HDF5 示例代码
Posted
技术标签:
【中文标题】HDF5 示例代码【英文标题】：HDF5 Example code 【发布时间】：2011-12-17 21:23:15 【问题描述】：使用HDF5DotNet，谁能指出示例代码，它将打开一个hdf5文件，提取数据集的内容，并将内容打印到标准输出？
到目前为止，我有以下内容：
H5.Open();
var h5 = H5F.open("example.h5", H5F.OpenMode.ACC_RDONLY);
var dataset = H5D.open(h5, "/Timings/aaPCBTimes");
var space = H5D.getSpace(dataset);
var size = H5S.getSimpleExtentDims(space);
然后就有点混乱了。
我实际上想对数据集的内容进行一些处理，但我认为一旦我将转储到标准输出，我就可以从那里解决它。
更新：我已经解决了这个问题，足以解决我自己的问题。我没有意识到数据集是一个多数组 - 我认为它更像是一个 db 表。万一有人感兴趣，
double[,] dataArray = new double[size[0], 6];
var wrapArray = new H5Array<double>(dataArray);
var dataType = H5D.getType(d);
H5D.read(dataset, dataType, wrapArray);
Console.WriteLine(dataArray[0, 0]);
【问题讨论】：
我想指出，将H5、h5、H5F、H5S 和H5D 作为变量名是不好的形式。前两个仅因情况而异。其他的只有一个字母。
是的。 H5、H5F、H5S 和 H5D 均由 API - HDF5DotNet - C++/CLI Wrapper of the HDF5 Library 提供。如果您点击链接，您还可以享受 H5A、H5E、H5G、H5I、H5L、H5O、H5P 和 H5R。我正在寻找代码示例，因为我不想花太多时间研究这个界面。
你为什么不添加你的破解/解决方案作为答案并接受它？
@Sebastian，我刚刚找到了用于 HDF5 的 this（商业许可）高级 API。它看起来不错，但它是一个大而昂贵的项目的一部分。
请注意，HDF5 项目 now recommends 使用他们的 P/Invoke 包装器 here，并且 HDF5DotNet 在 2012 年没有维护和冻结。
【参考方案1】：
试试这个：
using System;
using HDF5DotNet;
namespace CSharpExample1
class Program
// Function used with
static int myFunction(H5GroupId id, string objectName, Object param)
Console.WriteLine("The object name is 0", objectName);
Console.WriteLine("The object parameter is 0", param);
return 0;
static void Main(string[] args)
try
// We will write and read an int array of this length.
const int DATA_ARRAY_LENGTH = 12;
// Rank is the number of dimensions of the data array.
const int RANK = 1;
// Create an HDF5 file.
// The enumeration type H5F.CreateMode provides only the legal
// creation modes. Missing H5Fcreate parameters are provided
// with default values.
H5FileId fileId = H5F.create("myCSharp.h5",
H5F.CreateMode.ACC_TRUNC);
// Create a HDF5 group.
H5GroupId groupId = H5G.create(fileId, "/cSharpGroup", 0);
H5GroupId subGroup = H5G.create(groupId, "mySubGroup", 0);
// Demonstrate getObjectInfo
ObjectInfo info = H5G.getObjectInfo(fileId, "/cSharpGroup", true);
Console.WriteLine("cSharpGroup header size is 0", info.headerSize);
Console.WriteLine("cSharpGroup nlinks is 0", info.nHardLinks);
Console.WriteLine("cSharpGroup fileno is 0 1",
info.fileNumber[0], info.fileNumber[1]);
Console.WriteLine("cSharpGroup objno is 0 1",
info.objectNumber[0], info.objectNumber[1]);
Console.WriteLine("cSharpGroup type is 0", info.objectType);
H5G.close(subGroup);
// Prepare to create a data space for writing a 1-dimensional
// signed integer array.
ulong[] dims = new ulong[RANK];
dims[0] = DATA_ARRAY_LENGTH;
// Put descending ramp data in an array so that we can
// write it to the file.
int[] dset_data = new int[DATA_ARRAY_LENGTH];
for (int i = 0; i < DATA_ARRAY_LENGTH; i++)
dset_data[i] = DATA_ARRAY_LENGTH - i;
// Create a data space to accommodate our 1-dimensional array.
// The resulting H5DataSpaceId will be used to create the
// data set.
H5DataSpaceId spaceId = H5S.create_simple(RANK, dims);
// Create a copy of a standard data type. We will use the
// resulting H5DataTypeId to create the data set. We could
// have used the HST.H5Type data directly in the call to
// H5D.create, but this demonstrates the use of H5T.copy
// and the use of a H5DataTypeId in H5D.create.
H5DataTypeId typeId = H5T.copy(H5T.H5Type.NATIVE_INT);
// Find the size of the type
uint typeSize = H5T.getSize(typeId);
Console.WriteLine("typeSize is 0", typeSize);
// Set the order to big endian
H5T.setOrder(typeId, H5T.Order.BE);
// Set the order to little endian
H5T.setOrder(typeId, H5T.Order.LE);
// Create the data set.
