System Design (LLD)Test Doubles: Mocks, Stubs, Fakes

Test Doubles: Mocks, Stubs, Fakes

LevelIntermediate

Duration75 mins

TopicTest Doubles: Mocks, Stubs, Fakes

1 / 4

Stubs — Pre-Programmed Returns

The Problem of External Dependencies

Every non-trivial software system interacts with the outside world. Classes call databases, services invoke APIs, components query external systems. These interactions are essential for production functionality—but they create a fundamental tension with testing.

The testing dilemma:

When your OrderService class calls a PaymentGateway to process transactions, how do you test the order logic independently? You face two unpalatable options:

Use the real payment gateway — But this is slow, requires network connectivity, costs money per transaction, and introduces non-determinism into your tests
Don't test those code paths — But this leaves critical logic unverified, creating gaps that manifest as production bugs

Neither option is acceptable. What we need is a way to substitute the real dependency with something we control—something that behaves predictably without incurring the costs of the real implementation.

What You Will Learn

By the end of this page, you will understand stubs as the foundational test double for providing pre-programmed responses. You'll learn when stubs are the right choice, how to design them effectively, and the patterns that make stub-based testing maintainable at scale.

Introducing Test Doubles

The term test double comes from the film industry concept of a "stunt double"—a person who substitutes for an actor in scenes requiring special skills or involving risk. In software testing, a test double is any object that substitutes for a real component during testing.

Gerard Meszaros, in his seminal work xUnit Test Patterns, categorized test doubles into five types:

Dummy — Passed around but never actually used (fills parameter lists)
Stub — Provides canned answers to calls made during the test
Spy — A stub that also records information about how it was called
Mock — Pre-programmed with expectations; verifies correct interactions occurred
Fake — A working implementation with shortcuts unsuitable for production

In this module, we focus on three of the most important: stubs, mocks, and fakes. Understanding the distinctions between them—and when to use each—is essential for writing effective, maintainable tests.

Why the Terminology Matters

Many developers use 'mock' as a generic term for any test double. This imprecision causes confusion and leads to suboptimal testing strategies. A stub is fundamentally different from a mock—they serve different purposes and are used in different testing scenarios. Precision in terminology leads to precision in testing.

The fundamental question each double answers:

Test Double	Primary Question
Stub	"What should happen when my code calls this dependency?"
Mock	"Did my code call this dependency correctly?"
Fake	"How do I get realistic behavior without production complexity?"

This page focuses entirely on stubs—understanding their purpose, design, implementation, and best practices.

What Exactly Is a Stub?

A stub is a test double that provides predetermined responses to method calls. When you stub a dependency, you're saying: "When my code under test calls this method, return this specific value—and don't do anything else."

Stubs are fundamentally about controlling inputs to your system under test. They allow you to:

Simulate any state or response from a dependency
Test error handling by returning error responses
Control non-deterministic behavior (time, randomness, external data)
Eliminate slow, expensive, or unavailable dependencies

Key characteristic: Stubs are passive. They respond when called but don't verify anything about the interaction. A stub doesn't care how many times it was called, in what order, or with what arguments (unless that affects the return value).

stub_concept.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
// Real dependency - talks to external payment processor
interface PaymentGateway {
    processPayment(amount: number, currency: string): PaymentResult;
    refund(transactionId: string): RefundResult;
}
 
// Production implementation - expensive, slow, requires network
class StripePaymentGateway implements PaymentGateway {
    processPayment(amount: number, currency: string): PaymentResult {
        // Makes HTTP call to Stripe API
        // Takes 200-500ms
        // Costs money
        // Requires valid credentials
        return stripeClient.charges.create({ amount, currency });
    }
    
    refund(transactionId: string): RefundResult {
        return stripeClient.refunds.create({ charge: transactionId });
    }
}
 
// STUB - provides pre-programmed responses
class StubPaymentGateway implements PaymentGateway {
    private nextPaymentResult: PaymentResult = { 
        success: true, 
        transactionId: "txn_stub_123" 
    };
    
    // Configure what the stub will return
    willReturnPaymentResult(result: PaymentResult): void {
        this.nextPaymentResult = result;
    }
    
    processPayment(amount: number, currency: string): PaymentResult {
        // Ignores arguments (or could use them)
        // Returns exactly what we configured
        // No network call, no cost, instant
        return this.nextPaymentResult;
    }
    
    refund(transactionId: string): RefundResult {
        return { success: true, refundId: "ref_stub_456" };
    }
}

In this example, the StubPaymentGateway implements the same interface as the real implementation but returns pre-configured values instead of calling external services. The stub is:

Fast — No network calls, returns immediately
Deterministic — Always returns exactly what we configured
Free — No API costs or rate limits
Available — Works offline, in CI/CD, anywhere
Controllable — We can simulate success, failure, or any edge case

The Anatomy of a Well-Designed Stub

Not all stubs are created equal. A poorly designed stub can make tests brittle, confusing, and hard to maintain. A well-designed stub enables clear, focused testing.

Principles of stub design:

Essential Stub Design Principles

•Implement the same interface — The stub must be a valid substitution for the real dependency. This means implementing the same interface or extending the same base class.
•Configurable responses — Provide clean APIs to configure what the stub returns. This makes test setup explicit and readable.
•Sensible defaults — Provide reasonable default responses so tests don't need configuration for every method.
•Minimal behavior — Stubs should do as little as possible. No side effects, no complex logic, no state changes beyond what's configured.
•Focused scope — Each stub should represent one dependency. Don't create 'god stubs' that combine multiple roles.

well_designed_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
// ✅ Well-designed stub: Implements interface, configurable, sensible defaults
interface UserRepository {
    findById(id: string): User | null;
    findByEmail(email: string): User | null;
    save(user: User): User;
    delete(id: string): void;
}
 
class StubUserRepository implements UserRepository {
    // Configurable storage for test data
    private users: Map<string, User> = new Map();
    private emailIndex: Map<string, User> = new Map();
    
