Test driven development

Test-Driven Development

Overview

Write tests before implementation. The Red-Green-Refactor cycle — write a failing test (RED), write minimal code to pass (GREEN), improve the code without changing behavior (REFACTOR) — produces code that is correct by construction and has comprehensive test coverage from the start.

When to Use

• Implementing any new feature or behavior
• Fixing any bug (use the Prove-It pattern)
• Adding new business logic to ViewModels, use cases, or repositories
• Before refactoring (establish a test safety net first)

Skip when: Purely cosmetic UI changes with no behavioral logic.

Core Process

Red-Green-Refactor Cycle

RED: Write a Failing Test

1. Write the test first — describe what the code should do:

class TaskListViewModelTest {
    private val repository = mockk<TaskRepository>()
    private lateinit var viewModel: TaskListViewModel

    @Test
    fun `initial state is Loading`() {
        every { repository.getTasks() } returns flowOf(emptyList())

        viewModel = TaskListViewModel(GetTasksUseCase(repository))

        assertEquals(TaskListUiState.Loading, viewModel.uiState.value)
    }

    @Test
    fun `emits Success with tasks when repository returns data`() = runTest {
        val tasks = listOf(Task("1", "Buy groceries", false))
        every { repository.getTasks() } returns flowOf(tasks)

        viewModel = TaskListViewModel(GetTasksUseCase(repository))
        advanceUntilIdle()

        assertEquals(TaskListUiState.Success(tasks), viewModel.uiState.value)
    }

    @Test
    fun `emits Error when repository throws`() = runTest {
        every { repository.getTasks() } returns flow { throw IOException("Network error") }

        viewModel = TaskListViewModel(GetTasksUseCase(repository))
        advanceUntilIdle()

        val state = viewModel.uiState.value
        assertIs<TaskListUiState.Error>(state)
        assertEquals("Network error", state.message)
    }
}

2. Run the test — it must fail (./gradlew test)
   • If it passes immediately, the test doesn't test what you think

GREEN: Write Minimal Code to Pass

3. Implement just enough to make the test pass:

@HiltViewModel
class TaskListViewModel @Inject constructor(
    getTasks: GetTasksUseCase,
) : ViewModel() {

    val uiState: StateFlow<TaskListUiState> = getTasks()
        .map<List<Task>, TaskListUiState> { TaskListUiState.Success(it) }
        .catch { emit(TaskListUiState.Error(it.message ?: "Unknown error")) }
        .stateIn(viewModelScope, SharingStarted.WhileSubscribed(5_000), TaskListUiState.Loading)
}

4. Run the test — it must pass
5. Don't add code "you'll need later" — only what the test demands

REFACTOR: Improve Without Changing Behavior

6. Clean up while tests stay green:
   • Extract common patterns
   • Improve naming
   • Reduce duplication
   • Run tests after every change

The Prove-It Pattern (Bug Fixes)

7. For every bug, prove it exists before fixing:

// Step 1: Write a test that demonstrates the bug
@Test
fun `completed tasks should not appear in active filter`() {
    val tasks = listOf(
        Task("1", "Active task", completed = false),
        Task("2", "Done task", completed = true),
    )
    every { repository.getTasks() } returns flowOf(tasks)

    viewModel = TaskListViewModel(GetTasksUseCase(repository))
    viewModel.setFilter(TaskFilter.ACTIVE)
    advanceUntilIdle()

    val state = viewModel.uiState.value as TaskListUiState.Success
    assertEquals(1, state.tasks.size)
    assertEquals("Active task", state.tasks[0].title)
}

// Step 2: Run it — confirm it FAILS (proves the bug exists)
// Step 3: Fix the code
// Step 4: Run it — confirm it PASSES (proves the fix works)
// Step 5: Run full suite — no regressions

