projection-patterns

Build read models from event streams by defining a Projection subclass, wiring it to a Projector, and registering it to process events. This creates queryable materialized views for scalable CQRS read sides.

Live projections maintain real-time read models from active streams, catchup projections rebuild from historical events, persistent projections checkpoint state to resume after failures, and inline projections update read models in the same transaction as writes.

Use catchup projections to rebuild read models from historical events, or persistent projections to checkpoint state and resume processing from the last saved point after a failure occurs.

Yes, projection patterns provide reusable templates and ready-to-adapt projectors for Python implementations, supporting common domain scenarios like order processing, product catalogs, and customer analytics in event-sourced architectures.

Use inline projections when you need strong consistency by updating read models in the same transaction as writes. Use live projections when maintaining real-time read models from active event streams without strict transactional guarantees.

Projections for cross-stream analytics require transactional updates and checkpoint management to handle failures. Complex cross-stream scenarios may need persistent projection architectures to ensure state consistency and enable recovery.