>Why trying to understand implementation in the language that will never be used in the production?
Because the end goal is understanding the algorithm, NOT writing production code.
Those reading the article will never even write a FFT function anyway, since you can use ready made highly optimized versions. So knowing what the production code looks like is totally moot to them.
It's like you're arguing why ever use pseudo-code. Because it's readable, makes the process easy to follow, and doesn't contain all the boilerplate junk you'd need to get performance in a production implementation. Python serves this role very well, with the added benefit that it's not even pseudo.
Because the end goal is understanding the algorithm, NOT writing production code.
Those reading the article will never even write a FFT function anyway, since you can use ready made highly optimized versions. So knowing what the production code looks like is totally moot to them.
It's like you're arguing why ever use pseudo-code. Because it's readable, makes the process easy to follow, and doesn't contain all the boilerplate junk you'd need to get performance in a production implementation. Python serves this role very well, with the added benefit that it's not even pseudo.
Plus, ever heard of prototyping?