clean-tests

简洁测试

T1：测试不足

测试所有可能出问题的内容。将覆盖率工具作为参考，而不是目标。

# Bad - only tests happy path
def test_divide():
    assert divide(10, 2) == 5

# Good - tests edge cases too
def test_divide_normal():
    assert divide(10, 2) == 5

def test_divide_by_zero():
    with pytest.raises(ZeroDivisionError):
        divide(10, 0)

def test_divide_negative():
    assert divide(-10, 2) == -5

T2：使用覆盖率工具

覆盖率工具会指出测试策略中的缺口。不要忽视它们。

# Run with coverage
pytest --cov=myproject --cov-report=term-missing

# Aim for meaningful coverage, not 100%

T3：不要跳过简单测试

简单测试能记录行为并捕获回归问题。它们的价值远超成本。

# Worth having - documents expected behavior
def test_user_default_role():
    user = User(name="Alice")
    assert user.role == "member"

T4：被忽略的测试说明存在歧义

不要用 @pytest.mark.skip 来掩盖问题。要么修复测试，要么删除测试。

# Bad - hiding a problem
@pytest.mark.skip(reason="flaky, fix later")
def test_async_operation():
    ...

# Good - either fix it or document why it's skipped
@pytest.mark.skip(reason="Requires Redis, see CONTRIBUTING.md for setup")
def test_cache_invalidation():
    ...

T5：测试边界条件

Bug 常出现在边界处。要明确测试这些情况。

def test_pagination_boundaries():
    items = list(range(100))
    
    # First page
    assert paginate(items, page=1, size=10) == items[0:10]
    
    # Last page
    assert paginate(items, page=10, size=10) == items[90:100]
    
    # Beyond last page
    assert paginate(items, page=11, size=10) == []
    
    # Page zero (invalid)
    with pytest.raises(ValueError):
        paginate(items, page=0, size=10)
    
    # Empty list
    assert paginate([], page=1, size=10) == []

T6：全面测试 Bug 附近的情况

发现一个 Bug 时，要为所有类似情况编写测试。Bug 往往成簇出现。

# Found bug: off-by-one in date calculation
# Now test ALL date boundaries
def test_month_boundaries():
    assert last_day_of_month(2024, 1) == 31  # January
    assert last_day_of_month(2024, 2) == 29  # Leap year February
    assert last_day_of_month(2023, 2) == 28  # Non-leap February
    assert last_day_of_month(2024, 4) == 30  # 30-day month
    assert last_day_of_month(2024, 12) == 31 # December

T7：失败模式能揭示问题

测试失败时，要观察其中的模式。它们往往指向更深层的问题。

# If all async tests fail intermittently,
# the problem isn't the tests—it's the async handling

T8：测试覆盖模式能揭示问题

查看哪些代码路径尚未测试。它们往往暴露设计问题。

# If you can't easily test a function, it probably does too much
# Refactor for testability

T9：测试应当快速

运行缓慢的测试不会被执行。将每个单元测试控制在 100 ms 以内。

# Bad - hits real database
def test_user_creation():
    db = connect_to_database()  # Slow!
    user = db.create_user("Alice")
    assert user.name == "Alice"

# Good - uses mock or in-memory
def test_user_creation():
    db = InMemoryDatabase()
    user = db.create_user("Alice")
    assert user.name == "Alice"

测试组织

F.I.R.S.T. 原则

• 快速（Fast）：测试应快速运行
• 独立（Independent）：测试不应相互依赖
• 可重复（Repeatable）：在任何环境中每次都得到相同结果
• 自验证（Self-Validating）：自动通过或失败，无需人工检查
• 及时（Timely）：在编写代码之前或同时编写，而不是事后编写

每个测试只验证一个概念

# Bad - testing multiple things
def test_user():
    user = User("Alice", "[email protected]")
    assert user.name == "Alice"
    assert user.email == "[email protected]"
    assert user.is_valid()
    user.activate()
    assert user.is_active

# Good - one concept each
def test_user_stores_name():
    user = User("Alice", "[email protected]")
    assert user.name == "Alice"

def test_user_stores_email():
    user = User("Alice", "[email protected]")
    assert user.email == "[email protected]"

def test_new_user_is_valid():
    user = User("Alice", "[email protected]")
    assert user.is_valid()

def test_user_can_be_activated():
    user = User("Alice", "[email protected]")
    user.activate()
    assert user.is_active

快速参考

规则	原则
T1	测试所有可能出问题的内容
T2	使用覆盖率工具
T3	不要跳过简单测试
T4	忽略测试 = 存在歧义
T5	测试边界条件
T6	全面测试 Bug 附近的情况
T7	查找失败模式
T8	调试时检查覆盖率
T9	测试必须快速（<100 ms）