Refactoring Catalog

Systematic, low-risk transformations to improve code structure, readability, and maintainability without changing observable behavior.

Context

You are guiding code improvement through safe refactoring. Your role is to:

• Identify refactoring opportunities (duplication, poor names, long methods, complex conditionals)
• Apply proven transformations from Martin Fowler's catalog
• Use tests as a safety net to ensure behavior is preserved
• Keep changes small and focused (one refactoring per commit)
• Document the rationale for each refactoring

Refactoring is the art of cleaning house without breaking anything. Each refactoring is tiny, testable, and immediately verifiable.

Domain Context

Based on Martin Fowler's Refactoring and Michael Feathers' Working Effectively with Legacy Code:

• Red-Green-Refactor Cycle: Refactor only when tests are green. Tests are the safety net.
• Small Steps: Each refactoring should be a 5-minute task. If it takes longer, break it down.
• Behavior Preservation: Refactoring changes structure, never behavior. Observable inputs/outputs remain identical.
• Baby Steps: Commit frequently. Each commit should be revertible and self-contained.
• Duplication Is Evil: "Once is code, twice is coincidence, three times is a pattern. Extract."
• Names Matter: Spend time on clear, intention-revealing names. Better names = lower cognitive load.

When to Use This Skill

• Code has duplication across multiple locations
• Method names don't clearly express intent
• A function is doing multiple things (low cohesion)
• Conditional logic is deeply nested or complex
• Parameters are loosely related and could be grouped
• Coupling between classes is high; methods belong elsewhere
• After implementing a feature, code needs improvement (Refactor phase of TDD)
• Preparing code for a large change (improve structure first)

Prerequisites

Before refactoring, ensure:

• Tests Are Green: All tests pass. Refactoring is a destructive operation if tests fail.
• Test Coverage: Code should have good test coverage (70%+) in the area being refactored
• Small Scope: Refactoring only in the module/class at hand; don't refactor the whole system at once
• Clear Criteria: Know what "better" looks like (clarity, cohesion, reduced duplication)
• Time: Refactoring takes focus; don't multitask or rush

Instructions (Detailed)

1. Identify the Refactoring Opportunity

Read the code and ask:

• Are there similar code blocks in multiple places? → Extract Method / Consolidate Duplicate Code
• Does a variable have a vague name? → Rename Variable
• Is the function doing multiple things? → Extract Method / Extract Class
• Are parameters unrelated? → Introduce Parameter Object
• Is there a boolean flag parameter? → Replace Flag Argument with Methods
• Is there a complex conditional chain? → Replace Conditional with Polymorphism
• Does a method belong to a different class? → Move Method

Example identifying refactoring opportunities:

# Poor: Vague names, duplicated logic, multiple responsibilities
def process_order(order):
    subtotal = 0
    for item in order.items:
        subtotal += item.price * item.qty
    subtotal *= 0.95  # 5% discount (magic number!)
    if order.is_vip:
        subtotal *= 0.9  # another 10% for VIP
    tax = subtotal * 0.08
    total = subtotal + tax
    print(f"Total: {total}")
    if total > 100:
        order.flag_for_priority_shipping()
    return total

Refactoring opportunities:

• Extract calculate_subtotal() (duplicated calculation logic across similar methods)
• Extract apply_discount() (separate discount logic)
• Extract apply_tax() (separate tax logic)
• Replace magic numbers with named constants
• Extract is_priority_order() check (readability)

2. Write/Verify Tests Cover the Code

Before refactoring, ensure tests cover the behavior you'll change:

def test_process_order_calculates_correct_total():
    order = Order(items=[
        Item("Widget", 50, 2),   # 2x Widget @ $50 = $100
        Item("Gadget", 25, 1)    # 1x Gadget @ $25 = $25
    ], is_vip=False)
    # Subtotal: $125 → -5% discount = $118.75 → +8% tax = $128.25
    assert process_order(order) == 128.25

def test_process_order_vip_discount():
    order = Order(items=[
        Item("Widget", 100, 1)
    ], is_vip=True)
    # Subtotal: $100 → -5% = $95 → -10% = $85.50 → +8% tax = $92.34
    assert process_order(order) == 92.34

def test_process_order_flags_priority_for_large_orders():
    order = Order(items=[
        Item("Widget", 60, 2)  # $120
    ], is_vip=False)
    process_order(order)
    assert order.priority_shipping_flagged is True

Run these tests. They pass with the current (messy) code. Now refactor with confidence.

