan |
the page:tabs.type → tabStyle rename precedent exists | 5 occurrences in registry-inputs-spec-parity.test.ts (1,966 lines) |
⇒ The rename does not introduce a word. It promotes the spelling that already worked to being the declared one, and it follows an in-repo precedent that resolved the identical carrier collision by renaming rather than aliasing.
action-bar-member-type-resolution.test.tsx:26-27 says action:icon "read schema.typealone and dropped actionType entirely", which reads as contradicting the PR. It does not — that docblock is past tense about a defect since fixed, and action-icon.tsx:105-110 records the repair in its own words. The source, not the prose about the source, settles it.
⭐ The disclosure that deserves the most weight
The implementer caught and reported its own vacuous ablation:
The first attempt at this ran the mutation script and the test as
ablate.sh && vitest, so the trap restored the tree before vitest started and the ablation vacuously passed 4/4. It was re-run with the measurement inside the same process; that is the run reported above.
A restore trap that fires before the measurement produces a green that means nothing — and it looks exactly like a successful ablation. It found that, said so, and re-ran. ⭐ It also states plainly that rows 1-3 of the acceptance test are green on origin/main too and that only row 4 measures the rename, rather than presenting four green rows as four proofs.
Recorded, not reviewed — for the reviewer to weigh
⛔ Observations from the dispatching seat, not a tier verdict:
- The
|| schema.typefallback is removed, and the reasoning is worth confirming rather than assuming: post-rename that leg is precisely "the old spelling still works", which the ruling forbids; and on every authored path it now yields the component id, never an action type. A declaration omittingactionTypeforwardsundefined, soActionRunnerfalls through to its ownaction.nameleg instead of being handed'action:icon'— the objectui#6306 shape. - Bytes:
frameworkmoves zero. The change is entirely insidepackages/components, which lands inui-components, not inframework(packages/core|react|types). Measuredframework523,823 → 523,823;ui-componentsgzip unchanged at 396,598 while absorbing +8 raw bytes; aggregate −2 B gzip. The ±1-5 B wobble on other chunks is explained as gzip entropy with unchanged raw sizes, which is the right way to dismiss it. ⇒ [Decision] Theframeworkper-chunk eager-closure ceiling leaves 177 bytes for the whole repo — two ruled user-facing fixes cannot land, and no wording of either fits #7399 is untouched. - The census reports 0 authored sites across 4,797 files with a firing control on every single row — including the negative-space rows (no
props/propertiesbag, no JSXtype=attribute).
Generated by Claude Code
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
|
This was referenced Sep 3, 2026
os-project-manager
marked this pull request as ready for review
September 3, 2026 04:46
Uh oh!
There was an error while loading. Please reload this page.
os-project-manager
deleted the
claude/issue-7415-action-type-input-rename
branch
September 3, 2026 05:03
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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#7415
Implements objectstack#14490 ruling A (maintainer, 2026-09-02, decision batch #13 item 4, verbatim 「同意」): the collision between the component input named
typeonaction:button/action:iconand the SDUI envelope discriminatortypeis removed at the source. The input is renamed toactionType. No alias, no transition window (the standing 不渐进 rule). Landing this unblocks objectstack#14490, which ispm:blockedon this card.All readings below were taken at
09562aa26, the head of this branch.1. Census first — and it is not a formality
The question measured: how many authored sites set the component input named
typeonaction:buttonoraction:icon? Every zero below is reported with a control on the same query that fires non-empty.typeisaction:button/action:icontypekey; 127type: "button"nodestype: "button"nodes in the same blockstype,label,icon,action(×2) andtype,label,action(×3): the discriminator plus theactionchannelprops/propertiesbag carryingtypeon those two componentsActionButton/ActionIconJSX tag carrying atype=attributeresolveIconbindings or prose, verified individuallycontent/docstypeinput of either componentcontent/docs/core/enhanced-actions.mdxdocuments the legacyactionTypedialect — 30 hits, so the term firesapps/consolepublic-contract.test.tslists both ids, so the term firespackages/**non-test sourceaction:bar's own compositionschema.typeread as an action typeaction-button.tsx,action-icon.tsx)inputs+defaultPropsScanned 4,797 files across
