Replies: 1 comment
-
I did some investigation on this topic, and found this passage in Hadley's book on R package development (at https://r-pkgs.org/description.html#depends-and-linkingto) ============ The most legitimate current use of Depends is to state a minimum version for R itself, e.g. Depends: R (>= 4.0.0). Again, think carefully if you do this. This raises the same issues as setting a minimum version for a package you depend on, except the stakes are much higher when it comes to R itself. Users can’t simply consent to the necessary upgrade, so, if other packages depend on yours, your minimum version requirement for R can cause a cascade of package installation failures. The backports package is useful if you want to use a function like tools::R_user_dir(), which was introduced in 4.0.0 in 2020, while still supporting older R versions. The tidyverse packages officially support the current R version, the devel version, and four previous versions.3 We proactively test this support in the standard build matrix we use for continuous integration. Packages with a lower level of use may not need this level of rigor. The main takeaway is: if you state a minimum of R, you should have a reason and you should take reasonable measures to test your claim regularly. ============ Before we commit to setting an R version dependency, let's pause for a moment and think about how that might impact pacehrh users (not just ourselves). |
Beta Was this translation helpful? Give feedback.
-
Discussed briefly with @celiot-IDM
Currently the automated build and testing were using R 4.2 version
although we do not anticipate imcompatibility for older/newer versions but may be a good practice to declare a dependency to a minimum R version in DESCRIPTION file.
also we should discuss if build/test need to be updated for future R versions (e.g. 4.3)
Beta Was this translation helpful? Give feedback.
All reactions