[datatable-help] key<- messes with my factors

[Chris Neff]

key<- with a factor mixes up factor levels, and loses the is.ordered
property too.

Pretty simple example:

> dt=data.table(x=factor(1:20,ordered=TRUE),y=rnorm(20))
> dt$x
 [1] 1  2  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 20
Levels: 1 < 2 < 3 < 4 < 5 < 6 < 7 < 8 < 9 < 10 < 11 < 12 < 13 < 14 <
15 < 16 < 17 < 18 < 19 < 20
> dt$x
 [1] 1  10 11 12 13 14 15 16 17 18 19 2  20 3  4  5  6  7  8  9
Levels: 1 10 11 12 13 14 15 16 17 18 19 2 20 3 4 5 6 7 8 9


With integers I could of course just leave them as integers instead of
turning it into a factor, but the real case happening to me is I have
upper cutoffs of buckets of a float variable, so I can't keep the
simple numeric value for a key so it must be a factor. However, I want
it to be ordered so that the factors still go lowest bucket to
highest.


I think this is a bug but I want confirmation before filing.

-Chris