backend-patterns

Define resource-based URLs like GET /api/markets/:id, abstract data access behind a repository interface, and place business logic in a service class that depends on the repository. This keeps handlers thin and business rules testable.

Batch fetch related records instead of querying in a loop. Collect foreign keys from the main result set, run one query for all related rows, then join them in memory using a Map keyed by ID.

Use the cache-aside pattern: check Redis first, and on a miss fetch from the database and write the result back with a TTL such as 300 seconds. A decorator repository class can wrap the base repository to add caching transparently.

No. Per-process in-memory counters reset on deploy, split across replicas, and fail open in serverless environments. Use a shared store like Redis, a gateway, or the platform's native limiter instead.

Verify the JWT to extract the user's role, map roles to permission lists, and wrap handlers with a higher-order function that checks the required permission before executing. Return 403 when permissions are insufficient.

Use transactions when multiple writes must succeed or fail together, such as creating a market with its initial position. Implement them as Postgres functions via supabase.rpc so rollback happens automatically on error.