data.table: temporarily rename single column while keeping all others-CodePudding

Consider the table below:

DT <- data.table(x = 1, y = 10, z = 100, w = 1000)

I want to rename the column z to y, while keeping one or more columns in the table. The columns and renaming would be specified by variables. For example:

keep <- c("x", "w")
old <- "z"
new <- "y"
# Manually:
DT[, .(x, w, y = z)]

In addition to that, I want the original y column to still exist. That is, I want the renaming to be temporary, and to be able to go back to the original table.

Solutions I tried:

Partially manual with ... But returns something I did not expect and do not understand why that was returned.

DT[, .(..keep, y = z)]
>    ..keep   y
> 1:      x 100
> 2:      w 100

Selecting everything but the column that would be replaced, then renaming with setnames:

keep <- setdiff(colnames(DT), "y")
newDT <- DT[, ..keep]
data.table::setnames(new_DT, old, new)
print(newDT)
>   x   y    w
> 1: 1 100 1000

The second option works as expected, but it does not copy values by reference.

Using dplyr::select and dplyr::rename.

newDT <- dplyr::rename(dplyr::select(DT, -y), y = z)

Works as expected, but copies are not by reference, and I need dplyr instead of just using data.table.

Question: Is there a way of temporarily renaming one column while keeping other columns around, without copying data (done by reference)? That is, is there a way to recreate a table by reference for most columns, except change one of them.

Thanks for helping! Please let me know is something is not clear.

CodePudding user response：

We could use mget to return the values of 'keep' in a list and then concatenate (c) with the list renamed (.(y = z))

 DT[,  c(mget(keep), .(y = z))]

-output

      x     w     y
   <num> <num> <num>
1:     1  1000   100

CodePudding user response：

I could not find a one-liner that accomplishes the answer by reference, but I was able to use setnames to obtain a solution that does not copy the data.

Consider the original data:

DT <- data.table(x = 1, y = 10, z = 100, w = 1000)
.Internal(inspect(DT))
> @1b588e10 19 VECSXP g0c7 [OBJ,REF(7),ATT] (len=4, tl=1028)
  @1510aea0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1
  @1510b1b0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 10
  @1510b4f8 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 100
  @151079a0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1000

To rename z into y without losing y and not copying any data, we need to:

Rename the original y into something else with setnames (this is done by reference):

setnames(DT, "y", "y_original")
.Internal(inspect(DT))
> @1b588e10 19 VECSXP g0c7 [OBJ,REF(17),ATT] (len=4, tl=1028)
  @1510aea0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1
  @1510b1b0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 10
  @1510b4f8 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 100
  @151079a0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1000

Rename z into y with setnames:

setnames(DT, "z", "y")
.Internal(inspect(DT))
> @1b588e10 19 VECSXP g0c7 [OBJ,REF(27),ATT] (len=4, tl=1028)
  @1510aea0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1
  @1510b1b0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 10
  @1510b4f8 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 100
  @151079a0 14 REALSXP g0c1 [REF(2)] (len=1, tl=0) 1000

Then, the resulting table DT still has all columns in keep, the column z was renamed to y, the original column y is still available (but now y_original), and no data was copied in the process.