data.table - delete columns programmatically

Question

What is the best practice to delete columns programmatically in data.table?

The following works:

DT[, c("a", "b") := NULL]

But when trying to do this using a variable that stores column names

cols.to.del <- c("a", "b")
DT[, cols.to.del := NULL]

it fails badly as cols.to.del is not evaluated in the correct environment.

akrun · Accepted Answer · 2015-12-19T17:09:07.747

7

We can wrap it inside the brackets, and then assign (:=) to 'NULL' (preferred way)

DT[, (cols.to.del) := NULL]

Or another option (in case we don't want to wrap it with brackets) would be to loop over the 'cols.to.del' in a for loop and assign to NULL

for(j in seq_along(cols.to.del)){
    DT[, cols.to.del[j] := NULL]
}

Or for subsetting the columns, we can use setdiff along with with=FALSE.

DT[, setdiff(names(DT), cols.to.del), with=FALSE]

edited Dec 19 '15 at 17:09

answered Dec 19 '15 at 16:53

akrun

1

The third version is definitely my favorite. Thanks! – paljenczy Dec 19 '15 at 16:57
7

@paljenczy note that the third version doesn't remove the columns from `DT`; rather the output of that command is a _new_ `data.table` which you must _assign_ (by copying) to `DT`, which is likely inefficient. – MichaelChirico Dec 19 '15 at 17:49
1

@MichaelChirico noted, thanks! – paljenczy Dec 19 '15 at 20:13

1 Answers1