Abstract Feature Branch 1.6.0

Gem Version Build Status Coverage Status Code Climate

abstract_feature_branch is a Ruby gem that provides a unique variation on the Branch by Abstraction Pattern by Paul Hammant and the Feature Toggles Pattern by Martin Fowler to enhance team productivity and improve software fault tolerance.

It provides the ability to wrap blocks of code with an abstract feature branch name, and then specify in a configuration file which features to be switched on or off.

The goal is to build out upcoming features in the same source code repository branch (i.e. Continuous Integration and Trunk-Based Development), regardless of whether all are completed by the next release date or not, thus increasing team productivity by preventing integration delays. Developers then disable in-progress features until they are ready to be switched on in production, yet enable them locally and in staging environments for in-progress testing.

This gives developers the added benefit of being able to switch a feature off after release should big problems arise for a high risk feature.

abstract_feature_branch additionally supports Domain Driven Design's pattern of Bounded Contexts by allowing developers to configure context-specific feature files if needed.

abstract_feature_branch is one of the simplest and most minimalistic "Feature Flags" Ruby gems out there as it enables you to get started very quickly by simply leveraging YAML files without having to set up a data store if you do not need it (albeit, you also have the option to use Redis as a very fast in-memory data store).

Requirements

  • Ruby (between ~> 3.3.0 and ~> 1.9.1)
  • [Optional] Rails (between ~> 7.0 and ~> 2.0)
  • [Optional] Redis Server (between ~> 7.0 and ~> 2.0)
  • [Optional] Redis client gem (between ~> 5.0 and ~> 3.0)

Setup

Rails Application Use

  1. Configure Rubygem
    • With rails between ~> 7.0 and ~> 2.0: Add the following to Gemfile
      gem 'abstract_feature_branch', '~> 1.6.0'
    • With rails ~> 2.0 only: Add the following to config/environment.rb
      config.gem 'abstract_feature_branch', :version => '1.6.0'
  2. Generate config/initializers/abstract_feature_branch.rb, lib/tasks/abstract_feature_branch.rake, config/features.yml and config/features.local.yml in your Rails app directory by running
    rails g abstract_feature_branch:install
  3. [Optional] Generate config/features/[context_path].yml in your Rails app directory by running
    rails g abstract_feature_branch:context context_path
    (more details under instructions)
  4. [Optional] Customize configuration in config/initializers/abstract_feature_branch.rb (can be useful for changing location of feature files in Rails application, configuring Redis with a Redis or ConnectionPool instance to use for overrides, and scoped feature enablement (e.g. per-user), or troubleshooting a specific Rails environment feature configuration)
  5. [Optional] Redis Server (between ~> 7.0 and ~> 3.0): On the Mac, you can install simply via Homebrew with brew install redis
  6. [Optional] redis client gem (between ~> 5.0 and ~> 3.0): Add the following to Gemfile above abstract_feature_branch
    gem 'redis', '~> 5.0.5'

Ruby Application General Use

  1. gem install abstract_feature_branch -v 1.6.0
  2. Add code require 'abstract_feature_branch'
  3. Create config/features.yml under AbstractFeatureBranch.application_root and fill it with content similar to that of the sample config/features.yml mentioned under instructions.
  4. [Optional] Create config/features.local.yml under AbstractFeatureBranch.application_root (more details under instructions)
  5. [Optional] Create config/features/[context_path].yml under AbstractFeatureBranch.application_root (more details under instructions)
  6. [Optional] Add code AbstractFeatureBranch.application_root = "[your_application_path]" to configure the location of feature files (it defaults to '.')
  7. [Optional] Add code AbstractFeatureBranch.application_environment = "[your_application_environment]" (it defaults to 'development'). Alternatively, you can set ENV['APP_ENV'] before the require statement or an an external environment variable.
  8. [Optional] Add code AbstractFeatureBranch.logger = "[your_application_logger]" (it defaults to a new instance of Ruby Logger. Must use a logger with info and warn methods).
  9. [Optional] Add code AbstractFeatureBranch.cacheable = {[environment] => [true/false]} to indicate cacheability of loaded feature files for enhanced performance (it defaults to true for every environment other than development).
  10. [Optional] Add code AbstractFeatureBranch.load_application_features to pre-load application features for improved first-use performance
  11. [Optional] Add code AbstractFeatureBranch.feature_store = Redis.new(options) to configure Redis for overrides and/or scoped feature enablement (e.g. per-user)
  12. [Optional] Set AbstractFeatureBranch.feature_store_live_fetching = true to enable live fetching of features from store (e.g. Redis) to avoid need for app/server restart upon feature changes, with the trade-off of slightly more latency due to making calls to feature store over the network (this affects scoped feature enablement too, pre-caching all user IDs for features on app/server restart when disabled)

