fwf_positions() should enforce col_names to be a character

[richierocks]

If the col_names argument to fwf_positions() is a factor, then the return value of that function has a column named col_names that is also a factor, which causes read_fwf() to throw the error

Error: `col_names` must be TRUE, FALSE or a character vector

Reproducible example

# 1. Create a tibble with col_names as a factor
library(readr)
start_posns <- c(1, 2, 4, 8)
end_posns <- c(1, 3, 7, 15)
col_names <- factor(letters[1:4])
fwf_col_info <- fwf_positions(start_posns, end_posns, col_names)

glimpse(fwf_col_info)
## Observations: 4
## Variables: 3
## $ begin     <dbl> 0, 1, 3, 7
## $ end       <dbl> 1, 3, 7, 15
## $ col_names <fct> a, b, c, d

# 2. Write some data to file
tfile <- tempfile()
writeLines(
  c("1AA1111aaaaaaaa", "2BB2222bbbbbbbb", "3CC3333cccccccc", "4DD4444dddddddd"),
  tfile
)

# 3. Try to import it
read_fwf(tfile, fwf_col_info)
## Error: `col_names` must be TRUE, FALSE or a character vector

Thoughts on fixes

Either fwf_positions() should convert factor col_names to character, or read_fwf() should allow factors and do the conversion itself, or both.

[jimhester]

This looks like a legitimate bug, but unfortunately the reproducible example is not actually reproducible.

The best way to ensure it is reproducible is to use the reprex package from GitHub devtools::install_github("tidyverse/reprex").

[richierocks]

@jimhester Whoops! Good spot. I've updated the comment to add the missing line.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/