如何安装与 Python.h 关联的库

Question

I am working on calling C from Python. My "hello world" program references such functions as `PyUnicode_FromString', whose prototypes I believe are in the header file Python.h. My system has the header file but seems to lack the associated library. I do not know what to install. I already have done 'apt-get install python3-dev." Either I need to install the library or I if it is already there then I need to point to it when I compile. Here is my code:

#include <Python.h>
static PyObject* _hello_world(PyObject* self) {
    return PyUnicode_FromString("hello world");
}
static struct PyMethodDef methods[] ={
    {"hello_world", (PyCFunction)_hello_world, METH_NOARGS,},
    {NULL, NULL}
};

static struct  PyModuleDef module= {
    PyModuleDef_HEAD_INIT,
    "_hello",
    NULL,
    -1,
    methods
};

PyMODINIT_FUNC PyInit_hello(void) {
    return PyModule_Create(&module);
}

Answer 1

The library you're looking for is libpython , which should already be installed. On Debian and Ubuntu, it's in the package libpython3.8 (or similar, depending on the version).

However, you should only have to worry about this if you're embedding Python in a native program. If you're extending Python by writing a module in C, I don't think you need libpython at all, and in any case distutils / setuptools should take care of all those details for you if you're building your module in the usual way as described in the documentation .

---

Side note: it's generally not recommended these days to use the Python C/C++ API at all unless you have a very good reason. If you're just playing around and learning about how the internals of CPython work - fantastic, Carry on, and have fun!

But if you just have some C code that you want to call from a Python program, the best way to do this is generally either to call it with cffi or ctypes , or to write your extension module in Cython . All of these options are easier and less likely to break with future versions of Python, and the first two have the added benefit of working well on other implementations of Python (ie PyPy ).