[smc-discuss] Reordering question with dotreph

[Rajeesh K Nambiar]

On Tue, Aug 27, 2013 at 1:57 PM, Santhosh Thottingal
<santhosh.thottingal at gmail.com> wrote:
> Rajeesh,
>
> for dotreph+ത്തു as in കാൎത്തു,
> what will be the reordering by rendering engine
>
> This is what I observe:
>
> th1 xx th1 dotreph u1
>
> IMO, this must be:
>
> th1 xx th1 u1 dotreph for old orthography font so that we can create ത്തു
> and then put dotreph using GPOS.
>
> Some more context:
> (mlm2)$ hb-shape Meera/Meera.ttf "കാൎത്തു"
> [k1=0+1605|uni0D3E=0+848|dotrephth1th1u1=2+2461]
>
> I wanted to remove dotrephth1th1u1 glyph. Once I removed it,
>
> (mlm2)$ hb-shape Meera/Meera.ttf "കാൎത്തു"
> [k1=0+1605|uni0D3E=0+848|dotrephth1th1=2+2708|u1=2+639]
>
> I want to remove dotrephth1th1 glyph too, then I get
>
> (mlm2)$ hb-shape Meera/Meera.ttf "കാൎത്തു"
> [k1=0+1605|uni0D3E=0+848|th1th1=2+2708|dotreph=2+0|u1=2+639]
>
> This is wrong with ു sign coming at the end instead of forming ത്തു
>
> since dotreph comes in between th1th1 and u1, we cannot create ത്തു
> ligature.
>
> Do you think this as a bug in hb reordering?

It is the harbuzz behaviour to reorder dotreph after the base
syllable. Sureshettan raised this issue and Behdad says this is what
Uniscribe (windows 8) does as well. See
http://lists.freedesktop.org/archives/harfbuzz/2013-July/003433.html

That said, I think to handle reph sanely using GPOS, it is best to
keep dotreph at the beginning of the syllable. Otherwise, we'll never
be able to handle the cases like this (ൎത്തു). We should cite these
examples and request for the behaviour to be changed to keep dotreph
before base syllable.

>
> An analogy can be ത്തു + ോ .
> This is not real, but with hb, I get:v
>
> ((mlm2)$ hb-shape --show-unicode Meera/Meera.ttf "ത്തുൊ "
> <U+0D24,U+0D4D,U+0D24,U+0D41,U+0D4A>
> [uni0D46=0+1184|th1th1u1=0+2461|uni0D3E=0+848]
>
> With dotreph I expect a similar reordering.
>
> Santhosh
>
>



-- 
Cheers,
Rajeesh
http://rajeeshknambiar.wordpress.com