Instructions

config/features.yml contains the main configuration for the application features.

config/features.local.yml contains local overrides for the configuration, ignored by git, thus useful for temporary local feature switching for development/testing/troubleshooting purposes.

Optional context specific config/features/[context_path].yml contain feature configuration for specific application contexts. For example: admin, public, or even internal/wiki. Useful for better organization especially once config/features.yml grows too big (e.g. 20+ features)

Optional context specific config/features/[context_path].local.yml contain local overrides for context-specific feature configuration. These files are rarely necessary as any feature (even a context feature) can be overridden in config/features.local.yml, so these additional *.local.yml files are only recommended to be utilized once config/features.local.yml grows too big (e.g. 20+ features).

Here are the contents of the generated sample config/features.yml, which you can modify with your own features, each enabled (true) or disabled (false) per environment (e.g. production).

defaults: &defaults
  feature1: true
  feature2: true
  feature3: false
  feature4: scoped

development:
  <<: *defaults

test:
  <<: *defaults

staging:
  <<: *defaults
  feature2: false

production:
  <<: *defaults
  feature1: false
  feature2: false

Notice in the sample file how the feature "feature1" was configured as true (enabled) by default, but overridden as false (disabled) in production. This is a recommended practice.

  • Declaratively feature branch logic to only run when feature1 is enabled:

multi-line logic:

feature_branch :feature1 do
  # perform logic
end

single-line logic:

feature_branch(:feature1) { # perform logic }

Note that feature_branch returns nil and does not execute the block if the feature is disabled or non-existent.

feature_branch supports multi-threaded code (i.e. usage from multiple parallel threads, as possible in JRuby).

  • Imperatively check if a feature is enabled or not:
if feature_enabled?(:feature1)
  # perform logic
else
  # perform alternate logic
end

Note that feature_enabled? returns false if the feature is disabled and nil if the feature is non-existent (practically the same effect, but nil can sometimes be useful to detect if a feature is referenced).

feature_enabled? supports multi-threaded code (i.e. usage from multiple parallel threads, as possible in JRuby).

  • List all configured features for a particular environment:
AbstractFeatureBranch.environment_features('development')
# => {"feature1"=>true, "feature2"=>false, "feature3"=>false, "feature4"=>true}

Scoped Feature Enablement

It is possible to restrict enablement of features per specific users (or per entities of any kind) by setting a feature value to scoped, and then toggling features for specific users (or other entities).

  1. Use toggle_features_for_scope in Ruby code to enable features per scope ID (e.g. entity ID, comma-separated compound ID, JSON string, or value object), which must be a String or a value that is safely-convertable to a String like Integer (e.g. email address or database ID). This loads Redis client gem into memory and stores scoped feature configuration in Redis. In the example below, current_user is a method that provides the current signed in user (e.g. using Rails Devise library).
scope = current_user.email
AbstractFeatureBranch.toggle_features_for_scope(scope, :feature1 => true, :feature2 => false, :feature3 => true, :feature5 => true)

Use alternate version of feature_branch and feature_enabled? passing extra scope argument

Examples:

feature_branch :feature1, current_user.email do
  # THIS WILL EXECUTE
end

if feature_enabled?(:feature2, current_user.email)
  # THIS ONE WILL NOT EXECUTE
else
  # THIS ONE WILL EXECUTE
end

feature_branch :feature1, another_user.email do
  # THIS WILL NOT EXECUTE
end

if feature_enabled?(:feature2, another_user.email)
  # THIS ONE WILL EXECUTE (assuming feature2 is enabled in features.yml)
else
  # THIS ONE WILL NOT EXECUTE
end

Note:

If a feature is enabled as true or disabled as false in features.yml (or one of the overrides like features.local.yml or environment variable overrides), then it overrides toggled scoped feature restrictions, becoming enabled or disabled globally.

