fix(normalizr): do not mutate cached journey on result-cache hit by ntucker · Pull Request #3925 · reactive/data-client

ntucker · 2026-04-28T01:31:06Z

Summary

Companion to #3924 (which targets master), shaped for the argsKey infrastructure on cursor/scalar-schema-design-for-lenses-d6e3 (#3887).

The base PR (#3887) introduced function-typed (argsKey) deps and added a hasStringDeps filter pass to GlobalCache.getResults's hit branch — but kept the pre-existing paths.shift() mutation. That mutation operates on the journey array stored by reference on WeakDependencyMap.Link.journey, so every successive hit on the same cached entry shifts off one more real entity path, progressively corrupting the subscription list returned to consumers.

This PR fixes the bug and (in commit 5a63f74, originally #3928) moves journey materialization to write time so the hit branch returns the stored array by reference with zero post-processing:

@@ packages/normalizr/src/memo/WeakDependencyMap.ts
   set(
     dependencies: Dep<Path, K>[],
     value: V,
     args: readonly any[] = [],
+    /** Optional consumer-facing journey returned to `get()` callers verbatim.
+     * Defaults to `dependencies.map(d => d.path)`. The array becomes a shared
+     * reference held by every subsequent cache hit; callers MUST NOT mutate. */
+    journey?: Path[],
   ) {
     …
-    curLink.journey = dependencies.map(d => d.path) as Path[];
+    curLink.journey = journey ?? (dependencies.map(d => d.path) as Path[]);
   }

@@ packages/normalizr/src/memo/globalCache.ts (getResults)
     if (paths === undefined) {
       data = computeValue();
       paths = this.paths();   // already excludes placeholder + argsKey functions
       this.dependencies[0] = { path: { key: '', pk: '' }, entity: input };
-      this._resultCache.set(this.dependencies, data, this._args);
-    } else {
-      paths.shift();
-      if (this._resultCache.hasStringDeps) {
-        for (let i = 0; i < paths.length; i++) {
-          if (typeof paths[i] === 'function') {
-            paths = paths.filter(p => typeof p !== 'function') as EntityPath[];
-            break;
-          }
-        }
-      }
+      this._resultCache.set(this.dependencies, data, this._args, paths);
     }
+    // hit branch: `paths` aliases the stored Link.journey — return as-is.
     return { data, paths };
   }

The stored journey is built once at write time, in the consumer-facing shape (placeholder slot + function-typed argsKey deps already filtered out by paths()). No mutation, no per-hit allocation, no per-hit typeof === 'function' walk.

Why it matters

Two distinct downstream consumers of the corrupted paths, only one of which is GC-related:

Controller.getResponseMeta → entityExpiresAt(paths, state.entitiesMeta) → expiresAt — picks the earliest entity expiry across the path list. Returns Infinity once the journey is empty. Flows directly into useSuspense (and useCache / useFetch / useDLE / useLive)'s refetch gate if (Date.now() <= expiresAt && !forceFetch) return;, so once any entity is stale the hook never refetches. This bug fires under ImmortalGCPolicy too — createCountRef is a no-op there, but entityExpiresAt is unconditional.
Controller.getResponseMeta → gcPolicy.createCountRef({key, paths}) → GCPolicy.entityCount — under-counted entities get reaped while still referenced. This one is GC-only; ImmortalGCPolicy masks it.

Controller.get (used by useQuery) is the GC-only path — it only passes paths to createCountRef.

Empirical confirmation

A tight loop of result-cache hits against the same primed MemoCache drains the journey to zero on the buggy code; with the fix it stays intact:

iterations	paths.length (buggy)	paths.length (fix)
1	2826	2826
100	2727	2826
1,000	1827	2826
10,000	0	2826

Differences from #3924 (master version)

fix(normalizr): do not mutate cached journey on result-cache hit #3924 (master): minimal one-liner paths.shift() → paths = paths.slice(1). No argsKey infrastructure to coexist with.
This PR: collapses the existing hasStringDeps filter pass with the placeholder-strip, and additionally moves journey materialization to write time so the hit path is allocation-free.

