INSTALL_CONDA_VIRT_ENV.md

Install Python packages and netCDF fortran interfaces using Conda virtual environment

Here, we recommend using conda virtual environment to install the python packages and netCDF fortran interfaces in linux system.

Step 1: Install Conda (see here for more details)

First, download conda to the directory you will install conda using wget (please replace 2020.11 with your preferred version):

wget https://repo.continuum.io/archive/Anaconda3-2020.11-Linux-x86_64.sh

Then, install conda:

bash Anaconda3-2020.11-Linux-x86_64.sh

Once the conda is installed locally, set the package channel to conda-forge:

conda config --add channels conda-forge
conda config --set channel_priority strict

Step 2: Create a new virtual environment (see here for more details)

To create an environment with Python 3.7, use:

conda create --name dart-pflotran python=3.7

When conda asks you to proceed, type y:

proceed ([y]/n)?

Once the virtual environment is created, to activate this environment, use:

conda activate dart-pflotran

After you are done, you can deactivate an active environment by:

conda deactivate

Step 3: Install netCDF fortran interface and NCO utility

To install netCDF fortran interface, use (make sure your virtual environment is activated if you are using it and your gfortran compiler is consistent with you netCDF fortran interface):

conda install netcdf-fortran nco

For pinklady user, please install the following specific version of netcdf-fortran (to be compatible with gfortran on pinklady):

conda install netcdf-fortran=4.5.2 nco

Then, revise or add LD_LIBRARY_PATH to your virtual environment (so that when you run conda activate dart-pflotran, the environment variables are set to the values you wrote into the file; and when you run conda deactivate, those variables are erased.):

cd $CONDA_PREFIX # go to the path of the current virtual environment
mkdir -p ./etc/conda/activate.d
mkdir -p ./etc/conda/deactivate.d
touch ./etc/conda/activate.d/env_vars.sh
touch ./etc/conda/deactivate.d/env_vars.sh

Once two empty files ./etc/conda/activate.d/env_vars.sh and ./etc/conda/deactivate.d/env_vars.sh are created, add the following two lines to ./etc/conda/activate.d/env_vars.sh as follows:

#!/bin/sh

export OLD_LD_LIBRARY=${LD_LIBRARY_PATH}
export LD_LIBRARY_PATH=$CONDA_PREFIX/lib:${LD_LIBRARY_PATH}

and add the following two lines to ./etc/conda/deactivate.d/env_vars.sh as follows:

#!/bin/sh

unset LD_LIBRARY_PATH
export LD_LIBRARY_PATH=${OLD_LD_LIBRARY}

Once that is done, first deactivate the virtual environment using:

conda deactivate

and activate the virtual environment again using:

conda activate dart-pflotran

so that what have been written in env_vars.sh would be loaded.

Step 4: Install Python packages

Because we are working on the virtual environment, the following packages are not available now. Therefore, to install the required Python packages, use (make sure your virtual environment is activated if you are using it):

conda install numpy scipy f90nml h5py pandas netcdf4 jupyter jupyterlab matplotlib seaborn

Step 5: Start Jupyter notebook

Once the virtual environment is set and all the packages are installed. To start Jupyter notebook, this can be done either through using (if you installed the virtual environment locally):

jupyter notebook

or through this post (if you installed the virtual environment in a remote machine. Note that please avoid the following port numbers listed in this post.