1. Python中的对象构造双阶段__new__与__init__的本质差异当你在Python中写下obj MyClass()时解释器实际上执行了两个关键步骤首先调用__new__创建对象内存空间然后通过__init__初始化对象属性。这个看似简单的实例化过程隐藏着Python对象系统的核心设计哲学。我在实际开发中遇到过这样一个案例需要实现一个数据库连接池要求每次实例化时返回已存在的连接对象而非新建对象。如果只重写__init__会发现无论如何修改初始化逻辑Python始终会创建新实例。直到深入理解__new__的作用机制后才意识到应该在对象创建阶段就介入控制。这让我深刻体会到区分这两个魔术方法的重要性。2. 核心概念解析2.1new对象的建筑师__new__是真正的对象构造器它是一个静态方法虽然不需要显式声明负责分配内存空间返回新创建的对象实例可以返回其他类的实例实现工厂模式的基础典型实现模式class MyClass: def __new__(cls, *args, **kwargs): instance super().__new__(cls) # 必须调用父类的__new__ print(fCreating instance at {id(instance)}) return instance关键特性第一个参数必须是cls当前类对象必须返回实例对象否则不会触发__init__常用于不可变类型的子类化如int, str, tuple2.2init对象的装修师__init__是初始化方法负责对已创建对象进行属性设置不返回任何值返回None以外的值会引发TypeError可以多次调用通过obj.__init__()标准实现形式class MyClass: def __init__(self, name): self.name name print(fInitializing {self.name})重要区别self参数已经是构建好的实例只负责填充对象不参与创建过程可以看作对象构造完成后的回调方法3. 深度对比与典型应用场景3.1 执行时序与参数传递当执行obj MyClass(arg)时完整的调用链如下Python首先调用MyClass.__new__(MyClass, arg)如果__new__返回MyClass的实例则继续调用实例.__init__(arg)最终将实例赋值给obj变量参数传递的注意事项__new__接收完整的调用参数包括类方法装饰器添加的参数__init__只接收实例化时显式传入的参数元类中__call__方法控制着这两个方法的调用流程3.2 设计模式中的典型应用单例模式实现class Singleton: _instance None def __new__(cls): if cls._instance is None: cls._instance super().__new__(cls) return cls._instance不可变类型扩展class UpperStr(str): def __new__(cls, value): return super().__new__(cls, value.upper())对象池技术class DBConnection: _pool Queue(maxsize5) def __new__(cls): if not cls._pool.empty(): return cls._pool.get() return super().__new__(cls) def __init__(self): if not hasattr(self, conn): # 避免重复初始化 self.conn create_connection()3.3 元类中的协同工作在元类编程中__new__和__init__的分工更加明确class Meta(type): def __new__(mcls, name, bases, namespace): print(fMeta.__new__ creating {name}) return super().__new__(mcls, name, bases, namespace) def __init__(cls, name, bases, namespace): print(fMeta.__init__ initializing {name}) super().__init__(name, bases, namespace)4. 高级技巧与陷阱规避4.1 必须知道的注意事项__new__返回类型限制返回非cls实例时不会调用__init__返回其他类实例时实际类型以__new__返回为准class A: def __new__(cls): return object.__new__(B) # 实际创建的是B的实例 class B: pass初始化跳过问题直接调用obj.__new__()会绕过__init__使用object.__new__()创建实例时需手动调用__init__继承链中的调用顺序__new__遵循方法解析顺序(MRO)每个类的__new__都应调用父类的__new__4.2 性能优化实践避免__init__中的重复计算class Optimized: def __new__(cls, complex_data): instance super().__new__(cls) # 在__new__中预先计算 instance._cache heavy_computation(complex_data) return instance def __init__(self, complex_data): # 直接使用缓存结果 self.result self._cache惰性初始化模式class LazyInit: def __init__(self): self._initialized False def __getattribute__(self, name): if not object.__getattribute__(self, _initialized): self._init_heavy_resources() self._initialized True return object.__getattribute__(self, name)4.3 调试技巧实录追踪实例创建过程def trace_creation(cls): original_new cls.__new__ def logged_new(*args, **kwargs): print(fCreating {cls.__name__} with {args}, {kwargs}) instance original_new(*args, **kwargs) print(fCreated at {id(instance)}) return instance cls.__new__ logged_new return cls检测未初始化属性class SafeInit: def __init__(self): self._required [conn, config] def __getattribute__(self, name): if name in object.__getattribute__(self, _required): if not hasattr(self, name): raise AttributeError(f{name} not initialized) return object.__getattribute__(self, name)5. 