With EditMode::Vi, insert mode should not have keyseq_timeout applied when receiving KeyPress::Esc

[bobbygebert]

When in insert mode, you have to wait for 500ms (or whatever timeout was given in the editor config). I would expect hitting the escape key to return to command mode immediately.

[gwenn]

See #371

In vi mode, when you read an ESC from stdin, you don't know if it is a single escape or an escape sequence

[homburg]

@gwenn Maybe we could change the default value for vi mode timeout? There is not delay in zsh/bash vi mode.

[gwenn]

Are you sure ?
https://tiswww.case.edu/php/chet/readline/readline.html

int rl_set_keyboard_input_timeout (int u)

https://stackoverflow.com/questions/44890662/disable-timeout-on-esc-key

ttimeoutlen timeoutlen

https://python-prompt-toolkit.readthedocs.io/en/master/pages/advanced_topics/key_bindings.html?highlight=timeoutlen#timeouts

[homburg]

No, you are absolutely right.

Tested some different delays for switching from insert mode to command mode with escape:

environment	shell	delay	reference
macOS	bash	~0ms	?
ubuntu (docker)	bash	500ms	...
macOS	zsh	400ms	https://github.com/zsh-users/zsh/blob/b74635aa741d7e2d8d81f601f0192e79037aa991/Src/params.c#L823
alpine (docker)	ash	~100ms	?
*	fish shell	30ms	fish-shell/fish-shell#6590 :-)

An important distinction is that bash seems to recognize command movement keys immediately after escape, eg:

1. <insert mode>
2. input: Esc, h
3. (without delay) <command mode>, caret one step left

Not sure how that's implemented

[bobbygebert]

Yeah @homburg, the fact that movement keys aren't recognized after escape is what surprised me. Maybe that is the solution. We could add key sequence rules for those combinations pretty easily, right?

[homburg]

@bobbygebert Sounds like a good strategy, what do you think @gwenn ?

For reference, I think this is relevant bash readline logic: https://git.savannah.gnu.org/cgit/readline.git/tree/readline.c#n901 but I'm not familiar enough with the code to understand it.

[gwenn]

I am not sure to understand:

the fact that movement keys aren't recognized after escape is what surprised me.

Does it mean that movement keys are not recognized when timeout is too low ?
Or does it mean that movement keys are not recognized even when timeout is high ?

[homburg]

@gwenn it means that bash recognizes the movement keys, eg. j as <command-mode> keys during the timeout, ie. esc, j -> instant <command mode>

[homburg]

@gwenn I did a quick proof-of-concept, that seems to match the behavior: #426

[gwenn]

@homburg Ok, I understand now. We don't expect Meta-X in Vi mode only escape sequence. So Meta-X means Esc (exit insert mode) and do command associated to X. Looks good to me. We should be able to refactor/split vi_command method and call the bottom part from your patched block in vi_insert.
Thanks.

[gwenn]

6.3.0 released