Managing Python Versions

Most projects we work on require Python 3.8 or greater. A few projects may require an even later version. In either case it is necessary to be able to have multiple versions of Python installed and easily switch between them. Both can be accomplished using pyenv.

Installing pyenv

Use one of the recommended installation methods. If you are using Windows use the pyenv-win fork.

To verify that the installation was successful run the following and ensure that it prints out a version number:

pyenv --version

Installing Python Versions

Running the pyenv versions command will list the versions of Python that pyenv knows about. If you just installed pyenv the output should say that either no Python version are installed:

$ pyenv versions
Warning: no Python detected on the system

Or that a system version (not managed by pyenv) is installed and active (denoted by the asterisk):

$ pyenv versions
* system (set by [...]/.pyenv/version)

Use the pyenv install command to install new versions of Python that will be managed by pyenv. For example to install a Python 3.8 version1:

pyenv install 3.8.15

This can be repeated for any other versions you would like to install. The pyenv install --list command will show all the versions available for installation.

Once you have a few Python versions installed by pyenv, verify they are listed in the output of pyenv versions:

pyenv versions
* system (set by [...]/.pyenv/version)

Selecting a Python Version

pyenv allows setting a Python version at three levels: global, local, and shell2. When getting set up for the first time, the most important version to set is the global version. Depending on the projects you work on, you may never use local or shell versions at all.

To set the global Python version use the pyenv global <version> command. For example:

pyenv global 3.8.15

Verify that this version is now the default Python version by running:

$ python --version
Python 3.8.15

The selected version can also be verified with the pyenv versions. The asterisk indicates the currently selected Python version:

$ pyenv versions
* 3.8.15 (set by [...]/.pyenv/version)

Managing Virtual Environments

Dependencies for different projects need to managed independently in what Python calls virtual environments. Generally each project will have one virtual environment and we use Poetry to manage those environments.

Installing Poetry

First you should have at least one Python 3.7+ version installed and set as the global version through pyenv. Then follow Poetry’s recommended installation methods.

To verify that the installation was successful run the following and ensure that it prints out a version number:

poetry --version

Configuring Poetry

By default Poetry creates virtual environments in one central location. However, we find it more convenient to have the virtual environment created within the project directory itself (specifically in a directory called .venv). This behavior can be enabled by running:

poetry config true

By default Poetry has its own mechanism for locating (but not installing) different Python versions3. This can cause some confusion when used alongside pyenv. Instead of having Poetry attempt to find the right Python version when creating a virtual environment, we want to use pyenv to activate an appropriate version and then use that version to create the virtual environment4. To allow that to work set the following option:

poetry config virtualenvs.prefer-active-python true

Creating Virtual Environments

If you start working on an existing project that already uses Poetry (i.e. it has a pyproject.toml and poetry.lock file in the root of the project), first ensure that you have a compatible Python version. You can find out which Python versions are acceptable for the project by looking at the python entry in the tool.poetry.dependencies section of the pyproject.toml file. If necessary install and/or activate a compatible Python version using pyenv as described above.

Next create the virtual environment and install the projects dependencies into it by running:

poetry install

To start a new project that uses Poetry use either the poetry new or poetry init command. poetry new generates a simple project directory structure, including a pyproject.toml file, based on a name you provide. Using the --src option to create a src directory is recommended:

poetry new --src my-project
cd my-project
poetry install

poetry init will interactively prompt you to build a custom pyproject.toml file. It does not create any other files.

mkdir my-project
cd my-project
poetry init
poetry install

Adding New Dependencies

Use poetry add to add a new dependency to a project.

poetry add requests

This will install the requests package into the project’s virtual environment, add it to the pyproject.toml file, and update the poetry.lock file with information about the exact version installed.

In order to synchronize your poetry.lock file with any manual changes to pyproject.yaml directly, be sure to run:

poetry lock --no-update

Updating Dependencies

The poetry add command can also be used to upgrade a package to a particular version (or range of versions):

poetry add requests@^2.25.0

Or automatically upgrade to the latest version:

poetry add requests@latest

To update a dependency to the latest version that still respects what is specified in pyproject.toml use the poetry update command:

poetry update requests

If you need to debug a dependency tree, use the poetry show command:

poetry show --tree my_package

Running Commands

Use the poetry run command to run commands in the project’s virtual environment. For example to run an arbitrary Python module:

poetry run python src/

Or if your project (or one of its dependencies) provides a CLI (e.g. pytest):

poetry run pytest

For interactive sessions it may be more convenient to activate the virtual environment in which case poetry run will not be necessary. This can be accomplished by running:

poetry shell
python src/  # this will use the virtual environment's Python version
exit                # deactivate the virtual environment

Further Reading


  1. Use pyenv latest --known <prefix> to list latest version that matches <prefix>. For example: pyenv latest --known 3.8

  2. This guide mainly focuses on setting the global Python version. See the pyenv documentation for more information about setting local and shell versions. 

  3. For more information see the poetry env documentation, but it not recommended to use this feature in conjunction with pyenv. 

  4. If an incompatible Python version was active when you attempt to create a virtual environment via poetry install you will see a message like: The currently activated Python version 3.8.15 is not supported by the project (^3.9.0). Or, when attempting to run the Python executable: Current Python version (3.8.15) is not allowed by the project (^3.9.0). In this case, instead of using the poetry env command just delete the virtual environment (rm -rf .venv), ensure a compatible Python version is activated via pyenv (pyenv global 3.9.15), and recreate the virtual environment (poetry install