Python C API：py_limited_api的子分类（内置）类型

发布于 2025-02-01 05:18:45 字数 635 浏览 4 评论 0 原文

我正在尝试将现有的C Python扩展程序移植以使用 py_limited_api ，这样我们就不必为我们支持的所有Python版本构建扩展程序。

我面临的问题之一是，该库提供了一种类型，该类型属于Python float 类型。

最后的官方文档中带有示例（）没有提及有限的API，我面临的问题是， pyfloatObject 不是属于稳定的ABI，这是创建基本结构的必要

typedef struct {
    PyFloatObject base;
    int state;
} SubclassedFloat;

条件：使用 py_limited_api ，尤其是 float ，还是不可能？

原文

I am trying to port an existing C Python extension to use the Py_LIMITED_API, so that we don't have to build the extension for all the python versions we support.

One of the issues I'm facing is, that this library provides a type that subclasses the python float type.

The last official documentation with examples ( https://docs.python.org/3.5/extending/newtypes.html#subclassing-other-types) does not mention the limited API, and I'm facing the problem, that PyFloatObject is not part of the stable ABI, which would be necessary to create the base struct:

typedef struct {
    PyFloatObject base;
    int state;
} SubclassedFloat;

Is there a way to subclass a builtin type when using Py_LIMITED_API, especially float, or is this impossible?

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

み格子的夏天 2025-02-08 05:18:45

这在python c extensions Cornord上得到了回答：

基本上，您必须从 pyfloat_type 在运行时获取 __基础化__ 键入属性，以声明您的对象大小：

//Code simplified for brevity

struct SubclassedFloatObject
{
  // put your data members here
};

Py_ssize_t PyFloat_basicsize = PyLong_AsSsize_t(
    PyObject_GetAttrString(
        (PyObject*)&PyFloat_Type,
        "__basicsize__"));
PyType_Spec SubclassedFloat_spec = {
    "SubclassedFloat",                                                      //name
    static_cast<int>(PyFloat_basicsize + sizeof(SubclassedFloatObject)),    //basicsize
    0,                                                                      //itemsize
    Py_TPFLAGS_DEFAULT,                                                     //flags
    SubclassedFloat_slots                                                   //slots; /* terminated by slot==0. */
};

每当您访问对象时，您都必须偏移 pyobject*< pyobject*<<<<<<<<< /代码>此大小的指针：

int SomePythonFunc(PyObject* self)
{
    SubclassedFloat* p_data = (SubclassedFloat*)((uintptr_t)(self) + PyFloat_basicsize);
    // do something with p_data
}

This was answered on the Python c-extensions discord:
https://discord.com/channels/267624335836053506/728390945384431688/981133468652154920

Basically, you have to fetch the __basicsize__ type attribute from PyFloat_Type at runtime to declare your object size:

//Code simplified for brevity

struct SubclassedFloatObject
{
  // put your data members here
};

Py_ssize_t PyFloat_basicsize = PyLong_AsSsize_t(
    PyObject_GetAttrString(
        (PyObject*)&PyFloat_Type,
        "__basicsize__"));
PyType_Spec SubclassedFloat_spec = {
    "SubclassedFloat",                                                      //name
    static_cast<int>(PyFloat_basicsize + sizeof(SubclassedFloatObject)),    //basicsize
    0,                                                                      //itemsize
    Py_TPFLAGS_DEFAULT,                                                     //flags
    SubclassedFloat_slots                                                   //slots; /* terminated by slot==0. */
};

Whenever you access your object, you'll have to offset the PyObject* pointer for this size:

int SomePythonFunc(PyObject* self)
{
    SubclassedFloat* p_data = (SubclassedFloat*)((uintptr_t)(self) + PyFloat_basicsize);
    // do something with p_data
}

回复收藏 0 原文

~没有更多了~