perf(pm): extract requested core from full manifest parse

[elrrrrrrr]

Summary

Review-sized split from #3028 / #2948, stacked after #3031.

Adds the full-manifest parse helper used by the upcoming provider implementation:

• keeps parse_full_manifest_off_runtime as the compatibility wrapper;
• adds parse_full_manifest_with_core_off_runtime to parse the full manifest and optionally extract the current edge's core version in the same worker;
• preserves FullManifest::raw for later on-demand extraction;
• keeps native parsing on rayon and wasm parsing inline;
• adds a regression test for raw byte preservation and speculative latest extraction.

This PR does not change resolver scheduling yet; it only exposes the CPU helper consumed by the provider/job PR.

Size

• Diff against perf(pm): parse version manifests from vec buffers #3031: 1 file changed, 65 insertions(+), 15 deletions(-)

Validation

• cargo fmt
• cargo check -p utoo-ruborist
• cargo test -p utoo-ruborist service::manifest::tests::parse_full_manifest_with_core_extracts_requested_spec
• cargo clippy --all-targets -- -D warnings --no-deps

pack-napi still warns locally because next.js is a symlink in this worktree; clippy exits successfully.

[gemini-code-assist]

Code Review

This pull request refactors the manifest parsing logic to allow optional extraction of a core version manifest during the initial parsing task. It introduces parse_full_manifest_with_core_off_runtime and a corresponding synchronous helper. Feedback suggests optimizing the path where a version specifier is provided to avoid redundant JSON parsing and memory allocations by using a DOM-based approach to extract both the full manifest and the specific version manifest in a single pass.

[gemini-code-assist]

This implementation performs two full JSON parses and an extra memory allocation when a spec is provided. manifest.get_core_version calls extract_version, which clones manifest.raw into a new Vec and re-parses it using simd_json::to_borrowed_value.

Since this PR is focused on performance, you can optimize the Some(spec) path by parsing the JSON once into a simd_json::BorrowedValue DOM. From this DOM, you can both deserialize the FullManifest and extract the specific CoreVersionManifest without re-parsing or re-allocating the raw bytes.

For the None case, keeping the direct simd_json::serde::from_slice is preferred as it is generally faster than building the intermediate DOM.

fn parse_full_manifest_with_core_sync(
    raw_bytes: Bytes,
    spec: Option<String>,
) -> Result<FullManifestParseResult, anyhow::Error> {
    use serde::Deserialize;
    // simd_json mutates the parse buffer; copy inside the worker so
    // response-body bytes can stay immutable until parsing starts.
    let mut parse_buf = raw_bytes.to_vec();

    let (mut manifest, speculative) = if let Some(spec) = spec {
        use simd_json::prelude::ValueObjectAccess;
        let value = simd_json::to_borrowed_value(&mut parse_buf)
            .map_err(|e| anyhow!("JSON parse error: {e}"))?;

        let manifest = FullManifest::deserialize(&value)
            .map_err(|e| anyhow!("JSON mapping error: {e}"))?;

        let speculative = resolve_target_version((&manifest).into(), &spec)
            .ok()
            .and_then(|version| {
                let version_obj = value.get("versions")?.get(version.as_str())?;
                let core = CoreVersionManifest::deserialize(version_obj).ok()?;
                Some((spec, core))
            });

        (manifest, speculative)
    } else {
        let manifest: FullManifest = simd_json::serde::from_slice::<FullManifest>(&mut parse_buf)
            .map_err(|e| anyhow!("JSON parse error: {e}"))?;
        (manifest, None)
    };

    manifest.raw = raw_bytes;
    Ok((manifest, speculative))
}

[github-actions]

📊 pm-bench-phases · aa977d7 · linux (ubuntu-latest)

Workflow run — ant-design

PMs: utoo (this branch) · utoo-npm (latest published) · bun (latest)

npmjs.org

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	9.11s	0.03s	10.30s	10.02s	719M	337.8K
utoo-next	7.83s	0.29s	10.29s	12.00s	1.01G	123.0K
utoo-npm	8.07s	0.09s	10.61s	12.30s	1022M	123.2K
utoo	8.28s	0.44s	10.32s	11.96s	1010M	124.1K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	17.0K	19.0K	1.19G	7M	1.86G	1.75G	1M
utoo-next	129.1K	95.0K	1.16G	5M	1.71G	1.70G	2M
utoo-npm	143.7K	103.9K	1.16G	5M	1.71G	1.70G	2M
utoo	130.5K	80.4K	1.16G	6M	1.71G	1.70G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	2.06s	0.02s	3.90s	1.12s	495M	167.0K
utoo-next	2.89s	0.02s	5.00s	1.84s	615M	89.2K
utoo-npm	3.06s	0.05s	5.20s	2.21s	612M	82.3K
utoo	2.92s	0.09s	5.03s	1.86s	609M	77.6K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	10.3K	4.2K	202M	3M	107M	-	1M
utoo-next	51.0K	69.3K	200M	2M	7M	3M	2M
utoo-npm	77.0K	91.4K	200M	3M	7M	3M	2M
utoo	51.8K	71.4K	200M	3M	7M	3M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	6.73s	0.45s	6.29s	9.55s	610M	206.6K
utoo-next	7.48s	2.62s	5.00s	10.57s	475M	61.8K
utoo-npm	6.43s	0.68s	5.08s	10.43s	472M	60.8K
utoo	6.58s	0.82s	5.08s	10.27s	517M	64.6K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	6.8K	7.2K	1019M	4M	1.76G	1.76G	1M
utoo-next	116.2K	52.5K	988M	3M	1.70G	1.70G	2M
utoo-npm	107.8K	50.4K	988M	3M	1.70G	1.70G	2M
utoo	96.4K	53.3K	988M	3M	1.70G	1.70G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	3.35s	0.03s	0.20s	2.40s	133M	32.2K
utoo-next	2.15s	0.08s	0.47s	3.69s	80M	18.4K
utoo-npm	2.23s	0.07s	0.46s	3.71s	79M	18.6K
utoo	2.15s	0.04s	0.46s	3.70s	79M	18.2K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	216	25	5M	18K	1.91G	1.75G	1M
utoo-next	41.5K	19.0K	324K	24K	1.70G	1.70G	2M
utoo-npm	39.5K	18.3K	326K	9K	1.70G	1.70G	2M
utoo	40.2K	18.3K	325K	8K	1.71G	1.70G	2M

npmmirror.com: no output captured.