update to use new event handling

[mikemrm]

This adds the new x/events handling.

Note:

There are some issues with detecting Ack/Nack's on messages. Which has resulted in some tests not being able to be done that check for Nacked messages.

[jnschaeffer]

Looks decent so far - from what I understand we can't reintroduce the testing logic to check that a message was acked because the NATS server isn't directly exposed in x/testing/eventtools. Thus, most of my input is around the code itself.

[jnschaeffer]

So additionalSbjectIDs isn't the abstraction we want here - that's more of a generic "things that are related to this thing" list. I'm pretty sure what we actually want is a map[string]gidx.PrefixedID or something similar, where that map is built from (events.ChangeMessage).SubjectFields or (events.ChangeMessage).Changes as follows:

• For the given resource, look up its type in the engine schema
• For each field in the map, check to see if a relation exists with that name
• If it does, look up the target types for the relation
• If it does, ensure the prefix in the ID matches at least one prefix in the target types

As a concrete example, say we have a payload that look like so:

{
    "subjectID": "loadbal-abc123",
    "subjectFields": {
        "name": "my-cool-loadbalancer",
        "owner": "tnntten-abc123"
    }
}

Then, assume a policy exists (in YAML) like so:

resourceTypes:
  - name: tenant
    idPrefix: tnntten
  - name: loadbalancer
    idPrefix: loadbal
    relationships:
      - relation: owner
        targetTypeNames:
          - tenant

When we process the event, we should load the resource type definition from the schema. From there, we can see owner is a relation on the authorization policy for a loadbalancer resource, but name is not. Thus, we should persist anything called owner as a relationship in SpiceDB, and check that its ID prefix is tnntten before we do.

[mikemrm]

This makes sense. Right now the services aren't using the SubjectFields attribute and is something I think we need to discuss further. I believe once we've figured that out, this will be an easy transition. For now I've published a few changes to handle the AdditionalSubjectIDs better.

WONTFIX

[fishnix]

couple of minor comments, otherwise lgtm 👍

[fishnix]

IMO we should be structuring these with . notation instead of -, although I haven't looked around to see what's being done elsewhere so feel free to ignore if there's an established pattern.

[mikemrm]

I believe in most places we actually use _ which I can update this to use.

[fishnix]

Feels like the schema should be preloaded and this should be an error?

[mikemrm]

This is just for the mock. Most tests just define an empty mock Engine. So this simplifies to use the default schema if we aren't testing something specific.