bug#67925: 29.1; delete-rectangle fails on multi-column characters

[awrhygty@outlook.com]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: awrhygty@outlook.com
>> Cc: 67925@debbugs.gnu.org
>> Date: Thu, 21 Dec 2023 16:30:39 +0900
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> > Thanks.  Does the patch below give good results?
>> >
>> > diff --git a/lisp/rect.el b/lisp/rect.el
>> > index 8dc188b..9049e32 100644
>> > --- a/lisp/rect.el
>> > +++ b/lisp/rect.el
>> > @@ -212,7 +212,10 @@ rectangle-dimensions
>> >        (cons width height))))
>> >  
>> >  (defun delete-rectangle-line (startcol endcol fill)
>> > -  (when (= (move-to-column startcol (if fill t 'coerce)) startcol)
>> > +  ;; We use >= here, not =, for characters that use more than one
>> > +  ;; column on display, when STARTCOL is in the middle of such a
>> > +  ;; character.
>> > +  (when (>= (move-to-column startcol (if fill t 'coerce)) startcol)
>> >      (delete-region (point)
>> >  		   (progn (move-to-column endcol 'coerce)
>> >  			  (point)))))
>> 
>> This patch gives similar result of kill-rectangle.
>> In the example above, I want all '4' characters moved to same column.
>
> How can that be done, when the first character takes 2 or more
> columns?  Deleting the first character is IMO wrong, since the other
> lines leave the first character intact.  Adding SPC to other lines is
> also wrong, since delete-rectangle is not supposed to _add_ columns.

I think wide characters may be replaced with SPC like TAB.

>> And TAB characters crossing the end column are not edited correctly
>> with delete-rectangle if indent-tabs-mode is on.
>> 
>> 012345678
>> 	8(TAB at head)
>> 012345678
>> 
>> The text above is edited to the text below.
>> 45678
>> 	8(TAB at head)
>> 45678
>> 
>> If indent-tabs-mode is off, the result is the text below.
>> 45678
>>     8(four SPCs)
>> 45678
>
> This is a separate issue with the original code.  It also happens with
> kill-rectangle, btw.  We could fix it by temporarily binding
> indent-tabs-mode to nil inside these commands -- would that be
> acceptable?  The result will be that the killed rectangle includes
> spaces, not the leading TAB.

I prefer that the killed rectangle has same column for each line.
If TAB is included when yanking, the current column affects the width of
the yanked string for each line.