Hacker News Reader

This was explicitly addressed in the article: as soon as you have to do anything which isn't a trivial numpy operation, performance goes off a cliff, and that can be a problem.

This also isn't necessarily true. Take the example of TensorFlow. You build a representation of the computation you want to run, and then you can run nearly the whole thing end-to-end in native C++ using Eigen data structures, with occasional shuttling of data back into PyObjects (rare) or numpy (common) for metrics tracking.

Cython is a much more powerful tool than I think the author of this article realizes.

As with Python, the fast libraries written for R are usually implemented with something else under the hood. Take the data.table library, for example:

https://github.com/Rdatatable/data.table

It's wicked fast for many kinds of tasks, but its R API is just a thin layer on top of C.

or a cuda wrapper