Because that’s not at all the case and your rhetoric is what perpetuates the false image.
Somehow we ended up in a state where step one of doing anything new with python is to fire up an empty docker container. And I'm awfully tired of folks in the "python community" blaming the victims of the mess they made.
This approach seems insane but correct.
It will be less hard than using docker.
My recent example: last week, I was installing a new program that uses some python scripts during its operation. I could not get those scripts to run. I knew it was because of the usual python problem of the scripts not matching the version of python that was executing them, but figuring out how to fix that took me a full day.
Just to make Python work. Not even as a dev.
Wearing my user hat, this is a clusterfuck. There is no other language that I am exposed to that presents this sort of problem, and this is a very common issue with Python.
This is why I start to get sweaty any time that I'm using software that involves Python. It turns into a crapshoot and half the time, it's going to cost me a lot of time and stress.
> perpetuates the false image.
You can believe it's a false image if you like, but there are a whole lot of end user experiences that indicate it's very real.
It will help with exactly that.
Python is the only language I've encountered that causes this sort of trouble for ordinary end-users. As a dev, I find this frustrating because the python scripts in question are clearly only "glue" in the first place. That's a lot of burden to put on the end user for just glue.
As for my issue, I did make it work in the end. I'm 90% sure I didn't do it in the right way and it will cause me problems at some point in the future, but that will be a problem for future me.
Ideally it should be fixed at the installer level.
I will write about this at the end of the week.
- easy_install, distutils and eggs were deprecated
- ensurepip and venv were introduced
- the whole wheel ecosystem flourished
- a new dependency resolver was introduced to python
But:
- a lot of noise has accumulated. Hence the procedure in the previous article, to avoid the traps from all that noise and go back to what usually works.
- we have a long way to go, but it's not something easily fixed (hence the conclusion of the current article explaining why it's going to take a long time to get better).
Hopefully, the direction of that long way is relegate python (and javascript!) to legacy languages.
Something, hopefully, will come out of webassembly. Otherwise, the last decade only saw the rise of two fundamentally flawed languages as the most popular ones.
Ruby losing to both was not a good thing. And I'm not even a Ruby programmer! I hated the language back in the Rails vs Java days. But it is impossible to deny that great software came out of Ruby.
Javascript and Python has produced no great software, just ponderous stacks of crappy code and poorly organized libraries. They are popular because they are used for throwaway web applications and throwaway scientific / non-programmer code, and it shows.
The fact that golang is what is producing new "good" software: a language that is intentionally bad, as opposed to python and javascript being worse because of cruft accumulation or Eich writing it a weekend, is also not a great sign but it's not the step backward that javascript and python are.
People writing in JS are usually techies who could/would migrate. People writing in Python are on a wide spectrum that goes from "one-step-above-matlab" to "know-it-all-c-wizard". You won't actually get all these people to move on any reasonable timeframe.
HN is biased because we tend to love tech and a lot of us know several languages, but Python is much more democratized. My friends in physics, biotech, mechanical engineering, basically everyone in STEM is using Python and to them it's a tool that took a lot of time to learn and they won't drop it all because dependencies are hard.
Besides, code at a large scale in any language is always messy. There are many ways to manage a growing codebase but if your only approach is relying a "good" language with static typing, verbose/explicit syntax, and a better built-in package manager, you are doomed to fail anyways.
Perhaps this is a case of correlation=/=causation? Python is more accessible so less adept programmers are more likely to use it.