Test plan

Adds regression test in packages/normalizr/src/__tests__/globalCache.test.ts (does not mutate cached journey across repeated result-cache hits) — fails on the buggy version, passes with fix.
Adds Controller-level integration test packages/core/src/controller/__tests__/getResponseMeta-countRef.ts simulating three React subscribers to the same endpoint + args; asserts all three countRef closures correctly increment GCPolicy.entityCount for every entity, and that decrement restores counts to zero. Fails on buggy version (under-counts first entity); passes with fix.
yarn jest --selectProjects ReactDOM --testPathPatterns "packages/(normalizr|core|endpoint)" — 837 tests pass.
yarn jest --selectProjects ReactDOM --testPathPatterns "packages/react/.*useSuspense|useCache|useQuery" — 62 tests pass.

Performance

Methodology: Node 24.14.1, full examples/benchmark normalizr suite × 5 runs per branch + core suite × 3 runs per branch + focused write-only and drain-free fair microbenches. Trimmed mean (drop min/max for n=5; median for n=3). Three branches compared:

master — pre-fix, paths.shift() mutation. Cache-hit "throughput" reflects the bug draining the journey: by the time Benchmark.js samples have stabilized, the stored journey is empty and the hit branch returns no work.
fix only — paths.slice(1) / non-mutating hasStringDeps filter at hit time.
fix + journey-at-write — this PR's final state (commit 5a63f74). Hit branch is allocation-free.

Read path (cache-hit benchmarks, n=5 trimmed mean, ops/sec)

Benchmark	master (buggy)	fix only	this PR	Δ% (this vs master)	Δ% (this vs fix only)
`denormalizeLong withCache`	12,438	11,483	13,291	+6.86%	+15.75%
`denormalizeLong withCache (Scalar churn)`	12,446	11,495	13,277	+6.67%	+15.50%
`denormalizeLong All withCache`	10,087	9,742	11,907	+18.04%	+22.23%
`denormalizeLong Query-sorted withCache`	10,636	9,710	12,313	+15.77%	+26.81%
`denormalizeLong Scalar withCache` (hasStringDeps)	13,961	12,315	13,965	+0.03%	+13.40%
`denormalizeLong Scalar update withCache` (hasStringDeps)	7,600	6,978	7,549	-0.67%	+8.19%
`denormalizeLong Values withCache`	8,878	8,718	9,064	+2.10%	+3.97%
`denormalizeLongAndShort withEntityCacheOnly`	2,979	2,911	3,075	+3.22%	+5.63%
`query All withCache`	10,549	9,269	12,277	+16.38%	+32.46%
`queryShort 500x withCache`	5,601	5,337	5,504	-1.73%	+3.14%
`denormalizeShort 500x withCache`	13,157	13,417	13,217	+0.46%	-1.49%

The *withCache benchmarks against master look modest because master was cheating: its paths.shift() was mutating the stored journey on each hit, so Benchmark.js samples were measuring the cost of returning empty subscription lists (after the journey drained to zero), not the real per-hit work. The fix-only column shows what doing the correct amount of work consistently costs. Moving journey materialization to write time recovers that and exceeds master's apparent throughput on every hit-path benchmark, by removing the per-hit slice(1) / filter entirely.

