Add Bengali Glyphset

[justvanrossum]

As suggested here: https://github.com/googlefonts/fontra/discussions/1943#discussioncomment-13600811

• https://github.com/mitradranirban/fbf-bn-glyphset/
• https://github.com/mitradranirban/fbf-bn-glyphset/blob/main/fbf-bn-glyphset.csv

@mitradranirban, I'm looking at this glyphset, and wonder if can convince you to convert to the Glyphs App convention.

For simple characters, this looks like this:

• bn_a -> a-beng

It uses a script suffix with a dash, which only needs to be mentioned once for ligatures and the like. For example:

• bn_ja_bn_hasanta_bn_nya.akhn -> ja_halant_nya-beng (not sure if the .akhn suffix is really needed?)
• bn_ja_bn_hasanta_bn_nya_bn_hasanta.haln -> ja_halant_nya_halant-beng

It has the distinct advantage that the components of the ligature can be automatically parsed, which allows Fontra to display a (hopefully) meaningful placeholder (also in the font overview):

Additionally, sadly, there is currently a shortcoming in Fontra where if you double-click an encoded glyph in the font overview, it will show the right character, but takes the glyphname from the Glyphs App mapping. For example, if I open bn_a from the font overview, it will show as a-beng in the editor.

[mitradranirban]

Actually I have taken the name from how Fontforge names the conjunct according to the lookup used. The glyph name is useful for preview in fontra, however as I have to use Fontforge for visual preview of open type table I need to replace the hyphen with underscore as Fontforge follows open type specifications and don't accept hyphen in glyph name. And I found fontra continues to show preview even when '-beng' is replaced with '_beng'.

[justvanrossum]

The FontForge compat requirement makes me think this particular set is perhaps better as a personal or custom glyph set, and perhaps not as a "builtin" Fontra glyph set.

[mitradranirban]

As fontra has till now rudimentary opentype support and GSUB and GPos(Non kerning) support is essential for Indic fonts, the Fontforge compatibility will be required for developing Indic fonts using open source software.

[mitradranirban]

As for .akhn, .cjct suffixes, they indicate which GSUB lookup the ligature belong, which will help the developer create the lookup. Any error in lookup.
As per Glyph compatible naming of Bengali conjuncts, I can make a new CSV file for that and share that for inclusion.

[justvanrossum]

Note that while - is illegal in glyphnames in binary ttf/otf, it is not illegal in source data.

With UFO it is possible to provide custom "production" names via the public.postscriptNames font.lib key:

• https://unifiedfontobject.org/versions/ufo3/lib.plist/#publicpostscriptnames

This is the preferred way to map from "designer friendly names" and "production names". Fontmake will use this mapping for the final glyph names in the ttf/otf output.

[mitradranirban]

I have modified the glyphset to make a fontra specific one.

https://github.com/mitradranirban/fbf-bn-glyphset/raw/refs/heads/main/fontra-bn-glyphset

Hopefully it will work well with fontra

[justvanrossum]

It is an improvement, but due to Fontra limitations, it would work better if all base glyph names would be come from this sheet:

https://raw.githubusercontent.com/googlefonts/fontra/refs/heads/main/src-js/fontra-core/assets/data/glyph-data.csv

For example, instead of aakaar-beng, use aaMatra-beng.

[mitradranirban]

made the necessary changes as per Glyphs nomenclature
https://github.com/mitradranirban/fbf-bn-glyphset/blob/main/fontra-bn-glyphset.csv

[justvanrossum]

There are a few glyph names that contain accidental spaces.

There are also some glyph names that contain an extension as part of a ligature component, such as ka_ssa.akhn_halant_ma-beng. This doesn't quite work, as extension should by definition go to the end. The extension is split first, then the script, then the remaining base is split for ligature components. This example gets the correct (I assume) placeholder string if it were named ka_ssa_halant_ma-beng.

[mitradranirban]

I have expanded the components like ka_halant_ssa_halant_ma-beng
However ligatures with more that three components are not likely to be shown by placeholders as they are created after 2nd round of open type processing - like
ka+halant+ssa+halant+ma -> ka_halant_ssa.akahand +halant+ma -> (ka_halant_ssa_akahnd)_ma.conjunct

[justvanrossum]

There are still two glyph names with spaces:

• ta_halant _ta_halant_ra-beng, note the space after the first ta_halant
• ka_ halant_ra-beng.salt, space after ka_

[mitradranirban]

Removed those overlooked spaces
Also added few missing halant s and deleted a duplicate entry.
Unfortunately due to cache of old file in my fontra setup I can not test the modification

[justvanrossum]

I think it looks good now. The first column header was renamed to "oName", that may have been an accident?

You can avoid caching by being explicit about the commit. For example, go to the history page for the file, then click on the "View code at this point" button, and use that URL.

For the latest commit, that gives me https://github.com/mitradranirban/fbf-bn-glyphset/blob/042bc9cd183f87c5584fca5c4c5d5760aeeedd51/fontra-bn-glyphset.csv

[mitradranirban]

Except half forms and some conjuncts, most of the placeholders are working fine