- Add proper CORS headers for Blossom endpoints including X-SHA-256, X-Content-Length, X-Content-Type headers required by blossom-client-sdk - Add root-level Blossom routes (/upload, /media, /mirror, /report, /list/) for clients like Jumble that expect Blossom at root - Export BaseURLKey from pkg/blossom for use by app handlers - Make blossomRootHandler return URLs with /blossom prefix so blob downloads work via the registered /blossom/ route - Remove Access-Control-Allow-Credentials header (not needed for * origin) - Add Access-Control-Expose-Headers for X-Reason and other response headers Files modified: - app/blossom.go: Add blossomRootHandler, use exported BaseURLKey - app/server.go: Add CORS handling for blossom paths, register root routes - pkg/blossom/server.go: Fix CORS headers, export BaseURLKey - pkg/blossom/utils.go: Minor formatting - pkg/version/version: Bump to v0.36.12 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
22 KiB
DDD Anti-Patterns
This reference documents common anti-patterns encountered when implementing Domain-Driven Design, how to identify them, and remediation strategies.
Anemic Domain Model
Description
Entities that are mere data containers with getters and setters, while all business logic lives in "service" classes. The domain model looks like a relational database schema mapped to objects.
Symptoms
- Entities with only get/set methods and no behavior
- Service classes with methods like
orderService.calculateTotal(order) - Business rules scattered across multiple services
- Heavy use of DTOs that mirror entity structure
- "Transaction scripts" in application services
Example
// ANTI-PATTERN: Anemic domain model
class Order {
id: string;
customerId: string;
items: OrderItem[];
status: string;
total: number;
// Only data access, no behavior
getId(): string { return this.id; }
setStatus(status: string): void { this.status = status; }
getItems(): OrderItem[] { return this.items; }
setTotal(total: number): void { this.total = total; }
}
class OrderService {
// All logic external to the entity
calculateTotal(order: Order): number {
let total = 0;
for (const item of order.getItems()) {
total += item.price * item.quantity;
}
order.setTotal(total);
return total;
}
canShip(order: Order): boolean {
return order.status === 'PAID' && order.getItems().length > 0;
}
ship(order: Order, trackingNumber: string): void {
if (!this.canShip(order)) {
throw new Error('Cannot ship order');
}
order.setStatus('SHIPPED');
order.trackingNumber = trackingNumber;
}
}
Remediation
// CORRECT: Rich domain model
class Order {
private _id: OrderId;
private _items: OrderItem[];
private _status: OrderStatus;
// Behavior lives in the entity
get total(): Money {
return this._items.reduce(
(sum, item) => sum.add(item.subtotal()),
Money.zero()
);
}
canShip(): boolean {
return this._status === OrderStatus.Paid && this._items.length > 0;
}
ship(trackingNumber: TrackingNumber): void {
if (!this.canShip()) {
throw new OrderNotShippableError(this._id, this._status);
}
this._status = OrderStatus.Shipped;
this._trackingNumber = trackingNumber;
}
addItem(item: OrderItem): void {
this.ensureCanModify();
this._items.push(item);
}
}
// Application service is thin - only orchestration
class OrderApplicationService {
async shipOrder(orderId: OrderId, trackingNumber: TrackingNumber): Promise<void> {
const order = await this.orderRepository.findById(orderId);
order.ship(trackingNumber); // Domain logic in entity
await this.orderRepository.save(order);
}
}
Root Causes
- Developers treating objects as data structures
- Thinking in terms of database tables
- Copying patterns from CRUD applications
- Misunderstanding "service" to mean "all logic goes here"
God Aggregate
Description
An aggregate that has grown to encompass too much. It handles multiple concerns, has many child entities, and becomes a performance and concurrency bottleneck.
