test-driven-development

name	Test-Driven Development
description	Write tests before code using RED-GREEN-REFACTOR cycle

Test-Driven Development Skill

Overview

TDD transforms testing from an afterthought into the foundation of development. Tests are proof that code works correctly.

When to Invoke

• Writing new features
• Fixing bugs (Prove-It pattern)
• When asked to write tests
• Before any implementation work

---

Core Cycle: RED-GREEN-REFACTOR

RED Phase

Write a test that describes expected behavior. It must fail.

describe('calculateDiscount', () => {
  it('should apply 10% discount for orders over $100', () => {
    const result = calculateDiscount(150);
    expect(result).toBe(15);
  });
});

Run: npm test → Should FAIL

GREEN Phase

Write minimal code to pass the test. No extras.

function calculateDiscount(amount) {
  if (amount > 100) return amount * 0.1;
  return 0;
}

Run: npm test → Should PASS

REFACTOR Phase

Improve code while keeping tests green.

const DISCOUNT_THRESHOLD = 100;
const DISCOUNT_RATE = 0.1;

function calculateDiscount(amount) {
  if (amount <= DISCOUNT_THRESHOLD) return 0;
  return amount * DISCOUNT_RATE;
}

Run: npm test → Should still PASS

---

Prove-It Pattern (Bug Fixes)

Step 1: Write Failing Test

it('should handle empty cart without error (bug #456)', () => {
  const cart = new Cart([]);
  expect(() => cart.getTotal()).not.toThrow();
  expect(cart.getTotal()).toBe(0);
});

Step 2: Verify Failure

Run test → Confirms bug exists

Step 3: Fix Bug

getTotal() {
  if (this.items.length === 0) return 0;  // Fix
  return this.items.reduce((sum, item) => sum + item.price, 0);
}

Step 4: Verify Pass

Run test → Confirms fix works

Step 5: Run Full Suite

npm test  # No regressions

---

Test Pyramid

         ┌─────────┐
         │   E2E   │  5%  — Critical user flows
         │  Tests  │       Full system, minutes
         ├─────────┤
         │ Integr. │  15% — API + DB interactions
         │  Tests  │       Seconds
         ├─────────┤
         │  Unit   │  80% — Pure logic
         │  Tests  │       Milliseconds
         └─────────┘

---

Test Structure: AAA Pattern

it('should calculate tax for California', () => {
  // Arrange — Setup
  const order = createOrder({ state: 'CA', subtotal: 100 });
  
  // Act — Execute
  const tax = calculateTax(order);
  
  // Assert — Verify
  expect(tax).toBe(7.25);
});

---

DAMP > DRY in Tests

Tests should be Descriptive And Meaningful Phrases.

// ✅ DAMP — Self-contained and clear
it('should reject password without uppercase letter', () => {
  const result = validatePassword('lowercase123!');
  
  expect(result.valid).toBe(false);
  expect(result.errors).toContain('Must contain uppercase letter');
});

// ❌ Too DRY — Requires reading shared context
it('should reject invalid password', () => {
  expect(validate(INVALID_PASSWORD_NO_UPPER)).toBe(false);
});

---

Test Doubles (Preference Order)

1. Real implementations — Best, but may be slow
2. Fakes — In-memory DB, simplified implementations
3. Stubs — Return canned responses
4. Mocks — Verify interactions (use sparingly)

// ✅ Prefer: Real or fake
const db = createTestDatabase();
const user = await userService.create(db, userData);

// ⚠️ Use sparingly: Mocks
const mockDb = { create: vi.fn().mockResolvedValue(user) };

---

Naming Conventions

// Pattern: should [expected behavior] when [condition]

// ✅ Good
'should return empty array when no users exist'
'should throw ValidationError when email is invalid'
'should send welcome email after registration'

// ❌ Bad
'works correctly'
'test user creation'
'handles error'

---

Anti-Patterns

Pattern	Problem	Solution
Testing internals	Breaks on refactor	Test behavior, not implementation
Flaky tests	Erodes trust	Use deterministic data
Over-mocking	False confidence	Prefer real implementations
Snapshot abuse	Large diffs ignored	Use sparingly
Shared mutable state	Tests affect each other	Reset in beforeEach
Testing frameworks	Wasted effort	Only test your code

---

Verification Checklist

• All new code has tests
• Bug fixes have reproduction tests
• Tests describe behavior (readable names)
• No skipped tests
• Coverage maintained/improved
• Full suite passes