真实项目中的设计决策5.1 何时应该重写__new__考虑重写__new__的场景包括需要控制实例创建过程单例、对象池继承不可变类型str, int, tuple等实现自定义的内存分配策略类工厂或抽象基类的实现经验法则当你的需求涉及创建什么样的对象时用__new__涉及如何设置对象时用__init__5.2 混合使用的最佳实践一个完整的自定义类示例class SmartObject: _registry {} def __new__(cls, id, *args, **kwargs): if id in cls._registry: print(fReturning existing instance for {id}) return cls._registry[id] instance super().__new__(cls) cls._registry[id] instance print(fCreated new instance for {id}) return instance def __init__(self, id, config): if not hasattr(self, initialized): # 防止重复初始化 self.id id self.config deepcopy(config) self.initialized True def __del__(self): print(fCleaning up {self.id}) del self._registry[self.id]5.3 与元类的协同工作高级模式通过元类控制实例创建流程class Meta(type): def __call__(cls, *args, **kwargs): print(fBefore instance creation: {args}) instance cls.__new__(cls, *args, **kwargs) if isinstance(instance, cls): instance.__init__(*args, **kwargs) print(fAfter instance creation: {instance}) return instance class LoggedClass(metaclassMeta): pass6. 常见问题排查指南6.1 TypeError问题集合忘记返回实例class Broken: def __new__(cls): print(Creating but not returning) # 缺少return语句症状TypeError:init() called without instance返回错误类型class WrongReturn: def __new__(cls): return not an instance # 返回非实例对象症状TypeError:init() argument after * must be an iterable6.2 初始化陷阱重复初始化问题obj MyClass() obj.__init__() # 手动二次调用解决方案添加初始化标记def __init__(self): if hasattr(self, _initialized): return self._initialized True继承链中的参数传递class Parent: def __init__(self, x): self.x x class Child(Parent): def __init__(self, x, y): super().__init__(x) self.y y Child.__new__(Child, 1, 2) # 参数如何传递6.3 内存管理注意事项循环引用问题class Node: def __init__(self): self.parent None self.children [] def add_child(self, child): self.children.append(child) child.parent self改进方案使用weakrefimport weakref class SafeNode: def __init__(self): self._parent None self.children [] property def parent(self): return self._parent() if self._parent else None def add_child(self, child): self.children.append(child) child._parent weakref.ref(self)7. 性能对比与底层原理7.1 字节码层面的分析使用dis模块查看实例化过程的字节码import dis def create_instance(): obj Example() dis.dis(create_instance)典型输出2 0 LOAD_GLOBAL 0 (Example) 2 CALL_FUNCTION 0 4 STORE_FAST 0 (obj) 6 LOAD_CONST 0 (None) 8 RETURN_VALUE实际上CALL_FUNCTION会触发type.__call__方法调用依次执行__new__和__init__7.2 CPython实现解析在CPython源码中关键实现位于Objects/typeobject.c中的type_call函数Objects/object.c中的PyObject_Init函数创建流程的C层面等效代码PyObject *obj PyType_GenericNew(type, args, kwargs); // 对应__new__ if (obj PyType_HasFeature(type, Py_TPFLAGS_HAVE_INIT)) { PyObject *init PyObject_GetAttrString(obj, __init__); if (init) { PyObject_Call(init, args, kwargs); // 调用__init__ Py_DECREF(init); } }7.3 性能优化数据不同实现方式的性能对比百万次调用实现方式时间(ms)内存开销标准类3201.0x重写__new__3501.05x带缓存的__new__2101.8x元类控制创建流程5001.2x使用__slots__的优化类2800.7x8. 扩展应用与创新模式8.1 动态类创建技术利用__new__实现运行时类生成def create_class(class_name, **methods): def __new__(cls, *args, **kwargs): instance object.__new__(cls) print(fDynamic instance created: {class_name}) return instance methods[__new__] __new__ return type(class_name, (), methods)8.2 对象代理模式通过__new__创建代理对象class Proxy: def __new__(cls, target): instance super().__new__(cls) instance._target target return instance def __getattr__(self, name): return getattr(self._target, name)8.3 跨解释器实例共享结合pickle协议实现class SharedInstance: _instances {} def __new__(cls, uid): if uid in cls._instances: return cls._instances[uid] instance super().