xreadline by word movement

[doremiyeon]

is this useful? added by-word navigation and deletion bindings for xreadline()

• extend xreadline() with some by-word movement
• xreadline ctrl-w memmove less

[jarun]

Can you explain each of the behavioural changes for me to take a call?

[N-R-K]

Untested, but some cursory review.

[N-R-K]

I've pushed some changes:

1. Removed the superfluous checks. Just make the code gracefully work with 0.
2. isalnum (and all the other ctype.h functions) only accept argument in the rage [0, 255], we can't feed it wchar_t arguments. Added custom macro for it instead.
3. We might be dealing with unicode chars, so instead of isalnum I changed the checks to isspace which would work on unicode chars instead of just ascii.
4. I notice a bug in the current Ctrl+W handling, we only delete a single space if there's multiple before the cursor while readline will remove all. I've fixed it, but it also no longer treats / as a boundary (neither does readline).

@jarun do you want xreadline to follow readline's behavior or do you want to keep the custom / as word boundary behavior? If you want the custom behavior, do you want it just for Ctrl+w or the rest as well?

[doremiyeon]

Can you explain each of the behavioural changes for me to take a call?

I added Meta-b, Meta-f conforming to https://www.gnu.org/software/bash/manual/bash.html#index-forward_002dword-_0028M_002df_0029 as well as the corresponding deletion commands, M-d and M-bspc.

They move cursor by word boundary, where a word is defined to be composed of letters and digits.

[doremiyeon]

@N-R-K

I've pushed some changes:
3. We might be dealing with unicode chars, so instead of isalnum I changed the checks to isspace which would work on unicode chars instead of just ascii.

I hadn't considered unicode, but could isspace be turned into something to include all the other non-word (non alphanum) characters within ascii range as well?

[N-R-K]

but could isspace be turned into something to include all the other non-word (non alphanum) characters within ascii range as well?

What meaningful difference does it make?

[doremiyeon]

but could isspace be turned into something to include all the other non-word (non alphanum) characters within ascii range as well?

What meaningful difference does it make?

the word boundaries differ, do they not? many filenames have underscores in them, or dashes. navigating between those is the whole point of these movements. paths contain slashes, same thing.

also: strictly speaking, Ctrl-w explicitly differs from Meta-bspc by what it considers a word boundary. whitespace vs non-alphanumerical (including unicode, not sure if it distinguishes between unicode "special characters" and the rest off the top of my head).

edit: there's also iswalnum() that "should just work."

[N-R-K]

the word boundaries differ, do they not? many filenames have underscores in them, or dashes. navigating between those is the whole point of these movements.

Ok, I see. I've switched to iswalnum. Hopefully it doesn't contain any weird pitfalls like the ctype.h counterparts do.

strictly speaking, Ctrl-w explicitly differs from Meta-bspc by what it considers a word boundary.

I've left out ctrl-w change for now, I'll look into it later. But first I'm going to wait for @jarun to reply on whether he wants to keep the custom / behavior.

edit: there's also iswalnum() that "should just work."

Yes, I thought about it just now as well.

[jarun]

do you want xreadline to follow readline's behavior or do you want to keep the custom / as word boundary behavior? If you want the custom behavior, do you want it just for Ctrl+w or the rest as well?

We can keep it the same as readline.

[N-R-K]

We can keep it the same as readline.

Okay, done.

---

@doremiyeon can you test it out and make sure everything still works as intended?