Hacker News Reader

Comment by davidwihl | Hacker News Reader

For a statistician, R has a variety of packages that are unavailable in Python. RStudio is a very easy to use cross platform dev environment. With a different R kernel, you can take advantage of multiple cores.

View on HN

Which packages would you say are the most glaring omissions? I thought that R would become even more popular now it is on Azure ML - (if you haven't checked MSFT Azure ML out you really should - it's a blast).

I use both (cut-paste monkey!). There seems to be a bit more magic with R, less code. Shiny is pretty cool too.

btw - shinyproxy package was compromised recently - check your download hashes :(

The most glaring package omissions are related to Machine Learning, e.g. 1) an R equivalent of sci-kit learn 2) A native R equivalent than the buggy interface to TensorFlow 3) A higher level ML equivalent to Keras or PyTorch.

And the huge variety of presentation tools such as ggplot2, Rmarkdown, and Shiny still makes R a very attractive choice.

Tight integration with Rcpp also makes it very easy to speed up calculations when performance becomes an issue.

For ggplot2 there is plotnine https://github.com/has2k1/plotnine (disscussion https://news.ycombinator.com/item?id=14430447)

Also Python is not a "statistician" language so the adoption there is very slow.

For me Rmarkdown is such a win for R. Web deployment, especially the web servers to host shiny apps is a weak point for R (Linux specific, and really not many options even in the Linux arena). Then again, perhaps notebooks are a better way to handle online data exploration.

Reply on news.ycombinator.com