Allow overriding the Reline encoding

[sjanusz-r7]

Description

Currently the Reline library uses the Ruby's default encoding, but this is not configurable:

reline/lib/reline/io/ansi.rb

Lines 43 to 45 in 7534f7f

	def encoding
	Encoding.default_external
	end

It would be great to configure this value, i.e. in the scenario of having the encoding set to ASCII-8BIT globally - but wanting Reline to use UTF-8 encoding:

  ::Encoding.default_external = ::Encoding::ASCII_8BIT
  ::Encoding.default_internal = ::Encoding::ASCII_8BIT

  # Workaround to fix encoding crash:
  # class ::Reline::ANSI
  #   def encoding
  #     ::Encoding::UTF_8
  #   end
  # end

  ascii_text = Reline.readline('prompt > ', false)

  $stderr.puts ascii_text.encoding.to_s

Example

Pasting in the characters こんにちは

prompt > こんにちは

Current Behaviour

Crash: gems/reline-0.5.10/lib/reline/unicode.rb:112:in 'encode': "\\xE3" from ASCII-8BIT to UTF-8 (Encoding::UndefinedConversionError)

Desired Behaviour

Being able to configure the value on Reline:

  def encoding
    @encoding || Encoding.default_external
  end

Terminal Emulator

OSX

Version

0.5.10

[tompng]

The crash is fixed in #751

prompt > ???????????????
ASCII-8BIT

When Encoding.default_external is set to ascii-8bit, Reline thinks terminal emulator's encoding is ascii-8bit. Any bytes (not restricted to utf-8) could come to STDIN and all Reline should encode output to ascii-8bit before printing them to STDOUT, but "\x7f".."\xff" are not printable character.
If the assumption (terminal emulator's encoding is ascii-8bit) is wrong, I think changing Encoding.default_external is the right way.

If your request is to have an ascii-8bit mode like this, please open another issue.

ruby -Eascii-8bit
Pasting in the characters あ
prompt > \xE3\x81\x82
Delete key will delete single byte
prompt > \xE3\x81

[sjanusz-r7]

Hello! Thanks for getting back to us.

For context, for our project we set the default_encoding to be ASCII-8BIT as we deal with a lot of binary data blobs by default - and rarely UTF-8. However we want to configure Reline separately to use UTF-8.

For our scenario it would be great if Reline would either use the IO object's encoding or would allow for separate configuration to UTF-8, maybe something like:

Reline.encoding = Encoding::UTF_8

If the assumption (terminal emulator's encoding is ascii-8bit) is wrong, I think changing Encoding.default_external is the right way.

Hopefully the above context clarifies why we can't globally change Encoding.default_external to UTF-8 to use Reline, as that would break our other network protocol handling in our project

[tompng]

Thank you for the detailed scenario

Reline would either use the IO object's encoding or ...

The idea using IO object's encoding looks good.