tiny_skia_path/
path_geometry.rs

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
// Copyright 2006 The Android Open Source Project
// Copyright 2020 Yevhenii Reizner
//
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

//! A collection of functions to work with Bezier paths.
//!
//! Mainly for internal use. Do not rely on it!

#![allow(missing_docs)]

use crate::{Point, Transform};

use crate::f32x2_t::f32x2;
use crate::floating_point::FLOAT_PI;
use crate::scalar::{Scalar, SCALAR_NEARLY_ZERO, SCALAR_ROOT_2_OVER_2};

use crate::floating_point::{NormalizedF32, NormalizedF32Exclusive};
use crate::path_builder::PathDirection;

#[cfg(all(not(feature = "std"), feature = "no-std-float"))]
use crate::NoStdFloat;

// use for : eval(t) == A * t^2 + B * t + C
#[derive(Clone, Copy, Default, Debug)]
pub struct QuadCoeff {
    pub a: f32x2,
    pub b: f32x2,
    pub c: f32x2,
}

impl QuadCoeff {
    pub fn from_points(points: &[Point; 3]) -> Self {
        let c = points[0].to_f32x2();
        let p1 = points[1].to_f32x2();
        let p2 = points[2].to_f32x2();
        let b = times_2(p1 - c);
        let a = p2 - times_2(p1) + c;

        QuadCoeff { a, b, c }
    }

    pub fn eval(&self, t: f32x2) -> f32x2 {
        (self.a * t + self.b) * t + self.c
    }
}

#[derive(Clone, Copy, Default, Debug)]
pub struct CubicCoeff {
    pub a: f32x2,
    pub b: f32x2,
    pub c: f32x2,
    pub d: f32x2,
}

impl CubicCoeff {
    pub fn from_points(points: &[Point; 4]) -> Self {
        let p0 = points[0].to_f32x2();
        let p1 = points[1].to_f32x2();
        let p2 = points[2].to_f32x2();
        let p3 = points[3].to_f32x2();
        let three = f32x2::splat(3.0);

        CubicCoeff {
            a: p3 + three * (p1 - p2) - p0,
            b: three * (p2 - times_2(p1) + p0),
            c: three * (p1 - p0),
            d: p0,
        }
    }

    pub fn eval(&self, t: f32x2) -> f32x2 {
        ((self.a * t + self.b) * t + self.c) * t + self.d
    }
}

// TODO: to a custom type?
pub fn new_t_values() -> [NormalizedF32Exclusive; 3] {
    [NormalizedF32Exclusive::ANY; 3]
}

pub fn chop_quad_at(src: &[Point], t: NormalizedF32Exclusive, dst: &mut [Point; 5]) {
    let p0 = src[0].to_f32x2();
    let p1 = src[1].to_f32x2();
    let p2 = src[2].to_f32x2();
    let tt = f32x2::splat(t.get());

    let p01 = interp(p0, p1, tt);
    let p12 = interp(p1, p2, tt);

    dst[0] = Point::from_f32x2(p0);
    dst[1] = Point::from_f32x2(p01);
    dst[2] = Point::from_f32x2(interp(p01, p12, tt));
    dst[3] = Point::from_f32x2(p12);
    dst[4] = Point::from_f32x2(p2);
}

// From Numerical Recipes in C.
//
// Q = -1/2 (B + sign(B) sqrt[B*B - 4*A*C])
// x1 = Q / A
// x2 = C / Q
pub fn find_unit_quad_roots(
    a: f32,
    b: f32,
    c: f32,
    roots: &mut [NormalizedF32Exclusive; 3],
) -> usize {
    if a == 0.0 {
        if let Some(r) = valid_unit_divide(-c, b) {
            roots[0] = r;
            return 1;
        } else {
            return 0;
        }
    }

