English version: English
When to Use
- The aggregate boundaries have been designed, and the collaboration mechanisms between aggregates, and between aggregates and the outside, need to be defined.
- It is necessary to answer "how aggregates communicate, who orchestrates cross-aggregate processes, and what the persistence contracts are".
- As the final step of tactical modeling, complete all the "connecting tissue".
Input Requirements
- Required: Aggregate catalog and invariant table (from
ddd-aggregates), event candidate list (fromddd-discover)。 - Optional: Context mapping and contract information (from
ddd-context-map), context catalog and glossary (fromddd-contexts, used to distinguish the boundary between domain events and integration events).
Process
- Event normalization: Unify event naming rules (past tense, business-readable, no technical terms); define trigger conditions, source aggregate, key fields, and ordering requirements for each event.
- Distinguish scope: Distinguish domain events (within a context) from integration events (cross-context/published externally).
- Reliability strategy: Define idempotency keys, replay strategies, deduplication windows, and dead-letter handling for each event.
- Domain services: Identify operations that do not belong to any single aggregate but carry domain logic, and define them as domain services—clarifying inputs, outputs, aggregates used, and events.
- Repository interfaces: Define the semantic interface of the repository for each aggregate (load, persist, query)—define only the contract, without implementation.
- Factories: Identify complex aggregate creation logic and define factory methods—clarifying creation conditions, validation rules, and initial state.
- Subscribers and side effects: Clarify the consumers of each event, triggered actions, compensation strategies, and monitoring metrics.
Output
| Artifact | Structural Requirements |
|---|---|
| Domain event catalog | Table: event name, source aggregate, trigger condition, key fields, scope (Domain/Integration) |
| Integration event contracts | Table: event name, publisher, consumer, contract owner, version strategy, compatibility window |
| Idempotency and replay strategies | Table: event, idempotency key, deduplication window, replay rules, exception handling |
| Domain service definitions | Table: service name, responsibility, input, output, dependent aggregates/events, logic that should not be included |
| Repository interface list | Table: aggregate, method signature (semantic), semantic description, query boundary |
| Factory list | Table: factory name/method, creation target, creation conditions, validation rules, initial state |
| Subscribers and side effects | Table: event, subscriber, triggered action, compensation strategy, monitoring metrics |
Validation Checklist
- All event names use the past tense, with no technical term contamination
- Events published across contexts have a version strategy and compatibility window
- Key events provide idempotency keys and replay strategies
- Domain services do not contain business logic that should belong to aggregates (avoid anemic models)
- Repository interfaces define only semantic contracts, without implementation details (SQL, ORM)
- Every complex creation logic has a corresponding factory (no bare new in the application layer)
Retroactive trigger
- The event needs to carry private data of another aggregate (a clean event schema cannot be designed) → Retrospect to
ddd-aggregates(aggregate boundaries need to be adjusted to ensure the event is self-contained).
Example
@ddd-domain-interactions
基于以下聚合设计,帮我设计领域交互层:
- 聚合:Booking(根:Booking,VO:TimeSlot, BookingStatus)
- 聚合:RoomSchedule(根:RoomSchedule,VO:TimeSlotOccupancy)
- 事件候选:BookingConfirmed, BookingCancelled, CheckInRecorded
请输出领域事件目录、领域服务、仓储接口与订阅者清单。