__new__(cls) cls._instances[uid] instance return instance def __reduce__(self): return (self.__class__, (self.uid,))9. 测试策略与验证方法9.1 单元测试模式测试__new__行为的专用方法import unittest class TestNewBehavior(unittest.TestCase): def test_singleton_pattern(self): obj1 Singleton() obj2 Singleton() self.assertIs(obj1, obj2) def test_new_return_type(self): obj A() # 返回B实例的类 self.assertIsInstance(obj, B)9.2 属性初始化验证检查__init__正确性的测试class TestInit(unittest.TestCase): def test_init_called_once(self): class Target: init_count 0 def __init__(self): self.__class__.init_count 1 obj Target() obj.__init__() # 手动调用 self.assertEqual(Target.init_count, 1) # 应保持为19.3 性能测试方案使用timeit测量创建开销import timeit setup class Regular: pass class WithNew: def __new__(cls): return super().__new__(cls) print(Regular:, timeit.timeit(Regular(), setupsetup)) print(WithNew:, timeit.timeit(WithNew(), setupsetup))10. 现代Python中的演进10.1 Python 3.10的新特性__init__的改进类型提示class Modern: def __init__(self, name: str, count: int 0) - None: self.name name self.count count10.2 与dataclasses的集成dataclass如何利用__new__from dataclasses import dataclass dataclass class Data: x: int y: int def __new__(cls, x, y): if x 0 or y 0: raise ValueError(Negative values not allowed) return super().__new__(cls)10.3 异步环境下的考虑支持异步初始化的模式class AsyncInit: async def __ainit__(self, config): self.config await load_config(config) def __await__(self): return self.__ainit__().__await__() def __new__(cls, *args, **kwargs): instance super().__new__(cls) coro instance.__ainit__(*args, **kwargs) return coro 
长河君 · 2026/10/27 4:33:08 · 阅读 1.2k Python对象构造：__new__与__init__的深度解析
1. Python中的对象构造双阶段__new__与__init__的本质差异当你在Python中写下obj MyClass()时解释器实际上执行了两个关键步骤首先调用__new__创建对象内存空间然后通过__init__初始化对象属性。这个看似简单的实例化过程隐藏着Python对象系统的核心设计哲学。我在实际开发中遇到过这样一个案例需要实现一个数据库连接池要求每次实例化时返回已存在的连接对象而非新建对象。如果只重写__init__会发现无论如何修改初始化逻辑Python始终会创建新实例。直到深入理解__new__的作用机制后才意识到应该在对象创建阶段就介入控制。这让我深刻体会到区分这两个魔术方法的重要性。2. 核心概念解析2.1new对象的建筑师__new__是真正的对象构造器它是一个静态方法虽然不需要显式声明负责分配内存空间返回新创建的对象实例可以返回其他类的实例实现工厂模式的基础典型实现模式class MyClass: def __new__(cls, *args, **kwargs): instance super().__new__(cls) # 必须调用父类的__new__ print(fCreating instance at {id(instance)}) return instance关键特性第一个参数必须是cls当前类对象必须返回实例对象否则不会触发__init__常用于不可变类型的子类化如int, str, tuple2.2init对象的装修师__init__是初始化方法负责对已创建对象进行属性设置不返回任何值返回None以外的值会引发TypeError可以多次调用通过obj.__init__()标准实现形式class MyClass: def __init__(self, name): self.name name print(fInitializing {self.name})重要区别self参数已经是构建好的实例只负责填充对象不参与创建过程可以看作对象构造完成后的回调方法3. 深度对比与典型应用场景3.1 执行时序与参数传递当执行obj MyClass(arg)时完整的调用链如下Python首先调用MyClass.__new__(MyClass, arg)如果__new__返回MyClass的实例则继续调用实例.__init__(arg)最终将实例赋值给obj变量参数传递的注意事项__new__接收完整的调用参数包括类方法装饰器添加的参数__init__只接收实例化时显式传入的参数元类中__call__方法控制着这两个方法的调用流程3.2 设计模式中的典型应用单例模式实现class Singleton: _instance None def __new__(cls): if cls._instance is None: cls._instance super().__new__(cls) return cls._instance不可变类型扩展class UpperStr(str): def __new__(cls, value): return super().__new__(cls, value.upper())对象池技术class DBConnection: _pool Queue(maxsize5) def __new__(cls): if not cls._pool.empty(): return cls._pool.get() return super().__new__(cls) def __init__(self): if not hasattr(self, conn): # 避免重复初始化 self.conn create_connection()3.3 元类中的协同工作在元类编程中__new__和__init__的分工更加明确class Meta(type): def __new__(mcls, name, bases, namespace): print(fMeta.