Purging Aggregates
All three storage adapters implement the same API:
interface EventStore {
purgeAggregate(id: string): Promise<void>;
}The method accepts only the aggregate ID. It does not accept an aggregate class or aggregate name, and event deletion matches the ID across the adapter's event storage.
import { EVENT_STORE, type EventStore } from "@event-nest/core";
import { Inject, Injectable } from "@nestjs/common";
@Injectable()
export class PrivacyService {
constructor(@Inject(EVENT_STORE) private readonly eventStore: EventStore) {}
async eraseAccount(accountId: string): Promise<void> {
await this.eventStore.purgeAggregate(accountId);
}
}Deleted State
Each adapter deletes, in order, the aggregate's:
- Snapshots, when snapshot storage is enabled.
- Events.
- Aggregate version record.
Those deletions run in one adapter transaction. If a deletion fails, the transaction rejects instead of intentionally leaving a partially purged stream.
| Adapter | Transaction used | Deleted records |
|---|---|---|
| MongoDB | MongoDB client session with withTransaction | Snapshot documents, event documents, aggregate document |
| PostgreSQL | Knex database transaction | Snapshot rows, event rows, aggregate row |
| Microsoft SQL Server | Knex database transaction | Snapshot rows, event rows, aggregate row |
MongoDB deployments must support transactions for this operation, as required by MongoDB client sessions. Snapshot-disabled configurations are supported: the no-op snapshot store participates without requiring snapshot storage.
Idempotent Absence
Purging an unknown, adapter-valid ID succeeds and resolves to undefined. It does not affect unrelated aggregates. This makes repeated purge requests safe with respect to already-absent event-store data.
An invalid ID representation can still fail adapter validation. For example, the MongoDB adapter converts the supplied value to an ObjectId; it requires a 24-character hexadecimal ObjectId string and does not accept UUIDs.