Symptoms
- Aggregates with 10+ child entity types
- Long load times due to eager loading everything
- Frequent optimistic concurrency conflicts
- Methods that only touch a small subset of the aggregate
- Difficulty reasoning about invariants
Example
// ANTI-PATTERN: God aggregate
class Customer {
private _id: CustomerId;
private _profile: CustomerProfile;
private _addresses: Address[];
private _paymentMethods: PaymentMethod[];
private _orders: Order[]; // History of all orders!
private _wishlist: WishlistItem[];
private _reviews: Review[];
private _loyaltyPoints: LoyaltyAccount;
private _preferences: Preferences;
private _notifications: Notification[];
private _supportTickets: SupportTicket[];
// Loading this customer loads EVERYTHING
// Updating preferences causes concurrency conflict with order placement
}
Remediation
// CORRECT: Small, focused aggregates
class Customer {
private _id: CustomerId;
private _profile: CustomerProfile;
private _defaultAddressId: AddressId;
private _membershipTier: MembershipTier;
}
class CustomerAddressBook {
private _customerId: CustomerId;
private _addresses: Address[];
}
class ShoppingCart {
private _customerId: CustomerId; // Reference by ID
private _items: CartItem[];
}
class Wishlist {
private _customerId: CustomerId; // Reference by ID
private _items: WishlistItem[];
}
class LoyaltyAccount {
private _customerId: CustomerId; // Reference by ID
private _points: Points;
private _transactions: LoyaltyTransaction[];
}
Identification Heuristic
Ask: "Do all these things need to be immediately consistent?" If the answer is no, they probably belong in separate aggregates.
Aggregate Reference Violation
Description
Aggregates holding direct object references to other aggregates instead of referencing by identity. Creates implicit coupling and makes it impossible to reason about transactional boundaries.
Symptoms
- Navigation from one aggregate to another:
order.customer.address - Loading an aggregate brings in connected aggregates
- Unclear what gets saved when calling
save() - Difficulty implementing eventual consistency
Example
// ANTI-PATTERN: Direct reference
class Order {
private customer: Customer; // Direct reference!
private shippingAddress: Address;
getCustomerEmail(): string {
return this.customer.email; // Navigating through!
}
validate(): void {
// Touching another aggregate's data
if (this.customer.creditLimit < this.total) {
throw new Error('Credit limit exceeded');
}
}
}
Remediation
// CORRECT: Reference by identity
class Order {
private _customerId: CustomerId; // ID only!
private _shippingAddress: Address; // Value object copied at order time
// If customer data is needed, it must be explicitly loaded
static create(
customerId: CustomerId,
shippingAddress: Address,
creditLimit: Money // Passed in, not navigated to
): Order {
return new Order(customerId, shippingAddress, creditLimit);
}
}
// Application service coordinates loading if needed
class OrderApplicationService {
async getOrderWithCustomerDetails(orderId: OrderId): Promise<OrderDetails> {
const order = await this.orderRepository.findById(orderId);
const customer = await this.customerRepository.findById(order.customerId);
return new OrderDetails(order, customer);
}
}
Smart UI
Description
Business logic embedded directly in the user interface layer. Controllers, presenters, or UI components contain domain rules.
Symptoms
- Validation logic in form handlers
- Business calculations in controllers
- State machines in UI components
- Domain rules duplicated across different UI views
- "If we change the UI framework, we lose the business logic"
Example
// ANTI-PATTERN: Smart UI
class OrderController {
submitOrder(request: Request): Response {
const cart = request.body;
// Business logic in controller!
let total = 0;
for (const item of cart.items) {
total += item.price * item.quantity;
}
// Discount rules in controller!
if (cart.items.length > 10) {
total *= 0.9; // 10% bulk discount
}
if (total > 1000 && !this.hasValidPaymentMethod(cart.customerId)) {
return Response.error('Orders over $1000 require verified payment');
}
// More business rules...
const order = {
customerId: cart.customerId,
items: cart.items,
total: total,
status: 'PENDING'
};
this.database.insert('orders', order);
return Response.ok(order);
}
}
Remediation
// CORRECT: UI delegates to domain
class OrderController {
submitOrder(request: Request): Response {
const command = new PlaceOrderCommand(
request.body.customerId,
request.body.items
);
try {
const orderId = this.orderApplicationService.placeOrder(command);
return Response.ok({ orderId });
} catch (error) {
if (error instanceof DomainError) {
return Response.badRequest(error.message);
}
throw error;
}
}
}
// Domain logic in domain layer
class Order {
private calculateTotal(): Money {
const subtotal = this._items.reduce(
(sum, item) => sum.add(item.subtotal()),
Money.zero()
);
return this._discountPolicy.apply(subtotal, this._items.length);
}
}
class BulkDiscountPolicy implements DiscountPolicy {
apply(subtotal: Money, itemCount: number): Money {
if (itemCount > 10) {
return subtotal.multiply(0.9);
}
return subtotal;
}
}
Database-Driven Design
Description
The domain model is derived from the database schema rather than from domain concepts. Tables become classes; foreign keys become object references; database constraints become business rules.
