Posts Tagged ‘test-driven development’

Double-Blind Test-Driven Development in Rails 3: Part 3

February 2, 2011

  1. Simple Tests
  2. Double-Blind Tests
  3. Making it Practical with RSpec Matchers

This is the last article in this series describing the concept of double-blind test-driven development. This style of testing can add time to development, but this can be cut significantly using RSpec matchers.

If you’re not familiar with matchers, they’re the helpers that give RSpec its english-like syntax, and they can be a powerful tool speeding up all of your test-driven development – whether you follow the double-blind method or not.

If you’re using RSpec, you’re already using their built-in matchers. Say we have a Site model, and its url method takes the host attribute and appends the ‘http://’ protocol. Here’s a likely test:

describe Site, 'url'
  it "should begin with http://" do
    site = Site.new :host => 'example.com'
    site.url.should equal('http://example.com')
  end
end

The equal() method in the code above is the matcher. You can pass it to any of RSpec’s should or should_not methods, and it will magically work.

But the magic isn’t that hard, and you can harness it yourself for custom matchers that conform to your application.

The Many Faces of Custom RSpec Matchers

While I don’t want this article to turn into a primer on custom RSpec matchers (it’s a little off-topic), I’ll give you the three styles of defining them, and explain my recommendations. There are simple matchers, the Matcher DSL, and full RSpec matcher classes.

Let’s start by writing a test we want to run:

it "should be at least 5" do
  6.should be_at_least(5)
end

This test should always pass, provided we’ve defined our matcher correctly. The first way to do this is the simple matcher:

def be_at_least(minimum)
  simple_matcher("at least #{minimum}"){|actual| actual >= minimum}
end

As you might guess, actual represents the object that “.should” whatever – in this case “.should be_at_least(5)”. This version makes a lot of assumptions, including the auto-creation of generic pass and fail messages.

If you want a little more control, you can step up to RSpec’s Matcher DSL. This is the middle-of-the-road option for creating custom matchers:

RSpec::Matchers.define :be_at_least do |minimum|
  match do |actual|
    actual >= minimum
  end

  failure_message_for_should do |actual|
    "expected #{actual} to be at least #{minimum}"
  end

  failure_message_for_should_not do |actual|
    "expected #{actual} to be less than #{minimum}"
  end

  description do
    "be at least #{minimum}"
  end
end

Now we’re rocking custom failure messages, and test names. This is pretty cool, and honestly how I started out doing matchers. It’s also how I started out doing the matchers for double-blind testing.

The problem is that by skipping the creation of actual matcher classes, we lose the ability to do things like inheritance. Not a big deal if our matchers stay simple, but they won’t. Not if we use them as often as we should! I found myself re-defining the same helper methods in each matcher I defined this way.

So let’s see just how daunting a full-fledged custom matcher class really is:

module CustomMatcher  
  class BeAtLeast
    def initialize(minimum)  
      @minimum = minimum
    end  
  
    def matches?(actual)  
      @actual = actual
      @actual >= @minimum
    end  
  
    def failure_message_for_should  
      "expected #{@actual} to be at least #{@minimum}"  
    end  
  
    def failure_message_for_should_not  
      "expected #{@actual} to be less than #{@minimum}"  
    end  
  end  
  
  def be_at_least(expected)  
    BeAtLeast.new(expected)  
  end  
end  

This isn’t so bad! We’re defining a new class, but you can see it doesn’t have to inherit from anything, or use any unholy Ruby voodoo to work.

We just have to define four methods: initialize, match? (which returns true or false), and the two failure message methods. Along the way, we set some instance variables so we can access the data when we need it. Finally, we define a method that creates a new instance of this class, and that’s what RSpec will rely on.

You can add as many other methods as these four will rely on. But you also get other benefits over the DSL. You can use inheritance, moving common methods up the chain so you only have to define them once, instead of in each matcher definition. You can also write setup/teardown code in your parent classes, make default arguments a breeze, and standardize any error handling. I do all of these in the matchers I created for the example app.

