PDM It's a new one Python Package manager , Maybe you don't know its existence yet , But actually PDM It has been born for two years , And in 2021 Released in 1.0 edition , At present, the highest version is 1.12.8.
Just heard that PDM when , I subconsciously thought it was Python Development Manager, Another and Pipenv and Poetry The same virtual environment management tool for changing soup without changing medicine .
Until I turned to the author's blog , not have understood until then PDM The full name is Python Development Master, It's even better than I thought .
It is worth mentioning that ,PDM The author is PyPa member 、Pipenv One of the main maintainers at present , most important of all , He is a Chinese , So this is a tool developed by Chinese people .
Early package managers （ Such as Pipevn,Poetry）, Are based on virtual environments , Virtual environment is mainly used to isolate the project development environment , But if it comes to virtual Environment nested virtual environment , The problem is difficult , There are always problems .
PDM Benefit from one 2018 Year of PEP The proposal （PEP582,Python local packages directory）, Completely abandon the virtual environment .
From the author's blog , The reason why we had to build a wheel again , It's all because Pipenv and Poetry It's not easy to use , It happens that PEP582 , Can develop an epoch-making Python Package management tools , It is PDM .
PDM It includes the following features ：
PEP 582 Local project library Directory , Support installation and running commands , There's no need for a virtual environment .
A simple and relatively fast dependency parser , Especially for large binary packages .
compatible PEP 517 Build the back end of , For building release packages ( Source format and wheel Format )
Have a flexible and powerful plug-in system （ If there is a plug-in system, it will directly open a level ）
PEP 621 Metadata format
image pnpm The same centralized installation cache , Save disk space
Even though PDM It was developed by Chinese people , But considering internationalization , The official website documents are all in English .
I spent the whole day , After reading the document , Digested 70% Of PDM usage , Now share your experience , I'll start with you PDM To be helpful to .
About PDM, It's a lot of content , I intend to introduce it completely in two parts ：
Entry level tutorial for novices
A tutorial for ashes
This is the first one , Let's start with pdm There is a framework understanding of the basic usage of , and pdm Real competitiveness, please continue to pay attention to the follow-up articles .
PDM There are many ways to install , On the official website 6 Kind of , such as pip、pipx、homebrew etc.
In previous articles , I recommended pipx Tools , It is very easy to use when installing the package of command-line application .
details ： Use pipx Tool installation command line tool
And then PDM Is a command line tool , So I also recommend pipx install , It is convenient to manage the command line uniformly
perform pipx install pdm You can install
PDM Only Python 3.7+ Only the version of , Use other methods to install , Make sure your Python edition , But use pipx You don't need to worry about .
perform pdm init Will start initializing , When initializing , Some information that will let you choose the project
Whether to upload PyPI
Rely on the Python edition
License type
The author information
Email information
I have... On my machine Python 2.7 and Python 3.10 Two versions , When initializing the project, all the files on the machine will be deleted Python The versions have been scanned , Will let you choose the project Python edition .
When it's done ,PDM Will take your choice as toml Format to write pyproject.toml In profile .
pdm There are so many orders , Use -h You can see the help menu
and Poetry equally , The installation uses add command , but pdm Of add Than poetry To use , Mainly reflected in grouping , Please pay attention to the following articles
Use pdm list You can list the packages installed in the current environment in the form of a list
Add another --graph You can view in a tree , The hierarchy of direct and indirect dependency packages is clear at a glance
pdm list There are two more options ：
–freeze： With requirements.txt List installed packages in the format of
–json： With json List installed packages in the format of , But it has to be with --graph Use at the same time
To view the details of a package , Direct use pdm show that will do
Delete package using remove command
Without any parameters , You can print out the environment configuration of the project
If you want to change it , Just add key and value As a parameter , To modify pypi Take the mirror proxy as an example
It turned out that it was watercress , Now I want to change it to Ali source , Just execute the following command , Comparable poetry It's much more convenient ～
pdm config There are many configurations , If you want to find out one by one, you can go to the official website ：https://pdm.fming.dev/configuration/
Want to be in pdm Execute commands or projects in an environment , have access to run command , If the project is executed , There are many parameters , Can be in pyproject.toml Configure command aliases , The specific use , Please look back
Use info command , You can view the environment information of the current project
Updated words , In a simple scenario , Just use the following two
# Update all packages
pdm update
# Update a package
pdm update <pkg>
Complex scenes ,pdm It's all for you , It offers many options , It can be used as needed （ If there is any explanation error in the following , Please help to correct ）
–save-compatible： Projects rely on compatible versions
–save-wildcard： Save wildcard version （ I don't understand ）
–save-exact： Save the package with the specified exact version
–save-minimum： Keep the minimum version of the package
–update-reuse： Try to update only the packages specified on the command line , The dependent package cannot be updated without updating
–update-eager： Update a package and update its dependent packages （ Recursive upgrade ）
–prerelease： Allow early release （ I don't understand ）
–unconstrained： Ignore package version constraints , You can upgrade the package to the latest version
–top： Only updated in pyproject.toml My bag
–dry-run： A test run , Without modifying lock file
–no-sync： Update only lock file , But don't update the package
If your dependent package has set grouping , Group updates can also be specified
pdm update -G security -G http
You can also specify a group to update a package in the group
pdm update -G security cryptography
Add another -d You can specify dev rely on
# Update all dev rely on
pdm update -d
# to update dev A package that depends on the next group
pdm update -dG test pytest
similarly , You can also specify --prod perhaps --production Upgrade non dev （ That is production ） My bag .
