Anaconda

The Anaconda Distribution provides a popular method of installing Python, R and associated libraries and packages.

The Anaconda Distribution (https://www.anaconda.com/) is available for Linux, Windows and Apple systems, and will install all files within your home directory by default, and as such, installation does not require any elevated / administrative permissions.

Installation

There are various different Anaconda installers available, with the default installation containing Python version 2 or version 3 and many useful packages (https://www.anaconda.com/distribution/). However, it is worth noting that the installation can require several Gigabytes of disk space.

As an alternative the Miniconda installer (https://docs.conda.io/en/latest/miniconda.html) provides a minimal base installation which can be used to create environments and install required packages.

For Linux systems, the Miniconda 64 bit version 3 installer is recommended, which can be used to create Python version 2 and version 3 as well as R environments:

https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

As mentioned, Anaconda installations can get quite large, and will install in the home directory by default, so it is worth checking your quota (quota -s) or selecting a suitable location (for example, if you have access to some large volume disk space) before installation.

The Miniconda installer can be downloaded and made executable with the following commands:

$ wget 'https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh'
$ chmod +x Miniconda3-latest-Linux-x86_64.sh

The installer can then be run:

$ ./Miniconda3-latest-Linux-x86_64.sh

The installation process will ask for the terms of the license to be accepted, and then ask for an installation directory, defaulting to a directory named miniconda3 in the home directory. If you wish to select an alternative location, enter the full path to the required location. If installing in your home directory, a sensible path to choose might be /home/username/conda, replacing username with your own username.

After the installer has finished extracting the files in to the selected location, it will ask whether you wish to run conda init to initialise the installation. Selecting yes will update the ~/.bashrc so that the Anaconda programs will be available in future sessions.

Once the installation process has been completed, the installer is no longer required, and can be removed:

$ rm Miniconda3-latest-Linux-x86_64.sh

Initial Configuration

Once the installation is complete, if the ~/.bashrc file was updated successfully, the Miniconda installation will be active when you either run source ~/.bashrc, or open a new shell. You should see that the prompt is now prefixed with (base), to indicate that the base environment in the Miniconda installation is active:

(base) [see1-234:earxyz:1]$

At this point the python command will now default the version of Python installed with Miniconda, and the conda command can be used to install packages in the base environment:

(base) [see1-234:earxyz:3]$ which python
/home/earxyz/conda/bin/python
(base) [see1-234:earxyz:3]$ python -V
Python 3.7.4
(base) [see1-234:earxyz:4]$ conda install numpy

However, an alternative method may be more advisable. Anaconda environments can include a large number of different programs and libraries which can cause conflicts with system versions of the same files, so it can be better to create environments for specific purposes, activating and deactivating the environments as required.

To stop the base environment being automatically activated when a shell is opened, the following command can be run:

$ conda config --set auto_activate_base false

It may also be worth adding the conda-forge channel to configuration. The conda-forge channel is an additional repository from which packages can be installed, and includes many popular scientific packages, such as iris. The following commands will add the conda-forge channel:

$ conda config --add channels conda-forge

These changes update the file ~/.condarc, and after they have been run, and a new shell is opened, the base channel will no longer be activated by default, but the conda command will be available to allow creating new environments:

$ which python
/usr/bin/python
$ which conda
/home/earxyz/conda/condabin/conda
$ cat ~/.condarc
auto_activate_base: false
channels:
  - conda-forge
  - defaults

Creating Environments

The command conda create can be used to create environments containing the required packages. For example, to create an environment containing Python version 3 and the Spyder graphical development environment:

$ conda create -n py3_spyder python=3 spyder

Environments are created within the envs directory of the installation folder, and the environments can be activated and deactivated with the conda activate and conda deactivate commands:

[see1-234:earxyz:12]$ ls conda/envs/
py3_spyder
[see1-234:earxyz:12]$ conda activate py3_spyder
(py3_spyder) [see1-234:earxyz:13]$ which python
/home/earxyz/conda/envs/py3_spyder/bin/python
(py3_spyder) [see1-234:earxyz:14]$ python -V
Python 3.7.3
(py3_spyder) [see1-234:earxyz:15]$ which spyder
/home/earxyz/conda/envs/py3_spyder/bin/spyder
(py3_spyder) [see1-234:earxyz:16]$ conda deactivate
[see1-234:earxyz:18]$

To create a Python version 2 environment containing the numpy and scipy packages:

$ conda create -n py2_scipy python=2 numpy scipy

The available environments can be listed with conda env list:

$ conda env list
# conda environments:
#
base                  *  /home/earxyz/conda
py2_scipy                /home/earxyz/conda/envs/py2_spyder
py3_spyder               /home/earxyz/conda/envs/py3_spyder

Installing Packages

Once an environment has been activated, additional packages can be installed within that environment with the conda install command:

(py3_spyder) [see1-234:earxyz:24]$ conda install iris

The conda search command can be used to search for available packages:

(py3_spyder) [see1-234:earxyz:25]$ conda search 'obsp*'
Loading channels: done
# Name                       Version           Build  Channel             
obspy                          1.0.2          py27_0  conda-forge         
obspy                          1.0.2          py27_1  conda-forge
...

To install a specific version of a package, the version can be specified with ==:

(py3_spyder) [see1-234:earxyz:26]$ conda install obspy==1.0.3

It is also possible to install packages using the pip command, if a package is available in the PyPi repositories, but not available in the Anaconda channels (pip install packagename).

The packages which are currently installed in an environment, their version information and installation source can be viewed by running:

(py3_spyder) [see1-234:earxyz:27]$ conda list

Updating Packages

The conda update command can be used to update packages. For example, to update spyder:

(py3_spyder) [see1-234:earxyz:28]$ conda update spyder

To update all packages in an environment:

(py3_spyder) [see1-234:earxyz:29]$ conda update --all

If an environment is not updated for some time, changes to the environment (such as installing a new package) may cause version conflicts, so it may be wise to either create a new environment, or update all packages in the environment when making changes.

Installing R

As well as Python environments, the conda create command can also be used to create environments for running the R program.

To create an environment containing R and rstudio:

$ conda create -n R R rstudio-desktop

This will create an environment named R, which can be activated with conda activate R, and once the environment is active the R and rstudio programs will be available.