v0.2.0

[0.2.0] — 2026-04-06

Added

• --profile flag — profiling is now opt-in, no longer runs on every analysis
• --profile --fix shows before/after performance comparison
• --profile alone profiles the original file without applying fixes
• 5 new detectors (total: 24):
  • N+1 Query — flags ORM calls inside loops (Django, SQLAlchemy, general DB patterns)
  • Linear Search — flags .index() and .count() on lists inside loops (O(n) operations)
  • JS for...in on arrays — suggests for...of or .forEach() instead
  • Go goroutine in loops — flags unbounded goroutine spawning, suggests worker pool
  • C++ pass by value — flags large types (vector, string, map etc.) passed without const&

Fixed

• Integer += no longer flagged as string concatenation (COUNTER_NAMES heuristic)
• Integer + no longer flagged as list concatenation
• Mutable default detector no longer fires on type annotations (e.g. def foo(x: List[str] = None))
• ''.join() fix hint no longer suggested for JavaScript — now correctly suggests array.join('')
• Duplicate list concatenation warnings on chained expressions (e.g. a + b + c)
• get_fixed_code no longer called when --fix is not passed — eliminates wasteful API call

Improved

• Expanded CHEAP_CALLS — append, pop, keys, items, values, extend, update, get, add, remove, copy, clear, discard, and common builtins no longer flagged as expensive
• Dotted method calls (e.g. list.append) now correctly matched against CHEAP_CALLS
• Language passed to all detectors — enables language-aware warning messages