Dear all,
pmin/pmax used to work with ordered factors but fail now since this
summer when those functions have tried to handle more cases.
A simple way to trigger the issue is:
> min(ordered(c(1,5,6)))
[1] 1
Levels: 1 < 5 < 6
> pmin(ordered(c(1,5,6)), ordered(c(1,5,6)))
Error in `mostattributes<-`(`*tmp*`, value = attributes(elts[[1L]])) :
adding class "factor" to an invalid object
A simple fix is to explicitly test for the ordered class and use the
internal method in that case as proposed in the attached patch.
Yours,
Erwan
Index: src/library/base/R/pmax.R
===================================================================
--- src/library/base/R/pmax.R (revision 71694)
+++ src/library/base/R/pmax.R (working copy)
@@ -24,7 +24,7 @@
elts <- list(...)
if(length(elts) == 0L) stop("no arguments")
i4 <- isS4(elts[[1L]])
- if(!i4 && all(vapply(elts, function(x) is.atomic(x) && !is.object(x), NA))) {
+ if(!i4 && all(vapply(elts, function(x) is.atomic(x) && (!is.object(x) || is.ordered(x)), NA))) {
## NB: NULL passes is.atomic
mmm <- .Internal(pmax(na.rm, ...))
} else {
@@ -68,7 +68,7 @@
elts <- list(...)
if(length(elts) == 0L) stop("no arguments")
i4 <- isS4(elts[[1L]])
- if(!i4 && all(vapply(elts, function(x) is.atomic(x) && !is.object(x), NA))) {
+ if(!i4 && all(vapply(elts, function(x) is.atomic(x) && (!is.object(x) || is.ordered(x)), NA))) {
mmm <- .Internal(pmin(na.rm, ...))
} else {
mmm <- as.vector(elts[[1L]]) # attr(mmm, "dim") <- NULL # dim<- would drop names
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
