The Bash-Operated Reconciling Kludge

Project maintained by borksh Hosted on GitHub Pages — Theme by mattgraham

bork(1) – The Bash-Operated Reconciling Kludge


bork operation [config-file] [options]


Bork is a bash DSL for making declarative assertions about the state of a system.

Bork is written against Bash 3.2 and common unix utilities such as sed, awk and grep. It is designed to work on any UNIX-based system and maintain awareness of platform differences between BSD and GPL versions of unix utilities.


At the heart of bork is making assertions in a declarative manner via the ok and no functions. That is, you tell it what you want the system to look like instead of how to make it look like that. An assertion takes a type and a number of arguments. It invokes the type’s handler function with an action such as status, install, or upgrade, which determines the imperative commands needed to test the assertion or bring it up to date. There are a number of included types in the types directory, and bork makes it easy to create your own. The no function works as an opposite to ok – an ok assertion will require the presence of something, and a no assertion will require its absence.

When run, bork will test each ok/no assertion and determine if it’s met or not. If not, bork can go ahead and satisfy the assertion by installing, upgrading, removing, or otherwise altering the configuration of the item to match the assertion. It will then test the assertion again. Declarations are idempotent – if the assertion is already met, bork will not do anything.

When you’re happy with your config script, you can compile it to a standalone script which does not require bork to run. The compiled script can be passed around via curl, scp or the like and run on completely new systems.


You can run bork types from the command line to get a list of the assertion types and some basic information about their usage and options.


Per the usage guide, bork has a few main modes of operation:

status config-file

The status command will confirm that assertions are met or not, and output their status. It will not take any action to satisfy those assertions. There are a handful of statuses an assertion can return, and this since this mode is the closest bork can do to a true ‘dry run’*, you can use it to test a script against a pre-existing machine.

* Some types, such as git, need to modify local state by talking to the network (such as performing git fetch), without modifying the things the assertion aims to check.

The status command will give you output such as:

outdated: brew
ok: brew git
missing: brew zsh
ok: directory /Users/skylar/code
conflict (upgradable): github skylarmacdonald/dotfiles
local git repository has uncommitted changes
ok: symlink /Users/skylar/.gitignore /Users/skylar/code/dotfiles/configs/gitignore
conflict (clobber required): symlink /Users/skylar/.lein /Users/skylar/code/dotfiles/configs/lein
not a symlink: /Users/skylar/.lein
mismatch (upgradable): defaults tilesize integer 36
expected type: integer
received type: float
expected value: 36
received value: 55

Each item reports its status like so:

check assertion

The check command will take a single assertion on the command line and perform a status check as above for it.

satisfy config-file

The satisfy command is where the real magic happens. For every assertion in the config file, bork will check its status as described in the status command above, and if it is not ok or no it will attempt to make it ok or no, typically via installing, upgrading or removing something – but sometimes a conflict is detected which could lose data, such as a local git repository having uncommitted changes. In that case, bork will warn you about the problem and ask if you want to proceed. Sometimes conflicts are detected which bork does not know how to resolve — it will warn you about the problem so you can fix it yourself.

do assertion

The do command will take a single assertion on the command line and perform a satisfy operation on it as above.

compile config-file

The compile command will output to STDOUT a standalone shell script that does not require bork to run. You may pass this around as with any file via curl or scp or whatever you like and run it. Any sub-configs via include will be included in the output, and any type that needs to include resources to do what it does, such as the file type, will include their resources in the script as base64 encoded data.

inspect type

The inspect command will ask a type for a current inventory of how a system is configured, and output to STDOUT a Bork-compatible config file to configure the same state. For example, when used with the brew type, this will list all formulae installed with Homebrew and output a config file to check for those same formulae. Not all types will work with this command. Bork will exit with code 1 if a type has not implemented inspect.


Writing new types is pretty straightforward, and there is a guide to writing them in the docs/ directory. If you wish to use a type that is not in bork’s types directory, you can let bork know about it with the register declaration:

register etc/
ok pgdb my_app_db


You may compose config files into greater operations with the include directive with a path to a script relative to the current script’s directory.

# this is
include etc/
# this is etc/
# these will be read from the etc/ directory


Bork has two types of callback: before and after functions. These are only used when Bork is satisfying assertions (i.e. when running bork satisfy).

Until Bork starts processing an assertion made with ok or no, there’s no way to know if anything will change. Therefore, Bork will look for and execute functions with known names while it processes an assertion, before making the change.

The functions Bork expects are named:

Each of these will be unset by Bork after it has run them. You should only define these functions immediately before the assertion you wish to apply them to.

Bork will also call all of these functions with _any appended to the names (e.g. bork_will_change_any) – these callbacks will not be unset, and will be called every time it applies.

These are used as follows:

bork_will_install () {
  echo "callback says hello world"
ok directory foo

Bork will then output the following if (and only if) the directory foo has been newly created:

missing: directory foo
callback says hello world
verifying install: directory foo
* success

If the directory had already existed, the bork_will_install function would not have been called. Bork would also not have called the function if it had upgraded the state of the system, e.g. if the directory had existed but had the incorrect permissions.

After Bork has made a change, you may call a provided function in your script to determine the outcome of the change. These are used as follows:

ok brew fish
if did_install; then
  ok shells $(brew --prefix)/bin/fish
  chsh -s $(brew --prefix)/bin/fish

There are five functions to help you take further actions after a change:

Unlike with before callbacks, Bork will not call any functions after making a change. It is up to you to handle the logic however you wish. As with the before callbacks, you are strongly advised to use these functions immediately after the assertion you wish to check.


Bork documentation:


Bork’s lead maintainer is Skylar MacDonald. It was created and previously maintained by Matthew Lyon.

A full list of contributors is available on GitHub:


See our issues on GitHub: