domain-driven-design（领域驱动设计）

Domain-Driven Design Patterns

Patterns for modeling complex business domains with clear boundaries, enforced invariants, and appropriate consistency strategies.

When to Activate

• Modeling business domains and entities
• Designing aggregate boundaries
• Implementing complex business rules
• Planning data consistency strategies
• Establishing bounded contexts
• Designing domain events and integration

Strategic Patterns

Bounded Context

A bounded context defines the boundary within which a domain model applies. The same term can mean different things in different contexts.

Example: "Customer" in different contexts

┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐
│    Sales        │  │    Support      │  │    Billing      │
│    Context      │  │    Context      │  │    Context      │
├─────────────────┤  ├─────────────────┤  ├─────────────────┤
│ Customer:       │  │ Customer:       │  │ Customer:       │
│ - Leads         │  │ - Tickets       │  │ - Invoices      │
│ - Opportunities │  │ - SLA           │  │ - Payment       │
│ - Proposals     │  │ - Satisfaction  │  │ - Credit Limit  │
└─────────────────┘  └─────────────────┘  └─────────────────┘

Context Identification

Ask these questions to find context boundaries:

• Where does the ubiquitous language change?
• Which teams own which concepts?
• Where do integration points naturally occur?
• What could be deployed independently?

Context Mapping

Define how bounded contexts integrate:

Ubiquitous Language

The shared vocabulary between developers and domain experts:

Building Ubiquitous Language:

1. EXTRACT terms from domain expert conversations
2. DOCUMENT in a glossary with precise definitions
3. ENFORCE in code - class names, method names, variables
4. EVOLVE as understanding deepens

Example Glossary Entry:
┌─────────────────────────────────────────────────────────────┐
│ Term: Order                                                  │
│ Definition: A confirmed request from a customer to purchase │
│             one or more products at agreed prices.          │
│ NOT: A shopping cart (which is an Intent, not an Order)     │
│ Context: Sales                                              │
└─────────────────────────────────────────────────────────────┘

Tactical Patterns

Entities

Objects with identity that persists over time. Equality is based on identity, not attributes.

Characteristics:
- Has a unique identifier
- Mutable state
- Lifecycle (created, modified, archived)
- Equality by ID

Example:
┌─────────────────────────────────────────┐
│ Entity: Order                           │
├─────────────────────────────────────────┤
│ Identity: orderId (UUID)                │
│ State: status, items, total             │
│ Behavior: addItem(), submit(), cancel() │
└─────────────────────────────────────────┘

class Order {
  private readonly id: OrderId;      // Identity - immutable
  private status: OrderStatus;        // State - mutable
  private items: OrderItem[];         // State - mutable

  constructor(id: OrderId) {
    this.id = id;
    this.status = OrderStatus.Draft;
    this.items = [];
  }

  equals(other: Order): boolean {
    return this.id.equals(other.id);  // Equality by identity
  }
}

Value Objects

Objects without identity. Equality is based on attributes. Always immutable.

Characteristics:
- No unique identifier
- Immutable (all properties readonly)
- Equality by attributes
- Self-validating

Example:
┌─────────────────────────────────────────┐
│ Value Object: Money                     │
├─────────────────────────────────────────┤
│ Attributes: amount, currency            │
│ Behavior: add(), subtract(), format()   │
│ Invariant: amount >= 0                  │
└─────────────────────────────────────────┘

class Money {
  constructor(
    public readonly amount: number,
    public readonly currency: Currency
  ) {
    if (amount < 0) throw new Error('Amount cannot be negative');
  }

  add(other: Money): Money {
    if (!this.currency.equals(other.currency)) {
      throw new Error('Cannot add different currencies');
    }
    return new Money(this.amount + other.amount, this.currency);
  }

  equals(other: Money): boolean {
    return this.amount === other.amount &&
           this.currency.equals(other.currency);
  }
}

When to Use Value Objects

Aggregates

A cluster of entities and value objects with a defined boundary. One entity is the aggregate root.

Aggregate Design Rules:

1. PROTECT invariants at aggregate boundary
2. REFERENCE other aggregates by identity only
3. UPDATE one aggregate per transaction
4. DESIGN small aggregates (prefer single entity)

Example:
┌─────────────────────────────────────────────────────────────┐
│ Aggregate: Order                                            │
│ Root: Order (entity)                                        │
├─────────────────────────────────────────────────────────────┤
│  ┌─────────────────┐                                        │
│  │ Order (Root)    │◄── Aggregate Root                      │
│  │ - orderId       │                                        │
│  │ - customerId ───┼──► Reference by ID only                │
│  │ - status        │                                        │
│  └────────┬────────┘                                        │
│           │                                                 │
│  ┌────────▼────────┐                                        │
│  │ OrderItem       │◄── Inside aggregate                    │
│  │ - productId ────┼──► Reference by ID only                │
│  │ - quantity      │                                        │
│  │ - price (Money) │◄── Value Object                        │
│  └─────────────────┘                                        │
└─────────────────────────────────────────────────────────────┘

Aggregate Sizing

Start Small:
- Begin with single-entity aggregates
- Expand only when invariants require it

Signs of Too-Large Aggregate:
- Frequent optimistic lock conflicts
- Loading too much data for simple operations
- Multiple users editing simultaneously
- Transactional failures across unrelated data

Signs of Too-Small Aggregate:
- Invariants not protected
- Business rules scattered across services
- Eventual consistency where immediate is required

Domain Events

Represent something that happened in the domain. Immutable facts about the past.

