CRAN breaking monitor

[jarioksa]

Next vegan release after 2.6-x is scheduled to have changes that can break CRAN packages (discussion #645). This issue monitors breakage and the goal is to fix all these before the release. I run occasionally test of reverse dependencies agains CRAN release using tools package and post here the latest tests with comments. These tests are not completely fool-proof, but they are similar as ran by the CRAN team. 231 CRAN and Bioconductor packages depended on the vegan when 2.6-6 was released (but the number is volatile). I cannot build all reverse dependencies for reasons not dependent on vegan – Bioconductor packages are particularly painful.

Currently vegan master has these changes that can break other packages:

• summary.cca disallows extracting scores. Breaks some packages.
• summary.decorana is defunct. No breaks.
• adonis is defunct. Breaks several packages.
• orditkplot: no breaks.
• adding pca, ca & pco functions (Suggestion : Add functions for PCA and CA that just call rda() and cca() in the background #655) name clash breaks some packages (Adding pca and ca (wish of #655) breaks two packages in reverse dependencies tests #664)

Here test of CRAN and BioConductor 3.20 against vegan 2.6-8 (last tested with vegan 2.7-0 tagged as 2.6-7-368-gd3209c0f on 2025-01-11):

adonis (See #641 for fixing)

• betaper examples ERROR: patch emailed 2024-04-30
• GCalignR FIXED on CRAN
• hagis FIXED on CRAN
• mefa FIXED on CRAN
• microbial FIXED on CRAN
• MMUPHin FIXED (Bioconductor package)
• RVAideMemoire examples ERROR: author contacted 2024-04-24
• SigTree examples & re-building of vignette outputs ERROR: patch emailed 2024-04-30
• smartsnp examples ERROR: patch emailed 2024-04-30, maintainer responded

summary.cca (see #644 for fixing)

• BiodiversityR examples ERROR. Release 2.17-1 ChangeLog claims fixing this, but the package has unchanged files with error.
• bipartite FIXED on CRAN
• bipartiteD3 FIXED on CRAN
• LorMe PR on github 2024-11-26.
• pctax FIXED on CRAN

pca, ca & pco

• easyCODA Fixed on CRAN.
• PLSDAbatch Fixed on BioConductor 3.20.

There may also be cases where the use of defunct functions is not detected, because the code is not tested. From the CRAN point of view these do not bother us, but naturally that code should be fixed as well, because intended functionality fails. However, it is difficult to detect those cases. Scanning the reverse dependence test directory for possible cases of adonis found some packages that do not appear in the list above:

$ grep -l "adonis(" rdepends_*/00_pkg_src/*/R/*R | cut -d / -f 3,5
BiodiversityR/nested.npmanova.R
GUniFrac/GUniFrac.R
MicrobiotaProcess/method-binary_offspring.R
MicrobiotaProcess/method-mp_adonis_anosim.R
PCPS/FUN.ADONIS.R
QFASA/All_Rep_Code.R
RVAideMemoire/adonis.II.R
SigTree/adonis.tree.R
betaper/adonis_pertables.R
microbial/Functions.R
microbiomeExplorer/inter_module.R
multiWGCNA/diffModuleExpAnalysis.R
omicplotR/internal.R
smartsnp/smart_mva.R
smartsnp/smart_permanova.R

It seems that GUniFrac, PCPS, microbiomeExplorer, multiWGCNA and omicplot should be fixed and move from adonis to adonis2 although they do not fail in tests. Superficially it looks like most of these cases are trivial: internal structure is not accessed and all you need to do is to use vegan::adonis2. BiodiversityR (for adonis2), MicrobiotaProcess and QFASA seem to be OK, and the other packages are dealt with those that fail tests.