Data Scientists and AI engineers. Python has really good libraries for machine learning and other numerically intensive applications. Try launch deep learning models on your GPU with only shell scripting.
Do you do any scripts beyond extremely simple ones?
I do a fuck ton of scripting in both bash and python. I never want to do string manipulation in bash. As soon as string manipulation is required, I automatically choose python for a script. Also, if I need named arguments or multiple levels of arguments for subcommands. You can use sys.argv for basic args, which isn't any harder than bash arguments, but for complex inputs, argparse is a godsend. It has a bit of a learning curve, but it can handle anything. Bash requires you to write complex arg handling manually with loops and reducing.
You should check out Click. Way more user friendly than argparse imo. I agree with all of your points though, and I’d also add if you are working on a team that it will be infinitely easier for a co-worker to decipher your python code compared to a bash script. And you can write unit tests with py test, the list goes on and on. If the environment you are deploying to has the python interpreter, you should use python over bash.
I'll take a look at Click. Something better than argparse definitely piques my interest, because while the features are great, it feels a little cumbersome at times.
Tbh you’d still be better off writing them in python. They’ll be more maintainable, and you’ll learn valuable skills.
Also, since you commented that python was the slowest language ever, shell scripts are often significantly slower. This is due to the fact that shell commands are actually calling other programs, which is very very slow.
To be fair, sometimes that runtime difference matters. That’s why it’s C++ and python is a fairly common skill-combo amongst devs. But the fact that this dude is basically bragging about writing shell scripts as if that’s something an experienced dev couldn’t figure out tells me that they don’t really know anything about when you would choose either.
If they had mentioned the Global Interpreter Lock or dynamic typing maybe they would have had some sort of real case for why you should avoid python in certain situations.
Totally agreed the run time difference matters sometimes. In my experience it's hasn't been the case for scripts, as they are generally small in scope in comparison to C++ applications.
Maybe you should check how python compares relative to shell scripts before you comment. You’re making it very apparent that you don’t actually know what you’re talking about. Regardless of how slow python is, it is significantly faster than bash, or any other shell language purely by virtue of the fact that shell languages are primarily glue between other programs. Spawning a new process has a ton of overhead, which you would know if you were capable of doing anything other than projecting.
You’re also woefully unaware that it is completely possible to write python bindings for C++ code, which many popular libraries do. In practice python is not as slow as you think it is. That’s not even considering the fact that python 3.12 increased performance of the language.
It’s not perfect for everything, but this performance argument shows that you don’t know enough to understand why that isn’t really a drawback for writing scripts, which is undeniably an area that python excels at.