Ansible's declarative style makes it very easy to do configuration management. While the idea of ansible is a pretty good one, its YAML playbooks often confuse people. With the addition of blocks in Ansible 5.3, Ansible is looking more and more like a programming language.

However, readability is key and Ansible sometimes fails to deliver the user experience you would like to. Imagine a much simpler workflow. Instead of:

- name: Clone git repositories
  git: repo={{ item.repo }}
       dest={{ item.path }}
       clone=yes
       update=yes
  register: repository_update
  with_items:
    - { repo: "{{ repo1 }}", path: "/foo" }
    - { repo: "{{ repo2 }}", path: "/bar" }

- name: Next task

you could just write:

from pysible.modules import git

# Clone git repositories
for repo, dest in [(REPO, '/foo'), (REPO2, '/bar')]:
    last_update = git(repo=REPO, dest=REPO, clone=True, update=True)

# Next task

This allows for a lot more flexibility. Instead of includes you use simple functions and instead of roles... Well don't even get me started on that. I think that instead of using a pseudo language in YAML, using Python has the advantage of clean proper coding with all the advantages a language like Python comes in. Additionally we don't really have to give up Ansible features. We can use handlers, includes and modules like before. Just with a slightly different API.

It would be a brave new world, but finally there would be beautiful deployment code!

Running would be as simple as running Ansible. A simple command pysible <playbook> -i <hosts-file> would be enough to trigger execution on a few hosts. This command would ensure that host files could still be used (including host/group variables).

At the moment there is a proof of concept that I won't publish, yet. However if you would like to see this, start a conversation in the issue tracker or star the repository.

The current implementation is merely 30 lines long, because I'm using ansible's internal APIs. I'm thinking about publishing this tool.