    // use doubles so we don't overflow temporarily trying to compute R
    let mut dr = f64::from(b) * f64::from(b) - 4.0 * f64::from(a) * f64::from(c);
    if dr < 0.0 {
        return 0;
    }
    dr = dr.sqrt();
    let r = dr as f32;
    if !r.is_finite() {
        return 0;
    }

    let q = if b < 0.0 {
        -(b - r) / 2.0
    } else {
        -(b + r) / 2.0
    };

    let mut roots_offset = 0;
    if let Some(r) = valid_unit_divide(q, a) {
        roots[roots_offset] = r;
        roots_offset += 1;
    }

    if let Some(r) = valid_unit_divide(c, q) {
        roots[roots_offset] = r;
        roots_offset += 1;
    }

    if roots_offset == 2 {
        if roots[0].get() > roots[1].get() {
            roots.swap(0, 1);
        } else if roots[0] == roots[1] {
            // nearly-equal?
            roots_offset -= 1; // skip the double root
        }
    }

    roots_offset
}

pub fn chop_cubic_at2(src: &[Point; 4], t: NormalizedF32Exclusive, dst: &mut [Point]) {
    let p0 = src[0].to_f32x2();
    let p1 = src[1].to_f32x2();
    let p2 = src[2].to_f32x2();
    let p3 = src[3].to_f32x2();
    let tt = f32x2::splat(t.get());

    let ab = interp(p0, p1, tt);
    let bc = interp(p1, p2, tt);
    let cd = interp(p2, p3, tt);
    let abc = interp(ab, bc, tt);
    let bcd = interp(bc, cd, tt);
    let abcd = interp(abc, bcd, tt);

    dst[0] = Point::from_f32x2(p0);
    dst[1] = Point::from_f32x2(ab);
    dst[2] = Point::from_f32x2(abc);
    dst[3] = Point::from_f32x2(abcd);
    dst[4] = Point::from_f32x2(bcd);
    dst[5] = Point::from_f32x2(cd);
    dst[6] = Point::from_f32x2(p3);
}

// Quad'(t) = At + B, where
// A = 2(a - 2b + c)
// B = 2(b - a)
// Solve for t, only if it fits between 0 < t < 1
pub(crate) fn find_quad_extrema(a: f32, b: f32, c: f32) -> Option<NormalizedF32Exclusive> {
    // At + B == 0
    // t = -B / A
    valid_unit_divide(a - b, a - b - b + c)
}

pub fn valid_unit_divide(mut numer: f32, mut denom: f32) -> Option<NormalizedF32Exclusive> {
    if numer < 0.0 {
        numer = -numer;
        denom = -denom;
    }

    if denom == 0.0 || numer == 0.0 || numer >= denom {
        return None;
    }

    let r = numer / denom;
    NormalizedF32Exclusive::new(r)
}

fn interp(v0: f32x2, v1: f32x2, t: f32x2) -> f32x2 {
    v0 + (v1 - v0) * t
}

fn times_2(value: f32x2) -> f32x2 {
    value + value
}

// F(t)    = a (1 - t) ^ 2 + 2 b t (1 - t) + c t ^ 2
// F'(t)   = 2 (b - a) + 2 (a - 2b + c) t
// F''(t)  = 2 (a - 2b + c)
//
// A = 2 (b - a)
// B = 2 (a - 2b + c)
//
// Maximum curvature for a quadratic means solving
// Fx' Fx'' + Fy' Fy'' = 0
//
// t = - (Ax Bx + Ay By) / (Bx ^ 2 + By ^ 2)
pub(crate) fn find_quad_max_curvature(src: &[Point; 3]) -> NormalizedF32 {
    let ax = src[1].x - src[0].x;
    let ay = src[1].y - src[0].y;
    let bx = src[0].x - src[1].x - src[1].x + src[2].x;
    let by = src[0].y - src[1].y - src[1].y + src[2].y;

    let mut numer = -(ax * bx + ay * by);
    let mut denom = bx * bx + by * by;
    if denom < 0.0 {
        numer = -numer;
        denom = -denom;
    }

