Recommended way to manage several installed versions of Python (macOS)

Posted by Blakk_exe@reddit | Python | View on Reddit | 74 comments

When I use VS Code and select a version of Python on macOS, I have the following versions:

Python 3.12.8 ('3.12.8') \~/.pyenv/versions/3.12.8/bin/python
Python 3.13.2 /opt/homebrew/bin/python
Python 3.12.8 /usr/local/bin/python3
Python 3.9.6 /Library/Developer/CommandLineTools/usr/bin/python3
Python 3.9.6 /usr/bin/python3

I believe having this many versions of Python in different locations messes me up when trying to install packages (i.e. using brew vs pip3 vs pyenv), so I'm wondering what the best way is to clean this up and make package + version management easier?

[-]

usrname--@reddit

Use uv and create venv for every project.

[-]

matjam@reddit

This is the way

[-]

slowwolfcat@reddit

is it "better" than conda ?

[-]

Zer0designs@reddit

A 10000x better.

[-]

foobar93@reddit

One of us!

[-]

phant0md@reddit

There are literally dozens! DOZENS!

[-]

matjam@reddit

Gooble gobble, a loving cup, a loving cup!

[-]

unapologeticjerk@reddit

If uv could implement a uv shell equivalent to pipenv shell, I would have been in the KickstartFundMe Alpha of it and out preaching the word. Great tool, but man, I am lazy and used to pipenv.

[-]

chisoxaddict@reddit

Could you elaborate on what you mean? There is uv run python (and using --with package or project to run a python shell. Is that not what you're talking about?

[-]

twenty-fourth-time-b@reddit

Do you install pyright/mypy/flake8 for every product too?

[-]

flying-sheep@reddit

I prefer Hatch, which creates a venv for every tested configuration for every project:

hatch test -a

will create an environment for each Python version that matches your project.requires-python constraint using uv and then run pytest tests in it. (all configurable of course)

[-]

Chasian@reddit

This is totally out of the scope of what the person asked for. This is neat, but only applies to people trying to make python packages that work on all OS, all versions, etc

[-]

flying-sheep@reddit

Which is most people who make packages.

[-]

Chasian@reddit

Nobody talked about making a package. Nobody talked about testing.

All they did was ask for managing multiple python versions on their computer

[-]

emmzzss@reddit

This is the only correct answer

[-]

Still-Bookkeeper4456@reddit

Feels like those posts are just made so people can write 'uv'

[-]

No_Flounder_1155@reddit

its kind of weird. Never seen this for any other tool. Its not really that special. Starting to think they're either bots or schoolkids.

[-]

bulletmark@reddit

I've been using Python daily for 22 years. uv is the greatest thing since sliced bread.

[-]

No_Flounder_1155@reddit

and you still haven't learned one exception doesn't make the rule...

[-]

MaxDPS@reddit

Sometimes there is just a really obvious answer.

I haven’t switched over to UV, but if someone asked me this a few years ago, my answer would have been pyenv. And I’m sure a lot of other people would have had the same answer.

[-]

No_Flounder_1155@reddit

Its not about it being a good tool, its kind of extremem fan boy level. I wonder if its because people are also super junior? Someone else in the thread is writing an app to automate tailing docker logs, because they hate typing out the command...

[-]

Berkyjay@reddit

I know everyone always says pyenv or UV. But I still prefer installing my own versions and just creating aliases for each. Then I create alias a string of commands to create a venv for each version.

[-]

bulletmark@reddit

Why? uv venv -p 3.12 creates the venv for you and installs 3.12 on the fly (if not already installed in it's cache). All happens 1000 times faster than using pyenv.

[-]

Berkyjay@reddit

Dunno what to tell you. I've tried integrating Poetry +Pyenv and UV into my workflow. I like controlling my own installs I guess. If I want to start coding I run my vc312 alias and I'm ready to rock.

I also never got the speed thing. Unless you're dealing with deployment at scale, I don't see what a fraction of a second does for the individual user. It's nice UV is more efficient though, but it's not making me run out and installing it.

[-]

bulletmark@reddit

I said "1000 times" metaphorically of course. Just timed it, with no caches so in both cases 3.9 has to be fetched, and then `.venv` is built:

`pyenv install 3.9 && \~/.pyenv/versions/3.9.21/bin/python -m venv .venv` takes 1 min + 31 secs.

`uv venv -p 3.9` takes 3.3 secs.

[-]

coffeecoffeecoffeee@reddit

I just use pyenv.

[-]

bulletmark@reddit

We all used to use pyenv. But then uv came along and made pyenv redundant.

[-]

coffeecoffeecoffeee@reddit

something something xkcd standards

[-]

ritonlajoie@reddit

Nobody gave this another answer: use devcontainers

[-]

VindicoAtrum@reddit

Use uv, as everyone else has said. Ditch pyenv for uv.

[-]

lozinge@reddit

`rm -rf $(pyenv root)` everywhere

[-]

lozinge@reddit

Cringe pyenv users on suicide watch downvoting me

[-]

djavaman@reddit

Astral must pay a lot of money for reddit grassroots marketing. Its absurd.

[-]

papparmane@reddit

First make virtual environment. That's why they were created.

Second, You have so many versions, yet none of them are in the standard /Library/Framework/Python.framework/ location. Delete all of them and download from Python.org the installers you will get them cleanly installed in Franework/version.

Fro. There use python3.13 -m venv venv-3.12 or whatever to create a virtual environment.

