spread() with drop = FALSE applied to var column including NA values results an error

[ngr-t]

Below is an example data.frame.

> (df_with_na <- data.frame(var = factor(c(LETTERS[1:3], NA), LETTERS[1:4]), val = seq(4)))
   var val
1    A   1
2    B   2
3    C   3
4 <NA>   4
> df_with_na$var
[1] A    B    C    <NA>
Levels: A B C D

When I spread this df by the "var" column, call like

> tidyr::spread(df_with_na, var, val)
  A B C <NA>
1 1 2 3    4

In this case, I got an column "" originate from the row that contains NA in "var" column.
Then I added "drop = FALSE" option to keep all levels (including "D"), and got this:

> tidyr::spread(df_with_na, var, val, drop = FALSE)
Error in `colnames<-`(`*tmp*`, value = c("A", "B", "C", "D", "<NA>")) : 
  length of 'dimnames' [2] not equal to array extent

I think this is not an intended behavior and it'd better to return a data.frame with columns for all levels of "var" column and NA.

[hadley]

Could you please rework your reproducible example to use the reprex package ? That makes it easier to see both the input and the output, formatted in such a way that I can easily re-run in a local session

[ngr-t]

Is this alright?

# Prepare data.frame                                                                     
(df_with_na <- data.frame(var = factor(c(LETTERS[1:3], NA), LETTERS[1:4]), val = seq(4)))
#>    var val
#> 1    A   1
#> 2    B   2
#> 3    C   3
#> 4 <NA>   4
df_with_na$var                                                                           
#> [1] A    B    C    <NA>
#> Levels: A B C D
# No problem                                                                             
tidyr::spread(df_with_na, var, val)                                                      
#>   A B C <NA>
#> 1 1 2 3    4
# Bug?                                                                                   
tidyr::spread(df_with_na, var, val, drop = FALSE)                                        
#> Error in dimnames(x) <- dn: length of 'dimnames' [2] not equal to array extent