Pass runtime_fields from FieldCapabilitiesRequest to FieldCapabilitiesIndexRequest

[przemekwitek]

This PR makes the runtime_fields map passed down to remote FieldCapabilitiesRequests.
Additionally:

• makes runtimeFields member variable final in FieldCapabilitiesIndexRequest
• makes the style of reading field consistent in FieldCapabilitiesRequest

Relates to #68904

Marked as >non-issue as this is a fix for an unreleased feature.

[elasticmachine]

Pinging @elastic/es-search (Team:Search)

[cbuescher]

The change looks good to me, it would be great to have a way to test this in the "normal" case though (when the remote cluster is on 7.12 or above). I don't know if we have the appropriate test infra though, so I'm currently looking for more information. If this proves to be difficult we should probably merge this for now and open a follow up to add those tests later.

[przemekwitek]

I'm not sure about the test suite for _field_caps but there is a precedence of testing transform in multi-cluster env, see e.g.: x-pack/qa/multi-cluster-tests-with-security/src/test/resources/rest-api-spec/test/multi_cluster/80_transform.yml

[cbuescher]

@przemekwitek I found yaml tests for "field_caps" under ':qa:multi-cluster-search' and briefly played around with it (see cbuescher@93ca26c). Adding something basic like this with a runtime_mapping that changes a remote fields type should be enough to test the basic propagation I think.

[przemekwitek]

Adding something basic like this with a runtime_mapping that changes a remote fields type should be enough to test the basic propagation I think.

I verified that your test fails without the fix but passes with the fix.
I added the test to this PR.
Thank you for your help!

[cbuescher]

Great, thanks again for catching this. LGTM pending CI is green.

[javanna]

I am wondering how this is not throwing an error for trying to access doc_values of the field that is being defined. Would it make sense to use a different field name? You can potentially even skip the script.

[javanna]

Though I see that this is also testing shadowing the existing mapped field. Maybe that is not required though as it's besides the scope of this test. We could define a new field as part of runtime_mappings and check that field_caps returns it

[przemekwitek]

Done (skipped the script part).

I think the script is just not executed during _field_caps so it's cleaner not to provide it in the first place.

[javanna]

ah right that's it and that is why it was not causing an error :)