ラウンドトリップが保障されない文字を調べます。

diff.rb

def chr_unsafe(cp, enc)
  [cp.to_s(16)].pack("H*").force_encoding(enc)
end

enc = 'cp932'
count = 0
count2 = 0

(0x8000..0xffff).each do |cp|
  begin
    c = cp.chr(enc)
  rescue RangeError => ex
    count += 1
    unsafe = chr_unsafe(cp, enc)

    if unsafe.valid_encoding?
      print cp.to_s(16)
      print ' '
      count2 += 1
    end
  end
end

puts
puts count.to_s(10) + ' RangeError'
puts count2.to_s(10) + ' RangeError + valid_encoding?'

encoding.list.rb

Encoding.list.each do |enc|
  begin
    0x61.chr(enc).encode(enc)
    check(enc)
  rescue Encoding::ConverterNotFoundError => ex
    next
  end
end

result.txt

Big5
{:unsafe=>10, :error=>896}
Big5-HKSCS
{:unsafe=>10, :error=>165}
CP932
{:unsafe=>424, :error=>1777}
CP949
{:unsafe=>362, :error=>5641}

round-trip-safety.rb

def check(enc, max=0xFFFF)

  replace = "?"
  opt = {
    :invalid => :replace,
    :undef   => :replace,
    :replace => replace
  }

  unsafe = 0
  error = 0

  (0..max).each do |cp|

    c = [cp.to_s(16)].pack("H*").force_encoding(enc)

    unless c.valid_encoding?
      next
    end

    ret = c.encode("utf-8", opt).encode(enc, opt)

    if ret != c then
      if ret == replace then
        error += 1
      else
       unsafe += 1
      end
    end
  end

  puts enc
  puts "unsafe: #{unsafe}, error: #{error}"
end

list = [
  # Traditional Chinese
  "Big5",
  "Big5-HKSCS",
  # Japanese
  "CP932",
  # Korean
  "CP949"
]

list.each do |enc|
  check(enc)
end

round-trip-safety2.rb

def check(enc)
  replace = "?"
  opt = {
    :invalid => :replace,
    :undef   => :replace,
    :replace => replace
  }

  count = 0

  (0x8000..0xffff).each do |cp|
    begin
      c = cp.chr(enc)
    rescue RangeError => ex
      next
    end

    ret = c.encode('utf-8', opt).encode(enc, opt)

    if ret == c || ret == '?'
      next
    end

    #print "[#{cp.to_s(16)}: #{ret.unpack('H*').first}] "
    count += 1
  end

  puts enc
  puts "round-trip unsafe: #{count.to_s(10)}"
end


list = [
  # Traditional Chinese
  "Big5",
  # Japanese
  "CP932"
]

list.each do |enc|
  check(enc)
end

utils.rb

def chr_unsafe(cp, enc)
  [cp.to_s(16)].pack("H*").force_encoding(enc)
end

def valid_codepoint?(cp, enc)
  [cp.to_s(16)].pack("H*").force_encoding(enc).valid_encoding?
end