goccy/go-yaml deal with blank string uncorrectly

[polym]

goccy/go-yaml deal with blank string uncorrectly which is introduced by #242. And I met problem with parsing default containerd log.

CODE

package main

import (
	"fmt"

	errYaml "github.com/goccy/go-yaml"
	okYaml "gopkg.in/yaml.v2"
)

func main() {
	v := struct {
		Key string
	}{
		Key: " ",
	}
	d1, _ := okYaml.Marshal(v)
	fmt.Printf("%s\n%s\n", "YES", string(d1))

	d2, _ := errYaml.Marshal(v)
	fmt.Printf("%s\n%s\n", "NO", string(d2))

}

OUTPUT

YES
key: ' '

NO
key:

In what area(s)?

/area interceptor

What version of Loggie?

v1.2.0+

Expected Behavior

Load interceptor successfully

Actual Behavior

Got warnning log. get processor error: Key: 'SplitConfig.Separator' Error:Field validation for 'Separator' failed on the 'required' tag.

Steps to Reproduce the Problem

1. Config CRD Interceptor

    - type: normalize
      processors:
        - split:
            separator: ' '
            max: 4
            keys: [ "time", "stream", "F", "message" ]
        - drop:
            targets: [ "F", "body" ]
        - rename:
            convert:
              - from: "message"
                to: "body"
        - underRoot:
            keys:
              - kubernetes

2. kubectl delete pod and kubectl logs pod

[polym]

@ethfoo PTAL

[ethfoo]

Please refer to this document first when parsing the container log. The two YAML libraries do have some differences, and I'm still trying to figure out how to handle them.

[polym]

@ethfoo

Thanks. If I understand correctly, the document link you given seems to LogConfig. However, I use ClusterLogConfig to collect all pod's logs.

My CRD spec:

apiVersion: loggie.io/v1beta1
kind: ClusterLogConfig
metadata:
  name: all-pod-logs
spec:
  selector:
    type: cluster
    cluster: {{ .Values.config.loggie.discovery.kubernetes.cluster }}
  pipeline:
    sources: |
      - type: file
        name: container-logs
        paths:
          - /var/log/pods/**/*.log
    sinkRef: {{ .Values.sink.name }}
    interceptorRef: containerd

[OliverWoo]

I also encountered a similar problem when collecting container logs with k8s discovery enabled: if belongTo was specified explicitly, the belongTo field was not updated with the value podName/containerName/originalName. if rolling back to gopkg.in/yaml.v2, the belongTo field could be updated successfully.

apiVersion: loggie.io/v1beta1
kind: Interceptor
metadata:
  name: test
spec:
  interceptors: |
    - type: normalize
      name: test
      belongTo: ["test"]
      processors:
          ...

[ethfoo]

https://github.com/goccy/go-yaml/issues/309

[ethfoo]

Fixed: #317