Fix backwards character deletion on other whitespaces

[SoraTenshi]

This issue closes #2842

In summary:
delete_char_backward ignored any other whitespace other than ' ', which resulted in you not being able to backspace or C-h on insert mode when there's only whitespace with the character next to the cursor being a lexiographic whitespace.

[kirawi]

We actually probably just want to change this check here:

helix/helix-term/src/commands.rs

Line 2947 in 5b3b6ff

if !fragment.is_empty() && fragment.chars().all(|ch| ch.is_whitespace()) {

Most indentation will be using spaces or tabs, rather than full-width characters. I don't have a strong opinion on it though.

[SoraTenshi]

We actually probably just want to change this check here:

helix/helix-term/src/commands.rs

Line 2947 in 5b3b6ff

if !fragment.is_empty() && fragment.chars().all(|ch| ch.is_whitespace()) {

Most indentation will be using spaces or tabs, rather than full-width characters. I don't have a strong opinion on it though.

actually, i don't have any strong opinion either, however the change happened like that because this already looked rather wrong.

[archseer]

I think this won't work correctly with whitespace that has a wider width than 1 (eg. space used in Japanese text).

The body of this function should be extracted into helix-core and properly unit tested.

[SoraTenshi]

I think this won't work correctly with whitespace that has a wider width than 1 (eg. space used in Japanese text).

Well the Japanese text spaces are Ideographic Spaces.
I also filed this bug because i am using a japanese ime quite frequently.

This change has been tested locally, and works.

However i see what you mean, would it maybe make sense to something similar as the width variable for the substraction of the pos?

Some(c) if c.is_whitespace() => start -= c.width().unwrap_or(1),

[kirawi]

I think this is an unnecessary change, as aside from line endings, tab and space are the only two recognized as ascii whitespace.