Cannot use an empty map with kubebuilder:default marker

[Zeroji]

Hi,
I'm using kubebuilder to generate a CRD from the following type definition (minimized):

type CustomSpec struct {
    // +optional
    // +kubebuilder:default={}
    Foo FooType `json:"foo,omitempty"`
}

type FooType struct {
    // +optional
    // +kubebuilder:default=false
    Bar bool `json:"bar,omitempty"`

    // +optional
    // +kubebuilder:default=1
    Baz int `json:"baz,omitempty"`

    // +optional
    // +kubebuilder:default=QUX
    Qux string `json:"qux,omitempty"`
}

This gives the following CRD (comments added manually):

# This is .spec.versions[0].schema.openAPIV3Schema.properties.spec
# Indentation has been brought down
spec:
  properties:
    foo:
      # The following line is not present, but expected
      # default: {}
      properties:
        bar:
          default: false
          type: boolean
        baz:
          default: 1
          type: integer
        qux:
          default: QUX
          type: string
      type: object
  type: object

The issue I'm facing is that if I apply an empty spec for my custom resource, with spec: {} then foo only contains the golang type defaults, e.g. baz is 0 and qux is an empty string.
If I use spec: {foo: {}} then foo is populated with all the correct default values for its properties, however I don't want to force my clients to define foo.

If I modify the CRD manually (after controller-gen, but before kustomize) to add default: {} inside the CRD, everything behaves as expected, and an empty spec gives a correct CR when using kubectl get:

# kubectl get custom/object -o json | jq .spec
{
  "foo": {
    "bar": false,
    "baz": 1,
    "qux": "QUX"
  }
}

The current workaround I've found is to use the following annotation (which is non-ideal):

// +kubebuilder:default={baz: 1}
Foo FooType `json:"foo,omitempty"`

This has the effect of defining foo with the correct defaults, but foo.baz takes the value defined in foo's default over baz's default, so this can override default values on properties if the defaults are changed.

It seems the main issue is that controller-gen parses {} as an empty slice, which json.Marshal then marshalls as null, which gets ignored when writing the CRD. Here's the code that scans the slice, it appears this edge case was considered but not implemented yet:
https://github.com/kubernetes-sigs/controller-tools/blob/master/pkg/markers/parse.go#L297

Hopefully this is fixable without breaking stuff, or maybe a cleaner workaround can be found 🙂

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle rotten

[k8s-triage-robot]

The Kubernetes project currently lacks enough active 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:

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

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

/close

[k8s-ci-robot]

@k8s-triage-robot: Closing this issue.

In response to this:

The Kubernetes project currently lacks enough active 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:

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

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

/close

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.

[MaxThom]

/reopen

[k8s-ci-robot]

@MaxThom: You can't reopen an issue/PR unless you authored it or you are a collaborator.

In response to this:

/reopen

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.

[MaxThom]

Im facing the same exact issue.

[Danil-Grigorev]

/reopen

[k8s-ci-robot]

@Danil-Grigorev: Reopened this issue.

In response to this:

/reopen

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.