GCC12.2 C++11 智能指针代码走读
以引用计数为基础的智能指针,引用计数的管理逻辑如下:
- 除了初始化对象本身外,每个构造函数(拷贝构造函数除外)还要在堆上创建一个引用计数,用来记录有多少个对象共享状态。当我们创建一个对象时,只有一个对象共享状态,因此将引用计数初始化为 1;
- 拷贝构造函数不分配新的计数器,而是拷贝给定对象的数据成员,包括计数器。拷贝构造函数递增共享的计数器,标记给定对象的状态又被一个新用户所共享
- 析构函数递减计数器,标记共享状态的用户少了一个。如果计数器变为 0,则析构函数析构对象,包括引进计数变量;
- 拷贝赋值运算符递增右侧运算对象的计数器,递减左侧运算对象的计数器。如果左侧运算对象的计数器变为 0,意味着它的共享状态没有用户了,拷贝运算符就必须销毁状态;
- 计数器一般保存在动态内存中。当创建一个对象时,也分配一个新的计数器。当拷贝或赋值对象时,拷贝指向计数器的指针。使用这种方法,副本和原对象都会指向相同的计数器
1、std::shared_ptr
首先我们看看 shared_ptr 相关的类,主要是 __shared_ptr、__shared_count 和 、_Sp_counted_base 这三个。shared_ptr 继承 __shared_ptr,__shared_ptr 有两个成员变量:对象指针和引用计数。引用计数 __shared_count 主要借助 _Sp_counted_base 实现。
/// shared_ptr.h template<typename _Tp> class shared_ptr : public __shared_ptr<_Tp> { /* ... */ friend class weak_ptr<_Tp>; }; /// shared_ptr_base.h template<typename _Tp, _Lock_policy _Lp> class __shared_ptr : public __shared_ptr_access<_Tp, _Lp> { public: using element_type = typename remove_extent<_Tp>::type; /* ... */ element_type* _M_ptr; // Contained pointer. __shared_count<_Lp> _M_refcount; // Reference counter. }; /// shared_ptr_base.h template<_Lock_policy _Lp> class __shared_count { /* ... */ _Sp_counted_base<_Lp>* _M_pi; }; /// shared_ptr_base.h template<_Lock_policy _Lp = __default_lock_policy> class _Sp_counted_base : public _Mutex_base<_Lp> { /* .... */ }
shared_ptr 是一个模板类,继承 __shared_ptr 模板类,具体的实现都在基类,常用的接口如下
replaces the managed object (public member function) | |
swaps the managed objects (public member function) | |
returns the stored pointer (public member function) | |
dereferences the stored pointer (public member function) | |
[operator (C++17) | provides indexed access to the stored array (public member function) |
returns the number of
objects referring to the same managed object (public member function) | |
(until C++20) | checks whether the managed object is managed only by the current
instance (public member function) |
checks if the stored pointer is not null (public member function) | |
provides owner-based ordering of shared pointers (public member function) |
1.1、_Sp_counted_base
_Sp_counted_base 继承于 _Mutex_base,首先分析 _Mutex_base。
1.1.1、_Mutex_base
_Mutex_base 是为了处理多线程加锁问题,_Lock_policy 有三个取值
- _S_single:单线程,不考虑锁
- _S_mutex:使用互斥锁
- _S_atomic:使用原子操作
默认情况下,__default_lock_policy 的值是 _S_atomic(本文也只讨论 _Lock_policy 为 _S_atomic 的情况)
/// concurrence.h // Available locking policies: // _S_single single-threaded code that doesn't need to be locked. // _S_mutex multi-threaded code that requires additional support // from gthr.h or abstraction layers in concurrence.h. // _S_atomic multi-threaded code using atomic operations. enum _Lock_policy { _S_single, _S_mutex, _S_atomic }; // Compile time constant that indicates prefered locking policy in // the current configuration. static const _Lock_policy __default_lock_policy = #ifndef __GTHREADS _S_single; #elif defined _GLIBCXX_HAVE_ATOMIC_LOCK_POLICY _S_atomic; #else _S_mutex; #endif
弄清楚 _Lock_policy 和 __default_lock_policy 后,_Mutex_base 只有在模板参数 _Lp 为 _S_mutex,_S_need_barriers 才会被赋值为 1。
/// shared_ptr_base.h // Empty helper class except when the template argument is _S_mutex. template<_Lock_policy _Lp> class _Mutex_base { protected: // The atomic policy uses fully-fenced builtins, single doesn't care. enum { _S_need_barriers = 0 }; }; template<> class _Mutex_base<_S_mutex> : public __gnu_cxx::__mutex { protected: // This policy is used when atomic builtins are not available. // The replacement atomic operations might not have the necessary // memory barriers. enum { _S_need_barriers = 1 }; };
1.1.2、_Sp_counted_base
_Sp_counted_base 在定义时,模板参数 _Lp 被赋值为 __default_lock_policy,则 _S_need_barriers 为 0。
另外,_Sp_counted_base 有两个表示引用计数的成员变量:_M_use_count 和 _M_weak_count,分别表示当前有多少个 shared_ptr 和 weak_ptr 指向某对象。另外,如果 _M_use_count 不为 0,_M_weak_count 至少为 1。
_Sp_counted_base 不仅充当引用计数功能,还负责管理资源,即
- 当 _M_use_count 递减为 0 时,调用 _M_dispose() 释放 *this 管理的资源
- 当 _M_weak_count 递减为 0 时,调用 _M_destroy() 释放 *this 对象
_M_dispose() 是纯虚函数,由派生类实现。
