enhance(core): Check set() Union rows per member by ntucker · Pull Request #4179 · reactive/data-client

ntucker · 2026-10-05T07:44:15Z

Requested by Nathaniel · project thread

Follow-up to #4133 (unreleased typed set() values).

Motivation

#4133 merged a Union's members into one row type to speed up type checking. That made prev => ({ ...prev, field }) updaters on large Unions about 3x slower than v0.18, and it let rows mix fields from different members. The per-member rows it replaced weren't stricter: their { [k: string]: unknown } catch-all let a row fall through to another member's row.

const Feed = new schema.Union({ post: Post, comment: Comment }, 'type');
// TypeScript error: commentBody is a Comment field, not a Post field
ctrl.set(Feed, { id: '1', type: 'post' }, { type: 'post', commentBody: 'hi' });

Solution

Each Union member gets its own row, with no index signature. When each member declares its discriminator as a literal (readonly type = 'post'), TypeScript picks the member by that discriminator and checks the row against it. A Union row now takes only declared fields, like an Entity row. A key that a schemaAttribute function reads must therefore be declared on each member (decided: strict rows over keeping the catch-all).

TS 6 check time, 1000 calls on a 30-member Union (master, then this PR):

Shape	master	this PR
`prev => ({ ...prev, f })`	9.4s	3.3s (v0.18: 3.1s)
`prev => ({ id: prev.id, f })`	2.3s	2.3s
`() => ({ f })`	1.8s	1.7s
plain value	1.21s	0.91s

Zero loss: with @ts-expect-error neutralized, every error master reports still appears on TS 4.0, 4.5, 5.0, 5.4, 6 and 7. The new errors are the mixed-member and unknown-field rows. The v0.19 blog perf rows are re-measured against v0.18.1. The type-check budget is updated: union instantiations go up 19% from the per-member rows, but its check time doesn't change. setValues drops 71% and setUpdaters 49%.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

Note

Medium Risk
Compile-time-only change that may surface new TypeScript errors on Union set() calls; runtime store behavior is unchanged.

Overview
Tightens Controller.set() typing for Union schemas by making SetRow distributive over union members instead of merging all members into one permissive row. When each member uses a literal discriminator (e.g. readonly type = 'post'), TypeScript checks the value against the selected member only—mixing another member’s fields or unknown keys becomes a compile error, and keys used by schemaAttribute must be declared on each member.

Docs (Controller references, changeset, v0.19 blog) are updated to describe this behavior and refreshed perf numbers: set() plain values and updater functions on large unions are faster than the previous merged-row approach, with updated scripts/typeperf/budget.json instantiations. React hook tests add @ts-expect-error coverage for the new union rules.

^{Reviewed by Cursor Bugbot for commit 01577f0. Bugbot is set up for automated code reviews on this repo. Configure here.}

A Union row is now checked against the member its discriminator selects, without the index signature that let a row fall through to another member. Spread updaters on a 30-member Union check about 3x faster than before (TS 6: 9.4s to 3.3s per 1000 calls), and rows that mix members' fields or carry unknown keys are now errors. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

changeset-bot · 2026-10-05T07:44:19Z

⚠️ No Changeset found

Latest commit: 01577f0

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes changesets to release 7 packages

Name	Type
@data-client/core	Patch
@data-client/react	Patch
@data-client/vue	Patch
example-benchmark	Patch
example-benchmark-react	Patch
test-bundlesize	Patch
coinbase-lite	Patch

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

vercel · 2026-10-05T07:44:21Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
docs-site	Ready	Preview	Oct 5, 2026 5:56pm UTC

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

ntucker · 2026-10-05T07:45:47Z

Staff engineer (Cursor agent): LGTM on direction at f919dcc, FOLLOW_UP only, nothing here needs to change before merge.

This is a net simplification: SetItem and the FieldKeys/MemberField merge go away, SetRow just distributes, and the type gets both stricter and faster on the updater path. SetManyValue using SetRow directly is fine since E is already known not to be a union there.