    if numer <= 0.0 {
        return NormalizedF32::ZERO;
    }

    if numer >= denom {
        // Also catches denom=0
        return NormalizedF32::ONE;
    }

    let t = numer / denom;
    NormalizedF32::new(t).unwrap()
}

pub(crate) fn eval_quad_at(src: &[Point; 3], t: NormalizedF32) -> Point {
    Point::from_f32x2(QuadCoeff::from_points(src).eval(f32x2::splat(t.get())))
}

pub(crate) fn eval_quad_tangent_at(src: &[Point; 3], tol: NormalizedF32) -> Point {
    // The derivative equation is 2(b - a +(a - 2b +c)t). This returns a
    // zero tangent vector when t is 0 or 1, and the control point is equal
    // to the end point. In this case, use the quad end points to compute the tangent.
    if (tol == NormalizedF32::ZERO && src[0] == src[1])
        || (tol == NormalizedF32::ONE && src[1] == src[2])
    {
        return src[2] - src[0];
    }

    let p0 = src[0].to_f32x2();
    let p1 = src[1].to_f32x2();
    let p2 = src[2].to_f32x2();

    let b = p1 - p0;
    let a = p2 - p1 - b;
    let t = a * f32x2::splat(tol.get()) + b;

    Point::from_f32x2(t + t)
}

// Looking for F' dot F'' == 0
//
// A = b - a
// B = c - 2b + a
// C = d - 3c + 3b - a
//
// F' = 3Ct^2 + 6Bt + 3A
// F'' = 6Ct + 6B
//
// F' dot F'' -> CCt^3 + 3BCt^2 + (2BB + CA)t + AB
pub fn find_cubic_max_curvature<'a>(
    src: &[Point; 4],
    t_values: &'a mut [NormalizedF32; 3],
) -> &'a [NormalizedF32] {
    let mut coeff_x = formulate_f1_dot_f2(&[src[0].x, src[1].x, src[2].x, src[3].x]);
    let coeff_y = formulate_f1_dot_f2(&[src[0].y, src[1].y, src[2].y, src[3].y]);

    for i in 0..4 {
        coeff_x[i] += coeff_y[i];
    }

    let len = solve_cubic_poly(&coeff_x, t_values);
    &t_values[0..len]
}

// Looking for F' dot F'' == 0
//
// A = b - a
// B = c - 2b + a
// C = d - 3c + 3b - a
//
// F' = 3Ct^2 + 6Bt + 3A
// F'' = 6Ct + 6B
//
// F' dot F'' -> CCt^3 + 3BCt^2 + (2BB + CA)t + AB
fn formulate_f1_dot_f2(src: &[f32; 4]) -> [f32; 4] {
    let a = src[1] - src[0];
    let b = src[2] - 2.0 * src[1] + src[0];
    let c = src[3] + 3.0 * (src[1] - src[2]) - src[0];

    [c * c, 3.0 * b * c, 2.0 * b * b + c * a, a * b]
}

/// Solve coeff(t) == 0, returning the number of roots that lie withing 0 < t < 1.
/// coeff[0]t^3 + coeff[1]t^2 + coeff[2]t + coeff[3]
///
/// Eliminates repeated roots (so that all t_values are distinct, and are always
/// in increasing order.
fn solve_cubic_poly(coeff: &[f32; 4], t_values: &mut [NormalizedF32; 3]) -> usize {
    if coeff[0].is_nearly_zero() {
        // we're just a quadratic
        let mut tmp_t = new_t_values();
        let count = find_unit_quad_roots(coeff[1], coeff[2], coeff[3], &mut tmp_t);
        for i in 0..count {
            t_values[i] = tmp_t[i].to_normalized();
        }

        return count;
    }

    debug_assert!(coeff[0] != 0.0);

    let inva = coeff[0].invert();
    let a = coeff[1] * inva;
    let b = coeff[2] * inva;
    let c = coeff[3] * inva;

    let q = (a * a - b * 3.0) / 9.0;
    let r = (2.0 * a * a * a - 9.0 * a * b + 27.0 * c) / 54.0;

