.Rprofile crashes R Jupyter notebook

[nbafrank]

Hi all,

I am using a Jupyter notebook with an R kernel and we want to use renv with a particular project. After initializing the env, we find that the presence of .Rprofile in the folder crashes the R kernel in the notebook. If we remove .Rprofile from the folder and activate manually, everything works. This also works if we activate the renv manually from a different folder (e.g. using renv::activate("path to folder")), because the .Rprofile file is not there.

Any ideas?

Thank you so much for your time and help

[kevinushey]

Thanks for the bug report -- can you provide some instructions so that I might be able to easily reproduce myself?

[micdonato]

I was having the same issue. This seems to be due to Jupyter trying to load IRKernel from the renv, and if that has not been installed, it fails.

This was the code that broke it. This code needs to be in a cell in a notebook within the directory where renv is.

setwd('/home/micdonato/rtest') 
library(renv)
init()
renv::install('tinytest')

Trying to restart the kernel creates the error.

The error was this (part of it):

Error in loadNamespace(x) : there is no package called 'IRkernel'
Calls: loadNamespace -> withRestarts -> withOneRestart -> doWithOneRestart
Execution halted
[I 2022-02-03 19:21:15.261 SingleUserLabApp restarter:75] AsyncIOLoopKernelRestarter: restarting kernel (1/5), new random ports
Error in loadNamespace(x) : there is no package called 'IRkernel'
Calls: loadNamespace -> withRestarts -> withOneRestart -> doWithOneRestart
Execution halted

(mind that the notebook needs to be in the same directory of the renv)

Then I tried to check what the .libPaths() were. renv does not know about /usr/local/lib/R/site-library but it knows only about /usr/local/lib/R/library. I think that makes sense, to isolate the packages installed in the environment from the ones installed globally.

The solution then is either letting renv know about /usr/local/lib/R/site-library (removing the isolation) or installing IRkernel when setting up the environment with renv::install("IRkernel"). If I do that the kernel starts properly.

[hadley]

IMO this is more of a jupyter issue — failure to find IRKernel shouldn't cause jupyter to crash, but should instead generate a useful message. But I'll keep it open in case we want to do a patch to jupyter.

[hadley]

Actually we should solve this by adding IRKernel as a dependency when we see jupyter notebooks, and we should also parse .ipynbs for package dependencies.