examples/,apps/console/,content/docs/andpackages/.The md/mdx count of 5 is independently corroborated: this repo's own corpus census, recorded in
packages/react/src/SchemaRenderer.tsx, already reads "action:button(5 nodes)". Two different scanners, same number.⇒ The rename breaks no authored document in this repo. objectstack's html-tier census found zero; the react-tier population measured here is zero as well. A re-check of objectstack#14490's manifest scan would now return an empty list for inputs named
type— these two were the only ones of the 57 manifest components declaring one — while the stated control, thevariantscan, stays non-empty: both components still declarevariant, untouched by this change.2. The name
actionType, the ruling's working name — and the objectui component configuration does already have a convention, which points at the same string rather than a different one:action:barrenames the declared type as it spreads a member onto its child (type: componentType, actionType: action.type),ActionRunner.executeresolvesaction.type || action.actionType || action.name,ActionDefdeclaresactionType, and both renderers already readschema.actionTypefirst. The rename does not introduce a word; it makes the one spelling that already worked the declared one.page:tabs.typehit the identical carrier collision (a node is the object{ type: 'page:tabs', … }, andvalidate.tsliststypeinBASE_PROPS). Upstream retired it in favour oftabStyle(objectstack#6776);apps/console/src/__tests__/registry-inputs-spec-parity.test.tscarries that history in itsHARVESTEDtable.3. Why the old name could not be authored on any tier
type="api"parse.tscomposes{ type: tag, ...props }— props spread last — so the value replaced the component discriminator and the node stopped resolving to a component.validate.tscannot report it either:typeis inBASE_PROPS, skipped before the declared-input check runs. Two mechanisms, one outcome, no diagnostic. objectstack PR #14274 landed a refusal here whose prescription ("write the tag you meant") is wrong for these two components.type: taglast and parks the author's value underspecType(objectui#2880) — which neither action renderer reads.SchemaRenderertypeis the component id, and the renderer forwarded it as the action type.'action:button'binds no handler and no builtin: a click that did nothing, no error, no toast — the objectui#6306 shape.4. What changed
action-button.tsx,action-icon.tsx: registryinputsentrytype→actionType;defaultPropsfollows; the forward readsschema.actionTypealone.|| schema.typefallback is removed. Post-rename that leg is precisely "the old spelling still works", which the ruling forbids; and on every authored path it now yields the component id, never an action type. A declaration omittingactionTypeforwardsundefined, soActionRunnerfalls through to its ownaction.nameleg instead of being handed a component id — strictly better than the status quo.specType(objectui#2880) is left alone; it addresses other collisions.action:group/action:menuare unaffected: their members carry the specActionSchema.typeinside anactionsarray, a different surface. Theaction-onSuccess-forwardfixture now separates the two spellings explicitly rather than conflating them.@object-ui/componentsminor, naming the rename. It is the only@object-ui/*package whose published behaviour changes.⛔ Deliberately not in this PR (objectstack-side follow-ups, per the card): the
sdui.manifest.jsonpin bump, the html-tier refusal's component-aware special case, thereact-pages.mdxsentence.5. Acceptance
An html-tier page setting the renamed input parses and renders —
