[mule-aist:01294] Re: [Oriya-group] tables in utkalm.ttf

[TAKAHASHI Naoto]

Hello,

Are there any progress or comments concerning utkalm.ttf mentioned
below?

Kenichi Handa writes:

> Hi, my name is Kenichi Handa, a colleague of Dr. Takahashi.

> In article <1093417253.21067.203010049 at webmail.messagingengine.com>, "Rajesh Pradhan" <rajeshpradhan at speedymail.org> writes:
> [...]
>>>  I checked various combinations and found that it fails when
>>>  KA(U+0B15), GA(U+0B17) or NNA(U+0B23) occupies the second position in
>>>  a consonant cluster.
>>>  
>>>  I investigated the font and found that the "pres" GSUB feature has
>>>  three lookup list indices, namely 9, 8 and 7 in this order.  The three
>>>  consonants mentioned above receive special treatment in "lookup 9",
>>>  thus cannot be replaced with a proper ligature glyph by "lookup 8" or
>>>  "lookup 7".
>>>                                         
>>>  I believe this is why our renderer fails.  What is your intention for
>>>  the special treatment in "lookup 9"?  Does our renderer fail to
>>>  understand the OpenType tables correctly?

>> Could you please give example of the oriya word you are trying to type?
>> I am not able to reproduce the problem using yudit. I shall look into
>> the "lookup 9" feature.

> I checked the source code of yudit and found that it handles
> lookups of types 1 (Single Substitution) and 4 (Ligature
> Substitution) only; it simply ignores lookups of the other
> types.  And, the type of lookup 9 and 8 are both 6 (Chained
> Contextual Substitution).  Thus, yudit displays this
> sequence correctly:

>   U+0B38 U+0B4D U+0B15

> But our library handles a lookup of type 6.

> Here's the detailed flow.  At first the above character
> sequence is converted to this glyph code sequence by cmap.

>   004A 0058 0029

> Next, "haln" feature (lookup 4) is applied to the sequence
> 004A 0058, and it is converted to 00C3.  Thus now we have
> this sequence.

>   00C3 0029

> Next, we try "pres" feature (lookup 9 8 7 in this order).
> Lookup 9 (type 6) says that, while checking 0029, if the
> previous glyph is 00A3..00C4, try lookup 14.  And lookup 14
> (type 1) says that 0029 should be converted to 00C5.  Thus
> we have this sequence at last.

>   00C3 00C5

> On the other hand, lookup 7 (type 4) says that 00C3 0029
> sequence should be converted to 00E7, the correct final
> glyph.

> So, our question is why "pres" feature has lookups 9 8 7 in
> this order.

> ---
> Ken'ichi HANDA
> handa at m17n.org

-- 
TAKAHASHI Naoto
ntakahas at m17n.org
http://www.m17n.org/ntakahas/