Write path (write-then-discard benchmarks)

These benches do new MemoCache().denormalize(...) per iteration — exactly the "write but no hit" pattern. The optimization moves one allocation rather than adding work, so deltas are within run-to-run noise:

Benchmark	fix only	this PR	Δ% (this vs fix only)
`denormalizeLong` (1 write, N≈2826)	538	537	-0.06%
`denormalizeLong Values` (1 write, N≈2826)	477	476	-0.14%
`denormalizeLong with mixin Entity`	508	515	+1.31%
`denormalize bidirectional 50` (1 write, cycles)	9,783	9,599	-1.87%
`denormalizeShort 500x` (500 fresh writes/iter)	1,404	1,380	-1.71%

Dedicated write-only microbench (one fresh MemoCache, exactly one denormalize per iter, n=5 trimmed mean):

Benchmark	fix only	this PR	Δ%
`WRITE-only: ProjectSchema (1 miss)`	712	700	-1.64%
`WRITE-only: AllProjects (1 miss)`	711	708	-0.33%
`WRITE-only: StockSchema (1 miss, hasStringDeps)`	656	678	+3.30%

The optimization eliminates one redundant dependencies.map(d => d.path) allocation per top-level cache miss (the journey for the result cache was being computed twice — once as paths() for the consumer, once inside WeakDependencyMap.set). On entity-only schemas this is net-zero in the noise; on hasStringDeps it's a small win because the second dependencies.map was building a journey that included function entries which then got skipped at every hit. The inner per-entity cache (_getCache.set) is unchanged.

`core` suite (n=3, fresh-thermal re-run)

All within run-to-run noise (1–3% per packages/normalizr/AGENTS.md):

Benchmark	fix only	this PR	Δ%
`getResponse`	10,189	10,315	+1.24%
`getResponse Collection`	8,173	7,982	-2.34%
`get Collection`	7,628	7,699	+0.93%
`get Query-sorted`	7,662	8,997	+17.42%
`getSmallResponse`	7,694	7,594	-1.30%
`getSmallInferredResponse`	5,411	5,509	+1.81%
`setLong`	780	789	+1.15%
`setLongWithMerge`	448	451	+0.67%
`setLongWithSimpleMerge`	479	482	+0.63%
`setSmallResponse 500x`	1,667	1,692	+1.50%

^set* benchmarks don't go through MemoCache.set (it's only invoked from the denormalize cache-miss branch), so they don't move.

Drain-free fair microbench (one fresh `MemoCache` per iter → 1 miss + 1 hit)

Apples-to-apples since neither branch's bug can accumulate state across iterations:

Benchmark	master	fix only	this PR	Δ% (this vs master)
`ProjectSchema fresh-prime + 1 hit`	649	639	650	+0.15%
`AllProjects fresh-prime + 1 hit`	664	648	676	+1.81%
`StockSchema fresh-prime + 1 hit (hasStringDeps)`	597	608	635	+6.37%

V8 deopt analysis

node --trace-opt --trace-deopt on the denormalizeLong slice. Both getResults and paths reach stable TURBOFAN_JS optimization on every branch with zero deopts:

Metric	master	fix only	this PR
Total deopt events	68	71	71
Deopts on `getResults`	0	0	0
Deopts on `paths`	0	0	0
Stable tier for `getResults`	TURBOFAN_JS	TURBOFAN_JS	TURBOFAN_JS

Same set of pre-existing deopting functions across all branches (unvisit, getEntity, denormalize, visit, mergeEntity, arrayEach, etc.) — none introduced or removed by this PR. The simpler hit-branch shape on the final state lets getResults stabilize with fewer recompile cycles than master (22 vs 32 events through MAGLEV/TURBOFAN_JS).

Bundlesize

Net source change in packages/normalizr/src/memo/: the entire if (hasStringDeps) { …filter… } else { paths.slice(1) } block in getResults is removed (–14 LOC); WeakDependencyMap.set gains one optional parameter wrapped in JSDoc (+12 LOC). Roughly –80 bytes minified (CI compressed-size workflow on the rebase commit reported -37 B (-0.05%) on rdcClient.js for the fix-only state; the journey-at-write commit drops it slightly more).

Changeset

Included as .changeset/fix-journey-mutation.md. Version-linked packages: normalizr, core, endpoint, rest, graphql, react.

Note

Medium Risk
Touches core memoization/caching behavior used by all denormalize reads; while the change is small, it affects subscription paths used for GC ref-counting and expiry calculations and could surface subtle regressions in cache semantics.

