Add multiple columns to R data.table in one function call?
I have a function that returns two values in a list. Both values need to be added to a data.table in two new columns. Evaluation of the function is costly, so I would like to avoid having to compute the function twice. Here's the example:
library(data.table)
example(data.table)
DT
x y v
1: a 1 42
2: a 3 42
3: a 6 42
4: b 1 4
5: b 3 5
6: b 6 6
7: c 1 7
8: c 3 8
9: c 6 9
Here's an example of my function. Remember I said it's costly compute, on top of that there is no way to deduce one return value from the other given values (as in the example below):
myfun <- function (y, v)
{
ret1 = y + v
ret2 = y - v
return(list(r1 = ret1, r2 = ret2))
}
Here's my way to add two columns in one statement. That one needs to call myfun twice, however:
DT[,new1:=myfun(y,v)$r1][,new2:=myfun(y,v)$r2]
x y v new1 new2
1: a 1 42 43 -41
2: a 3 42 45 -39
3: a 6 42 48 -36
4: b 1 4 5 -3
5: b 3 5 8 -2
6: b 6 6 12 0
7: c 1 7 8 -6
8: c 3 8 11 -5
9: c 6 9 15 -3
Any suggestions on how to do this? I could save r2 in a separate environment each time I call myfun, I just need a way to add two columns by reference at a time.
You could store the output of your function call:
z <- myfun(DT$y,DT$v)
head(DT[,new1:=z$r1][,new2:=z$r2])
# x y v new1 new2
# [1,] a 1 42 43 -41
# [2,] a 3 42 45 -39
# [3,] a 6 42 48 -36
# [4,] b 1 4 5 -3
# [5,] b 3 5 8 -2
# [6,] b 6 6 12 0
but this also seems to work:
DT[, c("new1","new2") := myfun(y,v), with = FALSE]
New in data.table v1.8.3 on R-Forge, the with = FALSE is no longer needed here, for convenience :
DT[, c("new1","new2") := myfun(y,v)]
Up to the minute live NEWS is here.
To build on the previous answer, one can use lapply with a function that output more than one column. It's is then possible to use the function with more columns of the data.table.
myfun <- function(a,b){
res1 <- a+b
res2 <- a-b
list(res1,res2)
}
DT <- data.table(z=1:10,x=seq(3,30,3),t=seq(4,40,4))
DT
## DT
## z x t
## 1: 1 3 4
## 2: 2 6 8
## 3: 3 9 12
## 4: 4 12 16
## 5: 5 15 20
## 6: 6 18 24
## 7: 7 21 28
## 8: 8 24 32
## 9: 9 27 36
## 10: 10 30 40
col <- colnames(DT)
DT[, paste0(c('r1','r2'),rep(col,each=2)):=unlist(lapply(.SD,myfun,z),
recursive=FALSE),.SDcols=col]
## > DT
## z x t r1z r2z r1x r2x r1t r2t
## 1: 1 3 4 2 0 4 2 5 3
## 2: 2 6 8 4 0 8 4 10 6
## 3: 3 9 12 6 0 12 6 15 9
## 4: 4 12 16 8 0 16 8 20 12
## 5: 5 15 20 10 0 20 10 25 15
## 6: 6 18 24 12 0 24 12 30 18
## 7: 7 21 28 14 0 28 14 35 21
## 8: 8 24 32 16 0 32 16 40 24
## 9: 9 27 36 18 0 36 18 45 27
## 10: 10 30 40 20 0 40 20 50 30
The answer can not be used such as when the function is not vectorized.
For example in the following situation it will not work as intended:
myfun <- function (y, v, g)
{
ret1 = y + v + length(g)
ret2 = y - v + length(g)
return(list(r1 = ret1, r2 = ret2))
}
DT
# v y g
# 1: 1 1 1
# 2: 1 3 4,2
# 3: 1 6 9,8,6
DT[,c("new1","new2"):=myfun(y,v,g)]
DT
# v y g new1 new2
# 1: 1 1 1 5 3
# 2: 1 3 4,2 7 5
# 3: 1 6 9,8,6 10 8
It will always add the size of column g , not the size of each vector in g
A solution in such case is:
DT[, c("new1","new2") := data.table(t(mapply(myfun,y,v,g)))]
DT
# v y g new1 new2
# 1: 1 1 1 3 1
# 2: 1 3 4,2 6 4
# 3: 1 6 9,8,6 10 8