/// shared_ptr_base.h template<_Lock_policy _Lp = __default_lock_policy> class _Sp_counted_base : public _Mutex_base<_Lp> { public: _Sp_counted_base() noexcept : _M_use_count(1), _M_weak_count(1) { } virtual ~_Sp_counted_base() noexcept { } // Called when _M_use_count drops to zero, to release the resources // managed by *this. virtual void _M_dispose() noexcept = 0; // Called when _M_weak_count drops to zero. virtual void _M_destroy() noexcept { delete this; } /*** 释放引用计数变量内存 ***/ virtual void* _M_get_deleter(const std::type_info&) noexcept = 0; /* ... */ private: _Sp_counted_base(_Sp_counted_base const&) = delete; _Sp_counted_base& operator=(_Sp_counted_base const&) = delete; _Atomic_word _M_use_count; // #shared _Atomic_word _M_weak_count; // #weak + (#shared != 0) };
1)++_M_use_count
递增 _M_use_count 的接口有三个
- _M_add_ref_copy():在 use count 大于 0 时使用(调用者自己确保)
- _M_add_ref_lock():当 use count 为 0,抛出异常
- _M_add_ref_lock():当 use count 为 0,返回 false,否则递增,返回 true
/// shared_ptr_base.h // Increment the use count (used when the count is greater than zero). void _M_add_ref_copy() { __gnu_cxx::__atomic_add_dispatch(&_M_use_count, 1); } // Increment the use count if it is non-zero, throw otherwise. void _M_add_ref_lock() { if (!_M_add_ref_lock_nothrow()) __throw_bad_weak_ptr(); } // Increment the use count if it is non-zero. template<> inline bool _Sp_counted_base<_S_atomic>:: _M_add_ref_lock_nothrow() noexcept { // Perform lock-free add-if-not-zero operation. _Atomic_word __count = _M_get_use_count(); do { if (__count == 0) return false; // Replace the current counter value with the old value + 1, as // long as it's not changed meanwhile. } while (!__atomic_compare_exchange_n(&_M_use_count, &__count, __count + 1, true, __ATOMIC_ACQ_REL, __ATOMIC_RELAXED)); return true; }
2)++_M_weak_count
递增 _M_weak_count 的接口 _M_weak_add_ref。
/// shared_ptr_base.h // Increment the weak count. void _M_weak_add_ref() noexcept { __gnu_cxx::__atomic_add_dispatch(&_M_weak_count, 1); }
3)_M_release()
_M_release() 递减 _M_use_count,当 use count 为 0,调用 _M_release_last_use() 函数。
/// shared_ptr_base.h template<> inline void _Sp_counted_base<_S_atomic>::_M_release() noexcept { _GLIBCXX_SYNCHRONIZATION_HAPPENS_BEFORE(&_M_use_count); #if ! _GLIBCXX_TSAN /* ... */ #endif if (__gnu_cxx::__exchange_and_add_dispatch(&_M_use_count, -1) == 1) { _M_release_last_use(); } }
_M_release_last_use() 函数先调用 _M_dispose() 函数,然后递减 weak count。weak count 递减为 0,调用 _M_destroy() 函数释放 *this 对象。
为什么 user count 递减为 0,需要递减 weak count?原因在于 weak count 的计数方式 #weak + (#shared != 0),如果 user count 递减为 0,shared 就为 0,那么 weak count 应当减 1。
/// shared_ptr_base.h // Called by _M_release() when the use count reaches zero. void _M_release_last_use() noexcept { _GLIBCXX_SYNCHRONIZATION_HAPPENS_AFTER(&_M_use_count); _M_dispose(); /* ... */ if (__gnu_cxx::__exchange_and_add_dispatch(&_M_weak_count, -1) == 1) { _GLIBCXX_SYNCHRONIZATION_HAPPENS_AFTER(&_M_weak_count); _M_destroy(); } }
4)_M_weak_release()
_M_weak_release() 递减 weak count,如果 weak count 为 0,调用 _M_destroy() 释放 *this 对象。
/// shared_ptr_base.h // Decrement the weak count. void _M_weak_release() noexcept { // Be race-detector-friendly. For more info see bits/c++config. _GLIBCXX_SYNCHRONIZATION_HAPPENS_BEFORE(&_M_weak_count); if (__gnu_cxx::__exchange_and_add_dispatch(&_M_weak_count, -1) == 1) { _GLIBCXX_SYNCHRONIZATION_HAPPENS_AFTER(&_M_weak_count); if (_Mutex_base<_Lp>::_S_need_barriers) { // See _M_release(), // destroy() must observe results of dispose() __atomic_thread_fence (__ATOMIC_ACQ_REL); } _M_destroy(); } }
1.1.3、_Sp_counted_ptr
/// shared_ptr_base.h // Counted ptr with no deleter or allocator support template<typename _Ptr, _Lock_policy _Lp> class _Sp_counted_ptr final : public _Sp_counted_base<_Lp> { public: explicit _Sp_counted_ptr(_Ptr __p) noexcept : _M_ptr(__p) { } virtual void _M_dispose() noexcept { delete _M_ptr; } virtual void _M_destroy() noexcept { delete this; } virtual void* _M_get_deleter(const std::type_info&) noexcept { return nullptr; } _Sp_counted_ptr(const _Sp_counted_ptr&) = delete; _Sp_counted_ptr& operator=(const _Sp_counted_ptr&) = delete; private: _Ptr _M_ptr; }; template<> inline void _Sp_counted_ptr<nullptr_t, _S_atomic>::_M_dispose() noexcept { }