Symptoms
- Class names match table names exactly
- Foreign key relationships drive object graph
- ID fields everywhere, even where identity doesn't matter
nullabledatabase columns drive optional properties- Domain model changes require database migration first
Example
// ANTI-PATTERN: Database-driven model
// Mirrors database schema exactly
class orders {
order_id: number;
customer_id: number;
order_date: Date;
status_cd: string;
shipping_address_id: number;
billing_address_id: number;
total_amt: number;
tax_amt: number;
created_ts: Date;
updated_ts: Date;
}
class order_items {
order_item_id: number;
order_id: number;
product_id: number;
quantity: number;
unit_price: number;
discount_pct: number;
}
Remediation
// CORRECT: Domain-driven model
class Order {
private readonly _id: OrderId;
private _status: OrderStatus;
private _items: OrderItem[];
private _shippingAddress: Address; // Value object, not FK
private _billingAddress: Address;
// Domain behavior, not database structure
get total(): Money {
return this._items.reduce(
(sum, item) => sum.add(item.lineTotal()),
Money.zero()
);
}
ship(trackingNumber: TrackingNumber): void {
// Business logic
}
}
// Mapping is infrastructure concern
class OrderRepository {
async save(order: Order): Promise<void> {
// Map rich domain object to database tables
await this.db.query(
'INSERT INTO orders (id, status, shipping_street, shipping_city...) VALUES (...)'
);
}
}
Key Principle
The domain model reflects how domain experts think, not how data is stored. Persistence is an infrastructure detail.
Leaky Abstractions
Description
Infrastructure concerns bleeding into the domain layer. Domain objects depend on frameworks, databases, or external services.
Symptoms
- Domain entities with ORM decorators
- Repository interfaces returning database-specific types
- Domain services making HTTP calls
- Framework annotations on domain objects
import { Entity } from 'typeorm'in domain layer
Example
// ANTI-PATTERN: Infrastructure leaking into domain
import { Entity, Column, PrimaryColumn, ManyToOne } from 'typeorm';
import { IsEmail, IsNotEmpty } from 'class-validator';
@Entity('customers') // ORM in domain!
export class Customer {
@PrimaryColumn()
id: string;
@Column()
@IsNotEmpty() // Validation framework in domain!
name: string;
@Column()
@IsEmail()
email: string;
@ManyToOne(() => Subscription) // ORM relationship in domain!
subscription: Subscription;
}
// Domain service calling external API directly
class ShippingCostService {
async calculateCost(order: Order): Promise<number> {
// HTTP call in domain!
const response = await fetch('https://shipping-api.com/rates', {
body: JSON.stringify(order)
});
return response.json().cost;
}
}
Remediation
// CORRECT: Clean domain layer
// Domain object - no framework dependencies
class Customer {
private constructor(
private readonly _id: CustomerId,
private readonly _name: CustomerName,
private readonly _email: Email
) {}
static create(name: string, email: string): Customer {
return new Customer(
CustomerId.generate(),
CustomerName.create(name), // Self-validating value object
Email.create(email) // Self-validating value object
);
}
}
// Port (interface) defined in domain
interface ShippingRateProvider {
getRate(destination: Address, weight: Weight): Promise<Money>;
}
// Domain service uses port
class ShippingCostCalculator {
constructor(private rateProvider: ShippingRateProvider) {}
async calculate(order: Order): Promise<Money> {
return this.rateProvider.getRate(
order.shippingAddress,
order.totalWeight()
);
}
}
// Adapter (infrastructure) implements port
class ShippingApiRateProvider implements ShippingRateProvider {
async getRate(destination: Address, weight: Weight): Promise<Money> {
const response = await fetch('https://shipping-api.com/rates', {
body: JSON.stringify({ destination, weight })
});
const data = await response.json();
return Money.of(data.cost, Currency.USD);
}
}
Shared Database
Description
Multiple bounded contexts accessing the same database tables. Changes in one context break others. No clear data ownership.