API

AbstractFeatureBranch.toggle_features_for_scope(scope, feature1: true, feature2: false, ...): API method that toggles features (Strings or Symbols) for a scope

AbstractFeatureBranch.toggled_features_for_scope(scope): API method that returns toggled features for a scope (String)

AbstractFeatureBranch.scopes_for_feature(feature): API method that returns scopes for a (scoped) feature (String or Symbol)

Recommendations

  • Wrap routes in routes.rb with feature blocks to disable entire MVC feature elements by simply switching off the URL route to them. Example:
feature_branch :add_business_project do
  resources :projects
end
  • Wrap visual links to these routes in ERB views. Example:
<% feature_branch :add_business_project do %>
  <h2>Submit a Business</h2>
  <p>
    Please submit a business idea for review.
  </p>
  <ul>
    <% current_user.projects.each do |p| %>
    <li><%= link_to p.business_campaign_name, project_path(p) %></li>
    <% end %>
  </ul>
  <h4>
    <%= link_to('Start', new_project_path, :id => "business_background_invitation", :class => 'button') %>
  </h4>
<% end %>
  • In Rails 4, wrap newly added strong parameters in controllers for data security. Example:
params.require(:project).permit(
  feature_branch(:project_gallery) {:exclude_display},
  :name,
  :description,
  :website
)
  • In Rails 4 and 3.1+ with the asset pipeline, wrap newly added CSS or JavaScript using .erb format (gotcha and alternative solution). Example (renamed projects.css.scss to projects.css.scss.erb and wrapped CSS with an abstract feature branch block):
<% feature_branch :project_gallery do %>
.exclude_display {
  margin-left: auto;
  margin-right: auto;
  label {
    font-size: 1em;
    text-align: center;
  }
  height: 47px;
}
<% end %>
label {
  font-size: 1.5em;
  margin-bottom: -15px;
  margin-top: 3px;
  display: inline;
}
  • Once a feature has been released and switched on in production, and it has worked well for a while (e.g. for two consecutive releases), it is strongly recommended that its feature branching code is plucked out of the codebase to simplify and improve future maintainability given that it is no longer needed at that point.

  • Once config/features.yml grows too big (e.g. 20+ features), it is strongly recommended to split it into multiple context-specific feature files by utilizing the context generator mentioned above:

    rails g abstract_feature_branch:context context_path

  • When working on a new feature locally that the developer does not want others on the team to see yet, the feature can be enabled in config/features.local.yml only as it is git ignored while the feature is disabled in config/features.yml

  • When troubleshooting a deployed feature by simulating a non-development environment (e.g. staging or production) locally, the developer can disable it temporarily in config/features.local.yml (git ignored) under the non-development environment, perform tests on the feature, and then remove the local configuration once done.

Environment Variable Overrides

You can override feature configuration with environment variables by setting an environment variable with a name matching this convention (case-insensitive): ABSTRACT_FEATURE_BRANCH_[feature_name] and giving it the case-insensitive value "TRUE" or "FALSE"

Example:

export ABSTRACT_FEATURE_BRANCH_FEATURE1=TRUE
rails s

The first command adds an environment variable override for feature1 that enables it regardless of any feature configuration, and the second command starts the rails server with feature1 enabled.

To remove an environment variable override, you may run:

unset ABSTRACT_FEATURE_BRANCH_FEATURE1
rails s

The benefits can be achieved more easily via config/features.local.yml mentioned above. However, environment variable overrides are implemented to support overriding feature configuration for a Heroku deployed application more easily.

Redis Overrides

Prerequisites: Redis server and client (redis gem) and Redis configuration of AbstractFeatureBranch.feature_store in config/initializers/abstract_feature_branch.rb (Redis ConnectionPool instance is recommended for Production environments)

To be able to override feature configuration in a production environment, you can utilize Redis Overrides.

Alternatively, you may use Redis Overrides as your main source of feature configuration if you prefer that instead of relying on YAML files.