1.1.4、_Sp_counted_deleter
_Sp_counted_deleter::_Impl 继承 _Sp_ebo_helper 处理 Deleter 和 Alloc,然后定义一个成员变量 _M_ptr 保存传入指针。
构造和析构函数不同在意,特别分析 _M_dispose 和 _M_destroy 的实现。
/// shared_ptr_base.h // Support for custom deleter and/or allocator template<typename _Ptr, typename _Deleter, typename _Alloc, _Lock_policy _Lp> class _Sp_counted_deleter final : public _Sp_counted_base<_Lp> { class _Impl : _Sp_ebo_helper<0, _Deleter>, _Sp_ebo_helper<1, _Alloc> { typedef _Sp_ebo_helper<0, _Deleter> _Del_base; typedef _Sp_ebo_helper<1, _Alloc> _Alloc_base; public: _Impl(_Ptr __p, _Deleter __d, const _Alloc& __a) noexcept : _Del_base(std::move(__d)), _Alloc_base(__a), _M_ptr(__p) { } _Deleter& _M_del() noexcept { return _Del_base::_S_get(*this); } _Alloc& _M_alloc() noexcept { return _Alloc_base::_S_get(*this); } _Ptr _M_ptr; }; public: using __allocator_type = __alloc_rebind<_Alloc, _Sp_counted_deleter>; // __d(__p) must not throw. _Sp_counted_deleter(_Ptr __p, _Deleter __d) noexcept : _M_impl(__p, std::move(__d), _Alloc()) { } // __d(__p) must not throw. _Sp_counted_deleter(_Ptr __p, _Deleter __d, const _Alloc& __a) noexcept : _M_impl(__p, std::move(__d), __a) { } ~_Sp_counted_deleter() noexcept { } /* ... */ private: _Impl _M_impl; };
_M_dispose() 函数调用 user 传入的 Delter 处理 _M_ptr 指向的对象。_M_get_deleter() 也是返回 user 传入的 Deleter。
/// shared_ptr_base.h virtual void _M_dispose() noexcept { _M_impl._M_del()(_M_impl._M_ptr); } virtual void* _M_get_deleter(const type_info& __ti [[__gnu__::__unused__]]) noexcept { #if __cpp_rtti // _GLIBCXX_RESOLVE_LIB_DEFECTS // 2400. shared_ptr's get_deleter() should use addressof() return __ti == typeid(_Deleter) ? std::__addressof(_M_impl._M_del()) : nullptr; #else return nullptr; #endif }
_M_destroy() 没有特殊的处理,调用析构函数。
/// shared_ptr_base.h virtual void _M_destroy() noexcept { __allocator_type __a(_M_impl._M_alloc()); __allocated_ptr<__allocator_type> __guard_ptr{ __a, this }; this->~_Sp_counted_deleter(); }
1.1.5、_Sp_counted_ptr_inplace
_Sp_counted_ptr_inplace 和 _Sp_counted_deleter 的实现类似,唯一的差别是 _Sp_counted_ptr_inplace::_Impl 成员变量不是 user 传入的指针,而是一块足以放下 _Tp 的一块内存。
/// shared_ptr_base.h template<typename _Tp, typename _Alloc, _Lock_policy _Lp> class _Sp_counted_ptr_inplace final : public _Sp_counted_base<_Lp> { class _Impl : _Sp_ebo_helper<0, _Alloc> { typedef _Sp_ebo_helper<0, _Alloc> _A_base; public: explicit _Impl(_Alloc __a) noexcept : _A_base(__a) { } _Alloc& _M_alloc() noexcept { return _A_base::_S_get(*this); } __gnu_cxx::__aligned_buffer<_Tp> _M_storage; }; public: using __allocator_type = __alloc_rebind<_Alloc, _Sp_counted_ptr_inplace>; // Alloc parameter is not a reference so doesn't alias anything in __args template<typename... _Args> _Sp_counted_ptr_inplace(_Alloc __a, _Args&&... __args) : _M_impl(__a) { // _GLIBCXX_RESOLVE_LIB_DEFECTS // 2070. allocate_shared should use allocator_traits<A>::construct allocator_traits<_Alloc>::construct(__a, _M_ptr(), std::forward<_Args>(__args)...); // might throw } ~_Sp_counted_ptr_inplace() noexcept { } virtual void _M_dispose() noexcept { allocator_traits<_Alloc>::destroy(_M_impl._M_alloc(), _M_ptr()); } // Override because the allocator needs to know the dynamic type virtual void _M_destroy() noexcept { __allocator_type __a(_M_impl._M_alloc()); __allocated_ptr<__allocator_type> __guard_ptr{ __a, this }; this->~_Sp_counted_ptr_inplace(); /* 析构 _M_impl 对象 */ } private: friend class __shared_count<_Lp>; // To be able to call _M_ptr(). // No longer used, but code compiled against old libstdc++ headers // might still call it from __shared_ptr ctor to get the pointer out. virtual void* _M_get_deleter(const std::type_info& __ti) noexcept override { auto __ptr = const_cast<typename remove_cv<_Tp>::type*>(_M_ptr()); // Check for the fake type_info first, so we don't try to access it // as a real type_info object. Otherwise, check if it's the real // type_info for this class. With RTTI enabled we can check directly, // or call a library function to do it. if (&__ti == &_Sp_make_shared_tag::_S_ti() || #if __cpp_rtti __ti == typeid(_Sp_make_shared_tag) #else _Sp_make_shared_tag::_S_eq(__ti) #endif ) return __ptr; return nullptr; } _Tp* _M_ptr() noexcept { return _M_impl._M_storage._M_ptr(); } _Impl _M_impl; };
