editor: move completion functions

[flatcap]

NOTE: this is working, but I need to do some more testing.

Deep within libenter there are lots of dependencies caused by auto-completion and history.

Users call mw_get_field() with arguments like COMP_ALIAS which determine which completion function to use and where to save the history.
Worse still is the fact that the alias code is in libenter.
and browser code and notmuch code and pattern code...

This PR creates a basic Completion API that passes in a function to do the work.
It should help with: #3955 Neomutt completion API

All the functions on the right are now in more appropriate places.
The abstraction isn't perfect, but it's a step forward.

Source: graphviz, svg

[Amudtogal]

Hey flatcap,
this is amazing, thank you for the great refactor.
I will rebase my autocompletion branch on this, and this should speed things up.

[Amudtogal]

Excellent, this is now so much clearer (at least in my head).

It's great that each user (alias, browser etc) manages their own completion function, without needing too much logic.

[Amudtogal]

Excellent, this is now so much clearer (at least in my head).

It's great that each user (alias, browser etc) manages their own completion function, without needing too much logic.

[Amudtogal]

Excellent, this is now so much clearer (at least in my head).

It's great that each user (alias, browser etc) manages their own completion function, without needing too much logic.

[Amudtogal]

just one thing I spotted when compiling:

browser/complete.c:122:20: warning: implicit declaration of function ‘mutt_mb_is_shell_char’; did you mean ‘mutt_mb_is_pathname_char’? [-Wimplicit-function-declaration]
  122 |        (i > 0) && !mutt_mb_is_shell_char(wdata->state->wbuf[i - 1]); i--)
      |                    ^~~~~~~~~~~~~~~~~~~~~
      |                    mutt_mb_is_pathname_char

how would i best submit tiny patches like this?
is it best in a pull request, or assemble them along with the completion api branch?

[flatcap]

browser/complete.c:122:20: warning: implicit declaration of function ‘mutt_mb_is_shell_char’; did you mean ‘mutt_mb_is_pathname_char’? [-Wimplicit-function-declaration]

Hmm... I don't understand.

neomutt/browser/complete.c

Lines 121 to 124 in b99bce7

	for (i = wdata->state->curpos;
	(i > 0) && !mutt_mb_is_shell_char(wdata->state->wbuf[i - 1]); i--)
	{
	}

should be fine, because the file includes mutt/lib.h

neomutt/browser/complete.c

Line 33 in b99bce7

#include "mutt/lib.h"

which includes mbyte.h

neomutt/mutt/lib.h

Line 87 in b99bce7

#include "mbyte.h"

which declares mutt_mb_is_shell_char()

neomutt/mutt/mbyte.h

Line 49 in b99bce7

bool mutt_mb_is_shell_char(wchar_t ch);

I can't repeat the error myself with gcc or clang.

is it best in a pull request, or assemble them along with the completion api branch?

if you can figure out what's happened, push a commit to this branch.

how would i best submit tiny patches like this?

Generally, feel free to create/add to a branch [trivial] in the neomutt repo.
I'll merge it into my next bundle of tidying.

[Amudtogal]

Fair enough, it is gone on my side as well, after make clean && make.
Maybe it was a lingering object file...
If it comes back up, I will contribute to [trivial]

[Amudtogal]

Ah found it:
further down in my dev branch, I am renaming mutt_mb_is_shell_char to mutt_mb_is_pathname_char (more verbose, and it was only used in one spot anyway).
This was causing the compilation problem.

I have rebased my dev branch on this pull request to get started.