STAT 447 -- aka #DataScience#Programming Methods -- is coming to a close for this term, and I am quite pleased with how we ended with a story arc: from shell via other key tools, a lot of #Rstats to ... Github Actions using shell again to tie it all together (with the Spotify example I had posted about earlier). Slides are at https://stat447.com (while other material is for Illinois students only),
Long knives at CRAN. My (old) local script watching Rcpp usage changes see 135 fewer packages at CRAN (!!): 20594 versus 20729 (and 21 fewer Rcpp uses). That is a Yuge delta to yesterday, and it will take a few more (hourly) @CRANberriesFeed messages (which I cap to one removal message per run) to catch up.... #rstats
(Subsequent closer look confirms 139 removals which is likely a new record.)
Earlier @TreeStarMan asked 'why is it so hard to get tidyverse, sf, lme4 onto Linux for #RStats' leading to another quick demo -- the link should open an 80 second mp4 video in your browser showing how thanks to #r2u the packages tidyverse, sf and lme4 along with all their dependencies install in ... thirty seconds on Ubuntu from a single command.
With big thanks to @hrbrmstr for creating and curating the ulid package for so long, it is now back on CRAN ... which also deserves a big Thank You! for having such a well-oiled machine. From upload to newbies assignment to 'thanks, on its way to CRAN' in just over an hour (!!). I tried to do my homework here, and we had prior email, repo transfer and social media post showing I didn't hijack. All good! #rstats
You know, sometimes #Emacs function/variable names are hilarious and yet extremely unintuitive. I wanted to kill the window after eshell was exitted. I tried M-x eshell with kill, terminate, exit, or end, and I came up empty.
Out of frustration I searched online only to find that, lo and behold, the function I was trying to advise is 🥁 🥁 🥁 eshell-life-is-too-much 😂 and guess what is the description in the documentation:
"Kill the current buffer (or bury it). Good-bye Eshell."🤣
@Mehrad That strikes me as perfectly fine. You asked for a vertical split. You get to keep your vertical split.
I offered C-x k as alternative to the (humorous !!) function you listed. It works for me, and it sounds like you found something that works for you so all is good in the land of The One Editor.
Over on the dead bird site we have another BS thread claim that 'corporations will not touch software XYZ (here: #Rstats) ... because of GPL'.
Now, I have worked in industry since the 1990s: on Wall St, at Fortune 500 corp, at big Hedge Fund. GPL software is everywhere. Most companies do not sell or distribute software and are very well served by the GPL because it ensures these software resources will not go away -- but can, and are used and adapted at will. So just stop the FUD.
Getting a wee bit tired of grandstanding that "#Rstats cannot do reproducibility" because of what may be a misunderstanding: you do not "get" reproducibility "for free" by shoving your (semi-randomly?) collected dependencies into some sort of persistence envelop or environment.
You get reproducibility by carefilly writing code using stable interfaces. I have packages that have been on CRAN 20+ years and work now as they worked then. Still on .libPaths(), and no issues.
No disrespect to Wes McKinney (I don’t like #pandas, but I personally could have never done something like that myself), but there’s literally 0 reason (apart from running legacy code) to use #pandas now when there’s #polars on #Python. With #RStats, #dplyr is still the GOAT
It's complicated. It'll get there. But it may take a toolchain update (or two, or three, ...) to get there. But all things avant-garde become mainstream eventually due to the passage of time ...
@Cmastication I should get on with real work but maybe one big summarising post should focus on just how effin' awesome all these solutions are when they give us one-liners (data.table) or short expressions doing all this in friggin 30 seconds whereas in the original Java contest they have to work rather hard to beat that by large margins.
In short, it is beyond awesome to have a potent data language and with it even competing solutions for every style and preference!
@Cmastication I already did. It runs circles around the others (and for amusement I also did sqlite which craaaaaawls). But I don't think the comparison is fair when I first copy to an in memory table and don't count that time. So not sure how to account for this differece. Maybe your turn, mate! 😜
GHA power users. I have setup-r-dependencies workflow that installs all packages in DESCRIPTION. Well, one of those isn't available and won't ever be. How can I only install Imports packages?
@josi There is a lovely snippet I recently game across in a Dockerfile from the data.table repo. Base #Rstats only, and parses DESCRIPTION. Easy enough for you to adjust to only Imports as desired:
The visualization of the top nine songs among the Spotify TopUS (in the US) is now running on 'auto-pilot' as a simple demonstration of GitHub Actions. It updates the data, does a small transformation and updates chart each night. Currently very seasonal 😉
@dgkf That is something you (or we all) would have to take up with R Core. I just report what I see at the useful RSS summary of changes to NEWS (update daily) and/or the github mirror of the SVN.
The lack of clear communication from either CRAN or R Core is not exactly new, and why we have things like these bots.
So in sum I guess your best bet to faster updates may be to become an R Core member to see their emails 🙃