Booty is a language and command line utility for bootstrapping the setup of personal OS installs. Its goal is to execute all of the things that you do after your first boot, like install packages through your package manager, configure your shell, setup your terminal tools/IDE, and install SDKs for various programming languages, without you having to download everything manually and run various scripts in the right order.
You model your setup process as Targets and declare which ones depend on which ones, then execute booty
to set everything up, if
everything isn't already setup. Your install.booty file will specify various setup
and is_setup
methods that booty
uses to set your
system up, and to test if everything was actually setup.
Installing from pip
pip install booty-cli
Or download the appropriate binary from the latest release. This script just downloads it for you and marks it as executable, but you should
move it somewhere on your path. The script will drop it in your pwd
.
curl https://raw.githubusercontent.com/naddeoa/booty/master/scripts/booty-download-linux.sh | bash
curl https://raw.githubusercontent.com/naddeoa/booty/master/scripts/booty-download-mac-x86.sh | bash
curl https://raw.githubusercontent.com/naddeoa/booty/master/scripts/booty-download-mac-universal.sh | bash
Just execute booty
from a directory with an install.booty
file.
booty
You can run booty --help
to see all of the options. You'll generally cd to a directory with an install.booty
file and run booty
.
You'll see a table that shows the status of all of your targets and you'll be prompted to install all of the missing ones, which will
display in real-time as a second table with a row for each target getting setup.
Usage: booty [OPTIONS]
Options:
-c, --config TEXT Path to the booty file. Defaults to ./install.booty
-s, --status Check the status of all known targets
-i, --install Install all uninstalled targets
-d, --debug See the AST of the config file
-l, --log-dir TEXT Where to store logs. Defaults to ./logs
--no-sudo Don't allow booty to prompt with sudo -v. Instead, you
can manually run sudo -v before using booty to cache
credentials for any targets that use sudo. By default,
booty runs sudo -v upfront if you use sudo in any
targets.
-y, --yes Don't prompt for confirmation
--help Show this message and exit.
Setting your system up is inherently side effect prone. Every command from apt
to pip
is probably doing something to some part of your
system that you didn't realize. If you want to execute your install.booty
file without worrying about what might go wrong before you use
it for real then the best way to do that is via Docker (which is how this repo does "integration" testing).
There is a public docker image naddeoa/booty:ubuntu22.04 that you can
base your own Dockerfile on that mimics what a fresh install looks like for Ubuntu 22.04 users. You can create an image that does roughly
what you would do to your system to get the booty
command working, and then run it and make sure everything works as expected. The image's
user is named myuser
and its password is password
. You can see how its created in
Dockerfile.base.
This Dockerfile is what I use to test my own installs.
from naddeoa/booty:ubuntu22.04
RUN sudo apt-get install -y curl # Just like IRL, install curl first
RUN curl https://raw.githubusercontent.com/naddeoa/booty/master/scripts/booty-download-linux.sh | bash # MANUAL install booty without pip
# Copy my ssh keys in so I can clone my private repos
COPY ./ssh ./.ssh
RUN sudo chown -R myuser:myuser ./.ssh
RUN chmod 600 .ssh/id_rsa
COPY ./examples/install.booty ./
# This would be on my path IRL
ENV PATH="/home/myuser/.local/bin:${PATH}"
CMD ["bash"]
Then you build and run the image, and execute booty.
# Build the image
docker build . -t my-booty-test
# Run it
docker run --rm -it --entrypoint bash my-booty-test
## Inside the image now
./booty_linux_x86_64 -y
You can't test everything here but you can get pretty far. One thing I can't test here, for example, is chsh
because that requires logging
out, but I gets me confident enough to use it on a fresh install and know I won't mess things up. In practice, you do this sort of a test
infrequently, when you initially create your install.booty
file or when you make big changes to it.
See naddeoa/vim-booty for the syntax plugin.
For full examples, check out the examples folder in github.
The Booty language is a small language inspired primarily by make, with some additions to better address the problem space. There are a few different types of entities.
Recipes are reusable pieces of code that Targets and other Recipes can invoke. You can think of these as classes that implement a Recipe
by defining the setup
and is_setup
methods. Recipes are invoked with parameters being separated by commas, like recipe_name(foo, bar)
.
