Skittle

A tiny tool for simplifying system provisioning with simple bash scripts.

[] (http://travis-ci.org/d11wtq/skittle)

Overview

Skittle makes it really easy to script install procedures, configure computers or do just about anything that takes multiple steps using bash.

Installation

Because Skittle aims for zero dependencies, it is not intended that you install it in your system, instead prefering that you just add it to your project directly. It is less than 200 lines long (including whitespace and function definitions) and lives in a single file.

curl -O https://raw.githubusercontent.com/d11wtq/skittle/master/bin/skittle
chmod +x skittle

If you do want to add it to your system, go ahead place it on your $PATH.

Usage

There are very few concepts to learn when it comes to using Skittle. The first concept is knowing that the skittle executable takes a single argument, which is the name of a dependency it should resolve.

The second concept, is that a dependency is just a function that defines two other functions: one to indicate if it needs to run, and another that specifies how to run it.

Finally, dependencies may specify other depdencies that they require to run (i.e. they are recursively defined).

Let's look at a simple example of creating a log directory for a fictonal service called "turtle" (It was just a name that came to mind!)

We'll run this dependency like so.

bash-3.2$ ./skittle log_dir
+ log_dir
|
\ [fail] log_dir
Error: Cannot find dependency 'log_dir'

As you might expect, this produces an error, because we haven't actually specified what the log_dir dependency should do. We need to create a file named "./deps/log_dir.sh" and add a small amount of code.

# ./deps/log_dir.sh

log_dir() {
  dir_path=/var/log/turtle

  is_met() {
    ls $dir_path
  }

  meet() {
    sudo mkdir $dir_path
  }
}

Save the file and go ahead and run skittle log_dir again. You should see everything go green and an [ok] indicator.

bash-3.2$ ./skittle log_dir
+ log_dir
|
\ [ok] log_dir
bash-3.2$

Note sudo may prompt for your password when you run this dep.

If you take a look , /var/log/turtle should have been created.

Quite simply, Skittle looks for your depedency under ./deps, using the convention that it has a file name matching the dependency, but ending in ".sh".

Next, it evaluates the function—in a subshell—knowing that it produces is_met and meet functions.

If the is_met function returns a zero exit status, nothing happens. If, however it returns non-zero, it runs the meet function which should cause subsequent calls to is_met to return zero. Following this pattern makes Skittle dependencies idempotent. Go ahead and run it again. It still returns ok.

Chaining dependencies

Ok, so this example was a bit basic. Creating the directory alone is probably not enough. Let's ensure that directory is writable only to a user named 'turtle'. We now have to do four things to satisfy log_dir:

1. Ensure the directory exists.
2. Ensure the turtle user exists.
3. Ensure the directory is owned by turtle.
4. Ensure the directory has 755 permissions.

We'll use Skittle's require function for this.

# ./deps/log_dir.sh

log_dir() {
  dir_path=/var/log/turtle

  require dir_exists
  require dir_ownership
  require dir_permissions

  dir_exists() {
    is_met() {
      ls $dir_path
    }

    meet() {
      sudo mkdir $dir_path
    }
  }

  dir_ownership() {
    turtle_user_exists() {
      is_met() {
        id turtle
      }

      meet() {
        sudo useradd turtle
      }
    }

    is_met() {
      [[ `ls -ld $dir_path | awk '{print $3}'` = "turtle" ]]
    }

    meet() {
      sudo chown -R turtle: $dir_path
    }

    require turtle_user_exists
  }

  dir_permissions() {
    is_met() {
      [[ `ls -ld $dir_path | awk '{print $1}'` = "drwxr-xr-x" ]]
    }

    meet() {
      sudo chmod -R 0755 $dir_path
    }
  }
}

Now when we run skittle log_dir, we see it executes a tree. Again, this is idempotent—you can run it over and over just fine.

bash-3.2$ ./skittle log_dir
+ log_dir
|
+-+ dir_exists
| |
| \ [ok] dir_exists
|
+-+ dir_ownership
| |
| +-+ turtle_user_exists
| | |
| | \ [ok] turtle_user_exists
| |
| \ [ok] dir_ownership
|
+-+ dir_permissions
| |
| \ [ok] dir_permissions
|
\ [ok] log_dir
bash-3.2$

Notice that the original log_dir code moved into an inner depedency called dir_exists, since it was only checking if the log directory existed. The other requirements have then been stated using require.

Notice also that dir_ownership has a nested dependency turtle_user_exists. This is allowed with Skittle so that you can group related dependencies together. In this case however, turtle_user_exists probably doesn't belong here, as it is not directly related to the log directory. It is very simple to just move the function definition to ./deps/turtle_user_exists.sh. Try it, everything should work the same.

