HN Explorer

Ask HN: Rewriting any popular Python project in C/Rust/Haskell is better?

4 points by anyfactor ↗ HN
The internet's sentiment is that anything that is written in one of the "better language" is inherently superior than anything that is written in Python. The better languages are - Rust, C, C++, Lisp, Haskell and maybe Fortran, Zig, D, Nim.

I have seen this idea constantly popup in most discussion relating to Python and mathematical programming. Numpy being written in C justifies itself to be above criticism of Python's inefficiencies.

I sometimes am not sure if it is programming elitism or a valid argument.

If it is a valid argument, is Python should only be used as tool for proof of concept? And when there is enough traction the existing codebase should be rewritten to one of those better languages? Have you rewritten your codebase in this way and if so what was your thought process behind it?

5 comments

[ 1.5 ms ] story [ 18.6 ms ] thread
mc4ndr3
Rewriting Python in any compiled language is better.

Imagine binding typos keeping someone awake at night. It's 2022.

anyfactor
Have you done something like this before? I would love to know about the feature addition, maintainability, struggle (?) etc. behind that.
abrax3141
Rewriting python in anything is better. Esp. Julia, which is just uglified Lisp.
anyfactor
I am a data enthusiast, I have never thought about Julia as a viable option. R is the gold standard, Python has those C backed libraries.

I talked to couple of engineers about MATLAB vs Julia or X vs Julia. They said the company pays for whatever they do that includes computational time, error handling time and obviously paying for the software.

abrax3141
R is a horrible language. But is, indeed, the statistics gold standard. Mainly for the same reason as python, which is the only language worse than R. (Bcs PERL graciously died.) Both are piss slow (until you hit their C libraries.) But you can directly call python from Julia, so you get the libraries for free. Dunno about calling the R libraries, but that would be great.