Overview
Prevents GlobalCache.getResults from mutating cached dependency "journeys" on result-cache hits by materializing the consumer-facing paths once on cache write and storing it as the WeakDependencyMap journey returned verbatim on subsequent hits.

Adds regression coverage in normalizr and core to ensure repeated reads keep stable paths, expiresAt, and GCPolicy countRef tracking across multiple subscribers, and bumps @data-client/react peer ranges in @data-client/img and @data-client/test to include ^0.17.0.

^{Reviewed by Cursor Bugbot for commit 05e1298. Bugbot is set up for automated code reviews on this repo. Configure here.}

changeset-bot · 2026-04-28T01:31:10Z

🦋 Changeset detected

Latest commit: 05e1298

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 16 packages

Name	Type
@data-client/normalizr	Patch
@data-client/core	Patch
@data-client/endpoint	Patch
@data-client/rest	Patch
@data-client/graphql	Patch
@data-client/react	Patch
@data-client/img	Patch
@data-client/test	Patch
example-benchmark	Patch
normalizr-github-example	Patch
normalizr-redux-example	Patch
normalizr-relationships	Patch
@data-client/vue	Patch
example-benchmark-react	Patch
test-bundlesize	Patch
coinbase-lite	Patch

Not sure what this means? Click here to learn what changesets are.

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

vercel · 2026-04-28T01:31:11Z

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

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs-site	Ignored	Preview	Apr 28, 2026 1:08pm

codecov · 2026-04-28T01:33:55Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.21%. Comparing base (84078d7) to head (05e1298).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3925      +/-   ##
==========================================
- Coverage   98.21%   98.21%   -0.01%     
==========================================
  Files         154      154              
  Lines        3024     3018       -6     
  Branches      605      604       -1     
==========================================
- Hits         2970     2964       -6     
  Misses         11       11              
  Partials       43       43

☔ View full report in Codecov by Sentry.
📢 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.

github-actions

Benchmark React

Details

Benchmark suite	Current: `05e1298`	Previous: `84078d7`	Ratio
`data-client: getlist-100`	`132.46` ops/s (`± 4.6%`)	`142.86` ops/s (`± 5.0%`)	`1.08`
`data-client: getlist-500`	`37.04` ops/s (`± 6.6%`)	`42.38` ops/s (`± 6.4%`)	`1.14`
`data-client: update-entity`	`307.77` ops/s (`± 9.8%`)	`377.49` ops/s (`± 10.2%`)	`1.23`
`data-client: update-user`	`307.77` ops/s (`± 7.0%`)	`370.37` ops/s (`± 9.8%`)	`1.20`
`data-client: getlist-500-sorted`	`40.49` ops/s (`± 7.8%`)	`44.95` ops/s (`± 6.8%`)	`1.11`
`data-client: update-entity-sorted`	`256.41` ops/s (`± 5.8%`)	`312.5` ops/s (`± 4.9%`)	`1.22`
`data-client: update-entity-multi-view`	`274.02` ops/s (`± 6.0%`)	`333.33` ops/s (`± 10.7%`)	`1.22`
`data-client: list-detail-switch-10`	`6.45` ops/s (`± 6.4%`)	`7.18` ops/s (`± 5.8%`)	`1.11`
`data-client: update-user-10000`	`78.43` ops/s (`± 11.6%`)	`95.7` ops/s (`± 12.4%`)	`1.22`
`data-client: invalidate-and-resolve`	`33.56` ops/s (`± 4.8%`)	`39.22` ops/s (`± 5.3%`)	`1.17`
`data-client: unshift-item`	`200` ops/s (`± 4.8%`)	`243.9` ops/s (`± 4.3%`)	`1.22`
`data-client: delete-item`	`263.16` ops/s (`± 5.6%`)	`294.12` ops/s (`± 6.8%`)	`1.12`
`data-client: move-item`	`156.25` ops/s (`± 9.5%`)	`196.08` ops/s (`± 10.0%`)	`1.25`