packages/components/src/renderers/action/__tests__/action-type-input-html-tier.test.tsx, four rows through the realkind:'html'page path:action:buttontag written withlabel="Mark done" actionType="api" target="/api/v1/tasks/mark_done"compiles with no error panel, renders a real button, and on click handsActionRunnertype: 'api'with the authoredtarget.apihandler is not reached. That is what pins row 1's'api'to the authored prop rather than to a default, the tag, or the handler being the only one registered.type="api"is still the discriminator: the node's type becomes the bare stringapi, which no registry entry claims, so the page reports the compile error and renders no button.getJsxManifest()makes):actionTypeproduces nounknown-propdiagnostic on either component, while a bogus prop does — and the registryinputscontainactionTypeand no longer containtype.Rows 1-3 are green on
origin/maintoo, and that is stated rather than glossed: the renderer already readactionTypefirst, and an undeclared prop is only a warning, which does not stop a page compiling. Row 4 is the one that measures the rename.Reverse verification — two ablations, each mutated on disk, proven there, measured in the same process, and restored:
name: 'actionType'→name: 'type', both files)HEAD1 failed | 3 passed,AssertionError: expected [ Array(1) ] to deeply equal []|| schema.typeback onaction:icon)action-icon.tsxMUTATED,action-button.tsxUNCHANGED1 failed | 5 passed,expected "vi.fn()" to not be called at all, but actually been called 1 times— the component-id trap firesThe mutated files are imported by relative path from the tests, so they resolve to source and no rebuild is interposed; both legs restore via
git checkout HEAD -- ABSOLUTE_PATHunder anEXIT INT TERMtrap, andgit diff HEADis empty afterwards. (The first attempt at this ran the mutation script and the test asablate.sh && vitest, so the trap restored the tree before vitest started and the ablation vacuously passed 4/4. It was re-run with the measurement inside the same process; that is the run reported above.)6. Bytes — measured, not assumed
Two full console builds,
check:eager-closureon each, byte-exact fromapps/console/dist/eager-closure.json. Before =6411def25(this branch's base) in a dedicated comparison worktree; after =09562aa26.framework(gzip)ui-components(gzip)ui-components(raw)vendor-objectstack(gzip)Which chunk this lands in was the question, and the answer is
ui-components, notframework.frameworkispackages/core|react|typesand had 177 B (0.2 KB) of headroom with two PRs already parked on it (#7399); this change is entirely insidepackages/components, soframeworkmoves zero bytes — the +8 raw bytes land inui-components, which absorbs them without moving its gzip figure at all. The remaining ±1-5 B wobble on eight other chunks is gzip entropy: every one of them has an unchanged raw size.check:eager-closurepasses on both builds. ⛔ No ceiling raised, no baseline re-pinned, no gate weakened.7. Verification run
Union re-run at
09562aa26, the final commit (nothing landed after it).pnpm exec vitest run packages/components/ packages/core/src/registry/ apps/console/src/__tests__/— 258 files, 2,523 tests, all pass.packages/components(app-shellwidget/DOM sweeps,PageView,useConsoleActionRuntime×2,EnvironmentListToolbar×2,RecordDetailView, and the schema-catalog DOM-leak test) — 8 files, 275 tests, all pass.pnpm --filter @object-ui/components run type-check— exit 0 (the script is spelledtype-checkhere; the echoed script name confirms it ran rather than matching nothing).pnpm --filter @object-ui/components run lint— exit 0, 0 errors, 950 pre-existing warnings. Run as this repo spells it (eslint ., no--no-inline-config).check:action-forward-parity✅ ·check:control-bytes✅ (6,124 tracked text files) ·check:sdui-registration-pins✅ ·check:doc-types✅ ·check:doc-snippets✅ (417/417 blocks, after building what it names — its first exit 2 wasPRECONDITION NOT MET, not a red verdict) ·check:designer-field-key-parity✅ ·check:vi-mock-specifiers✅ ·check:vi-mock-inherit✅ ·check:eager-closure✅.Repo-wide
pnpm lintis CI's run and was not repeated locally; the narrowing above is the package the diff is confined to.8. Clause ② — contract review
A published component input is renamed, so
CONTRACT_REVIEW_TIERapplies. The shape is pinned by the ruling and implemented mechanically. This PR carriesneeds:contract-reviewfor the project director seat. ⛔ Not self-reviewed, auto-merge not enabled, not enqueued.🤖 Generated with Claude Code
https://claude.ai/code/session_01EMrWaQw3XS5DxTHxp4yRyC
Generated by Claude Code