The bottom line is this: defining your own matcher classes directly really DRY’s up your matchers, and that always makes life simpler. I think it’s the only way to go for serious and heavy RSpec users. It allows the class for my validate_presence_of matcher to be this short and sweet:

module DoubleBlindMatchers
  class ValidatePresenceOf < ValidationMatcher
    def default_options
      {:message => "can't be blank", :with => 'x'}
    end

    def match
      set_to @options[:with]
      @object.valid?
      check !@object.errors[@attribute].include?(@options[:message]), shouldnt_exist
      
      set_to nil
      check !@object.valid?, valid_when('nil')
      check @object.errors[@attribute].include?(@options[:message])
      
      set_to ""
      check !@object.valid?, valid_when("blank")
      check @object.errors[@attribute].include?(@options[:message])
    end
  end
  
  def validate_presence_of expected, options = {}
    ValidatePresenceOf.new expected, options
  end
end

And the Teacher model, which grew considerably during our double-blind testing, now looks like this (in its entirety):

# spec/models/teacher_spec.rb

require 'spec_helper'

describe Teacher do
  it {should have_many :subjects}
  
  it {should validate_presence_of :name}
  it {should validate_length_of :name, :maximum => 50, :message => "must be 50 characters or less"}
  
  it {should validate_presence_of :salary}
  it {should validate_numericality_of :salary, :within => (20_000..100_000), :message => "must be between $20K and $100K"}
end

Summary

Now that you’ve seen my entire proposal for double-blind testing, let me know what you think. Be cruel if you must, it’s the only way I’ll learn. I’ll do the best to explain (not defend) my reasoning, and keep an open mind to changes.

I’ll also be publishing my double-blind matchers as a gem so you can add them to your project.

Double-Blind Test-Driven Development in Rails 3: Part 2

February 1, 2011

  1. Simple Tests
  2. Double-Blind Tests
  3. Making it Practical with RSpec Matchers

The last article in this series defined the concept of double-blind test-driven development, but didn’t get much into real-world examples. In this article, we’ll explore several such examples.

The Example Application

This article includes a sample app that you can download using the link above. Be sure to checkout tag “double_blind_tests” to see the code as it appears in this article. The next article will have a lot of refactoring. I limited my samples to the model layer, where 100% coverage is a very realistic goal, and this is likely to be the greatest benefit.

I chose a simple high school scheduling app with teachers, the subjects they teach, students, and courses. In this case, I’m defining a course as a student’s participation in a subject. Teachers teach (ie, have) many subjects. Students take (have) many subjects, via courses. The course record contains that student’s grade for the given subject.

The database constraints are intentionally strict, and most of the validations in the models ensure that these constraints are respected in the application layer. We don’t want the user seeing an error page because of bad data. Depending on the application, that can be worse than actually having bad data creep in.

Associations

Here’s an example of a has_many association:

# excerpt from spec/models/teacher_spec.rb

describe Teacher do
  it "has many subjects" do
    teacher = Factory.create :teacher
    teacher.subjects.should be_empty

    subject = teacher.subjects.create Factory.attributes_for(:subject)
    teacher.subjects.should include(subject)
  end
end

In order to factor out our own assumptions, we have to ask what they are. The assumption is that the subject we add to the teacher’s subject list works because of the has_many relationship. So we’ll first test that teacher.subjects is, in fact, empty when we assume it would be. Then we’re free to test that adding a subject works as we expect.

Here’s a belongs_to association:

# excerpt from spec/models/subject_spec.rb

describe Subject do
  it "belongs_to a teacher" do
    teacher = Factory.create :teacher

    subject = Subject.new
    subject.teacher.should be_nil
    
    subject.teacher = teacher
    subject.teacher.should == teacher
  end
end

Again, we’re challenging the assumption that the association is nil by default, by testing against it before verifying that we can add a teacher. This tests that this is a true belongs_to association, and not simply an instance method. This is the kind of thing that can and will change over the life of an application.

Validations

Let’s test validates_presence_of:

# excerpt from spec/models/teacher_spec.rb

describe Teacher do
  describe "name" do
    it "is present" do
      error_message = "can't be blank"
      
      teacher = Teacher.new :name => 'Joe Example'
      teacher.valid?
      teacher.errors[:name].should_not include(error_message)

      teacher.name = nil
      teacher.should_not be_valid
      teacher.errors[:name].should include(error_message)

      teacher.name = ''
      teacher.should_not be_valid
      teacher.errors[:name].should include(error_message)
    end
  end
end

This example was actually explained in detail in the last article. Validate that the error doesn’t already exist before trying to trigger it. Don’t just test the default value when you create a blank object, test the likely possibilities. Refactor the error message to DRY up the test and add readability. And finally, test by modifying the object you already created (as little as possible) rather than creating a new object from scratch for each part of the test.

