Re: [emacs-bidi] diacritics, ligatures, etc.

[TAKAHASHI Naoto]

Eli Zaretskii writes:

>> I have heard that placing Hebrew diacritical marks at the correct
>> position (or at the position that looks good) is very difficult for
>> computers.

> Can you tell any details, or where to look to find them?

No, I can't.  Sorry, but I just don't know the details.  A professor
who studies Semitic languages once told me so, but his speciality is
Hebrew in the Bible.  Perhaps it has something to do with what he told
me.

> My experience with one (very old) DOS word processor that supported 
> diacriticals is that there's no problem to do that well enough for the 
> text to be readable.  After all, Emacs is not a typesetter, so we don't 
> need it to provide the same accuracy as we require from a typeset
> Bible ;-).

It is a very good news. :-)

> Could you tell a bit more about a possible use of Emacs compositions for 
> diacriticals, and especially about its limitations?

You can specify, to some extent, how the elements in a composition are
placed relative to one another.  For example, put the "xy" at the
beginning of a buffer and issue the following.

  M-: (compose-region 1 3 [?a (bottom-right . top-left) ?b)

The first two argument 1 and 3 specify the region to be composed.  The
third argument [?a (bottom-right . top-left) ?b] means

  "Instead of just overstriking the characters in the region (`x' and
   `y' in this case), display the characters `a' and `b' so that a's
   bottom-right and b's top-left share the same position."

`bottom-right', etc. can be abbriviated to `br', etc.  Also try the
followings to get the idea.

  M-: (compose-region 1 3 [?a (br . tl) ?b (br . tl) ?c])
  M-: (compose-region 1 3 [?a (tc . bc) ?b])
  M-: (compose-region 1 3 [?a (cl . cr) ?b (bc . tc) ?c])

See the description of the function compose-region and that of the
variable reference-point-alist, if you are interested.

The limitation is the coarseness of the reference points.  Also, the
composed glyphs heavily depend on the font.  Especially, the existence
of blank pixels surrounding the minimum BBX affects the result.  Be
careful when you want to distinguish sin and shin explicitly.

-- 
TAKAHASHI Naoto
address@hidden
http://www.m17n.org/ntakahas/