perf(signals): shared traps for merge()/omit() proxies, state on the target

[ryansolid]

Port of #3391 (main, mergeProps/splitProps) to the 2.0 primitives in @solidjs/signals.

What changes

Each merge() / omit() proxy used to allocate a target holding three fresh closures (get/has/keys) over that instance's sources, with the shared propTraps forwarding every trap call through them. Now the per-instance state lives on the target under symbol keys — { [$SOURCES]: flattened } for merge, { [$OMIT_PROPS]: props, [$OMIT_KEYS]: keys } for omit — and each primitive has one shared handler that reads it directly.

// before: 1 proxy + 1 target + 3 closures per instance, trap → closure → sources
new Proxy({ get(p) {…}, has(p) {…}, keys() {…} }, propTraps)
// after: 1 proxy + 1 one/two-slot object, trap → sources
new Proxy({ [$SOURCES]: flattened }, mergeTraps)

Semantics are unchanged: same right-to-left in lookup, same has/ownKeys filtering, $SOURCES still answers the flat list on a merge proxy (what mergeSources()/spread() read) and undefined on an omit proxy (#3014). The state keys are never reported by ownKeys nor answered by get/has, so they're invisible through the proxy; a string Object.defineProperty on the proxy lands on the target without touching them (before, defining "get" would have overwritten the target's method and broken the proxy). Proxy invariants hold: the target is extensible and its state keys configurable.

Tests

New test: state symbols not visible, Object.defineProperty(merged, "sources", …) / (rest, "props", …) don't affect reads or in/keys, live reads after a store write, descriptor getters work. Existing #3014 test and the whole merge/omit suite unchanged. Suites: signals 1768, solid 595, html 200, web dom 735 / server 795 / hydrate 169.

Performance

Measured on battery, so only directions are reliable (±20% run-to-run). Interleaved A/B, old vs new dist/prod in one process, 31 alternating rounds:

path	ratio new/old
retained 200-instance omit(merge(defaults, store), "id") batch	0.87 / 0.93 (two runs)
nested merge(mergeProxy, {x})	0.65 / 0.75
Object.keys / {...proxy}	0.90–0.98
single construct, single reads, in	within noise (process-isolated alternating runs show no regression)

A computed-symbol literal costs ~13 ns vs ~46 ns for the three-closure literal in isolation, so the construct win is real but small next to the Proxy itself; the retained batch is where it shows. CodSpeed's merge-proxy-keys / omit-proxy construct+read benches in utilities.bench.ts cover this on the PR.

[changeset-bot]

🦋 Changeset detected

Latest commit: 833cbdd

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
test-integration	Patch
@solidjs/web	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
@solidjs/diagnostics	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
solid-js	Patch
@solidjs/universal	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 34730721456

Coverage remained the same at 71.842%

Details

• Coverage remained the same as the base build.
• Patch coverage: No coverable lines changed in this PR.
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	1007
Covered Lines:	772
Line Coverage:	76.66%
Relevant Branches:	790
Covered Branches:	519
Branch Coverage:	65.7%
Branches in Coverage %:	Yes
Coverage Strength:	15.02 hits per line

---

💛 - Coveralls

[codspeed-hq]

Merging this PR will degrade performance by 24.76%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
❌ 9 regressed benchmarks
✅ 150 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	merge	274.5 µs	2,357.8 µs	-88.36%
❌	construct	34.8 µs	40.9 µs	-14.74%
❌	construct	34.8 µs	40.6 µs	-14.33%
❌	construct	37.7 µs	43.6 µs	-13.47%
❌	construct	46.4 µs	52.1 µs	-11%
❌	construct	56.6 µs	62.3 µs	-9.15%
❌	construct	56.8 µs	62.3 µs	-8.84%
❌	construct	57 µs	62.3 µs	-8.46%
❌	construct	58.6 µs	63.4 µs	-7.56%
⚡	merge	109 µs	86 µs	+26.69%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing perf/shared-prop-traps (833cbdd) with next (084e621)}

[ryansolid]

Closing: the premise doesn't hold up under CodSpeed. Against `next`, every shape this was meant to speed up regresses — `omit-proxy-store` construct −11% to −15%, `merge-proxy-keys-store` construct −7.5% to −9%, and `merge-merge-signal(deep 0,100,3,2)` at −88% (275 µs → 2.36 ms). The one win (`merge-static(0,100)` +27%) is on a path this PR doesn't touch, so it's noise. The local wall-clock A/B in the description was measured on battery at ±20%, which is wider than the effect it was trying to detect; the instruction-count numbers are the ones to trust here.

The plausible reason: three closures over locals in a fresh object literal are cheap and monomorphic, while symbol-keyed state read off `target` inside a shared handler adds a keyed load per trap and turns the target's shape into something V8 handles worse than the closure literal. The nested-merge outlier suggests the flatten path over `$SOURCES` became a slow-property read on every hop.

If someone wants to revisit, it needs to start from a CodSpeed-visible win, not a local timing run.

— Claude via Cursor