refactor（代码重构）

Refactoring Skill

Improve code structure and quality while preserving behavior.

Name

han-core:refactor - Restructure code to improve quality without changing behavior

Synopsis

/refactor [arguments]

Core Principle

Tests are your safety net. Never refactor without tests.

The Refactoring Cycle

1. Ensure tests exist and pass
2. Make ONE small change
3. Run tests (must still pass)
4. Commit (keep changes isolated)
5. Repeat

Each step must be reversible. If tests fail, revert and try smaller change.

Pre-Refactoring Checklist

STOP if any of these are false:

• Tests exist for code being refactored
• All tests currently pass
• Understand what code does
• External behavior will remain unchanged
• Have time to do this properly (not rushing)

If no tests exist:

1. Add tests first
2. Verify tests pass
3. THEN refactor

When to Refactor

Code Smells That Suggest Refactoring

Readability issues:

• Long functions (> 50 lines)
• Deep nesting (> 3 levels)
• Unclear naming
• Magic numbers
• Complex conditionals

Maintainability issues:

• Duplication (same code in multiple places)
• God classes (too many responsibilities)
• Feature envy (method uses another class more than its own)
• Data clumps (same groups of parameters passed around)

Complexity issues:

• Cyclomatic complexity > 10
• Too many dependencies
• Tightly coupled code
• Difficult to test

When NOT to Refactor

• No tests exist (add tests first)
• Under deadline pressure (defer to later)
• Code works and is readable (don't over-engineer)
• Changing external behavior (that's not refactoring, that's a feature/fix)
• Right before release (too risky)

Classic Refactorings

Extract Function

Problem: Function does too many things

// Before: Long function doing multiple things
function processOrder(order: Order) {
  // Validate order
  if (!order.items || order.items.length === 0) {
    throw new Error('Empty order')
  }
  if (!order.customer || !order.customer.email) {
    throw new Error('Invalid customer')
  }

  // Calculate totals
  let subtotal = 0
  for (const item of order.items) {
    subtotal += item.price * item.quantity
  }
  const tax = subtotal * 0.08
  const shipping = subtotal > 50 ? 0 : 9.99
  const total = subtotal + tax + shipping

  // Save to database
  return database.save({
    ...order,
    subtotal,
    tax,
    shipping,
    total
  })
}

// After: Extracted into focused functions
function processOrder(order: Order) {
  validateOrder(order)
  const totals = calculateTotals(order)
  return saveOrder(order, totals)
}

function validateOrder(order: Order): void {
  if (!order.items || order.items.length === 0) {
    throw new Error('Empty order')
  }
  if (!order.customer || !order.customer.email) {
    throw new Error('Invalid customer')
  }
}

function calculateTotals(order: Order) {
  const subtotal = order.items.reduce(
    (sum, item) => sum + item.price * item.quantity,
    0
  )
  const tax = subtotal * 0.08
  const shipping = subtotal > 50 ? 0 : 9.99
  const total = subtotal + tax + shipping

  return { subtotal, tax, shipping, total }
}

function saveOrder(order: Order, totals: Totals) {
  return database.save({ ...order, ...totals })
}

Benefits: Each function has single responsibility, easier to test, easier to understand

Extract Variable

Problem: Complex expression that's hard to understand

// Before: Dense, hard to parse
if (user.age >= 18 && user.country === 'US' && !user.banned && user.verified) {
  // ...
}

// After: Intent is clear
const isAdult = user.age >= 18
const isUSResident = user.country === 'US'
const hasGoodStanding = !user.banned && user.verified
const canPurchase = isAdult && isUSResident && hasGoodStanding

if (canPurchase) {
  // ...
}

Benefits: Self-documenting, easier to debug, easier to modify

Inline Function/Variable

Problem: Unnecessary indirection that doesn't add clarity

// Before: Over-abstraction
function getTotal(order: Order) {
  return calculateTotalAmount(order)
}

function calculateTotalAmount(order: Order) {
  return order.subtotal + order.tax
}

// After: Inline the unnecessary layer
function getTotal(order: Order) {
  return order.subtotal + order.tax
}

