Aruba is Cucumber extension for testing command line applications. Features at a glance:
- Test any command line application, implemented in any language
- Manipulate the file system
- Automatically reset state of file system between scenarios
If you have a Gemfile
, add aruba
. Otherwise, install it like this:
gem install aruba
Then, require
the library in one of your ruby files under features/support
(e.g. env.rb
)
require 'aruba/cucumber'
You now have a bunch of step definitions that you can use in your features. Look at lib/aruba/cucumber.rb
to see them all. Look at features/*.feature
for examples (which are also testing Aruba
itself).
Originally written for cucumber
, aruba
can be helpful in other contexts as
well. One might want to use it together with rspec
.
- Create a directory named
spec/support
- Create a file named
spec/support/aruba.rb
with:
require 'aruba/rspec'
- Add the following to your
spec/spec_helper.rb
Dir.glob(::File.expand_path('../support/*.rb', __FILE__)).each { |f| require_relative f }
- Add a type to your specs
RSpec.describe 'My feature', type: :aruba do
# [...]
end
aruba
provides a wonderful API to be used in your tests:
- Creating files/directories
- Deleting files/directories
- Checking file size
- Checking file existence/absence
- ...
A full documentation of the API can be found here.
Aruba has some default behaviour that you can change if you need to.
Per default Aruba will create a directory tmp/aruba
where it performs its file operations.
If you want to change this behaviour put this into your features/support/env.rb
:
Before do
@dirs = ["somewhere/else"]
end
Aruba will automatically add the bin
directory of your project to the PATH
environment variable for
the duration of each Cucumber scenario. So if you're developing a Ruby gem with a binary command, you
can test those commands as though the gem were already installed.
If you need other directories to be added to the PATH
, you can put the following in features/support/env.rb
:
ENV['PATH'] = "/my/special/bin/path#{File::PATH_SEPARATOR}#{ENV['PATH']}"
A process sometimes takes longer than expected to terminate, and Aruba will kill them off (and fail your scenario) if it is still alive after 3 seconds. If you need more time you can modify the timeout by assigning a different value to @aruba_timeout_seconds
in a Before
block:
Before do
@aruba_timeout_seconds = 5
end
Running processes interactively can result in race conditions when Aruba executes an IO-related step
but the interactive process has not yet flushed or read some content. To help prevent this Aruba waits
before reading or writing to the process if it is still running. You can control the wait by setting
@aruba_io_wait_seconds
to an appropriate value. This is particularly useful with tags:
Before('@slow_process') do
@aruba_io_wait_seconds = 5
end
Aruba defines several tags that you can put on on individual scenarios, or on a feature.
To get more information on what Aruba is doing, use these tags:
@announce-cmd
- See what command is run@announce-stdout
- See the stdout@announce-stderr
- See the stderr@announce-dir
- See the current directory@announce-env
- See environment variables set by Aruba@announce
- Does all of the above
You can hook into Aruba's lifecycle just before it runs a command:
Aruba.configure do |config|
config.before_cmd do |cmd|
puts "About to run '#{cmd}'"
end
end
Aruba clobbers all files in its working directory before each scenario. -Unless you tag it with @no-clobber
Aruba strips away ANSI escapes from the stdout and stderr of spawned child processes by default. It's usually rather cumbersome to
make assertions about coloured output. Still, there might be cases where you want to leave the ANSI escapes intact. Just tag your
scenario with @ansi
. Alternatively you can add your own Before
hook that sets @aruba_keep_ansi = true
.
If your CLI program is written in Ruby you can speed up your suite of scenarios by running
your CLI in the same process as Cucumber/Aruba itself. In order to be able to do this, the
entry point for your CLI application must be a class that has a constructor with a particular
signature and an execute!
method:
class MyMain
def initialize(argv, stdin=STDIN, stdout=STDOUT, stderr=STDERR, kernel=Kernel)
@argv, @stdin, @stdout, @stderr, @kernel = argv, stdin, stdout, stderr, kernel
end
def execute!
# your code here, assign a value to exitstatus
@kernel.exit(exitstatus)
end
end
Your bin/something
executable would look something like the following:
require 'my_main'
MyMain.new(ARGV.dup).execute!
Then wire it all up in your features/support/env.rb
file:
require 'aruba'
require 'aruba/in_process'
Aruba::InProcess.main_class = MyMain
Aruba.process = Aruba::InProcess
That's it! Everything will now run inside the same ruby process, making your suite a lot faster. Cucumber itself uses this approach to test itself, so check out the Cucumber source code for an example.
