disu.se

Inventory

Inventory keeps track of the contents of your Ruby projects. Such an inventory can be used to load the project, create gem specifications and gems, run unit tests, compile extensions, and verify that the project’s content is what you think it is.

Usage

Let’s begin by discussing the project structure that Inventory expects you to use. It’s pretty much exactly the same as the standard Ruby project structure:

├── README
├── Rakefile
├── lib
│   ├── foo-1.0
│   │   ├── bar.rb
│   │   └── version.rb
│   └── foo-1.0.rb
└── test
    └── unit
        ├── foo-1.0
        │   ├── bar.rb
        │   └── version.rb
        └── foo-1.0.rb

Here you see a simplified version of a project called “Foo”’s project structure. The only real difference from the standard is that the main entry point into the library is named “foo-1.0.rb” instead of “foo.rb” and that the root sub-directory of “lib” is similarly named “foo-1.0” instead of “foo”. The difference is the inclusion of the API version. This must be the major version of the project followed by a constant “.0”. The reason for this is that it allows concurrent installations of different major versions of the project and means that the wrong version will never accidentally be loaded with require.

There’s a bigger difference in the content of the files. Lib/foo-1.0/version.rb will contain our inventory instead of a String:

require 'inventory-1.0'

class Foo
  Version = Foo.new(1, 4, 0){
    authors{
      author 'A. U. Thor', 'a.u.thor@example.org'
    }

    homepage 'http://example.org/'

    licenses{
      license 'LGPLv3+',
              'GNU Lesser General Public License, version 3 or later',
              'http://www.gnu.org/licenses/'
    }

    def dependencies
      super + Dependencies.new{
        development 'baz', 1, 3, 0
        runtime 'goo', 2, 0, 0
        optional 'roo-loo', 3, 0, 0, :feature => 'roo-loo'
      }
    end

    def package_libs
      %w[bar.rb]
    end
  }
end

We’re introducing quite a few concepts at once, and we’ll look into each in greater detail, but we begin by setting the Version constant to a new instance of an Inventory with major, minor, and patch version atoms 1, 4, and 0. Then we add a couple of dependencies and list the library files that are included in this project.

The version numbers shouldn’t come as a surprise. These track the version of the API that we’re shipping using semantic versioning. They also allow the Inventory#to_s method to act as if you’d defined Version as '1.4.0'.

Next follows information about the authors of the project, the project’s homepage, and the project’s licenses. Each author has a name and an email address. The homepage is simply a string URL. Licenses have an abbreviation, a name, and a URL where the license text can be found.

We then extend the definition of dependencies by adding another set of dependencies to super. Super includes a dependency on the version of the inventory project that’s being used with this project, so you’ll never have to list that yourself. The other three dependencies are all of different kinds: development, runtime, and optional. A development dependency is one that’s required while developing the project, for example, a unit-testing framework, a documentation generator, and so on. Runtime dependencies are requirements of the project to be able to run, both during development and when installed. Finally, optional dependencies are runtime dependencies that may or may not be required during execution. The difference between runtime and optional is that the inventory won’t try to automatically load an optional dependency, instead leaving that up to you to do when and if it becomes necessary. By that logic, runtime dependencies will be automatically loaded, which is a good reason for having dependency information available at runtime.

The version numbers of dependencies also use semantic versioning, but note that the patch atom is ignored unless the major atom is 0. You should always only depend on the major and minor atoms.

As mentioned, runtime dependencies will be automatically loaded and the feature they try to load is based on the name of the dependency with a “-X.0” tacked on the end, where ‘X’ is the major version of the dependency. Sometimes, this isn’t correct, in which case the :feature option may be given to specify the name of the feature.

You may also override other parts of a dependency by passing in a block to the dependency, much like we’re doing for inventories.

The rest of an inventory will list the various files included in the project. This project only consists of one additional file to those that an inventory automatically include (Rakefile, README, the main entry point, and the version.rb file that defines the inventory itself), namely the library file bar.rb. Library files will be loaded automatically when the main entry point file loads the inventory. Library files that shouldn’t be loaded may be listed under a different heading, namely “additional_libs”. Both these sets of files will be used to generate a list of unit test files automatically, so each library file will have a corresponding unit test file in the inventory. We’ll discuss the different headings of an inventory in more detail later on.

Now that we’ve written our inventory, let’s set it up so that it’s content gets loaded when our main entry point gets loaded. We add the following piece of code to lib/foo-1.0.rb:

module Foo
  load File.expand_path('../foo-1.0/version.rb', __FILE__)
  Version.load
end

That’s all there’s to it.

The inventory can also be used to great effect from a Rakefile using a separate project called Inventory-Rake. Using it’ll give us tasks for cleaning up our project, compiling extensions, installing dependencies, installing and uninstalling the project itself, and creating and pushing distribution files to distribution points.

require 'inventory-rake-1.0'

load File.expand_path('../lib/foo-1.0/version.rb', __FILE__)

Inventory::Rake::Tasks.define Foo::Version

Inventory::Rake::Tasks.unless_installing_dependencies do
  require 'lookout-rake-3.0'
  Lookout::Rake::Tasks::Test.new
end

It’s Inventory::Rake::Tasks.define that does the heavy lifting. It takes our inventory and sets up the tasks mentioned above.

As we want to be able to use our Rakefile to install our dependencies for us, the rest of the Rakefile is inside the conditional #unless_installing_dependencies, which, as the name certainly implies, executes its block unless the task being run is the one that installs our dependencies. This becomes relevant when we set up Travis integration next. The only conditional set-up we do in our Rakefile is creating our test task via Lookout-Rake, which also uses our inventory to find the unit tests to run when executed.

Travis integration is straightforward. Simply put

before_script:
  - gem install inventory-rake -v '~> VERSION' --no-rdoc --no-ri
  - rake gem:deps:install

in the project’s .travis.yml file, replacing VERSION with the version of Inventory-Rake that you require. This’ll make sure that Travis installs all development, runtime, and optional dependencies that you’ve listed in your inventory before running any tests.

You might also need to put

env:
  - RUBYOPT=rubygems

in your .travis.yml file, depending on how things are set up.

API

If the guide above doesn’t provide you with all the answers you seek, you may refer to the API for more answers.

Financing

Currently, most of my time is spent at my day job and in my rather busy private life. Please motivate me to spend time on this piece of software by donating some of your money to this project. Yeah, I realize that requesting money to develop software is a bit, well, capitalistic of me. But please realize that I live in a capitalistic society and I need money to have other people give me the things that I need to continue living under the rules of said society. So, if you feel that this piece of software has helped you out enough to warrant a reward, please PayPal a donation to now@disu.se. Thanks! Your support won’t go unnoticed!

Reporting Bugs

Please report any bugs that you encounter to the issue tracker.

Authors

Nikolai Weibull wrote the code, the tests, the documentation, and this README.

Licensing

Inventory is free software: you may redistribute it and/or modify it under the terms of the GNU Lesser General Public License, version 3 or later, as published by the Free Software Foundation.