Using Ansible set_fact to create a dictionary from register results

Question

In Ansible I've used register to save the results of a task in the variable people. Omitting the stuff I don't need, it has this structure:

{
    "results": [
        {
            "item": {
                "name": "Bob"
            },
            "stdout": "male"
        },
        {
            "item": {
                "name": "Thelma"
            },
            "stdout": "female"
        }
    ]
}

I'd like to use a subsequent set_fact task to generate a new variable with a dictionary like this:

{
    "Bob": "male",
    "Thelma": "female"
}

I guess this might be possible but I'm going round in circles with no luck so far.

Answer 1

I think I got there in the end.

The task is like this:

- name: Populate genders
  set_fact:
    genders: "{{ genders|default({}) | combine( {item.item.name: item.stdout} ) }}"
  with_items: "{{ people.results }}"

It loops through each of the dicts (item) in the people.results array, each time creating a new dict like {Bob: "male"}, and combine()s that new dict in the genders array, which ends up like:

{
    "Bob": "male",
    "Thelma": "female"
}

It assumes the keys (the name in this case) will be unique.

---

I then realised I actually wanted a list of dictionaries, as it seems much easier to loop through using with_items:

- name: Populate genders
  set_fact:
    genders: "{{ genders|default([]) + [ {'name': item.item.name, 'gender': item.stdout} ] }}"
  with_items: "{{ people.results }}"

This keeps combining the existing list with a list containing a single dict. We end up with a genders array like this:

[
    {'name': 'Bob', 'gender': 'male'},
    {'name': 'Thelma', 'gender': 'female'}
]

Answer 2

Thank you Phil for your solution; in case someone ever gets in the same situation as me, here is a (more complex) variant:

---
# this is just to avoid a call to |default on each iteration
- set_fact:
    postconf_d: {}

- name: 'get postfix default configuration'
  command: 'postconf -d'
  register: command

# the answer of the command give a list of lines such as:
# "key = value" or "key =" when the value is null
- name: 'set postfix default configuration as fact'
  set_fact:
    postconf_d: >
      {{
        postconf_d |
        combine(
          dict([ item.partition('=')[::2]|map('trim') ])
        )
  with_items: command.stdout_lines

This will give the following output (stripped for the example):

"postconf_d": {
    "alias_database": "hash:/etc/aliases", 
    "alias_maps": "hash:/etc/aliases, nis:mail.aliases",
    "allow_min_user": "no", 
    "allow_percent_hack": "yes"
}

Going even further, parse the lists in the 'value':

- name: 'set postfix default configuration as fact'
  set_fact:
    postconf_d: >-
      {% set key, val = item.partition('=')[::2]|map('trim') -%}
      {% if ',' in val -%}
        {% set val = val.split(',')|map('trim')|list -%}
      {% endif -%}
      {{ postfix_default_main_cf | combine({key: val}) }}
  with_items: command.stdout_lines
...

"postconf_d": {
    "alias_database": "hash:/etc/aliases", 
    "alias_maps": [
        "hash:/etc/aliases", 
        "nis:mail.aliases"
    ], 
    "allow_min_user": "no", 
    "allow_percent_hack": "yes"
}

A few things to notice:

• in this case it's needed to "trim" everything (using the >- in YAML and -%} in Jinja), otherwise you'll get an error like:

  FAILED! => {"failed": true, "msg": "|combine expects dictionaries, got u\"  {u'...
  

• obviously the {% if .. is far from bullet-proof

• in the postfix case, val.split(',')|map('trim')|list could have been simplified to val.split(', '), but I wanted to point out the fact you will need to |list otherwise you'll get an error like:

  "|combine expects dictionaries, got u\"{u'...': <generator object do_map at ...
  

Hope this can help.