fread crash when using fill=TRUE option in a file with bad quotes

[AndeolEvain]

fread crashes when trying to call:

data.table::fread("testDoc.csv",
                         sep = ";",
                         quote = "\"",
                         header = TRUE,
                         fill = TRUE)

on this file, that I reduced to a minimal example: testDoc.csv

It seems to be due to the bad quotes in the last line, combined with fill = TRUE.
If the file is shorter, fread just returns an error without crashing, which is expected and not a problem.

The crash is consistently reproduced for me. I'm on windows 11, and tried it with data.table 1.14.0, as well as the latest 1.14.8, both on R.3.6.3 and R.4.2.2

I think it might be related to the bug report: #5110

[ben-schwen]

Could you specify what you mean by "crash"? Does it segfault or simply stop reading the file?

[sysilviakim]

I can report a similar problem.

• Segfault.
• If quote = "", it will solve the problem for the above reproducible example. For my example (not yet attached), it outputs a message such as

Stopped early on line 246469. Expected 39 fields but found 40. Consider fill=TRUE and comment.char=.

which is slightly odd error message to begin with, because fill = TRUE is already specified, as well as the fact that comment.char is no longer an argument of fread?

What's maddening about my case is that I can't actually find a bad quote in the said line, so I'm still trying to create a minimal reproducible example...

In any case, readr::read_csv didn't have this problem if this information benefits anyone.

[AndeolEvain]

Hi, sorry for the late reply, I had missed the notification.
Yes, it's an actual crash, the R process is killed (so segfault, or something equivalent). It doesn't just freeze or return.

[ben-schwen]

I can't reproduce it on the current dev. You can update to current dev by using data.table::update_dev_pkg()

[AndeolEvain]

Thanks! That worked.
A bit more details for people stumbling upon this in the future: using data.table::update_dev_pkg() got me to dev version 1.14.9. Now, using the same minimal example, I get the (very justified) warning:

"Warning message:
In data.table::fread("testDoc.csv", sep = ";", quote = "\"", header = TRUE,  :
  Found and resolved improper quoting out-of-sample. First healed line 102: <<"860003110";"S.A.R.L. \\\"\"LA GIBAUDERIE\\"\"\\"\"\"\"";"CLINIQUE";"Nouvelle-Aquitaine";NA;NA;NA;0;0;0;0;0;0;0;0>>. If the fields are not quoted (e.g. field separator does not appear within any field), try quote="" to avoid this warning."

And the dataframe is correctly read.

[ben-schwen]

Closing since this was fixed by #5046