Parameters can optionally container whitespace as well, meaning recipe_name(a b c)
is a single parameter a b c
. The body of the recipe
methods consists of one or more executable statements which can either be a line to invoke in shell or another recipe invocation.
This is a recipe named apt
who's setup
executes the shell command sudo apt-get install -y $((packages))
, where $((packages))
will be
substituted by the value of the parameter packages
when it's run. Its is_setup
executes a multiline shell statement. The shell is
invoked via bash -c ...
.
recipe apt(packages):
setup: sudo apt-get install -y $((packages))
is_setup:
for pkg in $(echo $((packages)) | tr " " "\n"); do
if ! dpkg -l "$pkg" &> /dev/null; then
echo "$pkg is not installed."
exit 1
fi
done
Targets are the main piece of a booty file. They invoke a recipe to accomplish their goal. This is a target named essentials
that invokes
the apt
recipe, passing it the parameter wget git vim autokey-gtk silversearcher-ag gawk xclip
.
essentials: apt(wget git vim autokey-gtk silversearcher-ag gawk xclip)
Custom Targets are similar to recipes, but they're defined inline. You would use this for something that didn't require any code shared between other recipes. The body of a Custom Target follows the same rules as a Recipe.
pyenv:
setup: curl https://pyenv.run | bash
is_setup: test -e ~/.pyenv/bin/pyenv
# Can also be multiline and invoke different commands/recipes.
pyenv:
setup:
apt(python3)
curl https://pyenv.run | bash
is_setup: test -e ~/.pyenv/bin/pyenv
Dependencies determine the execution order at setup time. The way that you declare this is flexible. You can use either the depends on
syntax (->
) or the depended upon
syntax (<-
). This is allowed because it's sometimes easier to manage a bunch of dependencies in a
single line when they're logically related. Its personal preference for how you maintain your install.booty file.
# the target `essentials` is depended upon by foo and bar.
essentials <- foo bar
# the target `baz` depends on bar.
baz -> bar
Certain recipes are included in booty by default. These include the following.
recipe apt(packages):
setup: sudo apt-get install -y $((packages))
is_setup:
for pkg in $(echo $((packages)) | tr " " "\n"); do
if ! dpkg -l "$pkg" &> /dev/null; then
echo "$pkg is not installed."
exit 1
fi
done
recipe ppa(name):
setup:
sudo add-apt-repository $((name))
sudo apt update
is_setup: grep $((name)) /etc/apt/sources.list
recipe pipx(packages):
setup: pipx install $((packages))
is_setup: pipx list | grep $((packages))
recipe git(repo dist):
setup: git clone $((repo)) $((dist))
is_setup: test -d $((dist))
recipe git_shallow(repo dist):
setup: git clone --depth 1 $((repo)) $((dist))
is_setup: test -d $((dist))
recipe ln(src dst):
setup:
mkdir -p $(dirname $((dst)))
ln -fs $((src)) $((dst))
is_setup: test -L $((dst)) && test -e $((dst))
recipe cp(src dst):
setup:
mkdir -p $(dirname $((dst)))
cp -r $((src)) $((dst))
is_setup: test -e $((dst))
Any of these can be referenced from any booty.install file, and you can redefine them locally if you want different recipe logic that uses the same name.
You can check the experiments folder to see what it looks like to implement the example booty file in make. It's a fair bit longer and clunkier for a few reasons:
- Every target in make is designed to be an actual file. This great for building things but it means that you end up marking a lot of
targets as
.PHONY
if they don't result in differences on the file system. - Implementing the
--status
flag from booty is very, very ugly. You'll just manually define two modes of each target and hard code a phony target that runs them all, with no particular attention paid to the output. - Same goes for running the setup code -- manually enumerating all targets as a phony target's dependency gets you the
--install
flag in booty. - Output in general isn't very digestible.
- Each line of a target is executed in a new shell. Good for sandboxing commands but it gets very ugly when you want to do something like an
if
statement.
There are a lot of good things about make though. My favorite relevant parts are:
- Easy, independent dependency specification
- Plain old shell for each target definition.