Downside: "^1.2.3" as default versioning for libraries. You just pinned a version? Oh great, now I can’t upgrade another library because you had to pin something in yours…
That non-standard syntax has been a PITA for the last few years. That being said: They created that syntax for regular applications (and not for libs) in a time when the pyproject.toml syntax was not anywhere near finalization.
I bet it’s darn amazing,
It is. In this older article (by Anna-Lena Popkes) uv is still not in the middle, but I would claim it’s the new King of Project Management, when it comes to Python.
uv init --name <some name> --package --app and you’re off to the races.
Are you cloning a repo that’s uv-enabled? Just uv sync and you’re done!
Heck, you can now add dependencies to a script and just uv run --script script.py (IIRC) and you don’t need to install anything - uv will take care of it all, including a needed Python version.
Only downside is that it’s not 1.0 yet, so the API can change at any update. That is the last hurdle for me.
pyproject.toml track the dependencies and dev-dependencies you actually care aboutuv.lock file that contains each and every lib that’s needed.uv sync and uv run <application> is pretty much all you need to get goingpip3 freeze > requirements.txt
I hate this. Because now I have a list of your dependencies, but also the dependencies of the dependencies, and I now have regular dependencies and dev-dependencies mixed up. If I’m new to Python I would have NO idea which libraries would be the important ones because it’s a jumbled mess.
I’ve come to love uv (coming from poetry, coming from pip with a requirements/base.txt and requirements/dev.txt - gotta keep regular dependencies and dev-dependencies separate).
uv sync
uv run <application>
That’s it. I don’t even need to install a compatible Python version, as uv takes care of that for me. It’ll automatically create a local .venv/, and it’s blazingly fast.
Python’s tooling is a mess.
Not only that. It’s a historic mess. Over the years, growing a better and better toolset left a lot of projects in a very messy state. So many answers on Stack Overflow that mention easy_install - I still don’t know what it is, but I guess it was some kind of proto uv.
Check if there’s a stub or types extension - https://pypi.org/search/?q=types
I can’t stand the bloat of virtual environments
Sucks to be you, but virtual envs are THE way to keep everything running on your own machine. You can try to keep everything in a single venv, but then you’ll start wondering why it doesn’t work on other’s machines.
built-in tools provided by the language
sqlite3 module comes with Python. As does the json module. Use pathlib’s Path object over os.whatever because the API is nicer to work with. abc (short for Abstract Base Class - abc, get it?) is useful for inheritance stuff. I like click, but there’s argparse if you need to write a CLI. Stay away from asyncio if you can - it’s usually not faster and a bigger PITA than it needs to be. Need to transport a binary, but you can only send text? base64. import datetime as dt if your new friend when it comes to datetime objects - to be aware to keep things timezone-native. You’ll need pytz for that - Python doesn’t update fast enough to keep up to date with all the new timezone shenanigants. always check for missing timezones; they’ll cap your kneecaps and kick you in the balls if you’re not careful. http can be used to make calls to APIs though requests is a favorite; slap urllib3 and you got yourself a sauce going on! Skip logging and jump to a lib that supports structured logging, python-json-logger is a nice beginner’s log lib, jump to structlog if you need the raw power (say you want control over other people’s logging output). math is a sleeper hit - powers, roots, floors, ceitings, and statistics is a nice addition on top of that together with random for probabilistic stuff. multiprocessing if you need to run something concurrently: ProcessPoolExecutor for IO, ThreadPoolExecutor for CPU bound stuff (IIRC). re for regex related stuff. uuid if you need a unique identifier (this can be slow in DBs, so take care!)
ecosystem
PyPI (Python Packaging Index) if your new friend - pai-pee-ai; not paipai - that’s the JIT implementation of Python: pypy.
Use pytest to run your tests, ruff for formatting (though I imagine it intimidating for beginners, better to bite the bullet (even if you nibble on it slowly), uv to track your dependencies and Python version.
Yes, figuring out a nice configuration for these tools is a pain, but that’s what ChatGPT is for. Just don’t go too wild on the settings, as there are MANY, and if something doesn’t work because ChatGPT got a little outdated, it’ll hurt.
Anyway, that’s about 4 years of experience concentrated in a single page. If you ever feel like an imposter: that’s a common feeling: https://en.wikipedia.org/wiki/Impostor_syndrome If it’s hitting you on the head, see if you can talk with your seniors about it (presuming they’re sane adults) - otherwise there’s always Lemmy to vent :P .
edit: if you need to handle tabular data: Polars. If anyone recommends Pandas over Polars, hit them with a newspaper until they crawl back under the rock they came from. Unless it’s for GeoPandas, then it’s OK.
The new king on the block: uv. It can do everything poetry does, while also using a standard pyproject.toml (no more weird ^), and it’ll handle the Python version for you, so no faffing about with manually installing anything. Just uv sync and off you go!
Downside: not compatible with virtualenvwrapper, as it’ll force its .venv in the local folder.
It’s also still under heavy development and breaking changes are still expected, but it’s already super nice to use.
Same guys (Astral) also made ruff the formatter/linter that they intend to eventually integrate into uv, IIRC.
I’m running all my personal projects under uv and am having a blast. It’s so fast.
Don’t worry, they’re planning JS0 and JSSugar (IIRC) so you’re forced to compile anyway :D
LineageOS, maybe? Still Android, but (AFAIK) more open to change than standard Android.
So sad when it happens…
I don’t follow - do people still seriously use SMS? I for one try to use it as little as possible.
It’s still a draft, but that’s just a weirdly unnecessary change, IMO. There is no need from anyone to have the versions shift by 11 versions.
Here’s a better overiew: https://devguide.python.org/versions/
Shouldn’t it? Yes, just like the ability to unit test, but that doesn’t stop schools from skipping over them either.
Learn how to test (which is also what the article recommends).
Write unit tests. Write property-based tests. I don’t care which ones, but automate the SHIT out of your tests.
Also curse at your school for not teaching you this from day one.
Alas, I don’t know any articles or books on how to learn, because school didn’t teach me either. I learned the theory from Uncle Bob, among some others, but had to learn the actual work from work…
Protip: pip install pyupgrade
And then find . -name '*.py' -not -path '*.tox*' -print0 | xargs -0 pyupgrade --py310-plus in your repo to update what can be updated.
BTW, pyupgrade’s creator, asottile (that’s his name) also has an informative channel: Anthony Writes Code where he explains Python features, or goes into interesting bugs he ran into, etc. The good stuff.
So I remember the plan to improve Python’s speed from 3.9 to 3.13… Has there been an updated plan since? I presume the JIT will likely be faster in 3.14 (it’;s already at parity - pretty impressive for a first release), but is there anything else planned?
VSCodium doesn’t have the Python plugin, does it? It also misses the config sync when you’re logged in (IIRC). Not the worst to miss if you start out, but I’ll take it over having to track EVERYTHING in my code-workspace file.
Nice! I didn’t go because I couldn’t find anything I really wanted to see, but my co-workers enjoyed themselves, which was nice :D
500MB for Ray, another 500MB for Polars (though that was a bug IIRC), a few more megs for whatever binaries to read out those weird weather files (NetCDF and Grib2).