Parameterized dependencies

Sometimes it is useful for dependencies to accept arguments, so that they become more general and re-usable. A great example of this is turtle_user_exists from our earlier example. The code here could work for any user, so we can generalize it and accept a username as an argument. Because dependencies are just bash functions, arguments are numbered $1, $2 etc.

# ./deps/user_exists.sh

user_exists() {
  username=$1

  is_met() {
    id $username
  }

  meet() {
    sudo useradd $username
  }
}

Now we can change log_dir to use this generalized dep instead.

# ./deps/log_dir.sh

log_dir() {
  dir_path=/var/log/turtle

  # ... snip ...

  dir_ownership() {
    require user_exists turtle

    is_met() {
      [[ `ls -ld $dir_path | awk '{print $3}'` = "turtle" ]]
    }

    meet() {
      sudo chown -R turtle: $dir_path
    }
  }

  # ... snip ...
}

Note It is important to store the function arguments to variables so they can be used inside is_met and meet.

Logging/debugging

Sometimes you'll want to provide some arbitrary output to the console while Skittle is processing the dependency tree. Using echo won't work, since Skittle gathers all output on stdout and stderr into the log file. Instead, Skittle provides log, which prints a message to the console and reflects which branch of the dependency tree this log output comes from.

You can use log anywhere inside the main function body of the dep, or inside is_met and meet.

For example, in our generic user_exists dep, it might be a good idea to show which user the dep is being run for.

# ./deps/user_exists.sh

user_exists() {
  username=$1

  log "Checking user: $username"

  is_met() {
    id $username
  }

  meet() {
    sudo useradd $username
  }
}

This will output something along the lines of.

bash-3.2$ ./skittle log_dir
+ log_dir
|
+-+ dir_exists
| |
| \ [ok] dir_exists
|
+-+ dir_ownership
| |
| +-+ user_exists
| | |
| | \ [..] Checking user: turtle
| | |
| | \ [ok] user_exists
| |
| \ [ok] dir_ownership
|
+-+ dir_permissions
| |
| \ [ok] dir_permissions
|
\ [ok] log_dir
bash-3.2$

Another suggested use of log is to provide feedback to the user during long running operations.

compile_source() {
  cd project_src/

  is_met() {
    [[ -f some_binary ]]
  }

  meet() {
    log "Compiling (may take some time)"
    make
  }
}

Grouping multiple deps together

Ok, so we can create a log directory for our turtle service just fine. You could go on to install the entire turtle service, by breaking the problem down into similarly small tasks and putting them all under a top level dependency, like so.

install_turtle() {
  require copy_files
  require runlevels
  require log_dir
  require data_dir
}

Running ./skittle install_turtle will run each of the deps listed in order. If any single dep fails, the entire install process will halt. It's not hard to see how something like this could be wrapped under a simple install.sh for users of the turtle service to run.

That's pretty much all there is to it!

Composing dependency trees

At some point you may want to break your dependencies down into separate projects in order to allow for re-use in different ways. In general, this is as simple as adding nested directories under ./deps, however, there is one caveat: if your deps are coded using relative paths to support files, such as the following, you may run into trouble nesting your deps, since $PWD is relative to the user running skittle, not relative to the dependency file itself.

etc_issue() {
  wanted=./deps/issue_file.txt

  is_met() {
    diff $wanted /etc/issue
  }

  meet() {
    sudo cp -f $wanted /etc/issue
  }
}

If this dep is nested at, say ./deps/common/deps/etc_issue.sh, it will not do what you expect due to the relative path.

For this purpose, Skittle provides $p, which is implicitly set for each dep being run. This special variable defines the directory from which the dependency file itself was loaded. You should use it everywhere you mean "relative to the current dependency".

etc_issue() {
  wanted=$p/issue_file.txt

  is_met() {
    diff $wanted /etc/issue
  }

  meet() {
    sudo cp -f $wanted /etc/issue
  }
}

Now the above dep will run anywhere it is copied to.

Tests

Skittle uses Skittle to test itself. To run the tests, run skittle with the 'tests' dep.

./bin/skittle tests

Reading through the test code (in the deps directory) is good way to see an example of Skittle code too.

Other features

• Dependencies can be stored in subdirectories
• Other supported locations for ./deps are ~/skittle-deps and ./skittle-deps
• The dependency function is unset before its is_met and meet functions are run. This makes it safe to name a dependency the same as a system binary (i.e. if you execute the system binary git in a dep named git, you will indeed execute /usr/bin/git, rather than recurse into the dep function)

Donations

I work on open source projects for free and because I genuinely enjoy giving to the community, but of course any donations are well-received. You can donate a small amount via gittip if you're feeling generous.

Copyright & Licensing

Copyright © 2014 Chris Corbyn. See the LICENSE file for details.