Personal EDN Ansible Playbooks


Here’s how most of the tasks look like:

[{:name "hi"
  :hosts ["localhost"]

  :vars {:user "_____"
          :home "/home/"}

  [{:file {:path "/"
            :state "directory"}
    :loop [".local/bin"


Using multiline strings is straightforward, despite not being the prettiest thing I’ve seen:

{:name "Sway environment variables"
  :blockinfile {:path "/etc/environment"
                :block "#


Nothing special, it’s just fun to write in Clojure :) I do generally dislike YAML because of its extreme, uncalled-for minimalism. It may or may not lead to me shooting myself in the foot (it does).

If anything goes south, I can always resort back to it, anyways:

cat playbook.edn | bb '(-> *in* slurp edn/read-string (yaml/generate-string :dumper-options {:flow-style :block}))'


Since every Ansible command I run requires a compilation step, I’ve written a short Babashka script to act as a middleman on each Ansible run. Luckily, Babashka comes with YAML (and obviously EDN) support built-in.

#!/bin/env bb
(import [ File])
(require '[clojure.core.match :refer [match]])
(require '[babashka.process :refer [process check]])

(defn edn->yaml
  (-> file-path
      (yaml/generate-string :dumper-options {:flow-style :block})))

(defn main
  [action edn-playbook extra-args]
  (let [exec      (format "ansible-%s" action)
        yaml-file (.getAbsolutePath (File/createTempFile "playbook" ".yaml"))
        cmd       (into [exec yaml-file] extra-args)]
    (spit yaml-file (edn->yaml edn-playbook))
    @(process cmd {:out :inherit})))

(match (vec *command-line-args*)
  [action playbook & args] (main action playbook args)
  [action playbook] (main action playbook [])
  :else (println "Usage: ednsible ACTION PLAYBOOK [EXTRA_ARGS]"))


I saved the script as ednsible and created these two aliases to avoid unfortunate scenarios:

alias ednsible-playbook="ednsible playbook"
alias ednsible-lint="ednsible lint"

Why convert to YAML instead of JSON?

An alternative path would be to generate JSON instead of YAML, since JSON is supposedly a subset of YAML 1.2. However:

  1. ansible-lint doesn’t work at all when given a JSON playbook, even when given a JSON playbook that works perfectly with ansible-playbook
  2. Seems like the subsettery topic is controversial. With my luck, I’m sure I’ll step on a mine in this area; So I’d rather take the path of least resistance as YAML support is baked into Babashka :)
  3. The YAML parser that comes with Babashka is clj-yaml, which is a wrapper around SnakeYAML, which is a 1.1 parser 🙃