When you are initializing pdm Project time , You have selected the current Python Version and available Python Version range , If you want to change later , have access to use command , However, the version is subject to the previously set version range .
Suppose the allowable range is python 3.9+, Currently using python 3.10, You can switch directly .
pdm use python3.9
stay pyproject.toml add to [tool.pdm.scripts] You can set the shortcut command alias , If the execution of the project has many parameters , This method of setting aliases will be useful .
[tool.pdm.scripts] There are two forms
# The first one is
[tool.pdm.scripts]
start = "python main.py"
# The first one is
[tool.pdm.scripts]
start = {cmd = "python main.py"}
But if you want to add comments to the parameters , You have to use the second method , Such as this
[tool.pdm.scripts]
start = {cmd = [
"flask",
"run",
# Important comment here about always using port 54321
"-p", "54321"
]}
except cmd outside , There are two more parameters
One is shell Parameters , From the output, you should see and cmd The difference between , and subprocess.Popen() with shell=True Almost one thing
One is env_file Parameters , You can specify the file for configuring environment variables
[tool.pdm.scripts]
start.cmd = "flask run -p 54321"
start.env_file = ".env"
If you want to limit the file of this environment variable to more than one command , It is pdm run overall situation , You can configure it like this
[tool.pdm.scripts]
_.env_file = ".env"
Add --list perhaps -l You can view the shortcut alias of all settings
For every shortcut command , All can be set pre and post command ：
pre command ： Execute... Before each shortcut command is executed
post command ： After each shortcut command is executed
[tool.pdm.scripts]
pre_compress = "{
{ Run BEFORE the `compress` script }}"
compress = "tar czvf compressed.tar.gz data/"
post_compress = "{
{ Run AFTER the `compress` script }}"
pdm Although there are many orders , But it's not complicated , There is little need to use automatic completion , If you really need to complete , It can also be realized .
For different shell, The configuration of automatic completion is different , This is described in detail on the official website .
If you use it like me zsh, You can refer to my configuration method .
One step in the middle of the screenshot is vim ~/.zshrc , Yes, it will pdm Plug in configuration to zsh in
plugins=(git z macos extract zsh-syntax-highlighting zsh-autosuggestions pdm)
pdm Good enough to use , And open enough , If you are currently using another package manager , such as pipenv ,poetry, Or still using the most primitive requirements.txt , You can also easily migrate to pdm In the to ：
Use pdm import -f {file} No initialization is required , Direct conversion
perform pdm init perhaps pdm install When , It will automatically identify your current dependency and convert
alike , You can also be pdm Managed projects , Export to other schemes
pyproject.toml and pdm.lock yes pdm Two core documents of .
pdm As a rising star , And I didn't forget my roots , It supports ：
take pyproject.toml Turn into setup.py
pdm export -f setuppy -o setup.py
take pdm.lock Turn into requirements.txt
pdm export -o requirements.txt
It took a lot of effort , Finally put PDM After introducing the basic usage of , I believe someone will question ： This is what you call An epoch-making package manager ？
actually , The above is just an introduction , and PDM Some of the core knowledge of , Considering the limited space , I will arrange the contents of these advanced classes in subsequent articles , It will include but not be limited to ：
PDM Analysis of the principle of ：PEP 582 The proposal
Build the release package ：PEP 517 The proposal
Hook Definition and use of script
Plug in management system and custom plug-ins
Introduction of cache management system
These are PDM At the heart of , Only by understanding these , You can really make good use of PDM, Then you will feel ： Why? Guido It's not a standard tool to pack ？
About PDM Follow up advanced articles , Please pay attention to my official account 《 Mingo who writes some code 》
I'll keep updating about... There PDM Advanced use of
About packages and virtual environment management tools , I have written special articles on many tools
Interested in learning systematically , You can visit this link ：https://python.iswbm.com/
Finally, make a little investigation
What package management tools do you usually use ？
1、requirements.txt
2、virtualenv
3、pipenv
4、poetry
5、venv
6、 other
With PDM, Do you still use these tools ？
1、 Meeting
2、 Can't
3、 Learn more about
Welcome to share in the comments section