Open question: I'd keep it strict (this PR) rather than restore the { [k: string]: unknown } catch-all. The catch-all is what let rows fall through to the wrong member, and declaring a discriminator the schemaAttribute function reads on the Entity is the same rule Entity rows already follow. Since #4133 is unreleased and v0.18 didn't type set() values at all, nobody loses anything that shipped.

FOLLOW_UP (docs nuance, take it or leave it): per-member checking only kicks in when the discriminator field has a literal type on each member, like the fixtures' readonly type = 'first'. If a user writes type = 'post' (no readonly) or type: string, TypeScript can't narrow, falls back to "the key exists on some member", and { type: 'post', commentBody: 'hi' } compiles again. The new notes in Controller.md, the skill references and the changeset read as if that always errors. One clause such as "when each member declares its discriminator as a literal (readonly type = 'post')" would make the claim exact, either here if docs get touched again or in a later docs pass.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

ntucker · 2026-10-05T07:46:49Z

Folded the docs nuance into this PR in cdf135d. I confirmed it first: with type = 'a' (no readonly), { type: 'a', bOnly: 1 } compiles. Controller.md, the changeset, the blog and the skill references now say per-member checking applies when each member declares its discriminator as a literal.

Generated by Claude Code

github-actions · 2026-10-05T07:49:46Z

Size Change: 0 B

Total Size: 103 kB

ℹ️ View Unchanged

Filename	Size
`examples/test-bundlesize/dist/App.js`	1.46 kB
`examples/test-bundlesize/dist/polyfill.js`	307 B
`examples/test-bundlesize/dist/rdcClient.js`	10.9 kB
`examples/test-bundlesize/dist/rdcEndpoint.js`	8.07 kB
`examples/test-bundlesize/dist/rdcNextjs.js`	12.3 kB
`examples/test-bundlesize/dist/rdcPipeableStream.js`	9.64 kB
`examples/test-bundlesize/dist/react.js`	59.7 kB
`examples/test-bundlesize/dist/webpack-runtime.js`	784 B

_{compressed-size-action}

github-actions

Benchmark React

Details

Benchmark suite	Current: `01577f0`	Previous: `eb8aacd`	Ratio
`data-client: getlist-100`	`143.89` ops/s (`± 5.3%`)	`132.46` ops/s (`± 4.8%`)	`0.92`
`data-client: getlist-500`	`44.05` ops/s (`± 4.9%`)	`43.29` ops/s (`± 5.0%`)	`0.98`
`data-client: update-entity`	`384.62` ops/s (`± 11.0%`)	`322.58` ops/s (`± 7.5%`)	`0.84`
`data-client: update-user`	`408.33` ops/s (`± 11.5%`)	`333.33` ops/s (`± 8.6%`)	`0.82`
`data-client: getlist-500-sorted`	`41.15` ops/s (`± 9.3%`)	`44.64` ops/s (`± 9.3%`)	`1.08`
`data-client: update-entity-sorted`	`370.37` ops/s (`± 7.4%`)	`303.03` ops/s (`± 6.6%`)	`0.82`
`data-client: update-entity-multi-view`	`370.37` ops/s (`± 7.5%`)	`357.14` ops/s (`± 9.9%`)	`0.96`
`data-client: list-detail-switch-10`	`10.31` ops/s (`± 9.0%`)	`9.23` ops/s (`± 8.8%`)	`0.90`
`data-client: update-user-10000`	`80.01` ops/s (`± 14.7%`)	`77.52` ops/s (`± 14.7%`)	`0.97`
`data-client: invalidate-and-resolve`	`38.31` ops/s (`± 6.1%`)	`34.42` ops/s (`± 3.7%`)	`0.90`
`data-client: unshift-item`	`224.75` ops/s (`± 9.2%`)	`194.19` ops/s (`± 7.9%`)	`0.86`
`data-client: delete-item`	`294.12` ops/s (`± 6.7%`)	`243.9` ops/s (`± 2.5%`)	`0.83`
`data-client: move-item`	`183.5` ops/s (`± 10.1%`)	`166.67` ops/s (`± 9.3%`)	`0.91`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Benchmark

