iyarlin/r-advantages-over-python

This repository enumerates all the reasons why R is better than python for DS

r-advantages-over-python's Issues

Numpy can work with different types

I am referring to this part of your comparison:

R-advantages-over-python/README.md

Lines 1316 to 1318 in 0633661

	Do note however that numpy can’t help you if the vector you’re working
	with is made up of different types of elements (e.g. strings and
	numbers). In R you can use lists and still enjoy vectorization:

It seems that you can easily do that:

values = np.array(['a', 1, (1, 2, 3), 'wow', ['sweet', 2]])
indices = np.arange(len(values))
values[indices > 2]

## array(['wow', list(['sweet', 2])], dtype=object)

Moving windows example

I'm not so sure it's objectively worse in Python. It's about the same number of lines, and the only difference seems to be a subjective judgement call about the ubiquitousness of functions. I'm not sure it's a strong enough example.

H2O.ai benchmark is out of date

Great repo by the way!

The benchmark has quite different results now that Polars and DataFrames.jl have matured a bit more, even datatable for python. I've been working on automl packages for both R and Python (the R one is much further along and the Python one is relatively new) and I can't state enough how almost everything I do to create these packages is more easily done in R.

Now that Python does have some promising data wrangling packages (polars / datatable) I think it's important to focus on the capabilities in those packages compared to data.table (which is still king in my book). The example you have with the case-when of dplyr vs pandas is a superb example of two things: the capabilities that exists and the ease of which to deploy them, and in both cases, R wins. I believe that you'll find this to be true to more you go digging into feature engineering capabilities from all those packages.

Anyway, great work