Troubleshooting

Known Issues & Solutions

If your issue is not listed here, please email Uri about it, or create an issue on groundhog’s github page.

Current issues

Issue #1. Error: package or namespace load failed for ‘groundhog’

For not yet identified reasons, some users get this error message when loading a freshly installed groundhog v>3.0.0 version. It seems to have to do with previous versions of groundhog not fully updating.  The full error/warning message reads:

library(“groundhog”)
Error: package or namespace load failed for ‘groundhog’:
.onLoad failed in loadNamespace() for ‘groundhog’, details:
call: readRDS(cran.times.path)
error: error reading from connection
In addition: Warning message:
In readRDS(cran.times.path) : lzma decoding result 10
Solution: Delete the folder with groundhog files for a clean re-install.
path1<-paste0(path.expand("~"),"/R_groundhog") #find path to folder
unlink(path1,recursive=TRUE)                   #delete it

Issue #2. Failure to install from source

There are two likely causes behind a failure to install from source.

Possible cause #1.  Missing package in the operating system.
A more likely problem is actually unrelated to groundhog. It involves not having a piece of software that R needs to install packages from ‘source’. If you read the error message in your console you are likely to find the piece of software that’s missing indicated.

    • For Unix there are a variety of separate packages that need to be installed, again, usually the error message you obtain when relying on groundhog.library(),   will indicate which package it is that you need.

MAC example
If on Mac Big Sur OSX 11, you run
groundhog.library('lme4', '2020-12-01')

the installation of the package Matrix will fail. Among the copious output in your console you will find this line indicating that gfortran is missing.

    ld: library not found for -lgfortran

You just need to install it from here https://mac.r-project.org/tools/, then Matrix, and thus lme4, will successfully install.

 

Solved issues, affecting older versions of groundhog

Old issue #1. Installation fails reading 'the condition has length > 1'

With groundhog 3.0.0 a bug was introduced that is manifested when groundhog relies on a backup method for moving packages from installation directory to the directory where packages are used. You can tell if you are relying on the backup method (which has the bug) if you get TRUE when running groundhog:::cookie.exists("copy_instead_of_renaming")

The bug has been solved but updated groundhog package with it has not been sent to CRAN yet, pending further testing on April 26th or 27th of 2023. The development version with that fix, and which differs from the CRAN version v3.0.0,  only in having a fix to this bug, is available from github and can also be installed with
install.packages('https://groundhogr.com/groundhog_3.0.0.9001.tar.gz')

Another option is to use the previous version of groundhog till release 3.0.1 hits CRAN. To do that
meta.groundhog('2023-04-01') will load groundhog 2.2.1

Old Issue #2. Persistent package version conflict

With groundhog v<2.0, conflicts with a different version of the same package would sometimes persist after restarting R session (when R Studio would automatically load needed packages) and the solution was to uninstall the packages creating the conflict. Since v2.0 that is handled automatically.

Old Issue #3. 'This does not look like a tar archive'

With groundhog v<1.5, older Windows machines would produce error ‘This does not look like a tar archive’  when installing binaries.