Details

Benchmark suite	Current: `01577f0`	Previous: `eb8aacd`	Ratio
`normalizeLong`	`418` ops/sec (`±3.90%`)	`410` ops/sec (`±3.15%`)	`0.98`
`normalizeLong Values`	`385` ops/sec (`±1.24%`)	`393` ops/sec (`±0.48%`)	`1.02`
`normalizeLong Scalar`	`349` ops/sec (`±3.21%`)	`353` ops/sec (`±3.42%`)	`1.01`
`normalizeLong Scalar update`	`915` ops/sec (`±0.17%`)	`896` ops/sec (`±0.90%`)	`0.98`
`denormalizeLong`	`242` ops/sec (`±6.44%`)	`226` ops/sec (`±6.16%`)	`0.93`
`denormalizeLong Values`	`224` ops/sec (`±5.22%`)	`214` ops/sec (`±5.18%`)	`0.96`
`denormalizeLong donotcache`	`1005` ops/sec (`±0.97%`)	`984` ops/sec (`±0.59%`)	`0.98`
`denormalizeLong Values donotcache`	`746` ops/sec (`±0.28%`)	`731` ops/sec (`±0.26%`)	`0.98`
`denormalizeLong Scalar donotcache`	`1052` ops/sec (`±0.22%`)	`1005` ops/sec (`±0.42%`)	`0.96`
`denormalizeShort donotcache 500x`	`1400` ops/sec (`±0.15%`)	`1397` ops/sec (`±0.14%`)	`1.00`
`denormalizeShort 500x`	`606` ops/sec (`±6.14%`)	`629` ops/sec (`±6.39%`)	`1.04`
`denormalizeShort 500x withCache`	`6837` ops/sec (`±0.13%`)	`6684` ops/sec (`±0.32%`)	`0.98`
`queryShort 500x withCache`	`3173` ops/sec (`±0.08%`)	`3083` ops/sec (`±0.59%`)	`0.97`
`buildQueryKey All`	`58819` ops/sec (`±1.48%`)	`57185` ops/sec (`±0.91%`)	`0.97`
`query All withCache`	`6342` ops/sec (`±2.69%`)	`5928` ops/sec (`±2.01%`)	`0.93`
`denormalizeLong with mixin Entity`	`220` ops/sec (`±7.81%`)	`203` ops/sec (`±6.10%`)	`0.92`
`denormalizeLong withCache`	`7754` ops/sec (`±0.18%`)	`6654` ops/sec (`±0.29%`)	`0.86`
`denormalizeLong withCache (Scalar churn)`	`7682` ops/sec (`±1.22%`)	`6500` ops/sec (`±0.74%`)	`0.85`
`denormalizeLong Values withCache`	`5251` ops/sec (`±1.74%`)	`4766` ops/sec (`±1.90%`)	`0.91`
`denormalizeLong Scalar withCache`	`7953` ops/sec (`±0.40%`)	`7050` ops/sec (`±0.26%`)	`0.89`
`denormalizeLong Scalar update withCache`	`4098` ops/sec (`±1.07%`)	`4028` ops/sec (`±0.23%`)	`0.98`
`denormalizeLong All withCache`	`6604` ops/sec (`±0.20%`)	`6216` ops/sec (`±0.39%`)	`0.94`
`denormalizeLong Query-sorted withCache`	`6635` ops/sec (`±1.98%`)	`6060` ops/sec (`±1.43%`)	`0.91`
`denormalizeLongAndShort withEntityCacheOnly`	`1834` ops/sec (`±0.23%`)	`1684` ops/sec (`±0.31%`)	`0.92`
`denormalize bidirectional 50`	`4708` ops/sec (`±8.90%`)	`4350` ops/sec (`±11.67%`)	`0.92`
`denormalize bidirectional 50 donotcache`	`41693` ops/sec (`±0.27%`)	`43258` ops/sec (`±0.58%`)	`1.04`
`getResponse`	`4462` ops/sec (`±3.96%`)	`4393` ops/sec (`±3.68%`)	`0.98`
`getResponse (null)`	`10265517` ops/sec (`±0.56%`)	`9494914` ops/sec (`±1.41%`)	`0.92`
`getResponse (clear cache)`	`208` ops/sec (`±7.98%`)	`190` ops/sec (`±9.48%`)	`0.91`
`getSmallResponse`	`3544` ops/sec (`±0.45%`)	`3361` ops/sec (`±1.09%`)	`0.95`
`getSmallInferredResponse`	`2894` ops/sec (`±0.23%`)	`2841` ops/sec (`±0.53%`)	`0.98`
`getResponse Collection`	`4226` ops/sec (`±4.75%`)	`4463` ops/sec (`±3.49%`)	`1.06`
`get Collection`	`2695` ops/sec (`±0.20%`)	`2791` ops/sec (`±0.37%`)	`1.04`
`get Query-sorted`	`5123` ops/sec (`±1.57%`)	`4921` ops/sec (`±1.86%`)	`0.96`
`setLong`	`430` ops/sec (`±0.36%`)	`435` ops/sec (`±0.26%`)	`1.01`
`setLongWithMerge`	`253` ops/sec (`±0.15%`)	`238` ops/sec (`±1.29%`)	`0.94`
`setLongWithSimpleMerge`	`267` ops/sec (`±0.57%`)	`264` ops/sec (`±0.78%`)	`0.99`
`setSmallResponse 500x`	`911` ops/sec (`±1.65%`)	`895` ops/sec (`±1.64%`)	`0.98`
`setMany 50x one-per-row`	`146` ops/sec (`±0.47%`)	`141` ops/sec (`±0.49%`)	`0.97`
`setMany 50 batch`	`3563` ops/sec (`±1.02%`)	`3415` ops/sec (`±2.44%`)	`0.96`
`setMany 500x one-per-row`	`14.93` ops/sec (`±0.66%`)	`14.24` ops/sec (`±0.70%`)	`0.95`
`setMany 500 batch`	`1400` ops/sec (`±0.62%`)	`1332` ops/sec (`±3.14%`)	`0.95`