3. Apply First Refactoring: Extract Method

Goal: Move duplicated or complex logic into a helper method.

Before:

def process_order(order):
    subtotal = 0
    for item in order.items:
        subtotal += item.price * item.qty
    subtotal *= 0.95
    if order.is_vip:
        subtotal *= 0.9
    tax = subtotal * 0.08
    total = subtotal + tax
    # ... more logic

After:

def process_order(order):
    subtotal = self._calculate_subtotal(order)
    subtotal = self._apply_discount(subtotal, order.is_vip)
    total = self._apply_tax(subtotal)
    # ... more logic

def _calculate_subtotal(self, order):
    """Sum of all item prices (unit price * quantity)."""
    subtotal = 0
    for item in order.items:
        subtotal += item.price * item.qty
    return subtotal

def _apply_discount(self, subtotal, is_vip):
    """Apply standard 5% discount, plus 10% for VIP."""
    subtotal *= 0.95
    if is_vip:
        subtotal *= 0.9
    return subtotal

def _apply_tax(self, subtotal):
    """Apply 8% sales tax."""
    return subtotal * 1.08

Run tests: All tests should pass. Behavior is identical.

Commit: refactor: extract discount and tax logic into helper methods

4. Apply Next Refactoring: Replace Magic Numbers

Goal: Replace hardcoded numbers with named constants.

Before:

def _apply_discount(self, subtotal, is_vip):
    subtotal *= 0.95     # What is 0.95?
    if is_vip:
        subtotal *= 0.9  # What is 0.9?
    return subtotal

def _apply_tax(self, subtotal):
    return subtotal * 1.08  # What is 1.08?

After:

STANDARD_DISCOUNT_RATE = 0.05  # 5% discount
VIP_DISCOUNT_RATE = 0.10       # Additional 10% for VIP
SALES_TAX_RATE = 0.08          # 8% sales tax

def _apply_discount(self, subtotal, is_vip):
    subtotal *= (1 - self.STANDARD_DISCOUNT_RATE)
    if is_vip:
        subtotal *= (1 - self.VIP_DISCOUNT_RATE)
    return subtotal

def _apply_tax(self, subtotal):
    return subtotal * (1 + self.SALES_TAX_RATE)

Run tests: All tests should pass.

Commit: refactor: replace magic numbers with named constants

5. Apply Next Refactoring: Replace Conditional with Method

Goal: Replace boolean flag parameters with separate, well-named methods.

Before:

def process_order(order, include_tax=True):  # Flag parameter; confusing
    subtotal = self._calculate_subtotal(order)
    subtotal = self._apply_discount(subtotal, order.is_vip)
    if include_tax:
        total = self._apply_tax(subtotal)
    else:
        total = subtotal
    return total

After:

def process_order(order):
    """Calculate order total including all discounts and taxes."""
    subtotal = self._calculate_subtotal(order)
    subtotal = self._apply_discount(subtotal, order.is_vip)
    total = self._apply_tax(subtotal)
    return total

def calculate_order_subtotal(order):
    """Calculate order total without taxes (for internal use)."""
    subtotal = self._calculate_subtotal(order)
    subtotal = self._apply_discount(subtotal, order.is_vip)
    return subtotal

Now callers choose the appropriate method name, making intent clear.

Run tests: All tests should pass.

Commit: refactor: replace include_tax flag parameter with distinct methods

6. Apply Refactoring: Move Method

Goal: Move a method to the class that uses it most, improving cohesion.

Before:

class Order:
    def process(self):
        # ...
        total = order_processor._apply_tax(subtotal)

class OrderProcessor:
    def _apply_tax(self, subtotal):
        return subtotal * (1 + TAX_RATE)

After (Tax logic belongs to Order, not OrderProcessor):

class Order:
    def apply_tax(self, subtotal):
        """Apply sales tax to subtotal."""
        return subtotal * (1 + TAX_RATE)

    def process(self):
        # ...
        total = self.apply_tax(subtotal)

class OrderProcessor:
    # _apply_tax removed; moved to Order

Run tests: All tests should pass.

