fix(InlineModelResolver): do not deduplicate distinctly-named component schemas (#24177)

[Shaun-3adesign]

Fixes #24177 (a follow-on to the #23856 deduplication feature and the later #24192 fix).

Problem

deduplicateComponents() was introduced in #23856 to collapse numbered duplicates — e.g. FlowSegment_1 sitting alongside FlowSegment — that the Swagger Parser / inline resolver emit when a shared multi-file schema fails to be reused. Its removal logic, however, matched any two titled, structurally-identical component schemas.

Since 7.23.0 this also silently removes schemas the user authored by hand with distinct, non-numbered names. The reporter's minimal spec:

components:
  schemas:
    upper1:
      type: object
      properties: { foo: { type: string } }
      title: foo title identical by purpose
    upper2:
      type: object
      properties: { foo: { type: string } }
      title: foo title identical by purpose

On 7.22.0 both Upper1 and Upper2 are generated; on 7.23.0 → master, Upper2 disappears. Because both schemas have a title, the earlier title != null guards (in flatten() pre-population, matchGenerated(), and #24192) do not catch this case.

Fix

Two component keys can only both be non-numbered when the user deliberately named them that way — the parser always numbers its own collisions (_1, _2, …). So restrict deduplicateComponents() removal to numbered names:

if (name.matches(".*_\\d+$")) {
    duplicateToCanonical.put(name, canonical);
}

The deduplication keeps collapsing generated artifacts exactly as before, but never deletes a distinctly-named user schema. No new option/flag is added.

Verification

• New regression test deduplicateComponentsKeepsDistinctlyNamedUserSchemas — red before the fix (logs Removed duplicate component schema 'upper2'), green after.
• Full InlineModelResolverTest suite: 57/57 pass (JDK 21), including every existing numbered-duplicate test (Widget_1, FlowSegment, StorageBackend_1, ContainerMapping_1, fix(InlineModelResolver): do not merge untitled inline schemas differing only by description #24192).
• fix(InlineModelResolver): prevent numbered duplicate models from multi-file OAS 3.1 specs #23856's motivating multi-file spec (TAMS) regenerates byte-for-byte identical output (modulo generation timestamps) — the numbered-duplicate dedup still fires exactly as before.
• End-to-end: generating the reporter's minimal foo.yaml now emits both Upper1.java and Upper2.java.

No .mustache templates are touched — this is a code-generator logic change only.

/cc @wing328 @seonwooj0810

PR checklist

• Read the contribution guidelines.
• Run the following to build the project and update samples:

  ./mvnw clean package || exit
  ./bin/generate-samples.sh ./bin/configs/*.yaml || exit
  ./bin/utils/export_docs_generators.sh || exit
  

  No templates changed and the fix only withholds dedup from hand-named schemas, so no sample output is expected to change; deferring to CI to confirm across all generators.
• If your PR is targeting a particular programming language, @mention the technical committee members — n/a, this is a core (language-agnostic) change.

---

Summary by cubic

Fixes a bug where InlineModelResolver removed user-defined component schemas with distinct, non-numbered names. Deduplication now only collapses parser-generated numbered duplicates (e.g., Foo_1), preserving authored schemas; fixes #24177.

• Bug Fixes
  • Only deduplicate schemas with numbered names (matches .*_\\d+$, e.g., Foo_1).
  • Added regression test deduplicateComponentsKeepsDistinctlyNamedUserSchemas.
  • No template changes; behavior for numbered duplicates is unchanged.

^{Written for commit 2d95832. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}