This comment was automatically generated by workflow using github-action-benchmark.

…ad-ii70fb

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

codecov · 2026-10-05T17:16:38Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.08%. Comparing base (711c741) to head (01577f0).
⚠️ Report is 6 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4179   +/-   ##
=======================================
  Coverage   98.08%   98.08%           
=======================================
  Files         165      165           
  Lines        3139     3139           
  Branches      625      625           
=======================================
  Hits         3079     3079           
  Misses         18       18           
  Partials       42       42

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

…ion rows #4179 changed set() row types after #4189 started checking that the playground editor types are current, so the editor-types check is red on master. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H

#4179 changed set() row types after #4189 added the editor-types check, so master's committed core.d.ts was stale. Claude-Session: https://claude.ai/code/session_01N2JUxvTqcHCdBX8Xi1nQq6 Co-authored-by: Claude <noreply@anthropic.com>

@see

#4181) * docs: Generate DevTools MCP and Vue testing skill references from docs The devtools-debugging skill reference (copied byte for byte into the React and Vue skills) and vue-test-utilities (a copy of packages/vue/src/test/README.md) were hand-written, so the build:skills drift check never covered them. - Move the DevTools MCP recipe to docs/core/guides/devtools-debugging.md (framework-aware) and generate both skills' copies from it - Add Vue "Unit testing components" (mountDataClient) and "Unit testing composables" (renderDataCompose) guides as .vue.md overrides of the React pages, and generate the vue-testing skill's references from them - Fix renderDataCompose examples in the vue-testing skill: it is async and `result` is the composable's return value (there is no `result.current`) - Point the test README and JSDoc @see links at the new guides Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H * docs: Simplify Vue testing skill and guides; link Vue testing guides Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H * docs: Link DevTools MCP debugging from Agent Skills Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H * internal(website): Regenerate playground core types for per-member Union rows #4179 changed set() row types after #4189 started checking that the playground editor types are current, so the editor-types check is red on master. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H * docs: Simplify agent debugging links and Vue testing skill Move the DevTools MCP page into getting-started (it is listed there), link it from the framework skill bullets instead of a second "Debugging with agents" section, link DataClientPlugin to its API page, and drop examples the Vue testing skill repeated. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H * docs(vue): Serve composables testing guide at /vue/guides/unit-testing-composables The Vue page keeps its unit-testing-hooks.vue.md filename so it shares the React page's doc id (sidebar and framework switcher), and sets a slug so its URL says composables. routeOf now honors absolute slugs so generated skill references link to the same URL. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EcBpA1YWwvXC94QHnzQG1H --------- Co-authored-by: Claude <noreply@anthropic.com>

