spread does not appear to work with a `value` parameter of type factor about tidyr

[sjackman]

In the following example, spread.integer has the expected shape—three columns y, a and b—but spread.factor does not have the expected shape. It has instead two columns, and the second column is a matrix.

> str(spread.integer)
'data.frame':   2 obs. of  3 variables:
 $ y: Factor w/ 2 levels "c","d": 1 2
 $ a: int  1 2
 $ b: int  3 4
> str(spread.factor)
'data.frame':   2 obs. of  2 variables:
 $ y      : Factor w/ 2 levels "c","d": 1 2
 $ ordered: factor [1:2, 1:2] w x y z
  ..- attr(*, "levels")= chr  "w" "x" "y" "z"
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr  "a" "b"

[hadley]

Slightly simpler MRE

library(dplyr)
library(tidyr)

data <- data.frame(x = c("a", "a", "b", "b"), y = c("c", "d", "c", "d"), z = c("w", "x", "y", "z"))
data %>% spread(x, z) %>% str()
data %>% mutate(z = as.integer(z)) %>% spread(x, z) %>% str()

[hadley]

Root cause is that as.data.frame() does not work nicely on 2-d factors:

x <- factor(letters[1:4])
dim(x) <- c(2, 2)
as.data.frame(x)

[hadley]

Better now, although the output columns are characters. I'm not sure if this reasonable, or if each column should be a factor with the same levels.

[sjackman]

Thanks, Hadley. Better would be that the output columns are factors with the same levels as the input.