Hi @matthias-da — vip maintainer here with a heads-up, since your package depends on it.
CRAN will archive vip (0.4.6) on 2026-07-13. Rather than keep chasing CRAN's policies, I've moved vip to r-universe, where the latest stable release (0.5.0) already lives:
install.packages("vip", repos = c("https://bgreenwell.r-universe.dev", "https://cloud.r-project.org"))Since vip sits in your Suggests, the main action item is making sure CRAN can still find it: use vip conditionally (requireNamespace("vip", quietly = TRUE)) in examples/tests/vignettes and add Additional_repositories: https://bgreenwell.r-universe.dev to your DESCRIPTION — CRAN accepts that for Suggests-level dependencies.
Separately, vip 0.5.0 is a breaking release (details in the migration notes):
vip() now draws lightweight base R graphics (via tinyplot) and invisibly returns the "vi" object — it no longer returns a ggplot, so vip(...) + theme_*()-style code breaks.vi() returns a plain data frame with class "vi" instead of a tibble ($Variable/$Importance access is unchanged).- ggplot2 and tibble were dropped from vip's dependencies entirely.
From a quick look, compare_feature_importance.R is the call site to check — vi() scores work the same in 0.5.0 (minus the tibble class), but any vip() plot handling would need updating.
Happy to help or send a PR if useful — just let me know! And apologies for the churn.
Hi @matthias-da — vip maintainer here with a heads-up, since your package depends on it.
CRAN will archive vip (0.4.6) on 2026-07-13. Rather than keep chasing CRAN's policies, I've moved vip to r-universe, where the latest stable release (0.5.0) already lives:
Since vip sits in your
Suggests, the main action item is making sure CRAN can still find it: use vip conditionally (requireNamespace("vip", quietly = TRUE)) in examples/tests/vignettes and addAdditional_repositories: https://bgreenwell.r-universe.devto your DESCRIPTION — CRAN accepts that for Suggests-level dependencies.Separately, vip 0.5.0 is a breaking release (details in the migration notes):
vip()now draws lightweight base R graphics (via tinyplot) and invisibly returns the"vi"object — it no longer returns a ggplot, sovip(...) + theme_*()-style code breaks.vi()returns a plain data frame with class"vi"instead of a tibble ($Variable/$Importanceaccess is unchanged).From a quick look,
compare_feature_importance.Ris the call site to check —vi()scores work the same in 0.5.0 (minus the tibble class), but anyvip()plot handling would need updating.Happy to help or send a PR if useful — just let me know! And apologies for the churn.