H5DataSetId dataSetId = H5D.create(fileId, "/csharpExample",
typeId, spaceId);
// Write the integer data to the data set.
H5D.write(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
new H5Array<int>(dset_data));
// If we were writing a single value it might look like this.
// int singleValue = 100;
// H5D.writeScalar(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
// ref singleValue);
// Create an integer array to receive the read data.
int[] readDataBack = new int[DATA_ARRAY_LENGTH];
// Read the integer data back from the data set
H5D.read(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
new H5Array<int>(readDataBack));
// Echo the data
for(int i=0;i<DATA_ARRAY_LENGTH;i++)
Console.WriteLine(readDataBack[i]);
// Close all the open resources.
H5D.close(dataSetId);
// Reopen and close the data sets to show that we can.
dataSetId = H5D.open(fileId, "/csharpExample");
H5D.close(dataSetId);
dataSetId = H5D.open(groupId, "/csharpExample");
H5D.close(dataSetId);
H5S.close(spaceId);
H5T.close(typeId);
H5G.close(groupId);
//int x = 10;
//H5T.enumInsert<int>(typeId, "myString", ref x);
//H5G.close(groupId);
H5GIterateDelegate myDelegate;
myDelegate = myFunction;
int x = 9;
int index = H5G.iterate(fileId, "/cSharpGroup",
myDelegate, x, 0);
// Reopen the group id to show that we can.
groupId = H5G.open(fileId, "/cSharpGroup");
H5G.close(groupId);
H5F.close(fileId);
// Reopen and reclose the file.
H5FileId openId = H5F.open("myCSharp.h5",
H5F.OpenMode.ACC_RDONLY);
H5F.close(openId);
// This catches all the HDF exception classes. Because each call
// generates unique exception, different exception can be handled
// separately. For example, to catch open errors we could have used
// catch (H5FopenException openException).
catch (HDFException e)
Console.WriteLine(e.Message);
Console.WriteLine("Processing complete!");
Console.ReadLine();
【讨论】：
我编辑了这个例子，因为它没有编译：H5G.create -> 跳过第三个参数，H5S.create_simple 必须得到 long[] 而不是 ulong[] 作为第二个参数，H5T.getSize 返回 int而不是 uint【参考方案2】：所以，你的开始很棒。我创建了一些可以帮助你的扩展。在您的代码中使用它，您应该能够在面向对象的语言中更有意义的事情，例如（在您的情况下）：
H5.Open();
var h5FileId= H5F.open("example.h5");
double[,] dataArray = h5FileId.Read2DArray<double>("/Timings/aaPCBTimes");
// or more generically...
T[,] dataArray = h5FileId.Read2DArray<T>("/Timings/aaPCBTimes");
这是不完整的扩展，我会考虑将它们添加到 HDF5Net...
public static class HdfExtensions
// thank you http://***.com/questions/4133377/splitting-a-string-number-every-nth-character-number
public static IEnumerable<String> SplitInParts(this String s, Int32 partLength)
if (s == null)
throw new ArgumentNullException("s");
if (partLength <= 0)
throw new ArgumentException("Part length has to be positive.", "partLength");
for (var i = 0; i < s.Length; i += partLength)
yield return s.Substring(i, Math.Min(partLength, s.Length - i));
public static T[] Read1DArray<T>(this H5FileId fileId, string dataSetName)
var dataset = H5D.open(fileId, dataSetName);
var space = H5D.getSpace(dataset);
var dims = H5S.getSimpleExtentDims(space);
var dataType = H5D.getType(dataset);
if (typeof(T) == typeof(string))
int stringLength = H5T.getSize(dataType);
byte[] buffer = new byte[dims[0] * stringLength];
H5D.read(dataset, dataType, new H5Array<byte>(buffer));
string stuff = System.Text.ASCIIEncoding.ASCII.GetString(buffer);
return stuff.SplitInParts(stringLength).Select(ss => (T)(object)ss).ToArray();
T[] dataArray = new T[dims[0]];
var wrapArray = new H5Array<T>(dataArray);
H5D.read(dataset, dataType, wrapArray);
return dataArray;
public static T[,] Read2DArray<T>(this H5FileId fileId, string dataSetName)
var dataset = H5D.open(fileId, dataSetName);
var space = H5D.getSpace(dataset);
var dims = H5S.getSimpleExtentDims(space);
var dataType = H5D.getType(dataset);
if (typeof(T) == typeof(string))
// this will also need a string hack...
T[,] dataArray = new T[dims[0], dims[1]];
var wrapArray = new H5Array<T>(dataArray);
H5D.read(dataset, dataType, wrapArray);
return dataArray;
【讨论】：
这比公认的答案更有帮助。【参考方案3】：这是一个工作示例：
using System.Collections.Generic;
using System;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using HDF5DotNet;
namespace HDF5Test
public class HDFTester
static int myFunction(H5GroupId id, string objectName, Object param)
Console.WriteLine("The object name is 0", objectName);
Console.WriteLine("The object parameter is 0", param);
return 0;
public static void runTest()
try
// We will write and read an int array of this length.
const int DATA_ARRAY_LENGTH = 12;
// Rank is the number of dimensions of the data array.
const int RANK = 1;
// Create an HDF5 file.
