[yangshun]

Problem

Static prerendering made two requests per page: one for the HTML, and a second with X-VoidPages to write _void/pages/*.json. The loader and middleware ran twice for every page.

Fix

Fixes #500

The HTML already embeds the page object in __VOID_PAGE_DATA__, so the prerenderer now writes the JSON from that script and skips the second request.

• Uses extractPageDataFromHtml from @void/isr/core, which ISR already uses to build JSON from page HTML.
• Pages with deferred props still make the second request. Their HTML has those props missing, because the values are streamed after the page data script. The prefetch request returns them filled in.
  • In practice, defer() is mainly useful for server-rendered pages. A fully static site rarely needs it, so on most static sites the two implementations behave the same. IMO, this is an acceptable fix. The alternative is to
• Island pages and non-page routes have no __VOID_PAGE_DATA__ script, so they still get no JSON file.
• The JSON is parsed and stringified again, so the files are byte-identical to before. The embedded copy escapes <, > and &.

Results

I ran two benchmarks, once with a 1000-page Void app (no actual network requests), another with a remote PayloadCMS database.

1000-page Void app

Measured with my own 1000-page Void app benchmark, static output. Old and new code ran alternately in one process on the same build:

	Before	After
Requests	2,002	1,001
Prerender phase, Node	422–477 ms	380–396 ms (−10% to −19%)
Prerender phase, Cloudflare	848–1,660 ms	630–1,149 ms (−26% to −31%)

On Node, the second request was an in-process call, so a full build barely changes: 4.99 s before vs 4.64 s after, median of 3, within noise. The gain is larger on Cloudflare, where each request goes through the local Workers runtime.

The built output (1,001 HTML and 1,001 JSON files) is byte-identical to before on both Node and Cloudflare.

100-page blog backed by remote Payload CMS

Tested in a more realistic scenario of a 100-post blog backed by a remote Payload CMS database, where each pre-rendered page fetches from Payload CMS:

	Before	After
Build time	265.00s	133.31s
Payload operations	1,439	722

Tests

• Updated two existing tests that served page data only through the second request.
• Added one test for deferred pages, which still need the second request, and for HTML without page data, which gets no JSON file.
• static-pages and node-static integration tests pass.

Alternative considered

Another approach handles deferred props in the same single request. During prerendering, the page handler copies each deferred value into the page object as it streams, then appends the completed object to the end of the HTML as a build-only script.

The prerenderer removes that script and writes it as the JSON file. This avoids the second request even for deferred pages, but it changes handlePageGet, which runs on every page request in every app, and the appended script has to pass through the app's HTML middleware unchanged.

I prefer to keep the change in build-time code instead. Deferred props are uncommon in fully static pages, so those pages keeping the second request costs little.

[Brooooooklyn]

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-20T09:22:02.077668Z	846205a	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Delightful!

Reviewed commit: 846205ab0c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".