Test Write

Test Writer — Travmaskinen

Write tests that match the project's established conventions exactly. The codebase has 2,032 backend tests and 101 frontend tests — new tests must be indistinguishable from existing ones in style.

Decision: Which Layer?

Coverage commands:

• Backend: pytest backend/tests/ --cov=backend --cov-report=term-missing
• Frontend: pnpm --filter web test:coverage

TDD Flow (Always)

1. Red: Write the test first — it MUST fail
2. Green: Write minimal code to make it pass
3. Refactor: Clean up without changing behavior
4. Run the full suite to confirm no regressions

Backend Tests (pytest)

File Placement

backend/tests/
├── unit/          # Module-level logic, mocked dependencies
├── integration/   # Full endpoint tests via TestClient
└── ml/            # Model invariants, feature parity

Name: test_{module_name}.py — mirrors the source file.

Shared Fixtures (from conftest.py)

Use these — don't recreate them:

• fixture_game (session): Full ATG game from scripts/fixtures/v85_2026-03-07.json
• fixture_race (function): First race from fixture_game
• fixture_horses (function): Horse entries from first race
• mock_intel (function): MagicMock of V85DatabaseIntelligence with sensible zero-data defaults for all 30+ methods
• client (function): FastAPI TestClient with mocked DB and settings

Patterns

Async endpoint/service test

@pytest.mark.asyncio
async def test_cache_miss_calls_predictor():
    """When cache is empty, predictor.predict() is called."""
    import backend.api.predict as mod

    mock_pred = MagicMock()
    mock_pred._loaded = True
    mock_pred.predict = AsyncMock(return_value=[{"horse_id": 1, "prob": 0.5}])

    with patch.object(mod, "predictor", mock_pred), \
         patch.object(mod, "cache_get", new_callable=AsyncMock, return_value=None), \
         patch.object(mod, "cache_set", new_callable=AsyncMock):
        result = await mod.predict_race("2026-03-15_1_1")

    mock_pred.predict.assert_called_once()
    assert result[0]["prob"] == 0.5

Pydantic validation

def test_invalid_race_id_rejected():
    from pydantic import ValidationError
    with pytest.raises(ValidationError):
        PredictRequest(race_id="bad-id")

Integration endpoint

def test_endpoint_returns_200(client):
    resp = client.get("/api/v1/races/today")
    assert resp.status_code == 200
    assert "legs" in resp.json()

ML invariant

def test_probabilities_sum_to_one(fixture_race, mock_intel):
    probs = predictor.predict(fixture_race, mock_intel)
    total = sum(p["win_probability"] for p in probs)
    assert abs(total - 1.0) < 0.01, f"Probabilities sum to {total}, expected ~1.0"

Rules

• asyncio_mode = "auto" — no need for @pytest.mark.asyncio on fixtures, but DO use it on async test functions
• Use AsyncMock for any async return value
• Use patch.object(module, "name") — never patch("string.path") (brittle)
• Test BOTH success AND failure paths (cache hit/miss, model loaded/not, 200/404/500)
• fail_under = 100 in pyproject.toml — every line must be covered or marked # pragma: no cover
• Exclude lines: if TYPE_CHECKING:, if __name__ == "__main__":, raise NotImplementedError

Frontend Tests (Vitest + RTL)

File Placement

Tests live next to source: ComponentName.test.tsx beside ComponentName.tsx.

Test Utilities (from lib/test-utils.tsx)

Use these factories — don't create ad-hoc objects:

• createHorse(overrides?) — full BackendHorse with sensible defaults
• createScratchedHorse(overrides?) — scratched variant
• createFreeSession() / createProSession() / createAdminSession() — auth sessions
• createUnauthenticatedSession() — no auth

Required Mocks

Most components need these at the top of the file:

vi.mock('next/navigation', () => ({
  useRouter: () => ({ push: vi.fn(), back: vi.fn(), forward: vi.fn() }),
  useSearchParams: () => new URLSearchParams(),
  usePathname: () => '/',
}))

vi.mock('next/link', () => ({
  default: ({ children, href, ...props }: any) => <a href={href} {...props}>{children}</a>,
}))

Patterns

Component render + content assertion

it('renders horse name', () => {
  const horse = createHorse({ horse: { name: 'Elitlansen', id: 101 } })
  render(<HorseCard horse={horse} />)
  expect(screen.getByText('Elitlansen')).toBeInTheDocument()
})

Negative assertion (element NOT present)

it('hides badge when scratched', () => {
  const horse = createScratchedHorse()
  render(<HorseCard horse={horse} />)
  expect(screen.queryByText('FAVORIT')).not.toBeInTheDocument()
})

User interaction

it('calls onToggle when clicked', () => {
  const
  render(<HorseCard horse={createHorse()} />)
  fireEvent.click(screen.getByText('Elitlansen').closest('[class*="rounded-lg"]')!)
  expect(onToggle).toHaveBeenCalledTimes(1)
})

API/fetch mocking

beforeEach(() => { vi.useFakeTimers() })
afterEach(() => { vi.useRealTimers(); vi.restoreAllMocks() })

it('fetches data with auth header', async () => {
  vi.stubGlobal('fetch', vi.fn().mockResolvedValue({
    ok: true,
    json: () => Promise.resolve({ data: [] }),
  }))

  await api.get('/api/v1/races', { token: 'jwt-123' })

  expect(fetch).toHaveBeenCalledWith(
    expect.any(String),
    expect.objectContaining({
      headers: expect.objectContaining({ Authorization: 'Bearer jwt-123' }),
    }),
  )
})

Rules

• vi.mock() calls must be at file top (before imports that use the mocked module)
• Always vi.restoreAllMocks() in afterEach
• Use screen.getByText() / screen.getByRole() — never container.querySelector() (fragile)
• Use queryByText() for "should NOT exist" — returns null instead of throwing
• Use expect.objectContaining() for partial object matching
• Import alias: @/lib/... maps to ./lib/...

E2E Tests (Playwright)

File Placement

apps/web/e2e/{feature}.spec.ts

Patterns

import { test, expect } from '@playwright/test'

test.describe('Game page interactions', () => {
  test('loads V85 game with predictions', async ({ page }) => {
    await page.goto('/spel/V85')
    await expect(page.locator('h1')).toBeVisible()
    await expect(page).toHaveURL(/spel\/V85/)
  })

  test('navigates between legs', async ({ page }) => {
    await page.goto('/spel/V85')
    const tab = page.getByRole('tab', { name: 'Avd 2' })
    await tab.click()
    await expect(tab).toHaveAttribute('aria-selected', 'true')
  })
})

Rules

• Tests run on 3 browsers: chromium, firefox, mobile-chrome
• baseURL = http://localhost:3000 — dev server auto-starts locally
• Use page.locator() for CSS, page.getByRole() for semantic elements
• Always await interactions and assertions
• Traces on first retry, screenshots only on failure

Checklist Before Done

1. All new tests pass: run the relevant suite
2. No existing tests broken: run the full suite
3. Coverage check: run with --cov / test:coverage
4. Test names are descriptive: test_cache_miss_returns_fresh_prediction not test_1
5. Both happy path AND error path covered
6. Factories/fixtures used (no inline 50-line objects)