backend-patterns

Define a repository interface for data access operations like findAll and findById, then create a service class that consumes the repository for business logic. This separates data access from business rules, making both layers independently testable and replaceable.

Batch fetch related records instead of querying inside loops. Collect all foreign key IDs, fetch the related records in one query, build a lookup map, and merge results in memory rather than executing one query per parent record.

The cache-aside pattern checks Redis first, falls back to the database on cache miss, then populates the cache with a TTL. Wrap your base repository in a cached decorator class and invalidate keys when underlying data changes.

Yes, extract the Bearer token from the authorization header, verify it with jsonwebtoken against your secret, and attach the decoded payload to the request. Wrap handlers in a higher-order middleware function to enforce authentication across routes.

In-memory rate limiting works for single-instance deployments and development, tracking request timestamps per identifier in a Map. For multi-instance production deployments, use Redis-backed limiting so limits are shared across all server instances.

Simple in-process queues catch errors per job but lack persistence, retry policies, and monitoring. Jobs are lost on process restart, so production systems should use durable queues with dead-letter handling and structured logging of failures.