1.1.6、总结
_Sp_counted_base 不仅充当引用计数功能,还充当内存管理功能。从上面的分析可以看到,_Sp_counted_base 负责释放 user 申请的内存,即
- 当 _M_use_count 递减为 0 时,调用 _M_dispose() 释放 this 管理的资源
- 当 _M_weak_count 递减为 0 时,调用 _M_destroy() 释放 this 对象
1.2、__shared_count
__shared_count 只有一个指针类型成员变量 _M_pi,是基类 _Sp_counted_base<_Lp> 类型的指针。
__shared_count 有意思的是构造函数的实现,通过分析我们将会看到传入不同的参数类型,_M_pi 将会指向具体对象:
- _Sp_counted_ptr,只有管理对象的指针,默认用 delete 释放管理的函数;
- _Sp_counted_deleter,除了管理对象的指针,还有善后函数 deleter(),在析构时调用 deleter(),不再调用 delete 释放对象,相当于用用户指定的方式释放对象;
- _Sp_counted_ptr_inplace,std::make_shared() 申请的对象,管理的对象和引用计数在同一块内存上;
1.2.1、传入裸指针
传入裸指针是指我们自己 new 一个对象,然后用该对象构造 shared_ptr,比如类似如下方式
class Foo { public: Foo(int val): val(val), next(nullptr) {} int val; Foo* next; }; std::shared_ptr<Foo> ptr(new Foo(-1));
此时,_M_pi 指向 _Sp_counted_ptr 类型的对象,调用构造函数如下
/// shared_ptr_base.h template<typename _Ptr> explicit __shared_count(_Ptr __p) : _M_pi(0) { __try { _M_pi = new _Sp_counted_ptr<_Ptr, _Lp>(__p); } __catch(...) { delete __p; __throw_exception_again; } }
1.2.2、指定 Deleter
在构造 shared_ptr 时如果指定了 Deleter,类似于如下方法
struct Deleter { void operator()(Foo* p) const { std::cout << "Call delete from function object...\n"; delete p; } }; std::shared_ptr<Foo> ptr(new Foo(-1), Deleter());
此时,_M_pi 指向 _Sp_counted_deleter 类型的对象,调用构造函数如下
/// shared_ptr_base.h template<typename _Ptr, typename _Deleter, typename _Alloc, typename = typename __not_alloc_shared_tag<_Deleter>::type> __shared_count(_Ptr __p, _Deleter __d, _Alloc __a) : _M_pi(0) { typedef _Sp_counted_deleter<_Ptr, _Deleter, _Alloc, _Lp> _Sp_cd_type; __try { typename _Sp_cd_type::__allocator_type __a2(__a); auto __guard = std::__allocate_guarded(__a2); _Sp_cd_type* __mem = __guard.get(); ::new (__mem) _Sp_cd_type(__p, std::move(__d), std::move(__a)); _M_pi = __mem; __guard = nullptr; } __catch(...) { __d(__p); // Call _Deleter on __p. __throw_exception_again; } }
1.2.3、make_shared
使用 make_shared 的方法构造 shared_ptr 对象时,类似如下
std::shared_ptr<Foo> ptr = std::make_shared<Foo>(-1);
此时,_M_pi 指向 _Sp_counted_ptr_inplace 类型的对象,调用构造函数如下
/// shared_ptr_base.h template<typename _Tp, typename _Alloc, typename... _Args> __shared_count(_Tp*& __p, _Sp_alloc_shared_tag<_Alloc> __a, _Args&&... __args) { typedef _Sp_counted_ptr_inplace<_Tp, _Alloc, _Lp> _Sp_cp_type; typename _Sp_cp_type::__allocator_type __a2(__a._M_a); auto __guard = std::__allocate_guarded(__a2); _Sp_cp_type* __mem = __guard.get(); auto __pi = ::new (__mem) _Sp_cp_type(__a._M_a, std::forward<_Args>(__args)...); __guard = nullptr; _M_pi = __pi; __p = __pi->_M_ptr(); }
1.2.4、std::unique_ptr
当从 unique_ptr 构造 shared_ptr,类似于如下方法
std::unique_ptr<Foo> ptr(new Foo(-1)); std::shared_ptr<Foo> ptr2(ptr);
此时,_M_pi 指向 _Sp_counted_ptr 类型的对象,调用构造函数如下
/// shared_ptr_base.h // Special case for unique_ptr<_Tp,_Del> to provide the strong guarantee. template<typename _Tp, typename _Del> explicit __shared_count(std::unique_ptr<_Tp, _Del>&& __r) : _M_pi(0) { // _GLIBCXX_RESOLVE_LIB_DEFECTS // 2415. Inconsistency between unique_ptr and shared_ptr if (__r.get() == nullptr) return; using _Ptr = typename unique_ptr<_Tp, _Del>::pointer; using _Del2 = __conditional_t<is_reference<_Del>::value, reference_wrapper<typename remove_reference<_Del>::type>, _Del>; using _Sp_cd_type = _Sp_counted_deleter<_Ptr, _Del2, allocator<void>, _Lp>; using _Alloc = allocator<_Sp_cd_type>; using _Alloc_traits = allocator_traits<_Alloc>; _Alloc __a; _Sp_cd_type* __mem = _Alloc_traits::allocate(__a, 1); // _GLIBCXX_RESOLVE_LIB_DEFECTS // 3548. shared_ptr construction from unique_ptr should move // (not copy) the deleter _Alloc_traits::construct(__a, __mem, __r.release(), std::forward<_Del>(__r.get_deleter())); _M_pi = __mem; }
1.2.5、拷贝赋值构造函数
从 __shared_count 的赋值构造函数,可以清楚的看到引起计数的管理:增右侧运算对象的计数器,递减左侧运算对象的计数器。
/// shared_ptr_base.h __shared_count& operator=(const __shared_count& __r) noexcept { _Sp_counted_base<_Lp>* __tmp = __r._M_pi; if (__tmp != _M_pi) { if (__tmp != nullptr) __tmp->_M_add_ref_copy(); if (_M_pi != nullptr) _M_pi->_M_release(); _M_pi = __tmp; } return *this; }
