Passing numpy array element (int) to c++ int using SWIG

Question

I'd like to pass an integer element from a numpy array in python to a c++ function that catches it as a c++ integer using SWIG.

What am I missing here?

add_vector.i

%module add_vector
%{
    #define SWIG_FILE_WITH_INIT
    #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION  // gets rid of warning
    #include "add_vector.h"
%}

%include "numpy.i"
%init %{
import_array();
%}

%include "add_vector.h"

add_vector.h

#include <iostream>

void print_int(int x);

add_vector.cpp

#include "add_vector.h"

void print_int(int x) {
    std::cout << x << std::endl;
}

tester.py

import add_vector as vec
import numpy as np

a = np.array([1,2,3])
print(a[1])
vec.print_int(a[1])

OUTPUT

2
Traceback (most recent call last):
  File "tester.py", line 6, in <module>
    vec.print_int(a[1])
TypeError: in method 'print_int', argument 1 of type 'int'

Reading from the numpy.i manual (https://docs.scipy.org/doc/numpy-1.13.0/reference/swig.interface-file.html#numpy-array-scalars-and-swig), I copied the pyfragments.swg file into my working directory, but nothing changed.

I've also tried a number of %apply directives both for passing an int and an int *, but that hasn't yet changed anything. I keep getting the same TypeError I listed above.

Versions: numpy 1.17.3 ; swig 2.0.12 ; python 3.7.3 ; numpy.i is being copied into my working directory from: /usr/lib/python2.7/dist-packages/instant/swig/numpy.i

That's the problem then, but I'm less sure what the right fix for numpy is. Maybe make the function take int64_t from #include <stdint.h>? — Flexo - Save the data dump
– Flexo - Save the data dump ♦, Commented Oct 23, 2019 at 20:07
you mean change my function to: void print_int(int64_t x); ? When I do that I just get this: TypeError: in method 'print_int', argument 1 of type 'int64_t' — Otherness
– Otherness, Commented Oct 23, 2019 at 20:25
Update: I've now copied other swig/numpy.i code on my machine that works for others and I always get the same TypeError. I'm thinking that it's very likely to be the fact that I'm copying over numpy.i from a python2.7 folder, or that there's some other version compatibility issue. — Otherness
– Otherness, Commented Oct 23, 2019 at 21:45

Otherness · Accepted Answer · 2019-10-24 22:19:28Z

4

Solved! There were 3 issues:

The numpy.i file I copied over isn't compatible, and the compatible version isn't included in the installation package when you go through anaconda (still not sure why they'd do that).

Answer: Find which version of numpy you're running, then go here (https://github.com/numpy/numpy/releases) and download the numpy-[your_version].zip file, then specifically copy the numpy.i file, found in numpy-[your_version]/tools/swig/. Now paste that numpy.i into your project working directory.

As a default, numpy makes integers of type long. So in tester.py file, I needed to write: a = np.array([1,2,3], dtype=np.intc)
Need to convert numpy int to c++ int in add_vector.i. This can be done by using the %apply directive right above the %include "add_vector.h" line: %apply (int DIM1) {(int x)};

answered Oct 24, 2019 at 22:19

Otherness

4053 silver badges25 bronze badges

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

Collectives™ on Stack Overflow

Passing numpy array element (int) to c++ int using SWIG

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related