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

github-actions · 2026-04-28T05:20:27Z

Size Change: -37 B (-0.05%)

Total Size: 81 kB

📦 View Changed

Filename	Size	Change
`examples/test-bundlesize/dist/rdcClient.js`	10.8 kB	-37 B (-0.34%)

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/rdcEndpoint.js`	8.03 kB
`examples/test-bundlesize/dist/react.js`	59.7 kB
`examples/test-bundlesize/dist/webpack-runtime.js`	726 B

_{compressed-size-action}

github-actions

Benchmark

Details

Benchmark suite	Current: `05e1298`	Previous: `84078d7`	Ratio
`normalizeLong`	`429` ops/sec (`±4.04%`)	`433` ops/sec (`±2.62%`)	`1.01`
`normalizeLong Values`	`400` ops/sec (`±0.32%`)	`395` ops/sec (`±0.27%`)	`0.99`
`normalizeLong Scalar`	`329` ops/sec (`±0.50%`)	`331` ops/sec (`±0.85%`)	`1.01`
`normalizeLong Scalar update`	`837` ops/sec (`±0.49%`)	`853` ops/sec (`±0.69%`)	`1.02`
`denormalizeLong`	`236` ops/sec (`±4.70%`)	`243` ops/sec (`±4.33%`)	`1.03`
`denormalizeLong Values`	`218` ops/sec (`±4.48%`)	`231` ops/sec (`±4.22%`)	`1.06`
`denormalizeLong donotcache`	`960` ops/sec (`±0.68%`)	`942` ops/sec (`±0.13%`)	`0.98`
`denormalizeLong Values donotcache`	`712` ops/sec (`±0.41%`)	`717` ops/sec (`±0.20%`)	`1.01`
`denormalizeLong Scalar donotcache`	`894` ops/sec (`±1.77%`)	`902` ops/sec (`±0.59%`)	`1.01`
`denormalizeShort donotcache 500x`	`1533` ops/sec (`±0.70%`)	`1540` ops/sec (`±0.10%`)	`1.00`
`denormalizeShort 500x`	`697` ops/sec (`±4.13%`)	`717` ops/sec (`±3.96%`)	`1.03`
`denormalizeShort 500x withCache`	`7055` ops/sec (`±0.17%`)	`6899` ops/sec (`±0.14%`)	`0.98`
`queryShort 500x withCache`	`2685` ops/sec (`±0.61%`)	`2819` ops/sec (`±0.17%`)	`1.05`
`buildQueryKey All`	`54261` ops/sec (`±0.78%`)	`53631` ops/sec (`±0.51%`)	`0.99`
`query All withCache`	`5300` ops/sec (`±0.24%`)	`6245` ops/sec (`±0.31%`)	`1.18`
`denormalizeLong with mixin Entity`	`227` ops/sec (`±4.04%`)	`233` ops/sec (`±3.82%`)	`1.03`
`denormalizeLong withCache`	`7019` ops/sec (`±0.19%`)	`6104` ops/sec (`±0.38%`)	`0.87`
`denormalizeLong withCache (Scalar churn)`	`7007` ops/sec (`±0.18%`)	`6090` ops/sec (`±0.23%`)	`0.87`
`denormalizeLong Values withCache`	`4929` ops/sec (`±0.31%`)	`5139` ops/sec (`±0.17%`)	`1.04`
`denormalizeLong Scalar withCache`	`7761` ops/sec (`±1.42%`)	`7615` ops/sec (`±2.80%`)	`0.98`
`denormalizeLong Scalar update withCache`	`4068` ops/sec (`±0.42%`)	`4011` ops/sec (`±1.12%`)	`0.99`
`denormalizeLong All withCache`	`6122` ops/sec (`±0.19%`)	`6420` ops/sec (`±0.23%`)	`1.05`
`denormalizeLong Query-sorted withCache`	`5379` ops/sec (`±0.21%`)	`6232` ops/sec (`±0.21%`)	`1.16`
`denormalizeLongAndShort withEntityCacheOnly`	`1669` ops/sec (`±0.83%`)	`1659` ops/sec (`±0.21%`)	`0.99`
`denormalize bidirectional 50`	`4922` ops/sec (`±4.58%`)	`5090` ops/sec (`±4.89%`)	`1.03`
`denormalize bidirectional 50 donotcache`	`40019` ops/sec (`±0.21%`)	`36895` ops/sec (`±0.16%`)	`0.92`
`getResponse`	`4586` ops/sec (`±0.87%`)	`4636` ops/sec (`±0.97%`)	`1.01`
`getResponse (null)`	`10060925` ops/sec (`±0.63%`)	`9348305` ops/sec (`±1.09%`)	`0.93`
`getResponse (clear cache)`	`223` ops/sec (`±3.99%`)	`227` ops/sec (`±4.11%`)	`1.02`
`getSmallResponse`	`3425` ops/sec (`±0.25%`)	`3556` ops/sec (`±0.35%`)	`1.04`
`getSmallInferredResponse`	`2558` ops/sec (`±0.08%`)	`2623` ops/sec (`±0.25%`)	`1.03`
`getResponse Collection`	`4559` ops/sec (`±0.48%`)	`4566` ops/sec (`±0.65%`)	`1.00`
`get Collection`	`4404` ops/sec (`±0.30%`)	`4607` ops/sec (`±0.27%`)	`1.05`
`get Query-sorted`	`5204` ops/sec (`±0.26%`)	`5287` ops/sec (`±0.40%`)	`1.02`
`setLong`	`440` ops/sec (`±0.38%`)	`431` ops/sec (`±0.19%`)	`0.98`
`setLongWithMerge`	`252` ops/sec (`±0.52%`)	`250` ops/sec (`±0.34%`)	`0.99`
`setLongWithSimpleMerge`	`268` ops/sec (`±0.66%`)	`266` ops/sec (`±0.46%`)	`0.99`
`setSmallResponse 500x`	`939` ops/sec (`±0.07%`)	`921` ops/sec (`±0.68%`)	`0.98`

