Question

Edit: This problem was due to an outdated version of data.table being installed.

I have a data.table like the following:

require(xts)
a <- data.table(colour=c("Red","Green","Blue","Blue","Black","Black"), date=c(as.Date("2011-07-04"),as.Date("2011-07-10"),as.Date("2011-07-09"),as.Date("2011-07-12"),as.Date("2011-07-04"),as.Date("2011-07-09")),daily.quantity=c(1,-1,2,-2,1,1))

     colour       date daily.quantity
[1,]    Red 2011-07-04              1
[2,]  Green 2011-07-10             -1
[3,]   Blue 2011-07-09              2
[4,]   Blue 2011-07-12             -2
[5,]  Black 2011-07-04              1
[6,]  Black 2011-07-09              1

I would like cumulative totals per colour, to look like this:

     colour       date daily.quantity cumulative.quantity
[1,]  Black 2011-07-04              1                   1
[2,]  Black 2011-07-09              1                   2
[3,]   Blue 2011-07-09              2                   2
[4,]   Blue 2011-07-12             -2                   0
[5,]  Green 2011-07-10             -1                  -1
[6,]    Red 2011-07-04              1                   1

However, if I try the following, I end up with cumulative totals that don’t take colour into account:

setkey(a,colour,date)
a[,cumulative.quantity := cumsum(daily.quantity)]

     colour       date daily.quantity cumulative.quantity
[1,]  Black 2011-07-04              1                   1
[2,]  Black 2011-07-09              1                   2
[3,]   Blue 2011-07-09              2                   4
[4,]   Blue 2011-07-12             -2                   2
[5,]  Green 2011-07-10             -1                   1
[6,]    Red 2011-07-04              1                   2

I tried the obvious, but unfortunately unimplemented:

> a[,cumulative.quantity := cumsum(daily.quantity),keyby="colour,date"]
Error in `[.data.table`(a, , `:=`(cumulative.quantity, cumsum(daily.quantity)),  : 
  Combining := in j with by is not yet implemented. Please let maintainer('data.table') know if you are interested in this.

So, can anyone suggest a workaround for this?

Answer 1

The by grouping should only be on colour:

a[,cumulative.quantity := cumsum(daily.quantity), by=colour]
   colour       date daily.quantity cumulative.quantity
1:  Black 2011-07-04              1                   1
2:  Black 2011-07-09              1                   2
3:   Blue 2011-07-09              2                   2
4:   Blue 2011-07-12             -2                   0
5:  Green 2011-07-10             -1                  -1
6:    Red 2011-07-04              1                   1