Question

I went back to some R code that I wrote last month, but it appears that the version of reshape (0.8.4, R 2.15.2) I am using has changed that functionality.

Here is a sample:

> library(reshape)
> so.test <- data.frame(
                       one = as.character(rnorm(750)),
                       two = as.character(rnorm(750)),
                       three = as.character(rnorm(750)), four = as.character(rnorm(750)))
> check <- melt(so.test)

Using one, two, three, four as id variables

This gives a data.frame equal to the original:

> table(so.test == check)

TRUE 
3000 

I have also tried this with reshape2::melt but I am getting the same result. Interestingly, the melt() function works as expected with a data.frame with numeric values:

> so.test2 <- data.frame(
                      one = rnorm(750),
                      two = rnorm(750),
                      three = rnorm(750), four = rnorm(750))
> check2 <- melt(so.test2)
Using  as id variables
> head(check2)
  variable      value
1      one  0.2471168
2      one -0.0663480
3      one -0.0251867
4      one  2.8786207
5      one -0.2586785
6      one -0.7508927

Answer 1

The documentation for both the reshape and the reshape2 version say:

If you supply neither, melt will assume factor and character variables are id variables, and all others are measured.

So melt is behaving as documented.