How to pass a list of numpy array with pybind

Question

I have a preprocessor written in python. This preprocessor calculate a unknown number of numpy arrays. They are stored inside a list. For further calculation I need to read these list of numpy array in cpp. I dont know howto transform an element of the list into the array type.

main.py

import numpy as np
import myCPPAlgo

my_list = [ np.zeroes(shape=(10, 10), dtype=np.uint32) for _ in range(10)]
myCPPAlgo.call(my_list)

main.cpp

#include <pybind11/pybind11.h>
#include <pybind11/numpy.h>
#include <pybind11/stl.h>

#include <iostream>


namespace py = pybind11;

int call(py::list listOfNumpyArrays)
{
    for( py::handle array: listOfNumpyArrays)
    {
        // howto transform?
        py::array_t<uint32_t> casted_array = ???

        auto requestCastedArray = casted_array.request();
        uint32_t nRows = requestCastedArray.shape[1];
        uint32_t nCols = requestCastedArray.shape[0];
        uint32_t* pBlockedArray = (uint32_t*) requestCastedArray.ptr;
    }
    return 0;
}


PYBIND11_MODULE(myCPPAlgo, m) {
    m.doc() = ""

    m.def("call", &call, "");
}

How can I transform the pybind::handle into the py::array_t?

I want to go the other way: return a list of arrays to Python. Suggestions? — taranaki
– taranaki, Commented Nov 11, 2021 at 19:31

Wim Lavrijsen · Accepted Answer · 2019-11-20 18:44:57Z

4

Simply cast to an array: py::array_t<uint32_t> casted_array = py::cast<py::array>(array);. Full working example (modulo error checking :) ) below.

#include <pybind11/pybind11.h>
#include <pybind11/numpy.h>
#include <pybind11/stl.h>

#include <iostream>

namespace py = pybind11;

int call(py::list listOfNumpyArrays)
{
    for( py::handle array: listOfNumpyArrays)
    {
        // howto transform?
        py::array_t<uint32_t> casted_array = py::cast<py::array>(array);

        auto requestCastedArray = casted_array.request();
        uint32_t nRows = requestCastedArray.shape[1];
        uint32_t nCols = requestCastedArray.shape[0];
        uint32_t* pBlockedArray = (uint32_t*) requestCastedArray.ptr;
        std::cerr << " rows x columns = " << nRows << " x " << nCols << std::endl;
        for (int i = 0; i < nCols; ++i) {
            for (int j = 0; j < nRows; ++j) {
                std::cerr << pBlockedArray[i*nRows + j] << " ";
            }
            std::cerr << '\n';
        }
    }
    return 0;
}

PYBIND11_MODULE(myCPPAlgo, m) {
    m.doc() = "";
    m.def("call", &call, "");
}

and test code:

import numpy as np
import myCPPAlgo

my_list = [ np.ones(shape=(3, 10), dtype=np.uint32) for _ in range(10)]
my_list[9][1, 0] = 42
myCPPAlgo.call(my_list)

answered Nov 20, 2019 at 18:44

Wim Lavrijsen

3,7981 gold badge12 silver badges25 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

SebastianH Over a year ago

Thanks a lot. I am pretty sure, that I have tried some kind of py::cast. Not sure where the differences was.

Collectives™ on Stack Overflow

How to pass a list of numpy array with pybind

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related