tebellox / pent

Alternative Command Line Interface for Pipenv

Pent: Alternative Command Line Interface for Pipenv

Pent is a thin wrapper around Pipenv that provides an alternative set of commands for the user to work with. It does not provide any features (everything is backed by Pipenv), but only wrap them so it fits a different kind of brains.

Why?

Pipenv brings order to the chaotic Python project management ecosystem, which is wonderful. With the ecosystem being so chaotic, however, everyone comes to Pipenv with their own set of management rules, and it is impossible to satisfy everyone. It is good to be opinionated, but we can agree to acknowledge we have different preferences, as long as the underlying tooling stay compatible. I want to demostrate how easy it is to put together an alternative solution. Consider doing the same if you don’t like Pipenv’s preferences, instead of trying to convince people to have the same preferences as you do.

As an official PyPA project, Pipenv also is required to be more conservative when it comes to tradeoffs. It still uses virtualenv (as of May 2018), a proven inferior tool to the new built-in venv, because it needs to support Python 2 projects, and lacks resources to support both virtualenv and venv. Pent gives me an oppertunity to move faster and experiement new things without needing to convince people. I can also deal with less consequence if I make any wrong decisions.

How do I use this thing?

Let’s walk through a project’s lifetime to illustrate how Pent’s workflow looks like.

New project

Unlike Pipenv, Pent does not create a Pipfile automatically when you start a new project. You need to call new explicitly. The new command will always create a Pipfile in the current working directory, if it does not exist.

Environment creation

Once you have a Pipfile (either from new or when you clone a repository from somewhere), you need a virtual environment. This is done with init. This command however requires you to pass the Python interpreter (or at least the version of it) you want to use. It does not guess.

Pent also tries to use the built-in venv to create the environment if possible.

Manage dependencies

This part is the same as Pipenv. Call lock to resolve Pipfile’s specifications into Pipfile.lock, sync to install dependencies from Pipfile.lock, install/uninstall to add/remove packages.

Pent uses add and remove to manage package addition and removal, inspired by Yarn. I think this naming makes more sense. They are also more deterministic—they work exactly the same as editing Pipfile and running sync yourself.

Note that there is not update in Pent. Use lock and sync instead.

Other commands

run, check, graph, and clean are identical to their Pipenv counterparts (Pent simply calls them).

shell works a little better than Pipenv’s because Pent detects your currently-running shell, instead of depending on the SHELL environment variable, and does not guess if it cannot make a decision. It also only uses the “fancy” shell, no matter what platform you’re on, and whether you set the environment variable PIPENV_SHELL_FANCY or not.

where works as a combination of pipenv --where and pipenv --venv.

Words of warning

As such is the nature of this project, Pent relies extremely heavily on Pipenv’s internals, and calls a bunch of its internal functions that can break anytime. This should be okay as long as I am actively involved in Pipenv’s development, but I don’t make promises. Use this at your own risk, and be ready to roll up your own sleeves. Do not expect me to fix things for you.