Commit: refactor: move tax calculation to Order class

7. Look for Remaining Duplication

Scan for copy-paste code:

Before:

def format_price_usd(price):
    return f"${price:.2f}"

def format_price_eur(price):
    return f"€{price:.2f}"

def format_price_gbp(price):
    return f"£{price:.2f}"

After:

def format_price(price, currency_symbol="$"):
    return f"{currency_symbol}{price:.2f}"

Or use a Currency object for better abstraction.

Run tests: All tests should pass.

Commit: refactor: consolidate currency formatting

8. Verify Final State

After a series of refactorings:

• Run full test suite
• Code is clearer and more maintainable
• No behavior changed (tests still pass)
• Each refactoring is reversible (can see in git history)

Output Format

When using this skill, deliver:

1. Refactoring Identified: What refactoring technique is being applied? Why?
2. Before Code: Original code showing the problem
3. After Code: Refactored code showing the improvement
4. Verification: Tests still pass; behavior unchanged
5. Commit Message: Conventional commit format

Example output:

## Refactoring: Extract Method

**Identified Problem**:
Discount calculation is duplicated across `process_order()` and `calculate_order_subtotal()`.

**Before**:
[code showing duplication]

**After**:
[code with helper method]

**Tests**: All pass. Behavior unchanged.

**Commit Message**:
refactor: extract discount calculation into helper method

Worked Example: Simplifying a Payment Processor

Starting code (smelly):

class PaymentProcessor:
    def process_payment(self, amount, customer_type, apply_fee=True):
        # Validate amount
        if amount <= 0:
            raise ValueError("Invalid amount")

        # Calculate fee based on customer type
        fee = 0
        if customer_type == "premium":
            fee = amount * 0.01
        elif customer_type == "standard":
            fee = amount * 0.025
        else:
            fee = amount * 0.05

        # Apply fee
        if apply_fee:
            total = amount + fee
        else:
            total = amount

        # Process
        print(f"Processing ${total} for {customer_type} customer")
        return total

Refactoring 1: Extract Method for validation

def _validate_amount(self, amount):
    if amount <= 0:
        raise ValueError("Invalid amount")

def process_payment(self, amount, customer_type, apply_fee=True):
    self._validate_amount(amount)
    fee = self._calculate_fee(amount, customer_type)
    if apply_fee:
        total = amount + fee
    else:
        total = amount
    print(f"Processing ${total} for {customer_type} customer")
    return total

Refactoring 2: Extract fee calculation

def _calculate_fee(self, amount, customer_type):
    """Calculate processing fee based on customer type."""
    fee_rates = {
        "premium": 0.01,
        "standard": 0.025,
        "standard": 0.05,  # default
    }
    rate = fee_rates.get(customer_type, 0.05)
    return amount * rate

Refactoring 3: Replace flag parameter with methods

def process_payment(self, amount, customer_type):
    """Process payment including fees."""
    self._validate_amount(amount)
    return self._calculate_total_with_fee(amount, customer_type)

def process_payment_without_fee(self, amount, customer_type):
    """Process payment without fees."""
    self._validate_amount(amount)
    return amount

def _calculate_total_with_fee(self, amount, customer_type):
    fee = self._calculate_fee(amount, customer_type)
    total = amount + fee
    print(f"Processing ${total} for {customer_type} customer")
    return total

Final refactored state:

class PaymentProcessor:
    FEE_RATES = {
        "premium": 0.01,
        "standard": 0.025,
        "basic": 0.05,
    }

    def process_payment(self, amount, customer_type):
        """Process payment with fees."""
        self._validate_amount(amount)
        fee = self._calculate_fee(amount, customer_type)
        total = amount + fee
        self._log_transaction(total, customer_type)
        return total

    def process_payment_without_fee(self, amount, customer_type):
        """Process payment without fees (e.g., internal transfers)."""
        self._validate_amount(amount)
        return amount

    def _validate_amount(self, amount):
        if amount <= 0:
            raise ValueError("Invalid amount")

    def _calculate_fee(self, amount, customer_type):
        rate = self.FEE_RATES.get(customer_type, 0.05)
        return amount * rate

    def _log_transaction(self, amount, customer_type):
        print(f"Processing ${amount} for {customer_type} customer")

Each refactoring:

• Has a clear purpose
• Is tested (behavior unchanged)
• Is committed separately
• Makes code easier to understand

Decision Framework

When refactoring, use these decisions:

• If duplicated >2 times: Extract to shared method or function
• If method >15 lines: Too long; extract helpers
• If variable name unclear: Rename it immediately (takes 30 seconds)
• If conditional logic nested >2 levels deep: Extract method or use guard clause
• If parameter list >4 arguments: Group related ones into Parameter Object
• If class >300 lines: Too large; consider splitting into multiple classes
• If tests take >5 seconds to run: Refactor towards faster unit tests (fewer integration tests)
• If refactoring takes >15 minutes: You're doing too much; break into smaller steps

Anti-Patterns (Expanded)

1. Refactoring Without Tests

Mistake: Change code structure without test coverage; hope nothing breaks.

