spread when id column has names

[wangyuchen]

tidyr/R/id.R

Line 41 in cbdd14e

if (!is_null(attr(x, "n")) && !drop) return(x)

I think you'd want to add exact = TRUE here to match attribute "n" exactly.

Otherwise, when drop = FALSE, the empty attribute condition gets used and it will partial match all attributes. It happened that I had a named vector in my data frame and it was passed into id_var(). It matched the "names" attribute and returned the variable without adding the "n" attribute.

[hadley]

Can you please provide a minimal reprex (reproducible example)? That will help us create a unit test to make sure we fix the bug.

[wangyuchen]

Yes of course. Please see if this works.

library(tidyr)
#> Warning: package 'tidyr' was built under R version 3.4.4
library(dplyr)
#> Warning: package 'dplyr' was built under R version 3.4.4
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

# named vector
id_col <- c(x = 1, y = 2, z = 3)

# spread key contains levels not included in data
# need to use drop = FALSE when spread
spread_df <- 
  data.frame(key = factor(1:3, 1:5, letters[1:5]),
             out = 1:3,
             id = id_col)  # name attribute will be dropped
spread_df
#>   key out id
#> x   a   1  1
#> y   b   2  2
#> z   c   3  3

# both work fine
spread_df %>%  
  spread(key, out, drop = TRUE) 
#>   id  a  b  c
#> 1  1  1 NA NA
#> 2  2 NA  2 NA
#> 3  3 NA NA  3

spread_df %>%  
  spread(key, out, drop = FALSE) 
#>   id  a  b  c  d  e
#> 1  1  1 NA NA NA NA
#> 2  2 NA  2 NA NA NA
#> 3  3 NA NA  3 NA NA


spread_df2 <- 
  data.frame(key = factor(1:3, 1:5, letters[1:5]),
             out = 1:3) %>% 
  mutate(id = id_col)  # the name attribute will be preserved in mutate

spread_df2 %>%  
  spread(key, out, drop = TRUE) 
#>   id  a  b  c
#> 1  1  1 NA NA
#> 2  2 NA  2 NA
#> 3  3 NA NA  3

spread_df2 %>%  
  spread(key, out, drop = FALSE) 
#> Error: Result 1 is not a length 1 atomic vector

^{Created on 2019-01-04 by the reprex package (v0.2.1)}