    // Clear, explicit API for test setup
    givenUserExists(user: User): this {
        this.users.set(user.id, user);
        this.emailIndex.set(user.email, user);
        return this;  // Fluent interface for chaining
    }
    
    givenNoUserExists(): this {
        this.users.clear();
        this.emailIndex.clear();
        return this;
    }
    
    // Interface implementation with minimal behavior
    findById(id: string): User | null {
        return this.users.get(id) ?? null;  // Returns configured data or null
    }
    
    findByEmail(email: string): User | null {
        return this.emailIndex.get(email) ?? null;
    }
    
    save(user: User): User {
        // Minimal behavior: just store and return
        this.users.set(user.id, user);
        this.emailIndex.set(user.email, user);
        return user;
    }
    
    delete(id: string): void {
        const user = this.users.get(id);
        if (user) {
            this.emailIndex.delete(user.email);
        }
        this.users.delete(id);
    }
}
 
// Usage in test
describe('UserService', () => {
    it('should return user profile when user exists', () => {
        // ARRANGE - Configure stub with test data
        const stubRepository = new StubUserRepository()
            .givenUserExists({
                id: 'user-123',
                email: 'alice@example.com',
                name: 'Alice Smith'
            });
        
        const userService = new UserService(stubRepository);
        
        // ACT
        const profile = userService.getProfile('user-123');
        
        // ASSERT
        expect(profile.name).toBe('Alice Smith');
    });
    
    it('should throw when user does not exist', () => {
        // ARRANGE - Stub configured to return no users
        const stubRepository = new StubUserRepository()
            .givenNoUserExists();
        
        const userService = new UserService(stubRepository);
        
        // ACT & ASSERT
        expect(() => userService.getProfile('nonexistent'))
            .toThrow(UserNotFoundException);
    });
});

Notice how the stub:

Uses Given/When/Then naming (givenUserExists) making tests read like specifications
Provides a fluent interface for easy configuration chaining
Has minimal logic — just stores and retrieves data
Makes test intent explicit — the setup clearly communicates what scenario is being tested

When to Use Stubs

Stubs are the right choice when you need to control what your code receives from a dependency, but you don't care about how your code interacts with that dependency. The focus is on testing your code's logic given certain inputs.

Use stubs when:

Ideal Scenarios for Stubs

•Testing conditional logic — You have if/else branches based on dependency responses. Stub different responses to test each branch.
•Testing error handling — You need to verify your code handles errors correctly. Stub error responses that would be hard to trigger naturally.
•Replacing slow dependencies — Database queries, API calls, file I/O. Stubs eliminate latency from test execution.
•Replacing non-deterministic dependencies — Time, random numbers, external data sources. Stubs provide deterministic values.
•Testing edge cases — Unusual or rare conditions that are difficult to reproduce with real dependencies.
•Isolating the unit — When you genuinely want to test your code in isolation from collaborators.

When NOT to Use Stubs

Avoid stubs when you need to verify your code calls a dependency correctly. If the correctness of your system depends on making specific calls with specific arguments, you need a mock (which we'll cover next). Similarly, if you need realistic behavior including persistence and complex logic, consider a fake.

stub_use_cases.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
// USE CASE 1: Testing conditional logic based on dependency response
interface InventoryService {
    checkStock(productId: string): number;
}
 
class StubInventoryService implements InventoryService {
    private stockLevel: number = 0;
    
    givenStockLevel(level: number): this {
        this.stockLevel = level;
        return this;
    }
    
    checkStock(productId: string): number {
        return this.stockLevel;
    }
}
 
describe('OrderProcessor - stock handling', () => {
    it('should process order when stock is available', () => {
        const inventory = new StubInventoryService().givenStockLevel(10);
        const processor = new OrderProcessor(inventory);
        
        const result = processor.processOrder({ productId: 'P1', quantity: 5 });
        
        expect(result.status).toBe('CONFIRMED');
    });
    
    it('should reject order when stock is insufficient', () => {
        const inventory = new StubInventoryService().givenStockLevel(2);
        const processor = new OrderProcessor(inventory);
        
        const result = processor.processOrder({ productId: 'P1', quantity: 5 });
        
        expect(result.status).toBe('REJECTED');
        expect(result.reason).toBe('Insufficient stock');
    });
});
 
// USE CASE 2: Testing error handling
interface ExternalPricingApi {
    getPrice(productId: string): Promise<number>;
}
 
class StubPricingApi implements ExternalPricingApi {
    private priceToReturn: number = 0;
    private shouldThrow: boolean = false;
    private errorToThrow: Error | null = null;
    
    givenPrice(price: number): this {
        this.priceToReturn = price;
        this.shouldThrow = false;
        return this;
    }
    
    givenNetworkError(): this {
        this.shouldThrow = true;
        this.errorToThrow = new Error('Network timeout');
        return this;
    }
    
    async getPrice(productId: string): Promise<number> {
        if (this.shouldThrow) {
            throw this.errorToThrow;
        }
        return this.priceToReturn;
    }
}
 
describe('PricingService - error handling', () => {
    it('should use cached price when API fails', async () => {
        const api = new StubPricingApi().givenNetworkError();
        const cache = new InMemoryCache();
        cache.set('P1', 99.99);
        
        const service = new PricingService(api, cache);
        
        const price = await service.getCurrentPrice('P1');
        
        expect(price).toBe(99.99);  // Falls back to cached value
    });
});
 
// USE CASE 3: Replacing non-deterministic dependencies
interface TimeProvider {
    now(): Date;
}
 
class StubTimeProvider implements TimeProvider {
    private frozenTime: Date = new Date();
    
    givenCurrentTime(time: Date): this {
        this.frozenTime = time;
        return this;
    }
    
    now(): Date {
        return this.frozenTime;
    }
}
 
describe('SubscriptionService - expiration', () => {
    it('should mark subscription as expired after end date', () => {
        const time = new StubTimeProvider()
            .givenCurrentTime(new Date('2024-02-01'));
        
        const subscription = new Subscription({
            startDate: new Date('2024-01-01'),
            endDate: new Date('2024-01-31'),
        });
        
        const service = new SubscriptionService(time);
        
        expect(service.isActive(subscription)).toBe(false);
    });
});

Stub Implementation Patterns

Several patterns have emerged for implementing stubs effectively. Each has its place depending on the complexity of your testing needs.