    let q3 = q * q * q;
    let r2_minus_q3 = r * r - q3;
    let adiv3 = a / 3.0;

    if r2_minus_q3 < 0.0 {
        // we have 3 real roots
        // the divide/root can, due to finite precisions, be slightly outside of -1...1
        let theta = (r / q3.sqrt()).bound(-1.0, 1.0).acos();
        let neg2_root_q = -2.0 * q.sqrt();

        t_values[0] = NormalizedF32::new_clamped(neg2_root_q * (theta / 3.0).cos() - adiv3);
        t_values[1] = NormalizedF32::new_clamped(
            neg2_root_q * ((theta + 2.0 * FLOAT_PI) / 3.0).cos() - adiv3,
        );
        t_values[2] = NormalizedF32::new_clamped(
            neg2_root_q * ((theta - 2.0 * FLOAT_PI) / 3.0).cos() - adiv3,
        );

        // now sort the roots
        sort_array3(t_values);
        collapse_duplicates3(t_values)
    } else {
        // we have 1 real root
        let mut a = r.abs() + r2_minus_q3.sqrt();
        a = scalar_cube_root(a);
        if r > 0.0 {
            a = -a;
        }

        if a != 0.0 {
            a += q / a;
        }

        t_values[0] = NormalizedF32::new_clamped(a - adiv3);
        1
    }
}

fn sort_array3(array: &mut [NormalizedF32; 3]) {
    if array[0] > array[1] {
        array.swap(0, 1);
    }

    if array[1] > array[2] {
        array.swap(1, 2);
    }

    if array[0] > array[1] {
        array.swap(0, 1);
    }
}

fn collapse_duplicates3(array: &mut [NormalizedF32; 3]) -> usize {
    let mut len = 3;

    if array[1] == array[2] {
        len = 2;
    }

    if array[0] == array[1] {
        len = 1;
    }

    len
}

fn scalar_cube_root(x: f32) -> f32 {
    x.powf(0.3333333)
}

// This is SkEvalCubicAt split into three functions.
pub(crate) fn eval_cubic_pos_at(src: &[Point; 4], t: NormalizedF32) -> Point {
    Point::from_f32x2(CubicCoeff::from_points(src).eval(f32x2::splat(t.get())))
}

// This is SkEvalCubicAt split into three functions.
pub(crate) fn eval_cubic_tangent_at(src: &[Point; 4], t: NormalizedF32) -> Point {
    // The derivative equation returns a zero tangent vector when t is 0 or 1, and the
    // adjacent control point is equal to the end point. In this case, use the
    // next control point or the end points to compute the tangent.
    if (t.get() == 0.0 && src[0] == src[1]) || (t.get() == 1.0 && src[2] == src[3]) {
        let mut tangent = if t.get() == 0.0 {
            src[2] - src[0]
        } else {
            src[3] - src[1]
        };

        if tangent.x == 0.0 && tangent.y == 0.0 {
            tangent = src[3] - src[0];
        }

        tangent
    } else {
        eval_cubic_derivative(src, t)
    }
}

fn eval_cubic_derivative(src: &[Point; 4], t: NormalizedF32) -> Point {
    let p0 = src[0].to_f32x2();
    let p1 = src[1].to_f32x2();
    let p2 = src[2].to_f32x2();
    let p3 = src[3].to_f32x2();

    let coeff = QuadCoeff {
        a: p3 + f32x2::splat(3.0) * (p1 - p2) - p0,
        b: times_2(p2 - times_2(p1) + p0),
        c: p1 - p0,
    };

