Re: [emacs-bidi] Points composition - status.

[Roozbeh Pournader]

On Tue, 11 Dec 2001, Eli Zaretskii wrote:

> > From: "Yair Friedman (Jerusalem)" <address@hidden>
> > 
> > 2. The input method needs to re-order the points if they are entered in
> >    the "wrong" order.  Reading more sections 4.2 and 3.10 I understand
> >    they have to be ordered by "Canonical Combining Classes" property of
> >    the characters (the 4th field in UnicodeData file), then by code.  Is
> >    this correct?  I want to verify before creating all of the decoding
> >    tables etc. What might be the best strategy to do reordering?
> 
> Why do we need to reorder?  Can't we simply tell people that points
> should be typed _after_ the base character?

I think Yair is talking about the ordering of "more than one" points after
a base letter, which is called normalization. If two string are "equal"
in their normalized form, they are "equivalent"; if not, they're not (at 
least by definition). I am not sure about the algorithm Yair is 
suggesting, but it would be good to also look at UAX #15, Unicode 
Normalization Forms, at:

        http://www.unicode.org/unicode/reports/tr15/

BTW, where in Emacs do you exactly need this?

> I think reordering can lead to problems.  For example, what happens
> if I want to add a diacritical to an already existing character, and
> then continue inserting other characters without moving the cursor?
> If you reorder in this situation, you will place the diacritical
> under the wrong character, won't you?

Simple: no one should put diacratics before letters. We are only talking
about the ordering of multiple diacratics after a base letter.

roozbeh

-- 
Note: If you want me to read a message, please make sure you include my
address in "To" or "CC" fields. I may not be able to follow all the
discussions on the mailing lists I subscribe. Sorry. (No, there's no problem
to receive duplicates.)