如何使用可变长度类型将包含多个 std::vector<float> 的结构写出到 HDF5？

Question

我正在尝试将 c 结构（描述虚拟 "event"）写入 HDF5 格式。该结构包含 4 个浮点向量，每个事件的长度不同（事件将在 HDF5 文件中组合在一起），因此应使用可变长度类型写入文件。然而，在使用 "buffer-struct" 写入数据后，只有每隔一个向量被写入文件（下面代码中的 1 和 3）。其他的是None。我做错了什么？

#include <iostream>
#include <vector>
#include <cstdlib>
#include <string>
#include "H5Cpp.h"

const char * FILE_NAME = "test.h5";

const double HI   =  300.0;
const double LO   = -20.0;

typedef struct
{
    int ID;
    float energy;
    std::vector<float> v1, v2, v3, v4;
} event_struct;

typedef struct
{
    int ID;
    float energy;
    hvl_t v1Handle, v2Handle, v3Handle, v4Handle;
} event_struct_buffer;

int main(void)
{
    event_struct* event = new event_struct();

    event_struct_buffer* event_buffer = new event_struct_buffer();

    event->ID = 5;
    event->energy = 23.45;

    int n1 = 10;
    int n2 = 8;
    int n3 = 4;
    int n4 = 2;

    for( int i = 0;  i < n1; ++i ){
        float x = LO + static_cast <float> (rand()) /( static_cast <float> (RAND_MAX/(HI-LO)));
        event->v1.push_back(x);
        std::cout << x << ' ';
        if( i < n2 ){
            event->v2.push_back(x/2);
            if( i < n3 ){
                event->v3.push_back(x/3);
                if( i < n4 ){
                    event->v4.push_back(x/5);
                }
            }
        }
    }// end for loop
    std::cout << std::endl;

    hid_t      event_tid;                          /* File datatype identifier */
    hid_t      file, dataset, space, vlen_tid;  /* Handles */
    hsize_t    dim[] = {1};                /* Dataspace dimensions */

    int rank = sizeof(dim) / sizeof(hsize_t);
    std::cout << "Rank = " << rank << std::endl;

    space = H5Screate_simple(rank, dim, NULL);

    file = H5Fcreate(FILE_NAME, H5F_ACC_TRUNC, H5P_DEFAULT, H5P_DEFAULT);

    vlen_tid = H5Tvlen_create(H5T_NATIVE_FLOAT);

    event_tid = H5Tcreate( H5T_COMPOUND, sizeof(event_struct_buffer) );
    H5Tinsert(event_tid, "Event_ID", HOFFSET(event_struct_buffer, ID), H5T_NATIVE_INT);
    H5Tinsert(event_tid, "Energy", HOFFSET(event_struct_buffer, energy), H5T_NATIVE_FLOAT);
    H5Tinsert(event_tid, "Image1", HOFFSET(event_struct_buffer, v1Handle), vlen_tid);
    H5Tinsert(event_tid, "Image2", HOFFSET(event_struct_buffer, v2Handle), vlen_tid);
    H5Tinsert(event_tid, "Image3", HOFFSET(event_struct_buffer, v3Handle), vlen_tid);
    H5Tinsert(event_tid, "Image4", HOFFSET(event_struct_buffer, v4Handle), vlen_tid);

    dataset = H5Dcreate(file, "/Events_List", event_tid, space, H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT);

    event_buffer->ID = event->ID;
    event_buffer->energy = event->energy;
    event_buffer->v1Handle.len = event->v1.size();
    event_buffer->v1Handle.p = event->v1.data();
    event_buffer->v2Handle.len = event->v2.size();
    event_buffer->v2Handle.p = event->v2.data();
    event_buffer->v3Handle.len = event->v3.size();
    event_buffer->v3Handle.p = event->v3.data();
    event_buffer->v4Handle.len = event->v4.size();
    event_buffer->v4Handle.p = event->v4.data();

    H5Dwrite(dataset, event_tid, H5S_ALL, H5S_ALL, H5P_DEFAULT, event_buffer);

    H5Tclose(event_tid);
    H5Tclose(vlen_tid);
    H5Sclose(space);
    H5Dclose(dataset);
    H5Fclose(file);
    return 0;
}

伪随机数列表为：

-19.9975 22.0921 221.794 126.768 150.486 50.0669 -4.94572 197.237 197.375 279.102

我用 h5py 读取 HDF5 文件时得到的输出是：

[ (5, 23.450000762939453, [-19.997495651245117, 22.09209442138672, 221.793701171875, 
126.76803588867188, 150.4855194091797, 50.06694030761719, -4.945722579956055, 
197.23670959472656, 197.37486267089844, 279.1017150878906],
[-6.665832042694092, 7.3640313148498535, 73.93123626708984, 42.256011962890625], None, None)]

感谢您的帮助。

Answer 1

问题实际上是使用 h5py（以及表格）读取 h5 文件。当运行 h5dump test.h5 时，所有向量和所有事件中的数据结构和值（我最终写了一个 event_struct 的向量，每个事件中有 4 个可变长度 std::vector ) 填写正确。我也只使用一个结构并将 hvl_t 句柄添加到 event_struct。

如何使用可变长度类型将包含多个 std::vector<float> 的结构写出到 HDF5？

How to write out struct containing multiple std::vector<float> to HDF5 using variable length type?

c++

stl

allocation

hdf5

dataset