Pattern 1: The Configurable Stub

The most common pattern—a class with setter methods to configure responses:

configurable_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
// Pattern 1: Configurable Stub
// Best for: Multiple test scenarios with different configurations
 
class ConfigurableEmailService implements EmailService {
    private sendResult: SendResult = { success: true, messageId: 'msg-1' };
    private shouldThrow: boolean = false;
    private capturedEmails: Email[] = [];  // Optional: capture for verification
    
    // Configuration methods
    willSucceed(): this {
        this.sendResult = { success: true, messageId: 'msg-' + Date.now() };
        this.shouldThrow = false;
        return this;
    }
    
    willFail(reason: string): this {
        this.sendResult = { success: false, error: reason };
        this.shouldThrow = false;
        return this;
    }
    
    willThrowError(error: Error): this {
        this.shouldThrow = true;
        return this;
    }
    
    // Interface implementation
    async send(email: Email): Promise<SendResult> {
        this.capturedEmails.push(email);  // Capture for optional inspection
        
        if (this.shouldThrow) {
            throw new Error('Email service unavailable');
        }
        return this.sendResult;
    }
    
    // Helper for tests that need to verify sent emails
    getSentEmails(): Email[] {
        return [...this.capturedEmails];
    }
}

Pattern 2: The Inline Anonymous Stub

For simple, one-off scenarios, you can inline the stub implementation:

inline_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
// Pattern 2: Inline Anonymous Stub
// Best for: Simple, one-off test scenarios
 
describe('DiscountCalculator', () => {
    it('should apply loyalty discount for premium members', () => {
        // Inline stub - no separate class needed
        const membershipService: MembershipService = {
            getMemberTier: (userId: string) => 'PREMIUM',
            getMemberSince: (userId: string) => new Date('2020-01-01'),
        };
        
        const calculator = new DiscountCalculator(membershipService);
        
        const discount = calculator.calculateDiscount('user-123', 100);
        
        expect(discount).toBe(15);  // Premium members get 15%
    });
});
 
// TypeScript's structural typing makes this particularly elegant.
// As long as the object has methods matching the interface, it works.

Pattern 3: The State Machine Stub

For testing workflows where dependency responses change over time:

state_machine_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
// Pattern 3: State Machine Stub
// Best for: Testing sequences/workflows where responses change
 
class StatefulPaymentGateway implements PaymentGateway {
    private responses: Queue<PaymentResult> = new Queue();
    
    // Configure a sequence of responses
    thenReturn(result: PaymentResult): this {
        this.responses.enqueue(result);
        return this;
    }
    
    // Shorthand for common sequences
    firstSucceedsThenFails(): this {
        return this
            .thenReturn({ success: true, transactionId: 'first' })
            .thenReturn({ success: false, error: 'Declined' });
    }
    
    processPayment(amount: number, currency: string): PaymentResult {
        if (this.responses.isEmpty()) {
            throw new Error('Stub exhausted: no more configured responses');
        }
        return this.responses.dequeue();
    }
}
 
describe('RetryablePaymentProcessor', () => {
    it('should retry once on failure then succeed', () => {
        const gateway = new StatefulPaymentGateway()
            .thenReturn({ success: false, error: 'Temporary failure' })
            .thenReturn({ success: true, transactionId: 'txn-456' });
        
        const processor = new RetryablePaymentProcessor(gateway, maxRetries: 2);
        
        const result = processor.processWithRetry(100, 'USD');
        
        expect(result.success).toBe(true);
        expect(result.transactionId).toBe('txn-456');
    });
});

Pattern 4: The Argument-Matching Stub

When the response should vary based on the arguments received:

argument_matching_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
// Pattern 4: Argument-Matching Stub
// Best for: When different inputs should produce different outputs
 
class ArgumentMatchingUserRepository implements UserRepository {
    private usersByIdResponse: Map<string, User | null> = new Map();
    private defaultResponse: User | null = null;
    
    // Configure specific responses for specific inputs
    givenUserForId(id: string, user: User | null): this {
        this.usersByIdResponse.set(id, user);
        return this;
    }
    
    // Configure default when no specific match
    givenDefaultUser(user: User | null): this {
        this.defaultResponse = user;
        return this;
    }
    
    findById(id: string): User | null {
        // Check for specific match first
        if (this.usersByIdResponse.has(id)) {
            return this.usersByIdResponse.get(id) ?? null;
        }
        // Fall back to default
        return this.defaultResponse;
    }
}
 
describe('AuthorizationService', () => {
    it('should authorize admin actions for admin users only', () => {
        const adminUser: User = { id: 'admin-1', role: 'ADMIN', ... };
        const regularUser: User = { id: 'user-1', role: 'USER', ... };
        
        const repository = new ArgumentMatchingUserRepository()
            .givenUserForId('admin-1', adminUser)
            .givenUserForId('user-1', regularUser);
        
        const authService = new AuthorizationService(repository);
        
        expect(authService.canDeleteUser('admin-1')).toBe(true);
        expect(authService.canDeleteUser('user-1')).toBe(false);
    });
});

Using Stubbing Libraries

While hand-written stubs offer complete control, stubbing libraries reduce boilerplate and provide powerful utilities for common scenarios. Most testing frameworks include stubbing capabilities.

