Should it be possible to use .N in column names?

[DavidArenburg]

library(data.table) #v1.9.6
DT <- fread("user_id  date  weight
12345    2016-03-07    160
12345    2016-03-06    158
12345    2016-03-05    156
12345    2016-03-04    154")

I'm not sure why I can't just do

DT[, paste0("weight_", 1:.N) := shift(weight, 1:.N)]
# Error in 1:.N : argument of length 0

Instead of

DT[, paste0("weight_", 1:nrow(DT)) := shift(weight, 1:.N)]

Is that a bug or I simply don't understand how .N works?

Motivation: http://stackoverflow.com/questions/35855934/using-dplyrmutate-to-apply-parameterizations-of-a-function-to-a-single-data-fr?noredirect=1

[arunsrinivasan]

No. only on the RHS. But maybe we can get this to work.

[MichaelChirico]

Related #1543

[MichaelChirico]

I think subsumed by #581