bork

The Bash-Operated Reconciling Kludge


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

Skylar MacDonald’s Bork Fork

I still use Bork in the year 2021, so I forked it to fix it.

Bork puts the ‘sh’ back into IT. Bork Bork Bork.

the Swedish Chef Puppet of Config Management

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.

Installation

From source

  1. Clone this repository:
    git clone https://github.com/borksh/bork /usr/local/src/bork
    
  2. Symlink the bork binaries into your $PATH:
      ln -sf /usr/local/src/bork/bin/bork /usr/local/bin/bork
    

via Homebrew (macOS)

  1. Install via Homebrew:
    brew install bork
    

via npm

  1. Install via npm:
    npm install -g @borksh/bork
    

via pre-built package

Starting with version 0.13.0, packages are available for a handful of operating systems on the GitHub releases page. These are generally from CI and built unsigned, but SHA-1 hashes are always available.

Updating

Bork can update itself as part of satisfying your config file. Your config file should look something like this to update via git:

ok github /usr/local/src/bork borksh/bork --branch=main
ok symlink /usr/local/bin/bork /usr/local/src/bork/bin/bork

(This example relies on you being able to write to /usr/local; if your Bork is installed elsewhere you should replace the paths above.)

If you have Homebrew available to you, you can do this instead:

ok brew bork

You can also specify the --HEAD option on the assertion to install Bork’s main branch via Homebrew:

ok brew bork --HEAD

This will always keep the latest commit installed. Note that the latest commit will contain unreleased code that might break, so take care when using it.

Using a package manager is the recommended way to install, as then you can ensure you’re only installing released versions of Bork, and rely on it to update Bork for you. If you prefer to use git, you can use bork version to show the status of your local repo or installation. This command should be able to tell you how you installed Bork (e.g. via git or Homebrew), and therefore how you should go about updating it.

Usage and Operations

Running bork without arguments will output some help:

bork usage:

bork operation [config-file] [options]

where "operation" is one of:

- check:      perform 'status' for a single command
    example:  bork check ok github skylarmacdonald/dotfiles
- compile:    compile the config file to a self-contained script output to STDOUT
    --conflicts=(y|yes|n|no)  If given, sets an automatic answer for conflict resolution.
    example:  bork compile dotfiles.sh --conflicts=y > install.sh
- do:         perform 'satisfy' for a single command
    example:  bork do ok github skylarmacdonald/dotfiles
- satisfy:    satisfy the config file's conditions if possible
- status:     determine if the config file's conditions are met
- inspect:    output a Bork config file based on a type's current configuration
- types:      list types and their usage information
- docgen:     generates documentation under docs/_types for newly-added types
- version:    get the currently installed version of bork

Let’s explore these in more depth:

Assertions and Config Files

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.

Here’s a basic example:

ok brew                                                # presence and updatedness of Homebrew
ok brew git                                            # presence and updatedness of Homebrew git package
ok directory $HOME/code                                # presence of the ~/code directory
ok github $HOME/code/dotfiles skylarmacdonald/dotfiles # presence, drift of git repository in ~/code/dotfiles
cd $HOME
for file in $HOME/code/dotfiles/configs/.[!.]*
do                                            # for each file in ~/code/dotfiles/configs,
  ok symlink "$(basename $file)" $file       # presense of a symlink to file in ~ with a leading dot
done

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.

Assertion Types

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.

If adding features to Bork core, you can also use the command bork docgen to generate GitHub Pages-compatible Markdown files based on how a type responds to the desc action.

Generic assertions

          check: runs a given command.  OK if returns 0, FAILED otherwise.

File System

      directory: asserts presence of a directory
           file: asserts the presence, checksum, owner and permissions of a file
       download: asserts the presence of a file compared to an http(s) url
        symlink: assert presence and target of a symlink

Source Control

            git: asserts presence and state of a git repository
         github: front-end for git type, uses github urls

Language Package Managers

            gem: asserts the presence of a gem in the environment's ruby
            npm: asserts the presence of a nodejs module in npm's global installation
            pip: asserts presence of packages installed via pip
           pip3: asserts presence of packages installed via pip3
          pipsi: asserts presence of pipsi or packages installed via pipsi
            apm: asserts the presence of an atom package
         go-get: asserts the presence of a go package

macOS specific

           brew: asserts presence of packages installed via Homebrew on macOS
       brew-tap: asserts a Homebrew formula repository has been tapped; does NOT assert updatedness of a tap's formula. Use `ok brew` for that.
           cask: asserts presence of apps installed via caskroom.io on macOS
       defaults: asserts settings for macOS's 'defaults' system
            mas: asserts a Mac app is installed and up-to-date from the App Store
                 via the 'mas' utility https://github.com/argon/mas
         scutil: verifies macOS machine name with scutil

Linux specific:

            apt: asserts packages installed via apt-get on Debian or Ubuntu Linux
            apk: asserts packages installed via apk (Alpine Linux)
            yum: asserts packages installed via yum on CentOS or RedHat Linux
         zypper: asserts packages installed via zypper (SUSE)

User management (currently Linux-only)

          group: asserts presence of a unix group (Linux only, for now)
           user: assert presence of a user on the system

UNIX utilities

       iptables: asserts presence of iptables rule
         shells: asserts presence of a shell in /etc/shells

Runtime Operations

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

bork status myconfig.sh

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 com.apple.dock tilesize integer 36
expected type: integer
received type: float
expected value: 36
received value: 55

Each item reports its status like so:

bork check ok github skylarmacdonald/dotfiles

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

bork satisfy myconfig.sh

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.

bork do ok github skylarmacdonald/dotfiles

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

bork compile myconfig.sh

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.

bork inspect brew

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.

Custom Types

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/pgdb.sh
ok pgdb my_app_db

Composing Config Files

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 main.sh
include databases.sh
include etc/projects.sh
# this is etc/projects.sh
include project-one.sh
include project-two.sh
# these will be read from the etc/ directory

Taking Further Action on Changes

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
fi

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.

Contributing

  1. Fork it
  2. Create your feature branch: git checkout -b feature/my-new-feature
  3. Commit your changes: git commit -am 'Add some feature'
  4. Push to the branch: git push origin feature/my-new-feature
  5. Submit a pull request

Contribution Guidelines

  1. Prefer clarity of intent over brevity. Bash can be an obtuse language, but it doesn’t have to be. Many people have said bork has some of the clearest bash code they’ve ever seen, and that’s a standard to strive for.

  2. Favor helper abstractions over arbitrary platform-specific checks. See md5cmd, http, and permission_cmd, and look at how they’re used.

  3. Types are independent, stateless, and atomic. Do not attempt to maintain a cache in a type file unless you’re talking to the network. An assertion is the whole of the assertion — don’t attempt to create a multi-stage assertion type that depends on maintaining state. Find a way to express the whole of the assertion in one go.

  4. Leave Dependency Management to the user. Is a needed binary not installed for a type? Return $STATUS_FAILED_PRECONDITION in your status check. Let the user decide the best way to satisfy any dependencies.

Community

Discuss on GitHub Discussions

Requirements / Dependencies

Version

0.14.0

License

Apache License 2.0