Popular stubbing libraries:

Stubbing Libraries by Ecosystem
Language	Library	Key Features
JavaScript/TypeScript	Jest, Sinon.JS, Vitest	Auto-mocking, stub creation, argument matching
Java	Mockito, EasyMock	when().thenReturn() syntax, argument matchers
Python	unittest.mock, pytest-mock	patch decorator, MagicMock
C#	Moq, NSubstitute	Fluent API, return value configuration
Go	testify/mock, gomock	Interface-based mocking

stubbing_with_jest.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
// Modern stubbing with Jest/Vitest
 
import { describe, it, expect, vi } from 'vitest';
 
describe('OrderService with library stubs', () => {
    it('should calculate total with tax from pricing service', async () => {
        // Create a stub from the interface
        const pricingService: PricingService = {
            getBasePrice: vi.fn().mockReturnValue(100),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0),
        };
        
        const orderService = new OrderService(pricingService);
        
        const total = await orderService.calculateTotal('product-1');
        
        expect(total).toBe(108);  // Base price + 8% tax
    });
    
    it('should apply discount when available', async () => {
        const pricingService: PricingService = {
            getBasePrice: vi.fn().mockReturnValue(100),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0.10),  // 10% discount
        };
        
        const orderService = new OrderService(pricingService);
        
        const total = await orderService.calculateTotal('product-1');
        
        // 100 * 0.90 (after 10% discount) * 1.08 (with tax) = 97.20
        expect(total).toBe(97.20);
    });
});
 
// More complex stubbing scenarios with Jest
 
describe('PaymentService with conditional stubs', () => {
    it('should handle different products differently', async () => {
        const pricingService = {
            getBasePrice: vi.fn().mockImplementation((productId: string) => {
                // Return different prices based on product
                const prices: Record<string, number> = {
                    'premium-product': 199.99,
                    'basic-product': 29.99,
                    'free-product': 0,
                };
                return prices[productId] ?? 49.99;  // Default price
            }),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0),
        };
        
        const orderService = new OrderService(pricingService);
        
        expect(await orderService.calculateTotal('premium-product')).toBeCloseTo(215.99);
        expect(await orderService.calculateTotal('basic-product')).toBeCloseTo(32.39);
        expect(await orderService.calculateTotal('free-product')).toBe(0);
    });
});

Hand-Written vs Library Stubs

Hand-written stubs are more verbose but more explicit and easier to debug. Library stubs are concise but can become opaque when complex. For core business logic testing, many teams prefer hand-written stubs for clarity. For simpler scenarios, library stubs reduce boilerplate effectively.

Common Stub Pitfalls

Stubs are powerful but easy to misuse. Recognizing common pitfalls helps you avoid tests that pass incorrectly or break unexpectedly.

Stub Anti-Patterns

•Pitfall 1: Stubs that lie — Stubs returning impossible values that the real implementation would never produce. Your tests pass, but production fails.
•Pitfall 2: Over-stubbing — Stubbing so much that you're testing stub behavior, not production code. If changing the stub changes test outcomes but production wouldn't change, you've over-stubbed.
•Pitfall 3: Stub maintenance burden — When interfaces change, all stubs must be updated. Too many duplicate stubs leads to painful refactoring.
•Pitfall 4: Ignoring stub state — Forgetting to reset stubs between tests, causing test interdependence.
•Pitfall 5: Testing the collaboration — Using stubs when you should be using mocks. If the interaction is what matters, stubs won't catch bugs.

stub_pitfalls.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
// ❌ PITFALL 1: Stub that lies
// This stub returns a scenario the real API would never produce
class LyingPaymentStub implements PaymentGateway {
    processPayment(): PaymentResult {
        return {
            success: true,
            transactionId: 'txn-123',
            amount: -500,  // ❌ Real API never returns negative amounts!
        };
    }
}
 
// Tests pass but code might fail in production when it assumes amount >= 0
 
// ❌ PITFALL 2: Over-stubbing - testing the stub, not the code
describe('over-stubbed test', () => {
    it('returns what the stub returns', () => {
        const stub = {
            calculate: () => 42,  // This IS the test
        };
        const service = new CalculatorService(stub);
        
        expect(service.getResult()).toBe(42);  // Testing stub, not service logic
    });
});
 
// ✅ BETTER: Stub provides input, test verifies actual logic
describe('properly stubbed test', () => {
    it('doubles the calculated value', () => {
        const stub = {
            calculate: () => 21,  // Stub provides raw input
        };
        const service = new DoublerService(stub);
        
        expect(service.getDoubledResult()).toBe(42);  // Tests actual doubling logic
    });
});
 
// ❌ PITFALL 3: Stub maintenance burden
// When interface changes, ALL these must update:
class StubUserRepoV1 implements UserRepository { ... }
class StubUserRepoV2 implements UserRepository { ... }
class TestableUserRepoStub implements UserRepository { ... }
class AnotherModuleUserStub implements UserRepository { ... }
 
// ✅ BETTER: Single shared stub, configured per-test
class SharedUserRepositoryStub implements UserRepository {
    // One stub, many configurations
}
 
// ❌ PITFALL 4: Stub state leaking between tests
let sharedStub = new ConfigurableStub();  // Danger: shared state
 
describe('test suite with state leak', () => {
    it('test 1 configures stub', () => {
        sharedStub.willReturn('value-1');
        // ... test ...
    });
    
    it('test 2 assumes default', () => {
        // This test might fail or pass depending on test order!
        // sharedStub still has test 1's configuration
    });
});
 
// ✅ BETTER: Fresh stub per test
describe('isolated tests', () => {
    let stub: ConfigurableStub;
    
    beforeEach(() => {
        stub = new ConfigurableStub();  // Fresh instance each time
    });
    
    it('test 1', () => { ... });
    it('test 2', () => { ... });  // Guaranteed isolated
});

Summary: Mastering Stubs

We've comprehensively covered stubs—the foundational test double for controlling what your code receives from dependencies. Let's consolidate the key insights:

Key Takeaways

•Stubs provide pre-programmed responses — They control inputs to your system under test, enabling deterministic, fast, isolated testing.
•Stubs are passive — Unlike mocks, stubs don't verify interactions. They simply return configured values when called.
•Use stubs for controlling inputs — When testing conditional logic, error handling, edge cases, or eliminating slow/non-deterministic dependencies.
•Design stubs for clarity — Fluent interfaces, sensible defaults, and 'Given' naming make tests readable and maintainable.
•Multiple patterns exist — Configurable stubs, inline stubs, state machine stubs, and argument-matching stubs each serve different needs.
•Avoid stub pitfalls — Don't create stubs that lie, don't over-stub, and ensure test isolation through fresh stub instances.

