[varLib instancer] should set up italic style linking as an option

[RosaWagner]

When slicing a VF with ital or slant axis, varLib instancer could set up proper italic style bits + italic angle + caret slope if italic != 0 and if slant != 0. For example with and additional argument like --is-italic followed by an integer for italic angle.

For example when running:
fonttools varLib.instancer "../fonts/variable/Nunito[ital,wght].ttf" ital=1 wght=200:1000 -o "../fonts/variable/Nunito-Italic[wght].ttf"

Then we have to patch this kind of metadata:

for font_path in glob.glob("../fonts/variable/*Italic*.ttf"):
        with open(font_path, "rb") as f:
            print("Fix italic style linking in {}".format(font_path))
            ttFont = TTFont(f)
            ttFont["head"].macStyle = MacStyle.ITALIC
            ttFont["post"].italicAngle = -9
            ttFont["hhea"].caretSlopeRise = 1000
            ttFont["hhea"].caretSlopeRun = 158
            ttFont.save(font_path)

[madig]

Here's some code to set up some kind of linking:

    # style mapping
    styleMapStyleName = font["name"].getName(2, 3, 1, 0x409).toStr().lower()
    macStyle = []
    if styleMapStyleName == "bold":
        macStyle = [0]
    elif styleMapStyleName == "bold italic":
        macStyle = [0, 1]
    elif styleMapStyleName == "italic":
        macStyle = [1]
    font["head"].macStyle = intListToNum(macStyle, 0, 16)

    selection = font["OS/2"].fsSelection
    selection &= ~(1 << 0)
    selection &= ~(1 << 5)
    selection &= ~(1 << 6)
    if styleMapStyleName == "regular":
        selection |= 1 << 6
    elif styleMapStyleName == "bold":
        selection |= 1 << 5
    elif styleMapStyleName == "italic":
        selection |= 1 << 0
    elif styleMapStyleName == "bold italic":
        selection |= 1 << 0
        selection |= 1 << 5
    font["OS/2"].fsSelection = selection

[anthrotype]

yeah, I like that.
basically what I suggested to do #1264 (comment) but never got around doing it

PRs welcome!

[madig]

@RosaWagner Can you please test with the latest fontTools release and report what else instancer should be doing?

[anthrotype]

Fixed with #2673 -- feel free to reopen if not

[RosaWagner]

Hi, sorry I was on holidays.

The PR did fixed the style linking issue but I still see 2 things that would influence user experience:

• italic angle is still 0
• caretSlopeRise and caretSlopeRun are still 1 and 0 respectively. They can be calculated from the italic angle.

[anthrotype]

we only set the post.italicAngle for "slnt" axis where the relationship is well defined, but not for "ital". How would you go about setting that for "ital" axis?

fonttools/Lib/fontTools/varLib/__init__.py

Lines 876 to 880 in 455158f

	if "slnt" in location and "post" in font:
	italicAngle = max(-90, min(location["slnt"], 90))
	if font["post"].italicAngle != italicAngle:
	log.info("Setting post.italicAngle = %s", italicAngle)
	font["post"].italicAngle = italicAngle

caretSlopeRise and caretSlopeRun

I don't think we touch those. Feel free to suggest how this should be updated and/or make a PR, thanks

[RosaWagner]

I use the ital axis in anticipation of slicing, to set up the stat, fvar and name tables correctly from scratch.

If I set up the VF with a slant axis, then the STAT in the output italic is wrong:

'slnt' should be be replaced by an 'ital' axis [0;1], Roman=0, Italic=1, and 0 linked to 1 in Roman.

caretSlopeRise and caretSlopeRun
I don't think we touch those. Feel free to suggest how this should be updated and/or make a PR, thanks

I don't know exactly how to do that, but this is the informations I found to calculate them (the OT spec is now very chatty about it): https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6hhea.html

[RosaWagner]

Coming back to this since we are trying to onboard a font which has a continuous italic axis that we need to slice.

I made 3 test cases that I attach here, and below is the result. I think that caret slope and italic angle should be set in all cases by varLib.instancer. Then it might make sense that the nameID25 and STAT table is not handled by fonttools here.