A more complex version is needed to validate the presence of an association:

# excerpt from spec/models/subject_spec.rb

describe Subject do
  describe "teacher" do
    it "is present" do
      error_message = "can't be blank"

      teacher = Factory.create(:teacher)
      subject = Factory.create(:subject, :teacher => teacher)
      subject.valid?
      subject.errors[:teacher].should_not include(error_message)
    
      subject.teacher = nil
      subject.should_not be_valid
      subject.errors[:teacher].should include(error_message)
    end
  end
end

While the test is more complex, the code to satisfy it is not:

# excerpt from app/models/subject.rb

validates_presence_of :teacher

testing validates_length_of:

# excerpt from spec/models/teacher_spec.rb

describe Teacher do
  describe "name" do
    it "is at most 50 characters" do
      error_message = "must be 50 characters or less"
      
      teacher = Teacher.new :name => 'x' * 50
      teacher.valid?
      teacher.errors[:name].should_not include(error_message)
      
      teacher.name += 'x'
      teacher.should_not be_valid
      teacher.errors[:name].should include(error_message)
    end
  end
end

And here’s the model code that satisfies the test:

# excerpt from app/models/teacher.rb

validates_length_of :name, :maximum => 50, :message => "must be 50 characters or less"

While you can definitely start to see a pattern in validation testing, this introduces a new element. Instead of freshly setting the name attribute to be 51 characters long, we test the valid edge case first and then add *just* enough to make it invalid – one more character.

This does two things: it verifies that our edge case was as “edgy” as it could be, and it makes our test less brittle. If we wanted to change the test to allow up to 100 characters, we’d only have to modify the test name and the initial set value.

validating a number’s range using validates_numericality_of:

# excerpt from spec/models/teacher_spec.rb

describe Teacher do
  describe "salary" do
    it "is at or above $20K" do
      error_message = "must be between $20K and $100K"
      
      teacher = Teacher.new :salary => 20_000
      teacher.valid?
      teacher.errors[:salary].should_not include(error_message)

      teacher.salary -= 0.01
      teacher.should_not be_valid
      teacher.errors[:salary].should include(error_message)
    end

    it "is no more than $100K" do
      error_message = "must be between $20K and $100K"

      teacher = Teacher.new :salary => 100_000
      teacher.valid?
      teacher.errors[:salary].should_not include(error_message)
      
      teacher.salary += 0.01
      teacher.should_not be_valid
      teacher.errors[:salary].should include(error_message)
    end
  end
end

And here’s the code that satisfies the test:

# excerpt from app/models/teacher.rb

validates_numericality_of :salary, :message => "must be between $20K and $100K",
  :greater_than_or_equal_to => 20_000, :less_than_or_equal_to => 100_000

We’re doing the same here as in our testing of name’s length. We’re setting the edge value that’s *just* within the allowed range, then adding or subtracting a penny to make it invalid. I split up the top and bottom edge tests, because it’s better to test as atomically as possible – one limit per test.

Defaults

Another tricky database constraint to test for is a default value:

# excerpt from spec/models/course_spec.rb

describe Course do
  describe "grade_percentage" do
    it "defaults to 1.0" do
      course = Course.new :grade_percentage => nil
      course.grade_percentage.should be_nil
      
      course = Course.new :grade_percentage => ''
      course.grade_percentage.should be_blank
      
      course = Course.new :grade_percentage => 0.95
      course.grade_percentage.should == 0.95
      
      course = Course.new
      course.grade_percentage.should == 1.0
    end
  end
end

In this case, we can’t avoid having to recreate the model from scratch, because the nature of the implementation. There’s no actual code in the model that makes this happen, it’s purely in the database schema. Why should we test it, then? Because we test any behavior we’re going to rely on in the application. The fact that this model behavior is implemented at the database level (and therefore, not purely TDD) is a small inconvenience.

What’s the assumption our double-blind test is verifying in this case? That the value is only set in the absence of other values being explicitly assigned. Testing with nil and blank values verifies that the default doesn’t override them – it only works in the complete absence of any assignment. I also test an arbitrary (but valid) value as the anti-assumption test before finally verifying that the default is setting to the correct value.

Most default tests verify only that the correct default value is set – the double-blind version verifies that it’s acting only as a default value in all cases.

