melt could collapse to multiple variable/value columns

[arunsrinivasan]

As an experiment so far.

require(data.table)
DT = data.table(id=1:5, a=6:10, b=11:15, c=letters[1:5])

# expected result
ans
#     id variable1 value1 variable2 value2
#  1:  1         a      6         c      a
#  2:  2         a      7         c      b
#  3:  3         a      8         c      c
#  4:  4         a      9         c      d
#  5:  5         a     10         c      e
#  6:  1         b     11         c      a
#  7:  2         b     12         c      b
#  8:  3         b     13         c      c
#  9:  4         b     14         c      d 
#10:  5         b     15         c      e

Idea:

melt(DT, id.var="id", measure.var=list(c("a", "b"), "c"))

This is particularly useful when:

• all elements in measure.var list are of same length and the columns we've to melt is of different type (ex: integer and character). Currently it'll all be coerced to the same type depending on hierarchy into one value column.

Similarly, cast should also be able to operate on multiple columns so that unnecessary melting can be completely avoided. But that's another FR for next release (1.9.6) #739.

[arunsrinivasan]

Implemented here. Will add README and then close. More specifically commit 9833fe0.

[brodieG]

Arun, I think you need to add one more column to the example, otherwise it will look like something like:

library(reshape2)
melt(DT, id.vars=c("id", "c"))

will do the job. The tricky case is:

DT <- data.table(id=1:5, a=6:10, b=11:15, c=letters[1:5], d=letters[6:10])
melt(DT, id.var="id", measure.var=list(c("a", "b"), c("c", "d")))

Took me a minute to figure out why this was needed.

[arunsrinivasan]

@brodieG thanks. The examples in ?melt cover this quite well, I believe (feedbacks welcome). That was just for me to get a better picture. On your example, right. That was one of the main reasons I wanted to implement this.

[arunsrinivasan]

• Update documentation
• Need for variable.name = NULL? (In discussion with Ananda).