Test Pyramid

    /‾‾‾‾‾‾‾‾‾\
   / UI Tests   \        ~5%  — Compose rules, Espresso, Maestro flows
  / (device)     \             see: android-device-testing, android-e2e-verification
 /‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾\
/ Integration Tests \     ~15% — Room in-memory DB, MockWebServer,
| (local/device)    |           Robolectric
|‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾|
|    Unit Tests       |   ~80% — JUnit5 + MockK
|  (fast, local)      |          ViewModels, UseCases, Repositories
 ‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾

Testing Patterns by Layer

8. ViewModel tests:

@OptIn(ExperimentalCoroutinesApi::class)
class ViewModelTest {
    @get:Rule
    val mainDispatcherRule = MainDispatcherRule() // Sets Dispatchers.Main to TestDispatcher

    @Test
    fun `action updates state correctly`() = runTest {
        // Arrange
        val viewModel = createViewModel()

        // Act
        viewModel.onEvent(TaskListEvent.ToggleTask("1"))
        advanceUntilIdle()

        // Assert
        val state = viewModel.uiState.value
        // ...assertions...
    }
}

9. Repository tests (integration):

class TaskRepositoryTest {
    private val dao = mockk<TaskDao>()
    private val api = mockk<TaskApi>()
    private val mapper = TaskMapper()
    private val repository = TaskRepositoryImpl(dao, api, mapper)

    @Test
    fun `syncTasks fetches from API and saves to database`() = runTest {
        val apiTasks = listOf(TaskResponse("1", "Test"))
        coEvery { api.getTasks() } returns apiTasks
        coEvery { dao.upsertAll(any()) } just Runs

        repository.syncTasks()

        coVerify { dao.upsertAll(match { it.size == 1 && it[0].id == "1" }) }
    }
}

10. Room DAO tests:

@RunWith(AndroidJUnit4::class)
class TaskDaoTest {
    private lateinit var db: AppDatabase
    private lateinit var dao: TaskDao

    @Before
    fun setup() {
        db = Room.inMemoryDatabaseBuilder(
            ApplicationProvider.getApplicationContext(),
            AppDatabase::class.java,
        ).build()
        dao = db.taskDao()
    }

    @After
    fun teardown() { db.close() }

    @Test
    fun upsert_and_observe_returns_task() = runTest {
        val task = TaskEntity("1", "Test", null, false)
        dao.upsert(task)

        val result = dao.observeAll().first()
        assertEquals(1, result.size)
        assertEquals("Test", result[0].title)
    }
}

Best Practices

• State over interactions — assert on outcomes, not method calls
• DAMP over DRY — each test should be self-contained and readable
• Prefer real > fakes > stubs > mocks — mock only at system boundaries
• Descriptive names — test names read like behavior specifications
• Use runTest — for coroutine testing with TestDispatcher
• Use advanceUntilIdle() — not delay() or Thread.sleep()

Common Rationalizations

Shortcut	Why It Fails
"I'll write tests after the code works"	You won't. And tests written after are weaker — they verify what you built, not what you intended.
"TDD is slow"	TDD prevents debugging. Net time is less.
"This is too simple to test"	Simple code becomes complex code. Tests catch the transition.
"Mocking is too complex for this"	If it's hard to mock, the design has too many dependencies. TDD surfaces this.
"The test passed on the first try, so it's good"	A test that never failed may not test what you think. Make it fail first.

Red Flags

• Code written before tests
• Tests that pass immediately (never saw RED)
• Thread.sleep or hardcoded delays in tests
• Tests that depend on execution order
• Mocking internal classes (mock boundaries only)
• No tests for error/edge cases
• @Ignore or @Disabled tests without issue references

Verification

• Tests written BEFORE implementation (Red-Green-Refactor)
• Every test seen failing before passing
• Bug fixes use the Prove-It pattern
• ./gradlew test passes (all unit tests)
• Test names describe behavior, not implementation
• No Thread.sleep in tests (use advanceUntilIdle)
• Test pyramid proportions reasonable (~80/15/5)
• Edge cases and error paths covered