test-driven-development（测试驱动开发）

Test-Driven Development (TDD) Workflow

Overview

Test-Driven Development enforces the RED → GREEN → REFACTOR discipline to ensure robust, well-tested code with clean design. This skill guides implementation of the three-phase TDD cycle for both Python and TypeScript/JavaScript codebases.

TDD Three-Phase Cycle

Phase 1: RED - Write Failing Test

Write the smallest possible test that fails for the right reason.

Principles:

• Test documents the intended behavior
• Test must fail initially (verify failure)
• Focus on one specific behavior per test
• Use descriptive test names that explain the "should"

Test Structure:

// Arrange - Set up test data
// Act - Execute the behavior
// Assert - Verify the result

Phase 2: GREEN - Make Test Pass

Write the minimal implementation to make the test pass.

Principles:

• Implement only what's needed for the test to pass
• Don't optimize or refactor yet
• Hardcode values if necessary
• Focus on correctness, not elegance

Phase 3: REFACTOR - Clean Up Code

Improve the code while keeping tests green.

Focus Areas:

• Extract repeated logic into functions
• Improve naming and readability
• Apply SOLID principles
• Remove duplication
• Optimize performance if needed

Safety Net: Run tests after each refactoring step.

Implementation Workflow

Step 1: Understand Requirements

• Clarify the feature's expected behavior
• Identify edge cases and error conditions
• Define acceptance criteria

Step 2: Write Integration Test First

• Start with a higher-level test that describes user behavior
• Use realistic test data and scenarios
• Focus on the public interface

Step 3: Follow TDD for Implementation

• Write unit test (RED)
• Implement minimal solution (GREEN)
• Refactor and clean up (REFACTOR)
• Repeat for each behavior

Step 4: Verify Complete Coverage

• Run coverage report
• Ensure all new code paths are tested
• Add tests for any missing edge cases

Test Categories

Unit Tests

• Test individual functions/methods in isolation
• Use mocks for external dependencies
• Fast execution (< 100ms each)
• High coverage of business logic

Integration Tests

• Test component interactions
• Use real database/API connections where appropriate
• Validate end-to-end workflows
• Slower but comprehensive

Edge Case Tests

• Null/undefined inputs
• Empty collections
• Boundary values (min/max)
• Error conditions and exceptions

Language-Specific Patterns

Python (pytest)

def test_should_calculate_total_with_tax():
    # Arrange
    cart = ShoppingCart()
    cart.add_item("item1", price=100.00)

    # Act
    total = cart.calculate_total(tax_rate=0.08)

    # Assert
    assert total == 108.00

TypeScript (Jest)

describe('ShoppingCart', () => {
  it('should calculate total with tax', () => {
    // Arrange
    const cart = new ShoppingCart();
    cart.addItem('item1', 100.00);

    // Act
    const total = cart.calculateTotal(0.08);

    // Assert
    expect(total).toBe(108.00);
  });
});

Quality Gates

Before Moving to GREEN

• [ ] Test fails for the right reason
• [ ] Test name clearly describes the behavior
• [ ] Test is minimal and focused
• [ ] All existing tests still pass

Before Moving to REFACTOR

• [ ] New test passes
• [ ] Implementation is minimal
• [ ] No over-engineering
• [ ] All tests still pass

After REFACTOR

• [ ] Code is clean and readable
• [ ] No duplication
• [ ] SOLID principles applied
• [ ] All tests pass
• [ ] Coverage maintained

Common Pitfalls to Avoid

Writing Too Much Code in GREEN

Problem: Implementing multiple features at once Solution: Write only enough code to make the current test pass

Skipping the RED Phase

Problem: Writing tests after implementation Solution: Always write failing test first, verify it fails

Not Refactoring

Problem: Accumulating technical debt Solution: Refactor immediately after GREEN, while context is fresh

Testing Implementation Details

Problem: Tests break when refactoring Solution: Test public behavior, not internal implementation

Tools and Commands

Running Tests

# Python
pytest --cov=src tests/
pytest -v --cov-report=html

# TypeScript/Jest
npm test
npm run test:coverage
npm run test:watch

Coverage Thresholds

• Lines: 80% minimum
• Branches: 80% minimum
• Functions: 90% minimum
• New code: 100% coverage required

Additional Resources

Reference Files

For detailed patterns and advanced techniques, consult:

• references/tdd-patterns.md - Common TDD patterns and best practices
• references/testing-pyramid.md - Testing strategy and test types
• references/mocking-strategies.md - Mock patterns for different scenarios

Example Files

Working TDD examples in examples/:

• examples/python-tdd-example.py - Complete Python TDD workflow
• examples/typescript-tdd-example.ts - TypeScript TDD implementation
• examples/api-endpoint-tdd.py - TDD for API endpoints

Scripts

Utility scripts in scripts/:

• scripts/run-tdd-cycle.sh - Automated TDD cycle runner
• scripts/coverage-check.sh - Coverage validation script
• scripts/test-watch.sh - Watch mode for continuous testing

TDD in Context

When to Use TDD

• New feature development
• Bug fixes with regression tests
• Refactoring existing code
• API design and validation

When to Consider Alternatives

• Exploratory coding/prototyping
• UI/UX experimentation
• Performance optimization spikes
• External system integration discovery

Success Metrics

Quantitative

• Code coverage > 80%
• Test execution time < 5 minutes
• Test failure rate < 5%
• Defect density reduction

Qualitative

• Confidence in refactoring
• Clear behavior documentation
• Faster debugging cycles
• Improved code design

Apply TDD discipline consistently to build robust, maintainable code with comprehensive test coverage and clean architecture.