Simple, Heroku-friendly Rails app configuration using ENV and a single YAML file

NOTE: If you're using Figaro 0.7 or prior, please refer to the appropriate documentation or upgrade to Figaro 1.0.

Why does Figaro exist?

Figaro was written to make it easy to securely configure Rails applications.

Configuration values often include sensitive information. Figaro strives to be secure by default by encouraging a convention that keeps configuration out of Git.

How does Figaro work?

Figaro is inspired by the Twelve-Factor App methodology, which states:

The twelve-factor app stores config in environment variables (often shortened to env vars or env). Env vars are easy to change between deploys without changing any code; unlike config files, there is little chance of them being checked into the code repo accidentally; and unlike custom config files, or other config mechanisms such as Java System Properties, they are a language- and OS-agnostic standard.

This is straightforward in production environments but local development environments are often shared between multiple applications, requiring multiple configurations.

Figaro parses a Git-ignored YAML file in your application and loads its values into ENV.

Getting Started

Add Figaro to your Gemfile and bundle install:

gem "figaro"

Figaro installation is easy:

$ bundle exec figaro install

This creates a commented config/application.yml file and adds it to your .gitignore. Add your own configuration to this file and you're done!

Example

Given the following configuration file:

# config/application.yml

pusher_app_id: "2954"
pusher_key: "7381a978f7dd7f9a1117"
pusher_secret: "abdc3b896a0ffb85d373"

You can configure Pusher in an initializer:

# config/initializers/pusher.rb

Pusher.app_id = ENV["pusher_app_id"]
Pusher.key    = ENV["pusher_key"]
Pusher.secret = ENV["pusher_secret"]

Please note: ENV is a simple key/value store. All values will be converted to strings. Deeply nested configuration structures are not possible.

Environment-Specific Configuration

Oftentimes, local configuration values change depending on Rails environment. In such cases, you can add environment-specific values to your configuration file:

# config/application.yml

pusher_app_id: "2954"
pusher_key: "7381a978f7dd7f9a1117"
pusher_secret: "abdc3b896a0ffb85d373"

test:
  pusher_app_id: "5112"
  pusher_key: "ad69caf9a44dcac1fb28"
  pusher_secret: "83ca7aa160fedaf3b350"

You can also nullify configuration values for a specific environment:

# config/application.yml

google_analytics_key: "UA-35722661-5"

test:
  google_analytics_key: ~

Using Figaro.env

Figaro.env is a convenience that acts as a proxy to ENV.

In testing, it is sometimes more convenient to stub and unstub Figaro.env than to set and reset ENV. Whether your application uses ENV or Figaro.env is entirely a matter of personal preference.

# config/application.yml

stripe_api_key: "sk_live_dSqzdUq80sw9GWmuoI0qJ9rL"

ENV["stripe_api_key"] # => "sk_live_dSqzdUq80sw9GWmuoI0qJ9rL"
ENV.key?("stripe_api_key") # => true
ENV["google_analytics_key"] # => nil
ENV.key?("google_analytics_key") # => false

Figaro.env.stripe_api_key # => "sk_live_dSqzdUq80sw9GWmuoI0qJ9rL"
Figaro.env.stripe_api_key? # => true
Figaro.env.google_analytics_key # => nil
Figaro.env.google_analytics_key? # => false

Required Keys

If a particular configuration value is required but not set, it's appropriate to raise an error. With Figaro, you can either raise these errors proactively or lazily.

To proactively require configuration keys:

# config/initializers/figaro.rb

Figaro.require_keys("pusher_app_id", "pusher_key", "pusher_secret")

If any of the configuration keys above are not set, your application will raise an error during initialization. This method is preferred because it prevents runtime errors in a production application due to improper configuration.

To require configuration keys lazily, reference the variables via "bang" methods on Figaro.env:

# config/initializers/pusher.rb

Pusher.app_id = Figaro.env.pusher_app_id!
Pusher.key    = Figaro.env.pusher_key!
Pusher.secret = Figaro.env.pusher_secret!

Deployment

Figaro is written with deployment in mind. In fact, Heroku's use of ENV for application configuration was the original inspiration for Figaro.

