refactor(app-browser-entry): dedupe encoded JSON header parsing

[james-elicx]

Summary

• Extract the repeated try { JSON.parse(decodeURIComponent(headerValue)) } catch { ... } pattern used to hydrate params from the X-Vinext-Params header into a single local helper, parseEncodedJsonHeader<T>(value: string | null): T | null.
• Both call sites in packages/vinext/src/server/app-browser-entry.ts (the initial RSC hydration path and the in-flight navigation path) now share the helper.
• Pure refactor — no behavior change.

Follow-up to #1049.

Files changed

• packages/vinext/src/server/app-browser-entry.ts — adds the helper at the top of the file and updates the two call sites.

Notes on behavior preservation

The two original blocks had a subtle difference:

• The first call site (initial RSC response) parsed the header and called applyClientParams(params) inside the same try, so the catch silently swallowed any error from applyClientParams as well.
• The second call site only parsed.

The helper itself only handles parsing. To preserve the original error-swallowing semantics at the first call site, applyClientParams is wrapped in a small dedicated try/catch there. The helper does not log or surface errors — silent by design, matching the pre-existing behavior.

A quick grep -rn "JSON.parse(decodeURIComponent" packages/vinext/src/ confirmed the pattern only appears in this file, so a module-private helper (rather than a shared utility) is sufficient.

Test plan

• pnpm vp test run tests/app-router.test.ts — 308 tests pass
• pnpm fmt --write on the touched file
• Manual diff review to confirm structural-only change

🤖 Generated with Claude Code

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/vinext@1070

commit: 7e0b27c

[github-actions]

Example	Preview	Production	Original
app-router-cloudflare	preview	production
pages-router-cloudflare	preview	production
app-router-playground	preview	production	original
realworld-api-rest	preview	production
nextra-docs-template	preview	production
benchmarks	preview	production
hackernews	preview	production	original

[james-elicx]

/bigbonk review

[ask-bonk]

Clean refactor. The helper is well-scoped, the two call sites read better, and the behavioral nuance (separating applyClientParams from the parse try/catch at the first call site) is handled correctly.

Two minor observations, neither blocking:

1. !value guard treats "" as missing — parseEncodedJsonHeader("") returns null because the empty string is falsy. That's fine here since headers.get() returns null for absent headers and an empty X-Vinext-Params value would fail JSON.parse anyway, so the short-circuit is just saving a wasted decode+parse. Worth noting in case the helper is ever promoted to a shared utility where an empty string might carry different semantics.

2. The generic as T cast is unchecked at runtime — the caller trusts that the JSON payload matches Record<string, string | string[]>. That's inherited from the original code and not a regression, just a note that if this helper ever gets reused for other header shapes, a runtime schema check (or at least a comment about the trust boundary) would be warranted.

LGTM — all CI green, behavior preserved, code is cleaner.

[ask-bonk]

Review posted and PR approved. The refactor is straightforward and correct:

• The parseEncodedJsonHeader<T> helper cleanly deduplicates the repeated try { JSON.parse(decodeURIComponent(...)) } catch pattern from both call sites.
• The subtle behavioral difference (first call site previously swallowed applyClientParams errors inside the same try/catch) is properly preserved with a dedicated try/catch wrapper.
• All 30 CI checks pass. No behavior change, just cleaner code.

github run