[BUG] RequestBodies schema type conflicts with the header

[zelinh]

What is the bug?

Separate from #633

For example /_bulk, we specify the requestBodies for bulk as an array which conflicts with the application/x-ndjson format, which uses newline-delimited JSON rather than a JSON array.

Ref:

opensearch-api-specification/spec/namespaces/_core.yaml

Lines 2256 to 2259 in 07e329e

	content:
	application/x-ndjson:
	schema:
	type: array

What is the expected behavior?

We should specify whichever schema for application/x-ndjson as a String.

[Xtansia]

We need to represent the structure that is being returned or accepted by the API. It is an array of objects, just encoded in a certain manner, which is indicated by what the content type is.
This is not much different to the way things like XML would be represented: https://swagger.io/docs/specification/v3_0/data-models/representing-xml/

[dblock]

@Xtansia Are you seeing this is a non-issue and represented correctly? @zelinh do you disagree? otherwise close this?

[zelinh]

@Xtansia Thanks for the reply. While I think the correct structure to represent this should be String. As we specify it as Newline delimited JSON, It's slightly different from the simple array or json structure. This can cause error when we run any schema validation.

[Xtansia]

@zelinh Which schema validation are you referring to? It is still structured data not a plain string, and the encoding is determined by what the content type is, which any validation or (de)serialization must take into account. Otherwise this also applies to cbor & smile.

Additionally we need the structure information for client code generation.

[zelinh]

I see. Thanks for the context here. I think we could close out this issue then as this is by design.