Anaconda¶

Anaconda is a distribution (or package manager) for data science. It’s typically associated with Python, but in actuality it installs a more broad range of languages than that, such as R and system packages.

Install¶

You’ll likely want to follow the install instructions for your distribution, or use miniconda for a smaller install. A headless “quick install” for Linux might look like this:

# Anaconda for Python
cd /tmp
wget https://repo.anaconda.com/archive/Anaconda3-2021.05-Linux-x86_64.sh

# Install to "anaconda3" in HOME, and add to your bash profile
bash Anaconda3-2021.05-Linux-x86_64.sh -p $HOME/anaconda3 -b
echo "PATH=\$HOME/anaconda3/bin:\$PATH" >> $HOME/.profile
echo export PATH >> $HOME/.profile

And you can browse the archive for more recent or earlier versions.

Once you have it installed, you can source your bash profile (or open up a new terminal) and it should be on your PATH.

$ which conda
$ which python
# /home/vanessa/anaconda3/bin/python

If you need Anaconda just for Python, you’re good to go! If you want to create environments for different projects, continue reading.

Create a Conda Environment¶

An Anaconda (or miniconda) environment is referred to as a conda environment. Yes, it has some snakey references. Here is how to create an environment:

$ conda create --name myenv
  
...
#
# To activate this environment, use
#
#     $ conda activate myenv
#
# To deactivate an active environment, use
#
#     $ conda deactivate

You can then follow the instruction to activate or deactivate the environment.

$ conda activate myenv

What happens when you activate? Well you’ll notice that your command prompt changes to indicate the name of the active environment:

(myenv) me@computer$

If you now do conda install, the software will be installed to that environment. If you have already installed software in that environment, it will be on your PATH (given an executable in the environment’s bin) and of course the Python libraries are available on your PYTHONPATH. As a bonus, you also get a bunch of freebie tools that come packaged with Anaconda (Jupyter, curl, openssl, etc.)

$ which curl
/home/vanessa/anaconda3/bin/curl

When you are done in your environment, just deactivate it:

$ conda deactivate

This environment is going to stay with your conda install until you delete it. If you can’t remember the names of your environments, just list them:

$ conda env list

# conda environments:
#
base                  *  /home/vanessa/anaconda3
myenv                    /home/vanessa/anaconda3/envs/myenv

To delete the environment (and make sure you have deactivated first):

$ conda env remove --name myenv

Install Software¶

Installing software is fairly straight forward. You can find packages from Anaconda or from conda-forge. Conda forge is a community package repository, and packages build from conda-feedstocks. If you’d like to contribute a package, you can read the instructions here. When you install from conda forge, you usually need to add the channel to search:

$ conda config --add channels conda-forge 

It’s a super cool package manager because the entire infrastructure is maintained over GitHub, and contributing is really easy. Let’s walk through installing something. Let’s say we want to search for a package first.

$ conda search requests
Loading channels: done
# Name                       Version           Build  Channel             
requests                      2.18.4  py27hc5b0589_1  pkgs/main           
...
requests                      2.25.1    pyhd3eb1b0_0  pkgs/main           
requests                      2.26.0    pyhd3eb1b0_0  pkgs/main   

You’ll search across your channels and see all the versions available. Just omit the version to get latest:

$ conda install requests

or choose a specific version!

$ conda install requests@2.26.0

And remember if you want to install from a different channel, you need to add it first. If you don’t want to add it, you can search and specify it directly (and this appears to be faster):

$ conda search --channel conda-forge deid

Sometimes conda forge will have a release of common software (e.g., R) before official Anaconda - the two communities collaborate together and conda-forge can be seen like a bit of a testing ground. Finally, remember that when you are interacting with your environment programmatically, such as in a script, you’ll need to activate the environment first and then interact with it.

Mamba¶

A Mamba is another fast-moving, venomous snake, and another cross platform package manager! Some people choose to use it because it’s faster to resolve environments than Anaconda or Miniconda. You can actually install it with conda, and this is the recommended approach:

$ conda install mamba --name mamba --channel conda-forge

And then install software as follows:

$ mamba install xtensor-r --channel conda-forge

That’s it! If you have any questions or requests for documentation, please don’t hesitate to open an issue.