Subject: Re: Indian Language in AbiWord
From: Tomas Frydrych (tomas@frydrych.uklinux.net)
Date: Sat Dec 29 2001 - 04:30:02 CST
> > The bidi version of AW does two-character ligatures, i.e., it can do
> > a+b-> c; we do not do more then two at the moment. In future we
> > will probably use Pango (www.pango.org) which will take care of
> > that. We also do basic initial/medial/final/isolate glyph shaping.
> > However, all of this only if the alternate glyphs have separate
> > codepoints in Unicode, and I am not sure Indic languages do).
>
> They don't. The alternate glyphs for the vowels do (one for the vowel in
> isolation, and one for the vowel in combination with a consonant), but the
> consonants don't. So if you have RKSSRI, that's stored in Unicode (which for
> the Indic scripts is several copies of ISCII, 128 codepoints each, at
> different offsets) as RA VIRAMA KA VIRAMA SSA VIRAMA RI, and you have to turn
> that into I KSS SLASHRA CANDRA, where I is the vowel that is written before
> the consonant it comes after, KSS is a ligature that looks nothing like K or
> SS, SLASHRA (of which there may need to be two or three variants) is a
> vertical bar with a slash attached, and CANDRA (the first R) is a crescent in
> the upper right corner. I has a codepoint, but KSS and SLASHRA do not.
>
> The rules are slightly different from one Nagari script to another. JI, for
> example, has its own glyph in Gujarati, but not in Devanagari.
>
I think the only solution to this is to use a proper external renderer,
such as Pango. I think when we start rewriting the layout engine
after the 1.0 release, we need to start working on this as well, or at
least examine the Pango api and make changes that would make
it easier to plug it in at a later stage.
Tomas
This archive was generated by hypermail 2b25 : Sat Dec 29 2001 - 04:30:42 CST