Scope of a single scenario - should I call my role multiple times in one scenario

[pbarriscale]

Hi,

I have a number of input variables for a role and perform some basic input validation on them before executing the main tasks in the role. See below.

# Role abc - input validation tasks
- name: Check for x
  when: x | default('', true) | trim == ''
  ansible.builtin.fail:
    msg: "x is required"

- name: Check for x if it is an integer
  when: x | default('', true) | trim != x | default('', true) | int | string
  ansible.builtin.fail:
    msg: "x must be a number"

- name: Check for y
  when: y is not defined
  ansible.builtin.fail:
    msg: "y is required"

- name: Check y is a mapping
  when: y is not mapping
  ansible.builtin.fail:
    msg: "y type must be a mapping"

I see two approaches to testing this input validation.

1. Create a scenario for each check. This would be 4 scenarios in total.
2. Create a single scenario and test all validation checks together. This would mean I would call the role in my converge step 4 times, each with different inputs set to trigger the specific condition.

Number 1 creates a lot of overhead even if I was to create share some of the molecule step playbooks/resources between scenarios.

Number 2 is the opposite end of the spectrum where a number of similar "scenarios" are bundled together.

Any thoughts or recommendations on how to proceed or is there some other alternative?

Thanks,
Paul

[ziegenberg]

That's what role argument validation is for. See: https://docs.ansible.com/ansible/latest/playbook_guide/playbooks_reuse_roles.html#specification-format

This is much more powerful than just asserting the right role variables. It also gives you documentation.

[pbarriscale]

This is great thank you. This is much more powerful. My original question would still be what would be the best way to test role argument validation ensuring expected conditions are met when invalid or missing inputs are supplied.

Thanks

[cidrblock]

IMHO, I would do 1 scenario for each check (set of vars). The reason is, it is easier to isolate, test, and see and recreate a failure. If you are using the pytest-ansible pluign, each scenario will show as an individual pytest test so when running in your CI it's easier to specific exactly which one failed.

In the end, having the tests is better than not having them, so which ever why you choose to go, everything will be better. :)

[pbarriscale]

Great thanks. To confirm my understanding for my example above you would recommend 4 scenarios correct?

• Check for x
• Check for x if it is an integer
• Check for y
• Check y is a mapping

Thanks