Mastodawn

Dr. Russell Pierce Nov 6, 2022

@grant thoughts are like... everybody has 'em.

Julia doesn't seem to have captured the popular imagination yet. I suspect this is in large part because speed of computation is often relatively low on people's needs list.

I expect we'll see a {reticulate} for Julia and from the R side that'll serve most people's needs.

I think the next time I'm tempted to write C++ for speed in R or Python, I might give Julia another look.

Show thread

Sharon Machlis Nov 6, 2022

@russellpierce @grant I haven’t tried it (I don't use Julia), but the {JuliaCall} #RStats 📦 lets you call #JuliaLang functions from within R @rstats https://hwborchers.github.io/

JuliaCall: Integrating R and Julia

[“A short description of how to use Julia functions from R.”]

[“Notes on JuliaCall”]

Show thread

Jesus Castagnetto 🇵🇪Nov 7, 2022

@smach @russellpierce @grant @rstats

I have used more #Rstats than #JuliaLang in my work, but I will like to change that in the future.

Not only can you call #JuliaLang code from #Rstats, you can also call #Rstats in your #JuliaLang code using the "Rcall" package.

Sometime last year I was playing with that, using a modification of a #JuliaLang benchmark from "Julia Academy". See: https://gist.github.com/jmcastagnetto/fce3cad5856517250ad1cdc468d49865

Notebook benchmarking C, Python, Julia and R

Notebook benchmarking C, Python, Julia and R. GitHub Gist: instantly share code, notes, and snippets.

Gist

Show thread

Dr. Russell Pierce Nov 7, 2022

@jmcastagnetto @smach @grant @rstats for the benchmark in calling R via RCall aren't you eating the overhead of the FFI? E.g. a copy from Julia to R and back again?

grant Nov 7, 2022

Show thread

Sharon Machlis

@russellpierce @jmcastagnetto @grant @rstats Yes, the {JuliaCall} 📦 author briefly discusses overhead on the package website, but I think you'd really need to benchmark a specific task both ways to see if there's any worthwhile performance gain. Might vary considerably based on the operation, size, data structure, and I/O. Might be interesting to R users pondering dabbling in Julia regardless, though.
https://hwborchers.github.io/#Appendix