ntucker self-assigned this Oct 5, 2026

docs(blog): Link the set() updater speedup

f919dcc

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

vercel Bot deployed to Preview October 5, 2026 07:45 View deployment

docs: Note Union rows need literal discriminators to check per member

cdf135d

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

vercel Bot deployed to Preview October 5, 2026 07:47 View deployment

github-actions Bot reviewed Oct 5, 2026

View reviewed changes

ntucker marked this pull request as ready for review October 5, 2026 07:56

ntucker mentioned this pull request Oct 5, 2026

ci: Add a type-check perf budget and path-type fuzz check #4182

Merged

claude added 2 commits October 5, 2026 17:08

Merge remote-tracking branch 'origin/master' into claude/project-thre…

2643de4

…ad-ii70fb

internal: Update type-check budget for per-member Union rows

f852788

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

vercel Bot deployed to Preview October 5, 2026 17:15 View deployment

docs(blog): Re-time Typical app row over 10 runs

945c14b

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

vercel Bot deployed to Preview October 5, 2026 17:45 View deployment

docs(blog): Re-time every type-check benchmark over 10 runs

01577f0

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GzkTc2gisCbeWKi5F3v472

vercel Bot deployed to Preview October 5, 2026 17:56 View deployment

ntucker merged commit 10e15cc into master Oct 5, 2026
30 checks passed

ntucker deleted the claude/project-thread-ii70fb branch October 5, 2026 18:00

This was referenced Oct 5, 2026

docs: Generate DevTools MCP and Vue testing skill references from docs #4181

Merged

internal: Regenerate playground editor types for set() row types #4190

Merged

Navigation Menu

Uh oh!

enhance(core): Check set() Union rows per member - #4179

enhance(core): Check set() Union rows per member#4179
ntucker merged 7 commits into
masterfrom
claude/project-thread-ii70fb

ntucker commented Oct 5, 2026 •

edited by cursor Bot

Loading

Uh oh!

changeset-bot Bot commented Oct 5, 2026 •

edited

Loading

Uh oh!

vercel Bot commented Oct 5, 2026 •

edited

Loading

Uh oh!

ntucker commented Oct 5, 2026

Uh oh!

ntucker commented Oct 5, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

github-actions Bot left a comment •

edited

Loading

Uh oh!

github-actions Bot left a comment •

edited

Loading

Uh oh!

codecov Bot commented Oct 5, 2026 •

edited

Loading

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

ntucker commented Oct 5, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Solution

Uh oh!

changeset-bot Bot commented Oct 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚠️ No Changeset found

Uh oh!

vercel Bot commented Oct 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ntucker commented Oct 5, 2026

Uh oh!

ntucker commented Oct 5, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

github-actions Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Benchmark React

Uh oh!

github-actions Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Benchmark

Uh oh!

codecov Bot commented Oct 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ntucker commented Oct 5, 2026 •

edited by cursor Bot

Loading

changeset-bot Bot commented Oct 5, 2026 •

edited

Loading

vercel Bot commented Oct 5, 2026 •

edited

Loading

github-actions Bot left a comment •

edited

Loading

github-actions Bot left a comment •

edited

Loading

codecov Bot commented Oct 5, 2026 •

edited

Loading