proposal: add encoding/cue to encode/decode cue code to/from a string

[Somefive]

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

In KubeVela, we use pure CUE file to write and organize render logics but the code is embedded as YAML object while deployed into Kubernetes cluster.

For example, the following CUE file will be converted into the following YAML object.

scaler: {
	type: "trait"
	annotations: {}
	labels: {}
	description: "Manually scale K8s pod for your workload which follows the pod spec in path 'spec.template'."
	attributes: {
		podDisruptive: false
		appliesToWorkloads: ["deployments.apps", "statefulsets.apps"]
	}
}
template: {
	parameter: {
		// +usage=Specify the number of workload
		replicas: *1 | int
	}
	// +patchStrategy=retainKeys
	patch: spec: replicas: parameter.replicas
}

apiVersion: core.oam.dev/v1beta1
kind: TraitDefinition
metadata:
  annotations:
    definition.oam.dev/description: Manually scale K8s pod for your workload which follows the pod spec in path 'spec.template'.
  name: scaler
  namespace: {{ include "systemDefinitionNamespace" . }}
spec:
  appliesToWorkloads:
    - deployments.apps
    - statefulsets.apps
  podDisruptive: false
  schematic:
    cue:
      template: |
        parameter: {
        	// +usage=Specify the number of workload
        	replicas: *1 | int
        }
        // +patchStrategy=retainKeys
        patch: spec: replicas: parameter.replicas

Reference:

• https://github.com/kubevela/kubevela/blob/master/vela-templates/definitions/internal/trait/scaler.cue
• https://github.com/kubevela/kubevela/blob/master/charts/vela-core/templates/defwithtemplate/scaler.yaml

Currently, we use vela def render command to do the conversion. The conversion logic includes:

• Find out CUE in template field and convert into string and fill it into spec.schematic.cue.template.
• Parse the rest struct into metadata or other fields.

Therefore, although the YAML object can be rendered from the CUE file, the CUE file itself is not actually cue native, as the fields in the template field are actually isolated from other fields.

Describe the solution you'd like
A clear and concise description of what you want to happen.

If we have some internal function like cue.Marshal from encoding/cue, just like encoding/json and json.Marshal, we can make the CUE rendering process more elegant. For example, with this function we can write,

// scaler.cue
import (
  "encoding/cue"
)

parameter: {
        name: "scaler"
	type: "trait"
	annotations: {}
	labels: {}
	description: "Manually scale K8s pod for your workload which follows the pod spec in path 'spec.template'."
	attributes: {
		podDisruptive: false
		appliesToWorkloads: ["deployments.apps", "statefulsets.apps"]
	}
        template: {
	        parameter: {
		        // +usage=Specify the number of workload
		        replicas: *1 | int
	        }
	        // +patchStrategy=retainKeys
	        patch: spec: replicas: parameter.replicas
        }
}

output: {
        apiVersion: "core.oam.dev/v1beta1"
        kind: strings.ToTitle(parameter.type) + "Definition"
        metadata: {
                name: parameter.name
                annotations: {
                        parameter.annotations
                        "definition.oam.dev/description": parameter.description
                }
                labels: parameter.labels
        }
        spec: {
                parameter.attributes
                schematic: cue: template: cue.Marshal(parameter.template)
        }
}

And then we could use cue eval scaler.cue -e output -d yaml | kubectl apply -f -n vela-system - to get the rendered YAML object and apply it into Kubernetes cluster directly.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

[myitcv]

Thanks, @Somefive.

Noting per our discussion that this is something that has been discussed before, although I can't find a good issue/discussion that captures that discussion. So thanks for creating this issue.

Noting some of the main challenges we have identified in these conversations:

• What API would encoding/cue provide?
• Does the API end up looking similar to the CLI API surface via cue (def|eval|export)?
• ...

[roman-mazur]

Does the API end up looking similar to the CLI API surface via cue (def|eval|export)?

The export does not make much sense since you already have the other encoding packages (encoding/yaml, encoding/json). I would start with having something that is equivalent to cue eval to get the "minimized" form of the input serialized.

[sdboyer]

This would be quite useful for Thema, as well.

[FogDong]

This one's priority is medium and we have some workaround. But the workarounds need to use syntax, and #2258 blocks the workaround.

Not user-facing.

[mvdan]

Noting per our discussion that this is something that has been discussed before, although I can't find a good issue/discussion that captures that discussion. So thanks for creating this issue.

I believe that might be #423, although given its inactivity, I think we should close that one as a duplicate of this one even though this one is newer.