ITAL = -12

Name

• ID25 appends Italic
• ID2,17,etc have Italic

Style linking bits

• post (italic angle is not 0)
• hhea (caret slope is not 1)
• head (macstyle is Italic)
• OS/2 (fsSelection is Italic)

Design space

• only italic instances
• no slant/ital slider
• ital axis in stat
• ital value is 1

ITAL = 1

Name

• ID25 appends Italic
• ID2,17,etc have Italic

Style linking bits

• post (italic angle is not 0)
• hhea (caret slope is not 1)
• head (macstyle is Italic)
• OS/2 (fsSelection is Italic)

Design space

• only italic instances
• no slant/ital slider
• ital axis in stat
• ital value is 1

SLANT = -12

Name

• ID25 appends Italic
• ID2,17,etc have Italic

Style linking bits

• post (italic angle is not 0)
• hhea (caret slope is not 1)
• head (macstyle is Italic)
• OS/2 (fsSelection is Italic)

Design space

• only italic instances
• no slant/ital slider
• ital axis in stat instead of slnt
• ital value is 1

Test-Slice.zip

[anthrotype]

sorry for the delay, I thought I had already replied here last week but apparently it got lost (maybe I closed the tab before hitting the green button).

a continuous italic axis

You mean you can have ital=0.5? That's a bit uncommon, I thought ital was meant to be a discrete axis, like a boolean 0 or 1.

I think that caret slope and italic angle should be set in all cases by varLib.instancer

as I said above, we only set the post.italicAngle for "slnt" axis where the relationship is well defined in the OT spec, but not for "ital". A font could arguably be "italic" in style without being slanted, or two fonts be equally considered "italic" while having different slant angles; there's no intrinsic relationship between "ital" and "slnt" other than the one the font designer specifies.
If you are suggesting that we also update the caret slope/rise based on the "slnt" axis then I think we could do that. I suggest you file a separate specific issue for that, as this one is already getting too broad, and already partly addressed (we already set the style linking flags in OS/2 and head table after #2673, but we only do that when --update-font-names option is passed [which in turn requires a full STAT table] and we only consider a font "bold" or "italic" if their subfamily name after instancing contains those particles).

regarding nameID 25 ("Variations PostScript Name Prefix" string), my understanding is that it is optional/cosmetic (to help squeeze long family names) and Adobe's "PostScript Name Generation Algorithm" document explains how to make one up if 25 is missing. I don't think it should be the job of the instancer to update that.

Coming back to ital vs slnt, does your font include both ital and slnt? The interaction between the two axes is a bit complicated, the spec uses these words (emphasis mine):

The Italic axis is distinct from the Slant axis ('slnt'). Fonts may use one or the other, depending on the nature of the design, but should rarely use both. While an italic design often incorporates some slant in the design, use of the Italic axis does not require use of the Slant axis. An italic font should not be characterized in the STAT table as being italic and also having some slant, unless the font family includes multiple italic designs with different amounts of slant.

https://learn.microsoft.com/en-us/typography/opentype/spec/dvaraxistag_ital

[RosaWagner]

You mean you can have ital=0.5? That's a bit uncommon, I thought ital was meant to be a discrete axis, like a boolean 0 or 1.

You are right. My goal is that once separated, a VF with a continuous ital or slant axis has all metadata corrected to be a family with proper Roman/Italic relationship. ITAL=1 is the only scenario where I end up with the 0/1 boolean between Roman and Italic style in STAT.

regarding nameID 25 ("Variations PostScript Name Prefix" string), my understanding is that it is optional/cosmetic (to help squeeze long family names) and Adobe's "PostScript Name Generation Algorithm" document explains how to make one up if 25 is missing. I don't think it should be the job of the instancer to update that.

Indeed. But not having it creates bugs in Indesign. I agree it is not the compiler job to do it, so we will have to patch that with gftools.

(we already set the style linking flags in OS/2 and head table after #2673, but we only do that when --update-font-names option is passed [which in turn requires a full STAT table] and we only consider a font "bold" or "italic" if their subfamily name after instancing contains those particles).

Exactly! This is what I am doing and aiming for. The problem is that for now it doesn't result in a fully functional Italic style. I understand it would be a bit too much for fonttools to also change the STAT for a slant axis (slnt,-12 should become ital,1), it may go against the designer's will. My thoughts would have been to integrate an argument like --is-italic, and if present then boom, changes everything.

Coming back to ital vs slnt, does your font include both ital and slnt?

It would be nice but no, it is too confusing for text editors to do that for now. But I can make a test; then ital would be using STAT format 4, right?

I'll open a new issue for caret.

[anthrotype]

ITAL=1 is the only scenario where I end up with the 0/1 boolean between Roman and Italic style in STAT.

STAT also has an axis value format that describes a range, maybe you can use that to call Italic anything that is ital>0? Not sure if that would make the instancer behave like you expect as regards setting the italic style linking bits, but I'd think so.

not having it creates bugs in Indesign.

I was dreading that answer... 😓

fonttools to also change the STAT for a slant axis (slnt,-12 should become ital,1)

hm, I'm not sure about that. 🤔