web-testing-vitest
Testing Standards
Quick Guide: E2E for user flows (Playwright). Unit for pure functions (Vitest). Integration tests okay but not primary (Vitest + RTL + network-level mocking). Current app uses integration tests with network-level API mocking.
<critical_requirements>
CRITICAL: Before Using This Skill
All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering,
import type, named constants)
(You MUST write E2E tests for ALL critical user workflows - NOT unit tests for React components)
(You MUST use Playwright for E2E tests and organize by user journey - NOT by component)
(You MUST only write unit tests for pure functions - NOT for components, hooks, or side effects)
(You MUST co-locate tests with code in feature-based structure - NOT in separate test directories)
(You MUST use network-level API mocking - NOT module-level mocks)
</critical_requirements>
this is a new change
Auto-detection: E2E testing, Playwright, test-driven development (Tester), Vitest, React Testing Library, test organization
When to use:
- Writing E2E tests for user workflows (primary approach with Playwright)
- Unit testing pure utility functions with Vitest
- Setting up network-level mocking for integration tests (current codebase approach)
- Organizing tests in feature-based structure (co-located tests)
When NOT to use:
- Unit testing React components (use E2E tests instead)
- Unit testing hooks with side effects (use E2E tests or integration tests)
- Testing third-party library behavior (library already has tests)
- Testing TypeScript compile-time guarantees (TypeScript already enforces)
- Creating stories for app-specific features (stories are for design system only)
Key patterns covered:
- E2E tests for user workflows (primary - inverted testing pyramid)
- Unit tests for pure functions only (not components)
- Integration tests with Vitest + React Testing Library + network-level mocking (acceptable, not ideal)
- Feature-based test organization (co-located with code)
Detailed Resources:
- For E2E and unit test examples, see examples/core.md
- For integration test examples, see examples/integration.md
- For anti-patterns (what NOT to test), see examples/anti-patterns.md
- For Ladle story examples, see examples/ladle-stories.md
- For decision frameworks, see reference.md
Testing Philosophy
PRIMARY: E2E tests for most scenarios
E2E tests verify actual user workflows through the entire stack. They test real user experience, catch integration issues, and provide highest confidence.
SECONDARY: Unit tests for pure functions
Pure utilities, business logic, algorithms, data transformations, edge cases.
Integration tests acceptable but not primary
React Testing Library + network-level mocking useful for component behavior when E2E too slow. Don't replace E2E for user workflows.
Testing Pyramid Inverted:
E2E Tests (Most) - Test real user workflows
Integration Tests (Some, acceptable) - Component behavior
Unit Tests (Pure functions only) - Utilities, algorithms
When to use E2E tests:
- All critical user-facing workflows (login, checkout, data entry)
- Multi-step user journeys (signup -> verify email -> complete profile)
- Cross-browser compatibility needs
- Testing real integration with backend APIs
When NOT to use E2E tests:
- Pure utility functions (use unit tests instead)
- Individual component variants in isolation (use Ladle stories for documentation)
When to use unit tests:
- Pure functions with clear input -> output
- Business logic calculations (pricing, taxes, discounts)
- Data transformations and formatters
- Edge cases and boundary conditions
When NOT to use unit tests:
- React components (use E2E tests)
- Hooks with side effects (use E2E tests or integration tests)
- API calls or external integrations (use E2E tests)
Core Patterns
Pattern 1: E2E Testing with Playwright (PRIMARY)
E2E tests verify complete user workflows through the entire application stack, providing the highest confidence that features work correctly.
Framework: Playwright (recommended) or Cypress
What to test end-to-end:
- ALL critical user flows (login, checkout, data entry)
- ALL user-facing features (forms, navigation, interactions)
- Multi-step workflows (signup -> verify email -> complete profile)
- Error states users will encounter
- Happy paths AND error paths
- Cross-browser compatibility (Playwright makes this easy)
What NOT to test end-to-end:
- Pure utility functions (use unit tests)
- Individual component variants in isolation (not user-facing)
Test Organization:
tests/e2e/directory at root or in each app- Test files:
*.spec.tsor*.e2e.ts - Group by user journey, not by component
See examples/core.md for complete E2E test examples.
Pattern 2: Unit Testing Pure Functions
Only write unit tests for pure functions with no side effects. Never unit test React components - use E2E tests instead.
What to test:
- Pure functions with clear input -> output
- Business logic calculations (pricing, taxes, discounts)
- Data transformations and formatters
- Edge cases and boundary conditions
What NOT to test:
- React components (use E2E tests)
- Hooks with side effects (use E2E tests or integration tests)
- API calls or external integrations (use E2E tests)
See examples/core.md for pure function test examples.
Pattern 3: Integration Testing with Network-Level Mocking (Current Approach)
The current codebase uses Vitest + React Testing Library + network-level mocking for integration tests. This is acceptable but not ideal compared to E2E tests.
When Integration Tests Make Sense:
- Component behavior in isolation (form validation, UI state)
- When E2E tests are too slow for rapid feedback
- Testing edge cases that are hard to reproduce in E2E
- Development workflow (faster than spinning up full stack)
Current Pattern:
- Tests in
__tests__/directories co-located with code - Network-level API mocking (intercepts HTTP requests)
- Centralized mock data in shared package
- Test all states: loading, empty, error, success
Benefits:
- Tests component with API integration (via network-level mocking)
- Tests all states: loading, empty, error, success
- Centralized mock handlers in shared package
- Shared between tests and development
Limitations:
- Doesn't test real API (mocks can drift)
- Doesn't test full user workflow
- Requires maintaining mock parity with API
See examples/integration.md for integration test examples.
Pattern 4: Feature-Based Test Organization
Co-locate tests with code in feature-based structure. Tests live next to what they test.
Direct Co-location (Recommended):
apps/client-react/src/
features/
auth/
components/
LoginForm.tsx
LoginForm.test.tsx # Test next to component
hooks/
useAuth.ts
useAuth.test.ts # Test next to hook
Alternative: __tests__/ Subdirectories:
apps/client-react/src/features/auth/
components/
LoginForm.tsx
__tests__/
LoginForm.test.tsx
E2E Test Organization:
apps/client-react/
tests/
e2e/
auth/
login-flow.spec.ts
register-flow.spec.ts
checkout/
checkout-flow.spec.ts
File Naming Convention:
*.test.tsx/*.test.tsfor unit and integration tests (Vitest)*.spec.tsfor E2E tests (Playwright)
Choose one pattern and be consistent across the codebase.
Pattern 5: Component Documentation with Ladle
Design system components MUST have .stories.tsx files. App-specific features do NOT need stories.
Where Stories are REQUIRED:
packages/ui/src/
primitives/ # Stories required
components/ # Stories required
patterns/ # Stories required
templates/ # Stories required
Where Stories are OPTIONAL:
apps/client-next/
apps/client-react/
# App-specific features don't need stories
See examples/ladle-stories.md for Ladle story examples.
<critical_reminders>
CRITICAL REMINDERS
All code must follow project conventions in CLAUDE.md
(You MUST write E2E tests for ALL critical user workflows - NOT unit tests for React components)
(You MUST use Playwright for E2E tests and organize by user journey - NOT by component)
(You MUST only write unit tests for pure functions - NOT for components, hooks, or side effects)
(You MUST co-locate tests with code in feature-based structure - NOT in separate test directories)
(You MUST use network-level API mocking - NOT module-level mocks)
Failure to follow these rules will result in fragile tests that break on refactoring, untested critical user paths, and false confidence from high coverage of low-value tests.
</critical_reminders>