[variations] Support multiple axes with same tag

[behdad]

Currently doesn't work.

[behdad]

Needs no API change though.

[jfkthame]

Do renderers actually support fonts having multiple axes with the same tag? I guess I've always assumed axis tags needed to be distinct, though now that you mention it, I don't know that the spec says anything about this.

[behdad]

Do renderers actually support fonts having multiple axes with the same tag? I guess I've always assumed axis tags needed to be distinct, though now that you mention it, I don't know that the spec says anything about this.

The geniuses that discovered this trick claim that it works in Firefox and Chrome. Which might mean just CoreText, or CoreText and FreeType. cc @drott @lemzwerg FYI. We should check FreeType.

But yeah, this would allow higher-order polynomial interpolations. I'm going to recommend that spec allow this as long as at most one axis per tag is non-HIDDEN.

[lemzwerg]

Is there a sample font for this?

[behdad]

Is there a sample font for this?

I'll send you privately.

[justvanrossum]

Fwiw, this is a nice write-up, and links to a test font: https://bungeman.github.io/hoi.html

[lemzwerg]

Indeed, thanks! And the demo font seems to exhibit a problem with FreeType: If you test with ftmulti (from current git), there is a jump of the point glyph at 'Weight' value 450...

[justvanrossum]

I also see the jump with FreeType 2.10.2.

[lemzwerg]

It' an artifact of the font, I guess – if you manually enter value 450 into Samsa, you see exactly the same jump. @bungeman, can you have a look, please?

[bungeman]

Yes, the example font purposely has the 'bad' way to encode multiple curves into a single contour on the 'negative' side of the axes. In fact that's why it has that strange range on the negative side, so that the most viewers would actually hit the discontinuity (it's really easy to make a font which doesn't show the discontinuity in most viewers since the viewer may not be able to exactly hit the point).

That discontinuity is the motivation for the 'second attempt' which is demonstrated on the 'positive' side of the axes which doesn't have any discontinuity between curves on the contour. So yes, this jump is expected. Maybe should be more explicit about that in the 'example' text.

[lemzwerg]

Thanks!