What's next:

Stubs answer the question "What should my code receive from this dependency?" But sometimes we need to answer a different question: "Did my code call this dependency correctly?"

In the next page, we explore mocks—test doubles that verify interactions. Mocks are essential when the correctness of your system depends not just on producing correct outputs, but on making correct calls to its collaborators.

Page Complete

You now have a comprehensive understanding of stubs—their purpose, design, implementation patterns, and pitfalls. You can confidently use stubs to control dependency behavior in your tests, enabling isolated, deterministic verification of your code's logic.

1 / 4

Loading learning content...

System Design (LLD)Test Doubles: Mocks, Stubs, Fakes

Test Doubles: Mocks, Stubs, Fakes

LevelIntermediate

Duration75 mins

TopicTest Doubles: Mocks, Stubs, Fakes

1 / 4

Stubs — Pre-Programmed Returns

The Problem of External Dependencies

The testing dilemma:

When your OrderService class calls a PaymentGateway to process transactions, how do you test the order logic independently? You face two unpalatable options:

Use the real payment gateway — But this is slow, requires network connectivity, costs money per transaction, and introduces non-determinism into your tests
Don't test those code paths — But this leaves critical logic unverified, creating gaps that manifest as production bugs

What You Will Learn

Introducing Test Doubles

Gerard Meszaros, in his seminal work xUnit Test Patterns, categorized test doubles into five types:

Dummy — Passed around but never actually used (fills parameter lists)
Stub — Provides canned answers to calls made during the test
Spy — A stub that also records information about how it was called
Mock — Pre-programmed with expectations; verifies correct interactions occurred
Fake — A working implementation with shortcuts unsuitable for production

Why the Terminology Matters

The fundamental question each double answers:

Test Double	Primary Question
Stub	"What should happen when my code calls this dependency?"
Mock	"Did my code call this dependency correctly?"
Fake	"How do I get realistic behavior without production complexity?"

This page focuses entirely on stubs—understanding their purpose, design, implementation, and best practices.

What Exactly Is a Stub?

Stubs are fundamentally about controlling inputs to your system under test. They allow you to:

Simulate any state or response from a dependency
Test error handling by returning error responses
Control non-deterministic behavior (time, randomness, external data)
Eliminate slow, expensive, or unavailable dependencies

stub_concept.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
// Real dependency - talks to external payment processor
interface PaymentGateway {
    processPayment(amount: number, currency: string): PaymentResult;
    refund(transactionId: string): RefundResult;
}
 
// Production implementation - expensive, slow, requires network
class StripePaymentGateway implements PaymentGateway {
    processPayment(amount: number, currency: string): PaymentResult {
        // Makes HTTP call to Stripe API
        // Takes 200-500ms
        // Costs money
        // Requires valid credentials
        return stripeClient.charges.create({ amount, currency });
    }
    
    refund(transactionId: string): RefundResult {
        return stripeClient.refunds.create({ charge: transactionId });
    }
}
 
// STUB - provides pre-programmed responses
class StubPaymentGateway implements PaymentGateway {
    private nextPaymentResult: PaymentResult = { 
        success: true, 
        transactionId: "txn_stub_123" 
    };
    
    // Configure what the stub will return
    willReturnPaymentResult(result: PaymentResult): void {
        this.nextPaymentResult = result;
    }
    
    processPayment(amount: number, currency: string): PaymentResult {
        // Ignores arguments (or could use them)
        // Returns exactly what we configured
        // No network call, no cost, instant
        return this.nextPaymentResult;
    }
    
    refund(transactionId: string): RefundResult {
        return { success: true, refundId: "ref_stub_456" };
    }
}

In this example, the StubPaymentGateway implements the same interface as the real implementation but returns pre-configured values instead of calling external services. The stub is:

Fast — No network calls, returns immediately
Deterministic — Always returns exactly what we configured
Free — No API costs or rate limits
Available — Works offline, in CI/CD, anywhere
Controllable — We can simulate success, failure, or any edge case

The Anatomy of a Well-Designed Stub

Not all stubs are created equal. A poorly designed stub can make tests brittle, confusing, and hard to maintain. A well-designed stub enables clear, focused testing.

Principles of stub design:

Essential Stub Design Principles

•Implement the same interface — The stub must be a valid substitution for the real dependency. This means implementing the same interface or extending the same base class.
•Configurable responses — Provide clean APIs to configure what the stub returns. This makes test setup explicit and readable.
•Sensible defaults — Provide reasonable default responses so tests don't need configuration for every method.
•Minimal behavior — Stubs should do as little as possible. No side effects, no complex logic, no state changes beyond what's configured.
•Focused scope — Each stub should represent one dependency. Don't create 'god stubs' that combine multiple roles.

well_designed_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
// ✅ Well-designed stub: Implements interface, configurable, sensible defaults
interface UserRepository {
    findById(id: string): User | null;
    findByEmail(email: string): User | null;
    save(user: User): User;
    delete(id: string): void;
}
 
class StubUserRepository implements UserRepository {
    // Configurable storage for test data
    private users: Map<string, User> = new Map();
    private emailIndex: Map<string, User> = new Map();
    
    // Clear, explicit API for test setup
    givenUserExists(user: User): this {
        this.users.set(user.id, user);
        this.emailIndex.set(user.email, user);
        return this;  // Fluent interface for chaining
    }
    
    givenNoUserExists(): this {
        this.users.clear();
        this.emailIndex.clear();
        return this;
    }
    
