[R-pkg-devel] Impact of Matrix requirement for R >= 4.4.0 on CRAN package ecosystem
Hi Ivan, Thank you so much for your answer. Everything is clear now. First, we have to ask ourselves if my package needs a more modern Matrix version. Then, if I need a newer Matrix than 1.7-0, I set the minimal required Matrix version and the corresponding R version. If I need Matrix, but the one provided for a specific R version (from R X version) is sufficient, I simply specify Matrix as a dependency, and it does not influence my minimal R requirement. The edge case is an R installation without "Recommended" packages, which can cause additional manual Matrix installation. KR Maciej
On Tue, 20 Jan 2026 at 13:04, Ivan Krylov <ikrylov at disroot.org> wrote:
Dear Maciej Nasinski, ? Tue, 20 Jan 2026 11:57:49 +0100 Maciej Nasinski <nasinski.maciej at gmail.com> ?????:
Because so many packages depend on Matrix, this change effectively raises the "floor" for a large portion of CRAN to R 4.4.0. Is it true?
Ideally, only for those depending on Matrix (>= 1.7-0), currently CoTiMA, FlexRL, gamm4, matrixCorr, organik, and their strong recursive reverse dependencies (29 more packages via gamm4). If the dependencies of the remaining packages are truly satisfied by Matrix < 1.7-0, installing them shouldn't pose a problem, as Matrix is likely to be pre-installed as a "Recommended" package. Even if it's not installed and the distribution source code is missing, it's still possible to download a matching version of Matrix from CRAN, e.g., from https://cran.r-project.org/src/contrib/4.0.0/Recommended/. Still, it's a reason to worry, because most developers don't test their packages with the minimal version of every dependency installed (some CRAN packages even have dependencies on minimal versions of R that don't exist!). There may be packages that effectively depend on a later version of Matrix and don't know it. -- Best regards, Ivan