Interpolation problems in `Recursive[CASL,CRSV,MONO,slnt,wght].ttf`

[behdad]

Hello!

This is an automatically-generated report about possible interpolation problems in Recursive[CASL,CRSV,MONO,slnt,wght].ttf, as found in the Google Fonts catalog.

The particular version of the font that was tested was archive:.

To download a PDF version of this report with helpful visuals of the problems, click here; Or to view it on the GitHub website, click here.

The report follows:

Glyph horncombo was not compatible:
  Masters: '', 'CASL=1.0':
    Contour 0 start point differs: 0 in '', 48 in 'CASL=1.0'; reversed: False
Glyph one was not compatible:
  Masters: '', 'CASL=1.0':
    Contour 1 start point differs: 0 in '', 2 in 'CASL=1.0'; reversed: False
Glyph one.pnum was not compatible:
  Masters: '', 'CASL=1.0':
    Contour 1 start point differs: 0 in '', 2 in 'CASL=1.0'; reversed: False
Glyph percent was not compatible:
  Masters: 'wght=800.0061', 'wght=1000.0':
    Contour 1 structures differ: 'wght=800.0061', 'wght=1000.0'
  Masters: 'slnt=-14.0497 wght=800.0061', 'slnt=-14.0497 wght=1000.0':
    Contour 1 structures differ: 'slnt=-14.0497 wght=800.0061', 'slnt=-14.0497 wght=1000.0'
    Contour 3 structures differ: 'slnt=-14.0497 wght=800.0061', 'slnt=-14.0497 wght=1000.0'
Glyph perthousand was not compatible:
  Masters: 'wght=800.0061', 'wght=1000.0':
    Contour 2 start point differs: 0 in 'wght=800.0061', 25 in 'wght=1000.0'; reversed: True
    Contour 4 start point differs: 0 in 'wght=800.0061', 25 in 'wght=1000.0'; reversed: True
    Contour 6 start point differs: 0 in 'wght=800.0061', 25 in 'wght=1000.0'; reversed: True
  Masters: 'slnt=-14.0497 wght=800.0061', 'slnt=-14.0497 wght=1000.0':
    Contour 2 start point differs: 0 in 'slnt=-14.0497 wght=800.0061', 25 in 'slnt=-14.0497 wght=1000.0'; reversed: True
    Contour 4 start point differs: 0 in 'slnt=-14.0497 wght=800.0061', 25 in 'slnt=-14.0497 wght=1000.0'; reversed: True
    Contour 6 start point differs: 0 in 'slnt=-14.0497 wght=800.0061', 25 in 'slnt=-14.0497 wght=1000.0'; reversed: True
Glyph uni031B was not compatible:
  Masters: '', 'CASL=1.0':
    Contour 0 start point differs: 0 in '', 42 in 'CASL=1.0'; reversed: False

This report was generated using the fonttools varLib.interpolatable tool. We understand that sometimes the tool generates false-positives. Particularly for more complicated font designs. If you did not find this report useful, please apologize and ignore & close it.

To give feedback about this report, please file an issue or open a discussion at fonttools.

[arrowtype]

Hey Behdad, this is a super cool report! At least in the PDF visuals, the suggested fixes all appear to be worse than the midway interpolations. But, Recursive is definitely a complex designspace with some unusual interpolations, so it's understandable that an automated tool might misinterpret a few of parts of it.

I'll close this issue, but please feel free to let me know if there's anything you think I might be missing.

Thanks!

[behdad]

Yeah the tool cannot always propose good solution. But the first glyph seems to have kinks in that are probably unintended.

[arrowtype]

Oh, upon second look, you are correct! Thanks for pointing that out.

[behdad]

Hi again,

I'm implementing a kink-detector. It's still far from complete. But I thought I get your feedback on it.

out.pdf

[behdad]

I think my question is, is this useful? Or those differences are small enough to let go? Basically, I'm trying to calibrate the tool.

[arrowtype]

Ooh, yes, this is amazing! Definitely a useful thing to find issues that were hard to locate, before. I especially want to try this out on Shantell Sans, next.

[arrowtype]

Some of the differences here are small enough to let go of, IMO. In particular, the /percent glyph is deliberately weird, and I don't see any super noticeable issues in the /one. But, seeing the issues in the the /Euro and /uni02C8 is very informative, and those are things I'd like to improve.

[behdad]

Here's a much more aggressive version. I'm still tuning. :)

out.pdf

[behdad]

Retweaked. Shantell Sans. Catches one that doesn't matter.

out.pdf

[arrowtype]

These are great! The more aggressive version is nice, imo. Does the command line version have an arg for threshold?

[behdad]

These are great! The more aggressive version is nice, imo.

Interesting. I thought it would be too busy. I'm still tweaking.

Does the command line version have an arg for threshold?

It will have, when I'm done with it.

[behdad]

Does the command line version have an arg for threshold?

It will have, when I'm done with it.

For now I'm reusing the existing --tolerance. But I'll probably add a separate knob for just the kink checker as well.

[behdad]

Here's what I'm happy with as default tolerance.

out.pdf

[behdad]

I settled on this for default, for now. There's --kinkiness parameter you can tweak. Might tweak the defaults again. Your thoughts?

out.pdf

[behdad]

That was way too much subtle details for the catalog. I'm trying this setting now.

out.pdf

You can still get the aggressive result if you want by setting --kinkiness to 1 or above. It defaults to .5 now.

[behdad]

How about this as default?

out.pdf

[behdad]

I made the default conservative, as across the catalog it was overwhelming the reports...

[behdad]

Here's what the current output looks like. out.pdf

[arrowtype]

Hey, this output is super helpful! Thanks again for the excellent testing tool. I’ll try to fix these kinks in the coming weeks, and make a new release early in the new year.