Re: Adjust point to move it off zero-width characters

[Eli Zaretskii <eliz@gnu.org>]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: emacs-devel@gnu.org
> Date: Sun, 07 Aug 2011 20:50:23 -0400
> 
> So you're saying that it should always be fused with the next char?
> Then my question becomes:
>  Is there a reason to fuse it with the next char rather than fusing
>  it with the preceding char? ;-)

There is a good reason for the directional control characters, see
below.  For others, I think it's immaterial.

> > There's one other complication: to _really_ DTRT, there's one
> > directional control that needs to be fused with the preceding
> > character: the PDF, u+202c (POP DIRECTIONAL FORMATTING).  That's
> 
> That was my impression as well: the fusing should not always be done in
> the same direction.
> I'd still like to hear your reasoning about why it only depends on the
> particular zero-width char and not on the surrounding context.

I explained that, for the directional control characters: the need to
be able to interactively copy/paste text and be sure it displays the
same.  The directional controls affect the _following_ characters.  If
we "fuse" LRM/RLM/LRO/RLO/LRE/RLE with the previous character, they
can only be copy/pasted with the text that precedes them, where they
have no effect whatsoever.  That's why PDF is different: it affects
the characters that precede it.

> > But hard-coding the special treatment of PDF on the C level sounds too
> > kludgey, no?
> 
> Without knowing why you think LRM needs to be fused one way and PDF in
> the other way, it's hard to know for sure whether hard-coding them is
> a kludge, but yes: it doe sound likely to be so.

So what are the alternatives?  Another char-table?