Fix MediaType implementation of OpenApi 3.1

[cavallium]

The internal __absolute_uri__ flag is breaking the code that checks if the model is empty

Motivation:

OpenAPI 3.1 specifications do not work because vert.x populates the model with its own fields, then checks if the model is empty

Conformance:

You should have signed the Eclipse Contributor Agreement as explained in https://github.com/eclipse/vert.x/blob/master/CONTRIBUTING.md
Please also make sure you adhere to the code style guidelines: https://github.com/vert-x3/wiki/wiki/Vert.x-code-style-guidelines

[pk-work]

Hi @cavallium thanks for finding this bug! Could you also provide a test case to ensure that we don't run into this issue again?

[cavallium]

A test can't guarantee this with the current code, since that variable can contain fields that do not exist on the defined specification.
I suggest to merge this temporarily, and then refactor the class in a way that makes it possible to check if the original unmodified model is really empty.

[pk-work]

I am convinced that we can write a test that will prevent exactly this problem from occurring again. We can also directly write a test, to ensure that the other two annotations __absolute_recursive_ref__ and __absolute_ref__ [1] will not result in the same problem.

I will not merge a fix without a test, because otherwise nobody will add a test later. MediaTypeImplTest would be a good place for testing this issue.

And yes, long term we should fix it in vertx-json-schema.

[1] https://github.com/eclipse-vertx/vertx-json-schema/blob/1ad80a82191aa3e6184ac642e3eaf9cc3aaec686/src/main/java/io/vertx/json/schema/impl/JsonObjectSchema.java#L22

[cavallium]

I am convinced that we can write a test that will prevent exactly this problem from occurring again. We can also directly write a test, to ensure that the other two annotations __absolute_recursive_ref__ and __absolute_ref__ [1] will not result in the same problem.

I will not merge a fix without a test, because otherwise nobody will add a test later. MediaTypeImplTest would be a good place for testing this issue.

And yes, long term we should fix it in vertx-json-schema.

[1] https://github.com/eclipse-vertx/vertx-json-schema/blob/1ad80a82191aa3e6184ac642e3eaf9cc3aaec686/src/main/java/io/vertx/json/schema/impl/JsonObjectSchema.java#L22

Should MediaType#getSchema() return the "__*" annotations?

[cavallium]

@pk-work I have fixed the existing test instead of creating a new one, since it was already checking for that behavior, but it wasn't checking for internal json-schema annotations.

[pk-work]

Cool! I think we could improve the readability of the code and avoid code duplication, be re-using (copying) the created JsonObjects. Instead of writing multipel times new JsonObject().put(ABS_URI, DUMMY_REF_VALUE) you could create a JsonObejct schemaWithABSUri and re-use it. In case you use merge call copy()before.

I think this will reduce the code in testGetters() a lot.

[cavallium]

Cool! I think we could improve the readability of the code and avoid code duplication, be re-using (copying) the created JsonObjects. Instead of writing multipel times new JsonObject().put(ABS_URI, DUMMY_REF_VALUE) you could create a JsonObejct schemaWithABSUri and re-use it. In case you use merge call copy()before.

I think this will reduce the code in testGetters() a lot.

Done, thanks for your suggestion, it's much better now

[pk-work]

Thank you very much for your contribution and your patience to add the best possible test!