Zero appears when converting armadillo vectors to ndarray

[HPDell]

Hi,

I'm working on a Python package based on nanobind and a 3rd-party C++ library libgwmodel. Everything else works fine but when I try to convert a vector of armadillo to ndarray, the first two elements both become 0. However, when a matrix is converted to ndarray, everything works normally. I tried to convert the armadillo vector to ndarry of shape (n, 1), the first two elements are still 0.

The following is a part of my code

using nvec = nb::ndarray<nb::numpy, const double, nb::shape<nb::any>>;
using nmat = nb::ndarray<nb::numpy, const double, nb::shape<nb::any, nb::any>, nb::f_contig>;

inline arma::mat as(nmat src)
{
    return arma::mat(src.data(), src.shape(0), src.shape(1));
}

inline arma::vec as(nvec src)
{
    return arma::vec(src.data(), src.shape(0));
}

inline auto wrap(const arma::mat& src)
{
    return nmat(src.memptr(), { src.n_rows, src.n_cols }, nb::handle(), { 1, int64_t(src.n_rows) });
}

inline auto wrap(const arma::vec& src)
{
    return nvec(src.memptr(), { src.n_elem }, nb::handle(), { 1 });
}

A vector should be

array([178500., 448000., 172000., 210000., 110000., 199500., 113000.,
       230000., 224000., 249950., 188000., 182000., 317500., 121000.,
       137000., 187500., 144500., 185000., 154000., 157500., 350000.,
       158000., 199000.,  75000., 350000., 118600.,  94000., 160000.,
        95000., 114111., 140000., 195500., 157000., 180000., 137000.,
        87000., 303000., 300000., 115500.,  87500., 124786., 152500.,
       110750., 113500., 150000., 110000., 138000.,  99000.,  87500.,
       112000., 195000., 145000., 147500., 159000., 180000., 425000.,
       225000., 133750., 126500., 180000.,  75000., 140000., 172000.,
       295000., 140000., 114000.,  89000., 285000., 134000., 125000.,
       177500., 137000., 207500., 195000., 123500., 150000., 104500.,
       525000., 185000.,  93000., 220000.,  55000., 360000., 165000.,
       135000., 207000., 203000., 209950., 145000., 255000., 215000.,
       279950., 153825., 189000., 120000., 147500., 278000., 168000.,
       389950., 110000.])

But after the conversion from armadillo vector to ndarray, it becomes

array([     0.,      0., 172000., 210000., 110000., 199500., 113000.,
       230000., 224000., 249950., 188000., 182000., 317500., 121000.,
       137000., 187500., 144500., 185000., 154000., 157500., 350000.,
       158000., 199000.,  75000., 350000., 118600.,  94000., 160000.,
        95000., 114111., 140000., 195500., 157000., 180000., 137000.,
        87000., 303000., 300000., 115500.,  87500., 124786., 152500.,
       110750., 113500., 150000., 110000., 138000.,  99000.,  87500.,
       112000., 195000., 145000., 147500., 159000., 180000., 425000.,
       225000., 133750., 126500., 180000.,  75000., 140000., 172000.,
       295000., 140000., 114000.,  89000., 285000., 134000., 125000.,
       177500., 137000., 207500., 195000., 123500., 150000., 104500.,
       525000., 185000.,  93000., 220000.,  55000., 360000., 165000.,
       135000., 207000., 203000., 209950., 145000., 255000., 215000.,
       279950., 153825., 189000., 120000., 147500., 278000., 168000.,
       389950., 110000.])

And when I read for multiple times, the returned matrices are not the same.

I tried to remove the wrap(vec&) to convert vectors like matrices (via wrap(mat&) as vec is a subclass of mat), but it won't help.

Could anyone give me some clues? I appreciate it very much.

[HPDell]

I solved this issue by create a custom caster for armadillo mat and use the move rv policy.