refactor-dev

Use the Strangler Fig pattern: place an API gateway in front of the legacy app, build one small feature in the new architecture, route traffic incrementally, and repeat until the legacy system shrinks away. Always confirm tests are green before starting.

N+1 queries occur when fetching related records in a loop, causing one query per parent record. Fix it by collecting parent IDs and running a single batched query with an IN clause, reducing dozens of round trips to two.

Replace nested conditionals with lookup tables, use early returns instead of deep nesting, extract complex conditions into named functions, and split large components by section, modal, or list item. Target max function complexity under 30.

Extract a hook when you find coupled useState groups, complex useEffect logic with multiple dependencies, reusable business logic, or data transformations. This separates state management from UI rendering and enables isolated hook testing.

Mean latency hides tail behavior; a system can have a fast average while P99 latency is seconds long. Always measure P95 and P99 percentiles with tools like k6 or Artillery to find the real user-facing breaking point.

Regex search produces false positives, such as matching words in comments, and cannot understand code structure. For reliable refactoring targets, use AST-based structural search tools like Comby or static analyzers like knip for dead code detection.