Luke Tierney

@antoine_fabri Not in R-devel

In #rstats an environment can be its own parent, or we can have longer cycles. I wonder if there's anything cool we can do with this. e1 <- new.env() e2 <- new.env() e3 <- new.env() parent.env(e1) <- e2 parent.env(e2) <- e3 parent.env(e3) <- e1

@antoine_fabri I wouldn’t count on this “working” in the future.

#rstats TIL

@tslumley Went through the same considerations eons ago with LispStat. Ended up with nonlinear models inheriting from linear in order to pick up things like diagnostics.

@eddelbuettel @mpiktas Actually, string hashing was added by @sfalcon , many moons ago as you say.

@mpiktas Character variables are internally hashed so you save on use of repeated values; that was an important change many (many!!) moons ago. We can all bow in the general direction of Iowa and thank @LukeTierney4 for this (and so many other internal improvements). #rstats

I have a 12mln row data.frame in R with 6 columns. 5 columns lots of repeating strings. Exported CSV weighs ~3Gb, but object.size shows that it uses only 0.6Gb, or 8 bytes exactly for each cell. Does R use some in-memory columnar storage? #rstats

@MyKo101AB callCC provides a higher level interface to this

You can force a parent function to exit early using `do.call()` by forcing a call to `return()` early #rstats #DataScience

@daniela_witten @aristeinberg @COPSSNews @AmstatNews Congratulations!

Now that the dust has settled… it is absolutely humbling to win this award. I am pinching myself! So grateful to have received it at #JSM2022 in person, with @aristeinberg and our three children cheering me on. @COPSSNews @AmstatNews

@ijlyttle So THAT’s the reason they didn’t have room for an ‘r’ in the new name. Makes sense now. Thanks!

in somewhat related news, thinking of trying "posit" as my new wordle starter

following a six-day hiatus, had a lucky second guess Wordle 406 3/6* ⬛⬛⬛🟨⬛ 🟩⬛⬛⬛🟨 🟩🟩🟩🟩🟩

@tslumley @hadleywickham That was the first one that went through my mind as well. Can't imagine why. Would have come with some good emoji too🍺🍻

@hadleywickham @asmae_toumi A garter snake might have been good enough if the new name had included an ‘r’ as well as a ‘p’ …

@asmae_toumi I realise now we should’ve done a bit where I wrestle an actual python

@jimhester_ Seems like a missed opportunity to provide a good quality mask with the conference badge printed on it

Underestimated how hard it would be to recognize people in masks at #rstudioconf2022!

@stephaniehicks @daniela_witten @jhubiostat @JHUDataScience @JohnsHopkinsSPH Congratulations!

I'm excited to share that I've been promoted to Associate Professor of Biostatistics @jhubiostat @JHUDataScience @JohnsHopkinsSPH 🎉 Thank you to all of my mentors, lab members, colleagues, collabs, friends & most importantly family who are my greatest supporters #TakesAVillage

@henrikbengtsson @jimhester_ @tslumley @ekuber @Thoughtfulnz Nobody is suggesting removing that. But neither python nor julia tell you how to get out in their startup messages, and that's OK: for both an end of file (^D) does the job, which the generic terminal REPL quit. And that works in R too.

@LukeTierney4 @jimhester_ @tslumley @ekuber @Thoughtfulnz I think R's instruction on how to quit R is appreciated by first timers and folks less familiar with terminal REPLs

@jimhester_ @tslumley @ekuber @Thoughtfulnz In any case, the startup message would most likely have scrolled off the screen at the point where 'type this' information would be useful. An IDE with a help menu that is always visible and provides this information would be more useful.

@jimhester_ @tslumley @ekuber @Thoughtfulnz One could argue that having a larger startup message that python, as R and Julia both do, might make it slightly more likely that a user would notice and look, but I doubt it.

@jimhester_ @tslumley @ekuber @Thoughtfulnz Unfortunately we live in a world where most system startup messages are not very useful so we've been pretty much conditioned to ignore the lot.

@tslumley @ekuber @Thoughtfulnz I think the problem is more that the rest of the default startup message is not very useful for most interactive sessions, and the help part is buried in the middle of a sentence so users just ignore all of it because it doesn’t seem relevant

@sid_mnnit @kategchalmers Using vectorized operations, if that is possible, will give the best performance. But rewriting a loop in terms of vectorized operations is not always easy.

@sid_mnnit @kategchalmers In principle, apply function calls could more easily be optimized to perform better, but this currently doesn't happen. Once you are used to them, using apply functions can produce cleaner code and simplify debugging.

I prefer loops in #RStats because I found them easier as a beginner. I sometimes feel like I’m a minority/missing out on something by not using apply tho. What do other people use? Is it worth making the switch? #R #DataScience #coding