1.3、__shared_ptr
__shared_ptr 有两个数据域,一个存放对象指针,一个存放引用计数相关的结构。
另外,__shared_ptr 继承于 __shared_ptr_access,__shared_ptr_access 是为了重载运算符 operator*、operator-> 和 operator[] 的,这里不做介绍。
/// shared_ptr_base.h template<typename _Tp, _Lock_policy _Lp> class __shared_ptr : public __shared_ptr_access<_Tp, _Lp> { public: using element_type = typename remove_extent<_Tp>::type; /* ... */ friend class __weak_ptr<_Tp, _Lp>; /* ... */ element_type* _M_ptr; // Contained pointer. __shared_count<_Lp> _M_refcount; // Reference counter. };
__shared_ptr 我们重点分析构造函数,在某些构造函数中,_M_enable_shared_from_this_with() 函数。主要是如下几种情况
/// shared_ptr_base.h template<typename _Yp, typename = _SafeConv<_Yp>> explicit __shared_ptr(_Yp* __p) : _M_ptr(__p), _M_refcount(__p, typename is_array<_Tp>::type()) { static_assert( !is_void<_Yp>::value, "incomplete type" ); static_assert( sizeof(_Yp) > 0, "incomplete type" ); _M_enable_shared_from_this_with(__p); } template<typename _Yp, typename _Deleter, typename = _SafeConv<_Yp>> __shared_ptr(_Yp* __p, _Deleter __d) : _M_ptr(__p), _M_refcount(__p, std::move(__d)) { static_assert(__is_invocable<_Deleter&, _Yp*&>::value, "deleter expression d(p) is well-formed"); _M_enable_shared_from_this_with(__p); } template<typename _Yp, typename _Deleter, typename _Alloc, typename = _SafeConv<_Yp>> __shared_ptr(_Yp* __p, _Deleter __d, _Alloc __a) : _M_ptr(__p), _M_refcount(__p, std::move(__d), std::move(__a)) { static_assert(__is_invocable<_Deleter&, _Yp*&>::value, "deleter expression d(p) is well-formed"); _M_enable_shared_from_this_with(__p); } template<typename _Yp, typename _Del, typename = _UniqCompatible<_Yp, _Del>> __shared_ptr(unique_ptr<_Yp, _Del>&& __r) : _M_ptr(__r.get()), _M_refcount() { auto __raw = __to_address(__r.get()); _M_refcount = __shared_count<_Lp>(std::move(__r)); _M_enable_shared_from_this_with(__raw); }
_M_enable_shared_from_this_with() 是 __weak_ptr 的成员函数,函数是为了实现 shared_from_this,稍后会有介绍。
虽然 __shared_count 也保存了对象指针,但是 __shared_ptr 并没有从 __shared_count 获取对象指针,而是使用自己保存的对象指针。
/// shared_ptr_base.h /// Return the stored pointer. element_type* get() const noexcept { return _M_ptr; } /// Return true if the stored pointer is not null. explicit operator bool() const noexcept { return _M_ptr != nullptr; }
另外,参数 __weak_ptr 的构造函数也需要先提示,当我看将 weak_ptr 提升为 shared_ptr 的时候,需要进入这个构造函数。
/// shared_ptr_base.h // This constructor is used by __weak_ptr::lock() and // shared_ptr::shared_ptr(const weak_ptr&, std::nothrow_t). __shared_ptr(const __weak_ptr<_Tp, _Lp>& __r, std::nothrow_t) noexcept : _M_refcount(__r._M_refcount, std::nothrow) { _M_ptr = _M_refcount._M_get_use_count() ? __r._M_ptr : nullptr; }
2、std::weak_ptr
不控制所指向对象生命周期的智能指针,它指向一个 shared_ptr 管理的对象
/// shared_ptr.h template<typename _Tp> class weak_ptr : public __weak_ptr<_Tp> { /* ... */ shared_ptr<_Tp> lock() const noexcept { return shared_ptr<_Tp>(*this, std::nothrow); } }; /// shared_ptr_base.h template<typename _Tp, _Lock_policy _Lp> class __weak_ptr { /* ... */ element_type* _M_ptr; // Contained pointer. __weak_count<_Lp> _M_refcount; // Reference counter. }; /// shared_ptr_base.h template<_Lock_policy _Lp> class __weak_count { /* ... */ _Sp_counted_base<_Lp>* _M_pi; };
和 shared_ptr 设计一样,weak_ptr 只是对 __weak_ptr 的封装,所有实现都在基类,常用的接口如下
releases the ownership of the managed object (public member function) | |
swaps the managed objects (public member function) | |
returns the number of
objects that manage the object (public member function) | |
checks whether the referenced object was already deleted (public member function) | |
creates a
that manages the referenced object (public member function) | |
provides owner-based ordering of weak pointers (public member function) |
2.1、__weak_count
只有一个数据成员 _M_pi,是 _Sp_counted_base 类型,和 __shared_count 一样
/// shared_ptr_base.h template<_Lock_policy _Lp> class __weak_count { /* ... */ private: friend class __shared_count<_Lp>; #if __cplusplus >= 202002L template<typename> friend class _Sp_atomic; #endif _Sp_counted_base<_Lp>* _M_pi; };
