code-optimization

Code Optimization

Works In Both Modes

• Repo mode — analyze actual source files
• Chat mode — analyze pasted code snippets

What to Look For

Memory Leaks

• Connections opened but never closed (DB, HTTP, file handles)
• Event listeners added but never removed
• Unbounded in-memory caches or queues (grows forever)
• Objects held in global scope that should be request-scoped
• Circular references preventing garbage collection

Inefficient Loops & Algorithms

• N+1 query patterns (DB query inside a loop)
• Nested loops over large datasets (O(n²) or worse)
• Repeated computation of the same value inside a loop (move outside)
• Sorting inside a loop (sort once, outside)
• Linear search where a map/set lookup would be O(1)

Blocking Operations

• Synchronous file I/O in an async/event-driven service
• Blocking network calls on the main thread
• CPU-intensive work blocking an event loop (Node.js, Python asyncio)
• Long-running DB queries without timeout

Container & Infrastructure Waste

• Oversized Docker images (unnecessary packages, dev deps in prod image)
• Redundant image layers (multiple RUN commands that could be combined)
• CPU/memory limits set far above actual usage
• Idle replicas with no autoscaling configured
• Logs written to disk inside container instead of stdout

Common Language Patterns

Python:

• list.append() in a loop where a list comprehension would do
• + string concatenation in a loop (use join)
• Reading entire file into memory when streaming is possible

JavaScript/Node:

• Promise chains that could run in parallel (Promise.all)
• console.log left in hot paths
• Synchronous fs methods (fs.readFileSync) in async handlers

Go:

• Goroutine leaks (goroutines started but never terminated)
• Mutex held too long — narrow the critical section
• Unnecessary allocations in hot paths

Reporting Format

For each finding:

[SEVERITY] Type — Description
  Location: file:line or "pasted code, line X"
  Impact: what this costs (memory, CPU, latency, etc.)
  Fix: specific change to make
  Example:
    Before: <bad code>
    After:  <fixed code>

Severity: HIGH (causes real degradation), MEDIUM (wastes resources), LOW (minor improvement).