Testing Patterns

Overview

Decision guide for testing strategies focusing on Vitest for unit/integration and Playwright for E2E tests.

Test Type Decision

Testing Strategy

Strategy is auto-detected based on project type and can be configured per-directory.

Automatic Strategy Detection

Check current project strategy:

# Via detect.sh
source ~/.claude/lib/testing/strategy-detector.sh
detect_testing_strategy "."              # Project-level
detect_directory_strategy "." "src/lib"  # Directory-level

Or check

settings.json

testing.strategy

Per-Directory Strategy Rules

lib/**

packages/**

utils/**

src/components/**

app/**

api/**

server/**

algorithms/**

core/**

Testing Pyramid (Traditional)

     /  E2E  \        Few - slow, brittle, high confidence
    /   Int   \       Some - moderate speed/confidence
   /   Unit    \      Many - fast, isolated, low confidence

Best for: Libraries, utilities, pure logic, microservices

Guidance:

• Test every public function with unit tests
• Cover edge cases extensively (null, empty, boundaries)
• Mock external dependencies
• Aim for >80% unit test coverage

Testing Trophy (Kent C. Dodds)

       E2E            Few - critical paths only
   Integration        MOST - best confidence/speed ratio
      Unit            Some - complex logic only
     Static           TypeScript, ESLint, etc.

Best for: React apps, user-facing features, full-stack apps

Guidance:

• Test user workflows, not implementation details
• Use Testing Library patterns (query by role, text)
• Mock only network/external services
• Unit test only complex business logic

When to Use Each

Override Strategy

Project-level - Add to

.claude/settings.json

{
  "testing": {
    "strategy": "pyramid"
  }
}

Directory-level - Add frontmatter to folder's

CLAUDE.md

---
testing_strategy: trophy
---

Vitest Patterns

Basic Test Structure

import { describe, it, expect, beforeEach, vi } from 'vitest';

describe('UserService', () => {
  let service: UserService;

  beforeEach(() => {
    service = new UserService();
    vi.clearAllMocks();
  });

  it('creates user with valid data', async () => {
    const user = await service.create({ email: 'test@example.com' });
    expect(user).toMatchObject({ email: 'test@example.com' });
  });

  it('throws on duplicate email', async () => {
    await service.create({ email: 'test@example.com' });
    await expect(service.create({ email: 'test@example.com' }))
      .rejects.toThrow('already exists');
  });
});

Mocking

// Mock module
vi.mock('@/lib/db', () => ({
  db: {
    user: {
      create: vi.fn(),
      findUnique: vi.fn(),
    },
  },
}));

// Mock implementation per test
it('handles not found', async () => {
  vi.mocked(db.user.findUnique).mockResolvedValue(null);
  await expect(getUser('123')).rejects.toThrow('Not found');
});

// Spy on existing function
const spy = vi.spyOn(console, 'error').mockImplementation(() => {});
expect(spy).toHaveBeenCalledWith('Error message');

React Component Testing

import { render, screen, waitFor } from '@testing-library/react';
import userEvent from '@testing-library/user-event';

it('submits form with valid data', async () => {
  const onSubmit = vi.fn();
  render(<LoginForm onSubmit={onSubmit} />);

  await userEvent.type(screen.getByLabelText(/email/i), 'test@example.com');
  await userEvent.type(screen.getByLabelText(/password/i), 'password123');
  await userEvent.click(screen.getByRole('button', { name: /submit/i }));

  await waitFor(() => {
    expect(onSubmit).toHaveBeenCalledWith({
      email: 'test@example.com',
      password: 'password123',
    });
  });
});

Playwright Patterns

Page Object Model

// tests/pages/login.page.ts
export class LoginPage {
  constructor(private page: Page) {}

  async goto() {
    await this.page.goto('/login');
  }

  async login(email: string, password: string) {
    await this.page.fill('[name="email"]', email);
    await this.page.fill('[name="password"]', password);
    await this.page.click('button[type="submit"]');
  }

  async expectError(message: string) {
    await expect(this.page.getByText(message)).toBeVisible();
  }
}

// tests/auth.spec.ts
test('user can login', async ({ page }) => {
  const loginPage = new LoginPage(page);
  await loginPage.goto();
  await loginPage.login('user@example.com', 'password');
  await expect(page).toHaveURL('/dashboard');
});