Symptoms
- Multiple services querying the same tables
- Fear of schema changes because "something else might break"
- Unclear which service is authoritative for data
- Cross-context joins in queries
- Database triggers coordinating contexts
Example
// ANTI-PATTERN: Shared database
// Sales context
class SalesOrderService {
async getOrder(orderId: string) {
return this.db.query(`
SELECT o.*, c.name, c.email, p.name as product_name
FROM orders o
JOIN customers c ON o.customer_id = c.id
JOIN products p ON o.product_id = p.id
WHERE o.id = ?
`, [orderId]);
}
}
// Shipping context - same tables!
class ShippingService {
async getOrdersToShip() {
return this.db.query(`
SELECT o.*, c.address
FROM orders o
JOIN customers c ON o.customer_id = c.id
WHERE o.status = 'PAID'
`);
}
async markShipped(orderId: string) {
// Directly modifying shared table
await this.db.query(
"UPDATE orders SET status = 'SHIPPED' WHERE id = ?",
[orderId]
);
}
}
Remediation
// CORRECT: Each context owns its data
// Sales context - owns order creation
class SalesOrderRepository {
async save(order: SalesOrder): Promise<void> {
await this.salesDb.query('INSERT INTO sales_orders...');
// Publish event for other contexts
await this.eventPublisher.publish(
new OrderPlaced(order.id, order.customerId, order.items)
);
}
}
// Shipping context - owns its projection
class ShippingOrderProjection {
// Handles events to build local projection
async handleOrderPlaced(event: OrderPlaced): Promise<void> {
await this.shippingDb.query(`
INSERT INTO shipments (order_id, customer_id, status)
VALUES (?, ?, 'PENDING')
`, [event.orderId, event.customerId]);
}
}
class ShipmentRepository {
async findPendingShipments(): Promise<Shipment[]> {
// Queries only shipping context's data
return this.shippingDb.query(
"SELECT * FROM shipments WHERE status = 'PENDING'"
);
}
}
Premature Abstraction
Description
Creating abstractions, interfaces, and frameworks before understanding the problem space. Often justified as "flexibility for the future."
Symptoms
- Interfaces with single implementations
- Generic frameworks solving hypothetical problems
- Heavy use of design patterns without clear benefit
- Configuration systems for things that never change
- "We might need this someday"
Example
// ANTI-PATTERN: Premature abstraction
interface IOrderProcessor<TOrder, TResult> {
process(order: TOrder): Promise<TResult>;
}
interface IOrderValidator<TOrder> {
validate(order: TOrder): ValidationResult;
}
interface IOrderPersister<TOrder> {
persist(order: TOrder): Promise<void>;
}
abstract class AbstractOrderProcessor<TOrder, TResult>
implements IOrderProcessor<TOrder, TResult> {
constructor(
protected validator: IOrderValidator<TOrder>,
protected persister: IOrderPersister<TOrder>,
protected notifier: INotificationService,
protected logger: ILogger,
protected metrics: IMetricsCollector
) {}
async process(order: TOrder): Promise<TResult> {
this.logger.log('Processing order');
this.metrics.increment('orders.processed');
const validation = this.validator.validate(order);
if (!validation.isValid) {
throw new ValidationException(validation.errors);
}
const result = await this.doProcess(order);
await this.persister.persist(order);
await this.notifier.notify(order);
return result;
}
protected abstract doProcess(order: TOrder): Promise<TResult>;
}
// Only one concrete implementation ever created
class StandardOrderProcessor extends AbstractOrderProcessor<Order, OrderResult> {
protected async doProcess(order: Order): Promise<OrderResult> {
// The actual logic is trivial
return new OrderResult(order.id);
}
}
Remediation
// CORRECT: Concrete first, abstract when patterns emerge
class OrderService {
async placeOrder(command: PlaceOrderCommand): Promise<OrderId> {
const order = Order.create(command);
if (!order.isValid()) {
throw new InvalidOrderError(order.validationErrors());
}
await this.orderRepository.save(order);
return order.id;
}
}
// Only add abstraction when you have multiple implementations
// and understand the variation points
Heuristic
Wait until you have three similar implementations before abstracting. The right abstraction will be obvious then.
