[BUG]: Converting Complex non-contiguous ND Arrays fails

[awni]

Problem description

Implictly converting non-contiguous nd arrays to contiguous ones works fine except for complex64 which fails with a type error. My understanding is nanobind should implicitly cast the non-contiguous array to a contiguous one (or at least the behavior should be consistent across types).

TypeError: convert(): incompatible function arguments. The following argument types are supported:
    1. convert(arg: ndarray[writable=False, order='C', device='cpu']) -> None

Invoked with types: ndarray

Reproducible example code

The extension (my_ext.cpp)

#include <iostream>                                  
#include <nanobind/nanobind.h>                       
#include <nanobind/ndarray.h>                        
                                                     
namespace nb = nanobind;                             
                                                     
NB_MODULE(my_ext, m) {                               
    m.def(                                           
        "convert",                                   
        [](nb::ndarray<nb::ro, nb::c_contig, nb::device::cpu>) {
          std::cout << "Worked" << std::endl;        
        },
        nb::arg()                                    
    );  
}  

The Python code:

import my_ext
import numpy as np

c_f32 = np.random.rand(10, 10)
c_c64 = c_f32.astype(np.complex64)

my_ext.convert(c_f32)
my_ext.convert(c_c64)

nc_f32 = c_f32.T
nc_c64 = c_c64.T

my_ext.convert(nc_f32)
my_ext.convert(nc_c64) # <-- Fails

[wjakob]

Thank you for catching this and posting a nice reproducer!

[awni]

Thanks for the quick fix!