search

w3c / font-text-cg

GitHub Pages
https://w3c.github.io/font-text-cg/
Other
28 stars 5 forks source link

[Myanmar] Shaping of "abbreviation" characters #58

Open simoncozens opened 3 years ago

simoncozens commented 3 years ago

Myanmar symbols ၌၍၎၏ (U+014C-U+104F) are considered by Unicode to be punctuation marks, and have no Indic_Syllabic_Category value. However, the symbols are pronounced, and may have tone marks - dot below, anusvara, visarga. There is apparently manuscript attestation for forms such as ၏့.

Because the symbol characters are not bases in the OpenType cluster model, they are not allowed to form syllables with marks, and so a dotted circle is inserted on shaping, as you can probably see above.

Supporting these forms correctly in OpenType would require a bigger change to the Myanmar cluster model than the recent medial-la issue, so I thought it would be worth discussing here first.

(@ohbendy, @peterconstable, @xadxura, @behdad)

behdad commented 3 years ago

If you want to open an issue against HarfBuzz, we will happily work out a solution there, and then communicate our findings back to @xadxura. I don't see the point of opening against this repository.

NorbertLindenberg commented 3 years ago

Opening an issue in this repository serves at least to alert people involved in Unicode or in shaping systems other that HarfBuzz that there is an issue.

I'm pretty sure the Unicode Script Ad Hoc would be interested in seeing the manuscript attestations, possibly with a proposal for how to update the Indic syllabic category or general category of these characters. (As the Myanmar shaping engine documentation doesn't reference the Indic syllabic category, it would need to be updated separately from Unicode data. Sadly, that documentation appears to be abandoned.)

r12a commented 3 years ago

Fwiw, I think that ၎ has the Indic_Syllabic_Category value Consonant_Placeholder which iiuc would allow it to serve as a base for placement of vowel marks and tones. See an example

The values for the others, i agree, are undefined.

PeterConstable commented 3 years ago

Agree with Norbert that it would be good to have a doc submitted to Unicode and discussed by the Script Ad Hoc group proposing InSC property values.

That said, the InSC property has no specification and is not tied to any Unicode algorithms. Microsoft's Myanmar implementation spec pre-dated Unicode's InSC property, but even with it available any values don't determine what an OT shaping implementation or spec would do. At best, they suggest, modulo how one wishes to interpret the property values.

(For my part, I think Unicode's lack of any spec for InSC is not a good thing.)