perf(server): cache App Router route wiring plans

[NathanDrake2406]

Overview

Item	Detail
Goal	Reduce App Router SSR route wiring overhead shown in CPU profiles.
Core change	Cache route-invariant App Router wiring plans in a WeakMap keyed by the generated route object.
Main boundary	Keep params, dependency barriers, and React element creation per request while hoisting stable layout/template/error/slot derivations.
Primary file	packages/vinext/src/server/app-page-route-wiring.tsx
Expected impact	Lower buildAppPageElements self-time on warm SSR requests.

Why

Generated App Router route objects are stable for the server process, but the SSR path rebuilt derived layout, template, error, tree-position, and slot metadata for every request. Those derivations are route-invariant, while params and render dependencies remain request-scoped.

Area	Principle / invariant	What this PR changes
Route wiring	Generated route structure is stable after startup.	Adds a per-route cached wiring plan for derived entries, maps, ids, slot entries, and source page data.
Request render state	Params, dependency barriers, and React elements are request-scoped.	Leaves these on the request path.
Slots/templates	Empty feature sets should not allocate support structures per request.	Avoids slot dependency snapshots and template maps when the matched route does not use them.

Profiling

Warm SSR profile of benchmarks/vinext at /, built with --sourcemap --minify false, using 1000 warmup requests plus 20000 measured requests.

Metric	Before	After	Change
buildAppPageElements self-time	85.83ms	38.75ms	-54.9%
app-page-route-wiring source self-time	104.88ms	48.62ms	-53.6%
20k measured HTTP wall time	13.49s	10.62s	-21.3%

CPU self-time is the primary signal here. HTTP timing is included as supporting context only.

Validation

Commands run

vp check tests/app-page-route-wiring.test.ts tests/app-page-element-builder.test.ts tests/app-static-siblings.test.ts
vp test run tests/app-page-route-wiring.test.ts tests/app-page-element-builder.test.ts tests/app-static-siblings.test.ts
vp run vinext#build
node benchmarks/generate-app.mjs
cd benchmarks/vinext && vp build --sourcemap --minify false

The final pre-commit hook also ran staged vp check --fix, full check, and knip.

Risk / Compatibility

Notes

• Runtime behaviour should be unchanged. The cache stores only route-invariant structure derived from the generated route object.
• The cache is a WeakMap, so route objects can still be collected if a dev/runtime graph drops them.
• Per-request state remains per request: matched params, slot override params, render dependencies, search params, and React elements are not cached.
• This does not include the separate safe JSON or Vary optimizations from PRs perf(app-router): trim AppElements Flight metadata #2176 and perf(server): skip duplicate App RSC vary merge #2177.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/@vinext/cloudflare@2179

npm i https://pkg.pr.new/vinext@2179

commit: 2f4da1e

[github-actions]

Performance benchmarks

Compared 2f4da1e against base 94eb44e using alternating same-runner rounds. Next.js was unchanged and skipped.

0 improved · 0 regressed · 6 within ±1.5%

Scenario	Framework	Baseline	Current	Change
Client bundle size (gzip)	vinext	132.8 KB	132.8 KB	⚫ -0.0%
Client entry size (gzip)	vinext	118.6 KB	118.6 KB	⚫ -0.0%
Dev server cold start	vinext	2.34 s	2.32 s	⚫ -0.5%
Production build time	vinext	2.79 s	2.81 s	⚫ +0.7%
RSC entry size (gzip)	vinext	60.1 KB	60.3 KB	⚫ +0.4%
Server bundle size (gzip)	vinext	175.4 KB	175.6 KB	⚫ +0.1%

View detailed results and traces

_{🟢 improvement · 🔴 regression · ⚫ change below 1.5% · paired base/head}