如何在 boost multi_array 容器中存储值？

Question

我正在努力访问这些值并将它们存储在 boost multi_array 容器中。我尝试使用索引方法（[] 和 .at()）访问元素，但抛出 error: no matching function for call to 'boost::multi_array::data(int )'，但是我可以打印数据（参见代码），但不知道如何存储它并再次访问它以进行进一步计算。数据是二维的（11214, 3），但与此同时，我只想将其展平并具有一系列值。所以我的问题是如何访问元素以及如何将它们存储在容器中？

#include <boost/multi_array.hpp>
#include <boost/timer/timer.hpp>
#include <boost/range/irange.hpp>
#include <h5xx/h5xx.hpp>
#include <iostream>
#include <vector>
#include <algorithm>
#include <iterator>
#include <string>

using array_2d_t = boost::multi_array<float, 2>;
//using array_2d_t = boost::multi_array<float, 3>;
template <typename T> 
void print_array(T const& array)
{
    for (auto const& row : array) 
        { for (auto v : row)
            printf("%10f ", v);
        printf("\n"); //prints a new line similar t0 \n
    }
}

h5xx::dataset open_dataset(std::string const& filename) {
    h5xx::file xaa(filename, h5xx::file::mode::in);
    h5xx::group g(xaa, "particles/lipids/box/positions");
    return h5xx::dataset(g, "value");
}

std::vector<float> cell_from_all_frames(h5xx::dataset& ds, size_t row, size_t col) {
    // determine dataset shape: frames, particle count, space dimension
    auto ds_shape = h5xx::dataspace(ds).extents<3>();
    std::vector<float> cells(ds_shape[0]); // number of frames

    std::vector<hsize_t> offsets{0, row, col};
    std::vector<hsize_t> counts{ds_shape[0], 1, 1};
    h5xx::slice slice(offsets, counts);

    h5xx::read_dataset(ds, cells, slice);
    return cells;
}

array_2d_t read_frame(std::string const& filename, unsigned frame_no) {
    //h5xx::file xaa("../../data/xaa.h5", h5xx::file::mode::in);
    h5xx::file xaa(filename, h5xx::file::mode::in);

    h5xx::group   g(xaa, "particles/lipids/box/positions");
    h5xx::dataset ds(g, "value");

    // determine dataset shape: frames, particle count, space dimension
    auto ds_shape = h5xx::dataspace(ds).extents<3>();
    array_2d_t arr(boost::extents[ds_shape[1]][ds_shape[2]]);

    std::vector<hsize_t> offsets{frame_no, 0, 0};
    std::vector<hsize_t> counts{1, arr.shape()[0], arr.shape()[1]};
    h5xx::slice slice(offsets, counts);

    h5xx::read_dataset(ds, arr, slice);
    return arr;
}

int main(int argc, char const* argv[])
{
    if (argc < 2) {
        std::cout << "Usage: " << argv[0] << " input.h5" << std::endl;
        return -1;
    }

    auto ds = open_dataset(argv[1]);
    std::vector<float> first_cells = cell_from_all_frames(ds, 0, 0);

    // set up multi-tau correlator for the computation of time correlation functions
    size_t nsamples = 10;            // FIXME obtain these parameters from HDF5 file. These 10 elements would be first element of first row, first element of 11214 row, first element of 11214*2 row, first element of 11214*3 row ,..., first element of 11214*10 row.
  return 0;
}

在 main() 中，我从函数 read_frame 读取数据并尝试将其传递给 nsamples。我已经尝试了一些东西，但它不起作用！

Answer 1

你可以选择。

存储它？

array_2d_t  frame = read_frame(filename, 1);

访问一个元素？

// access individual elements:
float ele = frame[0][3];

// or with index list:
std::array<int, 2> indices{0,3};
ele = frame(indices);

