bug#50679: 28.0.50; outline-minor-mode-cycle should be able to only cycle at the begining of a line

[Juri Linkov <juri@linkov.net>]

>>>         "<tab>" 'outline-cycle
>>>         "<backtab>" 'outline-cycle-buffer
>>>         "f" 'outline-forward-same-level
>>>         "b" 'outline-backward-same-level
>>>         "n" 'outline-next-visible-heading
>>>         "p" 'outline-previous-visible-heading
>>>         "u" 'outline-up-heading
>>>         "F" 'outline-move-subtree-down
>>>         "B" 'outline-move-subtree-up
>>>         "@" 'outline-mark-subtree
>>
>> But this means you can't type all these keys on heading lines?
>> This is why I confugured it not to use self-inserting keys:
>
> In my setup those keys only apply at BOL in a heading line.  I never
> type anything there.
>
> TAB for reindenting might be a bit trickier.  Apparently I never try to
> reindent a heading line when point is at BOL, because I never thought
> about a possible conflict, but I guess I would type SPC TAB in that
> case.

This is exactly what I do every time, i.e. first I type TAB to indent,
but it hides the body, then I type again TAB to unhide the body
to revert the mistake, then type SPC TAB to workaround this annoyance.

>>   (let ((map outline-mode-cycle-map)
>>         (cmds '(("M-<down>"  outline-next-visible-heading)
>>                 ("M-<up>"    outline-previous-visible-heading)
>>                 ("M-<left>"  outline-hide-subtree)
>>                 ("M-<right>" outline-show-subtree))))
>>     (dolist (command cmds)
>>       (define-key map (kbd (nth 0 command))
>>         `(menu-item
>>           "" ,(nth 1 command)
>>           :filter ,(lambda (cmd)
>>                      (when (outline-on-heading-p))
>>                        cmd))))))
>
> I can't find this block in outline.el, is this a suggestion for a
> possible default?

This is from my init file.  I don't suggest to add it to outline.el,
it only shows how it's possible to customize navigation keys.

>>> Looking at the code, it seems that adding bindings to
>>> `outline-mode-cycle-map` would achieve precisely that, but the name
>>> of the keymap doesn't reflect that.
>>
>> Do you suggest different keymaps?  And indeed, these two are
>> quite different beasts each with own problems.  So we need both
>> outline-mode-cycle-map and outline-minor-mode-cycle-map.
>
> Okay, those keymaps are active in the entire heading line, right?
>
> I would prefer to only have a special behavior at the first character of
> a heading line,

Philip already proposed a predicate that can be set to 'bolp'
to make the map active only on the first character.

> in exchange of allowing _any_ shortcut key with minimal
> interference.  Org mode already does that with a simple customization,
> (setq org-use-speed-commands t), so I don't think it would be too
> surprising for most users.

Adding more keys to this keymap has more problems.  For example,
in diff-mode TAB navigates to the next hunk, but with outline-minor-mode
TAB on the first character in a diff-mode buffer cycles the outline,
so another available key 'n' should be typed to navigate to the next hunk.
We can't sacrifice this key for outline-next-visible-heading,
because no more keys will remain for diff-hunk-next.