5. Manage Your Python Package Versions

Install¶

Install poetry with the following command: curl -sSL https://install.python-poetry.org | python3 -
Restart your terminal
Enter poetry and you should see a list of commands returned

NOTE: If that install doesn't work for you, use this workaround:

Activate the conda environment for your project
Run pip install poetry==1.7

Specify Project Dependencies¶

When you are first creating the project, you will use poetry init to interactively create the pyproject.toml file.

For example, to create specify this project's dependencies, I followed the steps below:

Create a conda environment with the same name as the project folder
- conda create -n python-dev-setup python=3.9
Run poetry init
- Press enter to accept the defaults until you get to the questions on defining dependencies interactivaly
- I find it easier to specificy the dependencies individually later vs. at this step. So, I enter no to both questions and then press enter to confirm generation.
You will now see a pyproject.toml file created

NOTE: If you are converting a project that currently uses a requirements.txt file, you can just fill in the template below instead of using poetry init:

    ```
    [tool.poetry]
    name = "{repo/project name}"
    version = "1.0.0"
    description = "{free text description}"
    readme = "README.md"
    repository = "{git repo link}"
    documentation = "{site url if applicable}"
    authors = [
        "{full name} <{email}>"
    ]

    [tool.poetry.dependencies]
    python = "=={Python version}"

    [tool.poetry.dev-dependencies]

    [build-system]
    requires = ["poetry-core>=1.0.0"]
    build-backend = "poetry.core.masonry.api"
    ```

Create poetry.lock file¶

The first time you run poetry install or add a package with poetry add {package name}, the poetry.lock file is created

After adding the dependencies needed with poetry add, push the pyproject.toml and poetry.lock files to the remote repo.

Install Project Dependencies¶

Now that the lock file is created, other team members can simply:

Create a conda environment using python version 3.9
Run poetry install

After these steps, they will be working with the same dependencies you are using!

Useful Poetry Commands¶

Install requirements¶

poetry install

Add a package¶

This will install the latest package version. The pyproject.toml will show a version greater than or equal to the version is required and the poetry.lock file will update to specify the specific version installed.

Add the latest version fo the package meeting other project depencies:
- poetry add {package name}
Install specific version of the package:
- poetry add {package name}={version}
Install a model from a URL:
- poetry add {package name}@{url}
Install an editable local package:
- poetry add --editable {relative path to package}

Remove a package¶

This will update the pyproject.toml and poetry.lock files

poetry remove {package name}

Convert `poetry.lock` file to `requirements.txt`¶

poetry export -f requirements.txt -o requirements.txt --without-hashes

Troubleshooting¶

If poetry is endlessly searching for a package to install, try deleting the poetry.lock file and running poetry install again