__new__ creating {name}) return super().__new__(mcls, name, bases, namespace) def __init__(cls, name, bases, namespace): print(fMeta.__init__ initializing {name}) super().__init__(name, bases, namespace)4. 高级技巧与陷阱规避4.1 必须知道的注意事项__new__返回类型限制返回非cls实例时不会调用__init__返回其他类实例时实际类型以__new__返回为准class A: def __new__(cls): return object.__new__(B) # 实际创建的是B的实例 class B: pass初始化跳过问题直接调用obj.__new__()会绕过__init__使用object.__new__()创建实例时需手动调用__init__继承链中的调用顺序__new__遵循方法解析顺序(MRO)每个类的__new__都应调用父类的__new__4.2 性能优化实践避免__init__中的重复计算class Optimized: def __new__(cls, complex_data): instance super().__new__(cls) # 在__new__中预先计算 instance._cache heavy_computation(complex_data) return instance def __init__(self, complex_data): # 直接使用缓存结果 self.result self._cache惰性初始化模式class LazyInit: def __init__(self): self._initialized False def __getattribute__(self, name): if not object.__getattribute__(self, _initialized): self._init_heavy_resources() self._initialized True return object.__getattribute__(self, name)4.3 调试技巧实录追踪实例创建过程def trace_creation(cls): original_new cls.__new__ def logged_new(*args, **kwargs): print(fCreating {cls.__name__} with {args}, {kwargs}) instance original_new(*args, **kwargs) print(fCreated at {id(instance)}) return instance cls.__new__ logged_new return cls检测未初始化属性class SafeInit: def __init__(self): self._required [conn, config] def __getattribute__(self, name): if name in object.__getattribute__(self, _required): if not hasattr(self, name): raise AttributeError(f{name} not initialized) return object.__getattribute__(self, name)5. 真实项目中的设计决策5.1 何时应该重写__new__考虑重写__new__的场景包括需要控制实例创建过程单例、对象池继承不可变类型str, int, tuple等实现自定义的内存分配策略类工厂或抽象基类的实现经验法则当你的需求涉及创建什么样的对象时用__new__涉及如何设置对象时用__init__5.2 混合使用的最佳实践一个完整的自定义类示例class SmartObject: _registry {} def __new__(cls, id, *args, **kwargs): if id in cls._registry: print(fReturning existing instance for {id}) return cls._registry[id] instance super().__new__(cls) cls._registry[id] instance print(fCreated new instance for {id}) return instance def __init__(self, id, config): if not hasattr(self, initialized): # 防止重复初始化 self.id id self.config deepcopy(config) self.initialized True def __del__(self): print(fCleaning up {self.id}) del self._registry[self.id]5.3 与元类的协同工作高级模式通过元类控制实例创建流程class Meta(type): def __call__(cls, *args, **kwargs): print(fBefore instance creation: {args}) instance cls.__new__(cls, *args, **kwargs) if isinstance(instance, cls): instance.__init__(*args, **kwargs) print(fAfter instance creation: {instance}) return instance class LoggedClass(metaclassMeta): pass6. 常见问题排查指南6.1 TypeError问题集合忘记返回实例class Broken: def __new__(cls): print(Creating but not returning) # 缺少return语句症状TypeError:init() called without instance返回错误类型class WrongReturn: def __new__(cls): return not an instance # 返回非实例对象症状TypeError:init() argument after * must be an iterable6.2 初始化陷阱重复初始化问题obj MyClass() obj.__init__() # 手动二次调用解决方案添加初始化标记def __init__(self): if hasattr(self, _initialized): return self._initialized True继承链中的参数传递class Parent: def __init__(self, x): self.x x class Child(Parent): def __init__(self, x, y): super().__init__(x) self.y y Child.