// The enumeration type H5F.CreateMode provides only the legal
// creation modes. Missing H5Fcreate parameters are provided
// with default values.
H5FileId fileId = H5F.create("myCSharp.h5",
H5F.CreateMode.ACC_TRUNC);
// Create a HDF5 group.
H5GroupId groupId = H5G.create(fileId, "/cSharpGroup");
H5GroupId subGroup = H5G.create(groupId, "mySubGroup");
// Demonstrate getObjectInfo
ObjectInfo info = H5G.getObjectInfo(fileId, "/cSharpGroup", true);
Console.WriteLine("cSharpGroup header size is 0", info.headerSize);
Console.WriteLine("cSharpGroup nlinks is 0", info.nHardLinks);
Console.WriteLine("cSharpGroup fileno is 0 1",
info.fileNumber[0], info.fileNumber[1]);
Console.WriteLine("cSharpGroup objno is 0 1",
info.objectNumber[0], info.objectNumber[1]);
Console.WriteLine("cSharpGroup type is 0", info.objectType);
H5G.close(subGroup);
// Prepare to create a data space for writing a 1-dimensional
// signed integer array.
long[] dims = new long[RANK];
dims[0] = DATA_ARRAY_LENGTH;
// Put descending ramp data in an array so that we can
// write it to the file.
int[] dset_data = new int[DATA_ARRAY_LENGTH];
for (int i = 0; i < DATA_ARRAY_LENGTH; i++)
dset_data[i] = DATA_ARRAY_LENGTH - i;
// Create a data space to accommodate our 1-dimensional array.
// The resulting H5DataSpaceId will be used to create the
// data set.
H5DataSpaceId spaceId = H5S.create_simple(RANK, dims);
// Create a copy of a standard data type. We will use the
// resulting H5DataTypeId to create the data set. We could
// have used the HST.H5Type data directly in the call to
// H5D.create, but this demonstrates the use of H5T.copy
// and the use of a H5DataTypeId in H5D.create.
H5DataTypeId typeId = H5T.copy(H5T.H5Type.NATIVE_INT);
// Find the size of the type
int typeSize = H5T.getSize(typeId);
Console.WriteLine("typeSize is 0", typeSize);
// Set the order to big endian
H5T.setOrder(typeId, H5T.Order.BE);
// Set the order to little endian
H5T.setOrder(typeId, H5T.Order.LE);
// Create the data set.
H5DataSetId dataSetId = H5D.create(fileId, "/csharpExample",
typeId, spaceId);
// Write the integer data to the data set.
H5D.write(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
new H5Array<int>(dset_data));
// If we were writing a single value it might look like this.
// int singleValue = 100;
// H5D.writeScalar(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
// ref singleValue);
// Create an integer array to receive the read data.
int[] readDataBack = new int[DATA_ARRAY_LENGTH];
// Read the integer data back from the data set
H5D.read(dataSetId, new H5DataTypeId(H5T.H5Type.NATIVE_INT),
new H5Array<int>(readDataBack));
// Echo the data
for (int i = 0; i < DATA_ARRAY_LENGTH; i++)
Console.WriteLine(readDataBack[i]);
// Close all the open resources.
H5D.close(dataSetId);
// Reopen and close the data sets to show that we can.
dataSetId = H5D.open(fileId, "/csharpExample");
H5D.close(dataSetId);
dataSetId = H5D.open(groupId, "/csharpExample");
H5D.close(dataSetId);
H5S.close(spaceId);
H5T.close(typeId);
H5G.close(groupId);
//int x = 10;
//H5T.enumInsert<int>(typeId, "myString", ref x);
//H5G.close(groupId);
H5GIterateCallback myDelegate;
myDelegate = myFunction;
int x = 9;
int start = 0;
int index = H5G.iterate(fileId, "/cSharpGroup",myDelegate, x, ref start);
// Reopen the group id to show that we can.
groupId = H5G.open(fileId, "/cSharpGroup");
H5G.close(groupId);
H5F.close(fileId);
// Reopen and reclose the file.
H5FileId openId = H5F.open("myCSharp.h5",
H5F.OpenMode.ACC_RDONLY);
H5F.close(openId);
// This catches all the HDF exception classes. Because each call
// generates unique exception, different exception can be handled
// separately. For example, to catch open errors we could have used
// catch (H5FopenException openException).
catch (HDFException e)
Console.WriteLine(e.Message);
Console.WriteLine("Processing complete!");
Console.ReadLine();
【讨论】：
【参考方案4】：我知道这已经过时了，但对于仍然需要处理 HDF5 文件的任何人，我都有一个 C# 包装器，它封装了 github 的大部分操作（基于其他人的原创作品）。
unit tes 项目中有很多例子。
【讨论】：
以上是关于HDF5 示例代码的主要内容，如果未能解决你的问题，请参考以下文章