More extensive to_kurbo

[madig]

Based on https://github.com/linebender/norad/blob/8c9038b/examples/letterspacer.rs#L523-L680, which is based on the fontTools' point pen to pen adapter pen.

This handles some more corner cases like e.g. converting successive curves without off-curves to lines, and all-off-curve paths.

I'll hold off of implementing this on Component and Glyph until I have a use-case.

[madig]

The ErrorKinds used here are really UFO specification violations. Maybe they can get their own enum later.

[github-actions]

🗜 Bloat check ⚖️

Comparing 477c428 against 3d502fe

target	old size	new size	difference
target/release/examples/open_ufo	1.71 MB	1.71 MB	---
target/debug/examples/open_ufo	7.14 MB	7.14 MB	---

[cmyr]

I would have many of the same comments on this patch as I had on #116, it might be worth reading through those?

[madig]

@cmyr i got the following, converting a QCurve with some off-curves:

let c1 = Contour::new(
    vec![
        ContourPoint::new(0.0, 0.0, PointType::OffCurve, false, None, None, None),
        ContourPoint::new(2.0, 2.0, PointType::OffCurve, false, None, None, None),
        ContourPoint::new(4.0, 4.0, PointType::OffCurve, false, None, None, None),
        ContourPoint::new(100.0, 100.0, PointType::QCurve, false, None, None, None),
    ],
    None,
    None,
);

assert_eq!(
    c1.to_kurbo().unwrap(),
    vec![
        kurbo::PathEl::MoveTo((100.0, 100.0).into()),
        kurbo::PathEl::QuadTo((0.0, 0.0).into(), (1.0, 1.0).into(),),
        kurbo::PathEl::QuadTo((2.0, 2.0).into(), (3.0, 3.0).into(),),
        kurbo::PathEl::QuadTo((4.0, 4.0).into(), (100.0, 100.0).into(),),
        kurbo::PathEl::ClosePath,
    ]
);

Does the PathEl::MoveTo((100.0, 100.0)) as the first element strike you as right or should it be (0.0, 0.0)? I copied this behavior from fontTools.

[anthrotype]

what's the status of this PR? Are you planning to finish this at some point?

[madig]

One day. I think I needed to validate that the decomposer actually does what kurbo expects, i.e. is the code right to put the MoveTo(100.0, 100.0) before the QuadTo((0.0, 0.0), (1.0, 1.0)) in https://github.com/linebender/norad/pull/132/files#diff-19795d9bd46090f88223ed97fc7eecb8758e48ce418c4b2dbed4f57bf7814c15R919-R923 or is that just how fontTools stuff does it?

[anthrotype]

I think that's correct, c1 is a closed path, so the last on-curve point (100,100) becomes the first MoveTo

[anthrotype]

(also whatever fonttools does is as normative (if not more) in UFO world as the ufo spec itself)

[madig]

Yes, but I don't know if kurbo wants something different. @cmyr?

[anthrotype]

why would kurbo ever want anything different? those conventions for converting between GLIF points to segment-oriented APIs like fonttools pens are well in place since ages and work, I don't see why diverging from them now