WIP: Support Glyphs3-style glyph data incl. writing direction

[anthrotype]

I just created a PR from the 'GlyphInfo3' branch. @yanone @schriftgestalt what's the status of this?
Can you give this a better PR title and add a description of why/what this is about? thanks

related to #778 and googlefonts/ufo2ft#658

[schriftgestalt]

I finished what I had planed to do.

[madig]

Note to self:

1. Iterate over all names and ensure that I get the same info from glyphsLib as from Glyphs 3
2. Grab a bunch of .glyphs files and iterate over their names and test the same

[schriftgestalt]

Iterate over all names and ensure that I get the same info from glyphsLib as from Glyphs 3

when you iterate over all names in the xml file you basically compare the .xml files. More interesting (and potentially error prone) is the info construction from ligatures and composite glyphs.

[madig]

More interesting (and potentially error prone) is the info construction from ligatures and composite glyphs.

Could you maybe share the code? Did you follow it for making this PR?

[schriftgestalt]

The code in Glyphs and in glyphsLib is very different (the code in Glyphs is very old and does slightly more). But the behavior should be the same as shown by the tests.

What I meant is that when you test the new code, one should focus in glyphs that are not in the glyphData and those are mostly ligatures.

One thing I might need to explain:
There is a difference between f_f.ss01 and f_f._ss01. The first glyph is a stylistic alternate of the f_f and the later is a ligature of f and f.ss01. So if the suffix has the same number of underscores, each substring between the underscores is the suffix for the corresponding glyph name, including empty strings.

[yanone]

@anthrotype

I'm cleaning Georg’s work up a bit. The linting test passes now. Could you please take a look at the 3.7 test? postscriptNames.get("A") returns uni0041 now instead of None which doesn't seem like a mistake to me.

And then there's the regression check. That's whole bunch.

[anthrotype]

postscriptNames.get("A") returns uni0041 now instead of None which doesn't seem like a mistake to me.

the test expects None because the CustomGlyphData.xml does not contain an entry for glyph 'A'. So something has changed in this PR that makes the test fail. I think we want to keep the current behaviour as regards custom glyph data.

[schriftgestalt]

My code should set the production name to "A", not the uni-name. I added a test to make sure. So the new behavior seems to come somewhere else?

[schriftgestalt]

Can this be revived? The RTL<>LTR stuff is needed for g3 files round tripping and the other improvements are also good to have.