Laravel Testing

Philosophy

Testing should be:

• Isolated - Test one thing at a time
• Reliable - Consistent results every time
• Maintainable - Easy to update when code changes
• Fast - Quick feedback loop
• Realistic - Use factories, not hardcoded values

The Triple-A Pattern

Every test should follow the Arrange-Act-Assert pattern:

1. Arrange the World

Set up all the data and dependencies needed using factories:

it('creates an order with items', function () {
    // Arrange: Create the world state
    $user = User::factory()->create();
    $product = Product::factory()->active()->create(['price' => 1000]);

    $data = CreateOrderData::from([
        'customer_email' => '[email protected]',
        'items' => [
            ['product_id' => $product->id, 'quantity' => 2],
        ],
    ]);

    // Act: Perform the operation
    $order = resolve(CreateOrderAction::class)($user, $data);

    // Assert: Verify the results
    expect($order)
        ->toBeInstanceOf(Order::class)
        ->and($order->items)->toHaveCount(1)
        ->and($order->total)->toBe(2000);
});

2. Act on the World

Perform the single operation you're testing:

// ✅ Good - Single, clear action
$order = resolve(CreateOrderAction::class)($user, $data);

// ❌ Bad - Multiple actions mixed with assertions
$order = resolve(CreateOrderAction::class)($user, $data);
expect($order)->toBeInstanceOf(Order::class);
$order->refresh();
expect($order->total)->toBe(2000);

3. Assert on the Results

Verify the outcomes of your action:

// ✅ Good - Clear, focused assertions
expect($order)
    ->toBeInstanceOf(Order::class)
    ->and($order->status)->toBe(OrderStatus::Pending)
    ->and($order->items)->toHaveCount(2);

assertDatabaseHas('orders', [
    'id' => $order->id,
    'user_id' => $user->id,
]);

// ❌ Bad - Testing implementation details
expect($order->getAttribute('status'))->toBe('pending');

Testing Actions

Actions are the heart of your domain logic and should be thoroughly tested in isolation.

Basic Action Test

use App\Actions\Order\CreateOrderAction;
use App\Data\CreateOrderData;
use App\Enums\OrderStatus;
use App\Models\User;
use function Pest\Laravel\assertDatabaseHas;

it('creates an order', function () {
    // Arrange
    $user = User::factory()->create();
    $data = CreateOrderData::testFactory()->make([
        'status' => OrderStatus::Pending,
    ]);

    // Act
    $order = resolve(CreateOrderAction::class)($user, $data);

    // Assert
    expect($order)->toBeInstanceOf(Order::class);
    assertDatabaseHas('orders', [
        'id' => $order->id,
        'user_id' => $user->id,
        'status' => OrderStatus::Pending->value,
    ]);
});

Testing Action Guard Methods

it('throws exception when user has too many pending orders', function () {
    // Arrange
    $user = User::factory()
        ->has(Order::factory()->pending()->count(5))
        ->create();

    $data = CreateOrderData::testFactory()->make();

    // Act & Assert
    expect(fn () => resolve(CreateOrderAction::class)($user, $data))
        ->toThrow(OrderException::class, 'Too many pending orders');
});

Testing Action Composition

Critical pattern: Always resolve actions from the container using resolve() so dependencies are recursively resolved. Use swap() to replace dependencies with mocked versions.

use function Pest\Laravel\mock;
use function Pest\Laravel\swap;

it('processes order and sends notification', function () {
    // Arrange
    $user = User::factory()->create();
    $order = Order::factory()->for($user)->create();

    // Mock the dependency actions and swap them into the container
    $calculateTotal = mock(CalculateOrderTotalAction::class);
    $calculateTotal->shouldReceive('__invoke')
        ->once()
        ->with($order)
        ->andReturn(10000);
    swap(CalculateOrderTotalAction::class, $calculateTotal);

    $notifyOrder = mock(NotifyOrderCreatedAction::class);
    $notifyOrder->shouldReceive('__invoke')
        ->once()
        ->with($order);
    swap(NotifyOrderCreatedAction::class, $notifyOrder);

    // Act - resolve() from container so mocked dependencies are injected
    $result = resolve(ProcessOrderAction::class)($order);

    // Assert
    expect($result->total)->toBe(10000);
});

Why this pattern:

• resolve() ensures the action is pulled from the container with all dependencies
• swap() replaces the dependency in the container with your mock
• Container handles recursive dependency resolution automatically
• If a dependency adds a new dependency, your tests don't break

Mocking Guidelines

Only Mock What You Own

Critical principle: Only mock code that you control. Never mock external services directly.