    // Interface implementation with minimal behavior
    findById(id: string): User | null {
        return this.users.get(id) ?? null;  // Returns configured data or null
    }
    
    findByEmail(email: string): User | null {
        return this.emailIndex.get(email) ?? null;
    }
    
    save(user: User): User {
        // Minimal behavior: just store and return
        this.users.set(user.id, user);
        this.emailIndex.set(user.email, user);
        return user;
    }
    
    delete(id: string): void {
        const user = this.users.get(id);
        if (user) {
            this.emailIndex.delete(user.email);
        }
        this.users.delete(id);
    }
}
 
// Usage in test
describe('UserService', () => {
    it('should return user profile when user exists', () => {
        // ARRANGE - Configure stub with test data
        const stubRepository = new StubUserRepository()
            .givenUserExists({
                id: 'user-123',
                email: 'alice@example.com',
                name: 'Alice Smith'
            });
        
        const userService = new UserService(stubRepository);
        
        // ACT
        const profile = userService.getProfile('user-123');
        
        // ASSERT
        expect(profile.name).toBe('Alice Smith');
    });
    
    it('should throw when user does not exist', () => {
        // ARRANGE - Stub configured to return no users
        const stubRepository = new StubUserRepository()
            .givenNoUserExists();
        
        const userService = new UserService(stubRepository);
        
        // ACT & ASSERT
        expect(() => userService.getProfile('nonexistent'))
            .toThrow(UserNotFoundException);
    });
});

Notice how the stub:

Uses Given/When/Then naming (givenUserExists) making tests read like specifications
Provides a fluent interface for easy configuration chaining
Has minimal logic — just stores and retrieves data
Makes test intent explicit — the setup clearly communicates what scenario is being tested

When to Use Stubs

Use stubs when:

Ideal Scenarios for Stubs

•Testing conditional logic — You have if/else branches based on dependency responses. Stub different responses to test each branch.
•Testing error handling — You need to verify your code handles errors correctly. Stub error responses that would be hard to trigger naturally.
•Replacing slow dependencies — Database queries, API calls, file I/O. Stubs eliminate latency from test execution.
•Replacing non-deterministic dependencies — Time, random numbers, external data sources. Stubs provide deterministic values.
•Testing edge cases — Unusual or rare conditions that are difficult to reproduce with real dependencies.
•Isolating the unit — When you genuinely want to test your code in isolation from collaborators.

When NOT to Use Stubs

stub_use_cases.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
// USE CASE 1: Testing conditional logic based on dependency response
interface InventoryService {
    checkStock(productId: string): number;
}
 
class StubInventoryService implements InventoryService {
    private stockLevel: number = 0;
    
    givenStockLevel(level: number): this {
        this.stockLevel = level;
        return this;
    }
    
    checkStock(productId: string): number {
        return this.stockLevel;
    }
}
 
describe('OrderProcessor - stock handling', () => {
    it('should process order when stock is available', () => {
        const inventory = new StubInventoryService().givenStockLevel(10);
        const processor = new OrderProcessor(inventory);
        
        const result = processor.processOrder({ productId: 'P1', quantity: 5 });
        
        expect(result.status).toBe('CONFIRMED');
    });
    
    it('should reject order when stock is insufficient', () => {
        const inventory = new StubInventoryService().givenStockLevel(2);
        const processor = new OrderProcessor(inventory);
        
        const result = processor.processOrder({ productId: 'P1', quantity: 5 });
        
        expect(result.status).toBe('REJECTED');
        expect(result.reason).toBe('Insufficient stock');
    });
});
 
// USE CASE 2: Testing error handling
interface ExternalPricingApi {
    getPrice(productId: string): Promise<number>;
}
 
class StubPricingApi implements ExternalPricingApi {
    private priceToReturn: number = 0;
    private shouldThrow: boolean = false;
    private errorToThrow: Error | null = null;
    
    givenPrice(price: number): this {
        this.priceToReturn = price;
        this.shouldThrow = false;
        return this;
    }
    
    givenNetworkError(): this {
        this.shouldThrow = true;
        this.errorToThrow = new Error('Network timeout');
        return this;
    }
    
    async getPrice(productId: string): Promise<number> {
        if (this.shouldThrow) {
            throw this.errorToThrow;
        }
        return this.priceToReturn;
    }
}
 
describe('PricingService - error handling', () => {
    it('should use cached price when API fails', async () => {
        const api = new StubPricingApi().givenNetworkError();
        const cache = new InMemoryCache();
        cache.set('P1', 99.99);
        
        const service = new PricingService(api, cache);
        
        const price = await service.getCurrentPrice('P1');
        
        expect(price).toBe(99.99);  // Falls back to cached value
    });
});
 
// USE CASE 3: Replacing non-deterministic dependencies
interface TimeProvider {
    now(): Date;
}
 
class StubTimeProvider implements TimeProvider {
    private frozenTime: Date = new Date();
    
    givenCurrentTime(time: Date): this {
        this.frozenTime = time;
        return this;
    }
    
    now(): Date {
        return this.frozenTime;
    }
}
 
describe('SubscriptionService - expiration', () => {
    it('should mark subscription as expired after end date', () => {
        const time = new StubTimeProvider()
            .givenCurrentTime(new Date('2024-02-01'));
        
        const subscription = new Subscription({
            startDate: new Date('2024-01-01'),
            endDate: new Date('2024-01-31'),
        });
        
        const service = new SubscriptionService(time);
        
        expect(service.isActive(subscription)).toBe(false);
    });
});

Stub Implementation Patterns

Several patterns have emerged for implementing stubs effectively. Each has its place depending on the complexity of your testing needs.

