4

Recently I wanted to call Python from Fortran (see here) using a C wrapper. Below I post a simpler example. Second call of the function results in Segmentation fault at pModule = PyImport_Import(pName). I figured out that the problem is with the from scipy.optimize import newton line - if I comment it everything works fine. Any ideas how to fix it?

rootC.c

#include "rootC.h"
#include <Python.h>

void root_(double* A, double* B, double* t, double* x)
{
    PyObject *pName, *pModule, *pFunc;
    PyObject *pArgs, *pValue, *sys, *path;

    Py_Initialize();

    sys  = PyImport_ImportModule("sys");
    path = PyObject_GetAttrString(sys, "path");
    PyList_Append(path, PyString_FromString("."));

    pName = PyString_FromString("rootPY");
    pModule = PyImport_Import(pName);

    if (!pModule)
    {
        PyErr_Print();
        printf("ERROR in pModule\n");
        exit(1);
    }

    pFunc = PyObject_GetAttrString(pModule, "root");
    pArgs = PyTuple_New(3);
    PyTuple_SetItem(pArgs, 0, PyFloat_FromDouble((*A)));
    PyTuple_SetItem(pArgs, 1, PyFloat_FromDouble((*B)));
    PyTuple_SetItem(pArgs, 2, PyFloat_FromDouble((*t)));

    pValue = PyObject_CallObject(pFunc, pArgs);
    *x     = PyFloat_AsDouble(pValue);

    Py_Finalize();
}

rootC.h

#ifndef ROOT_H_
#define ROOT_H_

void root_(double*, double*, double*, double*);

#endif

rootPY.py

from mpmath import polylog, exp
from scipy.optimize import newton

def root(A,B,t):   
    return 1

main.c

#include "rootC.h"
#include <stdio.h>

int main() 
{
    double A = 0.4, B = 0.3, t = 0.1, x = 0.0;

    root_(&A,&B,&t,&x);
    printf("x = %.15f\n", x);

    root_(&A,&B,&t,&x);
    printf("x = %.15f\n", x);

    return 0;
}

Makefile

CC = gcc
FC = gfortran
CFLAGS = -I/usr/include/python2.7
LFLAGS = -L/usr/local/lib -lpython2.7 -lm

.PHONY: all clean

all: main

main: main.o rootC.o
    $(CC) $^ -o $@ $(LFLAGS)

main.o: main.c
    $(CC) $(CFLAGS) -c $< -o $@

rootC.o: rootC.c
    $(CC) $(CFLAGS) -c $< -o $@

clean:
    rm -f *.o
4
  • Why are you calling PyImport_Import instead of PyImport_Module as you do for the other modules? Also, I would check functions returning pointers for NULL unless the documentation explicitly says that it cannot be NULL. Commented Dec 19, 2015 at 13:33
  • I changed line PyImport_ImportModule("rootPy"), but still the same problem and additionally it hangs a little during the second call before I get the result. Commented Dec 19, 2015 at 13:47
  • I see that you don't have warnings enabled in your flags, why? They can help you spot stupid mistakes (that everyone makes). Add -Wall -Wextra -Werror to your CFLAGS and see if it helps. Commented Dec 19, 2015 at 13:57
  • As noted here and here numpy doesn't like to be initialized twice. Commented Dec 19, 2015 at 17:39

2 Answers 2

3

One can also do something like this:

#include "rootC.h"
#include <Python.h>
#include <stdlib.h>

void  Initialize ()
{ 
    Py_Initialize();
}

void Finalize ()
{
    Py_Finalize();
}

void root_(double* A, double* B, double* t, double* x)
{
    PyObject *pName, *pModule, *pFunc;
    PyObject *pArgs, *pValue, *sys, *path;
    static int i;

    if (i == 0)
    {   
        ++i;
        Initialize();
        atexit(Finalize);
    }

    ....
}
Sign up to request clarification or add additional context in comments.

Comments

2

The issue is with initializing and finalizing your python shell every time the function is called. (it is trying to initialise a .dll more than once)

Edit: Link to resource

This is my fix... ( not the best code for the job) in your main.c

int main() 
{
    double A = 0.4, B = 0.3, t = 0.1, x = 0.0;
    bool Stop = false;
    root_(&A,&B,&t,&x,&Stop);
    printf("x = %.15f\n", x);

    Stop=true;
    root_(&A,&B,&t,&x,&Stop);
    printf("x = %.15f\n", x);



    return 0;
}

then in your rootC.h

void root_(double*, double*, double*, double*,bool*);

then in rootC.c

#include "rootC.h"
#include <Python.h>

void root_(double* A, double* B, double* t, double* x,bool* Stop)
{
    PyObject *pName, *pModule, *pFunc;
    PyObject *pArgs, *pValue, *sys, *path;
    if (*Stop==false)
    {
        Py_Initialize();
    }
    sys  = PyImport_ImportModule("sys");
    path = PyObject_GetAttrString(sys, "path");
    PyList_Append(path, PyString_FromString("."));

    pName = PyString_FromString("rootPY");
    pModule = PyImport_Import(pName);

    if (!pModule)
    {
        PyErr_Print();
        printf("ERROR in pModule\n");
        exit(1);
    }

    pFunc = PyObject_GetAttrString(pModule, "root");
    pArgs = PyTuple_New(3);
    PyTuple_SetItem(pArgs, 0, PyFloat_FromDouble((*A)));
    PyTuple_SetItem(pArgs, 1, PyFloat_FromDouble((*B)));
    PyTuple_SetItem(pArgs, 2, PyFloat_FromDouble((*t)));

    pValue = PyObject_CallObject(pFunc, pArgs);
    *x     = PyFloat_AsDouble(pValue);
    if (*Stop==true)
    {
        Py_Finalize();
    }
}

This should work :)

2 Comments

This answer show an alternative to the Stop variable. And do you really need to pass Stop by-reference?
i think the above is just what happens when i try to write code after 3am

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.