openPMD-api 0.11.0a0

C++ & Python API for Scientific I/O with openPMD

C++ & Python API for Scientific I/O with openPMD

This library provides a high-level API for writing and reading scientific data according to openPMD. It allows one to build logical file structures which drive scientific I/O libraries such as HDF5 and ADIOS through a common, intuitive interface. Where supported, openPMD-api implements both serial and MPI parallel I/O capabilities.

Usage

C++

#include <openPMD/openPMD.hpp>
#include <iostream>

// ...

auto s = openPMD::Series("samples/git-sample/data%T.h5", openPMD::AccessType::READ_ONLY);

for( auto const& i : s.iterations ) {
    std::cout << "Iteration: " << i.first << "\n";

    for( auto const& m : i.second.meshes ) {
        std::cout << "  Mesh '" << m.first << "' attributes:\n";
        for( auto const& val : m.second.attributes() )
            std::cout << "    " << val << '\n';
    }

    for( auto const& p : i.second.particles ) {
        std::cout << "  Particle species '" << p.first << "' attributes:\n";
        for( auto const& val : p.second.attributes() )
            std::cout << "    " << val << '\n';
    }
}

Python

import openpmd_api

# ...

series = openpmd_api.Series("samples/git-sample/data%T.h5", openPMD.Access_Type.read_only)

for k_i, i in series.iterations.items():
    print("Iteration: {0}".format(k_i))

    for k_m, m in i.meshes.items():
        print("  Mesh '{0}' attributes:".format(k_m))
        for a in m.attributes:
            print("    {0}".format(a))

    for k_p, p in i.particles.items():
        print("  Particle species '{0}' attributes:".format(k_p))
        for a in p.attributes:
            print("    {0}".format(a))

More!

Curious? Our manual shows full read & write examples, both serial and MPI-parallel!

Dependencies

Required:

• CMake 3.11.0+
• C++11 capable compiler, e.g. g++ 4.8+, clang 3.9+, VS 2015+

Shipped internally in share/openPMD/thirdParty/:

• MPark.Variant 1.4.0+ (BSL-1.0)
• Catch2 2.6.1+ (BSL-1.0)
• pybind11 2.3.0+ (new BSD)
• NLohmann-JSON 3.7.0+ (MIT)

I/O backends:

• JSON
• HDF5 1.8.13+ (optional)
• ADIOS1 1.13.1+ (optional)
• ADIOS2 2.5.0+ (optional)

while those can be built either with or without:

• MPI 2.1+, e.g. OpenMPI 1.6.5+ or MPICH2

Optional language bindings:

• Python:
  • Python 3.5 - 3.8
  • pybind 2.3.0+
  • numpy 1.15+
  • mpi4py 2.1+

Installation

Choose one of the install methods below to get started:

Spack

# optional:               +python +adios1 -adios2 -hdf5 -mpi
spack install openpmd-api
spack load -r openpmd-api

Conda

# optional:            OpenMPI support  =*=mpi_openmpi*
# optional:              MPICH support  =*=mpi_mpich*
conda install -c conda-forge openpmd-api

PyPI

Behind the scenes, this install method compiles from source against the found installations of HDF5, ADIOS1, ADIOS2, and/or MPI (in system paths, from other package managers, or loaded via a module system, ...). The current status for this install method is experimental. Please feel free to report how this works for you.

# we need pip 19 or newer
# optional:                   --user
python3 -m pip install -U pip

# optional:                        --user
python3 -m pip install openpmd-api

or with MPI support:

# optional:                                    --user
python3 -m pip install -U pip setuptools wheel
python3 -m pip install -U cmake

# optional:                                                                   --user
openPMD_USE_MPI=ON python3 -m pip install openpmd-api --no-binary openpmd-api

From Source

openPMD-api can also be built and installed from source using CMake:

git clone https://github.com/openPMD/openPMD-api.git

mkdir openPMD-api-build
cd openPMD-api-build

# optional: for full tests, with unzip
../openPMD-api/.travis/download_samples.sh

# for own install prefix append:
#   -DCMAKE_INSTALL_PREFIX=$HOME/somepath
# for options append:
#   -DopenPMD_USE_...=...
# e.g. for python support add:
#   -DopenPMD_USE_PYTHON=ON -DPYTHON_EXECUTABLE=$(which python3)
cmake ../openPMD-api

cmake --build .