__weak_count 主要是对 _M_weak_add_ref() 和 _M_weak_release() 操作引用计数。
/// bits/shared_ptr.h __weak_count(const __shared_count<_Lp>& __r) noexcept : _M_pi(__r._M_pi) { if (_M_pi != nullptr) _M_pi->_M_weak_add_ref(); } __weak_count(const __weak_count& __r) noexcept : _M_pi(__r._M_pi) { if (_M_pi != nullptr) _M_pi->_M_weak_add_ref(); } __weak_count(__weak_count&& __r) noexcept : _M_pi(__r._M_pi) { __r._M_pi = nullptr; } ~__weak_count() noexcept { if (_M_pi != nullptr) _M_pi->_M_weak_release(); } __weak_count& operator=(const __shared_count<_Lp>& __r) noexcept { _Sp_counted_base<_Lp>* __tmp = __r._M_pi; if (__tmp != nullptr) __tmp->_M_weak_add_ref(); if (_M_pi != nullptr) _M_pi->_M_weak_release(); _M_pi = __tmp; return *this; } __weak_count& operator=(const __weak_count& __r) noexcept { _Sp_counted_base<_Lp>* __tmp = __r._M_pi; if (__tmp != nullptr) __tmp->_M_weak_add_ref(); if (_M_pi != nullptr) _M_pi->_M_weak_release(); _M_pi = __tmp; return *this; } __weak_count& operator=(__weak_count&& __r) noexcept { if (_M_pi != nullptr) _M_pi->_M_weak_release(); _M_pi = __r._M_pi; __r._M_pi = nullptr; return *this; } long _M_get_use_count() const noexcept { return _M_pi != nullptr ? _M_pi->_M_get_use_count() : 0; }
2.2 __weak_ptr
有两个数据域,一个存放对象指针,一个存放引用计数相关的结构。
/// shared_ptr_base.h template<typename _Tp, _Lock_policy _Lp> class __weak_ptr { /* ... */ element_type* _M_ptr; // Contained pointer. __weak_count<_Lp> _M_refcount; // Reference counter. };
lock() 是为了提升为 __shared_ptr 对象
/// shared_ptr_base.h __shared_ptr<_Tp, _Lp> lock() const noexcept { return __shared_ptr<element_type, _Lp>(*this, std::nothrow); }
另外就是 _M_assign() 函数,__enable_shared_from_this 会使用它。可能比较好奇,为什么需要 use_count() 函数返回 0 才赋值。
/// shared_ptr_base.h // Used by __enable_shared_from_this. void _M_assign(_Tp* __ptr, const __shared_count<_Lp>& __refcount) noexcept { if (use_count() == 0) { _M_ptr = __ptr; _M_refcount = __refcount; } }
use_count() 是调用 _M_get_use_count() 函数,_M_get_use_count() 会判断 _M_pi 是否为空,为空直接返回 0。
/// shared_ptr_base.h long use_count() const noexcept { return _M_refcount._M_get_use_count(); }
use_count() 返回 0,表示 __weak_ptr 没有指向任何对象。
3、std::unique_ptr
拷贝赋值构造函数和赋值运算符是删除的(移动语义除外)
/// unique_ptr.h template <typename _Tp, typename _Dp = default_delete<_Tp>> class unique_ptr { unique_ptr& operator=(unique_ptr&&) = default; template<typename _Up, typename _Ep> _GLIBCXX23_CONSTEXPR typename enable_if< __and_< __safe_conversion_up<_Up, _Ep>, is_assignable<deleter_type&, _Ep&&> >::value, unique_ptr&>::type operator=(unique_ptr<_Up, _Ep>&& __u) noexcept { reset(__u.release()); get_deleter() = std::forward<_Ep>(__u.get_deleter()); return *this; } /// Reset the %unique_ptr to empty, invoking the deleter if necessary. _GLIBCXX23_CONSTEXPR unique_ptr& operator=(nullptr_t) noexcept { reset(); return *this; } // Disable copy from lvalue. unique_ptr(const unique_ptr&) = delete; unique_ptr& operator=(const unique_ptr&) = delete; };
析构函数释放对象
/// unique_ptr.h ~unique_ptr() noexcept { static_assert(__is_invocable<deleter_type&, pointer>::value, "unique_ptr's deleter must be invocable with a pointer"); auto& __ptr = _M_t._M_ptr(); if (__ptr != nullptr) get_deleter()(std::move(__ptr)); __ptr = pointer(); }
常用的接口如下:
returns a pointer to the managed object and releases the ownership (public member function) | |
replaces the managed object (public member function) | |
swaps the managed objects (public member function) | |
returns a pointer to the managed object (public member function) | |
returns the deleter that is used for destruction of the managed object (public member function) | |
checks if there is an associated managed object (public member function) | |
dereferences pointer to the managed object (public member function) |
4、std::enable_shared_from_this
enable_shared_from_this 字如其名,“只从 this 指针构造一个 shared_ptr 对象”。当然,并不是直接用 this 指针,而是在一个类的(成员函数)内部,构造该类的 shared_ptr 对象,并且和已有 shared_ptr 共享状态。
比如如下代码,直接使用 this 指针构造一个 shared_ptr 对象,bad0 和 bad1 并不是共享状态。