Pros:
- Very fast compared to spawning processes
- You can use libraries like
simplecov more easily, because
there is only one "process" which adds data to
simplecov
's database
Cons:
- You might oversee some bugs: You might forget to require libraries in your "production" code, because you have required them in your testing code
Improve startup time by disabling JIT and forcing client JVM mode. This can be accomplished by adding
require 'aruba/jruby'
or setting a hook like this example:
Aruba.configure do |config|
config.before_cmd do |cmd|
set_env('JRUBY_OPTS', "-X-C #{ENV['JRUBY_OPTS']}") # disable JIT since these processes are so short lived
set_env('JAVA_OPTS', "-d32 #{ENV['JAVA_OPTS']}") # force jRuby to use client JVM for faster startup times
end
end if RUBY_PLATFORM == 'java'
Note - no conflict resolution on the JAVA/JRuby environment options is done; only merging. For more complex settings please manually set the environment variables in the hook or externally.
A larger process timeout for java may be needed
Before do
@aruba_timeout_seconds = RUBY_PLATFORM == 'java' ? 60 : 10
end
Refer to http://blog.headius.com/2010/03/jruby-startup-time-tips.html for other tips on startup time.
Sometimes your tests need existing files to work - e.g binary data files you
cannot create programmatically. Since aruba
>= 0.6.3 includes some basic
support for fixtures. All you need to do is the following:
- Create a
fixtures
-directory - Create fixture files in this directory
The expand_path
-helper will expand %
to the path of your fixtures
directory:
expand_path('%/song.mp3')
# => /home/user/projects/my_project/fixtures/song.mp3
Example
- Create fixtures directory
cd project
mkdir -p fixtures/
# or
mkdir -p test/fixtures/
# or
mkdir -p spec/fixtures/
# or
mkdir -p features/fixtures/
- Store
song.mp3
infixtures
-directory
cp song.mp3 fixtures/
-
Add fixture to vcs-repository - e.g.
git
,mercurial
-
Create test
RSpec.describe 'My Feature' do
describe '#read_music_file' do
context 'when the file exists' do
let(:path) { expand_path('%/song.mp3') }
before :each do
in_current_directory { FileUtils.cp path, 'file.mp3' }
end
before :each do
run 'my_command'
end
it { expect(all_stdout).to include('Rate is 128 KB') }
end
end
end
Important - you need Pygments installed to use this feature.
Aruba can generate a HTML page for each scenario that contains:
- The title of the scenario
- The description from the scenario (You can use Markdown here)
- The command(s) that were run
- The output from those commands (in colour if the output uses ANSI escapes)
- The files that were created (syntax highlighted in in colour)
In addition to this, it creates an index.html
file with links to all individual report files.
Reporting is off by default, but you can enable it by defining the ARUBA_REPORT_DIR
environment variable, giving it the value
where reports should be written:
ARUBA_REPORT_DIR=doc cucumber features
This will use Aruba's built-in template by default (See the templates
folder). If you want to use your own template you can override its location:
ARUBA_REPORT_TEMPLATES=templates ARUBA_REPORT_DIR=doc cucumber features
The templates directory must contain a main.erb
and files.erb
template. It can also contain other assets such
as css, javascript and images. All of these files will be copied over to the report dir well.
There are some edge cases where Gherkin and Markdown don't agree. Bullet lists using *
is one example. The *
is also an alias for
step keywords in Gherkin. Markdown headers (the kind starting with a #
) is another example. They are parsed as comments by Gherkin. To use either of these, just escape them with a backslash. So instead of writing:
Scenario: Make tea
## Making tea
* Get a pot
* And some hot water
Given...
You'd write:
Scenario: Make tea
\## Making tea
\* Get a pot
\* And some hot water
Given...
This way Gherkin won't recognize these lines as special tokens, and the reporter will render them as Markdown. (The reporter strips away any leading the backslashes before handing it off to the Markdown parser).
Another option is to use alternative Markdown syntax and omit conflicts and escaping altogether:
Scenario: Make tea
Making tea
----------
- Get a pot
- And some hot water
Given...
Please see the CONTRIBUTING.md
.
Copyright (c) 2010,2011,2012,2013,2014 Aslak Hellesøy, David Chelimsky and Mike Sassak. See LICENSE for details.