cm-clean-code

Clean Code — Code Hygiene Gate

TL;DR

• Use after a feature lands, before PR review
• Detects: dead code, duplicates, naming smells, dependency drift
• Powered by TRIZ contradictions and Boy Scout rule
• Next: cm-code-review

Code that works is not enough. Code must be CLEAN.

Inspired by Clean Code (Robert C. Martin) + Refactoring (Martin Fowler) + TRIZ.

When to Use

ALWAYS when:

• After completing a feature (mandatory hygiene pass before PR)
• After cm-reactor migration (cleanup dead code from migration)
• Before code review (cm-code-review) — clean FIRST, review AFTER
• During technical debt sprints
• When code smells are detected (see Detection section)
• After AI-generated code sessions (AI tends to leave mess)
• When file grows beyond 300 lines

Run automatically after:

• cm-execution completes a task batch
• cm-reactor Phase 5 (post-migration cleanup)
• cm-tdd Refactor phase (Red → Green → Refactor)

Skip when:

• Quick hotfix (patch first, clean later — but schedule the cleanup!)
• Prototype/spike code (will be thrown away)

TRIZ Principles Applied

| # | Principle | How Applied | |---|-----------|-------------| | #1 | Segmentation | Break large files/functions into focused units | | #10 | Prior Action | Clean BEFORE it rots — don't wait for tech debt sprint | | #6 | Universality | One function should serve one purpose (SRP) | | #27 | Cheap Short-living | Quick small cleanups > expensive large refactors | | #2 | Taking Out | Extract what doesn't belong — separate concerns |

The 7-Point Hygiene Checklist

Run this checklist on every file touched. Each point has auto-detect criteria:

┌───┬──────────────────────┬──────────────────────────────┬────────────────────────┐
│ # │ Check                │ Auto-Detect                  │ Action                 │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 1 │ Dead Code            │ Unused exports, unreachable  │ DELETE — don't comment │
│   │                      │ branches, commented-out code │ out, DELETE            │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 2 │ Unused Imports       │ Import not used in file      │ REMOVE import line     │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 3 │ Magic Numbers        │ Literal numbers in logic     │ EXTRACT to named const │
│   │ & Strings            │ Repeated string literals     │                        │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 4 │ Naming               │ Single-letter vars (not i,j) │ RENAME to describe     │
│   │                      │ Abbreviations, inconsistent  │ intent clearly         │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 5 │ Single Responsibility│ Function does 2+ things      │ EXTRACT into separate  │
│   │ (SRP)                │ Class has 5+ public methods  │ focused units          │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 6 │ DRY (Don't Repeat)   │ Similar code blocks in 2+    │ EXTRACT shared logic   │
│   │                      │ places                       │ into reusable function │
├───┼──────────────────────┼──────────────────────────────┼────────────────────────┤
│ 7 │ Nesting Depth        │ if/for/while nested > 3      │ EXTRACT, early return, │
│   │                      │ levels deep                  │ guard clauses          │
└───┴──────────────────────┴──────────────────────────────┴────────────────────────┘

The Process

Phase 1: SCAN — Detect Code Smells

Goal: Find what's dirty before cleaning.

Automated scan (file-by-file):

For each file modified in current task:

  1. SIZE CHECK:
     IF lines > 300 → FLAG "Large file — consider splitting"
     IF any function > 50 lines → FLAG "Long function — extract methods"
  
  2. IMPORT CHECK:
     Scan imports → cross-reference with usage in file body
     Unused import → FLAG for removal
  
  3. DEAD CODE CHECK:
     Commented-out code blocks → FLAG for deletion
     Functions not called anywhere → FLAG (verify with codeintell)
     Unreachable code after return/throw → FLAG
  
  4. DUPLICATION CHECK:
     Similar code blocks (>5 lines identical/near-identical) → FLAG
     Copy-paste patterns → FLAG
  
  5. NAMING CHECK:
     Single-char variables (except loop vars i,j,k) → FLAG
     Inconsistent casing (camelCase vs snake_case in same file) → FLAG
     Generic names (data, result, temp, item, value, obj) → FLAG
  
  6. COMPLEXITY CHECK:
     Nesting > 3 levels → FLAG
     Function with > 4 parameters → FLAG
     Cyclomatic complexity > 10 → FLAG

Output: Smell Report

## Clean Code Scan: [filename]

| # | Smell | Line | Severity | Auto-fix? |
|---|-------|------|----------|-----------|
| 1 | Unused import: lodash | 3 | Low | ✅ Yes |
| 2 | Magic number: 86400 | 47 | Medium | ✅ Yes |
| 3 | Long function: processData (78 lines) | 23-101 | High | 🔧 Manual |
| 4 | Dead code: commented block | 112-125 | Low | ✅ Yes |

**Total smells: 4 | Auto-fixable: 3 | Manual: 1**

Phase 2: CLEAN — Apply Fixes

Goal: Fix each smell, one at a time, with tests passing between each fix.

Rules (from refactoring.guru + Martin Fowler):

1. Tests first: Ensure tests exist and pass BEFORE cleaning
2. One change at a time: Fix one smell → run tests → commit → next smell
3. Behavior preservation: Clean code MUST NOT change functionality
4. No feature additions: Cleaning and feature work are SEPARATE commits

Fix patterns:

Fix 1: Dead Code & Unused Imports

Action: DELETE (not comment out)
Rationale: Version control is your backup, not comments
Commit: "clean: remove dead code in [file]"

Fix 2: Magic Numbers → Named Constants

Before: if (retryCount > 3) { ... }
         setTimeout(fn, 86400000)

After:   const MAX_RETRIES = 3;
         const ONE_DAY_MS = 86_400_000;
         if (retryCount > MAX_RETRIES) { ... }
         setTimeout(fn, ONE_DAY_MS)

Commit: "clean: extract magic numbers in [file]"

Fix 3: Extract Method (TRIZ #2 Taking Out)

Before: 50+ line function doing validation + calculation + persistence

After:  function processOrder(order) {
          validateOrder(order);
          const total = calculateTotal(order);
          return persistOrder(order, total);
        }

Commit: "clean: extract methods from [function] in [file]"

Fix 4: Reduce Nesting (Guard Clauses)

Before: function foo(x) {
          if (x) {
            if (x.valid) {
              if (x.items.length > 0) {
                // actual logic buried 3 levels deep
              }
            }
          }
        }

After:  function foo(x) {
          if (!x) return;
          if (!x.valid) return;
          if (x.items.length === 0) return;
          // actual logic at top level
        }

Commit: "clean: reduce nesting with guard clauses in [file]"

Fix 5: DRY — Extract Shared Logic

Before: // In file_a.ts
        const formatted = `${date.getFullYear()}-${date.getMonth()+1}-${date.getDate()}`
        
        // In file_b.ts (same pattern)
        const formatted = `${d.getFullYear()}-${d.getMonth()+1}-${d.getDate()}`

After:  // In utils/date.ts
        export function formatDate(date: Date): string { ... }
        
        // Both files import formatDate()

Commit: "clean: extract shared date formatting to utils"

Fix 6: Improve Naming

Before: const d = getData();
        const r = process(d);
        const x = r.filter(i => i.v > 0);

After:  const orders = fetchPendingOrders();
        const processedOrders = applyDiscounts(orders);
        const validOrders = processedOrders.filter(order => order.total > 0);

Commit: "clean: improve naming in [file]"

Phase 3: VERIFY — Confirm Cleanliness

**Goal