[dinwwwh]

Requested by dinwwwh · project thread

What

Simplifies #312, following the request for the simplest approach with minimal logic.

#312 made a $ref into a tuple's items, or into an additionalProperties beside patternProperties or of an enum-keyed record, be replaced by its target, since the conversion puts another schema there and a $ref kept as written would read that one. To tell those apart from other items and additionalProperties, isIntoReplaced walked every matching segment of the pointer and resolved and inspected each holder.

Now every $ref with an /items or /additionalProperties segment is replaced by its target, with one regex test in isInlined. Replacing a $ref by its converted target accepts the same values, so nothing gets stricter.

README: the note moves from the prefixItems, patternProperties and enum-keyed propertyNames rows to the $defs row, which already says how a replaced $ref converts.

Tests

The #312 test now also expects a $ref into the items of a plain array to be replaced, and still expects one into a property named itemsLeft to stay. The enum-record test expects a $ref into a plain record's additionalProperties to be replaced too. pnpm check, pnpm test and coverage (100%) pass.

Tradeoffs

• Output. A $ref into the items or additionalProperties of any schema now comes out as a copy of its target instead of a $ref. Generators such as zod-to-json-schema point a reused schema at its first use this way, so a code generator may now see an inline schema where it saw a shared one. Semantics are unchanged.
• Speed. Under callgrind with CodSpeed's V8 flags, one warmed conversion of the 10-resource generated API goes from 17,788,323 to 17,459,992 instructions (−1.8%). The order schema is unchanged.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed]

Merging this PR will not alter performance

1 warning

⚠️ 1 benchmark is dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 29 untouched benchmarks
⏩ 2 skipped benchmarks¹

---

_{Comparing claude/project-thread-h3q238 (dd53f1c) with main (6c595c5)}

Footnotes

1. 2 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[pullfrog]

ℹ️ Minor suggestion only — the simplification itself looks sound.

Reviewed changes

• isInlined broadened — packages/downgrader/src/v3.1-to-v3.0.ts: replaces isIntoReplaced's per-segment pointer walk with INTO_REPLACED = /\/(?:items|additionalProperties)(?:\/|$)/, so any resolvable $ref with an /items or /additionalProperties path segment is now inlined rather than kept.
• README note relocated — packages/downgrader/README.md: the "$ref into items/additionalProperties is replaced by its target" line moves from the prefixItems/patternProperties/propertyNames rows to the $defs row.
• Tests updated — packages/downgrader/tests/v3.1-to-v3.0.test.ts: a $ref into a plain array's items now expects inlining, and a plain-record additionalProperties inline case is added.

I verified the simplification is value-preserving for the intended cases and that the prefix-collision guard (itemsLeft) works. The one edge I'd call out is the exact-name match, detailed inline.

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