implementing-query-pagination

Pagination in Prisma uses cursor-based or offset-based strategies. Cursor pagination provides stable performance regardless of dataset size by tracking a unique field position; offset pagination uses skip/take but degrades with large data. Choose cursor pagination for infinite scroll and offset for small, static datasets.

Cursor pagination marks a position using a unique field and orders results consistently, ideal for infinite scroll with reliable performance. Offset pagination skips N rows and fetches the next batch, simpler to implement but slower on large datasets since it rescans skipped rows.

Cursor pagination uses take to limit results, orderBy to establish consistent ordering, and cursor to specify the starting position. Include the cursor field in orderBy, pass the previous record's cursor value to resume pagination, and return nextCursor from results for the next page.

Offset pagination suits small, static datasets in admin dashboards or internal tooling where page jumps are required and performance is less critical. Cursor pagination is preferred by default for production APIs and infinite-scroll interfaces serving large datasets efficiently.

Composite cursors combine multiple fields to create a unique sort position when individual fields aren't unique. Include all ordering fields in the cursor definition and orderBy clause to ensure stable, reproducible pagination across requests without duplicate or skipped records.

Create database indexes on fields used in orderBy clauses to accelerate cursor lookups and range scans. For composite cursors, use compound indexes matching the orderBy field order; proper indexing prevents full table scans and ensures pagination remains fast as datasets grow.