Bug: duplicate error lines after upgrading to v3.0

[BrendonPierson]

I had a unit test checking the proper storing of errors that broken post upgrade from 2.5 to 3.0. The test tries to parse a file with known stray quote errors on line 2 and 4:

Jerry,31,comedian
Elaine,28,"unknwon ""
Kramer,40,kramerica
George,34,architect "vandalay" ind
Newman,37,"postal, worker"

Running the following I expect two errors (row 2 and 4) and 3 good rows (row 1,3,5).

File.stream!(three_good_two_bad.csv) |> CSV.decode |> Enum.map(& &1)

Instead I get 3 errors:

[
  ok: ["Jerry", "31", "comedian"],
  error: "Stray escape character on line 4:\n\n\"\nKramer,40,kramerica\nGeorge,34,architect \"vandalay\" ind\n\nThis error often happens when the wrong separator or escape character has been applied.\n",
  ok: ["Kramer", "40", "kramerica"],
  error: "Stray escape character on line 4:\n\narchitect \"vandalay\" ind\n\nThis error often happens when the wrong separator or escape character has been applied.\n",
  error: "Stray escape character on line 5:\n\narchitect \"vandalay\" ind\n\nThis error often happens when the wrong separator or escape character has been applied.\n",
  ok: ["Newman", "37", "postal, worker"]
]

I wouldn't expect multiple error tuples for a single row, but also the error tuple messages list lines 4, 4, and 5 when I would expect the errors to be listed on lines 2 and 4.

Thanks for the help in advance!

[beatrichartz]

Thanks for raising this, reparsing included the last line in this scenario. It is fixed in 3.0.3

As an aside, the first error is indicated as line 4 because line 2 starts a valid escape sequence and also escapes the quote at the end correctly by doubling it. The stray escape character happens at "vandalay in the sequence.

[BrendonPierson]

Ah that makes sense on the line number. Thanks for tackling this so fast!