Wierd behaviour on backslashes in literals

[mvanbrab]

Found on version 5.0.0 and 4.15.0:

Single backslashes are eaten, double backslashes are output as double backslashes, triple backslashes are output as double backslahes.

A testcase is provided in attached file.

Remark: version 4.12.0 always output twice the number of backslashes in the input.

issue.zip

[DylanVanAssche]

Hi @mvanbrab !

Do you get the same behavior if the data is in JSON?

[mvanbrab]

Well no, but in equivalent JSON input I have to provide two backslashes where I mean one, so this is the input then and the output is OK (also contains the same amount of backslashes as the input), but that seems like a no-brainer to me...

[
  {
    "id": "1",
    "description": "One backslash:     \\."
  },
  {
    "id": "2",
    "description": "Two backslashes:   \\\\."
  },
  {
    "id": "3",
    "description": "Three backslashes: \\\\\\"
  },
  {
    "id": "4",
    "description": "A backslash before the '{': a ∈ ℝ⁺₀\\{1}."
  }
]```

[DylanVanAssche]

Aha! That confirms my hypothesis: a few versions ago, we switched from Apache CSV to Open CSV to parse CSV files. That library is probably eating the \ characters for lunch.

[DylanVanAssche]

It seems that this is a common problem: https://dzone.com/articles/properly-handling-backslashes-using-opencsv

[DylanVanAssche]

Tested the solution from dzone.com. It works for this case but breaks other edge cases.