✅ Good - Mock Your Own Actions

use function Pest\Laravel\mock;
use function Pest\Laravel\swap;

// Mock an action you own and swap it into the container
$sendEmail = mock(SendWelcomeEmailAction::class);
$sendEmail->shouldReceive('__invoke')
    ->once()
    ->with(Mockery::type(User::class));
swap(SendWelcomeEmailAction::class, $sendEmail);

// Then resolve the action under test - it will receive the mocked dependency
$result = resolve(RegisterUserAction::class)($data);

✅ Advanced - Verify Mock Arguments with Assertions

Use withArgs() with a closure to verify the exact instances and values being passed:

it('processes match with correct arguments', function () {
    $matchAttempt = MatchAttempt::factory()->create();
    $data = MatchData::testFactory()->make();

    // Mock and verify exact arguments using expect() assertions
    $mockAction = mock(CreateMatchResultAction::class);
    $mockAction->shouldReceive('__invoke')
        ->once()
        ->withArgs(function (MatchAttempt $_matchAttempt, MatchData $_data) use ($data, $matchAttempt) {
            // Verify the exact model instance is passed
            expect($_matchAttempt->is($matchAttempt))->toBeTrue()
                // Verify the exact DTO value is passed
                ->and($_data)->toBe($data->matches->first());

            return true; // Return true to pass the assertion
        });
    swap(CreateMatchResultAction::class, $mockAction);

    // Act
    resolve(ProcessMatchAction::class)($matchAttempt, $data);
});

✅ Good - Mock Your Own Services (via Facade)

// Mock your own service through its facade
Payment::shouldReceive('createPaymentIntent')
    ->once()
    ->with(10000, 'usd')
    ->andReturn(PaymentIntentData::from([
        'id' => 'pi_test_123',
        'status' => 'succeeded',
    ]));

❌ Bad - Mocking External Libraries Directly

// ❌ DON'T DO THIS - Mocking Stripe SDK directly
$stripe = Mockery::mock(\Stripe\StripeClient::class);
$stripe->shouldReceive('paymentIntents->create')
    ->andReturn(/* ... */);

// This is brittle and breaks when Stripe updates their SDK

When You Need to Mock Something You Don't Own

If you find yourself needing to mock an external service, create an abstraction:

1. Create a Service Layer with the Manager pattern
2. Define a Driver Contract (interface)
3. Implement the Real Driver (wraps external API)
4. Create a Null Driver for testing
5. Add a Facade for convenience

See Services for complete implementation examples.

Using Null Drivers

The null driver pattern provides deterministic, fast tests without external dependencies:

it('processes payment successfully', function () {
    // Arrange - Use null driver (configured in phpunit.xml or .env.testing)
    Config::set('payment.default', 'null');

    $order = Order::factory()->create(['total' => 10000]);
    $data = PaymentData::from(['amount' => 10000, 'currency' => 'usd']);

    // Act - No mocking needed, null driver returns test data
    $payment = resolve(ProcessPaymentAction::class)($order, $data);

    // Assert
    expect($payment)
        ->toBeInstanceOf(Payment::class)
        ->and($payment->status)->toBe(PaymentStatus::Completed);
});

Benefits of null drivers:

• No mocking required
• Fast execution (no network calls)
• Deterministic results
• Can test error scenarios by extending null driver
• Matches real driver interface exactly

Testing Error Scenarios

Extend the null driver for specific test scenarios:

// tests/Fakes/FailingPaymentDriver.php
class FailingPaymentDriver implements PaymentDriver
{
    public function createPaymentIntent(int $amount, string $currency): PaymentIntentData
    {
        throw PaymentException::failedToCharge('Card declined');
    }
}

// In test
it('handles payment failure gracefully', function () {
    $this->app->bind(PaymentManager::class, function () {
        $manager = new PaymentManager($this->app);
        $manager->extend('failing', fn () => new FailingPaymentDriver);
        return $manager;
    });

    Config::set('payment.default', 'failing');

    $order = Order::factory()->create();
    $data = PaymentData::testFactory();

    expect(fn () => resolve(ProcessPaymentAction::class)($order, $data))
        ->toThrow(PaymentException::class, 'Card declined');
});

Using Factories

Factories create realistic, randomized test data that makes tests more robust.

