C++ wrapper for the HDF5 C-library

h5cpp is a new C++ wrapper for HDF5s C-API.

Motivation

HDF5 is a powerful binary file format. There is virtually nothing that cannot be stored within an HDF5 file. To read and write data from and to an HDF5 file you can either use the C-API or one of the wrappers for a scripting language (for Python there is for instance h5py or pytables). However, to keep their interfaces simple, many of these wrappers do not provide the full functionality HDF5 has to offer. If you want to use all features available the C-API is most probably the safest way to go.

h5cpp wants to provide both, an easy to use modern C++ interface but at the same time give you access to the full functionality HDF5 has to offer.

Complete and domain agnostic

Many HDF5 wrappers which are currently around have to major issues:

1. they do not provide the full functionality of the C-API
2. they are specifically made for a particular field of application

Our goal is to keep h5cpp as much as application agnostic as possible and provide the full functionality of the C-API.

Easy to use

HDF5 is a rather powerful file format and thus the C-API is rather complex. h5cpps approach to this degree of complexity can be best described by a quote of Alan Kay

Simple things should be simple, complex things should be possible.

That's what we try to achieve and hopefully we succeeded in it.

Support for major platforms

We currently support the following operating systems

• Linux
• Windows
• OSX

Since we are using cmake to build h5cpp and also provide a cmake package for the library, using it should be the same on each of these platforms.

How to build h5cpp

The minimum requirements for building the library are:

• a C++ compiler, gcc>=4.8 should do well
• the boost libraries
• the HDF5 C library (>=1.8.13 would do but >=1.10.0 is prefered)
• cmake >= 3.0

Building the library is standard cmake & make fare, out of source. For example, in linux shell, you would do the following:

git clone https://github.com/ess-dmsc/h5cpp.git
cd h5cpp
mkdir build
cd build
cmake ..
make

To install the library to system, you would follow this up with:

sudo make install

How to use h5cpp

With the library installed as above, you can bring it into your program by adding something like this to your CMakeLists.txt file:

find_package(h5cpp REQUIRED)
.
.
.
add_executable(some_target some_code.cpp)
target_link_libraries(some_target h5cpp)

and adding the following:

#include <h5cpp/hdf5.hpp>

to your your source file. Here is a small example of how to make use of the library in code:

using namespace hdf5;

// create a file
file::File f = file::create("writing_vector.h5",file::AccessFlags::TRUNCATE);

// create a group
node::Group root_group = f.root();
node::Group my_group = root_group.create_group("my_group");

// create a dataset
using data_type = std::vector<int>;
data_type data{1,2,3,4};
node::Dataset dataset = my_group.create_dataset("data",
                                                datatype::create<data_type>(),
                                                dataspace::create(data));

// write to dataset
dataset.write(data);

Alternate install directory

If you do not wish to install h5cpp to your system folders you can slightly modify the above steps. When building the library, invoke CMake with the following option:

cmake -DCMAKE_INSTALL_PREFIX=/home/user1/some/path ..

and accordingly, when building the client program:

cmake -Dh5cpp_DIR=/home/user1/some/path/lib/cmake/h5cpp-0.0.1 path/to/your/source

where version number may vary.

For OSX and Windows instructions, as well as instructions for building tests and documentation, see the online documentation.