xUnit Test Patterns

Test Goals

• Defect Localization: tells you what specifically went wrong
• Repeatable Test: can be run many times in a row
• Tests as Documentation: should be readable

Test Smells

• Conditional Test Logic: some test code may or may not be executed
  • Production Logic in Test: duplicates the production code, not verifying anything
  • Related: Flexible Test
• Erratic Test: sometimes pass and sometimes fail
  • A cause: Interacting Test: tests depend on other tests in some way. Goal: test isolation.
  • A cause: Nondeterministic Test
• Fragile Test: fails when seemingly-unrelated code is changed
• Hard-to-Test Code: (self-explanatory)
• High Test Maintenance Cost: effort keeping tests passing is more than the benefit
• Obscure Test: difficult to understand at a glance
  • Eager Test: verifies too much in one test
  • Incorrect amount of detail
    • Irrelevant Information: data not necessary to understand the test set up inside the test
    • Mystery Guest: data necessary to understand the test set up outside the test
  • Indirect Testing: checked via interacting with a separate object

Patterns

• Calculated Value: Meszaros considers it a pattern, I generally consider it a smell.
• Fixture
  • Fresh Fixture: setting up a new fixture for each test.
  • General Fixture: larger than needed to verify the functionality. (Could consider this a smell.)
  • Minimal Fixture: smallest and simplest fixture possibile to verify the functionality.
  • Shared Fixture: reusing the same fixture for multiple tests. (Could consider this a smell.)
• Hard-Coded Test Data: solves production logic. Meszaros says it’s a smell, but then says “better to enumerate the sets of precalculated values” to solve production logic.
• Setup
  • In-Line Setup: done in the test function itself
  • Delegated Setup: explicitly call functions to set up the data
    • Parameterized Creation Method: (factories)
  • Implicit Setup: setUp() / beforeEach()
• Single-Condition Tests: verify one condition per test