perf(compiler): SSR props literals with getters compile to hoisted constructors (hoistProps)

[ryansolid]

What

With hoistProps (default on) both compilers emit a component's props literal that has getters as a module-level constructor instead of an object literal:

// before (per instance: dictionary-mode object + a closure per getter)
createComponent(Button, { get active() { return sig(); }, variant: "ghost", children: i })

// after (per instance: one fast-mode object with symbol slots; getters shared on the prototype)
const _p$ = Symbol(); function _P$(_p) { this[_p$] = _p; this.variant = "ghost"; ... }
Object.defineProperty(_P$.prototype, "active", { get() { return this[_p$](); }, enumerable: true, configurable: true });
...
createComponent(Button, new _P$(sig))

Own keys, order, descriptors and prototype are unchanged, so Object.keys, spread, hasOwn, isStatic() and hasStaticKeys() answer as before. Sites that close over a reassigned or later-declared binding, this/super/arguments, or sit at module level keep the literal. Babel (src/ssr/props.ts) and the native compiler (src/ssr/props.rs) emit byte-identical output.

Tracker: #3389 · design/issue: #3511

The contract that changes

A props getter is defined only for a read through its own object. Copying its descriptor onto another object (Object.defineProperty(target, k, Object.getOwnPropertyDescriptor(props, k))) no longer works — the getter reads its capture through the receiver. In dev the getter throws naming the rule.

• Our primitives never copy descriptors, except omit()'s no-Proxy fallback, which signals: omit()'s copy path re-homes accessors with the source as receiver #3544 fixes to re-home accessors. Merge signals: omit()'s copy path re-homes accessors with the source as receiver #3544 first.
• next-yak's combineProps/targetProps do copy descriptors; the solid-merged yak-bench lane crashes under this output (Cannot read properties of undefined (reading 'group')). Their fix is a one-line re-home in copyProps, or moving to merge(). Heads-up to next-yak is timed with the release that flips this on.

Measured

yak-bench, SSR in-process renderToString, instances/sec, native compiler, mprim lane (Solid primitives), hoistProps on vs off, same run alternated. Round 1 is the clean pair; round 2 was disturbed mid-run (react control dropped 30% on tabs), so the two-round average is shown as the conservative figure.

case	off → on (round 1)	2-round avg
polymorphic-chain	1.31×	1.12×
product-grid	1.25×	1.09×
multifile-composition	1.27×	1.20×
multifile-shop	1.29×	1.23×
realistic-button	1.19×	1.09×
tabs	1.15×	1.15×
dyn-inline	1.46×	1.47×
btn-variant, dyn-fair, dyn-translate, button-variants*, compose-* (no getter literal on the path)	0.98–1.04×	0.88–1.13×
geomean (15)	1.12×	1.09×

Octane ssr-throughput (news-50/500) and streaming-ssr Solid fixtures: flat on/off — element-heavy, no getter literals on the hot path — so no regression on the board.

Against yak's own hand-rolled merged runtime (last clean rc.8 run, same machine, indicative): mprim + hoisting now leads on every multi-component page (polymorphic-chain 1.49×, product-grid 1.42×, multifile-composition 1.27×, tabs 1.24×) and trails on the single-styled-element micro-cases (btn-variant / dyn-fair / dyn-translate 0.57×, realistic-button 0.86×).

Merge gate

That last row is the honest cost of the contract: once descriptor copying is off the table, a library like yak has to come through merge()/ssrElement, and today those trail their eager copy on the micro-cases. A profile of btn-variant puts the gap entirely in the runtime (props view ~40%, ssrElement multi-source path ~23%, attribute-name escaping ~13%, server memo owner ~10%); none of it is the compiler. That runtime work is in progress with btn-variant parity against merged as the exit criterion. hoistProps should not default on in a published rc before that lands; if it doesn't make the same rc, flip the default to false in config.ts / config.rs before release.

Tests

• packages/babel-plugin/test/ssr-props.spec.js — hoisting rules, capture rules, name generation, and execution of the emitted code (getter semantics, own-key parity, dev receiver error).
• Fixture outputs updated in both compilers; the compiler parity suite covers the byte-identical output.
• pnpm test in packages/babel-plugin (264) and packages/compiler (5785) green.

[changeset-bot]

🦋 Changeset detected

Latest commit: 0f90eae

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

This PR includes changesets to release 11 packages

Name	Type
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
test-integration	Patch
@solidjs/diagnostics	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
@solidjs/signals	Patch
solid-js	Patch
@solidjs/universal	Patch
@solidjs/web	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 35401656264

Coverage remained the same at 71.366%

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:	1035
Covered Lines:	783
Line Coverage:	75.65%
Relevant Branches:	802
Covered Branches:	528
Branch Coverage:	65.84%
Branches in Coverage %:	Yes
Coverage Strength:	15.52 hits per line

---

💛 - Coveralls

[codspeed]

Merging this PR will degrade performance by 9.95%

⚠️ 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 regressed benchmark
✅ 171 untouched benchmarks
⏩ 3 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	polymorphic-chain: 200 rows (renderToString): chain	20.7 ms	23 ms	-9.95%

Tip

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

---

_{Comparing perf/ssr-props-constructors (0f90eae) with next (49887e1)}

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. ↩