Ansible json_query move dictionary key/value to subkey that is a list

I'm trying to edit a dictionary in ansible with a one line json_query (and/or other filters), but something is wrong and I can't figure it out. I have this structure (with example data)

seinfeld:
  - name: Jerry
    jobs:
      - comedian
    friends: 
      - name: Elaine
      - name: Kramer
  - name: George
    jobs:
      - yankees
    friends:
      - name: Jerry
      - name: Susan

And I need to create a new structure like this

new_list:
    - main: Jerry
      name: Elaine
    - main: Jerry
      name: Kramer
    - main: George
      name: Jerry
    - main: George
      name: Susan

The name key is added to every element of friends key and flatten it out to a new list. I already made some progress mixing jmespath map and merge, but at least one key is always null or invalid, some of random tests are [].merge({friends: friends},{main: name}) [*].map(&merge({main: name}, @), friends)

Solution

There are more options for getting the data. For example,

  tmp: "{{ seinfeld | json_query('[].[name, friends[].name]') }}"

  tmp: "{{ seinfeld | map(attribute='name') |
           zip(seinfeld | map(attribute='friends')
                        | map('map', attribute='name')) }}"

give the same list

  tmp:
  - - Jerry
    - - Elaine
      - Kramer
  - - George
    - - Jerry
      - Susan

Now, you would need to map the function product. Unfortunately, this is not possible. The filter json_query doesn't help you with the list of lists either. So, this is a dead end.

The only option is a Jinja template

  new_list: |
    {% filter from_yaml %}
    {% for name in tmp %}
    {% for friend in name.1 %}
    - {main: {{ name.0 }}, name: {{ friend }}}
    {% endfor %}
    {% endfor %}
    {% endfilter %}

gives what you want

  new_list:
  - main: Jerry
    name: Elaine
  - main: Jerry
    name: Kramer
  - main: George
    name: Jerry
  - main: George
    name: Susan

But, you can easily iterate the list with subelements. For example,

    - debug:
        msg: "main: {{ item.0.name }}, name: {{ item.1.name }}"
      loop: "{{ seinfeld | subelements('friends') }}"

gives (abridged)

  msg: 'main: Jerry, name: Elaine'
  msg: 'main: Jerry, name: Kramer'
  msg: 'main: George, name: Jerry'
  msg: 'main: George, name: Susan'

The decision of whether you need new_list is up to you.

^{Example of a complete playbook for testing
- hosts: localhost

vars:

seinfeld:
- name: Jerry
jobs:
- comedian
friends:
- name: Elaine
- name: Kramer
- name: George
jobs:
- yankees
friends:
- name: Jerry
- name: Susan

tmp1: "{{ seinfeld | json_query('[].[name, friends[].name]') }}"
tmp2: "{{ seinfeld | map(attribute='name') |
zip(seinfeld | map(attribute='friends')
| map('map', attribute='name')) }}"
new_list: |
{% filter from_yaml %}
{% for name in tmp1 %}
{% for friend in name.1 %}
- {main: {{ name.0 }}, name: {{ friend }}}
{% endfor %}
{% endfor %}
{% endfilter %}

tasks:

- debug:
var: tmp1
- debug:
var: tmp2
- debug:
var: new_list

- debug:
msg: "main: {{ item.0.name }}, name: {{ item.1.name }}"
loop: "{{ seinfeld | subelements('friends') }}"}