How can I reshape a data.table (long into wide) without doing a function like sum or mean?

Question

How can I reshape a data.table (long into wide) without doing a function like sum or mean? I was looking at dcast/melt/reshape/etc. But I don't get the desired results.

This is my data:

DT <- data.table(id = c("1","1","2","3"), score = c("5", "4", "5", "6"))

Original format:

> DT
id score
1  5 
1  4 
2  5 
3  6 

Desired format:

id score1 score2
1  5      4
2  5      NA
3  6      NA 

I now do the trick with:

DT <- DT[, list(list(score)), by=id]

But then the contents of the first cell is like:

c("5", "4")

And I need to split it (I use the package splitstackshape):

DT <- cSplit(DT, "V1", ",")

This is probably not the most efficient method... What is a better way?

Answer 1

You can use getanID to create a unique .id for the grouping variable id. Then, try with dcast.data.table (or simply dcast from versions 1.9.5 and beyond) and if needed change the column names using setnames

 library(splitstackshape)
 res <- dcast(getanID(DT, 'id'), id~.id,value.var='score')
 setnames(res, 2:3, paste0('score', 1:2))[]
 #    id score1 score2
 #1:  1      5      4
 #2:  2      5     NA
 #3:  3      6     NA

Or using only data.table

 dcast(DT[, .id:=paste0('score', 1:.N), by=id],
       id~.id, value.var='score')
 #   id score1 score2
 #1:  1      5      4
 #2:  2      5     NA
 #3:  3      6     NA

Or from the code you were using (less number of characters)

cSplit(DT[, toString(score), by=id], 'V1', ',')
#   id V1_1 V1_2
#1:  1    5    4
#2:  2    5   NA
#3:  3    6   NA