Virtual environments are a way to isolate your project from the rest of your system. This is important because it allows you to install packages that are specific to your project, without affecting the rest of your system.

There are several ways to create virtual environments. The most popular (and recommended) is with Anaconda. After installing Anaconda or Miniconda (light version), you create a new environment like so:

# create new environment, press enter to accept
conda create -n project_env python=3.9

# view available environments
conda info --envs

# activate environment
conda activate project_env

# deactivate environment
(project_env) conda deactivate

For machines really light on memory (e.g. Raspberry Pi), you can use Virtualenv:

# install library if not already
pip install virtualenv

# create virtual environment (creates folder called project_env)
python3 -m venv project_env

# activate virtual environment
source project_env/bin/activate

# deactivate virtual environment
(project_env) deactivate

Note that when the virtual environment is activated, it will typically appear in parenthesis in the command line.

Inside your virtual environment, you can install packages specific to your project. It is highly recommended to keep track of the packages you install, so that others (including yourself) can easily recreate the same virtual environment. There are three common approaches to storing and keeping track of packages:

• requirements.txt: This is a simple text file that lists all the packages you have installed. You can create this file by running:

  (project_env) pip freeze > requirements.txt

  You can then install all the packages in this file by running:

  (project_env) pip install -r requirements.txt

• environment.yml: This is a YAML file that lists all the packages you have installed. You can create this file by running:

  (project_env) conda env export > environment.yml

  You can simulatenously create the environment and install all the packages in this file by running:

  conda env create -f environment.yml

  You can check that the environment was created by running:

  conda env list

  The the name of the environment is specified at the top of environment.yml.

  Note that this approach is specific to Anaconda / Miniconda. More information can be found here.

• A more involved approach to package your project, such that it can be installed via pip with the necessary dependencies:

  # local install
  (project_env) pip install -e .
  
  # if on PyPi
  (project_env) pip install <PACKAGE_NAME>

  This approach is discussed in PACKAGING.