尝试传递数组时，PyObject_CallObject 返回 NULL

问题描述

我试图将二维数组从 C++ 传递给 Python 函数，但 PyObject_CallObject 函数返回 NULL。这种情况也发生在一维数组的情况下。当我不传递任何参数或者参数只是单个变量而不是数组时，它工作正常。我的代码如下：

#include <stdio.h>
#include <Python.h>
#include <pyhelper.hpp>
#include <string>
#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
#include <numpy/arrayobject.h> // For capturing the output values given by detector Python inference code

using namespace std;

PyObject *PythonInitialize(string script_name,string function_name);

int main(int argc,char *argv[])
{

    if (!Py_IsInitialized())
    {
        Py_InitializeEx(0); // Initialize the Python interpreter
    }

    PyObject *PythonDetectorFunction,*PythonDetectorFunctionArguments;
    PythonDetectorFunction = PythonInitialize("test","getsum"); //getint,getsum

    PythonDetectorFunctionArguments = PyTuple_New(1); // Reference count incremented. Need to handle the reference counting
    if (PythonDetectorFunctionArguments == NULL)
        PyErr_Print();

    if (PyArray_API == NULL)
    {
        import_array();
    }

    float data[2][4] = {{1.2,3.4,5.6,7.8},{1.2,7.8}};
    npy_intp dims[2] = {2,4};

    PythonDetectorFunctionArguments = PyArray_SimpleNewFromData(1,dims,NPY_FLOAT,data);

    PyObject *PythonDetectorFeatureMaps = PyObject_CallObject(PythonDetectorFunction,PythonDetectorFunctionArguments); // Will contain the output given by Python code

    if (PythonDetectorFeatureMaps != NULL)
        printf("PythonDetectorFeatureMaps: %p\n",PythonDetectorFeatureMaps);
    else
        printf("PythonDetectorFeatureMaps is NULL.\n");

    Py_DECREF(PythonDetectorFunction);
    Py_DECREF(PythonDetectorFunctionArguments);
    //Py_DECREF(PythonDetectorFeatureMaps);
}

PyObject *PythonInitialize(string script_name,string function_name)
{
    PyObject *pName,*pModule,*pFunc;
    string PythonCodeImportStatement = "sys.path.append(os.getcwd())";
    PyRun_SimpleString("import sys,os");
    PyRun_SimpleString(PythonCodeImportStatement.c_str()); // Relative path for the python code

    // pNmae is the name of the python script/module to be called
    pName = PyUnicode_FromString(script_name.c_str()); // Reference count incremented. Need to handle the reference counting
    if (pName == NULL)
        PyErr_Print();

    // Getting the Python module reference inside of the C code
    pModule = PyImport_Import(pName);
    if (pModule == NULL)
        PyErr_Print();

    // Python function reference
    pFunc = PyObject_GetAttrString(pModule,function_name.c_str()); // Reference count incremented. Need to handle the reference counting
    if (pFunc == NULL)
        PyErr_Print();

    // Refrence count clean-up
    Py_XDECREF(pName);
    Py_XDECREF(pModule);
    //Py_XDECREF(pFunc);
    return pFunc;
}

test.py 中的 Python 函数：

def getsum(tuple1):
    print('Python function getsum() called')
    return None

有人能指出错误吗？

解决方法

PyObject_CallObject 接受一个可调用对象和一个可调用对象的参数元组，或者 NULL 表示没有参数。您正在尝试将 NumPy 数组而不是参数元组传递给它。