testing-strategies（测试策略）

Testing Strategies

Build comprehensive, effective test suites by strategically selecting and implementing the right testing approaches across unit, integration, E2E, and contract testing levels.

Purpose

This skill provides strategic frameworks for:

• Test Type Selection: Determine when to use unit vs. integration vs. E2E vs. contract testing
• Test Pyramid Balancing: Optimize test distribution for fast feedback and reliable coverage
• Multi-Language Implementation: Apply consistent testing patterns across TypeScript, Python, Go, and Rust
• Test Data Management: Choose appropriate strategies (fixtures, factories, property-based testing)
• CI/CD Integration: Integrate tests into automated pipelines with optimal execution patterns

Testing is foundational to reliable software. With microservices architectures and continuous delivery becoming standard in 2025, strategic testing across multiple levels is more critical than ever.

When to Use This Skill

Invoke this skill when:

• Building a new feature that requires test coverage
• Designing a testing strategy for a new project
• Refactoring existing tests to improve speed or reliability
• Setting up CI/CD pipelines with testing stages
• Choosing between unit, integration, or E2E testing approaches
• Implementing contract testing for microservices
• Managing test data with fixtures, factories, or property-based testing

The Testing Pyramid Framework

Core Concept

The testing pyramid guides test distribution for optimal speed and confidence:

         /\
        /  \  E2E Tests (10%)
       /----\  - Slow but comprehensive
      /      \  - Full stack validation
     /--------\
    /          \  Integration Tests (20-30%)
   /            \  - Moderate speed
  /--------------\  - Component interactions
 /                \
/------------------\  Unit Tests (60-70%)
                      - Fast feedback
                      - Isolated units

Key Principle: More unit tests (fast, isolated), fewer E2E tests (slow, comprehensive). Integration tests bridge the gap.

Modern Adaptations (2025)

Microservices Adjustment:

• Add contract testing layer between unit and integration
• Increase integration/contract tests to 30% (validate service boundaries)
• Reduce E2E tests to critical user journeys only

Cloud-Native Patterns:

• Use containers for integration tests (ephemeral databases, test services)
• Parallel execution for fast CI/CD feedback
• Risk-based test prioritization (focus on high-impact areas)

For detailed pyramid guidance, see references/testing-pyramid.md.

Universal Testing Decision Tree

Which Test Type Should I Use?

START: Need to test [feature]

Q1: Does this involve multiple systems/services?
  ├─ YES → Q2
  └─ NO  → Q3

Q2: Is this a critical user-facing workflow?
  ├─ YES → E2E Test (complete user journey)
  └─ NO  → Integration or Contract Test

Q3: Does this interact with external dependencies (DB, API, filesystem)?
  ├─ YES → Integration Test (real DB, mocked API)
  └─ NO  → Q4

Q4: Is this pure business logic or a pure function?
  ├─ YES → Unit Test (fast, isolated)
  └─ NO  → Component or Integration Test

Test Type Selection Examples

For comprehensive decision frameworks, see references/decision-tree.md.

Testing Levels in Detail

Unit Testing (Foundation - 60-70%)

Purpose: Validate small, isolated units of code (functions, methods, components)

Characteristics:

• Fast (milliseconds per test)
• Isolated (no external dependencies)
• Deterministic (same input = same output)
• Broad coverage (many tests, small scope each)

When to Use:

• Pure functions (input → output)
• Business logic and algorithms
• Utility functions
• Component rendering (without integration)
• Validation logic

Recommended Tools:

• TypeScript/JavaScript: Vitest (primary, 10x faster than Jest), Jest (legacy)
• Python: pytest (industry standard)
• Go: testing package (stdlib) + testify (assertions)
• Rust: cargo test (stdlib)

For detailed patterns, see references/unit-testing-patterns.md.

Integration Testing (Middle Layer - 20-30%)

Purpose: Validate interactions between components, modules, or services

Characteristics:

• Moderate speed (seconds per test)
• Partial integration (real database, mocked external APIs)
• Focused scope (test component boundaries)
• API and database validation

When to Use:

• API endpoints (request → response)
• Database operations (CRUD, queries)
• Service-to-service communication
• Event handlers and message processing
• File I/O operations

Recommended Tools:

• TypeScript/JavaScript: Vitest + MSW (API mocking), Supertest (HTTP testing)
• Python: pytest + pytest-httpserver, pytest-postgresql
• Go: testing + httptest, testcontainers
• Rust: cargo test + mockito, testcontainers

For detailed patterns, see references/integration-testing-patterns.md.

