Reshaping data using the data.table package

Question

I have asked the same question a few days ago ( click here), but didn\'t mention that a result using data.table would be appreciated

The \"aggregate-soluti

Answer 1

Going to wide format is a little awkward currently in data.table, but I think this works:

library(data.table)
dt = data.table(x=c("p1","p1","p2"),y=c("a","b","a"),z=c(14,14,16))

setkey(dt, x, y)
dt[CJ(unique(x), unique(y)), list(.N, z)][,
   setNames(as.list(c(N, z[!is.na(z)][1])), c(y, 'z')), by = x]
#    x a b  z
#1: p1 1 1 14
#2: p2 1 0 16

The CJ part joins by all combinations of unique x and y, and then in that join there is a hidden by-without-by that's used to compute counts via .N. Once you have those it's just a matter of putting them horizontally for each x together with any non-NA z (I chose the first) and that's accomplished using as.list. Finally setNames sets the column names correctly.