Heroku

Heroku already makes setting application configuration easy:

$ heroku config:set google_analytics_key=UA-35722661-5

Using the figaro command, you can set values from your configuration file all at once:

$ figaro heroku:set -e production

For more information:

$ figaro help heroku:set

Other Hosts

If you're not deploying to Heroku, you have two options:

• Generate a remote configuration file
• Set ENV variables directly

Generating a remote configuration file is preferred because of:

• familiarity – Management of config/application.yml is like that of config/database.yml.
• isolation – Multiple applications on the same server will not produce configuration key collisions.

Is Figaro like dotenv?

Yes. Kind of.

Figaro and dotenv were written around the same time to solve similar problems.

Similarities

• Both libraries are useful for Ruby application configuration.
• Both are popular and well maintained.
• Both are inspired by Twelve-Factor App's concept of proper configuration.
• Both store configuration values in ENV.

Differences

• Configuration File
  • Figaro expects a single file.
  • Dotenv supports separate files for each environment.
• Configuration File Format
  • Figaro expects YAML containing key/value pairs.
  • Dotenv convention is a collection of KEY=VALUE pairs.
• Security vs. Convenience
  • Figaro convention is to never commit configuration files.
  • Dotenv encourages committing configuration files containing development values.
• Framework Focus
  • Figaro was written with a focus on Rails development and conventions.
  • Dotenv was written to accommodate any type of Ruby application.

Either library may suit your configuration needs. It often boils down to personal preference.

Is application.yml like secrets.yml?

Yes. Kind of.

Rails 4.1 introduced the secrets.yml convention for Rails application configuration. Figaro predated the Rails 4.1 release by two years.

Similarities

• Both are useful for Rails application configuration.
• Both are popular and well maintained.
• Both expect a single YAML file.

Differences

• Configuration Access
  • Figaro stores configuration values in ENV.
  • Rails stores configuration values in Rails.application.secrets.
• Configuration File Structure
  • Figaro expects YAML containing key/value string pairs.
  • Secrets may contain nested structures with rich objects.
• Security vs. Convenience
  • Figaro convention is to never commit configuration files.
  • Secrets are committed by default.
• Consistency
  • Figaro uses ENV for configuration in every environment.
  • Secrets encourage using ENV for production only.
• Approach
  • Figaro is inspired by Twelve-Factor App's concept of proper configuration.
  • Secrets are… not.

The emergence of a configuration convention for Rails is an important step, but as long as the last three differences above exist, Figaro will continue to be developed as a more secure, more consistent, and more standards-compliant alternative to secrets.yml.

For more information, read the original The Marriage of Figaro… and Rails blog post.

How do I upgrade to Figaro 1.0?

In most cases, upgrading from Figaro 0.7 to 1.0 is painless. The format expectations for application.yml are the same in 1.0 and values from application.yml are loaded into ENV as they were in 0.7.

However, there are breaking changes:

Figaro.env

In Figaro 0.7, calling a method on the Figaro.env proxy would raise an error if a corresponding key were not set in ENV.

In Figaro 1.0, calling a method on Figaro.env corresponding to an unset key will return nil. To emulate the behavior of Figaro 0.7, use "bang" methods as described in the Required Keys section.

NOTE: In Figaro 0.7, Figaro.env inherited from Hash but in Figaro 1.0, hash access has been removed.

Heroku Configuration

In Figaro 0.7, a Rake task existed to set remote Heroku configuration according to values in application.yml.

In Figaro 1.0, the Rake task was replaced by a command for the figaro executable:

$ figaro heroku:set -e production

For more information:

$ figaro help heroku:set

NOTE: The environment option is required for the heroku:set command. The Rake task in Figaro 0.7 used the default of "development" if unspecified.

Spring Configuration

If you're using Spring, either stop or add config/application.yml to the watch list:

# config/spring.rb

%w(
  ...
  config/application.yml
).each { |path| Spring.watch(path) }

Who wrote Figaro?

My name is Steve Richert and I wrote Figaro in March, 2012 with overwhelming encouragement from my employer, Collective Idea. Figaro has improved very much since then, thanks entirely to inspiration and contribution from developers everywhere.

Thank you!

How can I help?

Figaro is open source and contributions from the community are encouraged! No contribution is too small.

See Figaro's contribution guidelines for more information.