    Point::from_f32x2(coeff.eval(f32x2::splat(t.get())))
}

// Cubic'(t) = At^2 + Bt + C, where
// A = 3(-a + 3(b - c) + d)
// B = 6(a - 2b + c)
// C = 3(b - a)
// Solve for t, keeping only those that fit between 0 < t < 1
pub(crate) fn find_cubic_extrema(
    a: f32,
    b: f32,
    c: f32,
    d: f32,
    t_values: &mut [NormalizedF32Exclusive; 3],
) -> usize {
    // we divide A,B,C by 3 to simplify
    let aa = d - a + 3.0 * (b - c);
    let bb = 2.0 * (a - b - b + c);
    let cc = b - a;

    find_unit_quad_roots(aa, bb, cc, t_values)
}

// http://www.faculty.idc.ac.il/arik/quality/appendixA.html
//
// Inflection means that curvature is zero.
// Curvature is [F' x F''] / [F'^3]
// So we solve F'x X F''y - F'y X F''y == 0
// After some canceling of the cubic term, we get
// A = b - a
// B = c - 2b + a
// C = d - 3c + 3b - a
// (BxCy - ByCx)t^2 + (AxCy - AyCx)t + AxBy - AyBx == 0
pub(crate) fn find_cubic_inflections<'a>(
    src: &[Point; 4],
    t_values: &'a mut [NormalizedF32Exclusive; 3],
) -> &'a [NormalizedF32Exclusive] {
    let ax = src[1].x - src[0].x;
    let ay = src[1].y - src[0].y;
    let bx = src[2].x - 2.0 * src[1].x + src[0].x;
    let by = src[2].y - 2.0 * src[1].y + src[0].y;
    let cx = src[3].x + 3.0 * (src[1].x - src[2].x) - src[0].x;
    let cy = src[3].y + 3.0 * (src[1].y - src[2].y) - src[0].y;

    let len = find_unit_quad_roots(
        bx * cy - by * cx,
        ax * cy - ay * cx,
        ax * by - ay * bx,
        t_values,
    );

    &t_values[0..len]
}

// Return location (in t) of cubic cusp, if there is one.
// Note that classify cubic code does not reliably return all cusp'd cubics, so
// it is not called here.
pub(crate) fn find_cubic_cusp(src: &[Point; 4]) -> Option<NormalizedF32Exclusive> {
    // When the adjacent control point matches the end point, it behaves as if
    // the cubic has a cusp: there's a point of max curvature where the derivative
    // goes to zero. Ideally, this would be where t is zero or one, but math
    // error makes not so. It is not uncommon to create cubics this way; skip them.
    if src[0] == src[1] {
        return None;
    }

    if src[2] == src[3] {
        return None;
    }

    // Cubics only have a cusp if the line segments formed by the control and end points cross.
    // Detect crossing if line ends are on opposite sides of plane formed by the other line.
    if on_same_side(src, 0, 2) || on_same_side(src, 2, 0) {
        return None;
    }

    // Cubics may have multiple points of maximum curvature, although at most only
    // one is a cusp.
    let mut t_values = [NormalizedF32::ZERO; 3];
    let max_curvature = find_cubic_max_curvature(src, &mut t_values);
    for test_t in max_curvature {
        if 0.0 >= test_t.get() || test_t.get() >= 1.0 {
            // no need to consider max curvature on the end
            continue;
        }

        // A cusp is at the max curvature, and also has a derivative close to zero.
        // Choose the 'close to zero' meaning by comparing the derivative length
        // with the overall cubic size.
        let d_pt = eval_cubic_derivative(src, *test_t);
        let d_pt_magnitude = d_pt.length_sqd();
        let precision = calc_cubic_precision(src);
        if d_pt_magnitude < precision {
            // All three max curvature t values may be close to the cusp;
            // return the first one.
            return Some(NormalizedF32Exclusive::new_bounded(test_t.get()));
        }
    }

