R session crashes depending on the size of guess_max

[DavisVaughan]

Extracted from tidyverse/dplyr#5977

For some reason, the first call to vroom() doesn't crash, but the second does (with a larger guess_max). I'm not sure why this user is splitting by \n since the file uses ; delimiters, but I think they just wanted it to be read in as a single column of text data that they would post process manually.

I assume the problem is not really with guess_max, but somehow vroom is hitting something between lines 101-1000 that it can't handle for some reason.

library(vroom)

url <- "https://www.data.gouv.fr/fr/datasets/r/936f6d38-5969-46e5-8b9d-c7646d6390ec"
tf <- tempfile(fileext = ".txt")
download.file(url, tf)

# no crash - even on repeated calls
df <- vroom(
  file = tf, 
  delim = "\n", 
  col_names = FALSE, 
  locale = locale(encoding = "ISO-8859-1")
)

# crashes pretty reliably
df <- vroom(
  file = tf, 
  delim = "\n", 
  col_names = FALSE, 
  locale = locale(encoding = "ISO-8859-1"),
  guess_max = 1000L
)