End-to-End Testing (Top Layer - 10%)

Purpose: Validate complete user workflows across the entire application stack

Characteristics:

• Slow (minutes per test suite)
• Full integration (real browser, services, database)
• Wide scope (user journeys from start to finish)
• Prone to flakiness (requires careful design)

When to Use:

• Critical user journeys (login, checkout, payment)
• Cross-browser compatibility validation
• Real-world scenarios not covered by integration tests
• Regression prevention for core features

Best Practices:

• Limit E2E tests to high-value scenarios (not every edge case)
• Use stable selectors (data-testid, not CSS classes)
• Implement retry logic for network flakiness
• Run tests in parallel for speed

Recommended Tools:

• All Languages: Playwright (cross-browser, fast, Microsoft-backed)

For detailed patterns, see references/e2e-testing-patterns.md.

Contract Testing (Microservices)

Purpose: Validate service interfaces and API contracts without full integration

When to Use:

• Microservices architecture
• Service-to-service communication
• API contract validation
• Reducing E2E testing overhead

Recommended Tool: Pact (pact.io) - supports TypeScript, Python, Go, Rust

For detailed patterns, see references/contract-testing.md.

Test Data Management Strategies

When to Use Each Approach

Fixtures (Static Data):

• Pros: Deterministic, easy to debug
• Cons: Can become stale, doesn't test variety
• Use When: Testing known scenarios, regression tests

Factories (Generated Data):

• Pros: Flexible, generates variety
• Cons: Less deterministic, harder to debug
• Use When: Need diverse test data, testing edge cases

Property-Based Testing (Random Data):

• Pros: Finds edge cases not anticipated
• Cons: Can be slow, failures harder to reproduce
• Use When: Complex algorithms, parsers, validators

Recommended Combination:

• Unit Tests: Fixtures (known inputs) + Property-Based (edge cases)
• Integration Tests: Factories (flexible data) + Database seeding
• E2E Tests: Fixtures (reproducible scenarios)

Property-Based Testing Tools:

• TypeScript/JavaScript: fast-check
• Python: hypothesis (best-in-class)
• Go: gopter
• Rust: proptest (primary)

For detailed strategies, see references/test-data-strategies.md.

Mocking Decision Matrix

When to Mock vs. Use Real Dependencies

Guiding Principles:

1. Unit Tests: Mock everything external
2. Integration Tests: Use real database (ephemeral), mock external APIs
3. E2E Tests: Use real everything (or staging equivalents)
4. Contract Tests: Mock nothing (test real interfaces)

For detailed mocking patterns, see references/mocking-strategies.md.

Language-Specific Quick Starts

TypeScript/JavaScript

Unit Testing with Vitest:

import { describe, test, expect } from 'vitest'

test('calculates total with tax', () => {
  const items = [{ price: 10, quantity: 2 }]
  expect(calculateTotal(items, 0.1)).toBe(22)
})

Integration Testing with MSW:

import { setupServer } from 'msw/node'
import { http, HttpResponse } from 'msw'

const server = setupServer(
  http.get('/api/user/:id', ({ params }) => {
    return HttpResponse.json({ id: params.id, name: 'Test User' })
  })
)

E2E Testing with Playwright:

import { test, expect } from '@playwright/test'

test('user can checkout', async ({ page }) => {
  await page.goto('https://example.com')
  await page.getByRole('button', { name: 'Add to Cart' }).click()
  await expect(page.getByText('1 item in cart')).toBeVisible()
})

See examples/typescript/ for complete working examples.

Python

Unit Testing with pytest:

def test_calculate_total_with_tax():
    items = [{"price": 10, "quantity": 2}]
    assert calculate_total(items, tax_rate=0.1) == 22

Property-Based Testing with hypothesis:

from hypothesis import given, strategies as st

@given(st.lists(st.integers()))
def test_reverse_reverse_is_identity(lst):
    assert reverse(reverse(lst)) == lst

See examples/python/ for complete working examples.