Fixtures

// tests/fixtures.ts
import { test as base } from '@playwright/test';
import { LoginPage } from './pages/login.page';

type Fixtures = {
  loginPage: LoginPage;
  authenticatedPage: Page;
};

export const test = base.extend<Fixtures>({
  loginPage: async ({ page }, use) => {
    await use(new LoginPage(page));
  },
  authenticatedPage: async ({ page }, use) => {
    await page.goto('/login');
    await page.fill('[name="email"]', 'test@example.com');
    await page.click('button[type="submit"]');
    await page.waitForURL('/dashboard');
    await use(page);
  },
});

Running Long E2E Tests Efficiently

E2E tests often take 2-5+ minutes. Use the TaskOutput pattern to avoid token waste.

The Problem: Polling Waste

❌ INEFFICIENT - wastes tokens on empty polling:
1. Bash(test, run_in_background: true) → task_id
2. Bash(sleep 60 && tail output)  ← empty, wasted tokens
3. Bash(cat output)               ← still empty, wasted
4. Bash(ps aux | grep test)       ← check if running, wasted
5. Bash(sleep 90 && cat output)   ← wasted again
... repeat 5-10 times before getting results

The Solution: Block Until Complete

✅ EFFICIENT - one call, waits for completion:
1. Bash(test, run_in_background: true) → task_id
2. TaskOutput(task_id, block: true, timeout: 300000)
   ↳ Waits up to 5 min, returns results when done

That's it. Two tool calls instead of 6+.

Pattern: E2E Test Execution

// Step 1: Start test in background
Bash("pnpm test:e2e -- e2e/checkout.spec.ts", run_in_background: true)
// Returns: task_id = "abc123"

// Step 2: Wait for completion (up to 5 min)
TaskOutput(task_id: "abc123", block: true, timeout: 300000)
// Returns full output when test completes

Pattern: Parallel E2E with TaskOutput

// Start multiple tests in parallel
Bash("pnpm test:e2e auth.spec.ts", run_in_background: true)  // task_a
Bash("pnpm test:e2e cart.spec.ts", run_in_background: true)  // task_b
Bash("pnpm test:e2e checkout.spec.ts", run_in_background: true)  // task_c

// Wait for all to complete
TaskOutput(task_id: "task_a", block: true, timeout: 300000)
TaskOutput(task_id: "task_b", block: true, timeout: 300000)
TaskOutput(task_id: "task_c", block: true, timeout: 300000)

When to Use Non-Blocking

Use

block: false

// Quick status check (doesn't wait)
TaskOutput(task_id: "abc123", block: false, timeout: 5000)

// Returns immediately with current output
// Use this if you want to continue other work

Timeout Guidelines

Red Flags

• Using

  sleep && tail

  to poll → Use TaskOutput instead
• Multiple

  cat

  commands checking same file → Use TaskOutput with block: true
• Checking

  ps aux | grep

  for process → TaskOutput tells you if still running

Anti-Patterns

// BAD: Testing implementation
expect(component.state.isLoading).toBe(true);

// GOOD: Testing behavior
expect(screen.getByRole('status')).toHaveTextContent('Loading...');

// BAD: Hardcoded delay
await page.waitForTimeout(2000);

// GOOD: Wait for condition
await page.waitForSelector('[data-loaded="true"]');

Test Organization

tests/
├── unit/              # Pure function tests
├── integration/       # Component + dependency tests
├── e2e/              # Playwright tests
│   ├── fixtures/
│   ├── pages/        # Page objects
│   └── *.spec.ts
└── setup.ts          # Global setup

Red Flags

• Tests that pass when code is broken
• Tests that fail intermittently (flaky)
• Tests longer than 50 lines (decompose)
• Mocking everything (test becomes meaningless)
• No assertions (test does nothing)
• Duplicated setup across tests (use fixtures)

Quick Reference

// Vitest config
export default defineConfig({
  test: {
    globals: true,
    environment: 'jsdom',
    setupFiles: ['./tests/setup.ts'],
    coverage: { reporter: ['text', 'html'] },
  },
});

// Playwright config
export default defineConfig({
  testDir: './tests/e2e',
  use: { baseURL: 'http://localhost:3000' },
  webServer: { command: 'npm run dev', port: 3000 },
});

// Fast feedback: watch mode
vitest --watch
playwright test --ui