typemap not matching in overloaded methods

[karlnapf]

I have a set of methods of the form

class Foo
{
Foo(SGMatrix<float64_t> value);

void bar(const std::string& name, float64_t const & value);
void bar(const std::string& name, SGVector<float64_t> const & value);
void bar(const std::string& name, SGMatrix<float64_t> const & value); // the method I care about here
}

There are (say Python) typemaps for all of the SG*.

Now in Python I do

mat = np.zeros( (2,2), dtype=np.float64)
foo = Foo(mat) # works, i.e. typemap is matched
foo.bar("mat", mat) # does not work

The error message I get for the second call is in the lines of

417:     foo.put("mat", mat)
417: NotImplementedError: Wrong number or type of arguments for overloaded function 'SGObject_put'.
417:   Possible C/C++ prototypes are:
417:     shogun::Foo::bar(std::string const &, float64_t const &)
417:     shogun::Foo::bar(std::string const &, SGVector< float64_t > const &)
417:     shogun::Foo::bar(std::string const &, SGMatrix< float64_t > const &)

Indeed, checking the generated c++ interface code, I see that for the constructor, the typemap is used, in the lines of

SWIGINTERN int _wrap_new_Foo__SWIG_0(PyObject *self, PyObject *args) {
......
    matrix_from_numpy<float64_t>(arg3, obj2, NPY_FLOAT64)

but for the method, this is not the case

SWIGINTERN PyObject *_wrap_Foo_bar__SWIG_5(PyObject *self, PyObject *args) {
......
    arg3 = reinterpret_cast< SGMatrix< float64_t > * >(argp3);

I tried playing around with removing the const references and changing the order of the arguments, but I could not get swig to match the typemap. Is there anything obvious I am doing wrong?

Many thanks!

[karlnapf]

There is some other issue going on, I need to dig a bit before I can pin things down. Sorry