Model Factories

// Arrange with factories
$user = User::factory()->create();
$product = Product::factory()->active()->create();
$order = Order::factory()->for($user)->create();

// Factory with state
$pendingOrder = Order::factory()->pending()->create();
$paidOrder = Order::factory()->paid()->create();

// Factory with relationships
$user = User::factory()
    ->has(Order::factory()->count(3))
    ->create();

Declarative Factory Methods

Critical principle: Make tests declarative and readable by hiding database implementation details behind factory methods.

// ❌ Bad - Database schema leaks into test
$calendar = Calendar::factory()->create([
    'status' => 'accepted',
    'reminder_sent_at' => null,
    'approved_by' => User::factory()->create()->id,
    'approved_at' => now(),
]);

// ✅ Good - Declarative and readable
$calendar = Calendar::factory()->accepted()->create();

→ Complete declarative factory patterns: testing-factories.md

DTO Test Factories

DTOs should provide test factories for consistent test data:

class CreateOrderData extends Data
{
    public function __construct(
        public string $customerEmail,
        public OrderStatus $status,
        public array $items,
    ) {}

    public static function testFactory(): self
    {
        return new self(
            customerEmail: fake()->email(),
            status: OrderStatus::Pending,
            items: [
                [
                    'product_id' => Product::factory()->create()->id,
                    'quantity' => fake()->numberBetween(1, 5),
                ],
            ],
        );
    }
}

// Usage in tests
$data = CreateOrderData::testFactory();

Testing Strategy

Feature Tests (HTTP Layer)

Test the complete request/response cycle:

use function Pest\Laravel\actingAs;
use function Pest\Laravel\postJson;

it('creates an order via API', function () {
    $user = User::factory()->create();
    $product = Product::factory()->create();

    $response = actingAs($user)
        ->postJson('/api/orders', [
            'customer_email' => '[email protected]',
            'items' => [
                ['product_id' => $product->id, 'quantity' => 2],
            ],
        ]);

    $response->assertCreated()
        ->assertJsonStructure([
            'data' => ['id', 'status', 'items'],
        ]);
});

Unit Tests (Actions)

Test domain logic in isolation:

it('calculates order total correctly', function () {
    $order = Order::factory()->create();
    $order->items()->createMany([
        ['price' => 1000, 'quantity' => 2],
        ['price' => 1500, 'quantity' => 1],
    ]);

    $total = resolve(CalculateOrderTotalAction::class)($order);

    expect($total)->toBe(3500);
});

Avoiding Brittle Tests

Brittle tests break when implementation changes, even if behavior is correct.

Signs of Brittle Tests

• Too many mocks
• Testing implementation details
• Hardcoded values everywhere
• Complex setup with many steps
• Tests break with refactoring

How to Avoid Brittleness

1. Use Real Instances When Possible

// ✅ Good - Use real instances
it('calculates order total', function () {
    $order = Order::factory()->create();
    $order->items()->createMany([
        ['price' => 1000, 'quantity' => 2],
        ['price' => 500, 'quantity' => 1],
    ]);

    $total = resolve(CalculateOrderTotalAction::class)($order);

    expect($total)->toBe(2500);
});

// ❌ Bad - Mock everything
it('calculates order total', function () {
    $item1 = Mockery::mock(OrderItem::class);
    $item1->shouldReceive('getPrice')->andReturn(1000);
    // ... too much mocking
});

2. Test Behavior, Not Implementation

// ✅ Good - Test the behavior
it('sends welcome email when user registers', function () {
    Mail::fake();

    $data = RegisterUserData::testFactory();
    $user = resolve(RegisterUserAction::class)($data);

    Mail::assertSent(WelcomeEmail::class, function ($mail) use ($user) {
        return $mail->hasTo($user->email);
    });
});

// ❌ Bad - Test implementation details
it('sends welcome email when user registers', function () {
    $mailer = Mockery::mock(Mailer::class);
    $mailer->shouldReceive('send')
        ->with(Mockery::on(function ($email) {
            return $email->template === 'emails.welcome';
        }));
    // Too specific, breaks if template name changes
});

3. Use Factories Instead of Hardcoded Data

// ✅ Good - Use factories
$user = User::factory()->create();
$data = ProfileData::testFactory();

// ❌ Bad - Hardcoded data
$data = new ProfileData(
    firstName: 'John',
    lastName: 'Doe',
    phone: '555-1234',
    bio: 'Test bio',
);

4. Minimize Mocking

Rule of thumb: Mock collaborators, not data.

