bug#55692: 29.0.50; delete-selection-mode: Replace region only if set using the mouse

[Eli Zaretskii <eliz@gnu.org>]

> From: Visuwesh <visuweshm@gmail.com>
> Cc: 55692@debbugs.gnu.org
> Date: Sun, 29 May 2022 13:40:20 +0530
> 
> >>    By default, text insertion occurs normally even if the mark is
> >>  active---for example, typing @kbd{a} inserts the character @samp{a},
> >>  then deactivates the mark.  Delete Selection mode, a minor mode,
> >>  modifies this behavior: if you enable that mode, then inserting text
> >>  while the mark is active causes the text in the region to be deleted
> >> -first.  To toggle Delete Selection mode on or off, type @kbd{M-x
> >> -delete-selection-mode}.
> >> +first.  If you want to replace only temporary regions, set by
> >> +mouse-dragging or shift-selection, then change the variable
> >
> > There should be cross-references here to where these methods are
> > described in the manual.
> >
> 
> These methods as in "mouse-dragging" or "shift-selection".  I'm afraid I
> don't fully understand what you mean.

I mean to add @xref or @pxref to the places in the manual where mouse
selection and shift-selection are described.  Those places are,
respectively, the nodes "Setting Mark" and "Shift Selection".

> >> @@ -251,6 +259,10 @@ delete-selection-pre-hook
> >>  have this property won't delete the selection.
> >>  See `delete-selection-helper'."
> >>    (when (and delete-selection-mode (use-region-p)
> >> +             (if delete-selection-temporary-regions-only
> >> +                 (and (consp transient-mark-mode)
> >> +                      (eq (car transient-mark-mode) 'only))
> >> +               t)
> >
> > I wonder whether we should also optionally allow replacing text in
> > regions activated by "C-u C-x C-x" when transient-mark-mode is off.
> > Those cause transient-mark-mode to have the value 'lambda', not
> > '(only...)'.  So maybe the new defcustom should be a tristate, not a
> > boolean?
> >
> 
> Sure, that could be useful.  But I have transient-mark-mode turned on
> always so I can't comment much on it.  However, what would the third
> state be?

Some symbol.

> When the defcustom is t, we could have it check for '(only . ...)'
> and the value 'lambda', no?

Yes, we could have:

  . nil - always replace the region
  . t   - replace region from mouse, shift-selection, and "C-u C-x C-x"
  . selection - only replace  region from mouse and shift-selection

> > Also, can we have the above condition in a more elegant form?  In
> > general, whenever you have something like
> >
> >    (if SOMETHING foo t)
> >
> > it begs to be rewritten so the "t" part is not needed, because 'if'
> > itself already returns a boolean value.
> 
> Sure.  I will factor out the condition into a separate function.  Is
> that more elegant?

No, that's overkill.  I meant that if you need an 'if' to return t or
something else, you could instead make it return nil or something
else, as in

   (if SOMETHING foo)

and reverse the condition.

But if you still don't understand, just forget this comment.