# Refactoring Legacy Code: Part 5 - Game's Testable Methods

This post is part of a series called Refactoring Legacy Code.
Refactoring Legacy Code: Part 4 - Our First Unit Tests
Refactoring Legacy Code: Part 6 - Attacking Complex Methods

Old code. Ugly code. Complicated code. Spaghetti code. Gibberish nonsense. In two words, Legacy Code. This is a series that will help you work and deal with it.

In our previous tutorial, we tested our Runner functions. In this lesson, it is time to continue where we left off by testing our Game class. Now, when you start with such a big chunk of code like we have here, it is tempting to start to test in a top-down manner, method by method. This is, most of the time, impossible. It is much better to start testing it by its short, testable methods. This is what we'll do in this lesson: find and test those methods.

## Creating a Game

In order to test a class we need to initialize an object of that specific type. We can consider that our first test is to create such a new object. You will be surprised how many secrets constructors can hide.

To our surprise, Game can actually be created quite easily. No problems while running just new Game(). Nothing breaks. This is a very good start, especially considering that Game's constructor is quite large and it does a lot of things.

## Finding the First Testable Method

It is tempting to simplify the constructor right now. But we have only the golden master to make sure we do not break anything. Before we go to the constructor, we need to test most of the rest of the class. So, where should we start?

Look for the first method that returns a value and ask yourself, "Can I call and control the return value of this method?". If the answer is yes, it is a good candidate for our test.

What about this method? It seems like a good candidate. Only two lines and it returns a boolean value. But wait, it calls another method, howManyPlayers().

This is basically just a method that counts the elements in the class' players array. OK, so if we do not add any players, it should be zero. isPlayable() should return false. Let's see if our assumption is correct.

We renamed our previous test method to reflect what we really want to test. Then we just asserted the game is not playable. The test passes. But false positives are common in many cases. So for piece of mind, we can assert true and make sure the test fails.

And it does!

So far, pretty promising. We managed to test the method's initial return value, the value represented by the initial state of the Game class. Please note the emphasized word: "state". We need to find a way to control the state of the game. We need to change it, so it will have the minimum number of players.

If we analyze Game's add() method, we will see that it adds elements to our array.

Our assumption is enforced by the way the add() method is used in RunnerFunctions.php.

Based on these observations, we can conclude that by using add() twice, we should be able to bring our Game into a state with two players.

By adding this second test method, we can ensure isPlayable() returns true, if conditions are met.

Much better. All irrelevant code is in private methods, $game is initialized in setUp() and a lot of pollution was removed from the test methods. However, we did have to make a compromise here. In our first test, we start with an assertion. This presumes that setUp() will always create an empty game. This is OK for now. But at the end of the day, you must realize there is no such thing as perfect code. There is just code with compromises that you are willing to live with. ## The Second Testable Method If we are scanning our Game class from the top towards the bottom, the next method on our list is add(). Yes, the same method we used in our tests in the previous paragraph. But can we test it? Now this is a different way of testing objects. We call our method and then we verify the object's state. As add() always returns true, there is no way we can test its output. But we can start with an empty Game object and then check if there is a single user after we add one. But is that enough verification? Wouldn't it be better to also verify if there are no players before we call add()? Well, it may be a little bit too much here, but as you can see in the code above, we could do it. And whenever you are not sure of the initial state, you should do an assertion on it. This also protects you from future code changes that may change your object's initial state. But are we testing all of the things the add() method does? I say no. Besides adding a user, it also sets a lot of settings for it. We should also check for those. This is better. We verify each action that the add() method does. This time, I preferred to directly test the $players array. Why? We could have used the howManyPlayers() method which basically does the same thing, right? Well, in this case we considered that it is more important to describe our assertions by the effects that the add() method has on the state of the object. If we need to change add(), we would expect that the test which is testing its strict behavior, will fail. I've had endless debates with my colleagues at Syneto about this. Especially because this type of testing introduces a strong coupling between the test and how the add() method is actually implemented. So, if you prefer to test it the other way around, that does not mean your ideas are wrong.

We can safely ignore the testing of the output, the echoln() lines. They are just outputting content on the screen. We don't want to touch these methods, yet. Our golden master totally relies on this output.

### Refactoring Tests (Bis)

We have another tested method with a brand new passing test. It's time to refactor both of them, just a little bit. Let's start with our tests. Aren't the last three assertions a little bit confusing? They don't seem to be related strictly to adding a player. Let's change it:

That's better. The method is now more abstract, reusable, expressively named and hides all of the unimportant details.

### Refactoring the add() Method

We can do something similar with our production code.

We extracted the unimportant details into setDefaultPlayerParametersFor().

Actually this idea came to me after I wrote the test. This is another nice example of how tests force us to think about our code from a different point of view. This different angle on the problem, is what we must exploit and let our tests guide our design of the production code.

## The Third Testable Method

Let's find our third candidate for testing. howManyPlayers() is too simple and indirectly already tested. roll() is too complex to be tested directly. Plus it returns null. askQuestions() seems to be interesting at first sight, but it is all presentation, no return value.

currentCategory() is testable, but it is pretty difficult to test. It is a huge selector with ten conditions. We need a ten lines long test and then we need to seriously refactor this method and most certainly the tests as well. We should take note of this method and come back to it after we are finished with the easier ones. For us, this will be in our next tutorial.

wasCorrectlyAnswered() is to complicated again. We will need to extract from it, small pieces of code that are testable. However, wrongAnswer() seems promising. It outputs stuff on the screen, but it also changes the state of our object. Let's see if we can control it and test it.

Grrr... It was quite hard to write this test method. wrongAnswer() relies on $this->currentPlayer for its behavioral logic, but it also uses $this->players in its presentation part. One ugly example of why you should not mix logic and presentation. We will deal with this in a future tutorial. For now, we tested that the user enters the penalty box. We must also observe that there is an if() statement in the method. This is a condition that we do not yet test, as we only have a single player and thus we are not satisfying the condition. We could test for the final value of \$currentPlayer though. But adding this line of code to the test will make it fail.

A closer look at the private method shouldResetCurrentPlayer() reveals the problem. If the index of the current player equals with the number of players, it will be reset to zero. Aaaahhh! We actually enter the if()!

Good. We created a second test, to test the specific case when there are still players who did not play. We don't care about the inPenaltyBox state for the second test. We are only interested in the index of the current player.

## The Final Testable Method

The last method we can test and then refactor is didPlayerWin().

We can immediately observe that its code structure is very similar to isPlayable(), the method we tested first. Our solution should be something similar as well. When your code is so short, only two to three lines, doing more than one tiny step is not that big of a risk. In worst case scenarios, you revert three lines of code. So let's do this in a single step.

But wait! That fails. How is that possible? Shouldn't it pass? We provided the correct number of coins. If we study our method, we'll discover a little misleading fact.

The return value is actually negated. So the method is not telling us if a player won, it tells us if a player did not win the game. We could go in and find the places where this method is used and negate its value there. Then change its behavior here, to not falsely negate the answer. But it is used in wasCorrectlyAnswered(), a method we can not yet unit test. Maybe for the time being, a simple renaming to highlight the correct functionality will be enough.

## Thoughts & Conclusion

So this about wraps up the tutorial. While we do not like the negation in the name, this is a compromise we can make at this point. This name will surely change when we start refactoring other parts of the code. Additionally, if you take a look at our tests, they look odd now:

By testing false on a negated method, exercised with a value that suggests a true result, we introduced quite a lot of confusion to our codes readability. But this is good for now, as we do need to stop at some point, right?

In our next tutorial, we will start working on some of the more difficult methods within the Game class. Thank you for reading.