C-n and fields

C-n and fields

[Richard Stallman]

It's now clear what the behavior of C-n with fields needs to be.
It should keep the column number as close as possible to the desired one
but without moving across field boundaries.

Can someone please implement that, then ack?

Re: C-n and fields

[Chong Yidong]

Richard Stallman <rms@gnu.org> writes:

> It's now clear what the behavior of C-n with fields needs to be.
> It should keep the column number as close as possible to the desired one
> but without moving across field boundaries.
>
> Can someone please implement that, then ack?

The behavior was broken by my 2006-08-03 change to
line-move-to-column, which in retrospect is completely misguided.
I've now reverted it.

The change was originally aimed at a line-motion bug reported on
emacs-pretest-bug on 18 July ("C-n doesn't work in Customize Option
buffer in -nw with long value displayed").  The way to trigger this
bug was supplied by Johan BockgÃrd:

    (insert (propertize "x" 'field t)
            (make-string 89 ?y)
            "\n")

    Put point on the x (or the first y) and press C-n. Point doesn't
    move.

I now think the problem lies, not in line-move-to-column, but in
line-end-position.  After computing the line end position,
line-end-position calls constrain-to-field with a nil ESCAPE-FROM-EDGE
argument.  Thus, if point lies right at the edge of the field,

            X|YYYYY
 non-nil     ^        null   
 `field'    point     `field'
 property             property

Then line-end-position reports the field end position, and similarly
end-of-line won't move point.  This is counter-intuitive.

How bout making line-end-position pass a non-nil ESCAPE-FROM-EDGE
argument?  (This indeed fixes the bug in the test case.)

*** emacs/src/editfns.c.~1.428.~	2006-09-12 23:04:03.000000000 -0400
--- emacs/src/editfns.c	2006-09-16 10:42:43.000000000 -0400
***************
*** 874,880 ****
  
    /* Return END_POS constrained to the current input field.  */
    return Fconstrain_to_field (make_number (end_pos), make_number (orig),
! 			      Qnil, Qt, Qnil);
  }
  
  \f
--- 874,880 ----
  
    /* Return END_POS constrained to the current input field.  */
    return Fconstrain_to_field (make_number (end_pos), make_number (orig),
! 			      Qt, Qt, Qnil);
  }
  
  \f

Re: C-n and fields

[Richard Stallman]

How does line-end-position have an effect on C-n?
I can't find any calls to it from the line-move code.
If it does indeed get called, can you tell me how?

I need to see how it gets called, and from where,
in order to think about what is right to do there.

Re: C-n and fields

[Chong Yidong]

Richard Stallman <rms@gnu.org> writes:

> How does line-end-position have an effect on C-n?
> I can't find any calls to it from the line-move code.
> If it does indeed get called, can you tell me how?
>
> I need to see how it gets called, and from where,
> in order to think about what is right to do there.

It's the call to end-of-line in simple.el:3597:

    (let (done)
      (while (and (> arg 0) (not done))
	;; If the following character is currently invisible,
	;; skip all characters with that same `invisible' property value.
	(while (and (not (eobp)) (line-move-invisible-p (point)))
	  (goto-char (next-char-property-change (point))))
	;; Now move a line.
	(end-of-line)
        ...