Ruven Guna
Ruven Guna

Reputation: 424

Calling a C function in Python and returning 2 values

I am trying to figure out how to return 2 vales from a C function that I called in python. I have read through the material online and am using struct to output the two variables. I am able to output the variables when I call this function in the same C file. However, when I try to call it in python, it still only returns one value.

This is my C code:

struct re_val {

    double predict_label;
    double prob_estimates;  

};

struct re_val c_func(const char* dir, double a, double b, double c, double d )
{
    double x[] = {a,b,c,d};

    printf ("x[0].index: %d \n", 1);
    printf ("x[0].value: %f \n", x[0]);

    printf ("x[1].index: %d \n", 2);
    printf ("x[1].value: %f \n", x[1]);

    printf ("x[2].index: %d \n", 3);
    printf ("x[2].value: %f \n", x[2]);

    printf ("x[3].index: %d \n", 4);
    printf ("x[3].value: %f \n", x[3]);

    printf ("\nThis is the Directory: %s \n", dir); 

    struct re_val r;
    r.predict_label = 5.0;
    r.prob_estimates = 8.0;   

    return r;

}

This is my Python code:

calling_function = ctypes.CDLL("/home/ruven/Documents/Sonar/C interface/Interface.so")
calling_function.c_func.argtypes = [ctypes.c_char_p, ctypes.c_double, ctypes.c_double, ctypes.c_double, ctypes.c_double]
calling_function.c_func.restype =  ctypes.c_double
q = calling_function.c_func("hello",1.3256, 2.45, 3.1248, 4.215440)
print q

Currently, when I run my python file in the terminal it outputs this:

x[0].index: 1 
x[0].value: 1.325600 

x[1].index: 2 
x[1].value: 2.450000 

x[2].index: 3 
x[2].value: 3.124800 

x[3].index: 4 
x[3].value: 4.215440 

This is the Directory: hello 

5.0

Instead,I would like it to output this:

x[0].index: 1 
x[0].value: 1.325600 

x[1].index: 2 
x[1].value: 2.450000 

x[2].index: 3 
x[2].value: 3.124800 

x[3].index: 4 
x[3].value: 4.215440 

This is the Directory: hello 

5.0
8.0

Upvotes: 2

Views: 1508

Answers (1)

hetepeperfan
hetepeperfan

Reputation: 4431

Your C code is fine, the problem you are experiencing is in how you use python ctypes. You should tell that the function returns a struct re_val and not a double:

calling_function.c_func.restype =  ctypes.c_double

The above makes the function return a single double value in the eyes of ctypes. You should tell python that the function returns a structure:

import ctypes as ct

# Python representation of the C struct re_val
class ReVal(ct.Structure):
    _fields_ = [("predict_label", ct.c_double),("prob_estimates", ct.c_double)]

calling_function = ctypes.CDLL("/home/ruven/Documents/Sonar/C interface/Interface.so")
calling_function.c_func.argtypes = [ctypes.c_char_p, ctypes.c_double, ctypes.c_double, ctypes.c_double, ctypes.c_double]
# and instead of c_double use:
calling_function.c_func.restype = ReVal

This way you tell python's ctypes that the function returns a aggregate object that is a subclass of ctypes.Structure that matches the struct re_val from the c library.

NOTE Be very carefull with argtypes and restype, if you use these incorrectly it is easy to crash the python interpreter. Then you get a segfault instead of a nice traceback.

Upvotes: 3

Related Questions