When to inline: Abstraction doesn't add value, makes code harder to follow

Rename

Problem: Unclear or misleading names

// Before: Unclear
function proc(d: any) {
  const r = d.x * d.y
  return r
}

// After: Self-explanatory
function calculateArea(dimensions: Dimensions) {
  const area = dimensions.width * dimensions.height
  return area
}

Benefits: Code is self-documenting, no need to guess what variables mean

Replace Magic Number with Named Constant

Problem: Unexplained numbers in code

// Before: What's 0.08? What's 9.99?
const tax = subtotal * 0.08
const shipping = subtotal > 50 ? 0 : 9.99

// After: Clear meaning
const TAX_RATE = 0.08
const FREE_SHIPPING_THRESHOLD = 50
const STANDARD_SHIPPING_COST = 9.99

const tax = subtotal * TAX_RATE
const shipping = subtotal > FREE_SHIPPING_THRESHOLD ? 0 : STANDARD_SHIPPING_COST

Remove Duplication

Problem: Same code in multiple places

// Before: Duplication
function formatUserName(user: User) {
  return `${user.firstName} ${user.lastName}`.trim()
}

function formatAdminName(admin: Admin) {
  return `${admin.firstName} ${admin.lastName}`.trim()
}

function formatAuthorName(author: Author) {
  return `${author.firstName} ${author.lastName}`.trim()
}

// After: One implementation
function formatFullName(person: { firstName: string; lastName: string }) {
  return `${person.firstName} ${person.lastName}`.trim()
}

// Usage
formatFullName(user)
formatFullName(admin)
formatFullName(author)

Simplify Conditional

Problem: Complex nested if/else

// Before: Nested conditionals
function getShippingCost(order: Order) {
  if (order.total > 100) {
    return 0
  } else {
    if (order.items.length > 5) {
      return 5.99
    } else {
      if (order.weight > 10) {
        return 15.99
      } else {
        return 9.99
      }
    }
  }
}

// After: Early returns, flat structure
function getShippingCost(order: Order) {
  if (order.total > 100) return 0
  if (order.items.length > 5) return 5.99
  if (order.weight > 10) return 15.99
  return 9.99
}

// Or: Look-up table
const SHIPPING_RULES = [
  { condition: (o: Order) => o.total > 100, cost: 0 },
  { condition: (o: Order) => o.items.length > 5, cost: 5.99 },
  { condition: (o: Order) => o.weight > 10, cost: 15.99 },
]

function getShippingCost(order: Order) {
  const rule = SHIPPING_RULES.find(r => r.condition(order))
  return rule?.cost ?? 9.99
}

Replace Conditional with Polymorphism

Problem: Type checks scattered throughout code

// Before: Type checking everywhere
function calculatePrice(item: Item) {
  if (item.type === 'book') {
    return item.basePrice * 0.9  // 10% discount
  } else if (item.type === 'electronics') {
    return item.basePrice * 1.15  // 15% markup
  } else if (item.type === 'clothing') {
    return item.basePrice
  }
}

// After: Polymorphism
interface Item {
  calculatePrice(): number
}

class Book implements Item {
  calculatePrice() {
    return this.basePrice * 0.9
  }
}

class Electronics implements Item {
  calculatePrice() {
    return this.basePrice * 1.15
  }
}

class Clothing implements Item {
  calculatePrice() {
    return this.basePrice
  }
}

// Usage: No type checking needed
const price = item.calculatePrice()

Split Function

Problem: Function tries to do too many things

// Before: Does validation, calculation, and saving
function processPayment(payment: Payment) {
  // Validation
  if (!payment.amount || payment.amount <= 0) {
    throw new Error('Invalid amount')
  }
  if (!payment.method) {
    throw new Error('Payment method required')
  }

  // Calculation
  const fee = payment.amount * 0.029 + 0.30
  const total = payment.amount + fee

  // Persistence
  const record = database.save({
    amount: payment.amount,
    fee,
    total,
    method: payment.method,
    timestamp: Date.now()
  })

  // Notification
  notificationService.send({
    user: payment.user,
    message: `Payment of $${total} processed`
  })

  return record
}

// After: Separate concerns
function processPayment(payment: Payment) {
  validatePayment(payment)
  const totals = calculatePaymentTotals(payment)
  const record = savePayment(payment, totals)
  notifyPaymentProcessed(payment.user, totals.total)
  return record
}

