What is PDM?
PDM is meant to be a next generation Python package management tool. It was originally built for personal use. If you feel you are going well with Pipenv
or Poetry
and don't want to introduce another package manager, just stick to it. But if you are missing something that is not present in those tools, you can probably find some goodness in pdm
.
PEP 582 proposes a project structure as below:
foo
__pypackages__
3.8
lib
bottle
myscript.py
There is a __pypackages__
directory in the project root to hold all dependent libraries, just like what npm
does. Read more about the specification here.
Highlights of features
- PEP 582 local package installer and runner, no virtualenv involved at all.
- Simple and relatively fast dependency resolver, mainly for large binary distributions.
- A PEP 517 build backend.
- A full-featured plug-in system.
- PEP 621 project metadata format.
Why not virtualenv?
The majority of Python packaging tools also act as virtualenv managers to gain the ability to isolate project environments. But things get tricky when it comes to nested venvs: One installs the virtualenv manager using a venv encapsulated Python, and create more venvs using the tool which is based on an encapsulated Python. One day a minor release of Python is released and one has to check all those venvs and upgrade them if required.
PEP 582, on the other hand, introduces a way to decouple the Python interpreter from project environments. It is a relative new proposal and there are not many tools supporting it (one that does is pyflow, but it is written with Rust and thus can't get much help from the big Python community and for the same reason it can't act as a PEP 517 backend).
Installation
PDM requires python version 3.7 or higher.
Like Pip, PDM provides an installation script that will install PDM into an isolated environment.
For Linux/Mac
curl -sSL https://raw.githubusercontent.com/pdm-project/pdm/main/install-pdm.py | python -
For Windows
(Invoke-WebRequest -Uri https://raw.githubusercontent.com/pdm-project/pdm/main/install-pdm.py -UseBasicParsing).Content | python -
The installer will install PDM into the user site and the location depends on the system:
$HOME/.local/bin
for Unix%APPDATA%\Python\Scripts
on Windows
You can pass additional options to the script to control how PDM is installed:
usage: install-pdm.py [-h] [-v VERSION] [--prerelease] [--remove] [-p PATH] [-d DEP]
optional arguments:
-h, --help show this help message and exit
-v VERSION, --version VERSION | envvar: PDM_VERSION
Specify the version to be installed, or HEAD to install from the main branch
--prerelease | envvar: PDM_PRERELEASE Allow prereleases to be installed
--remove | envvar: PDM_REMOVE Remove the PDM installation
-p PATH, --path PATH | envvar: PDM_HOME Specify the location to install PDM
-d DEP, --dep DEP | envvar: PDM_DEPS Specify additional dependencies, can be given multiple times
You can either pass the options after the script or set the env var value.
Alternative Installation Methods
If you are on MacOS and using homebrew
, install it by:
$ brew install pdm
If you are on Windows and using Scoop, install it by:
PS> scoop bucket add frostming https://github.com/frostming/scoop-frostming.git
PS> scoop install pdm
Otherwise, it is recommended to install pdm
in an isolated environment with pipx
:
$ pipx install pdm
Or you can install it under a user site:
$ pip install --user pdm
Quickstart
Initialize a new PDM project
$ pdm init
Answer the questions following the guide, and a PDM project with a pyproject.toml
file will be ready to use.
Install dependencies into the __pypackages__
directory
$ pdm add requests flask
You can add multiple dependencies in the same command. After a while, check the pdm.lock
file to see what is locked for each package.
Run your script with PEP 582 support
Suppose you have a script app.py
placed next to the __pypackages__
directory with the following content(taken from Flask's website):
from flask import Flask
app = Flask(__name__)
@app.route('/')
def hello_world():
return 'Hello World!'
if __name__ == '__main__':
app.run()
If you are a Bash user, set the environment variable by eval "$(pdm --pep582)"
. Now you can run the app directly with your familiar Python interpreter:
$ python /home/frostming/workspace/flask_app/app.py
* Serving Flask app "app" (lazy loading)
...
* Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
Ta-da! You are running an app with its dependencies installed in an isolated place, while no virtualenv is involved.
For Windows users, please refer to the doc about how to make it work.
If you are curious about how this works, check this doc section for some explanation.
Docker image
$ docker pull frostming/pdm
Badges
Tell people you are using PDM in your project by including the markdown code in README.md:
[![pdm-managed](https://img.shields.io/badge/pdm-managed-blueviolet)](https://pdm.fming.dev)
PDM Eco-system
Awesome PDM is a curated list of awesome PDM plugins and resources.
FAQ
__pypackages__
?
1. What is put in PEP 582 is a draft proposal which still needs a lot of polishing. For instance, it doesn't mention how to manage CLI executables. PDM makes the decision to put bin
and include
together with lib
under __pypackages__/X.Y
.
2. How do I run CLI scripts in the local package directory?
The recommended way is to prefix your command with pdm run
. It is also possible to run CLI scripts directly from the outside. PDM's installer has already injected the package path to the sys.path
in the entry script file.
3. What site-packages will be loaded when using PDM?
Packages in the local __pypackages__
directory will be loaded before the system-level site-packages
for isolation.
__pypackages__
folder for deployment?
4. Can I relocate or move the You'd better not. The packages installed inside __pypackages__
are OS dependent. Instead, you should keep pdm.lock
in VCS and do pdm sync
on the target environment to deploy.
pdm
to manage a Python 2.7 project?
5. Can I use Sure. The pdm
itself can be installed under Python 3.7+ only, but it doesn't restrict the Python used by the project.
Credits
This project is strongly inspired by pyflow and poetry.
License
This project is open sourced under MIT license, see the LICENSE file for more details.