eventmap-codegen

Event Mapping Code Generation Example

This example demonstrates how to use the eventmap-gen tool to generate type-safe mapping code between domain events and event sourcing types.

Directory Structure

eventmap-codegen/
├── domain/
│   └── user/
│       └── events/
│           ├── v1/
│           │   └── user_events.go    # Initial event versions
│           └── v2/
│               └── user_events.go    # Evolved UserRegistered event
├── infrastructure/
│   └── persistence/
│       └── (generated code goes here)
└── main.go                           # Example usage

Running the Example

1. Generate the Mapping Code

From the repository root:

go run ./cmd/eventmap-gen \
  -input ./examples/eventmap-codegen/domain/user/events \
  -output ./examples/eventmap-codegen/infrastructure/persistence \
  -package persistence \
  -module github.com/getpup/pupsourcing/examples/eventmap-codegen/domain/user/events

This will create event_mapping.gen.go in the infrastructure/persistence directory.

2. Run the Example

go run ./examples/eventmap-codegen

What This Example Shows

1. Pure Domain Events - Events in domain/user/events have no dependencies on infrastructure
2. Versioned Events - UserRegistered exists in both v1 and v2 with different schemas
3. Type-Safe Mapping with Generics - Generated code provides compile-time safety with generic type parameters
4. Round-Trip Conversion - Domain → ES → Domain conversion maintains data integrity
5. Options Pattern - Metadata injection (CausationID, CorrelationID, TraceID)
6. Automatic Test Generation - Unit tests are generated alongside the mapping code

Key Takeaways

• Domain events are pure Go structs with JSON tags
• Version directories (v1/, v2/) determine EventVersion
• Event type (e.g., "UserRegistered") is constant across versions
• Generated code handles serialization, deserialization, and version routing
• No runtime reflection - everything is explicit

Clean Architecture

This example follows clean architecture principles:

• Domain Layer: Pure business logic, no infrastructure dependencies
• Infrastructure Layer: Generated mapping code, depends on both domain and pupsourcing
• Clear Boundaries: Domain events remain testable in isolation

Schema Evolution

The example shows how to handle breaking changes:

1. Create new version directory (v2/)
2. Add evolved event struct with new fields
3. Regenerate mapping code
4. Both versions remain available for replaying historical events

Old events (v1) and new events (v2) can coexist in the same event stream!