When does R Studio autoload packages?

jimvine · February 19, 2021, 12:30pm

I have encountered a problem using the groundhog package, which I think we have tracked down to the behaviour of R Studio in auto-loading packages.

See the entries from 2021-02-19 here for further background info on how this issue presents itself:

github.com/CredibilityLab/groundhog

`groundhog.library()` intermittent failure when loading `Matrix`

opened 11:36AM - 11 Feb 21 UTC

closed 07:14AM - 14 Nov 21 UTC

jimvine

I was going to title this indicating that the failure occurred when `pkg` or `da…te` area specified with double quotes but succeeds with single quotes. However, having restarted my R session and re-run the command a bunch more times, I'm now not sure whether that is a reliable pattern: I have seen the command working with the package name specified in double quotes, and I have seen it fail with the package name in single quotes. So this may be a red herring. --- System info: `Loaded 'groundhog' (version:1.2.0) using R-4.0.3` Windows 10. R Studio 1.4.1103. --- I have this code at the head of my file: ``` library(groundhog) groundhog.day <- "2021-02-02" groundhog.library("Matrix", groundhog.day) ``` This (mostly) fails with an error: ``` groundhog says: 1 of the 2 packages needed by 'Matrix_1.3-2' are currently loaded, but not with the version that is needed. Loaded: Matrix_1.2-18 Needed: Matrix_1.3-2 To solve this: restart the R session. Note: you will need to do 'library(groundhog)' again. In R Studio press: CTRL/CMD-SHIFT-F10 The package 'Matrix_1.3-2' was *NOT* loaded IMPORTANT: if you see this message after having restarted your R session, it means one of the packages you are loading does not properly reference a package it needs (in its 'DESCRIPTION' file). To solve this, you need to restart the session again, CTRL-SHIFT-F10, then run the code below **BEFORE** loading the packages you want to load: groundhog.library('Matrix','2021-02-02') ``` I looked at the command suggested in the last line of the error message and thought that it seemed unlikely to help since it looked equivalent to what my code said. But still, I tried it and was pretty surprised to find it worked. Over a few more restarts of the R session, I narrowed down the differences (trying both the date field and the pkg field). In my experiments, it seemed like I found that if either the pkg or date value were specified with double quotes, the command failed with the same error. If the date was specified as a named variable it succeeds (even if that was defined with double quotes). ``` groundhog.day <- "2021-02-02" # MOSTLY FAILS: groundhog.library('Matrix', "2021-02-02") groundhog.library("Matrix", "2021-02-02") groundhog.library("Matrix", '2021-02-02') groundhog.library("Matrix", groundhog.day) # MOSTLY SUCCEEDS: groundhog.library('Matrix', '2021-02-02') groundhog.library('Matrix', groundhog.day) ``` As I say, this association may be spurious. I am certain that the package loading sometimes works and sometimes does not work using variants of the command that I would normally expect to be equivalent. I know that I have seen a tendency (but not, now, a hard-and-fast rule) that it seems to fail more often with double quotes than single. But I have weak confidence in whether that's actually the material factor in why this sometimes succeeds and sometimes fails. I should also note that I am loading about 10 packages using groundhog. It is only Matrix that is throwing this error. The others are all loading fine. (For the rest I use the names in double quotes and the date as the `groundhog.day` variable I've specified.)

When I open R Studio afresh it is set to open my existing project. In Tools > Global Options... I have "Restore .RData into workspace at startup" set to false (unchecked). In Tools > Project Options this is set to "No".

Opening R Studio and immediately entering sessionInfo() at the Console, I get:

> sessionInfo()
R version 4.0.2 (2020-06-22)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19041)

Matrix products: default

locale:
[1] LC_COLLATE=English_United Kingdom.1252  LC_CTYPE=English_United Kingdom.1252   
[3] LC_MONETARY=English_United Kingdom.1252 LC_NUMERIC=C                           
[5] LC_TIME=English_United Kingdom.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.6      digest_0.6.25   crayon_1.4.1    lifecycle_0.2.0 magrittr_2.0.1  evaluate_0.14  
 [7] pillar_1.4.7    rlang_0.4.10    vctrs_0.3.6     ellipsis_0.3.1  rmarkdown_2.3   forcats_0.5.0  
[13] tools_4.0.2     hms_0.5.3       yaml_2.2.1      xfun_0.16       compiler_4.0.2  pkgconfig_2.0.3
[19] htmltools_0.5.0 knitr_1.29      haven_2.3.1     tibble_3.0.6

I have not consciously done anything to load the set of packages listed at the end of that sessionInfo() report.

I have checked the three locations mentioned in this Stack Overflow answer, and do not see a .Rprofile file in any of them.

If I restart my R session (Ctrl+Shift+F10) and re-run sessionInfo(), the packages are not loaded:

Restarting R session...

> sessionInfo()
R version 4.0.2 (2020-06-22)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19041)

Matrix products: default

locale:
[1] LC_COLLATE=English_United Kingdom.1252  LC_CTYPE=English_United Kingdom.1252   
[3] LC_MONETARY=English_United Kingdom.1252 LC_NUMERIC=C                           
[5] LC_TIME=English_United Kingdom.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] compiler_4.0.2 tools_4.0.2

However, when I run a chunk in my Rmd file that ostensibly doesn't have anything to do with most of these packages, it seems to trigger loading of them. This, however, is intermittent. Sometimes the packages are loaded and sometimes they are not.

If the chunk I am running is the first chunk in the Rmd, that seems from my experiments to not trigger the loading of these packages.

If I have a chunk above my intended chunk headed {r setup, include=FALSE} containing only knitr::opts_chunk$set(echo = TRUE), this seems to normally be enough to cause the packages to load.

If, however, I have a chunk above my intended chunk headed just {r} and containing only an empty line, it seems to cause the packages to be loaded about half the time, and not be loaded the other half.

I would like to understand what it is that causes R Studio to automatically load or not load packages. I would ideally like to know what to set to stop it from doing these things, but if there's no setting then understanding what the triggers are so that I can work around them would be good.

For example, I've restarted R and re-run my code probably a couple of dozen times in different variants. I think I see a pattern where the presence of a setup chunk pretty consistently triggers the loading of packages, and where the presence of an untitled empty chunk sometimes but not always triggers it. But the number of repetitions I can manually run in each configuration means that this could be a coincidence.

I also tried creating a MWE, and the presence of at least one subsequent chunk seemed to be relevant for whether the loading was triggered. But again, I have somewhat low confidence that that was a real effect rather than coincidence.

And I seemed to get the packages autoloaded if I ran a chunk (using the play button) or if I selected the code and hit Ctrl+Enter, but not if I copied the code and pasted it into the Console. (EDIT: I'm less certain on this observation now. I have been trying to work around the autoloading by running the relevant code by pasting it into the Console, and I think it does sometimes still trigger some packages to be loaded.)

Any information on what's going on here would be much appreciated. Thanks in advance!

system · March 12, 2021, 12:30pm

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.