How to select the columns by the content in another column in data.table of R?

Question

I have a data.table like this:

> dt<-data.table(F=rep(c('a','b','c'),each=2), fix=c(10:15),a=c(1:6), b=c(2:7),c=c(3:8))
> dt
   F fix a b c
1: a  10 1 2 3
2: a  11 2 3 4
3: b  12 3 4 5
4: b  13 4 5 6
5: c  14 5 6 7
6: c  15 6 7 8

The column F stores the column to be extract for each row. So for the first 2 records column, the column 'a' will be extract while the column 'b' will be extracted for the 3 and 4 record. The 'fix' will be kept for each record. The ideal output should be like this:

   F  fix new
1: a  10   1 
2: a  11   2
3: b  12   4
4: b  13   5
5: c  14   7
6: c  15   8

Answer 1

You can use .BY to extract the column from .SD

dt[, .(fix = fix, new = .SD[[.BY$F]]), by = F]
##    F fix new
## 1: a  10   1
## 2: a  11   2
## 3: b  12   4
## 4: b  13   5
## 5: c  14   7
## 6: c  15   8

Answer 2

You could also do

 melt(dt, id.var=c('F', 'fix'), value.name='new')[F==variable][, 
                         variable:=NULL][]
 #   F fix    new
 #1: a  10     1
 #2: a  11     2
 #3: b  12     4
 #4: b  13     5
 #5: c  14     7
 #6: c  15     8

Answer 3

One possibility is using lapply to each element of the new column:

dt$new = unlist(lapply(1:nrow(dt), function(ii){ return(dt[ii, dt$F[ii]])}))

Answer 4

If you convert this to a data.frame, you can use the following:

> x <- as.data.frame(dt)
> within(x, new <- x[cbind(1:6, match(x$F, names(x)))])[,c(1,2,6)]
  F fix new
1 a  10   1
2 a  11   2
3 b  12   4
4 b  13   5
5 c  14   7
6 c  15   8