Installation#

Install from PyPI#

The recommended installation uses the precompiled binary wheel from PyPI:

python -m pip install krigekit

The wheel includes the compiled Fortran kriging engine. A Fortran compiler, manual DLL installation, and a source checkout are not required.

Precompiled wheels are available for supported Python versions on:

Platform

Architecture

Linux

x86_64

macOS

arm64 (Apple Silicon) and x86_64

Windows

x86_64

Precompiled wheels support Python 3.10 through 3.14. pip installs the required NumPy, pandas, SciPy, and matplotlib dependencies automatically.

Tip

Upgrade pip first if it does not select a binary wheel:

python -m pip install --upgrade pip
python -m pip install krigekit

Verify the installation#

python -c "import krigekit; print(krigekit.__version__)"

You can also run a small import check:

from krigekit import Kriging, SpaceTimeKriging, VariogramModel

k = Kriging(ndim=2, nvar=1)
del k

Install in a conda environment#

Create an environment and install the precompiled PyPI wheel with pip:

conda create -n krigekit python=3.12
conda activate krigekit
python -m pip install krigekit

mamba can be substituted for conda.

Build from source#

Building from source is intended for contributors, custom compiler settings, or platforms without a compatible wheel. It requires a Fortran compiler such as gfortran, Intel ifx, or Intel ifort.

git clone https://github.com/ougx/krigekit.git
cd krigekit

Compile the shared library:

Linux or macOS with gfortran

python build_lib.py --compiler gfortran

Windows with Intel ifx

call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
python build_lib.py --compiler ifx

Windows with MinGW gfortran

python build_lib.py --compiler gfortran

The build script places libkriging.so, libkriging.dylib, or kriging.dll inside src/krigekit/. Then install the local package:

python -m pip install -e ".[dev]"
python -m pytest

Useful build options:

python build_lib.py --opt debug
python build_lib.py --no-openmp
python build_lib.py --hcache 0

Build the documentation#

From a source checkout:

python -m pip install -e ".[docs]"
python -m sphinx -b html docs docs/_build/html