[-]

DrShts@reddit

I use pyenv.

[-]

codingjerk@reddit

I've used it too, but now I recommend switching to uv. Way better DX and it manages venvs for you too

[-]

crying_lemon@reddit

always homebrew. then :
or the old way: pyenv install 3.xx.xx then in the folder pyenv local 3.xx.xx then python -m venv .venv And finally source .venv/bin/activate

or if you want you can try UV

[-]

zurtex@reddit

Btw, this is literally this xkcd from 2018: https://xkcd.com/1987/

Except now we have pyenv, uv, and multiple ways to manage conda to mix things in.

[-]

crippledgiants@reddit

We use Poetry for this and it's a breeze. Honestly surprised it hasn't been mentioned yet.

[-]

_MicroWave_@reddit

Because uv is better. I say that having migrated.

[-]

Prestigious_Run_4049@reddit

Because uv replaces poetry, but uv handles the python version as well.

So if a project uses python 3.x, it will download it into the venv, poetry doesn't do that

[-]

crippledgiants@reddit

So the reason to switch to uv is because it saves me from running a one time install command?

[-]

Prestigious_Run_4049@reddit

Yes! I never want to think about python versions again. Uv just let's you get rid of so much python tooling

pipx/pyenv/poetry/virtualenv -> uv

[-]

lozinge@reddit

But what if you want to try another version of python etc \~ https://docs.astral.sh/uv/guides/install-python/#installing-a-specific-version

uv is very impressive - I wasn't really interested but has totally changed how I use python and can't recommend it enough

[-]

w0m@reddit

One time per python version, uv is a superset of poetry. Dependency resolution is also an order of magnitude faster

[-]

Key-Half1655@reddit

I was using pyenv but just switched to uv today

[-]

EthanBradb3rry@reddit

Pyenv is the only answer here

[-]

Beregolas@reddit

As all the others have said: You are supposed to use one virtual environment per project, so you never install packages into system wide python installations. You can get started with that here: https://docs.python.org/3/library/venv.html

In addition, on macOS I strongly suggest to let homebrew as a package manager install and manage your python versions: https://brew.sh/ https://docs.brew.sh/Homebrew-and-Python

I used that setup for years professionally and it never failed me. It is also simple to understand and a similar workflow to other Linux/Unix based systems.

[-]

gnomonclature@reddit

This is how I do it as well. Since I write things I want to test under multiple versions of Python, I have Python 3.9, 3.10, 3.11, 3.12, and 3.13 on my box, and I don’t run into any problems.

[-]

Bitopium@reddit

You can do that with UV + nox just fine. Works without needing to have native python versions installed

` import nox

PYTHON_VERSIONS = ["3.11", "3.12", "3.13"]

nox.options.default_venv_backend = "uv"

@nox.session(python=PYTHON_VERSIONS) def tests(session): session.run_install( "uv", "sync", env={"UV_PROJECT_ENVIRONMENT": session.virtualenv.location}, ) session.run( "pytest", "--cov-report=term-missing", "--cov-fail-under=0", *session.posargs, ) `

[-]

i_dont_wanna_sign_in@reddit

Let brew install all the versions and then point whatever environment manager you're using at them. Then everything lives in /opt/homebrew and you aren't screwing around with multiple version managers.

Better yet, install a docker controller (docker desktop is still free for personal use) and develop on containers and manage those via DockerFile and never worry about dependencies on your host system

[-]

rainnz@reddit

Docker

Switch to Podman - https://podman.io

[-]

rainnz@reddit

uv - https://youtu.be/qh98qOND6MI

[-]

zacky2004@reddit

Spack

[-]

eddaz7@reddit

uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv uv

[-]

njnrj@reddit

You can install and other versions using brew itself. """ brew install python@3.12 brew install python@3.13 etc., """

The binaries will be "" python3.12 python3.13 ""

And for every project you could create an venv using these interpreters. Uv also can use these binaries.

[-]

riklaunim@reddit

Outside mentioned uv you can also look into using Docker. You should be able to also drop minor versions and have one 3.9 or 3.12 instead of two.

[-]

GrandBIRDLizard@reddit

I recommend learning how to set up your own virtual environment using virtualenv, pip, and pyenv. https://docs.python.org/3/library/venv.html

[-]

Mevrael@reddit

Create projects in the VS Code with the PM extension, and create uv/arkalos project for each project you are working on, let say inside your home \~dev/python folder.

Here are the recommended VS Extensions and the guide.

https://arkalos.com/docs/installation/

[-]

wineblood@reddit

Use a venv and try to make your projects work on as few versions of python as possible. Others have mentioned tools to manage multiple python versions but I believe that's just masking the problem, you'd be better off with everything running on 3.12, having a single python 3.12 version installed and using that to create a venv per project.

[-]

zaviex@reddit

Conda or uv

[-]

cgoldberg@reddit

I use pyenv, but I don't know how Mac support is.

Without additional tooling, just make sure you create your virtual env and install packages with the interpreter you want.

python3.9 -m venv venv
source venv/bin/activate
pip install package

That would create the venv with a specific interpreter, and then it will use the correct pip to install packages.

[-]

ogMasterPloKoon@reddit

UV or Miniconda ...

[-]

ComfortableFig9642@reddit

uv is the best solution if you'll never really need to manage anything other than Python

Mise (https://mise.jdx.dev/) is awesome as soon as you need to manage more than just Python, and can manage Python as well

[-]