vi-mode-switching reexecutes the prompt, which is slow

[tsoernes]

Setting fish_escape_delay_ms to 10 produces no change; exiting vi-mode is still very slow.

fish 2.7.1, kernel 5.0.3-200.fc29.x86_64, tried both tilix and urxvt

[faho]

It works for me - set -g fish_escape_delay_ms 300 makes things work quite slowly.

What exactly did you execute? Did you set it in config.fish or elsewhere?

[tsoernes]

I've tried both setting manually and in the config. It is not correct of me to say that the setting is not respected. Setting to 5 seconds makes it take 5 seconds. But setting to 10 msec and it's still like 400-600 msec. I do not have this problem in Bash vi mode with `set keyseq-timeout 0.01` in `.inputc`. There the delay is unnoticable. With fish and 10 msec it's nearly half a second. I've also noticed now that entering insert mode from command mode also takes a very long time, same as escape; at least 300 msec. So this issue is perhaps not related to the aformentioned setting.

[faho]

I just reinstalled 2.7.1 (I recommend upgrading to 3.0.2 btw!), and I can't say I can reproduce this.

What does bind \e; bind -M insert \e say?

[tsoernes]

bind \e cancel
bind -M insert \e 'if commandline -P; commandline -f cancel; else; set fish_bind_mode default; commandline -f backward-char force-repaint; end'

[tsoernes]

This is the full extent of my fish config:

fish_vi_key_bindings

set -U fish_escape_delay_ms 10

function fish_prompt
    powerline-shell --shell bare $status
end

set tacklebox_path ~/.tackle ~/.tacklebox
. ~/.tacklebox/tacklebox.fish

[faho]

Ah, okay.

Yeah, see that "force-repaint" there? That forces fish to repaint the prompt. And since it doesn't know which part will be different, it has to execute the entire thing.

Which means it'll call fish_prompt, fish_right_prompt and fish_mode_prompt. One of those must be slow for you. Make your prompt faster and the delay should disappear.

I also recommend upgrading to 3.0.2, since that's generally quite a bit faster.

[tsoernes]

So regular bash don't have to repaint the prompt? I'm using the same prompt there

[faho]

Yes, bash does not have the assumption that the prompt can change because of vi-mode. It doesn't feature a mode indicator.

[tsoernes]

Problem solved by changing to a lighter prompt. Is there a way to get a thin and blocky cursor for insert mode and command mode respectively? That would diminish the need to have the vi mode displayed in the prompt.

[faho]

Is there a way to get a thin and blocky cursor for insert mode and command mode respectively?

If your terminal supports it, set -g fish_cursor_default block; set -g fish_cursor_insert line.

[faho]

Okay, I've now added a commit that makes it so this only reexecutes the fish_mode_prompt function.

That should cut the time down significantly.

[faho]

After some careful thought, I've added 99dd6d7 which only uses fish_mode_prompt if it exists. Otherwise it acts like a normal repaint.

I'd assume there are people out there who have a mode indicator at another place in their prompt, and they seem to be okay with the time it takes. So they'd simply have to not define a fish_mode_prompt function (i.e. an empty file, not just an empty function) at all.

[floam]

Huh I didn't even know about fish_cursor_default - lets feature this more prominently in the documentation and a selector in fish_config.

[neon64]

@faho this has broken my config on the latest master - in which I have my own mode prompt inside fish_right_prompt (so it displays N, I, V etc.. on the right hand side of the screen). Is this regression intended, and if so, is there a workaround to manually re-execute fish_right_prompt every time my mode changes?

[faho]

Remove fish_mode_prompt. I.e. create an empty file at ~/.config/fish/functions/fish_mode_prompt.fish

[neon64]

Thanks that worked! - perhaps this new behavior ought to be mentioned in the docs somewhere though

[zanchey]

Updated to make even clearer in c5e535e.