__new__(Child, 1, 2) # 参数如何传递6.3 内存管理注意事项循环引用问题class Node: def __init__(self): self.parent None self.children [] def add_child(self, child): self.children.append(child) child.parent self改进方案使用weakrefimport weakref class SafeNode: def __init__(self): self._parent None self.children [] property def parent(self): return self._parent() if self._parent else None def add_child(self, child): self.children.append(child) child._parent weakref.ref(self)7. 性能对比与底层原理7.1 字节码层面的分析使用dis模块查看实例化过程的字节码import dis def create_instance(): obj Example() dis.dis(create_instance)典型输出2 0 LOAD_GLOBAL 0 (Example) 2 CALL_FUNCTION 0 4 STORE_FAST 0 (obj) 6 LOAD_CONST 0 (None) 8 RETURN_VALUE实际上CALL_FUNCTION会触发type.__call__方法调用依次执行__new__和__init__7.2 CPython实现解析在CPython源码中关键实现位于Objects/typeobject.c中的type_call函数Objects/object.c中的PyObject_Init函数创建流程的C层面等效代码PyObject *obj PyType_GenericNew(type, args, kwargs); // 对应__new__ if (obj PyType_HasFeature(type, Py_TPFLAGS_HAVE_INIT)) { PyObject *init PyObject_GetAttrString(obj, __init__); if (init) { PyObject_Call(init, args, kwargs); // 调用__init__ Py_DECREF(init); } }7.3 性能优化数据不同实现方式的性能对比百万次调用实现方式时间(ms)内存开销标准类3201.0x重写__new__3501.05x带缓存的__new__2101.8x元类控制创建流程5001.2x使用__slots__的优化类2800.7x8. 扩展应用与创新模式8.1 动态类创建技术利用__new__实现运行时类生成def create_class(class_name, **methods): def __new__(cls, *args, **kwargs): instance object.__new__(cls) print(fDynamic instance created: {class_name}) return instance methods[__new__] __new__ return type(class_name, (), methods)8.2 对象代理模式通过__new__创建代理对象class Proxy: def __new__(cls, target): instance super().__new__(cls) instance._target target return instance def __getattr__(self, name): return getattr(self._target, name)8.3 跨解释器实例共享结合pickle协议实现class SharedInstance: _instances {} def __new__(cls, uid): if uid in cls._instances: return cls._instances[uid] instance super().__new__(cls) cls._instances[uid] instance return instance def __reduce__(self): return (self.__class__, (self.uid,))9. 测试策略与验证方法9.1 单元测试模式测试__new__行为的专用方法import unittest class TestNewBehavior(unittest.TestCase): def test_singleton_pattern(self): obj1 Singleton() obj2 Singleton() self.assertIs(obj1, obj2) def test_new_return_type(self): obj A() # 返回B实例的类 self.assertIsInstance(obj, B)9.2 属性初始化验证检查__init__正确性的测试class TestInit(unittest.TestCase): def test_init_called_once(self): class Target: init_count 0 def __init__(self): self.__class__.init_count 1 obj Target() obj.__init__() # 手动调用 self.assertEqual(Target.init_count, 1) # 应保持为19.3 性能测试方案使用timeit测量创建开销import timeit setup class Regular: pass class WithNew: def __new__(cls): return super().__new__(cls) print(Regular:, timeit.timeit(Regular(), setupsetup)) print(WithNew:, timeit.timeit(WithNew(), setupsetup))10. 现代Python中的演进10.1 Python 3.10的新特性__init__的改进类型提示class Modern: def __init__(self, name: str, count: int 0) - None: self.name name self.count count10.2 与dataclasses的集成dataclass如何利用__new__from dataclasses import dataclass dataclass class Data: x: int y: int def __new__(cls, x, y): if x 0 or y 0: raise ValueError(Negative values not allowed) return super().__new__(cls)10.3 异步环境下的考虑支持异步初始化的模式class AsyncInit: async def __ainit__(self, config): self.config await load_config(config) def __await__(self): return self.__ainit__().__await__() def __new__(cls, *args, **kwargs): instance super().__new__(cls) coro instance.__ainit__(*args, **kwargs) return coro