Keep in mind that by default, Redis Overrides are fetched on app/server start to pre-cache for better performance.

To enable live fetching of Redis Overrides, set AbstractFeatureBranch#feature_store_live_fetching to true (e.g. in config/initializers/abstract_feature_branch.rb), but keep in mind the trade-off with more latency due to making calls to Redis Server over the network.

You can override feature configuration with Redis hash values by calling AbstractFeatureBranch#set_store_feature in rails console (or irb after requiring redis and abstract_feature_branch):

AbstractFeatureBranch.set_store_feature('feature1', true)

Behind the scenes, that is the equivalent of the following Redis client invocation, which stores a hash value in a 'abstract_feature_branch' key:

AbstractFeatureBranch.configuration.feature_store.hset('abstract_feature_branch', 'feature1', 'true')

To remove a Redis override, you can run the following in rails console (or irb):

AbstractFeatureBranch.delete_store_feature('feature1')

To get a Redis override value, you can run the following in rails console (or irb):

AbstractFeatureBranch.get_store_feature('feature1')

To get an array of all Redis Override features, you can run the following in rails console (or irb):

AbstractFeatureBranch.get_store_features

Heroku

Environment variable overrides can be extremely helpful on Heroku as they allow developers to enable/disable features at runtime without a redeploy.

Examples

Enabling a new feature without a redeploy:

heroku config:add ABSTRACT_FEATURE_BRANCH_FEATURE3=true -a heroku_application_name

Disabling a buggy recently deployed feature without a redeploy:

heroku config:add ABSTRACT_FEATURE_BRANCH_FEATURE2=false -a heroku_application_name

Removing an environment variable override:

heroku config:remove ABSTRACT_FEATURE_BRANCH_FEATURE2 -a heroku_application_name

Recommendation

It is recommended that you use environment variable overrides on Heroku only as an emergency or temporary measure. Afterward, make the change official in config/features.yml, deploy, and remove the environment variable override for the long term.

Gotcha with abstract feature branching in CSS and JS files

If you've used abstract feature branching in CSS or JS files via ERB, setting environment variable overrides won't affect them as you need asset recompilation in addition to it, which can only be triggered by changing a CSS or JS file and redeploying on Heroku (even if it's just a minor change to force it). In any case, environment variable overrides have been recommended above as an emergency or temporary measure. If there is a need to rely on environment variable overrides to alter the style or JavaScript behavior of a page back and forth without a redeploy, one solution is to do additional abstract feature branching in HTML templates (e.g. ERB or HAML to link to different stylesheets and JS files, use different CSS classes, or invoke different JavaScript methods per branch of HTML for example.)

Feature Configuration Load Order

For better knowledge and clarity, here is the order in which feature configuration is loaded, with the latter sources overriding the former if overlap in features occurs:

  1. Context-specific feature files: config/features/*/.yml
  2. Main feature file: config/features.yml
  3. Context-specific local feature file overrides: config/features/*/.local.yml
  4. Main local feature file override: config/features.local.yml
  5. Environment variable overrides
  6. Redis overrides

Rails Initializer

Here is the content of the generated initializer with redis client gem added, which contains instructions on how to customize via dependency injection:

# Storage system for features (other than YAML/Env-Vars). Right now, only Redis and ConnectionPool are supported.
# AbstractFeatureBranch.feature_store = Redis.new

# Storage can be a Redis ConnectionPool instance
# AbstractFeatureBranch.feature_store = ConnectionPool.new { Redis.new }

# The following example line works with Heroku Redis To Go while still operating on local Redis for local development
# AbstractFeatureBranch.feature_store = Redis.new(:url => ENV['REDISTOGO_URL'])

# Enable live fetching of feature configuration from storage system, to update features without app/server restart.
# false by default to only load features on app/server start for faster performance (requires restart on change)
AbstractFeatureBranch.feature_store_live_fetching = false

# Application root where config/features.yml or config/features/ is found
AbstractFeatureBranch.application_root = Rails.root

# Application environment (e.g. "development", "staging" or "production")
AbstractFeatureBranch.application_environment = Rails.env.to_s

# Abstract Feature Branch logger
AbstractFeatureBranch.logger = Rails.logger

# Cache feature files once read or re-read them at runtime on every use (helps development).
# Defaults to true if environment not specified, except for development, which defaults to false.
AbstractFeatureBranch.cacheable = {
  :development => false,
  :test => true,
  :staging => true,
  :production => true
}

# Pre-load application features to improve performance of first web-page hit
AbstractFeatureBranch.load_application_features unless Rails.env.development?

