backend-patterns

Wrap your handler in a withAuth middleware that extracts the Bearer token, verifies it with jwt.verify, and attaches the user to the request. The pattern returns 401 responses for missing or invalid tokens before the handler runs.

Batch fetch related records instead of querying in a loop. Collect all foreign keys, run one query to fetch the related entities, build a Map keyed by id, and merge results in memory.

The repository pattern abstracts data access behind an interface like MarketRepository with findAll, findById, and create methods. Concrete implementations such as SupabaseMarketRepository handle the actual queries, keeping business logic independent of the database.

Yes, cache-aside works with any database. Check Redis for a cached key first, and on a miss fetch from Supabase, then write the result back to Redis with a TTL such as 300 seconds.

Use a transaction when multiple writes must succeed or fail together, such as creating a market and its initial position. A Postgres function via Supabase rpc wraps both inserts atomically with automatic rollback on error.

In-memory rate limiters store request timestamps per process, so limits reset on restart and do not coordinate across multiple server instances. Distributed deployments need a shared store like Redis for accurate limiting.