perf(signals,web): omit over a merge holds the merge record — one record per layer, one-pass owners walk for ssrElement

[ryansolid]

Part of #3389 (item 3 of the plan: "omit over a merge: nested record vs per-leaf flattening").

What

An omit() over a merge() used to flatten at construction: one OmitView plus one combined hidden-key list per flattened leaf, and the next merge() copied those entries into its own arrays. On a Kobalte-shaped component chain (defaults + omit + spread, four layers) that was ~19 records and as many list copies per element — the largest allocation of the render. #3487 tried to make those copies cheaper and could not beat slice+push on instruction count; this PR does not make them.

The omit now holds the MergeView record itself (new source kind SOURCE_MERGE) and is one record however many leaves the merge has. A later merge() carries it as one entry; a later omit() folds into it. Nothing on the way is a proxy trap: sourceKeys/sourceHas/sourceGet, descriptors, hasStaticKeys and the resolved tables recurse into the record by function call. The property #3454 established — consumers read the leaves, never through a proxy — is kept; the per-leaf copies are gone.

Three things had to hold for it to pay, each found by measuring a variant that didn't:

1. One walk per read. A nested entry answers presence and value together (MISSING sentinel) — has-then-get per level was (2^{depth}) walks.
2. Inner records don't count reads toward their own 16-read table threshold, and the outer view's table is collected in one pass over the leaves (collectTable with a filter stack) — the naive recursion built a Map per layer (14 at depth 7 where flat built 2).
3. sourceOwners(source, keys, owners) (@internal): every key of a plain object, store, or view in merged order with its owning object, in one pass. ssrElement collects any spread that isn't plain-objects-only — a view, a store, or the array form with one among them — this way and reads owners[i][keys[i]] directly: the flat form's read cost without its construction cost. pushEntry is gone from server.ts.

Semantics to note in review:

• An omit's $SOURCES answers nothing now (previously its filtered leaf views); consumers reach the record via viewOf and walk it as one filtered entry. merge() re-merging an omit carries the record.
• Reads through an omit no longer count toward the inner merge's table threshold — the view that was asked decides for the whole tree. On the client that is the outer view the component reads, which still crosses the threshold and builds its (single-pass) table.

Measured

Against next, interleaved, min of N, quiet machine, compiled control within ±1.3%:

	next → this
tier-1 polymorphic-chain SSR, allocation	14.6 → 12.8 KB/row (−12%)
tier-1 polymorphic-chain SSR, time	−5.6% interpreter · −4.8% Sparkplug · −2.0% Maglev · −2.4% TurboFan
props-chain build (depth 1/3/7, interpreter)	−35% / −48% / −65%
props-chain build+consume (depth 7)	−31% interpreter · −26% Sparkplug · −9% TurboFan
omit/merge micro-suite (min of 3)	22 shapes better >5%, 77 within ±5%, 2 in the noise floor on code-identical paths
yak-bench SSR, all-primitives lane	+7% geomean; multifile-composition +36%, tabs +35%, polymorphic-chain +20%

On yak-bench that brings the all-Solid-primitives lane to parity with yak's hand-rolled runtime on the composition cases (0.74× → 0.99×) and moves the React gap on them from 0.36–0.39× to 0.50–0.51× (polymorphic-chain 0.24× → 0.29×). Element-dense cases move 0–4%: they are one or two layers deep and had little nesting to save.

The interpreter-tier results are the closest local proxy to CodSpeed's instruction counts (where #3487 lost); this is the tier where the change wins most.

Tests

Three structural tests in utilities.test.ts that asserted the flattened shape are rewritten to the nested one, with new assertions for filter scoping (a layer's filter applies to its own contribution only), merged key order through nesting, $SOURCES not leaking the unfiltered merge (#3014), one table per asked view, and sourceOwners. All suites green: signals 2 734, web server 986 / client 825 / hydrate 185, solid 620, universal 56. Size scenarios within limits.

[changeset-bot]

🦋 Changeset detected

Latest commit: 328580f

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
@solidjs/web	Patch
test-integration	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 35162887846

Warning

Build has drifted: This PR's base is out of sync with its target branch, so coverage data may include unrelated changes.
Quick fix: rebase this PR. Learn more →

Coverage remained the same at 71.46%

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:	1018
Covered Lines:	772
Line Coverage:	75.83%
Relevant Branches:	790
Covered Branches:	520
Branch Coverage:	65.82%
Branches in Coverage %:	Yes
Coverage Strength:	14.93 hits per line

---

💛 - Coveralls

[codspeed]

Merging this PR will degrade performance by 2.86%

⚠️ 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

⚡ 29 improved benchmarks
❌ 11 regressed benchmarks
✅ 135 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	merge	23 µs	1,441 µs	-98.4%
❌	omit	25.9 µs	50.8 µs	-49.06%
❌	merge	24.5 µs	30.8 µs	-20.61%
❌	merge	24.1 µs	30.3 µs	-20.54%
❌	merge	24 µs	29.9 µs	-19.74%
❌	ownKeys	254 µs	282.1 µs	-9.96%
❌	merge	23.4 µs	25.3 µs	-7.61%
❌	merge	23.4 µs	25.4 µs	-7.59%
❌	construct	28.8 µs	31.1 µs	-7.55%
❌	merge	23 µs	24.4 µs	-5.7%
❌	merge	24.2 µs	25.6 µs	-5.44%
⚡	ownKeys	2,773.1 µs	565.2 µs	×4.9
⚡	build + consume	841 µs	518 µs	+62.36%
⚡	build	283.2 µs	203.9 µs	+38.9%
⚡	build	163.2 µs	132.1 µs	+23.5%
⚡	merge	31.6 µs	26.2 µs	+20.34%
⚡	projection derive: write one NESTED field (reference)	2.1 ms	1.8 ms	+16.75%
⚡	build	155.2 µs	134.7 µs	+15.2%
⚡	build + consume	401.6 µs	360.5 µs	+11.39%
⚡	merge	28.7 µs	26.3 µs	+9.05%
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

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

---

_{Comparing perf/omit-nested-record (328580f) with next (d741d1e)}