Event Structure:
┌─────────────────────────────────────────┐
│ Event: OrderPlaced                      │
├─────────────────────────────────────────┤
│ eventId: UUID                           │
│ occurredAt: DateTime                    │
│ aggregateId: orderId                    │
│ payload:                                │
│   - customerId                          │
│   - items                               │
│   - totalAmount                         │
└─────────────────────────────────────────┘

Naming Convention:
- Past tense (OrderPlaced, not PlaceOrder)
- Domain language (not technical)
- Include all relevant data (event is immutable)

class OrderPlaced implements DomainEvent {
  readonly eventId = uuid();
  readonly occurredAt = new Date();

  constructor(
    readonly orderId: OrderId,
    readonly customerId: CustomerId,
    readonly items: OrderItemData[],
    readonly totalAmount: Money
  ) {}
}

Event Patterns

Repositories

Abstract persistence, providing collection-like access to aggregates.

Repository Principles:
- One repository per aggregate
- Returns aggregate roots only
- Hides persistence mechanism
- Supports aggregate reconstitution

interface OrderRepository {
  findById(id: OrderId): Promise<Order | null>;
  findByCustomer(customerId: CustomerId): Promise<Order[]>;
  save(order: Order): Promise<void>;
  delete(order: Order): Promise<void>;
}

// Implementation hides persistence details
class PostgresOrderRepository implements OrderRepository {
  async findById(id: OrderId): Promise<Order | null> {
    const row = await this.db.query('SELECT * FROM orders WHERE id = $1', [id]);
    return row ? this.reconstitute(row) : null;
  }

  private reconstitute(row: OrderRow): Order {
    // Rebuild aggregate from persistence
  }
}

Consistency Strategies

Transactional Consistency (ACID)

Use for invariants within an aggregate:

Rule: One aggregate per transaction

// Good: Single aggregate updated
async function addItemToOrder(orderId: OrderId, item: OrderItem) {
  const order = await orderRepo.findById(orderId);
  order.addItem(item);  // Business rules enforced
  await orderRepo.save(order);
}

// Bad: Multiple aggregates in one transaction
async function createOrderWithInventory() {
  await db.transaction(async (tx) => {
    await orderRepo.save(order, tx);
    await inventoryRepo.decrement(productId, quantity, tx);  // Don't do this
  });
}

Eventual Consistency

Use for consistency across aggregates:

Pattern: Domain Events + Handlers

// Order aggregate publishes event
class Order {
  submit(): void {
    this.status = OrderStatus.Placed;
    this.addEvent(new OrderPlaced(this.id, this.customerId, this.items));
  }
}

// Separate handler updates inventory (eventually)
class InventoryHandler {
  async handle(event: OrderPlaced): Promise<void> {
    for (const item of event.items) {
      await this.inventoryService.reserve(item.productId, item.quantity);
    }
  }
}

Saga Pattern

Coordinate multiple aggregates with compensation:

Saga: Order Fulfillment

┌─────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────┐
│ Create  │────►│ Reserve     │────►│ Charge      │────►│ Ship    │
│ Order   │     │ Inventory   │     │ Payment     │     │ Order   │
└────┬────┘     └──────┬──────┘     └──────┬──────┘     └─────────┘
     │                 │                   │
     │ Compensate:     │ Compensate:       │ Compensate:
     │ Cancel Order    │ Release Inventory │ Refund Payment
     ▼                 ▼                   ▼

On failure at any step, execute compensation in reverse order.

Choosing Consistency

Anti-Patterns

Anemic Domain Model

// Anti-pattern: Logic outside domain objects
class Order {
  id: string;
  items: Item[];
  status: string;
}

class OrderService {
  calculateTotal(order: Order): number { ... }
  validate(order: Order): boolean { ... }
  submit(order: Order): void { ... }
}

// Better: Logic inside domain objects
class Order {
  private items: OrderItem[];
  private status: OrderStatus;

  get total(): Money {
    return this.items.reduce((sum, item) => sum.add(item.subtotal), Money.zero());
  }

  submit(): void {
    this.validate();
    this.status = OrderStatus.Submitted;
  }
}

Large Aggregates

// Anti-pattern: Everything in one aggregate
class Customer {
  orders: Order[];           // Could be thousands
  addresses: Address[];
  paymentMethods: PaymentMethod[];
  preferences: Preferences;
  activityLog: Activity[];   // Could be millions
}

// Better: Separate aggregates referenced by ID
class Customer {
  id: CustomerId;
  defaultAddressId: AddressId;
  defaultPaymentMethodId: PaymentMethodId;
}

class Order {
  customerId: CustomerId;    // Reference by ID
}

Primitive Obsession

// Anti-pattern: Primitive types for domain concepts
function createOrder(
  customerId: string,
  productId: string,
  quantity: number,
  price: number,
  currency: string
) { ... }

// Better: Value objects
function createOrder(
  customerId: CustomerId,
  productId: ProductId,
  quantity: Quantity,
  price: Money
) { ... }

Implementation Checklist

Aggregate Design

• Single entity can be aggregate root
• Invariants are protected at boundary
• Other aggregates referenced by ID only
• Fits in memory comfortably
• One transaction per aggregate

Entity Implementation

• Has unique identifier
• Equality based on ID
• Encapsulates business rules
• State changes through methods

Value Object Implementation

• All properties immutable
• Equality based on attributes
• Self-validating
• Operations return new instances

Repository Implementation

• One per aggregate
• Returns aggregate roots only
• Hides persistence details
• Supports queries needed by domain

References

• Pattern Implementation Examples - Code examples in multiple languages
• Aggregate Design Guide - Detailed aggregate sizing heuristics