Breaking up with Python

> It's really a shame that data science, ML, and notebooks are so wrapped up in it. Otherwise we could jettison the whole thing into space

Although I personally feel Python has its place, I contribute to a project that hopes to diversify the ML/scientific computing space with a TypeScript tensor lib called Shumai: https://github.com/facebookresearch/shumai

People joke that Python programming is 'pip install solution', but it's honestly not too far from the truth. Until Go, Rust, or whatever fancy new language can match the wealth of well maintained libraries that Python has, I don't see it ever getting replaced for the quick scripting that the author talks about.

I don't disagree with any of his points, actually, and agree with almost all of them. The standard library has some holes - though I think that's true everywhere. I don't know that eg Ruby or java don't have holes either.

I do think the packaging Story in python is nuts. Why does gem just work for Ruby but python's story is such a disaster? Though imo my modern advice is "just use poetry and ignore everything else".

Pythons main draw for me is some combination of:

Type hinting is good enough. It has a big escape hatch to be able to do C stuff if you need to. Numpy. The language is still straightforward enough that beginners can approach it and the experienced can wax reasonably eloquent with it. And it's pervasive.

So when I sit here and I'm like "ugh I need a thing to sit here and hit an api and dump it to a file every day in a cron job, and I don't want to spend a whole day writing this". I go for python. And while go and java and rust etc would be just fine, I still end up taking longer than I'd like.

Somehow python is just the shortest, sanest route to just getting a small to medium sized doodad out the door and running reliably in a small footprint of time without needing to remember boilerplate (java) or think way lower level than I'd like to (go/rust). Python just lets me do it with low cognitive overhead. Sometimes while I'm half-listening to a meeting I'm in!

Honestly my biggest gripe with python is it's really hard to tell early enough when your program is on a trajectory to be too big for python. And I don't mean from a perf perspective or anything, I just mean when it goes from being "easy python" to "oops it's spaghetti now." There's a tipping point and I think even seasoned people end up well past it by the time they pause and rub their eyes and think "ugh I wish this were in java."

Last time I argued, like the TFA does, that Python's typing system is both too complex for casual users and useless for power users ("it doesn't do anything" is about right), and that this makes it hard to evangelize its use to other devs who aren't sold on type systems ("but this doesn't do anything!"), I got into a 30-level nesting flame war with some HN regular who just wouldn't let go.

So that's still my line of thought, but I'm already regretting typing this on HN.

Node is an interpreter just like Python, gives access to OS features like sockets, file systems and processes, so yes, you can.

However I’d rather not if you value your sanity and want to use typescript, because then you introduce packages and compilers and the whole npm ecosystem which makes building C++ look easy and sane.