Installing wflow on Ubuntu 18.04 LTS

[Sebastian]

Hello,

I am not sure if this is the right place to post this (in the Issues section). However, since it took me some time to install wflow on Ubuntu 18.04 LTS, I figured I'll share this, maybe it helps somebody. For my specific application, I want the Python distribution to be as lean as possible. Therefore, I was not going for the Anaconda environment solution.

Compile PCRaster from source:

sudo apt install cmake gcc g++ git qtbase5-dev libncurses5-dev libqwt-qt5-dev libxerces-c-dev libboost-all-dev libgdal-dev python3-numpy python3-docopt
wget http://pcraster.geo.uu.nl/pcraster/4.2.1/pcraster-4.2.1.tar.bz2
tar xf pcraster-4.2.1.tar.bz2 && cd pcraster-4.2.1
mkdir build && cd build
cmake -DFERN_BUILD_ALGORITHM:BOOL=TRUE -DCMAKE_INSTALL_PREFIX:PATH=$HOME/pcraster -DPYTHON_EXECUTABLE:FILEPATH=/usr/bin/python3 ..
cmake --build .
make install
export PATH=$PATH:$HOME/pcraster/bin
export PYTHONPATH=$PYTHONPATH:$HOME/pcraster/python

Since the automatic installation of the required Python packages did not work for me, I installed them manually using pip

sudo apt install python3-pip
pip3 install numpy
pip3 install scipy
pip3 install GDAL==$(gdal-config --version | awk -F'[.]' '{print $1"."$2}') --global-option=build_ext --global-option="-I/usr/include/gdal"
pip3 install netcdf4
pip3 install pyproj
sudo apt-get install libllvm-9-ocaml-dev libllvm9 llvm-9 llvm-9-dev llvm-9-doc llvm-9-examples llvm-9-runtime
export LLVM_CONFIG='/usr/bin/llvm-config-9'
pip3 install numba
pip3 install python-dateutil

Afterwards I was cloning the git repository to my $Home directory

cd $HOME
git clone https://github.com/openstreams/wflow.git

I had to excluded the required packages in the setup.py file, which is located in $HOME/wflow/setup.py

install_requires=[
    #   "numpy",
    #    "scipy",
   #     "gdal",
   #     "netCDF4",
   #     "xarray",
    #    "cftime",
   #     "numba",
   #     "pyproj",
    #    "python-dateutil",
    ], 

At last wflow can be installed by using pip

cd wflow
pip3 install -e .

Cheers, Sebastian

[Martijn Visser]

Thanks for taking the time to write this down, I'm sure it will help some people!

Shall I link to this issue from https://github.com/openstreams/wflow#install-using-pip and close this issue?

For my specific application, I want the Python distribution to be as lean as possible. Therefore, I was not going for the Anaconda environment solution.

I understand. I guess Miniconda was also not an option for this case? The PCRaster Conda build makes installation quite a bit easier. Nevertheless, it's good to document the pip install as well

[Sebastian]

Hello Martijn,

Miniconda would have been an option as well. However, since I have to run many Python based software packages for my application, I prefer to have all the packages in the systems Python environment and keep it as slim as possible ;)

Yes, of course you can link the post.

Cheers, Sebastian