Python cType,C++对象破坏

你可以实现 __del__ 方法调用您必须定义的析构函数:

C++

class A
{
public:
    void someFunc();
};

A* A_new() { return new A(); }
void delete_A(A* obj) { delete obj; }
void A_someFunc(A* obj) { obj->someFunc(); }

from ctypes import cdll

libA = cdll.LoadLibrary(some_path)

class A:
    def __init__(self):
        fun = libA.A_new
        fun.argtypes = []
        fun.restype = ctypes.c_void_p
        self.obj = fun()

    def __del__(self):
        fun = libA.delete_A
        fun.argtypes = [ctypes.c_void_p]
        fun.restype = None
        fun(self.obj)

    def some_func(self):
        fun = libA.A_someFunc
        fun.argtypes = [ctypes.c_void_p]
        fun.restype = None
        fun(self.obj)

self __init__ 方法。此外,还必须显式指定返回类型/参数类型,因为ctypes默认为32位整数,而在现代系统中,指针可能是64位的。

有人认为 __德尔__ with 语法:

class A:
    def __init__(self):
        fun = libA.A_new
        fun.argtypes = []
        fun.restype = ctypes.c_void_p
        self.obj = fun()

    def __enter__(self):
        return self

    def __exit__(self):
        fun = libA.delete_A
        fun.argtypes = [ctypes.c_void_p]
        fun.restype = None
        fun(self.obj)

    def some_func(self):
        fun = libA.A_someFunc
        fun.argtypes = [ctypes.c_void_p]
        fun.restype = None
        fun(self.obj)

with A() as a:
    # Do some work
    a.some_func()

一般来说,DLL应该提供一种方法来清理它们创建的对象。这样,内存分配就封装在dll中。这意味着,您的dll可能应该公开一个类似 void A_delete(A*) .