Skip to content

v4.0.0-rc5 (internal release candidate)

Pre-release
Pre-release

Choose a tag to compare

@ehrlinger ehrlinger released this 11 Sep 15:45
· 86 commits to main since this release
5ea540f

Internal release candidate on the v4.0.0 line. NOT a CRAN release: CRAN is on
3.5.3 (published 2026-09-11) and the checklist's Release status stays HOLD.

Verified on 5ea540f, the commit tagged (origin/main):
lintr 0 lints; NOT_CRAN=true VDIFFR_RUN_TESTS=true devtools::test()
841 tests, 2256 expectations, 0 fail / 0 error / 6 skips / 58
pre-existing warnings; 62 vdiffr baselines intact with an empty git status before and
after; clean git archive build; R CMD check --as-cran with the manual at
1 NOTE / 0 WARNING / 0 ERROR (CRAN incoming feasibility only: 0 days
since last update, 9 updates in the past 6 months); tests 103s and
vignette rebuild 217s elapsed; tarball 4.0 MB, no stray hidden files (.Rinstignore only), no
cran-comments, all 13 varpro.html figures present. DESCRIPTION Date is
unchanged at 2026-08-31. The GCC UBSAN gate was not re-run for this RC.

Since v4.0.0-rc4: 71 commits across 19 PRs, #266 to #289.

Public API. gg_ale_rfsrc() (#271): accumulated local effects for rfsrc
forests, with plot, autoplot, print and summary methods for gg_ale_rfsrc and
gg_ale_interaction. gg_brier() gains a crps_std attribute (#283).
plot.gg_vimp(relative = TRUE) now plots relative VIMP, and nvar keeps the top
variables rather than the top rows, so classification plots keep their class
panels (#285). One returned-object change: the randomForest no-importance NA
placeholder from gg_vimp() no longer carries an all-NA rel_vimp column (#289).

Dependencies. randomForestRHF 2.0.3 cumulative-hazard semantics adopted, with
a guard for older versions (#268, #270); five survival vdiffr baselines
regenerated for randomForestSRC 3.7.0 (#272); the sdependent test mock made
compatible with varPro 3.2.1 (#275).

Docs. New model-explainability vignette (#274). Survival and regression
vignettes seeded, with their prose matched to the seeded fits (#278, #286,
#287). A no-ai-slop pass (#282, #288) and the factual corrections it surfaced
(#285); crps_integrated documented as the raw Brier integral (#280).

Tarball is 4.0 MB, up from 3.59 MB at rc4; headroom against CRAN's 5 MB limit
is 1.0 MB. The growth is not yet attributed; the new vignette is the likely
source.