Reputation: 975
I have a c++ class that I am trying to wrap for Python using SWIG. I am having trouble trying to wrap one of the functions which takes an array as input.
Here's the header file,
class dyndiff_data_t
{
private:
double H[3];
double GAMMA;
double k;
double P;
public:
dyndiff_data_t(double H_[3],
const double GAMMA_,
const double k_,
const double P_);
void test();
};
and here's the swig interface file,
%module twowave
%{
#define SWIG_FILE_WITH_INIT
#include <twowave.h>
%}
%include "numpy.i"
%init %{
import_array();
%}
%apply (double IN_ARRAY1[3]) {(double H_[3])};
%include <twowave.h>
The problem is that for the array input, SWIG complains that there is no typemap. I don't understand why. The numpy.i file was taken from here and the typemap I am using is described here
Any help would be appreciated.
Upvotes: 5
Views: 3579
Reputation: 1
This works for me:
for
void get_position(double outarray[3])
use
%apply (double ARGOUT_ARRAY1[ANY]) {(double outarray[3])};
(note [ANY] and [3])
result in
a.get_position()
Out[2]: array([0., 0., 0.])
Upvotes: 0
Reputation: 7946
The problem is that the typemap in numpy.i defines a two argument typemap, and you're trying to apply it to a single argument. This would work if you had parameters int len1, and double* vec1 in your function:
%apply (int DIM1, double* IN_ARRAY1) {(int len, double* H_)}
Rather than writing your own typemap, just use carrays.i.
If you WERE to write a typemap, e.g. to take a tuple of doubles as input, it would look something like:
%typemap(in) double TUPLE[ANY]
{
...
}
in which case you would apply it to your function the way you expect.
%apply double TUPLE[3] {double H_[3]}
A good place to start when trying to figure out why you can't use a typemap is to run SWIG with the -tmsearch
option. It will tell you what it's looking for when trying to match your function parameters.
Upvotes: 3