1016
kids are gowing up faster and faster
(mander.xyz)
A place for majestic STEMLORD peacocking, as well as memes about the realities of working in a lab.
Rules
Eli5?
R is a complicated statistics programming language usually used by people in undergraduate and post graduate STEM degrees.
I have been IT support for medical researchers and that application is a BEAST. Installing and configuring it can be a nightmare. Especially when the researchers aren't proficient in it already. Watching someone who is good at it is like watching the pinball wizard.
What's nightmarish about the installation? Is it because medical stuff is still on like Win XP?
Installing on more modern Windows systems is pretty simple. Install the R distro from CRAN + almost certainly RStudio from Posit. Should be pretty plug-and-play. Not nearly as fiddly as LaTeX installs are.
Anything more than the most basic bare-bones install of Workbench (formerly Rstudio) quickly turns nightmarish. Try setting it up on a Linux dedicated server with AD auth with auto-mounting of network shares per-user. Posit's documentation isn't great (or even agrees with itself across pages) even in the simplest best-case scenario, and if you deploy anything that's even slightly complicated, it turns into a Hellscape. There's a good chance you will end up on one of the Posit employees' blog to read an incomplete explanation of setting up a feature because it's entirely missing or incomplete in the documentation. This isn't some crazy off-the-wall edge scenario either, it's an (allegedly) supported configuration and would be a typical deployment scenario in a multi-user R environment.
Their support is absolute shit too, it's truly fucking atrocious. First-level support will not solve your issue, I promise you that, and you won't get anyone who actually knows WTF they're talking about until you're escalated at least twice. And even then, they are very much up their own ass and have a VERY snobby attitude about the product, and always assume that it's the user at fault, even when you provide absolute 100% proof that it's their product at fault. It obviously couldn't possibly be their Super Precious Perfect-in-Every-Way Golden God Product, because as we've previously established before, it is a Perfect Product Which Does No Wrong, Ever. They also love to try and shirk responsibility and say that X is not a supported configuration for literally everything, and then claim that the documentation must be wrong when you point out in their documentation that it is.
Don't even get me started on the Lovecraftian nightmare that is R package management. It's even worse than the essay I just typed out, and they want to charge you essentially the entire Workbench license cost x2 to make it usable. Their logging is useless too, it has basically two settings, one of which is essentially "nothing," and the other is "firehose of bullshit that you need to follow along in their source code to try to find anything useful." That's not an exaggeration, I actually had to do that to diagnose an issue and provide proof to them that yeah, it is your half-assed shit product that's the problem.
So yeah, if you're not just Click-Click-Click-Next installing it, it very quickly becomes nightmarish. Posit desperately needs competition in the space, because they're absolute shit, but they can be absolute shit with impunity since they don't have any real competition.
Can confirm, it's not just you. I had no problem with it in college and my first job because I had full access to my system. My latest two jobs, though, fucking hell is it a pain in my ass, especially since I'm not required to use it but am doing so voluntarily to up the quality of our data analysis (it used to be terrible descriptive statistics and incorrectly performed Excel t tests). So IT doesn't want to even touch it yet I need their constant password pasting to make it go.
I just bring my own laptop when I need to do stats to things.
My condolences. And as much as I hate Posit/Workbench, some of that is on your IT department. If I'm reading your comment correctly, you're having issues because it needs admin privileges to update some things, namely packages. That's honestly a very simple fix, they just need to grant your user NTFS write permissions to the Rstudio/Workbench install directory locally (and maybe some registry keys, but that's not definite). That's it. It's a 10-second permanent fix and no more UAC prompts for you.
You are 100% correct. Unfortunately, due to the nature of my business, my access to sensitive information, and the mind blowing incompetence of whoever wrote our IT security policies, they won't do so. My PC is locked down tight to keep sensitive data from wandering off... Except for how I can connect my phone (not a USB drive, those are blocked), transfer data to it, analyze it on my PC, and transfer the results back using my phone. That is actually the IT sanctioned workaround. It makes my brain hurt.
I guess I must've saved my IT department some headache when me and my colleagues just asked for RStudio IDE. Everything runs perfectly well, no need for any of the garbage you just described. I literally just need an IDE to write scripts in. I'd say I don't even need the IDE, except I do use
rstudioapi::getActiveDocumentContext()$path
to set the working directory and work with relative paths. Plus I just like it, barebones as it is (and dark mode).Sorry for the nightmare you clearly had to go through. :(
I'll admit it's been awhile but about 10 years ago, it could be a thing. One guy needed almost a week of hands on to get it right.
How is Pak better than pacman, which is my preferred package management solution for R.
Idk, that was someone else's comment. I just
install.packages()
andlibrary()
.I've almost entirely switched to using pak for managing my R packages. I'm not 100% sure what the magic is in pak, but it's really made my life easier when installing packages.
I barely even write base R anymore, I mostly use it for data wrangling these days so my code is almost entirely tidyverse. Every once in a while I get to bust out some statistics, but rarely.
I'm a tidyverse zealot and I just cannot stand fixing people's 300 line base R spaghetti that can easily be refactored into 10 lines of dplyr. Especially annoying when researchers can't move away from doing everything in matrix format (when it's unnecessary).
I use python as my main programming language, I'm doing an MBA in actuarial sciences and all my professors use R, so all the classes and exercises are in R. They are kind enough to accept my exercises and exams in python, but I spent half my time translating R functions to python. This pass week I found the first function that doesn't exists in python and had to learn how to run R code inside python. Just the cell of that function took 6hs processing, because of the back and forth between python to R to python again.
I'm not sure if you're using Quarto or not, but use it almost daily and frequently write R and Python in the same file without any noticeable overhead.
See everyone I know who uses R ends up at that point when there’s just no way around learning it and using it 😂
... complicated? it's a pretty easy language in comparison to others, it's simple to use (although it can be quite terse)
For the untrained eye can look complicated because it had a pretty unorthodox syntaxes, like <- to variable assign, c() to create a vector, df$column... and other little R specific things that are not common on other languages.
R is made infinitely better with tidyverse.
Thank you very much
ELI4?