# optional
ctest

# sudo might be required for system paths
cmake --build . --target install

The following options can be added to the cmake call to control features. CMake controls options with prefixed -D, e.g. -DopenPMD_USE_MPI=OFF:

CMake Option	Values	Description
openPMD_USE_MPI	AUTO/ON/OFF	Parallel, Multi-Node I/O for clusters
openPMD_USE_HDF5	AUTO/ON/OFF	HDF5 backend (.h5 files)
openPMD_USE_ADIOS1	AUTO/ON/OFF	ADIOS1 backend (.bp files up to version BP3)
openPMD_USE_ADIOS2	AUTO/ON/OFF	ADIOS2 backend (.bp files in BP3, BP4 or higher)
openPMD_USE_PYTHON	AUTO/ON/OFF	Enable Python bindings
openPMD_USE_INVASIVE_TESTS	ON/OFF	Enable unit tests that modify source code 1
openPMD_USE_VERIFY	ON/OFF	Enable internal VERIFY (assert) macro independent of build type 2
PYTHON_EXECUTABLE	(first found)	Path to Python executable

¹ e.g. changes C++ visibility keywords, breaks MSVC ² this includes most pre-/post-condition checks, disabling without specific cause is highly discouraged

Additionally, the following libraries are shipped internally. The following options allow to switch to external installs:

CMake Option	Values	Library	Version
openPMD_USE_INTERNAL_VARIANT	ON/OFF	MPark.Variant	1.4.0+
openPMD_USE_INTERNAL_CATCH	ON/OFF	Catch2	2.6.1+
openPMD_USE_INTERNAL_PYBIND11	ON/OFF	pybind11	2.3.0+
openPMD_USE_INTERNAL_JSON	ON/OFF	NLohmann-JSON	3.7.0+

By default, this will build as a shared library (libopenPMD.[so|dylib|dll]) and installs also its headers. In order to build a static library, append -DBUILD_SHARED_LIBS=OFF to the cmake command. You can only build a static or a shared library at a time.

By default, the Release version is built. In order to build with debug symbols, pass -DCMAKE_BUILD_TYPE=Debug to your cmake command.

By default, tests, examples and command line tools are built. In order to skip building those, pass -DBUILD_TESTING=OFF, -DBUILD_EXAMPLES=OFF, or -DBUILD_CLI_TOOLS=OFF to your cmake command.

Linking to your project

The install will contain header files and libraries in the path set with -DCMAKE_INSTALL_PREFIX.

CMake

If your project is using CMake for its build, one can conveniently use our provided openPMDConfig.cmake package which is installed alongside the library.

First set the following environment hint if openPMD-api was not installed in a system path:

# optional: only needed if installed outside of system paths
export CMAKE_PREFIX_PATH=$HOME/somepath:$CMAKE_PREFIX_PATH

Use the following lines in your project's CMakeLists.txt:

# supports:                       COMPONENTS MPI NOMPI HDF5 ADIOS1 ADIOS2
find_package(openPMD 0.9.0 CONFIG)

if(openPMD_FOUND)
    target_link_libraries(YourTarget PRIVATE openPMD::openPMD)
endif()

Alternatively, add the openPMD-api repository source directly to your project and use it via:

add_subdirectory("path/to/source/of/openPMD-api")

target_link_libraries(YourTarget PRIVATE openPMD::openPMD)

For development workflows, you can even automatically download and build openPMD-api from within a depending CMake project. Just replace the add_subdirectory call with:

include(FetchContent)
set(CMAKE_POLICY_DEFAULT_CMP0077 NEW)
set(BUILD_TESTING OFF)
set(BUILD_EXAMPLES OFF)
set(openPMD_USE_PYTHON OFF)
FetchContent_Declare(openPMD
  GIT_REPOSITORY "https://github.com/openPMD/openPMD-api.git"
  GIT_TAG        "dev")
FetchContent_MakeAvailable(openPMD)

Manually

If your (Linux/OSX) project is build by calling the compiler directly or uses a manually written Makefile, consider using our openPMD.pc helper file for pkg-config which are installed alongside the library.

