[QUESTION] How to get specific column from multidimensional py::array

[scotie]

Hi,

I have a python script with a function that returns a 2-dimentional array and looks something like this:
[[ 269.83 1391. ] [ 270.17347367 1483. ] [ 270.51702984 1492. ] ... [1118.45796377 1537. ] [1118.93168884 1541. ] [1119.40545879 1525. ]]

I want to be able to get the whole column 1 ( data[:,1] ), preferably without iterating over the array and convert it to std::vector.

Thank you

[EricCousineau-TRI]

Have you tried it out via the Python API via C++?
https://pybind11.readthedocs.io/en/stable/advanced/pycpp/object.html
Somewhat related: https://pybind11.readthedocs.io/en/stable/advanced/pycpp/numpy.html#ellipsis

You should be able to use slice(None) as a stand-in for :. Then leveraging example from above docs for NumPy:

py::array a = ...;
py::array b = a[py::make_tuple(py::slice(nullopt, nullopt, nullopt), 1)]; 

[scotie]

Thanks, I'll give it a go.

For now, I did something like
`auto r = x.unchecked<2>();

for (py::ssize_t i = 0; i < r.shape(0); i++)
{
vec1.push_back(r(i, 0));
vec2.push_back(r(i, 1));
}`

But it seems hacky to me.