problems reading from .zip but not from decompressed file

[benzipperer]

Hi, thank you for the excellent package! I am having some problems with expected != actual columns after reading a .csv file. I've narrowed it down to only experiencing the problems reading the original compressed .zip archive of the file, rather than the manually decompressed .csv file.

library(vroom)

# download zipped data
url <- "https://data.bls.gov/cew/data/files/2017/csv/2017_qtrly_singlefile.zip"
temp_dir <- tempdir()
data_zip <- tempfile(tmpdir = temp_dir, fileext = ".zip")
download.file(url, data_zip)

# decompress data
csv_basename <- unzip(data_zip, list=TRUE)$Name[1]
unzip(data_zip, files = csv_basename, exdir = temp_dir, overwrite=TRUE)
data_csv <- file.path(temp_dir, csv_basename)

# confirm problems with zip file only
vroom_from_zip <- vroom(data_zip)
problems(vroom_from_zip)
# A tibble: 1,058 x 5
     row   col expected   actual     file 
   <int> <int> <chr>      <chr>      <chr>
 1     2    33 42 columns 33 columns ""   
 2     2    33 42 columns 33 columns ""   
 3     2    33 42 columns 33 columns ""   
 4     2    33 42 columns 33 columns ""   
 5     2    33 42 columns 33 columns ""   
 6     2    33 42 columns 33 columns ""   
 7     2    33 42 columns 33 columns ""   
 8     2    33 42 columns 33 columns ""   
 9     2    33 42 columns 33 columns ""   
10     2    33 42 columns 33 columns ""   
# … with 1,048 more rows

vroom_from_csv <- vroom(data_csv)
problems(vroom_from_csv)
# A tibble: 0 x 5
# … with 5 variables: row <int>, col <int>, expected <chr>,
#   actual <chr>, file <chr>

Do you have any insights?

[jimhester]

Thank you for opening the issue and for supplying a reproducible example, it is a big help!

This was a recent regression when reading files from connections with windows line endings when the bytes of the line ending spanned two different connection buffers.

It should now be fixed.