// ✅ Good - Mock the notification service (collaborator)
$notifier = mock(NotificationService::class);
$notifier->shouldReceive('send')->once();
swap(NotificationService::class, $notifier);

resolve(ShipOrderAction::class)($order);

// ❌ Bad - Mock the data (order, user)
$order = Mockery::mock(Order::class);
// ... mocking data objects makes test brittle

Common Testing Patterns

Testing State Transitions

it('transitions order from pending to paid', function () {
    $order = Order::factory()->pending()->create();

    resolve(MarkOrderAsPaidAction::class)($order);

    expect($order->fresh()->status)->toBe(OrderStatus::Paid)
        ->and($order->fresh()->paid_at)->not->toBeNull();
});

Testing Relationships

it('creates order with items', function () {
    $user = User::factory()->create();
    $products = Product::factory()->count(3)->create();

    $data = CreateOrderData::from([
        'customer_email' => '[email protected]',
        'items' => $products->map(fn ($p) => [
            'product_id' => $p->id,
            'quantity' => 2,
        ])->all(),
    ]);

    $order = resolve(CreateOrderAction::class)($user, $data);

    expect($order->items)->toHaveCount(3);
});

Testing Transactions

it('rolls back transaction on failure', function () {
    $user = User::factory()->create();

    $data = CreateOrderData::from([
        'customer_email' => '[email protected]',
        'items' => [
            ['product_id' => 99999, 'quantity' => 1], // Non-existent product
        ],
    ]);

    expect(fn () => resolve(CreateOrderAction::class)($user, $data))
        ->toThrow(Exception::class);

    assertDatabaseCount('orders', 0);
    assertDatabaseCount('order_items', 0);
});

Testing Email/Notifications

use Illuminate\Support\Facades\Mail;

it('sends welcome email to new user', function () {
    Mail::fake();
    $data = RegisterUserData::testFactory();

    $user = resolve(RegisterUserAction::class)($data);

    Mail::assertSent(WelcomeEmail::class, function ($mail) use ($user) {
        return $mail->hasTo($user->email);
    });
});

Testing Jobs

use Illuminate\Support\Facades\Queue;

it('dispatches job to process order', function () {
    Queue::fake();
    $order = Order::factory()->create();

    resolve(ProcessOrderAction::class)($order);

    Queue::assertPushed(ProcessOrderJob::class, function ($job) use ($order) {
        return $job->order->id === $order->id;
    });
});

Best Practices Summary

✅ Do This

• Follow triple-A pattern - Arrange, Act, Assert
• Use factories for all test data
• Create declarative factory methods - Calendar::factory()->accepted() not ['status' => 'accepted']
• Test actions in isolation - Unit test your domain logic
• Mock what you own - Actions, services you control
• Create abstractions when you need to mock external services
• Use null drivers for external service testing
• Test behavior, not implementation
• Keep tests simple - One concept per test
• Use DTO test factories for consistent data

❌ Don't Do This

• Mock external libraries - Create service layer instead
• Hardcode test data - Use factories
• Leak database schema into tests - Use declarative factory methods
• Test implementation details - Test behavior
• Create brittle tests - Too many mocks, too specific
• Skip factories - Always use factories for models and DTOs
• Mix arrange and act - Keep them separate
• Over-mock - Use real instances when possible

Quick Reference

Test Structure

it('does something', function () {
    // Arrange - Set up the world with declarative factories
    $model = Model::factory()->active()->create();
    $data = Data::testFactory();

    // Act - Perform the operation
    $result = resolve(Action::class)($model, $data);

    // Assert - Verify the results
    expect($result)->/* assertions */;
});

Mocking Pattern

use function Pest\Laravel\mock;
use function Pest\Laravel\swap;

// Mock a dependency action
$mockAction = mock(YourDependencyAction::class);
$mockAction->shouldReceive('__invoke')
    ->once()
    ->with(/* expected params */)
    ->andReturn(/* return value */);

// Swap into container
swap(YourDependencyAction::class, $mockAction);

// Resolve action under test - container injects mocked dependencies
$result = resolve(ActionUnderTest::class)(/* params */);

Database Assertions

use function Pest\Laravel\assertDatabaseHas;
use function Pest\Laravel\assertDatabaseCount;

assertDatabaseHas('orders', ['id' => $order->id]);
assertDatabaseCount('orders', 1);