Dec 02, 2022


Install is a bit wonky, they want you to do a user installation:

pip install --user pipenv

Which installs pipenv into a bin directory inside site.USER_BASE which for me is $HOME/.local; if you're unsure what yours is you can do python -c "import site; print(site.USER_BASE)". To access it, I had to add $HOME/.local/bin to my $PATH variable, since I had never used anything in that directory before.

Side note: from my normal asdf-installed python, I get ~/.local as the user base; if I call the python command above from inside a pipenv-activated shell, I get ~/Library/Python/3.10, despite that fact that both pythons are the same. I don't understand why

Once it is installed, you can do pipenv install <package>, and it will add it to a file called Pipfile, and store a lock file in Pipfile.lock. If you're familiar with bundler or npm, these should be familiar to you.

Here's what the first Pipfile I made, after pipenv install nba_api pandas pyarrow fastparquet:

url = "https://pypi.org/simple"
verify_ssl = true
name = "pypi"

nba-api = "*"
pandas = "*"
pyarrow = "*"
fastparquet = "*"


python_version = "3.10"

It has a bit of metadata, then specifies the packages I have installed, and the version of python it requires.

By default, pipenv stores your packages in a virtualenv it creates in {site.USER_BASE}/share/virtualenvs/{directory_name}-{hash}; for me in this case it is stored in ~/.local/share/virtualenvs/test_nba_api-xg-_2li1.

That name is not stored in the Pipfile, so if you were to move your code or rename the directory it's in, you would lose access to the virtualenv that Pipenv created.

side note: I'm not sure why they don't just store that metadata in the Pipfile!

If you don't want that, you can set an environment variable PIPENV_VENV_IN_PROJECT=1, in which case it will store the virtualenv in a directory called .venv in the directory you do your first pipenv install from.


↑ up