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

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit fc6aeba. Configure here.}

* pkg: Bump peerdeps of @data-client/react to support 0.17 * pkg: Update yarn.lock for peerdep bump

Move the per-hit paths.slice(1) (and the hasStringDeps filter loop) out of GlobalCache.getResults and into the cache write. GlobalCache.paths() already produces the placeholder-free, function-free shape every consumer needs. Hand it to WeakDependencyMap.set as the journey, and the cache-hit branch can return that array by reference - no per-hit allocation, no per-hit typeof === 'function' walk. Safety: paths is now a shared reference held by every subsequent hit. The contract that consumers must not mutate it was established by the journey-mutation fix (PR #3925) and is exercised by the existing globalCache.test.ts regression test.

…#3929) * test(core): replace getResponseMeta-countRef with getResponseMeta-paths The previous integration test poked GCPolicy['entityCount'] directly to prove the journey-mutation bug. Replace it with a test that asserts the public-API consequence the bug creates outside of GC: every subscriber to the same endpoint must observe the same expiresAt from Controller.getResponseMeta(). This is the property the bug actually broke for non-GC users: with the buggy paths.shift(), entityExpiresAt(paths, …) iterates a progressively-shorter list, dropping the entity with the earliest expiry first. Subscriber 2 observes a too-late expiresAt; subscriber 3+ observe Infinity and never refetch. Fires under ImmortalGCPolicy too, since entityExpiresAt is unconditional whenever the endpoint has no top-level meta.expiresAt — typical for state populated via controller.set(Entity, …), SSR hydration, or useQuery. Verified the assertion fails on the buggy paths.shift() (m3.expiresAt returns FOO_2_EXPIRY instead of FOO_1_EXPIRY) and passes on the fix. * test(core): keep existing countRef integration test alongside paths test Restore the GC-side getResponseMeta-countRef.ts integration test that the prior commit replaced. The two tests cover the journey-mutation bug from complementary angles: - getResponseMeta-countRef.ts: GC consumer of paths (entityCount under-counting → premature reaping under default GCPolicy). - getResponseMeta-paths.ts: non-GC consumer of paths (entityExpiresAt → suppressed entity-expiry refetch; fires under ImmortalGCPolicy too). Both pass on the fix; both fail on the buggy paths.shift().