Go and Rust

See examples/go/ and examples/rust/ for complete working examples in these languages.

Coverage and Quality Metrics

Meaningful Coverage Targets

Recommended Targets:

• Critical Business Logic: 90%+ coverage
• API Endpoints: 80%+ coverage
• Utility Functions: 70%+ coverage
• UI Components: 60%+ coverage (focus on logic, not markup)
• Overall Project: 70-80% coverage

Anti-Pattern: Aiming for 100% coverage leads to testing trivial code and false confidence.

Coverage Tools:

• TypeScript/JavaScript: Vitest coverage (c8/istanbul)
• Python: pytest-cov
• Go: go test -cover
• Rust: cargo-tarpaulin

Mutation Testing (Validate Test Quality)

Mutation testing introduces bugs and verifies tests catch them.

Tools:

• TypeScript/JavaScript: Stryker Mutator
• Python: mutmut
• Go: go-mutesting
• Rust: cargo-mutants

For detailed coverage strategies, see references/coverage-strategies.md.

CI/CD Integration Patterns

Fast Feedback Loop Strategy

Stage 1: Pre-Commit (< 30 seconds)

• Lint and format checks
• Unit tests (critical paths only)

Stage 2: On Commit (< 2 minutes)

• All unit tests
• Static analysis

Stage 3: Pull Request (< 5 minutes)

• Integration tests
• Coverage reporting

Stage 4: Pre-Merge (< 10 minutes)

• E2E tests (critical paths)
• Cross-browser testing

Parallel Execution

Speed Up Test Runs:

• Vitest: vitest --threads
• Playwright: playwright test --workers=4
• pytest: pytest -n auto (requires pytest-xdist)
• Go: go test -parallel 4

Common Anti-Patterns to Avoid

The Ice Cream Cone (Inverted Pyramid)

Problem: Too many E2E tests, too few unit tests Result: Slow test suites, flaky tests, long feedback loops Solution: Rebalance toward more unit tests, fewer E2E tests

Testing Implementation Details

Problem: Tests coupled to internal structure, not behavior Solution: Test behavior (inputs → outputs), not implementation

Over-Mocking in Integration Tests

Problem: Mocking everything defeats purpose of integration tests Solution: Use real databases (ephemeral), mock only external APIs

Flaky E2E Tests

Problem: Tests fail intermittently due to timing issues Solutions:

• Use auto-wait features (Playwright has built-in auto-wait)
• Avoid hardcoded waits (sleep(1000))
• Use stable selectors (data-testid)
• Implement retry logic for network requests

Integration with Other Skills

For form testing: See building-forms skill for form-specific validation and submission testing patterns.

For API testing: See api-patterns skill for REST/GraphQL endpoint testing and contract validation.

For CI/CD pipelines: See building-ci-pipelines skill for test automation, parallel execution, and coverage reporting.

For data visualization testing: See visualizing-data skill for snapshot testing chart configurations and visual regression testing.

Reference Documentation

For deeper exploration of specific topics:

• references/testing-pyramid.md - Detailed testing pyramid framework and balancing strategies
• references/decision-tree.md - Comprehensive decision frameworks for test type selection
• references/unit-testing-patterns.md - Unit testing patterns across languages
• references/integration-testing-patterns.md - Integration testing with databases, APIs, and containers
• references/e2e-testing-patterns.md - E2E testing best practices and Playwright patterns
• references/contract-testing.md - Consumer-driven contract testing with Pact
• references/test-data-strategies.md - Fixtures, factories, and property-based testing
• references/mocking-strategies.md - When and how to mock dependencies
• references/coverage-strategies.md - Meaningful coverage metrics and mutation testing

Working Examples

Complete, runnable code examples are available in the examples/ directory:

• examples/typescript/ - Vitest unit/integration, Playwright E2E, MSW mocking, fast-check property tests
• examples/python/ - pytest unit/integration, fixtures, Playwright E2E, hypothesis property tests
• examples/go/ - stdlib testing, testify assertions, httptest integration
• examples/rust/ - cargo test unit tests, proptest property tests, integration patterns

All examples include dependencies, usage instructions, and error handling.