Installation

SciNumTools v3 can be installed in several ways, depending on how it is intended to be used. Command-line users can download a precompiled snt executable from GitHub Releases. Python users can install the Python bindings from PyPI or Conda, while C++ users can use vcpkg, Conan, Homebrew, or build the library directly from source.

For most users, installing a pre-built package is recommended. Building from source is useful when developing SciNumTools itself, requiring specific build options, or integrating a development version of the library.

Precompiled snt executable

Download the archive for your system from the latest GitHub release. The v0.9.3 archives are:

• Linux x86-64

• macOS ARM64

• Windows x86-64

On macOS, download and unpack the archive, then run the executable:

curl -fLO https://github.com/scinumtools/snt3/releases/download/v0.9.3/snt-v0.9.3-macos-arm64.tar.gz
tar -xzf snt-v0.9.3-macos-arm64.tar.gz
./snt-v0.9.3-macos-arm64/snt --version

On Linux, use the corresponding archive:

curl -fLO https://github.com/scinumtools/snt3/releases/download/v0.9.3/snt-v0.9.3-linux-x86_64.tar.gz
tar -xzf snt-v0.9.3-linux-x86_64.tar.gz
./snt-v0.9.3-linux-x86_64/snt --version

The Linux executable requires system runtime libraries, including libcurl and libaec. Keep the extracted directory together when moving it, because the macOS archive includes libraries beside snt. To call snt by name in the current shell, add the extracted directory to PATH (use linux-x86_64 in this command on Linux):

export PATH="$PWD/snt-v0.9.3-macos-arm64:$PATH"

For a permanent installation, move the entire extracted directory to a stable location and add its absolute path to your shell’s PATH configuration.

On Windows, extract the ZIP archive with File Explorer or PowerShell:

Invoke-WebRequest https://github.com/scinumtools/snt3/releases/download/v0.9.3/snt-v0.9.3-windows-x86_64.zip -OutFile snt-v0.9.3-windows-x86_64.zip
Expand-Archive snt-v0.9.3-windows-x86_64.zip -DestinationPath .
.\snt-v0.9.3-windows-x86_64\snt.exe --version

The archives contain the command-line application, not the C++ development libraries or Python bindings. The snt hub command additionally requires Git and Python 3 to fetch and run project adapters.

Python

The Python bindings are available as a regular Python package on PyPI:

pip install scinumtools3

SciNumTools is also available through conda-forge:

conda install conda-forge::scinumtools3

After installation, the Python interface can be imported directly:

from scinumtools3.puq import Quantity
from scinumtools3.dip import DIP

C++

For C++ projects, SciNumTools can be installed using several package managers, including vcpkg and Conan.

vcpkg

SciNumTools is available as a vcpkg package:

git clone https://github.com/microsoft/vcpkg.git
cd vcpkg

./bootstrap-vcpkg.sh

./vcpkg install scinumtools3

The local vcpkg port builds report support with the reports feature: ./vcpkg install 'scinumtools3[reports]'. The feature fetches Brief++.

On Windows, use bootstrap-vcpkg.bat instead of bootstrap-vcpkg.sh.

Conan

SciNumTools can also be created as a local Conan package directly from the source repository:

git clone --recurse-submodules https://github.com/scinumtools/snt3.git
cd snt3

conan create .

The Conan package installs the C++ libraries and the snt executable with server, viewer, and report support enabled by default. The viewer requires the platform’s OpenGL and window-system development libraries. Conan options with_server=False, with_viewer=False, and with_reports=False disable the corresponding feature and its bundled dependency. For example, create a package without the three features using:

conan create . -o 'scinumtools3/*:with_server=False' -o 'scinumtools3/*:with_viewer=False' -o 'scinumtools3/*:with_reports=False'

To put snt on PATH in a consuming project, add scinumtools3/0.9.0 under [tool_requires] in its conanfile.txt and generate VirtualBuildEnv. A regular library requirement still provides the executable in the package’s bin directory, but does not add that directory to the runtime environment.

macOS / Homebrew

On macOS, SciNumTools can be installed using the project Homebrew tap:

brew tap scinumtools/tap
brew install scinumtools3

Building from Source

SciNumTools is built using CMake and requires a C++17-compatible compiler. Building from source provides full access to the C++ library, command-line applications, tests, and other components of the project. For a summary of required and included build dependencies, see Dependencies.

With DIP enabled (the default), install HDF5 development headers and the HDF5 C library before configuring. The HDF5 C++ library is not required. If HDF5 is installed outside the standard search paths, pass its installation prefix as -DHDF5_ROOT=/path/to/hdf5 when running CMake. The commands below also require the Ninja build tool.

Clone the repository with its submodules and configure the build. The briefpp submodule provides the header-only renderer for DIP reports. Alternatively, set SNT_BRIEFPP_INCLUDE_DIR to a directory containing briefpp/report.hpp. The cpp-httplib submodule provides the optional REST server. Alternatively, install cpp-httplib 0.46.0 or newer and set SNT_HTTPLIB_INCLUDE_DIR to its header directory. Builds with -DENABLE_SNT_SERVER=OFF do not need cpp-httplib. Builds with -DENABLE_SNT_REPORT=OFF omit report generation and do not need Brief++. The report option defaults to on when DIP is enabled.

git clone --recurse-submodules https://github.com/scinumtools/snt3.git
cd snt3

cmake -G Ninja -B build
cmake --build build

The test suite can then be executed using CTest:

ctest --test-dir build

Finally, install the compiled library and associated components:

cmake --install build

Using the Setup Script

The repository also provides a convenience setup script for building, testing, and installing SciNumTools:

sudo ./setup.sh -b -t -i

Here -b builds the project, -t runs the tests, and -i installs the resulting components.

Using SciNumTools from CMake

Once installed, SciNumTools can be discovered from another CMake project using find_package:

find_package(snt REQUIRED)

Individual SciNumTools modules can then be linked to an executable:

add_executable(${EXEC_NAME} ${SOURCE_FILES})

target_link_libraries(
    ${EXEC_NAME}
    PRIVATE
    snt-exs
    snt-puq
    snt-dip
)

This modular approach allows an application to link only the SciNumTools components it requires.

Choosing an Installation Method

The recommended installation method depends on the intended use:

• Python application: install scinumtools3 from PyPI or Conda.

• Command-line application: download the precompiled snt executable from GitHub Releases.

• C++ application: use vcpkg, Conan, Homebrew, or a system installation.

• SciNumTools development: build directly from source.

• Reproducible development environment: use the Docker integration.

After installation, continue with Quick Start to start using SciNumTools.