apimachinery: fix UnsafeGuessKindToResource for some classes of words

[michaelbeaumont]

What type of PR is this?

/kind bug

What this PR does / why we need it:

Fixes guesses that apimachinery makes about resource names from a GVK.

Which issue(s) this PR fixes:

Fixes kubernetes/client-go#1082

Special notes for your reviewer:

Does this PR introduce a user-facing change?

NONE

[k8s-ci-robot]

Hi @michaelbeaumont. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[roycaihw]

/assign @apelisse
/triage accepted

[apelisse]

/ok-to-test
/lgtm

Not sure if this breaks compatibility with anything.

[AlexanderYastrebov]

The https://github.com/kubernetes/code-generator uses https://pkg.go.dev/k8s.io/gengo/namer#NewAllLowercasePluralNamer so another option might be to use it here or make client-go use it instead of UnsafeGuessKindToResource that is documented as broken.

[evankanderson]

FWIW, this affects https://gateway-api.sigs.k8s.io/, which is sort of annoying (spent about 2-3 hours debugging why a test framework using .Add wasn't finding my Gateway).

👍 on this change from me.

[ncdc]

/uncc

[evankanderson]

/assign @deads2k

[pmalek]

Also hit this when using Gateway API's objects with dynamic fake client: k8s.io/client-go/dynamic/fake's NewSimpleDynamicClientWithCustomListKinds (just as @evankanderson described it).

Any (even rough) ETA for this? Any workarounds before this gets merged?

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle stale
• Mark this issue or PR as rotten with /lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[evankanderson]

/remove-lifecycle stale

[pmalek]

Any chance that we can get someone else to look at this and review it (as @deads2k and @apelisse seem not be able to pick this up)?

[apelisse]

Why would you throw me under the bus? I've lgtm'd 6 months ago! 😄

[pmalek]

@apelisse sorry 😅 Didn't notice that (and the fact that still @deads2k is assigned as reviewer: so I guess we need lgtm from him). My bad.

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: michaelbeaumont, pmalek
Once this PR has been reviewed and has the lgtm label, please ask for approval from deads2k by writing /assign @deads2k in a comment. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• staging/src/k8s.io/apimachinery/pkg/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[evankanderson]

/assign caesarxuchao

[lavalamp]

I think we might have this code locked because changing it could break existing things and pluralizing english is hopeless unless we can call out to GPT-3 anyway.

There should be fields to define both in CRDs to avoid this code being invoked, if that's not the case please let us know.

[liggitt]

/hold

This isn't a compatible change to make to this function. Anything using this function to auto-generate plural names from singular names would break when re-generating an existing object that this change impacts.

I think we should remove all non-test in-tree calls to this function (I only see a single call from DefaultRESTMapper#Add, which I would also consider removing), mark it deprecated, file issues to public callers we find, and consider removing it in a future release.

If we want a new function that takes a singular name and pluralizes it that has this expanded behavior from the start (with a signature like AttemptToPluralize(singular string) (plural string)), I guess I wouldn't object, but that doesn't need to live in k8s.io/apimachinery

[evankanderson]

I believe that the use that bit me was in test code without an apiserver loaded. I can try to dig out the code in question; I believe it tries to pluralize "Gateway" as "Gatewaies", which was very confusing when I finally debugged it.

Removing this function would also be acceptable to me, but it's currently a trap for anyone testing the Gateway API.

[pmalek]

@evankanderson Can you dig that test code up? I'm positive I've also encountered this issue recently but can't find the relevant test code that would manifest the problem itself (other than calling UnsafeGuessKindToResource with Gateway type).

https://grep.app/search?q=.UnsafeGuessKindToResource%28&case=true doesn't show that many public callers 👍

[evankanderson]

I believe that we were getting hung up in tracker.Add from this code:

https://github.com/kubernetes/client-go/blob/master/testing/fixture.go#L341

I also see some questionable usage in https://github.com/knative/pkg/blob/main/resolver/addressable_resolver.go#L167 and https://github.com/knative/pkg/blob/main/kref/knative_reference.go#L53, but the usage of ObjectTracker from client-go/testing seems most likely to cause problems, because I'm guessing that package is used many places.