    None
}

// Returns true if both points src[testIndex], src[testIndex+1] are in the same half plane defined
// by the line segment src[lineIndex], src[lineIndex+1].
fn on_same_side(src: &[Point; 4], test_index: usize, line_index: usize) -> bool {
    let origin = src[line_index];
    let line = src[line_index + 1] - origin;
    let mut crosses = [0.0, 0.0];
    for index in 0..2 {
        let test_line = src[test_index + index] - origin;
        crosses[index] = line.cross(test_line);
    }

    crosses[0] * crosses[1] >= 0.0
}

// Returns a constant proportional to the dimensions of the cubic.
// Constant found through experimentation -- maybe there's a better way....
fn calc_cubic_precision(src: &[Point; 4]) -> f32 {
    (src[1].distance_to_sqd(src[0])
        + src[2].distance_to_sqd(src[1])
        + src[3].distance_to_sqd(src[2]))
        * 1e-8
}

#[derive(Copy, Clone, Default, Debug)]
pub(crate) struct Conic {
    pub points: [Point; 3],
    pub weight: f32,
}

impl Conic {
    pub fn new(pt0: Point, pt1: Point, pt2: Point, weight: f32) -> Self {
        Conic {
            points: [pt0, pt1, pt2],
            weight,
        }
    }

    pub fn from_points(points: &[Point], weight: f32) -> Self {
        Conic {
            points: [points[0], points[1], points[2]],
            weight,
        }
    }

    fn compute_quad_pow2(&self, tolerance: f32) -> Option<u8> {
        if tolerance < 0.0 || !tolerance.is_finite() {
            return None;
        }

        if !self.points[0].is_finite() || !self.points[1].is_finite() || !self.points[2].is_finite()
        {
            return None;
        }

        // Limit the number of suggested quads to approximate a conic
        const MAX_CONIC_TO_QUAD_POW2: usize = 4;

        // "High order approximation of conic sections by quadratic splines"
        // by Michael Floater, 1993
        let a = self.weight - 1.0;
        let k = a / (4.0 * (2.0 + a));
        let x = k * (self.points[0].x - 2.0 * self.points[1].x + self.points[2].x);
        let y = k * (self.points[0].y - 2.0 * self.points[1].y + self.points[2].y);

        let mut error = (x * x + y * y).sqrt();
        let mut pow2 = 0;
        for _ in 0..MAX_CONIC_TO_QUAD_POW2 {
            if error <= tolerance {
                break;
            }

            error *= 0.25;
            pow2 += 1;
        }

        // Unlike Skia, we always expect `pow2` to be at least 1.
        // Otherwise it produces ugly results.
        Some(pow2.max(1))
    }

    // Chop this conic into N quads, stored continuously in pts[], where
    // N = 1 << pow2. The amount of storage needed is (1 + 2 * N)
    pub fn chop_into_quads_pow2(&self, pow2: u8, points: &mut [Point]) -> u8 {
        debug_assert!(pow2 < 5);

        points[0] = self.points[0];
        subdivide(self, &mut points[1..], pow2);

        let quad_count = 1 << pow2;
        let pt_count = 2 * quad_count + 1;
        if points.iter().take(pt_count).any(|n| !n.is_finite()) {
            // if we generated a non-finite, pin ourselves to the middle of the hull,
            // as our first and last are already on the first/last pts of the hull.
            for p in points.iter_mut().take(pt_count - 1).skip(1) {
                *p = self.points[1];
            }
        }

        1 << pow2
    }

    fn chop(&self) -> (Conic, Conic) {
        let scale = f32x2::splat((1.0 + self.weight).invert());
        let new_w = subdivide_weight_value(self.weight);

        let p0 = self.points[0].to_f32x2();
        let p1 = self.points[1].to_f32x2();
        let p2 = self.points[2].to_f32x2();
        let ww = f32x2::splat(self.weight);

        let wp1 = ww * p1;
        let m = (p0 + times_2(wp1) + p2) * scale * f32x2::splat(0.5);
        let mut m_pt = Point::from_f32x2(m);
        if !m_pt.is_finite() {
            let w_d = self.weight as f64;
            let w_2 = w_d * 2.0;
            let scale_half = 1.0 / (1.0 + w_d) * 0.5;
            m_pt.x = ((self.points[0].x as f64
                + w_2 * self.points[1].x as f64
                + self.points[2].x as f64)
                * scale_half) as f32;

            m_pt.y = ((self.points[0].y as f64
                + w_2 * self.points[1].y as f64
                + self.points[2].y as f64)
                * scale_half) as f32;
        }

