Keep Critical Values In Your Tests

One of the keys to a well-written unit test is clarity. A great unit test should be short, focused on a single use case, easy to read, and free from noise or complex setup. Here, I will talk briefly about where the critical values should live, inside a test.

A knight in a full suit of armor, exemplifying the scenario in the follow unit test.

Consider the following unit test:

public void ArmorFlat_ApplyEffect()
    AllEffects.Apply(data, performer, new MemberAsTarget(target));

    Assert.AreEqual(6, target.State.Armor());

While it makes the test generally easy to read with the data built outside of the test, there is one quesion that can’t be answered by reading this test.

Why 6 Armor?

What is special about exactly 6? Would 5 be okay? Why can’t it be 4?

With a bit of effort, we find the answer in the surrounding context.

private EffectData data = new EffectData 
  EffectType = EffectType.ArmorFlat, 
  FloatAmount = new FloatReference(1) 

private Member performer = new Member(
    "Good Dummy One",
    new StatAddends()

private Member target = new Member(
    "Good Dummy Two",
    "Wooden Dummy",
    new StatAddends().With(StatType.Armor, 5)

It’s a bit buried in there, but a little bit of reading eventually makes this clear. This test is expecting that if the target already has 5 Armor and there is an effect to change that by +1 Armor, the target should then have 6 Armor.

What can we do to make this test immediately clear?

Let’s refactor the key values back into the test.

public void ArmorFlat_ApplyEffect_ArmorIsChangedCorrectly()
    var addArmorEffect = ChangeArmorBy(1);
    var target = MemberWithStartingArmor(5);
    AllEffects.Apply(addArmorEffect, anyPerformer, new MemberAsTarget(target));
    Assert.AreEqual(6, target.State.Armor());

Now, the test makes it obvious why 6 Armor is the correct amount.

When writing tests, it’s definitely valuable to factor out tangential setup into helper methods and builders. But, it’s essential to leave the critical values that matter to this particular test to remain in the test itself.

Always co-locate the critical values inside the test itself.

Reading the test alone should make the assertion completely clear.