Canonical Repository: https://gitlab.com/envied/envied/tree/master#envied
For the rationale behind this project, see this blogpost.
- check for presence and correctness of ENV-variables
- access to typed ENV-variables (integers, booleans etc. instead of just strings)
- check the presence and correctness of a Heroku config
- provide or load ENV-values
- Quickstart
- Installation
- Configuration
- Command-line interface
- Best Practices
- FAQ
- Testing
- Development
- Contributing
After successful installation, define some variables in Envfile
:
# file: Envfile
variable :FORCE_SSL, :boolean
variable :PORT, :integer
# during initialization
ENVied.require
This will throw an error if:
- one of
ENV['FORCE_SSL']
,ENV['PORT']
is absent. - or: their values cannot be coerced (resp. to boolean and integer).
Variables accessed via ENVied are of the correct type:
ENVied.PORT # => 3001
ENVied.FORCE_SSL # => false
Add this line to your application's Gemfile:
gem 'envied'
...then bundle:
$ bundle
...then for Rails applications:
$ bundle exec envied init:rails
...or for non-Rails applications:
$ bundle exec envied init
The following types are supported:
:array
(e.g. 'tag1,tag2' becomes['tag1', 'tag2']
):boolean
(e.g. '0'/'1', 'f'/'t', 'false'/'true', 'off'/'on', 'no'/'yes' for resp. false and true):date
(e.g. '2014-3-26'):env
(similar to:string
, but accessible via ENV - see Key alias for details):float
:hash
(e.g. 'a=1&b=2' becomes{'a' => '1', 'b' => '2'}
):integer
:string
(implied):symbol
:time
(e.g. '14:00'):uri
(e.g. 'http://www.google.com' becomes result ofURI.parse('http://www.google.com')
)
By default the value for variable FOO
should be provided by ENV['FOO']
. Sometimes though it's convenient to let a different key provide the value, based on some runtime condition. A key-alias will let you do this.
Consider for example local development where REDIS_URL
differs between the development and test environment. Normally you'd prepare different shells with different values for REDIS_URL
: one shell you can run tests in, and other shells where you'd run the console/server etc. This is cumbersome and easy to get wrong.
With a key alias that's calculated at runtime (e.g. Rails.env
) you'd set values for both REDIS_URL_TEST
and REDIS_URL_DEVELOPMENT
and the right value will be used for test and development.
Full example:
# file: Envfile
key_alias! { Rails.env }
variable :REDIS_URL, :uri
Source the following in your environment:
# file: .envrc
export REDIS_URL_DEVELOPMENT=redis://localhost:6379/0
export REDIS_URL_TEST=redis://localhost:6379/1
Now commands like rails console
and rails test
automatically point to the right redis database.
Note that ENV['REDIS_URL']
is still considered but REDIS_URL_<key_alias>
takes precedence.
Also: any truthy value provided as key_alias is converted to an upcased string.
Finally: this setting is optional.
Variables of type :env
take the key alias into account when accessing ENV['FOO']
.
Say, your application uses ENV['DATABASE_URL']
(wich you can't change to ENVied.DATABASE_URL
). Normally this would mean that the key alias has no effect. For env-type variables however, the key alias is taken into account:
# file: Envfile
key_alias! { Rails.env }
variable :DATABASE_URL, :env
The following now works:
$ DATABASE_URL_DEVELOPMENT=postgres://localhost/blog_development rails runner "p ENV['DATABASE_URL']"
"postgres://localhost/blog_development"
Note: this also works for ENV.fetch('FOO')
.
Also: no coercion is done (like you would expect when accessing ENV-values directly).
This means that for Rails applications when you set values for DATABASE_URL_DEVELOPMENT
and DATABASE_URL_TEST
, you no longer need a config/database.yml
.
Groups give you more flexibility to define when variables are needed. It's similar to groups in a Gemfile:
# file: Envfile
variable :FORCE_SSL, :boolean
group :production do
variable :SECRET_KEY_BASE
end
group :development, :staging do
variable :DEV_KEY
end
# For local development you would typically do:
ENVied.require(:default) #=> Only ENV['FORCE_SSL'] is required
# On the production server:
ENVied.require(:default, :production) #=> ...also ENV['SECRET_KEY_BASE'] is required
# You can also pass it a string with the groups separated by comma's:
ENVied.require('default, production')
# This allows for easily requiring groups using the ENV:
ENVied.require(ENV['ENVIED_GROUPS'])
# ...then from the prompt:
$ ENVIED_GROUPS='default,production' bin/rails server
# BTW the following are equivalent:
ENVied.require
ENVied.require(:default)
ENVied.require('default')
ENVied.require(nil)
For help on a specific command, use envied help <command>
.
