Fixtures aren't fun. Machinist is.
Machinist makes it easy to create test data within your tests. It generates data for the fields you don't care about, and constructs any necessary associated objects, leaving you to only specify the fields you do care about in your tests. For example:
describe Comment do
before do
# This will make a Comment, a Post, and a User (the author of
# the Post), and generate values for all their attributes:
@comment = Comment.make(:spam => true)
end
it "should not include comments marked as spam in the without_spam named scope" do
Comment.without_spam.should_not include(@comment)
end
end
You tell Machinist how to do this with blueprints:
require 'machinist/active_record'
require 'sham'
require 'faker'
Sham.name { Faker::Name.name }
Sham.email { Faker::Internet.email }
Sham.title { Faker::Lorem.sentence }
Sham.body { Faker::Lorem.paragraph }
User.blueprint do
name
email
end
Post.blueprint do
title
author
body
end
Comment.blueprint do
post
author_name { Sham.name }
author_email { Sham.email }
body
end
./script/plugin install git://github.com/notahat/machinist.git
sudo gem install machinist --source http://gemcutter.org
Create a blueprints.rb file to hold your blueprints in your test (or spec) directory. It should start with:
require 'machinist/active_record'
require 'sham'
Substitute data_mapper or sequel for active_record if that's your weapon of choice.
Require blueprints.rb in your test_helper.rb (or spec_helper.rb):
require File.expand_path(File.dirname(__FILE__) + "/blueprints")
Set Sham to reset before each test. In the class ActiveSupport::TestCase block in your test_helper.rb, add:
setup { Sham.reset }
or, if you're on RSpec, in the Spec::Runner.configure block in your spec_helper.rb, add:
config.before(:all) { Sham.reset(:before_all) }
config.before(:each) { Sham.reset(:before_each) }
Sham lets you generate random but repeatable unique attributes values.
For example, you could define a way to generate random names as:
Sham.name { (1..10).map { ('a'..'z').to_a.rand }.join }
Then, to generate a name, call:
Sham.name
So why not just define a helper method to do this? Sham ensures two things for you:
- You get the same sequence of values each time your test is run
- You don't get any duplicate values
Sham works very well with the excellent Faker gem by Benjamin Curtis. Using this, a much nicer way to generate names is:
Sham.name { Faker::Name.name }
Sham also supports generating numbered sequences if you prefer.
Sham.name {|index| "Name #{index}" }
If you want to allow duplicate values for a sham, you can pass the :unique option:
Sham.coin_toss(:unique => false) { rand(2) == 0 ? 'heads' : 'tails' }
You can create a bunch of sham definitions in one hit like this:
Sham.define do
title { Faker::Lorem.words(5).join(' ') }
name { Faker::Name.name }
body { Faker::Lorem.paragraphs(3).join("\n\n") }
end
A blueprint describes how to generate an object. The idea is that you let the blueprint take care of making up values for attributes that you don't care about in your test, leaving you to focus on the just the things that you're testing.
A simple blueprint might look like this:
Post.blueprint do
title { Sham.title }
author { Sham.name }
body { Sham.body }
end
You can then construct a Post from this blueprint with:
Post.make
When you call make, Machinist calls Post.new, then runs through the attributes in your blueprint, calling the block for each attribute to generate a value. The Post is then saved and reloaded. An exception is thrown if Post can't be saved.
You can override values defined in the blueprint by passing a hash to make:
Post.make(:title => "A Specific Title")
If you don't supply a block for an attribute in the blueprint, Machinist will look for a Sham definition with the same name as the attribute, so you can shorten the above blueprint to:
Post.blueprint do
title
author { Sham.name }
body
end
If you want to generate an object without saving it to the database, replace make with make_unsaved. (make_unsaved also ensures that any associated objects that need to be generated are not saved - although not if you are using Sequel. See the section on associations below.)
You can refer to already assigned attributes when constructing a new attribute:
Post.blueprint do
title
author { Sham.name }
body { "Post by #{author}" }
end
Named blueprints let you define variations on an object. For example, suppose some of your Users are administrators:
User.blueprint do
name
email
end
User.blueprint(:admin) do
name { Sham.name + " (admin)" }
admin { true }
end
Calling:
User.make(:admin)
will use the :admin blueprint.
Named blueprints call the default blueprint to set any attributes not specifically provided, so in this example the email attribute will still be generated even for an admin user.
You must define a default blueprint for any class that has a named blueprint, even if the default blueprint is empty.
If you're generating an object that belongs to another object, you can generate the associated object like this:
Comment.blueprint do
post { Post.make }
end
Calling Comment.make will construct a Comment and its associated Post, and save both.
If you want to override the value for post when constructing the comment, you can do this:
post = Post.make(:title => "A particular title)
comment = Comment.make(:post => post)
Machinist will not call the blueprint block for the post attribute, so this won't generate two posts.
Machinist is smart enough to look at the association and work out what sort of object it needs to create, so you can shorten the above blueprint to:
Comment.blueprint do
post
end
For has_many and has_and_belongs_to_many associations, ActiveRecord insists that the object be saved before any associated objects can be saved. That means you can't generate the associated objects from within the blueprint.
The simplest solution is to write a test helper:
def make_post_with_comments(attributes = {})
post = Post.make(attributes)
3.times { post.comments.make }
post
end
Note here that you can call make on a has_many association. (This isn't yet supported for DataMapper.)
Make can take a block, into which it passes the constructed object, so the above can be written as:
def make_post_with_comments
Post.make(attributes) do |post|
3.times { post.comments.make }
end
end
The plan method behaves like make, except it returns a hash of attributes, and doesn't save the object. This is useful for passing in to controller tests:
test "should create post" do
assert_difference('Post.count') do
post :create, :post => Post.plan
end
assert_redirected_to post_path(assigns(:post))
end
plan will save any associated objects. In this example, it will create an Author, and it knows that the controller expects an author_id attribute, rather than an author attribute, and makes this translation for you.
You can also call plan on has_many associations, making it easy to test nested controllers:
test "should create comment" do
post = Post.make
assert_difference('Comment.count') do
post :create, :post_id => post.id, :comment => post.comments.plan
end
assert_redirected_to post_comment_path(post, assigns(:comment))
end
(Calling plan on associations is not yet supported in DataMapper.)
Machinist also works with plain old Ruby objects. Let's say you have a class like:
class Post
attr_accessor :title
attr_accessor :body
end
You can then do the following in your blueprints.rb:
require 'machinist/object'
Post.blueprint do
title "A title!"
body "A body!"
end
You can always find the latest version on GitHub.
If you have questions, check out the Google Group.
File bug reports and feature requests in the issue tracker.
Machinist is maintained by Pete Yandell ([email protected], @notahat)
Other contributors include:
Marcos Arias, Jack Dempsey, Clinton Forbes, Perryn Fowler, Niels Ganser, Jeremy Grant, Jon Guymon, James Healy, Evan David Light, Chris Lloyd, Adam Meehan, Kyle Neath, Lawrence Pit, T.J. Sheehy, Roland Swingler, Gareth Townsend, Matt Wastrodowski, Ian White
Thanks to Thoughtbot's Factory Girl. Machinist was written because I loved the idea behind Factory Girl, but I thought the philosophy wasn't quite right, and I hated the syntax.