Mode and position off after Atuin history recall with tab

[miltieIV2]

$ ble version
ble.sh, version 0.4.0-nightly+31f264a (noarch)

~/.blerc

#bleopt keymap_vi_mode_string_nmap=$'\e[1m-- NORMAL --\e[m'

bleopt prompt_ps1_final='$(starship module character)'
bleopt prompt_rps1_final='$(starship module time)'
bleopt indent_offset=2
#ble-bind -m 'vi_imap' -f 'C-c' 'vi_imap/normal-mode-without-insert-leave'
ble-bind -m vi_imap -f 'C-c' discard-line
ble-bind -m vi_nmap -f 'C-c' discard-line

# https://github.com/akinomyoga/ble.sh/issues/537
bleopt history_default_point=begin

1. Use vi mode.
2. Enter atuin search.
3. Select an item from the history.
4. Hit to recall (not accept) the line to ble.sh.
5. Observe that: the line is recalled, the cursor is one space past the end of the line, INSERT mode is off - for INSERT mode to be off, the cursor should be on the last character, not past the last character.
6. Press - the cursor moves to where it should be (on the last character).
7. It is behaving as if the line was recalled and "A" was pressed (append to end of line), but ble.sh is not in INSERT mode.
8. Always typing "A" or "a" after tab recall puts ble.sh in the expected mode and position.

(@akinomyoga Not sure if this is related to changes performed in #537)

[miltieIV2]

It seems like this recall behavior should follow the history_default_point behavior.
This is what BASH "/ does - recall history with cursor at the beginning, in command mode.

[akinomyoga]

Thanks for the report. OK, I'll fix it. I was expecting the plugins such as Atuin to properly place the cursor position based on the current mode, but Atuin seems to try to put the cursor at the end of the string. However, Readline seems to automatically adjust the cursor position when a plugin places the cursor at an unallowed position. ble.sh can follow the Readline behavior.

It seems like this recall behavior should follow the history_default_point behavior. This is what BASH "/ does - recall history with cursor at the beginning, in command mode.

Atuin's selection of the history entry is unrelated to the history movement in the line editor. Atuin picks up the command from its own database and inserts the string into the line editor. From the line-editor side (such as Readline and ble.sh), there is no way to tell whether a string inserted by a random plugin would be considered the history movement. If you want to change the cursor position that Atuin sets, you should find or request the corresponding setting in Atuin.

Also, this behavior seems to be the same in plain Bash plus Atuin. Even without ble.sh, the cursor position after the insertion by Atuin is at the end of the text. It's unrelated to ble.sh.

[akinomyoga]

Atuin doesn't seem to support the option for the cursor position. It's hardcoded to be at the end of the string. Issue atuinsh/atuin#893 discusses a related extension.

[miltieIV2]

Thanks!
I obviously don't have a good understand of which parts are doing what.

[akinomyoga]

Thanks for the report. OK, I'll fix it. I was expecting the plugins such as Atuin to properly place the cursor position based on the current mode, but Atuin seems to try to put the cursor at the end of the string. However, Readline seems to automatically adjust the cursor position when a plugin places the cursor at an unallowed position. ble.sh can follow the Readline behavior.

I adjusted the behavior in commit 36ab934.