Pattern 1: The Configurable Stub

The most common pattern—a class with setter methods to configure responses:

configurable_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
// Pattern 1: Configurable Stub
// Best for: Multiple test scenarios with different configurations
 
class ConfigurableEmailService implements EmailService {
    private sendResult: SendResult = { success: true, messageId: 'msg-1' };
    private shouldThrow: boolean = false;
    private capturedEmails: Email[] = [];  // Optional: capture for verification
    
    // Configuration methods
    willSucceed(): this {
        this.sendResult = { success: true, messageId: 'msg-' + Date.now() };
        this.shouldThrow = false;
        return this;
    }
    
    willFail(reason: string): this {
        this.sendResult = { success: false, error: reason };
        this.shouldThrow = false;
        return this;
    }
    
    willThrowError(error: Error): this {
        this.shouldThrow = true;
        return this;
    }
    
    // Interface implementation
    async send(email: Email): Promise<SendResult> {
        this.capturedEmails.push(email);  // Capture for optional inspection
        
        if (this.shouldThrow) {
            throw new Error('Email service unavailable');
        }
        return this.sendResult;
    }
    
    // Helper for tests that need to verify sent emails
    getSentEmails(): Email[] {
        return [...this.capturedEmails];
    }
}

Pattern 2: The Inline Anonymous Stub

For simple, one-off scenarios, you can inline the stub implementation:

inline_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
// Pattern 2: Inline Anonymous Stub
// Best for: Simple, one-off test scenarios
 
describe('DiscountCalculator', () => {
    it('should apply loyalty discount for premium members', () => {
        // Inline stub - no separate class needed
        const membershipService: MembershipService = {
            getMemberTier: (userId: string) => 'PREMIUM',
            getMemberSince: (userId: string) => new Date('2020-01-01'),
        };
        
        const calculator = new DiscountCalculator(membershipService);
        
        const discount = calculator.calculateDiscount('user-123', 100);
        
        expect(discount).toBe(15);  // Premium members get 15%
    });
});
 
// TypeScript's structural typing makes this particularly elegant.
// As long as the object has methods matching the interface, it works.

Pattern 3: The State Machine Stub

For testing workflows where dependency responses change over time:

state_machine_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
// Pattern 3: State Machine Stub
// Best for: Testing sequences/workflows where responses change
 
class StatefulPaymentGateway implements PaymentGateway {
    private responses: Queue<PaymentResult> = new Queue();
    
    // Configure a sequence of responses
    thenReturn(result: PaymentResult): this {
        this.responses.enqueue(result);
        return this;
    }
    
    // Shorthand for common sequences
    firstSucceedsThenFails(): this {
        return this
            .thenReturn({ success: true, transactionId: 'first' })
            .thenReturn({ success: false, error: 'Declined' });
    }
    
    processPayment(amount: number, currency: string): PaymentResult {
        if (this.responses.isEmpty()) {
            throw new Error('Stub exhausted: no more configured responses');
        }
        return this.responses.dequeue();
    }
}
 
describe('RetryablePaymentProcessor', () => {
    it('should retry once on failure then succeed', () => {
        const gateway = new StatefulPaymentGateway()
            .thenReturn({ success: false, error: 'Temporary failure' })
            .thenReturn({ success: true, transactionId: 'txn-456' });
        
        const processor = new RetryablePaymentProcessor(gateway, maxRetries: 2);
        
        const result = processor.processWithRetry(100, 'USD');
        
        expect(result.success).toBe(true);
        expect(result.transactionId).toBe('txn-456');
    });
});

Pattern 4: The Argument-Matching Stub

When the response should vary based on the arguments received:

argument_matching_stub.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
// Pattern 4: Argument-Matching Stub
// Best for: When different inputs should produce different outputs
 
class ArgumentMatchingUserRepository implements UserRepository {
    private usersByIdResponse: Map<string, User | null> = new Map();
    private defaultResponse: User | null = null;
    
    // Configure specific responses for specific inputs
    givenUserForId(id: string, user: User | null): this {
        this.usersByIdResponse.set(id, user);
        return this;
    }
    
    // Configure default when no specific match
    givenDefaultUser(user: User | null): this {
        this.defaultResponse = user;
        return this;
    }
    
    findById(id: string): User | null {
        // Check for specific match first
        if (this.usersByIdResponse.has(id)) {
            return this.usersByIdResponse.get(id) ?? null;
        }
        // Fall back to default
        return this.defaultResponse;
    }
}
 
describe('AuthorizationService', () => {
    it('should authorize admin actions for admin users only', () => {
        const adminUser: User = { id: 'admin-1', role: 'ADMIN', ... };
        const regularUser: User = { id: 'user-1', role: 'USER', ... };
        
        const repository = new ArgumentMatchingUserRepository()
            .givenUserForId('admin-1', adminUser)
            .givenUserForId('user-1', regularUser);
        
        const authService = new AuthorizationService(repository);
        
        expect(authService.canDeleteUser('admin-1')).toBe(true);
        expect(authService.canDeleteUser('user-1')).toBe(false);
    });
});

Using Stubbing Libraries

While hand-written stubs offer complete control, stubbing libraries reduce boilerplate and provide powerful utilities for common scenarios. Most testing frameworks include stubbing capabilities.

