python 将结构和指针传递给 c 函数以进行输入和输出

Question

我按如下操作，报错

使用ctypes有什么问题吗，或者我传递的参数类型有问题？

我觉得可能有三点需要思考：

在python.
将结构类型的指针传递给c函数。
正确实例化结构取决于给定的数组，如果在python代码中需要结构。

我是 c 的新手，尤其是指针。

c code:the 结构，ansfer.h

typedef unsigned char boolean_T;


struct emxArray_real_T
{


    double *data;
    int *size;
    int allocatedSize;
    int numDimensions;
    boolean_T canFreeData;
};

python、ansfer.c

调用的函数

void ansferA(const emxArray_real_T *dataArray, double H, emxArray_real_T *TE,
                 emxArray_real_T *Lag)
{

...

P2 += dataArray->data[a + dataArray->size[0] * b];

...

TE->data[a + TE->size[0] * b] = te;

Lag->data[a + Lag->size[0] * b] = h;

...
}

然后，gcc -o libansfer.so -shared -fPIC *.c

libansfer.so生成。

我根据建议重写了test.py。添加结构如下。

python代码：test.py

import numpy as np
import ctypes

c_double_p = ctypes.POINTER(ctypes.c_double)
c_int_p = ctypes.POINTER(ctypes.c_int)

class emxArray_real_T(ctypes.Structure):
    _fields_ = [
        ("data",          c_double_p),  
        ("size",          c_int_p), 
        ("allocatedSize", ctypes.c_int),
        ("numDimensions", ctypes.c_int),
        ("canFreeData",   ctypes.c_bool)
    ]

indata = np.array([[1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1,1.1],[2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2,2.2],[3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,3.3,2.2,2.2,2.2]])

LL, CC = indata.shape 

TE = np.zeros((LL,LL), dtype=np.double)
Lag = np.zeros((LL,LL), dtype=np.double)

SS = np.array([LL, CC])
instruct = emxArray_real_T()
instruct.data = c_double_p(ctypes.c_double(indata.ctypes.data))
instruct.size = c_int_p(ctypes.c_int(SS.ctypes.data))
instruct.allocatedSize = ctypes.c_int(LL*CC)
instruct.numDimensions = ctypes.c_int(2)
instruct.canFreeData = ctypes.c_bool(0)

ss = np.array([LL, LL])
outstruct1 = emxArray_real_T()
outstruct1.data = c_double_p(ctypes.c_double(TE.ctypes.data))
outstruct1.size = c_int_p(ctypes.c_int(ss.ctypes.data))
outstruct1.allocatedSize = ctypes.c_int(LL*LL)
outstruct1.numDimensions = ctypes.c_int(2)
outstruct1.canFreeData = ctypes.c_bool(0)

outstruct2 = emxArray_real_T()
outstruct2.data = c_double_p(ctypes.c_double(Lag.ctypes.data))
outstruct2.size = c_int_p(ctypes.c_int(ss.ctypes.data))
outstruct2.allocatedSize = ctypes.c_int(LL*LL)
outstruct2.numDimensions = ctypes.c_int(2)
outstruct2.canFreeData = ctypes.c_bool(0)

lib = ctypes.cdll.LoadLibrary('./libansfer.so')

lib.ansferA(instruct,ctypes.c_double(3), outstruct1, outstruct2)

但类似的错误

Traceback (most recent call last):
  File "C:\Python27\testfile\test.py", line 106, in <module>
    lib.ansferA(instruct,ctypes.c_double(3), outstruct1, outstruct2)
WindowsError: exception: access violation writing 0x0000000000000000

Answer 1

这不是答案，但格式规则不允许我将其作为评论（并且仍然可读）。

该结构看起来像这样：

c_double_p = ctypes.POINTER(ctypes.c_double)
c_int_p = ctypes.POINTER(ctypes.c_int)

class emxArray_real_T(ctypes.Structure):
    _fields_ = [
        ("data",          c_double_p),  
        ("size",          c_int_p), 
        ("allocatedSize", ctypes.c_int),
        ("numDimensions", ctypes.c_int),
        ("canFreeData",   ctypes.c_bool)
    ]

我不得不猜测 canFreeData 的类型，boolean_T 不是标准的 C 类型。

然后用类似的东西初始化结构：

emxArray_struct = emxArray_real_T( insert values here as parameters )

我没有使用过 numpy，所以我不知道 if/how 它的类型映射到 ctypes 或常规 python 变量类型。

python 将结构和指针传递给 c 函数以进行输入和输出

python pass struct and pointer to c function for input and output

c

c++

python

ctypes