Summary

The point of double-blind testing is bullet-proof tests, that can’t be reasonably thwarted by antagonistic coding – whether that’s your anti-social pairing partner, or yourself several months down the road. The bottom line is this: test all assumptions.

That being said, this is very time consuming, and we can see a ton of repetition even in this small test suite. What we need is a way to get back to speedy testing before our boss/client notices it now takes an hour to implement one validation.*

*Even if you work for a government owned/regulated institution that actually digs that kind of non-agile perversion, you WILL eventually go insane. Even in this small sample app, the voices in my head had to talk me off a building ledge twice.

The answer lies in RSpec matchers, which are easy to implement, and can grow with your application. The benefit is not just speedier development – it’s also consistency across your application. We’ll explore that in the last article of this series.

Double-Blind Test-Driven Development in Rails 3: Part 1

January 31, 2011

This is a three-part series introducing the concept of double-blind test-driven development in Rails. This post defines the concept itself, and lays the groundwork by showing the way tests are more commonly written. The next couple posts will show how to double-blind test various common rails elements, and how to make this added layer of protection automatic and quick.

  1. Simple Tests
  2. Double-Blind Tests
  3. Making it Practical with RSpec Matchers

Looking at a rails application that was built with test-driven development, you might expect to see something like this:

# spec/models/teacher_spec.rb

describe Teacher do
  it "has many subjects" do
    teacher = Factory.create :teacher
    subject = teacher.subjects.create Factory.attributes_for(:subject)

    teacher.subjects.should include(subject)
  end
  
  describe "name" do
    it "is present" do
      teacher = Teacher.new

      teacher.should_not be_valid
      teacher.errors[:name].should include("can't be blank")
    end
    
    it "is at most 50 characters" do
      teacher = Teacher.new :name => 'x' * 51
      
      teacher.should_not be_valid
      teacher.errors[:name].should include("must be 50 characters or less")
    end
  end
end

Truth be told, if you’re seeing this in the wild the app is probably doing pretty good. This level of testing works great during the early stages of an app, when things are simple. But as things grow and/or multiple developers become involved, you need more.

Consider models where the associations and validations stretch into the dozens of lines. The more careful and specific you are about validations, the easier it is to get conflicting or overlapping validations. I actually came up with the concept of double-blind testing while retro-testing models in a client app that previously had no validation specs.

What is Double-Blind Testing?

In the world of scientific studies, you always need a control group. One set of participants gets the latest and greatest new diet pill, while the other gets a placebo. Researchers used to think this was good enough, and probably pretty funny to watch the placebo users rave about their shrinking waistlines. But it turns out studies like this still allowed some bias – as researchers observed the effects, their *own* preconceived notions tainted results. Enter the double-blind study.

In a double-blind study, the researchers themselves are unaware of which participants are in the control group, and which are being tested. Both sides are “blind”. They may have lost funny patient anecdotes, but they gained research reliability.

Applying the Lessons of Double-Blind Studies to Test-Driven Development

As I said, in the early stages of an app the tests I showed above work great, as long as you’re using TDD and the red-green-refactor cycle. This means you write the test, run it, and it fails. Then you write the simplest code that will make the test pass, run the test again, and confirm that it passes. Most testing tools will literally show red or green as you do this. Then, as you start to amass tests, you’re free to refactor your code (abstracting common code into helper methods, changing for readability, etc) and run the tests again at any time. You will see failures if you broke anything. If not, you’ve more or less guaranteed your code refactoring works properly.

The problem comes in when you start changing old code, or adding tests to processes that didn’t initially happen. What I’m calling double-blind testing is this:

each test needs to verify the object’s behavior before testing what changes.

As an example, let’s rewrite one of the tests from above:

# original test

describe "name" do
  it "is present" do
    teacher = Teacher.new

    teacher.should_not be_valid
    teacher.errors[:name].should include("can't be blank")
  end
end
# modified to be double-blind

describe "name" do
  it "is present" do
    error_message = "can't be blank"

    teacher = Teacher.new :name => 'Joe Example'
    teacher.valid?
    teacher.errors[:name].should_not include(error_message)

    teacher.name = nil
    teacher.should_not be_valid
    teacher.errors[:name].should include(error_message)

    teacher.name = ""
    teacher.should_not be_valid
    teacher.errors[:name].should include(error_message)
  end
end

This is the basic pattern for all double-blind testing. We’re not leaving anything to chance. In the original version, we expected our object to be invalid, we treated it as such, and we got the result we expected. Do you see the problem with this?

