In ?data.table about .SDcols argument:

May be character column names or numeric positions.

However, it seems like also a logical vector is possible, e.g.

d <- data.table(x = 1, y = "a")
d[ , .SD, .SDcols = sapply(d, is.numeric)]
#    x
# 1: 1

If selecting columns with a logical vector in .SDcols is intended to work, should this possibility be added to the help text?

Please also note two differences with j in this regard:

(1) Using a logical vector longer than the number of columns errors in j, but not in .SDcols

d[ , c(TRUE, TRUE, TRUE)]
# Error in `[.data.table`(d, , c(TRUE, TRUE, TRUE)) : 
#  Item 3 of j is 3 which is outside the column number range [1,ncol=2]

d[ , .SD, .SDcols = c(TRUE, TRUE, TRUE)]
#    x y
# 1: 1 a

(2) If a logical vector with length k < ncol(x) is passed to j, it will be filled to length ncol(x) with FALSE. In .SDcols, the logical vector is recycled.

d[ , TRUE]
#    x
# 1: 1

d[ , .SD, .SDcols = TRUE]
#    x y
# 1: 1 a

data.table 1.12.9 IN DEVELOPMENT built 2019-12-13 23:32:57 UTC