Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign upNew behaviour of droplevels in Rdevel #193
Comments
|
It seems that |
|
Right; I was in two minds whether to continue this when I saw the R-devel thread. However, the proposed changes don't break anything nor rely on whatever behaviour RCore ultimately settles upon, but they do make explicit the behaviour we want/expect/anticipate and so should be robust to RCore tweaking. I would suggest we don't push out a 2.4-1 just for this though if we don't continue to generate errors when R-devel version settles down. We can hold off till a future 2.5-0 release if that is preferable? |
|
It seems that
From that point of view, we need not push a new minor release. However, there are other bug fixes that may call for a minor release like discussed in issue #194. |
|
Merged into |
R Core have altered the behaviour of
droplevels(). For some initial discussions see #192. The basic issue is that nowdroplevelsis defined asfactor(x, ..., exclude = NULL)and this coupled with changes tofactor()whenexclude = NULLmean behaviour may have changed. To maintain behaviour of R v3.3 we should usedroplevels(x, exclude = NA). This issue exists to track changes to the range of functions in vegan that usedroplevels.Current list of functions known to use
droplevelscentroids.ccaenvfit.defaultfactorfitbetadisper(Fixed in ae67eba, see #192)hiersimu.defaultadipart.defaultmultipart.defaultI'm tracking this on the issue-#193 branch.