Installation

Quickstart

The fastest way to install nansat:

  • Install Miniconda on your platform of choice.

# create environment with key requirements
conda create -y -n py3nansat gdal=2.4.2 numpy pillow netcdf4 scipy
# activate environment
source activate py3nansat
# install nansat
pip install nansat

Nansat is now installed. For more details and other methods of installing Nansat, see below.

Requirements

Nansat requires the following packages:

The following packages are optional:

  • Scipy 0.18.1

    • Some mappers will not work without scipy. E.g. sentinel1_l1

  • Matplotlib >=2.1.1,<3.9

    • matplotlib is required for Nansat methods digitize_points() and crop_interactive()

  • Basemap >=1.0.8

    • basemap is required in write_domain_map()

The most tricky to compile yourself is GDAL and Basemap. But one can find pre-built binaries available for different platforms. We recommend to install all dependencies with Conda, from the conda-forge channel. See instructions on this below.

Installing Requirements

You have three main options on how to install the requirements. These are described in the following three sections.

Install dependencies from Anaconda

This is the recommended approach for installing dependencies.

  • Download Miniconda for your platform of choice.

  • Install Miniconda

    • When you install Miniconda on Windows, you will get a new app called “Anaconda Prompt”. Run this to access the conda installation.

    • On Linux/OS X use a regular terminal and make sure PATH is set to contain the installation directory as explained by the installer.

  • Run the following three commands:

    • conda create -n nansat Python=3.6

    • source activate nansat

      • On windows you would omit ‘source’ and just run ‘activate nansat’

    • conda install –yes -c conda-forge pythesint numpy scipy=0.18.1 matplotlib basemap netcdf4 gdal pillow urllib3

Install Pre-built Binaries

One can find pre-built binaries available for different platforms. We do not have an overview over all the possible repositories where you can find binaries. But if you e.g. are on Ubuntu, the following procedure can be used to install dependencies with apt and pip.

sudo apt install -y virtualenv libgdal-dev gdal-bin python3
cd
virtualenv nansat_env
source ~/nansat_env/bin/activate
pip install pythesint pillow netcdf4 urllib3 requests "gdal==$(gdal-config --version)" numpy \
 scipy 'matplotlib<3.9' setuptools basemap

Compile and Build Yourself

If you have the technical expertise to build all dependencies, and need to do it yourself, feel free to do so. If you need some aid, we would recommend you to look at how the corresponding conda-forge feedstocks have been built.

Installing Nansat

Install with pip (preferred method)

Run the following command:

pip install nansat

Nansat will then be added to your site-packages and can be used like any regular Python package.

Install Nansat from source

If you want to install Nansat from source, you first need to install all requirements. Then proceed with one of the following methods

Install from git repository

git clone the master (most stable) or develop (cutting edge) branch, and install:

git clone https://github.com/nansencenter/nansat.git
cd nansat
git checkout master (or a specific tag or branch)
pip install .

Nansat will then be added to your site-packages and can be used like any regular Python package.

Special install for Nansat Developers

If you are working directly on the Nansat source, you need to install Nansat in the following way.

Git clone the branch you are working on, and run the following command from the nansat directory:

pip install -e .

The pixel functions C module is then compiled but no code is copied to site-packages and no linking is performed. Make sure to follow the Nansat conventions if you want to contribute to Nansat.

In addition to the regular dependencies, developers also need to install nose and mock. This can easily be done with

pip install mock

Use a self-provisioned Virtual Machine

Another option to install Nansat in a controlled environment is to use a virtual machine. Configuration for Vagrant and Ansible that brings up and provision a VirtualBox machine is provided in Nansat repository. To start the machine you need to install Vagrant and VirtualBox on your computer; clone or download the nansat source code; and start the machine:

# download nansat source code
git clone https://github.com/nansencenter/nansat.git
cd nansat

#start virtual machine
vagrant up

That’s it! The virtual machine will be started and all software will be installed automatically. To start using Nansat you need to log in to the virtual machine and start Python from the conda environment:

vagrant ssh
source activate py3nansat
python

Use Docker

Docker is a platform for developers and sysadmins to develop, deploy, and run applications with containers (Get started with Docker). We have developed an image that containes compiled Nansat an a number of Python libraries needed for development and running of Nansat. A user can start using the production version of Nansat Docker image:

docker run --rm -it -v /path/to/data:/data nansencenter/nansat ipython

This will mound directory /path/to/data on your host to the directory /data in the container and launch IPython where Nansat is available.

For developing Nansat you needs access to the code both from the container (to run it) and from the host (to edit it). For this purpose you should clone Nansat repository and do the following steps: 1. Build pixelfunctions inplace

docker run --rm -it -v `pwd`:/src nansencenter/nansat_base pip install -e .

2. Run container with mounting of the current directory into /src. In this case Python will use Nansat from /src/nansat (the directory shared between host and container):

# launch Python with Nansat in container
docker run --rm -it -v `pwd`:/src nansencenter/nansat_base python

# ...or run tests
docker run --rm -it -v `pwd`:/src nansencenter/nansat_base python -m unittest discover nansat.tests

Alternatively you can run the script build_container.sh. The script will build the image with Python libraries from Anaconda, compile the Nansat code inplace and create a container for running Nansat. You can then start container:

docker start -i nansat
# and run tests:
(base) root@d1625f2ce873:~# python -m unittest discover nansat.tests