Upward navigation through completion results not working as expected

[mjgiarlo]

Description

Navigating through multi-line autocomplete results in IRB using the Tab key works as expected. Using S-Tab takes the current selection instead of navigating back up. I have S-Tab configured for menu-complete-backward in .inputrc (see below), and it works fine when I turn multi-line off and use readline (see irb_info pastes below).

When I look at https://github.com/ruby/reline/blob/master/lib/reline/line_editor.rb#L1137, the value of key.char comes through as the symbol :menu_complete_backward, which reline does not seem to know how to handle. If I re-map S-Tab from menu-complete-backward to completion-journey-up, upward navigation works in IRB but breaks upward/backward navigation in other command-line tools, e.g., file completion, likely because completion-journey-up does not seem to be defined or used beyond reline (as far as I can tell!).

Terminal Emulator, Shell, etc.

• Terminal: GNOME Terminal 3.49.92 using VTE 0.74.0 +BIDI +GNUTLS +ICU +SYSTEMD
• Shell: GNU bash, version 5.2.15(1)-release (x86_64-pc-linux-gnu)
• OS: Ubuntu 23.10
• irb_info:

  Ruby version: 3.2.2
  IRB version: irb 1.12.0 (2024-03-06)
  InputMethod: ReadlineInputMethod with ext/readline 8.2 and /home/mjg/.inputrc
  Completion: RegexpCompletor
  .irbrc paths: /home/mjg/.rvm/rubies/ruby-3.2.2/.irbrc, /home/mjg/.irbrc
  RUBY_PLATFORM: x86_64-linux
  LANG env: en_US.UTF-8
  East Asian Ambiguous Width: 1
  

• irb_info with readline (USE_MULTILINE=false):

  Ruby version: 3.2.2
  IRB version: irb 1.12.0 (2024-03-06)
  InputMethod: RelineInputMethod with Reline 0.5.1 and /home/mjg/.inputrc
  Completion: Autocomplete, RegexpCompletor
  .irbrc paths: /home/mjg/.rvm/rubies/ruby-3.2.2/.irbrc, /home/mjg/.irbrc
  RUBY_PLATFORM: x86_64-linux
  LANG env: en_US.UTF-8
  East Asian Ambiguous Width: 1
  

• .inputrc:

  set editing-mode emacs
  set show-all-if-ambiguous on
  set completion-ignore-case on
  set menu-complete-display-prefix on
  TAB: menu-complete
  "\e[Z": menu-complete-backward
  "\e[5~": history-search-backward
  "\e[6~": history-search-forward
  "\e[1;5C": forward-word
  "\e[1;5D": backward-word
  "\e[3;5~": kill-word
  

[mjgiarlo]

I'm not sure if this is what recommended, but if I tweak my .inputrc to include:

$if Reline
  "\e[Z": completion-journey-up
$else
  "\e[Z": menu-complete-backward
$endif

That allows upward navigation to work in IRB without borking e.g. bash file autocompletion. I'm still fuzzy on where "completion-journey-up" comes from or whether configuration like I pasted above should be necessary navigate results in both directions.

[ima1zumi]

The issue arises because Reline does not support the setting for menu-complete-backward. Although the setting cannot be configured, the functionality exists and seems to work with completion-journey-up. Here, the hyphen is converted to an underscore and is being called.

reline/lib/reline/line_editor.rb

Line 1137 in d348df9

elsif @config.editing_mode_is?(:emacs, :vi_insert) and key.char == :completion_journey_up

Reline should support configuring menu-complete-backward.

As a workaround, using $if Reline for branching, as you wrote, seems like a good idea. Please let me know if it doesn't work.

[mjgiarlo]

@ima1zumi 💬

...
Reline should support configuring menu-complete-backward.

I'd be happy to work on a PR for this to scratch my own itch, if that'd be valuable. It would help me to know if there are any similar examples in the codebase already for related configuration.

As a workaround, using $if Reline for branching, as you wrote, seems like a good idea. Please let me know if it doesn't work.

It does work, yes. 🙂

[ima1zumi]

@mjgiarlo

Thank you for your offer!

In this case, probably we need to turn this part into a method.

reline/lib/reline/line_editor.rb

Lines 1137 to 1142 in d348df9

	elsif @config.editing_mode_is?(:emacs, :vi_insert) and key.char == :completion_journey_up
	if not @config.disable_completion and @config.autocompletion
	process_insert(force: true)
	@completion_state = CompletionState::NORMAL
	completion_occurs = move_completed_list(:up)
	end

The reason is because Reline converts commands read from .inputrc from hyphens to underscores and then calls the method as a Symbol. For example, for beginning-of-line (C-a), beginning_of_line is called, which is an alias for ed_move_to_beg.

reline/lib/reline/line_editor.rb

Lines 1533 to 1536 in d348df9

	private def ed_move_to_beg(key)
	@byte_pointer = 0
	end
	alias_method :beginning_of_line, :ed_move_to_beg

Similarly, for other commands, you can find them by converting hyphens in Readline commands to underscores. If you can't find it even after converting hyphens to underscores, it's likely not implemented.

If you need to test, please refer to this file. It's a test using a testing framework called yamatanooroti.
https://github.com/ruby/reline/blob/d348df90d266e44f321810bcf272841fb9669f72/test/reline/yamatanooroti/test_rendering.rb

Here's how to run the tests:
https://github.com/ruby/reline?tab=readme-ov-file#run-tests

Feel free to ask me if you have any questions.