fix: check for type specified on referenced class

[MikeEdgar]

Fixes #2688

[phillip-kruger]

Looks good to me. The fix addresses the root cause: a @Schema on the field without type caused readSchemaAnnotatedField to apply the default type: object next to the $ref. I built the branch and ran the core tests (410 run, 0 failures). I also checked the issue's example in both scan orders (Zeta, Beta, Alpha and Alpha, Beta, Zeta). Both give zeta: {$ref, description} now, where main emitted type: object in both. Zeta[], @Schema(description) Zeta[] and @Schema(description) List<Zeta> produce the same output as before.

A few minor suggestions, none blocking:

1. Use the Annotations abstraction. The new lookup in TypeUtil.isTypeOverridden calls clazz.declaredAnnotation(SchemaConstant.DOTNAME_SCHEMA) directly. The rest of TypeUtil goes through context.annotations().getAnnotation(...) / getAnnotationValue(...) (e.g. getSchemaAnnotation, isIncludedAllOf). Using that here would keep it consistent and could reduce the block to a single getAnnotationValue(clazz, DOTNAME_SCHEMA, PROP_TYPE) call, assuming the enum conversion works there.

2. Array types also hit the class lookup. AugmentedIndexView.getClass(Type) goes through TypeUtil.getName, which unwraps array types to the element type. For a Zeta[] field, the lookup finds Zeta's @Schema(type = STRING), so isTypeOverridden now returns true (STRING vs. the array format). The generated output is still correct because array typing is applied elsewhere, but that seems accidental. Maybe limit the lookup to CLASS/PARAMETERIZED_TYPE kinds?

3. Comment on the typeList == null || change. If I read it right, this is needed because the type schema no longer gets a default type when the class declares one, and without the change such fields would fall through to checkRegistration. A short comment would help the next reader.

4. Javadoc wording. "If the classType does not itself define a type" reads as if it should say that the schemaAnnotation doesn't define one, with the class being the fallback.

5. Test both scan orders. The issue is about ordering, so asserting the reversed order against the same expected JSON would be cheap:

   assertJsonEquals("components.schemas.type-retained-from-class-annotation.json", Alpha.class, Beta.class, Zeta.class);

6. Parameters. AbstractParameterProcessor also calls isTypeOverridden, so something like @QueryParam("z") @Schema(description = "...") Zeta z will no longer get the default type: object. That's probably the desired outcome, but I didn't manage to verify the resulting parameter schema, so a test there may be worthwhile.

Also, for the release notes: the "expected" output in #2688 (type: string alongside $ref) is no longer produced either. Fields now get only the $ref plus their own attributes, which I think is better, but it will show up in users' spec diffs.

[MikeEdgar]

• ✔️ Use the Annotations abstraction
• ✔️ limit the lookup to CLASS/PARAMETERIZED_TYPE kinds
• ✔️ Comment on the typeList == null || change
• ✔️ Javadoc wording.
• ⚠️ skipped for now - Test both scan orders.
• ⚠️ skipped for now - Parameters

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud