[dinwwwh]

Requested by dinwwwh · project thread

What

map in packages/downgrader/src/shared.ts collected [key, value] pairs and then called Object.fromEntries (since #50). It now writes each entry straight into {} with setOwn, the same way convertObject and copy already do. setOwn keeps a __proto__ key as an own property, and key order is unchanged.

The existing __proto__ test in v3.1-to-v3.0.test.ts now also puts an integer key beside __proto__ in a properties map and checks the key order and that the map keeps a normal prototype. It fails if map uses a plain out[key] = converted.

Measurements

Wall time with the JIT, median of shuffled rounds with gc between runs:

• 3.1 to 3.0: bench doc about −15%, Discord −8%, Figma and GitHub −5%
• 3.2 to 3.1: bench doc about −20%
• graph benches (path item chain, reference diamond): same or faster

Output is byte-identical on 67 real docs and 800 fuzzed docs. 254 tests and pnpm run check pass.

---

Generated by Claude Code

[pullfrog]

✅ No new issues found.

Reviewed changes

• map builds output directly — packages/downgrader/src/shared.ts now writes each converted entry into a plain Record via setOwn instead of collecting pairs and calling Object.fromEntries, matching the existing convertObject/copy pattern.
• __proto__ test strengthened — the keeps own __proto__ keys as keys test now also places an integer key beside __proto__ and pins key order (['1','__proto__']) and the normal prototype.

The two paths are equivalent: Object.fromEntries also defines own data properties, so __proto__ was already handled as an own key, and integer-index enumeration order is spec-defined independently of insertion order. out was already a fresh local object, so returning it directly adds no aliasing. No concerns.

  ^{｜ View workflow run ｜ Using deepseek-v4.1-flash (free via Pullfrog for OSS) ｜ 𝕏}

[codspeed]

Merging this PR will improve performance by 13.3%

1 warning

⚠️ Base and head ran on different runner types

Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.

Learn more about Macro Runners

Open the report in CodSpeed to investigate

⚡ 4 improved benchmarks
✅ 17 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	dereferenced diamond, 64 levels	687.3 µs	583.1 µs	+17.86%
⚡	dereferenced diamond, 64 levels	1,114.2 µs	987.8 µs	+12.8%
⚡	cyclic callback graph in removed webhooks, 8 path items	1.2 ms	1.1 ms	+12.39%
⚡	path item $ref chain in removed components, 200 hops	5.7 ms	5.2 ms	+10.29%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing claude/project-thread-7aadhr (026d908) with main (d14fb04)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!