You can see my work-around here:

https://github.com/knative-sandbox/net-gateway-api/blob/main/pkg/reconciler/ingress/ingress_test.go#L404

		// The fake tracker's `Add` method incorrectly pluralizes "gatewaies" using UnsafeGuessKindToResource,
		// so create this via explicit call (per note in client-go/testing/fixture.go in tracker.Add)
		fakeCreates := []runtime.Object{}
		for _, x := range tr.Objects {
			myGw, ok := x.(*gatewayapi.Gateway)
			if ok {
				fakegwapiclientset.Get(ctx).GatewayV1beta1().Gateways(myGw.Namespace).Create(ctx, myGw, metav1.CreateOptions{})
				tr.SkipNamespaceValidation = true
				fakeCreates = append(fakeCreates, myGw)
			}
		}
		tr.WantCreates = append(fakeCreates, tr.WantCreates...)

[liggitt]

/close

Changing this function is not the right direction. I would be in favor of marking as deprecated and removing all in-tree calls to it, and adding the ability for fake clients to be told the accurate plural/singular mappings if needed.

[k8s-ci-robot]

@liggitt: Closed this PR.

In response to this:

/close

Changing this function is not the right direction. I would be in favor of marking as deprecated and removing all in-tree calls to it, and adding the ability for fake clients to be told the accurate plural/singular mappings if needed.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[pmalek]

@liggitt I'd like to move forward with this somehow but I'm having hard time understanding your proposal.

E.g. how would you go about removing in-tree calls of this? There are multiple places where it's used in user facing code e.g.

kubernetes/staging/src/k8s.io/apimachinery/pkg/api/meta/restmapper.go

Lines 99 to 113 in 694698c

	func (m *DefaultRESTMapper) Add(kind schema.GroupVersionKind, scope RESTScope) {
	plural, singular := UnsafeGuessKindToResource(kind)
	m.AddSpecific(kind, plural, singular, scope)
	}
	func (m *DefaultRESTMapper) AddSpecific(kind schema.GroupVersionKind, plural, singular schema.GroupVersionResource, scope RESTScope) {
	m.singularToPlural[singular] = plural
	m.pluralToSingular[plural] = singular
	m.resourceToKind[singular] = kind
	m.resourceToKind[plural] = kind
	m.kindToPluralResource[kind] = plural
	m.kindToScope[kind] = scope
	}

Would you suggest to mark func (m *DefaultRESTMapper) Add(kind schema.GroupVersionKind, scope RESTScope) as deprecated as well due to its usage of UnsafeGuessKindToResource and suggest users to use AddSpecific() instead?

[pmalek]

It seems that using k8s.io/gengo/namer's NewAllLowercasePluralNamer() would work:

func (m *DefaultRESTMapper) Add(kind schema.GroupVersionKind, scope RESTScope) {
	n := namer.NewAllLowercasePluralNamer(nil)
	t := types.Ref(kind.GroupVersion().String(), kind.Kind)
	plural := kind.GroupVersion().WithResource(n.Name(t))
	singular := kind.GroupVersion().WithResource(strings.ToLower(kind.Kind))

	m.AddSpecific(kind, plural, singular, scope)
}

If this would be accepted and is perceived as the direction that we can move forward with, I can submit a PR replacing usages of UnsafeGuessKindToResource with the above.

[liggitt]

Would you suggest to mark func (m *DefaultRESTMapper) Add(kind schema.GroupVersionKind, scope RESTScope) as deprecated as well due to its usage of UnsafeGuessKindToResource and suggest users to use AddSpecific() instead?

yes, mark deprecated and mark callers that rely on it as deprecated as well, pointing people to AddSpecific

[pmalek]

@liggitt I've pushed #119924 to try to address the above PTAL.