关于gcc以及python的版本配置建议大家去看我之前的一篇文章,里面比较详细地点明了关于32位和64的一些坑。
接下来将简单介绍python如何调用C和C++。包括调用整型,数组等情况。
python 调用C
C函数返回整型int
c代码 test.c
#include <stdio.h>
#include <stdlib.h>
int func(int a, int b)
{
int ans = a + b;
printf("You input %d and %d\n", a, b);
printf("The result is %d\n", ans);
return ans;
}
编译成so包
gcc -o testpycll_64.so -shared -fPIC test.c
其中参数‘-fPIC’:当产生共享库的时候,应该创建位置无关的代码,这会让共享库使用任意的地址而不是固定的地址,要实现这个功能,需要使用-fPIC参数。
python代码
import ctypes
import platform
ll = ctypes.cdll.LoadLibrary
lib = ll("./testpycll_64.so")
lib.func(1,3)
# platform.architecture()
运行之后的结果:
这里有一个问题,我是在jupyter notebook下的,只显示了C函数的返回值,对于过程打印语句并没有输出。但在命令行下是可以输出的,如下所示:
mypy.py 是python文件名。
C函数返回整型数组 int[]
C代码:
#include <stdio.h>
#include <stdlib.h>
int* func(int a, int b)
{
int* res = (int *)malloc(3*sizeof(int));
int ans = a + b;
printf("You input %d and %d\n", a, b);
printf("The result is %d\n", ans);
res[0] = a;
res[1] = b;
res[2] = ans;
return res;
//return ans;
}
python 代码:需要用到numpy包
(这里原先一直返回失败,也是试了非常多的方法才解决)
import ctypes
import platform
from numpy.ctypeslib import ndpointer
ll = ctypes.cdll.LoadLibrary
lib = ll("./testpycll_64.so")
lib.func.restype = ndpointer(dtype=ctypes.c_int, shape=(3,))
res = lib.func(1,3)
print(res)
运行之后的结果:
C函数传入整型数组 int[]
C代码
#include <stdio.h>
#include <stdlib.h>
int func(int a[], int n)
{
int sum = 0;
for(int i=0; i<n; i++)
sum += a[i];
return sum;
}
python代码:
有多种写法,这里给出几种:
(1)
ll = ctypes.cdll.LoadLibrary
lib = ll("./testpycll_64.so")
INPUT = ctypes.c_int * 4;
# 实例化一个长度为2的整型数组
inputdata = INPUT();
inputdata[0] = 1;
inputdata[1] = 2;
inputdata[2] = 3;
inputdata[3] = 4;
lib.func.restype = ctypes.c_int;
res = lib.func(inputdata, 4);
# res = lib.func2(1,2);
print(res)
(2)
ll = ctypes.cdll.LoadLibrary
lib = ll("./testpycll_64.so")
pyarray = [1, 2, 3, 4, 5]
carray = (ctypes.c_int * len(pyarray))(*pyarray)
res = lib.func(carray, len(pyarray))
print(res)
(3)搭配numpy数组
import ctypes
import platform
from numpy.ctypeslib import ndpointer
import numpy as np
ll = ctypes.cdll.LoadLibrary;
lib = ll("./testpycll_64.so");
pyarray = np.array([1,2,3,4,5,6,7,8], dtype="int32");
lib.func.argtypes = [ndpointer(ctypes.c_int, shape=(8, )), ctypes.c_int];
lib.func.restype = ctypes.c_int;
res = lib.func(pyarray, len(pyarray))
# print (np.array(pyarray))
print(res)
其余的数据类型就举一反三啦,ctypes下有很多数据类型,例如c_int, c_bool等,大家可以自己去尝试。
python 调用C++
调用方法和C类似,只是需要加上
extern "C"
{
...
}
C++ 代码
#include <iostream>
#include <string>
using namespace std;
class TestClass
{
public:
char* func();
};
char* TestClass::func()
{
cout << "In func without argument" << endl;
char* temp = "In func without argument";
return temp;
}
extern "C"
{
TestClass A;
char* myfunc1()
{
return A.func();
}
}
生成so包指令:注意变成g++
g++ -o testpycll_64.so -shared -fPIC test.cpp
python代码
import ctypes
import platform
from numpy.ctypeslib import ndpointer
import numpy as np
ll = ctypes.cdll.LoadLibrary;
lib = ll("./testpycll_64.so");
lib.myfunc1.restype = ctypes.c_uint64;
res = lib.myfunc1();
res = ctypes.string_at(res);
print(res);
运行结果:
以上大致是python调用C/C++的方法,有不足之处还望大家指出。