Python 3

On Ubuntu, the default python interpreter (/usr/bin/python3) produces a segmentation fault because it is linked statically to libpython3.x.so. As a work-around, one can use the python interpreter from e.g. miniconda available from https://conda.io/miniconda.html. In the following, miniconda is installed in /home/abarth/opt/miniconda2 and a conda environment mytest is created.

export PATH="/home/abarth/opt/miniconda2/bin/:$PATH"
export PYTHONNOUSERSITE=1
conda create --name mytest python=3
source activate mytest

The command which python3 should return something like /home/abarth/opt/miniconda2/envs/mytest/bin/python3.

• Install the dependencies

conda install pytest
conda install numpy
conda install netCDF4
export PYTHONPATH="$HOME/src/divand.py/divand:$PYTHONPATH"
export JULIA_PKGDIR=$HOME/Test/julia-v0.6-dev

Setting the variable JULIA_PKGDIR is only necessary if you installed divand.jl in a non-standard location.

• Install pyjulia

git clone https://github.com/JuliaPy/pyjulia
cd pyjulia/
python setup.py clean # unless it is a clean check-out
python setup.py install
cd -

• Re-build the julia package PyCall with the python interpreter from miniconda.

export PYTHON=$(which python3)
julia --eval 'Pkg.build("PyCall")'

The command python -c 'import julia' should return no error.

Run the test suite:

$ pytest

It should return no error:

=============================== test session starts ===============================
platform linux -- Python 3.6.2, pytest-3.2.1, py-1.4.34, pluggy-0.4.0
rootdir: /home/abarth/src/divand.py, inifile:
collecting 0 items

collected 4 items

tests/test_divand.py ..
tests/test_divand_run.py ..

============================ 4 passed in 49.99 seconds ============================

• Create a kernel for jupyter: install ipykernel and create the new kernel:

conda install ipykernel
python -m ipykernel install --user --name myenv --display-name "Python (myenv)"