Why LLMs make this: Confidence in code changes without safety net.

Guard: Before refactoring any code, ensure tests pass and cover the area being refactored (70%+ coverage).

Example:

• Bad: Refactor calculate_discount() without tests; verify manually
• Good: Ensure tests pass for calculate_discount() before refactoring

---

2. Mixing Refactoring and Feature Work

Mistake: In one commit/PR, both refactor code AND add a new feature.

Why LLMs make this: Feels efficient to clean up while implementing.

Guard: Separate refactoring commits from feature commits. Reviewers can't distinguish intent if mixed.

Example:

• Bad: PR: "Extract tax logic AND add new tax rate" (1 commit)
• Good: PR 1: "Extract tax logic" (refactoring), PR 2: "Add new tax rate" (feature)

---

3. Extracting Methods Too Early (YAGNI)

Mistake: Create helper methods for code that isn't duplicated.

Why LLMs make this: Speculative abstractions feel like good engineering.

Guard: Wait until you see duplication 2-3 times before extracting. One method alone is just splitting for no reason.

Example:

• Bad: Extract _format_debug_message() when used once
• Good: Use the code inline; extract when used in 2+ places

---

4. Large, Unfocused Refactorings

Mistake: Refactor many things in one commit (rename variables, extract methods, move classes).

Why LLMs make this: Completing the whole improvement feels productive.

Guard: One refactoring per commit. If it takes >15 minutes, break into smaller commits.

Example:

• Bad: "refactor: extract methods, rename variables, reorganize" (30 files)
• Good: "refactor: extract calculate_tax method" (2 files)

---

5. Not Running Tests Between Refactorings

Mistake: Make multiple refactorings, then run tests once.

Why LLMs make this: Batching feels efficient.

Guard: Run tests after each refactoring (every 5-15 minutes). Rapid feedback reveals problems immediately.

Example:

• Bad: Rename 5 variables, extract 3 methods, run tests once
• Good: Rename 1 variable, run tests; extract 1 method, run tests; repeat

---

6. Ignoring Performance Impact

Mistake: Refactor for clarity; accidentally introduce performance regression.

Why LLMs make this: Focus on readability; miss runtime implications.

Guard: Profile before and after refactoring. "Replace Temp with Query" introduces function calls; measure impact.

Example:

• Bad: Extract calculate_total() that's called in a loop 10,000x
• Good: Profile first; if in hot path, keep inline or cache result

---

Quality Checklist

Before considering a refactoring complete, verify:

• All tests pass — Run full suite; no failures or errors
• No behavior change — Inputs and outputs are identical to before
• Code is clearer — Names are better, structure is simpler, duplication is reduced
• One concern per commit — Refactoring only; no feature changes mixed in
• Refactoring is small — Can be reviewed quickly; <50 lines changed ideally
• Not over-engineered — Didn't extract prematurely; waited for duplication to emerge
• Performance verified — Profiled hot paths; no regressions
• Commit message is clear — Explains what refactoring and why
• Revertible — Can be undone if needed; has test safety net
• Readable by others — Another engineer can understand the refactoring intent

Further Reading

• Martin Fowler, Refactoring: Improving the Design of Existing Code (2nd ed., 2018) — The canonical reference. Read Chapters 6-9 for specific refactorings with detailed before/after.
• Michael Feathers, Working Effectively with Legacy Code (2004) — How to refactor code without tests (Characterization Tests).
• Kent Beck, Implementation Patterns (2007) — Philosophy and patterns for clean code structure.
• Sandro Mancuso, The Software Craftsman (2014) — Professionalism and disciplined refactoring as a daily practice.
• Robert C. Martin, Clean Code (2008) — Naming, function design, error handling; all driven by refactoring.
• Joshua Kerievsky, Refactoring to Patterns (2004) — How refactoring evolves code toward design patterns naturally.