split.data.table does not respect factor ordering

[MichaelChirico]

Pretty self-explanatory. MWE:

DT = data.table(a = factor(c('a', 'b'), levels = c('b', 'a')))
split(DT, by = 'a', sorted = TRUE)
# $a
#    a
# 1: a

# $b
#    a
# 1: b
split(DT, f = DT$a, sorted = TRUE)
# $b
#    a
# 1: b

# $a
#    a
# 1: a
split.data.frame(DT, f = DT$a)
# $b
#    a
# 1: b

# $a
#    a
# 1: a

[jangorecki]

AFAIR when using f argument then sorted is ignored. I don't see the point in extending f use case features when it is recommended to use by. f is less efficient, it is there so data.frame code can work with data.table method. Using by here is solution good enough, correct?

[MichaelChirico]

@jangorecki I should have added more exposition. the focus is the first example, where I use by but the output order is incorrect. But the order IS correct when I use f not by

[MichaelChirico]

@jangorecki check my PR. You're right that if we use f it's going to be less efficient, since internally it basically just passes to split.data.frame in this case.

However as mentioned, what I was intending to do was indeed use by, but it was giving incorrect output. That should be fixed now.