Rake Task

Abstract Feature Branch comes with a rake task to beautify feature files that have grown unorganized by sorting features by name and getting rid of extra empty lines. It does so per section, without affecting the order of the sections themselves.

For example, here is content before and after beautification.

Before:

defaults: &defaults


  gallery: true

  carousel: true

  third_party_integration: false
  caching: true

development:
  <<: *defaults

test:
  <<: *defaults

  caching: false

staging:
  <<: *defaults
  third_party_integration: true
  caching: true

production:
  <<: *defaults
  third_party_integration: false

  caching: false

After:

defaults: &defaults
  caching: true
  carousel: true
  gallery: true
  third_party_integration: false

development:
  <<: *defaults

test:
  <<: *defaults
  caching: false

staging:
  <<: *defaults
  caching: true
  third_party_integration: true

production:
  <<: *defaults
  caching: false
  third_party_integration: false

This is very useful in bigger applications that have scores of features since it allows a developer to quickly scan for alphabetical sorted feature names. Although file find is an alternative solution, having tidy organized feature names can help increase overall team productivity in the long term.

For Rails application use, the rake task is generated under lib/tasks/abstract_feature_branch.rake.

For Ruby application general use, here is the content of the rake task:

require 'abstract_feature_branch'
namespace :abstract_feature_branch do

  desc "Beautify YAML of specified feature file via file_path argument or all feature files when no argument specified (config/features.yml, config/features.local.yml, and config/features/**/*.yml) by sorting features by name and eliminating extra empty lines"
  task :beautify_files, :file_path do |_, args|
    AbstractFeatureBranch::FileBeautifier.process(args[:file_path])
  end

end

The rake task may be invoked in a number of ways:

  • rake abstract_feature_branch:beautify_files beautifies all feature files under [application_root]/config
  • rake abstract_feature_branch:beautify_files[file_path] beautifies a single feature file
  • rake abstract_feature_branch:beautify_files[directory_path] beautifies all feature files under directory path recursively

Note that the beautifier ignores comments at the top, but deletes entire line comments in the middle of a YAML file, so after invoking the rake task, verify that your feature file contents are to your satisfaction before committing the task changes.

Feature Branches vs Branch by Abstraction

Although feature branches and branching by abstraction are similar, there are different situations that recommend each approach.

Feature branching leverages your version control software (VCS) to create a branch that is independent of your main branch. Once you write your feature, you integrate it with the rest of your code base. Feature branching is ideal for developing features that can be completed within the one or two iterations. But it can become cumbersome with larger features due to the fact your code is isolated and quickly falls out of sync with your main branch. You will have to regularly rebase with your main branch or devote substantial time to resolving merge conflicts.

Branching by abstraction, on the other hand, is ideal for substantial features, i.e. ones which take many iterations to complete. This approach to branching takes place outside of your VCS. Instead, you build your feature, but wrap the code inside configurable flags. These configuration flags will allow for different behavior, depending on the runtime environment. For example, a feature would be set to "on" when your app runs in development mode, but "off" when running in "production" mode. This approach avoids the pain of constantly rebasing or resolving a myriad of merge conflicts when you do attempt to integrate your feature into the larger app.

Contributing to abstract_feature_branch

  • Check out the latest master to make sure the feature hasn't been implemented or the bug hasn't been fixed yet.
  • Check out the issue tracker to make sure someone already hasn't requested it and/or contributed it.
  • Fork the project.
  • Start a feature/bugfix branch.
  • Commit and push until you are happy with your contribution.
  • Make sure to add tests for it. This is important so I don't break it in a future version unintentionally.
  • Please try not to mess with the Rakefile, version, or history. If you want to have your own version, or is otherwise necessary, that is fine, but please isolate to its own commit so I can cherry-pick around it.

Committers

Contributors

Copyright (c) 2012-2023 Andy Maleh. See LICENSE.txt for further details.