Build Status Gem Version Dependency Status Code Climate Coverage Status


Using ENV in Ruby is like using raw SQL statements - it feels wrong, because it is.

If you agree, this gem is for you.

The benefits over using ENV directly:

  • much friendlier stubbing in tests
  • you no longer have to care whether false is "0" or "false" or whatever
  • keys become methods
  • namespaces which can be passed around as objects
  • you can subclass!
  • you can marshal/unmarshal your own types automatically!
  • strict mode saves you from doing validation yourself
  • and there's more to come...

Other benefits (and compared to other solutions):

  • should still work with Ruby 1.8 (in case anyone is still stuck with it)
  • it's designed to be as lightweight and as fast as possible compared to ENV
  • designed to be both hackable and convenient


Add this line to your application's Gemfile:

gem 'nenv'

And then execute:

$ bundle

Or install it yourself as:

$ gem install nenv

Examples !!!

Automatic booleans

You no longer have to care whether the value is "0" or "false" or "no" or "FALSE" or ... whatever

# Without Nenv
t.verbose = (ENV['CI'] == 'true')
ENV['DEBUG'] = "true"

now becomes:

t.verbose =
gemdeps = Nenv.rubygems_gemdeps? || Nenv.bundle_gemfile?
Nenv.debug = true


# Without Nenv

now becomes:

git = Nenv :git
puts git.browser
puts git.pager
puts git.editor

Custom type handling

# Code without Nenv
paths = [ENV['GEM_HOME`]] + ENV['GEM_PATH'].split(':')
enable_logging if Integer(ENV['WEB_CONCURRENCY']) > 1
mydata = YAML.load(ENV['MY_DATA'])
ENV['VERBOSE'] = debug ? "1" : nil

can become:

# setup
gem = Nenv :gem
gem.instance.create_method(:path) { |p| p.split(':') }

web = Nenv :web
web.instance.create_method(:concurrency) { |c| Integer(c) }

my = Nenv :my
my.instance.create_method(:data) { |d| YAML.load(d) }

Nenv.instance.create_method(:verbose=) { |v| v ? 1 : nil }

# and then you can simply do:

paths = [gem.home] + gem.path
enable_logging if web.concurrency > 1
mydata =
Nenv.verbose = debug

Automatic conversion to string

ENV['RUBYGEMS_GEMDEPS'] = 1  # TypeError: no implicit conversion of Fixnum (...)

Nenv automatically uses to_s:

Nenv.rubygems_gemdeps = 1  # no problem here

Custom assignment

data = YAML.load(ENV['MY_DATA'])
data[:foo] = :bar
ENV['MY_DATA'] = YAML.dump(data)

can now become:

my = Nenv :my
my.instance.create_method(:data) { |d| YAML.load(d) }
my.instance.create_method(:data=) { |d| YAML.dump(d) }

data =
data[:foo] = :bar = data

Strict mode

# Without Nenv
fail 'home not allowed' if ENV['HOME'] = Dir.pwd  # BUG! Assignment instead of comparing!
puts ENV['HOME'] # Now contains clobbered value

Now, clobbering can be prevented:

env =

fail 'home not allowed' if env.home = Dir.pwd  # Fails with NoMethodError
puts env.home # works

Mashup mode

You can first define all the load/dump logic globally in one place

Nenv.instance.create_method(:web_concurrency) { |d| Integer(d) }
Nenv.instance.create_method(:path) { |p| Pathname(p.split(File::PATH_SEPARATOR)) }
Nenv.instance.create_method(:path=) { |array| }

# And now, anywhere in your app:

Nenv.web_concurrency += 3
Nenv.path += Pathname.pwd + "foo"

MyEnv = do
end'my').foo? # same as ENV['MY_FOO'][/^(?:false|no|n|0)/i,1].nil?

class MyEnv < Nenv::Environment
  def initialize
end # same as ENV['MY_FOO'][/^(?:false|no|n|0)/i,1].nil?


Still, avoid using environment variables if you can.

At least, avoid actually setting them - especially in multithreaded apps.

As for Nenv, while you can access the same variable with or without namespaces, filters are tied to instances, e.g.:

Nenv.instance.create_method(:foo_bar) { |d| Integer(d) }
Nenv('foo').instance.create_method(:bar) { |d| Float(d) }
env = { |e| e.create_method(:bar) }

all work on the same variable, but each uses a different filter for reading the value.

What's wrong with ENV?

Well sure, having ENV act like a Hash is much better than calling "getenv".

Unfortunately, the advantages of using ENV make no sense:

  • it's faster but ... environment variables are rarely used thousands of times in tight loops
  • it's already an object ... but there's not much you can do with it (try ENV.class)
  • it's globally available ... but you can't isolate it in tests (you need to reset it every time)
  • you can use it to set variables ... but it's named like a const
  • it allows you to use keys regardless of case ... but by convention lowercase shouldn't be used except for local variables (which are only really used by shell scripts)
  • it's supposed to look ugly to discourage use ... but often your app/gem is forced to use them anyway
  • it's a simple class ... but either you encapsulate it in your own classes - or all the value mapping/validation happens everywhere you want the data (yuck!)

But the BIGGEST disadvantage is in specs, e.g.:

allow(ENV).to receive(:[]).with('MY_VARIABLE').and_return("old data")
allow(ENV).to receive(:[]=).with('MY_VARIABLE', "new data")

which could instead be completely isolated as:

let(:env) { instance_double(Nenv::Environment) }
before { allow(Nenv::Environment).to receive(:new).with(:my).and_return(env) }

allow(env).to receive(:variable).and_return("old data")
allow(env).to receive(:variable=).with("new data")


  1. Fork it ([my-github-username]/nenv/fork )
  2. Create your feature branch (git checkout -b my-new-feature)
  3. Commit your changes (git commit -am 'Add some feature')
  4. Push to the branch (git push origin my-new-feature)
  5. Create a new Pull Request