Anaconda Spyder erratically reset by distutils C extension

[OTAGOHARBOUR]

I am using Anaconda Spyder 3.3.6 (Python 3.7) on CentOS 7. I would like to pass 3D numpy arrays to a C function that processes the input and returns a numpy array as output. The C code is in a module, surfaceModules, which has the following form.

#include <stdio.h>
#include <stdlib.h>
#include "/home/peter/anaconda3/include/python3.7m/Python.h"
#include "numpy/arrayobject.h"

#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION

static PyObject *  squeezeSurfaces(PyObject* self, PyObject* args)
{
    PyArrayObject *matin, *matout;
    double ***cin, ***cout, dfactor;
    int i,j,k,n,m,l, dims[3], ifactor;

    // Parse tuples separately since args will differ between C fcns
    if (!PyArg_ParseTuple(args, "O!id",
        &PyArray_Type, &matin, &ifactor, &dfactor))  return NULL;
    if (NULL == matin)  return NULL;

    return PyArray_Return(matin);
}

static PyMethodDef surfaceMethods[] = {
    {"squeezeSurfaces", squeezeSurfaces, METH_VARARGS, "Squeezes surfaces along normal     vectors"},
    {NULL, NULL, 0, NULL}
};

static struct PyModuleDef surfaceModules = {
    PyModuleDef_HEAD_INIT,
    "surfaceModules",
    "Surface function Module",
    -1,
    surfaceMethods
};

PyMODINIT_FUNC PyInit_surfaceModules(void){
    import_array();
    return PyModule_Create(&surfaceModules);
}

The following code

import surfaceModules as sm
import numpy as NP

x=NP.zeros(27)
for i in range(0,27): x[i]=i 
x=x.reshape(3,3,3)
jfac=2
xfac=1.5
y=sm.squeezeSurfaces(x, jfac, xfac)

sometimes produces this

In [1]: import surfaceModules as sm
import numpy as NP

x=NP.zeros(27)
for i in range(0,27): x[i]=i 

x=x.reshape(3,3,3)
jfac=2
xfac=1.5
y=sm.squeezeSurfaces(x, jfac, xfac)

In [1]: y
Out[2]: 
array([[[ 0.,  1.,  2.],
    [ 3.,  4.,  5.],
    [ 6.,  7.,  8.]],

   [[ 9., 10., 11.],
    [12., 13., 14.],
    [15., 16., 17.]],

   [[18., 19., 20.],
    [21., 22., 23.],
    [24., 25., 26.]]])

which is what I want. However, sometimes, it produces this.

In [4]: import surfaceModules as sm
import numpy as NP

x=NP.zeros(27)
for i in range(0,27): x[i]=i 

x=x.reshape(3,3,3)
jfac=2
xfac=1.5
y=sm.squeezeSurfaces(x, jfac, xfac)

In [1]: y
Traceback (most recent call last):

  File "<ipython-input-1-9063a9f0e032>", line 1, in <module>
    y

NameError: name 'y' is not defined

Note In [1]: y instead of In [5]: y. Seems to reset the Spyder kernel. But seems erratic and unpredictable. I would not want this happening when it is part of larger code.

The same erratic behavior also happens with this code.

import surfaceModules as sm
import numpy as NP

x=NP.zeros(27)
jfac=2
xfac=1.5
y=sm.squeezeSurfaces(x, jfac, xfac)

[OTAGOHARBOUR]

I tried running the following code.

import surfaceModules as sm
import numpy as NP

x=NP.zeros(27)
for i in range(0,27): x[i]=i 
x=x.reshape(3,3,3)
jfac=2
xfac=1.5
y=sm.squeezeSurfaces(x, jfac, xfac)
y

in Jupyter Notebook. Sometimes, I got the desired result, which is a 3D numpy array. Other times I got '_execution_count_validate'. Again, the result appeared to be unpredictable and random.

[casevh]

I think the issue is with your use of PyArg_ParseTuple. The command O! returns a borrowed reference via matin. You are returning the borrowed reference without incrementing it.

Untested...

static PyObject *  squeezeSurfaces(PyObject* self, PyObject* args)
{
    PyArrayObject *matin, *matout;
    double ***cin, ***cout, dfactor;
    int i,j,k,n,m,l, dims[3], ifactor;
 
    // Parse tuples separately since args will differ between C fcns
    if (!PyArg_ParseTuple(args, "O!id",
        &PyArray_Type, &matin, &ifactor, &dfactor))  return NULL;
    if (NULL == matin)  return NULL;
 
    Py_INCREF(matin);
    return PyArray_Return(matin);
}

casevh

[OTAGOHARBOUR]

@casevh. That appears to have worked. Thanks very much

[ThomasL]

x=NP.zeros(27)
for i in range(0,27): x[i]=i 
x=x.reshape(3,3,3)

This code makes my brain bleeding.
1) NO ONE imports numpy as NP, instead it´s worldwide common use to import numpy as np
2) make it a oneliner

x = np.arange(27, dtype=np.float).reshape(3,3,3)