Starting with an anaconda 3.7 env on linux: https://www.anaconda.com/download/#linux
- install anaconda
conda install netCDF4 pyproj pysal rtree shapely gdal fiona
(this can be quite slow!)
Add this to your .profile, .bash_profile, .zshenv,..etc:
declare -x PYTHONPATH=/path/to/RiverObs/src/:$PYTHONPATH
- Updating the below instructions along with
setup.py
would be a great project for a contributor!
These are the instructions for installing the RiverObs package written by Ernesto Rodriguez in a Unix (linux or Mac) machine with an anaconda python setup. The nominal installation instructions have been tested with python3.6, but should also work with python2.7. Future developments may stop supporting python2.7, as it is no longer the community standard.
In what follows, it is assumed that the environment variable RIVER_DIR has been set to point to the root directory of the RiverObs package cloned by git. For instance, using bash
export RIVER_DIR=/home/erodrigu/SWOT/RiverObs
Note that the dependence on scikit-image is optional and required only if one wants to vectorize GWDLR data. In that case, a working grass installation is required (tested with grass 6.4; grass70 beta has a bug in r.to.vector as of this writing).
To make sure that you are retrieving the same version packages as have been used for testing, make sure that the conda-forge channel is added to your conda configuration. This can be done by issuing the command
conda config --add channels conda-forge
or modifying your ~/.condarc file to look something like this:
channels:
- conda-forge
- defaults
show_channel_urls: true
To create an anaconda virtual environment, execute (Simplest):
conda create -n RiverObs python=3.6 numpy jupyter notebook matplotlib
gdal scipy pip scikit-image statsmodels pysal pandas pytables
shapely netcdf4 sphinx numpydoc rtree pyproj
some thrid party packages may have trouble with the newer python 3.6, if you have trouble you can try with 3.5. Also, it may be necessary to use version 8d version of jpeg. If so try the following:
conda create -n RiverObs python=3.5 numpy jupyter notebook matplotlib gdal scipy pip scikit-image statsmodels pysal pandas pytables shapely netcdf4 sphinx numpydoc rtree pyproj jpeg=8d
Here is what I got working on a linux box with all the versions explicitly stated:
conda create -n RiverObs python=3.5 numpy=1.13.1 jupyter=1.0.0 notebook=5.0.0 matplotlib=2.0.2 gdal=2.1.0 libgdal=2.1.0 scipy=0.19.1 pip=9.0.1 scikit-image=0.13.0 statsmodels=0.8.0 pysal=1.13.0 pandas=0.20.3 pytables=3.4.2 shapely=1.5.16 netcdf4=1.2.4 sphinx=1.6.3 numpydoc=0.7.0 rtree=0.8.3 pyproj=1.9.5.1 jpeg=8d
or, if you want to keep the code and executables under the RiverObs folder:
cd $RIVER_DIR
conda create -p $RIVER_DIR/anaconda python=3.6 numpy jupyter notebook matplotlib
gdal scipy pip scikit-image statsmodels pysal pandas pytables
shapely netcdf4 sphinx numpydoc rtree pyproj
Note: if you must run python 2.7, substitute python=2.7 in the lines above (not recommended).
To activate this environment, if the first option was used, type
source activate RiverObs
or, if building in the RiverObs folder,
source activate $RIVER_DIR/anaconda
if anaconda/bin is in your path. Otherwise, use /path/to/anaconda/bin/source.
To deactivate this environment, type
source deactivate
If you would like to use jupyter notebooks within the RiverObs environment, issue the following command while inside the environment:
python -m ipykernel install --user
Then, to build the RiverObs and associated packages:
cd $RIVER_DIR
python setup.py install --force
For an anaconada local virtual environment, this will install the libraries in
$RIVER_DIR/anaconda/python3.6/site-packages
and the executables in
$RIVER_DIR/anaconda/bin
Otherwise, they are in similar directories in ~/anaconda/envs/RiverObs