Popular stubbing libraries:

Stubbing Libraries by Ecosystem
Language	Library	Key Features
JavaScript/TypeScript	Jest, Sinon.JS, Vitest	Auto-mocking, stub creation, argument matching
Java	Mockito, EasyMock	when().thenReturn() syntax, argument matchers
Python	unittest.mock, pytest-mock	patch decorator, MagicMock
C#	Moq, NSubstitute	Fluent API, return value configuration
Go	testify/mock, gomock	Interface-based mocking

stubbing_with_jest.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
// Modern stubbing with Jest/Vitest
 
import { describe, it, expect, vi } from 'vitest';
 
describe('OrderService with library stubs', () => {
    it('should calculate total with tax from pricing service', async () => {
        // Create a stub from the interface
        const pricingService: PricingService = {
            getBasePrice: vi.fn().mockReturnValue(100),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0),
        };
        
        const orderService = new OrderService(pricingService);
        
        const total = await orderService.calculateTotal('product-1');
        
        expect(total).toBe(108);  // Base price + 8% tax
    });
    
    it('should apply discount when available', async () => {
        const pricingService: PricingService = {
            getBasePrice: vi.fn().mockReturnValue(100),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0.10),  // 10% discount
        };
        
        const orderService = new OrderService(pricingService);
        
        const total = await orderService.calculateTotal('product-1');
        
        // 100 * 0.90 (after 10% discount) * 1.08 (with tax) = 97.20
        expect(total).toBe(97.20);
    });
});
 
// More complex stubbing scenarios with Jest
 
describe('PaymentService with conditional stubs', () => {
    it('should handle different products differently', async () => {
        const pricingService = {
            getBasePrice: vi.fn().mockImplementation((productId: string) => {
                // Return different prices based on product
                const prices: Record<string, number> = {
                    'premium-product': 199.99,
                    'basic-product': 29.99,
                    'free-product': 0,
                };
                return prices[productId] ?? 49.99;  // Default price
            }),
            getTaxRate: vi.fn().mockReturnValue(0.08),
            getDiscountRate: vi.fn().mockReturnValue(0),
        };
        
        const orderService = new OrderService(pricingService);
        
        expect(await orderService.calculateTotal('premium-product')).toBeCloseTo(215.99);
        expect(await orderService.calculateTotal('basic-product')).toBeCloseTo(32.39);
        expect(await orderService.calculateTotal('free-product')).toBe(0);
    });
});

Hand-Written vs Library Stubs

Common Stub Pitfalls

Stubs are powerful but easy to misuse. Recognizing common pitfalls helps you avoid tests that pass incorrectly or break unexpectedly.

Stub Anti-Patterns

•Pitfall 1: Stubs that lie — Stubs returning impossible values that the real implementation would never produce. Your tests pass, but production fails.
•Pitfall 2: Over-stubbing — Stubbing so much that you're testing stub behavior, not production code. If changing the stub changes test outcomes but production wouldn't change, you've over-stubbed.
•Pitfall 3: Stub maintenance burden — When interfaces change, all stubs must be updated. Too many duplicate stubs leads to painful refactoring.
•Pitfall 4: Ignoring stub state — Forgetting to reset stubs between tests, causing test interdependence.
•Pitfall 5: Testing the collaboration — Using stubs when you should be using mocks. If the interaction is what matters, stubs won't catch bugs.

stub_pitfalls.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
// ❌ PITFALL 1: Stub that lies
// This stub returns a scenario the real API would never produce
class LyingPaymentStub implements PaymentGateway {
    processPayment(): PaymentResult {
        return {
            success: true,
            transactionId: 'txn-123',
            amount: -500,  // ❌ Real API never returns negative amounts!
        };
    }
}
 
// Tests pass but code might fail in production when it assumes amount >= 0
 
// ❌ PITFALL 2: Over-stubbing - testing the stub, not the code
describe('over-stubbed test', () => {
    it('returns what the stub returns', () => {
        const stub = {
            calculate: () => 42,  // This IS the test
        };
        const service = new CalculatorService(stub);
        
        expect(service.getResult()).toBe(42);  // Testing stub, not service logic
    });
});
 
// ✅ BETTER: Stub provides input, test verifies actual logic
describe('properly stubbed test', () => {
    it('doubles the calculated value', () => {
        const stub = {
            calculate: () => 21,  // Stub provides raw input
        };
        const service = new DoublerService(stub);
        
        expect(service.getDoubledResult()).toBe(42);  // Tests actual doubling logic
    });
});
 
// ❌ PITFALL 3: Stub maintenance burden
// When interface changes, ALL these must update:
class StubUserRepoV1 implements UserRepository { ... }
class StubUserRepoV2 implements UserRepository { ... }
class TestableUserRepoStub implements UserRepository { ... }
class AnotherModuleUserStub implements UserRepository { ... }
 
// ✅ BETTER: Single shared stub, configured per-test
class SharedUserRepositoryStub implements UserRepository {
    // One stub, many configurations
}
 
// ❌ PITFALL 4: Stub state leaking between tests
let sharedStub = new ConfigurableStub();  // Danger: shared state
 
describe('test suite with state leak', () => {
    it('test 1 configures stub', () => {
        sharedStub.willReturn('value-1');
        // ... test ...
    });
    
    it('test 2 assumes default', () => {
        // This test might fail or pass depending on test order!
        // sharedStub still has test 1's configuration
    });
});
 
// ✅ BETTER: Fresh stub per test
describe('isolated tests', () => {
    let stub: ConfigurableStub;
    
    beforeEach(() => {
        stub = new ConfigurableStub();  // Fresh instance each time
    });
    
    it('test 1', () => { ... });
    it('test 2', () => { ... });  // Guaranteed isolated
});

Summary: Mastering Stubs

We've comprehensively covered stubs—the foundational test double for controlling what your code receives from dependencies. Let's consolidate the key insights:

Key Takeaways

•Stubs provide pre-programmed responses — They control inputs to your system under test, enabling deterministic, fast, isolated testing.
•Stubs are passive — Unlike mocks, stubs don't verify interactions. They simply return configured values when called.
•Use stubs for controlling inputs — When testing conditional logic, error handling, edge cases, or eliminating slow/non-deterministic dependencies.
•Design stubs for clarity — Fluent interfaces, sensible defaults, and 'Given' naming make tests readable and maintainable.
•Multiple patterns exist — Configurable stubs, inline stubs, state machine stubs, and argument-matching stubs each serve different needs.
•Avoid stub pitfalls — Don't create stubs that lie, don't over-stub, and ensure test isolation through fresh stub instances.

What's next:

Stubs answer the question "What should my code receive from this dependency?" But sometimes we need to answer a different question: "Did my code call this dependency correctly?"

Page Complete

1 / 4