unmarshal with comments provide wrong structure

[RaynDol]

This is the version

module mta_parse

go 1.12

require gopkg.in/yaml.v3 v3.0.0-20190502103701-55513cacd4ae

and run the following code which includes valid yaml

package main

import (
	"fmt"
	"log"

	"gopkg.in/yaml.v3"
)

var (
	sourceYaml = `version: 1
type: verbose
kind : bfr

# my list of applications
applications:

#  First app
  - name: app1
    kind: nodejs
    path: app1
    exec:
      platforms: k8s
      builder: test
`
)

func main() {
	t := yaml.Node{}

	err := yaml.Unmarshal([]byte(sourceYaml), &t)
	if err != nil {
		log.Fatalf("error: %v", err)
	}

	b, err := yaml.Marshal(&t)
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println(string(b))
}

The output is

version: 1
type: verbose
kind: bfr


# my list of applications
applications:
  - #  First app
    name: app1
    kind: nodejs
    path: app1
    exec:
        platforms: k8s
        builder: test


Process finished with exit code 0

please see that the comment is starting the array and not the name property which should start the array .

expected yaml

 #  First app
 -  name: app1
    kind: nodejs
    path: app1
    exec:
        platforms: k8s
        builder: test

if I remove the comment (first app) it looks ok, but with the comment change the yaml to invalid yaml, I think the comment and the state of the yaml is one of the killer feature in V3

[niemeyer]

Per note in the other issue, that yaml seems valid.

Which parser is failing to handle it?

[RaynDol]

@niemeyer no parser, just when you look on it seems to be wrong , it should be the same as before the unmarseling process, what do you think ?

[niemeyer]

The description above states it's invalid yaml. That looks and parses as valid yaml representing the same data that was unmarshalled.

We also currently do not attempt to preserve 100% of the formatting of the yaml parsed. The intent is to produce a format that most people would find pleasant by default. For example, in your input the comment is indented at a different column than the content it's documenting. At least to my taste that feels awkward.

[superbogy]

I had same issue. Read from file then Unmarshal it and the structure changed. It seems that the pkg sort it when Unmarshal.

yaml file:

openapi: 3.0.0
info:
  version: 1.0.0
  title: Swagger Petstore
  license:
    name: MIT
servers:
  - url: http://petstore.swagger.io/v1
paths:
  bar: foo
  a: 1
  z: 2
  b: 3
components:
  baz: true

after Unmarshal:

components:
  baz: true
info:
  license:
    name: MIT
  title: Swagger Petstore
  version: 1.0.0
openapi: 3.0.0
paths:
  a: 1
  b: 3
  bar: foo
  z: 2
servers:
- url: http://petstore.swagger.io/v1

@niemeyer

[mikefarah]

One of my users has reported that this is breaking his cassandra's docker entry point

mikefarah/yq#364

[niemeyer]

Sorry, I still don't understand what the issue is here. The two maps reported above are exactly the same yaml. That's also completely unrelated to comments since neither of them have any comments whatsoever.

If someone wants to report an issue, please do so in a slightly more readable form.