Big Ball of Mud
Description
A system without clear architectural boundaries. Everything depends on everything. Changes ripple unpredictably.
Symptoms
- No clear module boundaries
- Circular dependencies
- Any change might break anything
- "Only Bob understands how this works"
- Integration tests are the only reliable tests
- Fear of refactoring
Identification
# Circular dependency example
OrderService → CustomerService → PaymentService → OrderService
Remediation Strategy
- Identify implicit contexts - Find clusters of related functionality
- Define explicit boundaries - Create modules/packages with clear interfaces
- Break cycles - Introduce events or shared kernel for circular dependencies
- Enforce boundaries - Use architectural tests, linting rules
// Step 1: Identify boundaries
// sales/ - order creation, pricing
// fulfillment/ - shipping, tracking
// customer/ - customer management
// shared/ - shared kernel (Money, Address)
// Step 2: Define public interfaces
// sales/index.ts
export { OrderService } from './application/OrderService';
export { OrderPlaced, OrderCancelled } from './domain/events';
// Internal types not exported
// Step 3: Break cycles with events
class OrderService {
async placeOrder(command: PlaceOrderCommand): Promise<OrderId> {
const order = Order.create(command);
await this.orderRepository.save(order);
// Instead of calling PaymentService directly
await this.eventPublisher.publish(new OrderPlaced(order));
return order.id;
}
}
class PaymentEventHandler {
async handleOrderPlaced(event: OrderPlaced): Promise<void> {
await this.paymentService.collectPayment(event.orderId, event.total);
}
}
CRUD-Driven Development
Description
Treating all domain operations as Create, Read, Update, Delete operations. Loses domain intent and behavior.
Symptoms
- Endpoints like
PUT /orders/{id}that accept any field changes - Service methods like
updateOrder(orderId, updates) - Domain events named
OrderUpdatedinstead ofOrderShipped - No validation of state transitions
- Business operations hidden behind generic updates
Example
// ANTI-PATTERN: CRUD-driven
class OrderController {
@Put('/orders/:id')
async updateOrder(id: string, body: Partial<Order>) {
// Any field can be updated!
return this.orderService.update(id, body);
}
}
class OrderService {
async update(id: string, updates: Partial<Order>): Promise<Order> {
const order = await this.repo.findById(id);
Object.assign(order, updates); // Blindly apply updates
return this.repo.save(order);
}
}
Remediation
// CORRECT: Intent-revealing operations
class OrderController {
@Post('/orders/:id/ship')
async shipOrder(id: string, body: ShipOrderRequest) {
return this.orderService.ship(id, body.trackingNumber);
}
@Post('/orders/:id/cancel')
async cancelOrder(id: string, body: CancelOrderRequest) {
return this.orderService.cancel(id, body.reason);
}
}
class OrderService {
async ship(orderId: OrderId, trackingNumber: TrackingNumber): Promise<void> {
const order = await this.repo.findById(orderId);
order.ship(trackingNumber); // Domain logic with validation
await this.repo.save(order);
await this.publish(new OrderShipped(orderId, trackingNumber));
}
async cancel(orderId: OrderId, reason: CancellationReason): Promise<void> {
const order = await this.repo.findById(orderId);
order.cancel(reason); // Validates cancellation is allowed
await this.repo.save(order);
await this.publish(new OrderCancelled(orderId, reason));
}
}
Summary: Detection Checklist
| Anti-Pattern | Key Question |
|---|---|
| Anemic Domain Model | Do entities have behavior or just data? |
| God Aggregate | Does everything need immediate consistency? |
| Aggregate Reference Violation | Are aggregates holding other aggregates? |
| Smart UI | Would changing UI framework lose business logic? |
| Database-Driven Design | Does model match tables or domain concepts? |
| Leaky Abstractions | Does domain code import infrastructure? |
| Shared Database | Do multiple contexts write to same tables? |
| Premature Abstraction | Are there interfaces with single implementations? |
| Big Ball of Mud | Can any change break anything? |
| CRUD-Driven Development | Are operations generic updates or domain intents? |