feat(AIP-203): add resource name IDENTIFIER enforcement

[noahdietz]

Enforces AIP-203 guidance added in aip-dev/google.aip.dev#1201 that the resource's name field must have google.api.field_behavior = IDENTIFIER.

[sam-utila]

This breaks the lint for me.
I have an identifier, but it still fails on linting.

edit: This code explicitly check for "IDENTIFIER" and fails on anything else (i.e. OUTPUT_ONLY, REQUIRED, etc...)

I wonder how it did not fail any other test

[noahdietz]

@sam-utila can you open an issue instead of commenting on the PR? Also please include the proto that is failing.

worth noting this hasn't been released yet so you are using main which is probably not advisable.

[sam-utila]

Pardon me - I wasn't aware a new IDENTIFIER value is now allowed, should probably update my googleapis deps.

However - I wonder if this will continue to work for names that were marked as OUTPUT_ONLY and are filtered automatically from requests by tooling such as grpc-gateway/v2/protoc-gen-openapiv2.

Anyway I had to use main instead of latest since latest was broken for me and was fixed with the recent https://github.com/googleapis/api-linter/pull/1239 (specifically this commit: jhump/protoreflect@f770609).

[noahdietz]

I wonder if this will continue to work for names that were marked as OUTPUT_ONLY and are filtered automatically from requests by tooling such as grpc-gateway/v2/protoc-gen-openapiv2.

I can't speak to grpc-gateway's implementation of filtering based on field_behavior, that is their own implementation derived from presence of field_behavior annotations. I'd recommend filing an issue on their repo.

Please read the AIP-203 changes for the use of and reasoning for adding IDENTIFIER.

was broken for me and was fixed with the recent

I see makes sense! Consider using a specific version instead of latest so your API doesn't suddenly have issues if you are blocking development based on linter findings.

[sam-utila]

No problem, considering I can always put new linting rules on disabled.

Just wondering that sometimes IDENTIFIERs can also be OUTPUT_ONLY and this is something that is now not encouraged by this AIP - haven't sinked in yet :-) .

[flyfire2002]

I too am interested in the implication for name fields currently marked as OUTPUT_ONLY. I guess semantically IDENTIFIER doesn't preclude a behavior that is OUTPUT_ONLY ("Server ignores it if set as input"). However now we don't have a language for https://google.aip.dev/203#output-only scenario 3 where the name is the immutable field.