Base automatically changed from cursor/scalar-schema-design-for-lenses-d6e3 to master April 28, 2026 05:10

chore: rebase PR #3925 onto latest master

fc6aeba

ntucker force-pushed the fix/journey-mutation-on-result-cache-hit-scalar branch from ce58191 to fc6aeba Compare April 28, 2026 05:16

github-actions Bot reviewed Apr 28, 2026

View reviewed changes

cursor Bot reviewed Apr 28, 2026

View reviewed changes

Comment thread packages/endpoint/src/schemas/Scalar.ts Outdated

pkg: Bump peerdeps of @data-client/react to support 0.17 (#3927)

5ea71dc

* pkg: Bump peerdeps of @data-client/react to support 0.17 * pkg: Update yarn.lock for peerdep bump

ntucker mentioned this pull request Apr 28, 2026

perf(normalizr): store consumer-facing journey at write time #3928

Merged

ntucker mentioned this pull request Apr 28, 2026

fix(normalizr): do not mutate cached journey on result-cache hit #3924

Closed

2 tasks

fix: Scalar reversion

059b0e3

ntucker mentioned this pull request Apr 28, 2026

test(core): add getResponseMeta-paths integration test (non-GC angle) #3929

Merged

ntucker merged commit 6e8e499 into master Apr 28, 2026
27 checks passed

ntucker deleted the fix/journey-mutation-on-result-cache-hit-scalar branch April 28, 2026 13:16

This was referenced Apr 28, 2026

Version Packages #3926

Merged

Version Packages erickirt/data-client#82

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(normalizr): do not mutate cached journey on result-cache hit#3925

fix(normalizr): do not mutate cached journey on result-cache hit#3925
ntucker merged 5 commits intomasterfrom
fix/journey-mutation-on-result-cache-hit-scalar

ntucker commented Apr 28, 2026 •

edited by cursor Bot

Loading

Uh oh!

changeset-bot Bot commented Apr 28, 2026 •

edited

Loading

Uh oh!

vercel Bot commented Apr 28, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Apr 28, 2026 •

edited

Loading

Uh oh!

github-actions Bot left a comment •

edited

Loading

Uh oh!

github-actions Bot commented Apr 28, 2026 •

edited

Loading

Uh oh!

github-actions Bot left a comment •

edited

Loading

Uh oh!

cursor Bot left a comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

ntucker commented Apr 28, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Why it matters

Empirical confirmation

Differences from #3924 (master version)

Test plan

Performance

Read path (cache-hit benchmarks, n=5 trimmed mean, ops/sec)

Write path (write-then-discard benchmarks)

core suite (n=3, fresh-thermal re-run)

Drain-free fair microbench (one fresh MemoCache per iter → 1 miss + 1 hit)

V8 deopt analysis

Bundlesize

Changeset

Uh oh!

changeset-bot Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

vercel Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

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 commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

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!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

ntucker commented Apr 28, 2026 •

edited by cursor Bot

Loading

`core` suite (n=3, fresh-thermal re-run)

Drain-free fair microbench (one fresh `MemoCache` per iter → 1 miss + 1 hit)

changeset-bot Bot commented Apr 28, 2026 •

edited

Loading

vercel Bot commented Apr 28, 2026 •

edited

Loading

codecov Bot commented Apr 28, 2026 •

edited

Loading

github-actions Bot left a comment •

edited

Loading

github-actions Bot commented Apr 28, 2026 •

edited

Loading

github-actions Bot left a comment •

edited

Loading