Refactoring Golden Rules

Safety first:

• Tests exist and pass before starting
• Make one change at a time
• Run tests after each change
• Behavior must remain unchanged
• Commit after each successful refactoring

Refactoring Workflow

Step-by-Step Process

# 1. Ensure tests pass
npm test
# All tests passing

# 2. Make ONE refactoring change
# Example: Extract function

# 3. Run tests immediately
npm test
# Still passing

# 4. Commit with descriptive message
git add .
git commit -m "refactor: extract validateOrder function"

# 5. Repeat for next refactoring
# Make another small change, test, commit

If Tests Fail After Refactoring

# Tests failed after refactoring

# Option 1: Revert and try smaller change
git reset --hard HEAD
# Make smaller, safer change

# Option 2: Debug and fix
# Find what broke
# Fix it
# Run tests again

Refactoring Strategies

The Boy Scout Rule

"Leave code better than you found it"

When touching code for any reason:

1. Fix obvious issues you see
2. Improve naming
3. Extract complex expressions
4. Add missing tests
5. Remove commented code

Small improvements accumulate

Preparatory Refactoring

Before adding feature, refactor to make it easy

1. Need to add feature
2. Current code structure makes it hard
3. Refactor first to make space
4. Then add feature in clean code

Quote: "Make the change easy, then make the easy change"

Opportunistic Refactoring

Fix things you notice while working

• Fixing bug? Clean up surrounding code
• Adding feature? Improve structure
• Reading code? Fix confusing names

Planned Refactoring

Dedicated time to improve code health

• Tech debt tickets
• Refactoring sprints
• Clean-up sessions

Refactoring Safety Checklist

Before every change:

• Tests exist
• Tests pass
• Understand what code does

After every change:

• Tests still pass
• No functionality changed
• Code is clearer
• Ready to commit

If tests fail:

• Understand why
• Fix or revert
• Never commit broken tests

Common Refactoring Pitfalls

Refactoring Without Tests

Risk: Change behavior without noticing

Solution: Add tests first, then refactor

Too Many Changes at Once

Risk: Hard to debug if something breaks

Solution: One refactoring at a time, commit frequently

Changing Behavior

Risk: It's not refactoring if behavior changes

Solution: Tests must still pass, functionality unchanged

Over-Engineering

Risk: More complex after "refactoring"

Solution: Simpler is better, don't add unnecessary abstraction

Refactoring Under Pressure

Risk: Mistakes due to rushing

Solution: Defer to when you have time to do it right

Measuring Refactoring Success

Good refactoring results in:

• Easier to understand
• Easier to modify
• Easier to test
• Fewer lines of code (usually)
• Lower complexity
• Same or better performance
• All tests still pass

If any test fails, it wasn't successful refactoring

Output Format

After refactoring:

## Refactoring: [Brief description]

### Before
[Description of code smell or issue]

### Changes Made
- [Change 1 with reasoning]
- [Change 2 with reasoning]
- [Change 3 with reasoning]

### After
[How the code is better now]

### Verification
[Evidence that behavior unchanged - use proof-of-work skill]
- All tests pass: [test output]
- No functionality changed
- Code is more [readable/maintainable/simple]

Examples

When the user says:

• "This function is too long and hard to understand"
• "Clean up this messy code"
• "Remove duplication between these modules"
• "Simplify this nested if/else logic"
• "Break this god class into smaller pieces"

Tools

Automated refactoring tools:

• IDE refactoring commands (safe)
• Rename variable/function (safe)
• Extract method (safe)
• Move file (safe)

Manual refactoring:

• Make small changes
• Test frequently
• Commit after each change
• Use version control as safety net

Integration with Other Skills

• boy-scout-rule - Leave code better than found
• simplicity-principles - KISS, YAGNI, simple is better
• solid-principles - Single Responsibility, etc.
• structural-design-principles - Composition, encapsulation
• test-driven-development - Add tests if missing
• proof-of-work - Verify tests still pass
• code-review - Review refactored code

Remember

1. Tests first - No refactoring without tests
2. Small steps - One change at a time
3. Test after each step - Must stay green
4. Commit frequently - Each safe change gets a commit
5. Behavior unchanged - If behavior changes, it's not refactoring

Refactoring is about improving structure without changing what the code does.