common-domain-driven-design

common-domain-driven-design

Apply Domain-Driven Design (DDD) to model complex business domains, enforce clear boundaries, and produce maintainable, intent-revealing code.

When to use

Use this skill when:

• Designing or restructuring a domain model
• Creating new bounded contexts, aggregates, entities, or value objects
• Implementing domain events, repositories, or application services
• Reviewing code for proper separation of domain logic from infrastructure
• Discussing or documenting ubiquitous language with stakeholders

Core concepts

Ubiquitous Language

• Use a single, shared vocabulary between domain experts and developers.
• Every class, method, and variable in the domain layer must use terms from the ubiquitous language.
• If a term is ambiguous, clarify it and document the agreed definition.

Bounded Contexts

• A bounded context is an explicit boundary within which a domain model is defined and consistent.
• Different bounded contexts may use the same term with different meanings — that is expected.
• Communication between bounded contexts happens through well-defined interfaces (context mapping): Anti-Corruption Layer, Shared Kernel, Open Host Service, or Published Language.

Building blocks

Building block	Purpose
Entity	An object with a unique identity that persists over time.
Value Object	An immutable object defined only by its attributes, no identity.
Aggregate	A cluster of entities/value objects with a single root entity as entry.
Aggregate Root	The only entity through which external code may reference the aggregate.
Domain Event	A record of something meaningful that happened in the domain.
Repository	Abstracts persistence; retrieves and stores aggregates.
Domain Service	Stateless operation that doesn't naturally belong to an entity or VO.
Application Service	Orchestrates use cases; delegates to domain objects, no business logic.
Factory	Encapsulates complex creation logic for aggregates or entities.

Instructions

1. Identify the bounded context. Before writing code, determine which bounded context you are working in. Name it explicitly. If one does not exist yet, propose one using the ubiquitous language.

2. Define the ubiquitous language. List the key domain terms relevant to the task. Use these terms consistently in code, comments, commit messages, and documentation.

3. Model aggregates carefully.

   • Each aggregate should be as small as possible — only include what is needed to enforce invariants.
   • Reference other aggregates by identity (ID), not by direct object reference.
   • All state changes go through the aggregate root.
   • Protect invariants inside the aggregate — never allow external code to put it in an invalid state.

4. Use value objects liberally.

   • Prefer value objects over primitives for domain concepts (e.g., Money, EmailAddress, DateRange).
   • Value objects are immutable: operations return new instances.
   • Implement equality by comparing all attributes.

5. Raise domain events for side effects.

   • When an aggregate completes a meaningful state change, emit a domain event.
   • Name events in past tense using ubiquitous language (e.g., OrderPlaced, PaymentReceived).
   • Keep events minimal: include only the data a consumer needs.
   • Process side effects (notifications, projections, integrations) in event handlers, not in the aggregate.

6. Separate layers clearly.

   • Domain layer: entities, value objects, aggregates, domain events, repository interfaces, domain services. No framework or infrastructure imports.
   • Application layer: application services (use-case orchestrators), command/query handlers. Depends on domain layer only.
   • Infrastructure layer: repository implementations, messaging adapters, ORM mappings, external API clients. Implements interfaces defined in domain/application.
   • Presentation/API layer: controllers, serializers, DTOs. Translates between external input and application commands/queries.

7. Repository rules.

   • One repository per aggregate root.
   • Repository interface lives in the domain layer; implementation lives in infrastructure.
   • Repositories return and accept only whole aggregates (not partial entities).

8. Apply context mapping when crossing boundaries.

   • Use an Anti-Corruption Layer (ACL) to translate between your model and an external/legacy model.
   • Define explicit DTOs or integration events at context boundaries — never leak internal domain objects.
   • If two teams share a small model, use Shared Kernel with clear ownership rules.

9. Folder/module structure. Organize code to mirror bounded contexts:

   src/
     <bounded-context>/
       domain/
         model/          # Aggregates, entities, value objects
         event/          # Domain events
         service/        # Domain services
         repository/     # Repository interfaces
       application/
         command/        # Command handlers / use cases
         query/          # Query handlers / read models
       infrastructure/
         persistence/    # Repository implementations
         messaging/      # Event bus adapters
       presentation/     # Controllers, DTOs
   

10. Validate the design.

    • Every public method on an aggregate root should correspond to a domain action in the ubiquitous language.
    • If you cannot name something using domain terms, reconsider the model.
    • Ask: "Can this aggregate enforce its own invariants without reaching outside itself?" If no, redraw boundaries.
    • Confirm no business logic leaked into application services, controllers, or infrastructure.