First set the following environment hint if openPMD-api was not installed in a system path:

# optional: only needed if installed outside of system paths
export PKG_CONFIG_PATH=$HOME/somepath/lib/pkgconfig:$PKG_CONFIG_PATH

Additional linker and compiler flags for your project are available via:

# switch to check if openPMD-api was build as static library
# (via BUILD_SHARED_LIBS=OFF) or as shared library (default)
if [ "$(pkg-config --variable=static openPMD)" == "true" ]
then
    pkg-config --libs --static openPMD
    # -L/usr/local/lib -L/usr/lib/x86_64-linux-gnu/openmpi/lib -lopenPMD -pthread /usr/lib/libmpi.so -pthread /usr/lib/x86_64-linux-gnu/openmpi/lib/libmpi_cxx.so /usr/lib/libmpi.so /usr/lib/x86_64-linux-gnu/hdf5/openmpi/libhdf5.so /usr/lib/x86_64-linux-gnu/libsz.so /usr/lib/x86_64-linux-gnu/libz.so /usr/lib/x86_64-linux-gnu/libdl.so /usr/lib/x86_64-linux-gnu/libm.so -pthread /usr/lib/libmpi.so -pthread /usr/lib/x86_64-linux-gnu/openmpi/lib/libmpi_cxx.so /usr/lib/libmpi.so
else
    pkg-config --libs openPMD
    # -L${HOME}/somepath/lib -lopenPMD
fi

pkg-config --cflags openPMD
# -I${HOME}/somepath/include

Author Contributions

openPMD-api is developed by many people. It was initially started by the Computational Radiation Physics Group at HZDR as successor to libSplash, generalizing the successful HDF5 & ADIOS1 implementations in PIConGPU. The following people and institutions contributed to openPMD-api:

• Axel Huebl (HZDR, now LBNL): project lead, releases, documentation, automated CI/CD, Python bindings, installation & packaging, prior reference implementations
• Fabian Koller (HZDR): initial library design and implementation with HDF5 & ADIOS1 backend
• Franz Poeschel (HZDR): added JSON & ADIOS2 backend, data staging/streaming
• Junmin Gu (LBNL): non-collective parallel I/O fixes, ADIOS improvements

Further thanks go to improvements and contributions from:

• Carsten Fortmann-Grote (EU XFEL GmbH, now MPI-EvolBio): draft of our Python unit tests
• Dominik Stańczak (Warsaw University of Technology): documentation improvements
• Ray Donnelly (Anaconda, Inc.): support on conda packaging and libc++ quirks
• James Amundson (FNAL): compile fix for newer compilers
• René Widera (HZDR): design improvements for initial API design
• Erik Zenker (HZDR): design improvements for initial API design
• Sergei Bastrakov (HZDR): documentation improvements (windows)
• Rémi Lehe (LBNL): package integration testing on macOS and Linux
• Lígia Diana Amorim (LBNL): package integration testing on macOS
• Kseniia Bastrakova (HZDR): compatibility testing
• Richard Pausch (HZDR): compatibility testing
• Paweł Ordyna (HZDR): report on NVCC warnings

Grants

The openPMD-api authors acknowledge support via the following programs. This project has received funding from the European Unions Horizon 2020 research and innovation programme under grant agreement No 654220. Supported by the Consortium for Advanced Modeling of Particles Accelerators (CAMPA), funded by the U.S. DOE Office of Science under Contract No. DE-AC02-05CH11231. Supported by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of two U.S. Department of Energy organizations (Office of Science and the National Nuclear Security Administration).

Transitive Contributions

openPMD-api stands on the shoulders of giants and we are grateful for the following projects included as direct dependencies:

• ADIOS1 and ADIOS2 by S. Klasky (ORNL), team, collaborators and contributors
• Catch2 by Phil Nash, Martin Hořeňovský and contributors
• HDF5 by the HDF group and community
• json by Niels Lohmann and contributors
• pybind11 by Wenzel Jakob (EPFL) and contributors
• all contributors to the evolution of modern C++ and early library preview developers, e.g. Michael Park (Facebook)
• the CMake build system and contributors
• packaging support by the conda-forge, PyPI and Spack communities, among others
• the openPMD-standard by Axel Huebl (HZDR, now LBNL) and contributors