Closed jarioksa closed 8 years ago
2.4-0 (master
) and 2.3 (cran-2.3
) branches have diverged so much that many changes cannot be easily cherry-picked from master
to cran-2.3
. Actually, most changes in master
concern db-RDA re-write (dbrda
, adonis
, varpart
) and its collateral effects. One large thing that could be cherry-picked and would not touch many files outside its realm is Strona's Curveball nullmodel algorithm. I have no firm opinion of adding this. The main reason in favour is that it would narrow the gap between 2.3-5 and 2.4-0. The main reason against is that in general we avoid adding major new features in minor releases (but I think this is pretty safe).
Release v2.3-5 made on April 8, 2016.
I had expected to release vegan 2.4-0 this northern spring, and assumed there would be no minor 2.3-5 release. However, R development version introduced a change that breaks the current vegan release when tested in R-devel, and the good people in CRAN suggested we release a fix.
R svn commit 70391 (maechler Wed, 30 Mar 2016) added argument
rank.
toprcomp
to limit the number of eigenvalues displayed. With this he changedstats:::print.summary.prcomp
that we had (ab)used to print the results ofvegan:::summary.eigenvals
, and this broke vegan. The solution was to write our ownvegan:::print.summary.eigenvals
which is similar to the oldstats:::
function. This fix will work both in R-devel and in the currrent R release.There are a couple of other fixes in the
cran-2.3
branch:calibrate.cca
failed if user gaverank
(this was a plain bug)anova.cca
methods and some other methods for the results ofcca
and friends worked only if the original model was fitted with a formula interface. We warned on this in documentation, but not specifically, and the methods just gave an obscure error message without checking the object. Now they check the object and give a clearer error message. I added this because some people claimed a bug when they analysed models fitted without a formula.