[Glyphs3] support alternate layers

[RosaWagner]

Font doesn't build if has alternate layers.

Easy to go back to format 2 using intermediate layers, but not for alternate ones.
I'll try to find a test case in Google Fonts and link it here.

[RosaWagner]

Update: using this branch, I confirm it works with intermediate layers (build+interpolation). I can't confirm for alternate ones.

[RosaWagner]

Update: when using format 3 and this main branch of GlyphsLib, I confirm the font builds if it has alternate layers, but the interpolation of the said glyphs (with alternate layers) don't work.

[RosaWagner]

This is a test case for @simoncozens.
Bold style should be a square (like when exporting with Glyphs), but with fontmake, it is still a circle.
alternate.zip

[simoncozens]

Thank you!

[RosaWagner]

(It may be possible to have alternate layers on other axes than the first without too much trouble, but I deliberately want to do this in fairly small steps.) (#729)

My branch of Literata has sources with alternate layers for next step:

• alternate layers for 2 axes
• min or max value only on each axis

This is a minimal test-case:
alt-2axes-test.zip

A: no layers
B: alternate layers with location axis0: min, 120, axis1: min, 600
C: alternate layers with location axis0: 120, max, axis1: 600, max
D: alternate layers with location axis0: min, 120, axis1: 600, max

[simoncozens]

OK. Thanks for this; that's perfect for working on. Now I have to work on it...

@madig @belluzj, as you know, my interest is in using glyphsLib for font compilation, for which we don't need round-tripping between UFO and Glyphs. I am happy to provide a PR which implements multi-axis alternate layers for G3. I am not interested, at this stage, in making that round-trip from UFO to Glyphs; it isn't something I need, and things are going to be mind-bendingly complicated enough without having to worry about doing it all again backwards.

So: Do you need G3 alternate layer round-tripping? If so, are you prepared to implement it? Will you approve a PR which provides G3->UFO alternate layers but not the other way around?

(This is basically another symptom of #680.)

[anthrotype]

I think it's fine if at least initially we only support these when glyphs -> UFO.

[madig]

Ok with me, we can add more support later.

[simoncozens]

I have something which works for everything apart from D. The problem is that we split the bracket rules into separate buckets for forward and reverse rules, and D is a forward rule in parts and a reverse rule in parts. I think the solution is to get rid of the concept of "reverse brackets" - I guess that doing that would probably also help us be able to implement rules which have both a min and a max range - but I am not really sure whether knocking down that wall will make the whole building collapse.

Draft PR at #767.

[schriftgestalt]

I removed the backwards bracket layer rules in Glyphs 3. Too complicated. What I do now in cases where one would use a backwards bracket (]100]) in Glyphs 2, I add the rule to the master layer and have an empty rule in the alternate.

[emmamarichal]

Hello!

I have a file where some glyphs have three values in the alternate layer.

Even if I simplify it and remove extreme values, this is not supported yet and the export did not work.

So I tried to find another solution, by adding alternate glyphs and make the substitution directly with the rvrn feature. But now I have this error message:

ufo/features.fea:4:1: Expected glyph class definition or statement: got NAME condition.

Adding a "name" block in the feature from Glyphs had no impact, but maybe I do something wrong, any idea?

Here is the test file:
Playfair test.zip

Thanks!

[vv-monsalve]

An error is reported when building a font with Alternate layers.

Error reported: ValueError: Alternate rules on non-first axis not yet supported
Axis using the bracket layers: CRSV. (Originally was second axis, moved it to be the first one, but didn't work)
Command used: fontmake -g filename.glyphs -o variable

Alternate layers are used here to receive the Cursive shapes (two-stories a -> schoolbook a) instead rvrn and rlig since the Glyphs specific variable feature syntax is not supported yet #800

Test file (a and ã have the bracket applied):
Geo-bracket-test.glyphs.zip

[emmamarichal]

@vv-monsalve @simoncozens

I have it also:

raise ValueError("Alternate rules on non-first axis not yet supported")
ValueError: Alternate rules on non-first axis not yet supported

the weird thing is that I already managed to export once, the error only appeared recently

Files: https://github.com/emmamarichal/Playfair/tree/master/sources

I'll dig a bit more to see if I didn't do any mistakes

[schriftgestalt]

Glyphs 3 supports alternate rules for all axes. So this is a remnant from Glyphs 2 times.

[simoncozens]

ValueError: Alternate rules on non-first axis not yet supported

I fixed this in November. Problem is, the last glyphsLib release was in October.

[anthrotype]

we shall make a new release soon (by end of this week)