InvalidEncodingError reports line 1 when the invalid byte is in the row separator detection sample (File/IO input)

Aberta
#385 1 comentário 0 reações 0 responsáveis Ver no GitHub

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
3/5
Tempo estimado
1-2 dias
Facilidade para iniciantes
68/100
Tipo de issue
Bug
Clareza
Claramente especificada
Status de atividade
Ativa
Stack de tecnologia
ruby
Domínio
backend

Direção de pesquisa

Start in lib/csv/parser.rb at resolve_row_separator and build_scanner, then trace how sampled input reaches InputsScanner and raises InvalidEncodingError. Add a regression test for invalid bytes on a later line in File input, comparing its reported line number with String input. Done when both report the correct line and the relevant CSV tests pass.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

When the input is a File (or another IO) and an invalid byte sequence appears inside the sample read for row separator detection, CSV::InvalidEncodingError always reports line 1, even if the bad byte is on a later line. The same data given as a String or StringIO reports the correct line.

Reproduce
require "csv"
require "tempfile"

def try(label)
  yield
rescue CSV::MalformedCSVError => e
  puts "#{label}: #{e.class}: #{e.message} (line_number=#{e.line_number})"
end

small = "a,b\n\xFF,1\n".b
large = ("a,b\n\x80,1\n" + "x,y\n" * 16384).b  # 64 KiB, no 0xFF byte

[["small", small], ["large", large]].each do |name, data|
  Tempfile.create(["t", ".csv"]) do |f|
    f.binmode
    f.write(data)
    f.close
    try("#{name} String") { CSV.parse(data.dup.force_encoding("UTF-8")) }
    try("#{name} File  ") { CSV.read(f.path, encoding: "UTF-8") }
  end
end
Actual

csv 3.3.7 (Ruby 4.1.0dev) and current main:

small String: CSV::InvalidEncodingError: Invalid byte sequence in UTF-8 in line 2. (line_number=2)
small File  : CSV::InvalidEncodingError: Invalid byte sequence in UTF-8 in line 1. (line_number=1)
large String: CSV::InvalidEncodingError: Invalid byte sequence in UTF-8 in line 2. (line_number=2)
large File  : CSV::InvalidEncodingError: Invalid byte sequence in UTF-8 in line 1. (line_number=1)

CSV.foreach(path), CSV.new(File.open(path)) and CSV.new(io) for a pipe give the same result. I see the same with csv 3.1.9 (Ruby 3.0.0, MalformedCSVError), 3.2.8 (Ruby 3.3.0) and 3.3.5 (Ruby 4.0.0). If the bad byte is on line 3, the file case still says line 1.

Expected

line 2. / line_number=2 for the file input too, as for String/StringIO.

Cause

With row_sep: :auto, resolve_row_separator reads a sample with @input.gets(nil, chunk_size) (32 KiB for File, 1 KiB for other IOs):
https://github.com/ruby/csv/blob/v3.3.7/lib/csv/parser.rb#L703-L713

build_scanner computes the line number of an invalid byte only when the sample is the whole input:
https://github.com/ruby/csv/blob/v3.3.7/lib/csv/parser.rb#L886-L901

Otherwise the sample is wrapped in a StringIO for InputsScanner, whose constructor calls read_chunk, which raises InvalidEncoding for the whole sample before any row is parsed:
https://github.com/ruby/csv/blob/v3.3.7/lib/csv/parser.rb#L313-L316

@scanner is still nil at that point, so parse reports @lineno + 1, i.e. 1:
https://github.com/ruby/csv/blob/v3.3.7/lib/csv/parser.rb#L426-L433

So the line is wrong whenever the bad byte is in a sample that doesn't reach EOF (the "large" case above).

The "small" case hits this too, because IO#gets(nil, limit) stops right after a 0xFF byte. For rs = nil, rb_io_getline_0 passes newline = -1 to appendline, and search_delim does memchr(p, -1, len), which matches 0xFF. So the sample is only "a,b\n\xFF", not the whole file:
https://github.com/ruby/ruby/blob/v4.0.0/io.c#L4200
https://github.com/ruby/ruby/blob/v4.0.0/io.c#L3885

File.binwrite("t", "ab\xFFcd")
File.open("t", "rb") { |f| f.gets(nil, 100) } #=> "ab\xFF"

That part looks like a Ruby core issue (I can report it separately), but csv would still report the wrong line for large inputs even after it is fixed.

Possible fix

Run the same per-line check as the "whole input" branch on the samples before building InputsScanner:

         if string
-          unless string.valid_encoding?
-            index = string.lines(@row_separator).index do |line|
-              !line.valid_encoding?
-            end
-            if index
-              raise InvalidEncodingError.new(@encoding, @lineno + index + 1)
-            end
-          end
+          check_sample_encoding(string)
           Scanner.new(string)
         else
+          check_sample_encoding(@samples.join(""))
           inputs = @samples.collect do |sample|
             StringIO.new(sample)
           end
@@
+    def check_sample_encoding(sample)
+      return if sample.valid_encoding?
+      index = sample.lines(@row_separator).index do |line|
+        !line.valid_encoding?
+      end
+      if index
+        raise InvalidEncodingError.new(@encoding, @lineno + index + 1)
+      end
+    end

With this change both cases above report line 2, and the test suite passes with and without CSV_PARSER_SCANNER_TEST=yes. Like the existing string branch, it raises before yielding the rows that come before the bad line in the sample, which is what happens today anyway. I can open a PR with a test if this direction is OK.

🤖 Generated with Claude Code

Linguagem predominante
Ruby
Estrelas
220
Forks
139
Merge médio
1d 1h
PRs com merge (30d)
2

Preparar o ambiente

Este projeto não oferece contêiner de desenvolvimento, Dockerfile nem guia de contribuição, então a configuração fica por sua conta: comece pelo README e veja nosso guia da primeira contribuição para os passos gerais.

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de ruby/csv

Todas as issues de ruby/csv

Issues semelhantes

Mais issues de Ruby

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.