python为输入和输出传递结构和指向c函数的指针

Question

我这样做，但得到了错误。在

使用ctypes有什么错误吗，或者我传递的参数类型有什么问题吗？在

我认为应该考虑三点：

1. 在python中生成传递给c函数的正确参数。

2. 将struct类型的指针传递给c函数。

3. 如果需要python代码中的struct，则右实例化struct依赖于给定的数组。

我是c新手，尤其是指针。在

c代码：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调用的函数ansfer.c

^{pr2}$

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

在利班斯福生成。在

我重写了测试.py根据建议。按如下方式添加结构。在

python代码：测试.py在

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

这不是一个答案，但格式化规则不允许我将其作为注释放置（并且仍然可读）。在

结构如下所示：

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。在

然后用如下方法初始化结构：

^{pr2}$

我没有使用过numpy，所以我不知道它的类型是否/如何映射到ctypes或常规python变量类型。在