Here’s an exercise: can you make the original test pass, even though the object validation is not working correctly? There’s actually a style of pair programming that routinely does exactly this. One developer writes the test, and the other writes just enough code to make it pass, with the good-natured intention of tripping up the first developer whenever possible. If you wrote the original test, I could satisfy it by just adding the error message to every record on validation, regardless of whether it’s true! Your test would pass, but the app would fail.

The test is now “double-blind” in the sense that we as testers have factored out our own expectations from the test. In this case, we expect the error message to not be there until we initialize the object a certain way, and this can be bad. It may sound far-fetched or paranoid*, but in large codebases your original tests are often abused in this very way. The “you” that writes new code today is often at odds with the “you” from three months ago that wrote the older code with a different understanding of the problem at hand.

*Plus, everybody knows it’s not paranoia when the world really is out to get you. I’ve discussed this at length with the voices in my head, and they all agree. Except Javier. That guy’s a jerk.

Now that I’ve laid out the justification, let’s take a closer look at how the test changed. The first thing I did was create a version of the object that I believe should NOT trigger the error message. Then I run through two cases that should. You can see right away, I was forced to be more *specific* about what should trigger an error. Instead of just a blank object with no values set, I’ve proactively set the attribute in question to both nil and blank. A key element here is to try to work with the *same* object, modifying between tests, rather than creating a new object each time. My test wouldn’t have been as specific if I’d just recreated a blank Teacher object and run a single validation check.

Also, with the increased code comes the increased chance of typos. We don’t want to DRY test code up too much, because a good rule is to keep your tests are readable (non-abstract) as possible. But I’ve specified the error message at the top of the test, and reused that string over and over. I did this in a way that DRY’s the code and adds readability. You can see at a glance that all three tests are checking for the same error.

Finally, the first time I run the object’s validation, notice I’m not asserting that it should be valid. If I had written teacher.should be_valid on line 8 of the double-blind test, I’d have to take the extra time to make sure every other part of the object was valid. Not only is this time-consuming, it’s very brittle. Any future validations would break this test.

If you use factories often, you may suggest setting it up that way since a factory-generated object should always be valid. Then you could assert validity. However, this only slows down your test suite. it’s enough just to run valid? on the object, which triggers all the validation checks to load up our errors hash.

Summary

I believe this is a new concept – I was already coding most of my tests this way, but it didn’t dawn on me how valuable it was until I started retro-testing previously testless code. The value showed itself right away.

I would love to hear feedback on this – if you think it’s unnecessary (I tend to be very rainman-ish about my testing code) or even detrimental. However, if you think it’s too much work, I ask you to hold your criticism until you’ve read part 3 of this article, where I show how to use your own RSpec matchers to greatly speed this process.

Lazy Loading Ruby Methods the Smart Way

September 3, 2010

I just spent an hour trying to solve a seemingly simple problem. Luckily, test-driven development is the reason I spotted this problem at all! Here’s where I started:

  class Survey < ActiveRecord::Base
    SUBMISSION_STATII = ['submitting', 'submitted']

    def in_submission?
      SUBMISSION_STATII.include? self.status
    end
  end

I want the in_submission? method to cache its response in an instance variable, so it doesn’t have to run the check every time. I plan on this method getting called a lot. I wrote my code to test, using shoulda and mocha:

  should "cache the result" do
    Survey.SUMISSION_STATII.expects(:include?).with('new').returns(false).once
    survey = Survey.new :status => 'new'

    survey.in_submission?
    survey.in_submission?
  end

This is simply verifying that the status is checked for inclusion only once, even if we call the method multiple times. And here’s the code that I thought should satisfy the test:

  def in_submission?
    @in_submission ||= SUBMISSION_STATII.include?(self.status)
  end

I only added the @in_submission ||= part. This is my standard trick, but in this case it failed. Why? because whenever @in_submission is set to false, it will trigger the full code to run again next time. No “caching” happens. We should check whether the instance variable is defined, not whether it equates to true:

  def in_submission?
    return @in_submission if defined?(@in_submission)
    @in_submission = SUBMISSION_STATII.include?(self.status)
  end

Sure, we lose our sweet one-liner, but we gain an hour of productivity when our tests pass the first try. Or even worse, you didn’t bother to test caching at all, so it’s just slowing down your app and possibly even mucking things up. But you are using TDD, aren’t you?