Inconsistent NA handling in createDataPartition

[MichaelChirico]

caret/pkg/caret/R/createDataPartition.R

Line 122 in c98cc1a

xtab <- table(y)

As I'm finding in work towards #1382, missing inputs are handled a bit inconsistently by createDataPartition():

library(withr)

# y is numeric, missing --> error
createDataPartition(c(NA, 1))
# Error in quantile.default(y, probs = seq(0, 1, length = groups)) : 
#   missing values and NaN's not allowed if 'na.rm' is FALSE

# y is character, level with only one observation --> warning
with_seed(78458, createDataPartition(c("a", "b", "b")))
# $Resample1
# [1] 1 3
# Warning message:
# In createDataPartition(c("a", "b", "b")) :
#   Some classes have a single record ( a ) and these will be selected for the sample

# y is character, missing level has only one observation --> no warning
with_seed(78458, createDataPartition(c(NA, "b", "b")))
# $Resample1
# [1] 1 3

What is the proper handling of missing y? Should we just error if y has any missing observations? That would certainly make the migration away from dlply() simpler (c.f. 5805d30).