Syriac Abbreviation Mark rendering

[KrasnayaPloshchad]

The Syriac block have a control character “Syriac Abbreviation Mark” (U+070F), as Unicode Standard, this mark should be used for generating a special overline on all subsequent Syriac script letters, until the first character which is not a Syriac letter or diacritic.

I copied the text “ܐ܏ܒܓܕ” from a Wikipedia article, when I paste into BabelPad (this is using Uniscribe), I can see this overline, this line has a point at each edge and a point in the middle, but before I copy this text, I can see just a substitution glyph between the letters. I’m using Firefox, which using your text shaping engine on all platforms. So how can I see this overline?

[behdad]

We like to implement that but haven't done so far.

[behdad]

cc @jfkthame

[KrasnayaPloshchad]

cc @khaledhosny

[jfkthame]

It's not clear to me what harfbuzz can/should do about this. It's not a shaping issue; it needs to be handled during drawing of the shaped text. Is there anything to be done in HB that would facilitate that for clients?

[khaledhosny]

MS introduced a new feature for this, http://www.microsoft.com/typography/otspec/features_pt.htm#stch and https://www.microsoft.com/typography/OpenTypeDev/syriac/intro.htm#features, so probably HarfBuzz can support it. Segoe UI Historic from Windows 10 seems to include it.

[khaledhosny]

Here the result of hb-shape with the string above:

$ hb-shape seguihis.ttf 'ܐ܏ܒܓܕ'
[gid2404=4+723|gid2399=3+1589|gid2395=2+956|gid2334=1+41|gid2335=0+1524]
hb-shape seguihis.ttf 'ܐ܏ܒܓܕ' --features="+stch"
[gid2404=4+723|gid2399=3+1589|gid2395=2+956|gid2486=1+0|gid2487=1+0|gid2488=1+0|gid2487=1+0|gid2489=1+0|gid2335=0+1524]

So HarfBuzz is doing the substitution (the feature is implementing as a contextual substitution applying a one to many lookup), but it obviously does not handle the special positioning and does not enable it by default.

[KrasnayaPloshchad]

@khaledhosny Yes, I’m using Windows 10, these sequences works for me in BabelPad owe to Segoe UI Historic font.

[jfkthame]

Oh, right - I was forgetting about the 'stch' feature. Yes, I suppose we can do that in HB, so it does become part of shaping after all.

[KrasnayaPloshchad]

I can also reproduce this issue in Firefox. First, I open this page in Firefox, click “HTML” button:
http://wymeditor.github.io/wymeditor/dist/examples/01-basic.html
Then I input following code:

<p style="font-family:Segoe UI Historic; font-feature-settings:'stch'">ܐ܏ܒܓܕ</p>
<p style="font-family:Segoe UI Historic; font-feature-settings:'stch' 0">ܐ܏ܒܓܕ</p>

The result for me is:
If the 'stch' feature is closed, only substitution glyph appearing.
If the 'stch' feature is opened, a short overline appearing at the position of SAM, with a dot in the middle.

[behdad]

I'll give it a try soon...

Can someone with hb-uniscribe on Win 8+ send me expected hb-shape results for a few sequences?

[behdad]

Implementing it is not that hard actually. Working on it.

[behdad]

I also wonder if the mark should be skipped for other GSUB/GPOS processing.

[behdad]

Also, what does windows do if the mark is at the end of the word?

[khaledhosny]

Seeing that the it is contextual substitution, for this particular font I think it will not happen unless the mark if followed by at least one Syriac character, but I don’t know what it will do in case of a simple substitution (no Windows 10 here to test, and one needs to craft a special font any way).