        (
            Conic {
                points: [self.points[0], Point::from_f32x2((p0 + wp1) * scale), m_pt],
                weight: new_w,
            },
            Conic {
                points: [m_pt, Point::from_f32x2((wp1 + p2) * scale), self.points[2]],
                weight: new_w,
            },
        )
    }

    pub fn build_unit_arc(
        u_start: Point,
        u_stop: Point,
        dir: PathDirection,
        user_transform: Transform,
        dst: &mut [Conic; 5],
    ) -> Option<&[Conic]> {
        // rotate by x,y so that u_start is (1.0)
        let x = u_start.dot(u_stop);
        let mut y = u_start.cross(u_stop);

        let abs_y = y.abs();

        // check for (effectively) coincident vectors
        // this can happen if our angle is nearly 0 or nearly 180 (y == 0)
        // ... we use the dot-prod to distinguish between 0 and 180 (x > 0)
        if abs_y <= SCALAR_NEARLY_ZERO
            && x > 0.0
            && ((y >= 0.0 && dir == PathDirection::CW) || (y <= 0.0 && dir == PathDirection::CCW))
        {
            return None;
        }

        if dir == PathDirection::CCW {
            y = -y;
        }

        // We decide to use 1-conic per quadrant of a circle. What quadrant does [xy] lie in?
        //      0 == [0  .. 90)
        //      1 == [90 ..180)
        //      2 == [180..270)
        //      3 == [270..360)
        //
        let mut quadrant = 0;
        if y == 0.0 {
            quadrant = 2; // 180
            debug_assert!((x + 1.0) <= SCALAR_NEARLY_ZERO);
        } else if x == 0.0 {
            debug_assert!(abs_y - 1.0 <= SCALAR_NEARLY_ZERO);
            quadrant = if y > 0.0 { 1 } else { 3 }; // 90 / 270
        } else {
            if y < 0.0 {
                quadrant += 2;
            }

            if (x < 0.0) != (y < 0.0) {
                quadrant += 1;
            }
        }

        let quadrant_points = [
            Point::from_xy(1.0, 0.0),
            Point::from_xy(1.0, 1.0),
            Point::from_xy(0.0, 1.0),
            Point::from_xy(-1.0, 1.0),
            Point::from_xy(-1.0, 0.0),
            Point::from_xy(-1.0, -1.0),
            Point::from_xy(0.0, -1.0),
            Point::from_xy(1.0, -1.0),
        ];

        const QUADRANT_WEIGHT: f32 = SCALAR_ROOT_2_OVER_2;

        let mut conic_count = quadrant;
        for i in 0..conic_count {
            dst[i] = Conic::from_points(&quadrant_points[i * 2..], QUADRANT_WEIGHT);
        }

        // Now compute any remaing (sub-90-degree) arc for the last conic
        let final_pt = Point::from_xy(x, y);
        let last_q = quadrant_points[quadrant * 2]; // will already be a unit-vector
        let dot = last_q.dot(final_pt);
        debug_assert!(0.0 <= dot && dot <= 1.0 + SCALAR_NEARLY_ZERO);

        if dot < 1.0 {
            let mut off_curve = Point::from_xy(last_q.x + x, last_q.y + y);
            // compute the bisector vector, and then rescale to be the off-curve point.
            // we compute its length from cos(theta/2) = length / 1, using half-angle identity we get
            // length = sqrt(2 / (1 + cos(theta)). We already have cos() when to computed the dot.
            // This is nice, since our computed weight is cos(theta/2) as well!
            let cos_theta_over_2 = ((1.0 + dot) / 2.0).sqrt();
            off_curve.set_length(cos_theta_over_2.invert());
            if !last_q.almost_equal(off_curve) {
                dst[conic_count] = Conic::new(last_q, off_curve, final_pt, cos_theta_over_2);
                conic_count += 1;
            }
        }

