Circumvent package size limit

[vincentarelbundock]

Hi,

Thanks for your essential work on R-Universe!

My marginaleffects package repository includes largish data files which are required for continuous integration testing on Github Actions. When R-Universe tries to build the Windows binary, the process errors because the package exceeds the default 100M size limit (error pasted below).

I know that CRAN imposes a similar limit. When I submit packages there, I manually add large files to .Rbuildignore.

Is it possible to do something like that on R-Universe? Can I customize the build actions to either allow bigger packages or ignore some files? How do you recommend I circumvent the file size limit?

Thanks!

Run if [ -n "$(find "marginaleffects_0.21.0.14.zip" -prune -size +100M)" ]; then
  if [ -n "$(find "marginaleffects_0.21.0.14.zip" -prune -size +100M)" ]; then
    echo "File marginaleffects_0.21.0.14.zip is larger than 100 MB. This is not allowed"
    exit 1
  fi
  shell: C:\Program Files\Git\bin\bash.EXE --noprofile --norc -e -o pipefail {0}
  env:
    UNIVERSE_NAME: vincentarelbundock
    MY_UNIVERSE: https://vincentarelbundock.r-universe.dev/
    GH_APP_ID: 87942
    RTOOLS42_HOME: C:\rtools44
    RTOOLS43_HOME: C:\rtools44
    _R_INSTALL_TIME_PATCHES_: https://www.r-project.org/nosvn/winutf8/ucrt3/
    JAVA_HOME: C:\hostedtoolcache\windows\Java_Temurin-Hotspot_jdk\17.0.12-7\x64
File marginaleffects_0.21.0.14.zip is larger than 100 MB. This is not allowed
Error: Process completed with exit code 1.

[jeroen]

I think the same way as CRAN? If you only need those files for CI, add them to .Rbuildignore such that they do not get included in the deployed packages on r-universe.

[vincentarelbundock]

Thanks. The problem is that if I add them to .Rbuildignore in the main repository, my CI test suite will no longer run on Github Actions.

[jeroen]

Perhaps you can add them to .Rbuildignore and add a step to your own CI workflow to remove this line? Or you can condition something on the presence of the CI environment variable, which is set in GHA but not on CRAN/R-universe?

[vincentarelbundock]

lol, I feel ridiculous for not having thought of this obvious solution. Thanks a lot for your time and suggestion. And thanks again for your work on R-universe!