new version breaks pivot_wider

[dcaud]

Latest version of tidyr returns error about not being able to subset names that don't exist.

Reverting to ver. 1.1.4 makes the code work again.

I've no time to create a reprex, but here's essentially what breaks:

temp.pivot <- pivot_wider(
   output.df,
   id_cols = "boom",
   values_from = c(paste0(r,"some_suffix_name")), 
   names_from = "boom"
 )

[DavisVaughan]

We are going to need more information to be able to help you. Please provide a reprex.

[dcaud]

I'll try to provide a reprex. In the meantime, were there any breaking changes with the new version of tidyr?

[dcaud]

Here's a reprex:

require(devtools)
install_version("tidyr", version = "1.1.4", repos = "http://cran.us.r-project.org")

This works:

tidyr::pivot_wider(
  iris,
  id_cols = "Species",
  values_from = Sepal.Length, 
  names_from = "Species"
)

restart R

require(devtools)
install_version("tidyr", version = "1.2.0", repos = "http://cran.us.r-project.org")

This gives an error

tidyr::pivot_wider(
  iris,
  id_cols = "Species",
  values_from = Sepal.Length, 
  names_from = "Species"
)

[tjmahr]

I had this same error. Basically, I had used a column name in both id_cols and in names_from. That worked in 1.1.3 but now fails in 1.2.0. I fixed my code by removing the column name from id_cols.

[DavisVaughan]

Okay yea this was on purpose. Previously, when you duplicate a column name in both id_cols and names_from, it was getting used by names_from but silently dropped from the id_cols. This is almost always a typo on the user's part, you can't use the same name in both arguments.

So now id_cols is restricted to the subset of data after both names_from and values_from have been removed, which is why you get an error. This matches the default behavior if you don't supply id_cols at all, you get all the "remaining" columns as id_cols, so it makes sense that that should be all you have available to select from as well.

library(tidyr)

df <- tibble(
  names = c("x", "y"),
  value = c(1, 2)
)
df
#> # A tibble: 2 × 2
#>   names value
#>   <chr> <dbl>
#> 1 x         1
#> 2 y         2

# before
pivot_wider(
  df,
  id_cols = names, # <- this ends up getting ignored silently
  values_from = value, 
  names_from = names # <- this gets used
)
#> # A tibble: 1 × 2
#>       x     y
#>   <dbl> <dbl>
#> 1     1     2

# after
pivot_wider(
  df,
  id_cols = names, 
  values_from = value, 
  names_from = names
)
#> Error in `chr_as_locations()`:
#> ! Can't subset columns that don't exist.
#> x Column `names` doesn't exist.

I'll add an extra note to id_cols saying that if you supply a tidy selection, it is evaluated on data after removing names_from and values_from (it already mentions the default behavior, so this is just one step on top of that)