fix(store): shallow store leaves stay raw in drafts, shadows and SSR snapshots (#3498)

[ryansolid]

The shallow half of #3498, extracted from #3499 (@Monkeylordz — the fix and the tests are yours, credited as co-author) and reduced to its minimum. The hybrid hydration resequencing in #3499 is deliberately not here; see the note at the end.

What was wrong

{ shallow: true } promises raw leaf references, but three projection paths treated a shallow store as deep:

• Draft. wrapDraft wrapped every nested value read through the derive's draft in another proxy — leaf identity lost, and an object-valued non-configurable property on a frozen leaf trips a Proxy invariant error.
• Copies. The loading shadow (seedLoadingValue), its commit copy, the SSR draft/snapshots and the hydration replay shadow were JSON.parse(JSON.stringify(...)) of the whole tree: Date → string, NaN → null, undefined properties dropped, BigInt and cycles throw.
• SSR draft. createDeepProxy proxied shallow leaves too.

Fix (3 source files, ~40 lines net)

• wrapDraft(..., shallow): a boolean that short-circuits the recursion. No indirect call; the nested recursion never happens for a shallow store, so nothing is passed down.
• cloneState(v, shallow): root container alone for a shallow store (slice() / spread), the tree otherwise. Used at the two client sites and four server sites. Mirrored in the solid server runtime rather than exported — an implementation detail with an ugly contract, and that layer is by design a reimplementation that already carries its own twins of this kind of logic.
• createDeepProxy(..., shallow) skips child proxies; createShadowDraft(draft, shallow) copies the root.

Deep stores are unchanged — this is a strict narrowing to make the projection paths obey what shallow already means for plain stores. It doesn't preclude replacing the shadow with staged writes or switching the deep copy to a value-preserving clone later; both would subsume this.

Perf

For shallow stores every touched path does strictly less work: no per-read proxy allocation in the draft (previously one Proxy + traps object per nested access), and a root slice()/spread instead of a tree clone in loading windows and hydration. Deep stores pay one target.s read per derive run. CodSpeed on #3499 was red on projection derive: delete + set one ROOT key, but cross-CPU (Xeon baseline vs EPYC head); this PR measures on one environment.

Verification

• 15 tests (his, minus the hybrid describes): 8 signals (shallow-loading), 7 solid (shallow-hydration replay, server/shallow-projection). All 15 fail on unfixed next, all pass here.
• Full suites: signals 3461, solid 633, web client 833 / server 1005 / hydrate 190. pnpm types clean.
• Size (brotli): hydrating (no stores) 20,208 → 20,260 B; hydrating + stores 30,303 → 30,469 B — caps ratcheted to 20.3 / 30.5 KB with notes. signals+store 16.65 → 16.72 kB under its 16.75 cap. CSR, CSR-observe and the signal-only floor flat.

Not included: hybrid hydration resequencing (#3499's second half)

Waiting for hydration end ∧ first server answer before takeover, committing later runs' first yields, and keeping a rejected server answer visible until refresh are rules about what a hybrid store shows during handoff. That needs a ruling before it can be reviewed as a bug fix; #3499 stays open for it, and its tests for that path are the right pins if it's ratified. #3498 therefore stays open after this merges, with its shallow section resolved.

Co-authored-by: Leon Schiffler
Co-authored-by: Claude via Cursor

[changeset-bot]

🦋 Changeset detected

Latest commit: e87d694

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

This PR includes changesets to release 11 packages

Name	Type
@solidjs/signals	Patch
solid-js	Patch
test-integration	Patch
@solidjs/web	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
@solidjs/universal	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
@solidjs/diagnostics	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

[coveralls]

Coverage Report for CI Build 35319995646

Coverage increased (+0.06%) to 71.366%

Details

• Coverage increased (+0.06%) from the base build.
• Patch coverage: 3 of 3 lines across 1 file are fully covered (100%).
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	1035
Covered Lines:	783
Line Coverage:	75.65%
Relevant Branches:	802
Covered Branches:	528
Branch Coverage:	65.84%
Branches in Coverage %:	Yes
Coverage Strength:	15.53 hits per line

---

💛 - Coveralls

[codspeed]

Merging this PR will not alter performance

✅ 172 untouched benchmarks
⏩ 3 skipped benchmarks¹

---

_{Comparing fix/shallow-store-leaves (e87d694) with next (e171fc2)}

Footnotes

1. 3 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