[dinwwwh]

Requested by dinwwwh · project thread

Docs only. Five README notes for packages/downgrader, all found by the issue hunt on main 81010aa:

• Limitations, cycles: a not or oneOf outside a cycle can also reject values, when its $ref targets a cycle schema that lost a restriction only through a back $ref. Whether it does depends on which schema of the cycle comes first.
• Limitations, $defs size: each $ref into $defs repeats its target once serialized. SchemaStore's Kestra flow schema, 4 MB as compact JSON, would serialize to about 84 TB. In a document, components/schemas avoids this.
• swagger_dart_code_generator: takes only { type: "null" } as null, so a nullable $ref or enum (a oneOf with the "null" branch) becomes dynamic. Its 3.1 path keeps the class but stops on boolean schemas and fails on a nullable array.
• swagger_parser: in an object with no required list, one allOf property (a $ref with siblings) makes every property required and non-null, so decoding a value without an optional property throws. Its 3.1 path stops on boolean schemas.
• OpenApiSpex joins kin-openapi and openapi_parser on the line about rejecting null for an untyped schema.

All changed lines are list items, so there are no table cells to break. pnpm run check passes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01H41Joj1bcf6cWAwxSpC2c7

[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

✅ 21 untouched benchmarks

---

_{Comparing claude/project-thread-hjcqmq (912ae87) with main (81010aa)}

[pullfrog]

ℹ️ Minor suggestions only.

Reviewed changes — docs-only update to packages/downgrader/README.md; I verified the new third-party claims against each tool's source and confirmed the repo formatter still passes.

• $defs serialized size — new bullet explaining that inlining every $ref into $defs/definitions makes layered definitions expand when serialized (Kestra example), and recommending components/schemas.
• Cycle order — extends the existing cycle not/oneOf limitation to a not/oneOf outside the cycle, over a $ref to a cycle schema loosened only through such a $ref.
• swagger_dart_code_generator — new bullet on a 3.0 nullable $ref/enum becoming dynamic, a boolean-schema crash, and a nullable array that fails to build.
• swagger_parser — new bullet on a single allOf property forcing every property required and non-null, plus a boolean-schema crash.
• OpenApiSpex — added to the kin-openapi/openapi_parser bullet on rejecting null for an untyped {}.

I checked the three new third-party bullets against source and they hold: the dart generator only treats a literal type: "null" as null in its union path (swagger_models_generator.dart), and crashes casting a boolean items (swagger_schema.g.dart:20); swagger_parser's hasAllOfKey forces required/non-null (open_api_parser.dart:768-824) and crashes on a boolean schema (open_api_parser.dart:1324); OpenApiSpex's current Cast path rejects null for {} (cast.ex:157, ahead of the type: nil wildcard at :208). The only open point is the input size on the new $defs bullet, noted inline.

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

[pullfrog]

✅ No new issues found.

Reviewed changes — the delta since the prior pullfrog review is a single-line clarification to the $defs size bullet in packages/downgrader/README.md.

• $defs size bullet figures — states the Kestra input as "4 MB as compact JSON" and revises the projected output to "about 84 TB", resolving the prior review's note that the published kestra-0.19.0.json is ~7.6 MB and only ~3.98 MB once minified. The author's figures (3,975,214 bytes in; 83,683,942,216,949 bytes out) are internally consistent with the new wording, and no other lines changed.

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