struct Bad { std::shared_ptr<Bad> getptr() { return std::shared_ptr<Bad>(this); } ~Bad() { std::cout << "Bad::~Bad() called\n"; } }; void testBad() { // Bad, each shared_ptr thinks it's the only owner of the object std::shared_ptr<Bad> bad0 = std::make_shared<Bad>(); std::shared_ptr<Bad> bad1 = bad0->getptr(); std::cout << "bad1.use_count() = " << bad1.use_count() << '\n'; } // UB: double-delete of Bad int main() { testBad(); }
将出现 doule-free 问题。
Bad::~Bad() called Bad::~Bad() called *** glibc detected *** ./test: double free or corruption
enable_shared_from_this 就是为了解决这个问题,使得可以在类内部构造一个 shared_ptr 对象,和已经存在的 shared_ptr 共享状态。使用 enable_shared_from_this,我们可以这样实现
class Good : public std::enable_shared_from_this<Good> { public: std::shared_ptr<Good> getptr() { return shared_from_this(); } }; void testGood() { // Good: the two shared_ptr's share the same object std::shared_ptr<Good> good0 = std::make_shared<Good>(); std::shared_ptr<Good> good1 = good0->getptr(); std::cout << "good1.use_count() = " << good1.use_count() << '\n'; } int main() { testGood(); }
good1.user_count() 的值为 2,符合预期。
good1.use_count() = 2
enable_shared_from_this 是如何实现的呢?enable_shared_from_this 拥有一个 weak_ptr
/// shared_ptr_base.h template<typename _Tp> class enable_shared_from_this { protected: constexpr enable_shared_from_this() noexcept { } enable_shared_from_this(const enable_shared_from_this&) noexcept { } enable_shared_from_this& operator=(const enable_shared_from_this&) noexcept { return *this; } ~enable_shared_from_this() { } /* ... */ private: template<typename _Tp1> void _M_weak_assign(_Tp1* __p, const __shared_count<>& __n) const noexcept { _M_weak_this._M_assign(__p, __n); } // Found by ADL when this is an associated class. friend const enable_shared_from_this* __enable_shared_from_this_base(const __shared_count<>&, const enable_shared_from_this* __p) { return __p; } template<typename, _Lock_policy> friend class __shared_ptr; mutable weak_ptr<_Tp> _M_weak_this; };
比较好奇的是,_M_weak_this 什么时候初始化。需要回到 __shared_ptr 构造函数,会调用 _M_enable_shared_from_this_with() 函数。
如果某个类继承于 enable_shared_from_this,_M_enable_shared_from_this_with() 函数定义如下,调用
/// shared_ptr.h template<typename _Yp, typename _Yp2 = typename remove_cv<_Yp>::type> typename enable_if<__has_esft_base<_Yp2>::value>::type _M_enable_shared_from_this_with(_Yp* __p) noexcept { if (auto __base = __enable_shared_from_this_base(_M_refcount, __p)) __base->_M_weak_assign(const_cast<_Yp2*>(__p), _M_refcount); }
如果没有继承 enable_shared_from_this,_M_enable_shared_from_this_with() 函数什么都不做。
/// shared_ptr_base.h template<typename _Yp, typename _Yp2 = typename remove_cv<_Yp>::type> typename enable_if<!__has_esft_base<_Yp2>::value>::type _M_enable_shared_from_this_with(_Yp*) noexcept { }
_M_weak_assign() 函数调用 __weak_ptr::_M_assign() 函数初始化。
/// share_ptr_base.h _M_weak_assign(_Tp1* __p, const __shared_count<>& __n) const noexcept { _M_weak_this._M_assign(__p, __n); }
shared_from_this() 函数就不言而喻了,从 weak_ptr 构造一个 shared_ptr 对象。
/// shared_ptr.h shared_ptr<_Tp> shared_from_this() { return shared_ptr<_Tp>(this->_M_weak_this); } shared_ptr<const _Tp> shared_from_this() const { return shared_ptr<const _Tp>(this->_M_weak_this); }
weak_from_this() 返回 weak_ptr
/// shared_ptr.h weak_ptr<_Tp> weak_from_this() noexcept { return this->_M_weak_this; } weak_ptr<const _Tp> weak_from_this() const noexcept { return this->_M_weak_this; }
从 shared_from_this() 定义可以看到,如果 _M_weak_this 没有初始化,shared_from_this() 会出现错误,比如如下使用方式,shared_from_this() 会抛出异常。
void misuseGood() { // Bad: shared_from_this is called without having std::shared_ptr owning the caller try { Good not_so_good; std::shared_ptr<Good> gp1 = not_so_good.getptr(); } catch(std::bad_weak_ptr& e) { // undefined behavior (until C++17) and std::bad_weak_ptr thrown (since C++17) std::cout << e.what() << '\n'; } }
更好的方法时将构造函数定义为私有,提供统一的创建方法,确保调用 shared_from_this() 之前,一定存在至少一个 shared_ptr 共享对象。
