Grouping key/value columns into single rows

Question

I'm trying to take key-value combinations and put all the values on the same row as the keys. I'm pretty sure I knew how to do this at one point (I think with data.table) and I've been looking at the usual suspects reshape2, tidyr, data.table, etc, but I can't seem to figure out a simple solution.

key1 = c(1,1,1,1,2,2,2,2)
key2 = c("A","A","B","B","C","C","D","D")
value = c("a","b","c","d","e","f","g","h")
kvframe = data.frame(key1,key2,value)

#  key1 key2 value
#1    1    A     a
#2    1    A     b
#3    1    B     c
#4    1    B     d
#5    2    C     e
#6    2    C     f
#7    2    D     g
#8    2    D     h

Here's what I would like the table to look like:

# key1 key2 value1 value2
#    1    A      a      b
#    1    B      c      d
#    2    C      e      f
#    2    D      g      h

Most of the key1,key2 pairs have the same number of corresponding values, but not all of them do. I'm hoping for a solution with the number of value columns equal to the max number of values for any given key set, where any pairs with fewer values are filled in with NA.

Answer 1

You need a sequence column for the group 'key1/key2'.

library(data.table) # v1.9.5+
setDT(kvframe)[, Seq := paste0('value', 1:.N), by = .(key1, key2)] # generate Seq
dcast(kvframe, key1 + key2  ~Seq, value.var = 'value') # cast from long to wide

#   key1 key2 value1 value2
#1:    1    A      a      b
#2:    1    B      c      d
#3:    2    C      e      f
#4:    2    D      g      h

Or using reshape from base R

 d1 <- transform(kvframe, Seq=ave(seq_along(value),
              key1, key2, FUN=seq_along))
 reshape(d1, idvar=c('key1', 'key2'), timevar='Seq', direction='wide')
 #  key1 key2 value.1 value.2
 #1    1    A       a       b
 #3    1    B       c       d
 #5    2    C       e       f
 #7    2    D       g       h

Or

library(tidyr)
spread(d1, Seq, value)