CODEWITHSUNDEEP
rsortingdata.table
Cath
Cath

Reputation: 24074

Reorder rows in data.table in a specific order

I have a data.table dumdt:

set.seed(123)
dumdt <- data.table(v1=sample(1:10, 5), v2=1:5)

whose rows I'd like to reorder in this specific indices order (so first the third observation, then the fifth, then second, etc.): 

to_ord <- c(3, 5, 2, 1, 4)

So I'd like dumdt to be the result of dumdt[to_ord] but I also would like to do it by reference and avoid doing dumdt <- dumdt[to_ord].

I know I can reorder rows by reference with setorder (or setorderv) but only according to one or several variables, in ascending or descending order, not in a customised order.
However, if I wanted to reorder the columns, not the rows, in a customised order, I could use setcolorder.

So here comes my question : is there a function that would work like setcolorder but with the rows (or a way to use setorder to do the same) ?

My desired output would be something like

setroworder(x=dumdt, neworder=to_ord)
dumdt
   # v1 v2
# 1:  4  3
# 2:  6  5
# 3:  8  2
# 4:  3  1
# 5:  7  4

Upvotes: 28

Views: 12969

Answers (2)

jangorecki
jangorecki

Reputation: 16697

This capability is not (yet) exported. After looking at the source of setorderv I was able to extract required call to C function which does what you need and supply it with custom order. 

library(data.table)
set.seed(123)
dumdt <- data.table(v1=sample(1:10, 5), v2=1:5)
print(dumdt)
#   v1 v2
#1:  3  1
#2:  8  2
#3:  4  3
#4:  7  4
#5:  6  5
setroworder <- function(x, neworder) {
    .Call(data.table:::Creorder, x, as.integer(neworder), PACKAGE = "data.table")
    invisible(x)
}
to_ord <- c(3, 5, 2, 1, 4)
setroworder(x=dumdt, neworder=to_ord)
print(dumdt)
#   v1 v2
#1:  4  3
#2:  6  5
#3:  8  2
#4:  3  1
#5:  7  4

Yet the solution proposed by Frank looks a little bit nicer.

Upvotes: 19

Frank
Frank

Reputation: 66819

If I understand correctly, you can just add a col and then order by it:

setorder(dumdt[, .r := order(to_ord)], .r)[, .r := NULL]

   v1 v2
1:  4  3
2:  6  5
3:  8  2
4:  3  1
5:  7  4

Upvotes: 25

Related Questions