This is an (oversimplified) tutorial on getting started on conda. It works for Linux and macOS. All scripts are in bash.

What is and why conda?

Since we are going to discuss issues of anaconda (conda, hereafter), why are we interested in conda? OK, first of all, what is conda? Here is what I cited from wiki.

Anaconda is a distribution of the Python and R programming languages for scientific computing (data science, machine learning applications, large-scale data processing, predictive analytics, etc.), that aims to simplify package management and deployment.”

In short words, the first and foremost reason for me to turn to conda is to download and install packages (and their dependencies and dependencies of dependencies and depend… ) in a single easy step. A second but equally important reason is to manage environments for different projects. For example, I frequently use some tools on python 2, Julia 0.5 or R 2.X. Obviously, I don’t want those tools and their packages to mess up my current important project, so I can use conda to create different environments to separate those tools and side projects. Hence, I avoid conflicts and confusion.

Initiation

You need to do initiation only once.

Type conda info in your terminal. If you see any errors instead of normal information about your conda environment, you need initiation. If you see conda: command not found, you need initiation. Depending on what computers you are using, you may do the following:

If on an administered server or supercomputer

Your administrator has likely installed conda for all users. Try to load the module. For example, on Penn State Roar Supercomputer, do module load anaconda3. Consult your administrator on how to load the module.

# load conda module
module load anaconda3
# init for your $SHELL
conda init bash

If installed locally or on a personal computer

First, install conda from https://www.anaconda.com/ (or it might have been installed by your labmate). Then:

# go to the conda directory
cd /path/to/conda/dir
cd bin
# init
./conda init bash

After proper init, try conda info again. If no error pops, you are good. You need to do initialization only once.

Configure default pkg path

This step is optional but probably saves you from future troubles.

For most of the time, the $HOME directory has limited space, e.g. around 100M-10G. If your environment builds up, it runs out of space quickly. I would recommend avoiding storing any data, performing any experiments, or doing anything with large files in the $HOME directory. Only small configure files should be in $HOME.

However, the default path of conda package cache is in $HOME/.conda/pkg. BAD! I recommend changing the default package cache directory to somewhere else on the disk. See https://docs.anaconda.com/anaconda/user-guide/tasks/shared-pkg-cache/

Open and edit the conda configuration file ($HOME/.condarc. If not present, create one ) to add the following lines:

pkgs_dirs:  
 - /path/to/somewhere/on/disk

Optionally, you can add your favorite channels to the configuration file.

channels:  
 - bioconda   
 - conda-forge  

Create an environment

You can create a new environment as needed, e.g. for a new project, for a tool with a conflicting version of dependencies, or for an older version of a package.

Command to create an environment: (not recommended)

conda create -n env_name

This command creates a new environment named env_name (-n means name) in the default directory $HOME/.conda/envs. As mentioned, this might quickly exhaust all the space in your $HOME dir. I don’t like it.

It is more recommended to create a new environment somewhere else (-p means prefix):

conda create -p /some/other/dir/env_name

Optionally, you can specify the version of your packages during creation, e.g. python 2.7 or R 3.4:

conda create -p /some/other/dir/env_name_py2 	 python=2.7
conda create -p /some/other/dir/env_name_R34	 R=3.4

Activate & Deactivate

You need to activate an environment before using it.

# activate
conda activate env_name
# or equivalently
conda activate /some/other/dir/env_name

To exit an environment, deactivate it:

conda deactivate

Install packages

Google for commands to install your packages. Most commands are like conda install pkg_name. Sometimes there is a -c channel_name argument to specify which channel to use. For example, to install scallop transcriptome assembler, you can use:

conda install -c bioconda scallop

Summary

# create new environment
conda create -p /some/other/dir/env_name

# use environment
conda activate env_name
conda install pkg_name
conda deactivate

Read more about using conda https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html.

Misc

Sometimes conda may throw an error without further information (e.g. conda unexpected error has occurred). It could be because your ~/.condarc is misconfigured. Make sure .condarc is in the correct form and try conda again. If the problem persists, you might want to clear index cache (conda clean -i).