salt renderer for extremly lazy python dev

salt, renderer
pip install dawdaw==0.1.2



Dawdaw is an experiment to make a SaltStack custom renderer (the stuff that allows you to write your states in yaml/jinja2/mako/other) in an attempt to solve those problems:

  • current states are too verbose to write
  • you often repeat yourself too much
  • really have a linear states declaration for requires
  • explicit requires on included states, to avoid globals
  • namespacing all the things, to avoid globals
  • indirectly trying to solve the "salt states are very hard to redistribute" problem by going full python, you can now use setup.py and pypi/pip to redistribute you work¹

Disadvantages: you move await from full declarative code (which you were already doing in fact with jinja2 templates) to go back to python code. This is, on one hand very powerful, on the other hand probably too powerful (and may be way less easy to understand for devops that don't come from a programming background). That works for me because I'm a python dev and I'm using this for my personal usages, but that might not fit your case.


Move from:

  - dotfiles

    - name: wyrd
    - require:
      - sls: dotfiles

    - name: ssh://git@bitbucket.org/psycojoker/reminds.git
    - runas: psycojoker
    - target: /home/psycojoker/reminds/
    - require:
      - pkg: git

cd /home/psycojoker/reminds/ && bash init:
    - unless: ls /home/psycojoker/.reminders /home/psycojoker/.reminders.gcl
    - user: psycojoker
    - require:
      - git: reminds.git



from dawdaw.states import pkg, git, cmd, include
from dawdaw.utils import default, test, debug

dotfiles = include("dotfiles")

with default(user="psycojoker", runas="psycojoker"):
                  require=[dotfiles.get("pkg", "dotfiles-pkgs")])

    if not test("ls /home/psycojoker/.reminders /home/psycojoker/.reminders.gcl"):
        cmd.run("cd /home/psycojoker/reminds/ && bash init")


pip install dawdaw

# this is how you install a renderer in salt
# if you know a better way to distribute it, plz tell me

# adapt the path to the location of your salt data
mkdir -p /srv/salt/_renderers
touch /srv/salt/_renderers/__init__.py

curl "https://raw.githubusercontent.com/Psycojoker/dawdaw/master/dawdaw_template.py" > /srv/salt/_renderers/dawdaw_template.py

# if you use salt in master/slave
salt '*' saltutil.sync_renderers
# or locally
salt-call --local saltutil.sync_renderers

Once it's done, you can normally run highstates, this will handle dawdaw_template like any regular other state.


Once you have installed dawdaw (see previous section), to use it, you simply need to put this as the first line of your file (dawdaw_template being the name of the file under which you have redirected the curl command bellow):



Using states is extremely simple: just import the state module and call the corresponding function like a python function.


    - argument_1: value_1
    - argument_2: value_2
    - argument_3: value_3


from dawdaw.states import state_module


Another example

    - target: /tmp/dawdaw


from dawdaw.states import git

git.latest("https://github.com/Psycojoker/dawdaw", target="/tmp/dawdaw")

The 'default' context manager

In salt, you often end up repeating the same arguments a lot, like settings the prioprietary of the file to the same user a lot. This is boring and not error proof. Sure, the 'use' exists, but it's awkward and no one knows about it. Thanks to python, we have context managers and we can use the with keyword to handle that.

The default context manager create a context in which every command that waits for some specific keywords will be called with it.


from dawdaw.states import git, file
from dawdaw.utils import default

with default(makedirs=True): 
    # git won't received the 'makedirs' keyword
    git.latest("https:/...", target="/some/stuf")

    # file will received it
    file.managed("/some/stuff/subdir/settings_prod.py", source="...")

I often end up using it to settings user and groups:

with default(user='psycojoker', group='psycojoker', runas='psycojoker'): 
    # ...


(The stuff you use in the CLI like salt '*' cmd.run "ls /tmp"). As simple as states, just import it and call it like normal python code (and play with it's return like in normal python):

from dawdaw.modules import cmd

for f in cmd.run("ls /tmp"):
    # do some stuff with 'f'

The 'test' helper

Sometime, you need to test if a command return the code '0', you can do it using cmd.retcode("...") but that's quite boring. Dawdaw provides a simple helper to do that for you:

from dawdaw.utils import test

if test("ls /tmp/this_file_exist"):
    # do some stuff


In dawdaw, you don't have to care that much about requisites, a linear execution of the states in the order in which they are called is enforced. This mean, that, in this example, module.b will have a require on module.a and module.c will have a require on module.a and module.b:


The requires are only set if the state is actually called, so you can use 'if' and other control flow structure the way you want like in normal python code.

If you stil need/want to set explicit requires, every state return a reference to itself once it is called, so you can simply do it this way:

a = module.a("...")
module.b("...", require=[a])  # remember, requires are set in a list!

Namespacig, watch or more generally: how to refer to a state

In dawdaw, every state has its name namespaced with the name of the file it is stored in and the module from which it's called. For example, this state: git.latest("https://github.com/Psycojoker/dawdaw") in the file dawdaw.sls will have the name dawdaw_git_https://github.com/Psycojoker/dawdaw. **Keep this in mind if you want to refer to other states in non-dawdaw states.

But when you are in dawdaw you don't have to care about that: every state returns a reference to itself once called, you can use that without caring about how it is done and without the risk of making stupide mistake or having to rename it everywhere. For example:

a = module.a("...")
module.b("...", watch=[a])  # remember, watchs are set in a list!

Works for watch, watch_in, require, require_in, prereq, the other requisites etc ... Basically everytime you need to reference a state.

If you really need to do that by hand (don't), in reality, the reference is just a dict, so you can do this this way (don't forget about the namespacing!):

# in file example.sls

module.b("...", watch=[{"module": "example_module_some_name"}])  # remember, watchs are set in a list!

But don't do that.


include works nearly the same than in salt. The only difference is that you only include one state at once, not a list of states. This allows the include to return a representation of included sls file to reference states from this sls file.

In the same fashion than state, every state that follows an include will require on it to enforce linear execution.


from dawdaw.states import include



An include can be use to reference a state of the included sls file (and it's recommand to to avoid global namespaced reference) using the .get method. .get takes 2 parameters: the module and the name.


from dawdaw.states import include, pkg

some_state = include("some_state")

pkg.installed("stuff", require=[some_state.get("a_module", "a_name")])

If the included sls file is not a dawdaw file, you must pass the argument in_dawdaw=False to include because of namespacing.


from dawdaw.states import include, pkg

some_state = include("some_state", in_dawdaw=False)

pkg.installed("stuff", watch=[some_state.get("a_module", "a_name")])

Pillar, grains and opts

All those 3 salt artifacts are accessible very easily by simply importing them and they will behave the same way than they behave in jinja2 templates (hint: they are dictionaries):

from dawdaw import pillar, grains, opts



Dawdaw comes with a helper debug to debug what it does. This helper will simply print on the shell the generated yaml (you'll see it in the logs or if you run salt locally using "salt-call --local").


from dawdaw.utils import debug


You can pass a boolean argument to debug activated/desactivate debugging:

from dawdaw.utils import debug


if some_stuff:
    # finally don't need to debug

Also, since this is full python you can drop in ipdb to just debug your code. Be sure to only do that if you run salt locally.


Belgian Beerware.


I've had fun writing it, hopes you'll have using it. You don't want to know how it's made.

¹: I have another experiment that try to solve this problem, but I'm not writing enough salt right now to move on it.