或者，如您所愿，提供阵列的平面视图：

boost::multi_array_ref<float, 1> sequence(frame.origin(), boost::extents[frame.num_elements()]);
fmt::print("Sum of all {} elements: {}\n",
        sequence.size(),
        std::accumulate(sequence.begin(), sequence.end(), 0.f));

事实上，你可能会就地重塑，但是你不能改变维度，所以你会得到所有单元格的 1“行”：

frame.reshape(std::array<size_t, 2> {1, frame.num_elements()});
// now the first "row" is the full sequence:
auto&& sequence = frame[0];
fmt::print("Sum of all {} elements: {}\n",
        sequence.size(),
        std::accumulate(sequence.begin(), sequence.end(), 0.f));

有大量的切片/重新索引选项，有或没有跨步，但我refer to the Boost documentation 以防止在这里不必要地复杂化。

现场演示

#include <boost/multi_array.hpp>
#include <fmt/ranges.h>
#include <h5xx/h5xx.hpp>
#include <iostream>
#include <iterator>

using array_2d_t = boost::multi_array<float, 2>;

h5xx::dataset open_dataset(std::string const& filename) {
    h5xx::file xaa(filename, h5xx::file::mode::in);
    h5xx::group g(xaa, "particles/lipids/box/positions");
    return h5xx::dataset(g, "value");
}

array_2d_t read_frame(h5xx::dataset& ds, unsigned frame_no) {
    auto ds_shape = h5xx::dataspace(ds).extents<3>();
    array_2d_t arr(boost::extents[ds_shape[1]][ds_shape[2]]);

    std::vector<hsize_t> offsets{frame_no, 0, 0};
    std::vector<hsize_t> counts{1, arr.shape()[0], arr.shape()[1]};
    h5xx::slice slice(offsets, counts);

    h5xx::read_dataset(ds, arr, slice);
    return arr;
}

std::vector<float> cell_from_all_frames(h5xx::dataset& ds, size_t row, size_t col) {
    // determine dataset shape: frames, particle count, space dimension
    auto ds_shape = h5xx::dataspace(ds).extents<3>();
    std::vector<float> cells(ds_shape[0]); // number of frames

    std::vector<hsize_t> offsets{0, row, col};
    std::vector<hsize_t> counts{ds_shape[0], 1, 1};
    h5xx::slice slice(offsets, counts);

    h5xx::read_dataset(ds, cells, slice);
    return cells;
}

int main(int argc, char const* argv[])
{
    if (argc < 2) {
        std::cout << "Usage: " << argv[0] << " input.h5" << std::endl;
        return -1;
    }

    auto ds = open_dataset(argv[1]);
    array_2d_t  frame = read_frame(ds, 1);

    // access individual elements:
    [[maybe_unused]] float ele = frame[0][2];

    // or with index list:
    std::array<int, 2> indices{0,2};
    ele = frame(indices);

    {
        boost::multi_array_ref<float, 1> sequence(frame.origin(), boost::extents[frame.num_elements()]);
        fmt::print("Sum of all {} elements: {}\n",
                sequence.size(),
                std::accumulate(sequence.begin(), sequence.end(), 0.f));
    }

    {
        // in fact yuou might reshape in-place, but then you cannot change dimensiaonality
        frame.reshape(std::array<size_t, 2> {1, frame.num_elements()});
        // now the first "row" is the full sequence:
        auto&& sequence = frame[0];
        fmt::print("Sum of all {} elements: {}\n",
                sequence.size(),
                std::accumulate(sequence.begin(), sequence.end(), 0.f));
    }

    {
        std::vector<float> first_cells = cell_from_all_frames(ds, 0, 0);

        fmt::print("Sum of all {} first cells: {}\n",
                first_cells.size(),
                std::accumulate(first_cells.begin(), first_cells.end(), 0.f));
    }
}

不久前使用您的 xaa.h5：

Sum of all 33642 elements: 737589.1
Sum of all 33642 elements: 737589.1
Sum of all 75 first cells: 6053.3496