database-best-practices

Multi-tenant data isolation in Prisma requires always filtering queries by shopId at the schema and query level. Implement indexed shopId fields, enforce filtering in all reads and writes, use transactions for consistency, and validate isolation in migrations to prevent cross-shop data exposure.

Design shop-scoped models with shopId as a required, indexed field. Use composite indexes on shopId plus business keys, implement upsert patterns for idempotent updates, and define explicit relationships that respect tenant boundaries to ensure queries remain shop-isolated.

Safe migrations apply schema changes without downtime or data loss. Use transactional patterns, test migrations on staging data first, avoid breaking changes mid-deployment, and verify shopId filtering remains enforced after each migration to maintain isolation.

Prisma transactions batch multiple operations atomically, preventing partial updates and race conditions. Combine transactions with pagination, select only needed fields, and eager-load related data to eliminate N+1 queries while maintaining multi-tenant isolation.

Forgetting shopId filters, selecting all fields instead of required ones, missing indexes on tenant keys, and running unbatched operations trigger data leaks and slow queries. Always filter by shopId, use pagination, batch bulk operations, and index composite keys for performance and safety.

JSON fields store nested data without new tables; query them with type-safe filters. Bulk operations batch inserts or updates; wrap them in transactions, maintain shopId filtering on each record, and validate input to prevent injection or isolation breaches.