$ envied help
Commands:
envied check # Checks whether you environment contains required variables
envied check:heroku # Checks whether a Heroku config contains required variables
envied check:heroku:binstub # Generates a shell script for the check:heroku-task
envied extract # Grep code to find ENV-variables
envied help [COMMAND] # Describe available commands or one specific command
envied init # Generates a default Envfile in the current working directory
envied init:rails # Generate all files needed for a Rails project
envied version, --version, -v # Shows version number
Some best practices when using ENVied or working with env-configurable applications in general.
While ENVied will warn you when you start an application that is 'under-configured', it won't tell users what good default values are. To solve this add a file to the root of your project that contains sane defaults and instructions:
# file: .envrc.sample
# copy this file to .envrc and adjust values if needed
# then do `source .envrc` to load
export DATABASE_URL=postgres://localhost/blog_development
# export FORCE_SSL=true # only needed for production
# you can find this token on the Heroku-dashboard
export DEPLOY_TOKEN=1234-ABC-5678
let direnv manage your environment
direnv will auto-(un)load values from .envrc
when you switch folders.
As a bonus it has some powerful commands in it's stdlib.
For example:
# this adds the project's bin-folder to $PATH
PATH_add bin
# so instead of `./bin/rails -h` you can do `rails -h` from anywhere (deep) in the project
# the following will use the .envrc.sample as a basis
# when new variables are introduced upstream, you'll automatically use these defaults
if [ -f .envrc.sample ]; then
source_env .envrc.sample
fi
...your overrides
# a variant of this is source_up
# an .envrc in a subfolder can load the .envrc from the root of the project and override specific values
# this would allow e.g. for a specific test-environment in the subfolder:
# in my-project/test/.envrc
source_up .envrc
export DATABASE_URL=the-test-db-url
$ bundle exec envied extract
This comes in handy when you're not using ENVied yet. It will find all ENV['KEY']
and ENV.fetch('KEY')
statements in your project.
It assumes a standard project layout (see the default value for the globs-option).
The easiest/quickest is to run:
$ heroku config --json | bundle exec envied check:heroku
This is equivalent to having the heroku config as your local environment and running envied check:heroku --groups default production
.
You want to run this right before a deploy to Heroku. This prevents that your app will crash during bootup because ENV-variables are missing from heroku config.
You can turn the above into a handy binstub like so:
$ bundle exec envied check:heroku:binstub
# created bin/heroku-env-check
This way you can do stuff like:
$ ./bin/heroku-env-check && git push live master
The short version: simplicity, i.e. the best tool for the job.
In the early days of ENVied it was possible to provide default values for a variable.
While convenient, it had several drawbacks:
- it would introduce a value for ENVied.FOO, while ENV['FOO'] was nil: confusing and a potential source of bugs.
- it hides the fact that an application can actually be configged via the environment.
- it creates an in-process environment which is hard to inspect (as opposed to doing
printenv FOO
in a shell, after or before starting the application). - there are better ways: e.g. a sample file in a project with a bunch of exports (ie
export FOO=sane-default # and even some documentation
) that someone can source in their shell (see Best Practices). - made the code quite complex.
As an alternative include a file .envrc.sample
in the root of your project containing default values (ie export FOO=bar
) that users can source in their shell. See also Best Practices.
$ ./bin/setup
# run tests
$ ./bin/rspec
# hack with pry
$ ./bin/console
# run CLI:
$ ./bin/envied
There's a .envrc.sample
included that can be used in combination with direnv.
To suggest a new feature, open an Issue before opening a PR.
- Fork it: https://gitlab.com/envied/envied/-/forks/new
- Create your feature branch:
git checkout -b my-new-feature
- Commit your changes:
git commit -am 'Add some feature'
- Push to the branch:
git push origin my-new-feature
- Create a new pull request for your feature branch