Isolating programming environments for Python using Conda

These are instructions to create blank-slate programming environments Python reproducibility packages using Conda. They require Anaconda or Conda to be previusly installed.

  1. Open the Anaconda terminal
  2. Create a new environment with the command: conda env create --name <env-name> python=3.7
    • Replace <env-name> with your desired environment name
    • You can also specify other versions of Python for the environment. For example: python=3.10
  3. Accept any prompt you might receive to download or install packages
  4. After your environment is created, activate it with conda activate <env-name>
  5. Install all libraries required for your reproducibility package via conda or pip
    • Conda: conda install <package-name>=<version>. For example: conda install pandas=1.4.1. You can include multiple package names and versions in a single prompt
    • Pip: pip install <package-name>==<version> (notice the double equal sign). You can include multiple package names and versions in a single prompt
    • Not including a version installs the laters version available for the Python version of the environment and OS
    • Important: Incompatibilities between packages from conda and pip are a documented problem for Anaconda. This can be avoided by always installing first all packages required from conda and then installing any packages from pip
    • When in doubt between installing a package from conda or pip, use the first. Conda repositories work better for stable and tested versions of packages
  6. If you’re using a jupyter notebook, keep in mind that conda environments don’t automatically work with them. You need to install the packages notebook and nb_conda to be able to use environments in Jupyter notebooks

Some additional useful commands: