[HarfBuzz] Tai Tham Shaping Question #2 : MEDIAL RA

[Behdad Esfahbod]

On 13-01-07 05:04 AM, Theppitak Karoonboonyanan wrote:
> On Wed, Dec 26, 2012 at 3:12 PM, Theppitak Karoonboonyanan
> <thep at linux.thai.net> wrote:
>> On Mon, Dec 10, 2012 at 1:50 PM, Behdad Esfahbod <behdad at behdad.org> wrote:
>>
>>> Or just ship one that has both a DFLT table, and a non-DFLT table.
>>
>> How is this one?
>>
>>   http://linux.thai.net/~thep/esaan-scripts/fonts/Khottabun-hb.ttf

I didn't get to look into it.  Will do.

> Note that only 'ccmp' is used in the font, as there seems to be no
> indic_configs[] entry for Tai Tham yet. And 'blwf' just doesn't work
> without the virama info.

We don't necessarily need indic_config for Tai Tham.


> Here is my test result:
> - tham-test.txt is the input text.
> - tham-test.png is what hb-view renders.
> - tham-test-expected.png is the expected result.
> 
> Issues found:
> 
> - NA (ᨶ) + vowel AA (ᩣ) ligature is not always shaped if there is
>   something in between, such as subjoined consonant (line 1).
>   (For this one, I'm not sure what's the right thing to do, between
>   using GSUB and letting HB do appropriate preprocessing.

That's easy to fix.  Check MATRA_POS_BOTTOM in
hb-ot-shape-complex-indic-private.h.

>   How is this usually done in, say, Khmer "បុប្ផា"?)
> 
> - Leading vowels are correctly reordered, but medial RA (U+1A55)
>   is not (line 2-3).

I'll fix that.  We have the same issue in Myanmar and Cham too.  What you need
to do to get it reordered (after I fix the code) is to have the 'pref' feature
apply to the medial RA.  It can be a ContextSubst that just applies and has no
recursive lookups.  That's how HarfBuzz detects whether this character needs
to be reordered to pre-base position.


> - Final NGA (U+1A59) with virama following is not reordered after
>   the next base consonant (at the end of line 4).

Oh, that's new.  We need to figure out how to implement that.  That one will
be tricky.

b

> Regards,
> 

-- 
behdad
http://behdad.org/