class Best : public std::enable_shared_from_this<Best> { public: std::shared_ptr<Best> getptr() { return shared_from_this(); } // No public constructor, only a factory function, // so there's no way to have getptr return nullptr. [[nodiscard]] static std::shared_ptr<Best> create() { // Not using std::make_shared<Best> because the c'tor is private. return std::shared_ptr<Best>(new Best()); } private: Best() = default; }; void testBest() { // Best: Same but can't stack-allocate it: std::shared_ptr<Best> best0 = Best::create(); std::shared_ptr<Best> best1 = best0->getptr(); std::cout << "best1.use_count() = " << best1.use_count() << '\n'; // Best stackBest; // <- Will not compile because Best::Best() is private. } int main() { testBest(); }
best1.use_count() = 2
5、std::make_shared()
调用 allocate_shared 构造一个 shared_ptr
/// bits/shared_ptr.h template<typename _Tp, typename... _Args> inline shared_ptr<_NonArray<_Tp>> make_shared(_Args&&... __args) { using _Alloc = allocator<void>; _Alloc __a; return shared_ptr<_Tp>(_Sp_alloc_shared_tag<_Alloc>{__a}, std::forward<_Args>(__args)...); }
6、shared_ptr 指针转换
创建新的 std::shared_ptr 的实例,将管理对象的类型从 _Tp1 转换成 _Tp。底层仍然共享管理的对象
/// bits/shared_ptr.h /// Convert type of `shared_ptr`, via `static_cast` template<typename _Tp, typename _Up> inline shared_ptr<_Tp> static_pointer_cast(const shared_ptr<_Up>& __r) noexcept { using _Sp = shared_ptr<_Tp>; return _Sp(__r, static_cast<typename _Sp::element_type*>(__r.get())); } /// Convert type of `shared_ptr`, via `const_cast` template<typename _Tp, typename _Up> inline shared_ptr<_Tp> const_pointer_cast(const shared_ptr<_Up>& __r) noexcept { using _Sp = shared_ptr<_Tp>; return _Sp(__r, const_cast<typename _Sp::element_type*>(__r.get())); } /// Convert type of `shared_ptr`, via `dynamic_cast` template<typename _Tp, typename _Up> inline shared_ptr<_Tp> dynamic_pointer_cast(const shared_ptr<_Up>& __r) noexcept { using _Sp = shared_ptr<_Tp>; if (auto* __p = dynamic_cast<typename _Sp::element_type*>(__r.get())) return _Sp(__r, __p); return _Sp(); }
7、std::shared_ptr<void> 工作方式
下面的方式是可以正常工作的
class Foo { public: Foo(int val): val(val), next(nullptr) { std::cout << "Foo\n"; } ~Foo() { std::cout << "~Foo\n"; } int val; Foo* next; }; struct Deleter { void operator()(Foo* p) const { std::cout << "Call delete from function object...\n"; delete p; } }; int main() { shared_ptr<void> ptr; ptr.reset(new Foo(-1)); return 0; }
即使定义的时候,std::shared_ptr 的类模板类型是 void 类型,我们在 reset() 函数中传入一个 Foo 类型的指针,std::shared_ptr 也可以自动地析构 Foo 的对象。如果是 std::shared_ptr<int> 没有这种用法。
shared_ptr<int> ptr; ptr.reset(new Foo(-1)); // cannot convert ‘Foo*’ to ‘int*’ in initialization
根据分析的继承关系,shared_ptr 继承于 __shared_ptr,回头看一下 __shared_ptr 的实现
/// shared_ptr_base.h template<typename _Tp, _Lock_policy _Lp> class __shared_ptr : public __shared_ptr_access<_Tp, _Lp> { public: using element_type = typename remove_extent<_Tp>::type; /* ... */ element_type* _M_ptr; // Contained pointer. __shared_count<_Lp> _M_refcount; // Reference counter. };
可以看到,__shared_ptr::_M_ptr 跟模板参数类型相关,而 __shared_ptr::_M_refcount 跟模板参数是无关的。所以当模板参数是 void 的时候,void 指针可以指向任何对象,而其他指针则不行。
根据前面的分析 _M_refcount 是负责释放管理的对象的,那即使定义为 std::shared_ptr<void>,也可以释放对象,它是如何做到的?在回头看一下 __shared_count 的定义
/// shared_ptr_base.h template<_Lock_policy _Lp> class __shared_count { /* ... */ _Sp_counted_base<_Lp>* _M_pi; };
_Sp_counted_base 是一个基类,只有两个表示引用计数的成员。如前面所说,_Sp_counted_base 的 _M_release() 会调用派生类的 _M_dispose() 进行对象的释放。并且 __shared_count 会根据不同的传入参数,创建不同的 _Sp_counted_base 对象。接下来分析三个派生类的构造函数。
首先是 _Sp_counted_ptr,_Sp_counted_deleter 和 _Sp_counted_ptr_inplace 三个派生类都是模板类,模板参数 _Ptr 就是实际管理的对象的类型指针。所以即使在定义 std::shared_ptr 指定类模板参数为 void。可以看到 reset() 函数也是模板函数
/// shared_ptr_base.h template<typename _Yp> _SafeConv<_Yp> reset(_Yp* __p) // _Yp must be complete. { // Catch self-reset errors. __glibcxx_assert(__p == nullptr || __p != _M_ptr); __shared_ptr(__p).swap(*this); } template<typename _Yp, typename _Deleter> _SafeConv<_Yp> reset(_Yp* __p, _Deleter __d) { __shared_ptr(__p, std::move(__d)).swap(*this); } template<typename _Yp, typename _Deleter, typename _Alloc> _SafeConv<_Yp> reset(_Yp* __p, _Deleter __d, _Alloc __a) { __shared_ptr(__p, std::move(__d), std::move(__a)).swap(*this); }
不仅如此,__shared_ptr、__shared_count 的有参构造函数都是模板函数。所以通过模板推断,可以推断出 reset() 传入指针的类型,然后传入相应的派生类,因此可以正常析构。
欢迎关注 “源知源为”,阅读更多技术干货
#C++11##智能指针##23届找工作求助阵地#C/C++ 语言基础