        // now handle counter-clockwise and the initial unitStart rotation
        let mut transform = Transform::from_sin_cos(u_start.y, u_start.x);
        if dir == PathDirection::CCW {
            transform = transform.pre_scale(1.0, -1.0);
        }

        transform = transform.post_concat(user_transform);

        for conic in dst.iter_mut().take(conic_count) {
            transform.map_points(&mut conic.points);
        }

        if conic_count == 0 {
            None
        } else {
            Some(&dst[0..conic_count])
        }
    }
}

fn subdivide_weight_value(w: f32) -> f32 {
    (0.5 + w * 0.5).sqrt()
}

fn subdivide<'a>(src: &Conic, mut points: &'a mut [Point], mut level: u8) -> &'a mut [Point] {
    if level == 0 {
        points[0] = src.points[1];
        points[1] = src.points[2];
        &mut points[2..]
    } else {
        let mut dst = src.chop();

        let start_y = src.points[0].y;
        let end_y = src.points[2].y;
        if between(start_y, src.points[1].y, end_y) {
            // If the input is monotonic and the output is not, the scan converter hangs.
            // Ensure that the chopped conics maintain their y-order.
            let mid_y = dst.0.points[2].y;
            if !between(start_y, mid_y, end_y) {
                // If the computed midpoint is outside the ends, move it to the closer one.
                let closer_y = if (mid_y - start_y).abs() < (mid_y - end_y).abs() {
                    start_y
                } else {
                    end_y
                };
                dst.0.points[2].y = closer_y;
                dst.1.points[0].y = closer_y;
            }

            if !between(start_y, dst.0.points[1].y, dst.0.points[2].y) {
                // If the 1st control is not between the start and end, put it at the start.
                // This also reduces the quad to a line.
                dst.0.points[1].y = start_y;
            }

            if !between(dst.1.points[0].y, dst.1.points[1].y, end_y) {
                // If the 2nd control is not between the start and end, put it at the end.
                // This also reduces the quad to a line.
                dst.1.points[1].y = end_y;
            }

            // Verify that all five points are in order.
            debug_assert!(between(start_y, dst.0.points[1].y, dst.0.points[2].y));
            debug_assert!(between(
                dst.0.points[1].y,
                dst.0.points[2].y,
                dst.1.points[1].y
            ));
            debug_assert!(between(dst.0.points[2].y, dst.1.points[1].y, end_y));
        }

        level -= 1;
        points = subdivide(&dst.0, points, level);
        subdivide(&dst.1, points, level)
    }
}

// This was originally developed and tested for pathops: see SkOpTypes.h
// returns true if (a <= b <= c) || (a >= b >= c)
fn between(a: f32, b: f32, c: f32) -> bool {
    (a - b) * (c - b) <= 0.0
}

pub(crate) struct AutoConicToQuads {
    pub points: [Point; 64],
    pub len: u8, // the number of quads
}

impl AutoConicToQuads {
    pub fn compute(pt0: Point, pt1: Point, pt2: Point, weight: f32) -> Option<Self> {
        let conic = Conic::new(pt0, pt1, pt2, weight);
        let pow2 = conic.compute_quad_pow2(0.25)?;
        let mut points = [Point::zero(); 64];
        let len = conic.chop_into_quads_pow2(pow2, &mut points);
        Some(AutoConicToQuads { points, len })
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn eval_cubic_at_1() {
        let src = [
            Point::from_xy(30.0, 40.0),
            Point::from_xy(30.0, 40.0),
            Point::from_xy(171.0, 45.0),
            Point::from_xy(180.0, 155.0),
        ];

        assert_eq!(
            eval_cubic_pos_at(&src, NormalizedF32::ZERO),
            Point::from_xy(30.0, 40.0)
        );
        assert_eq!(
            eval_cubic_tangent_at(&src, NormalizedF32::ZERO),
            Point::from_xy(141.0, 5.0)
        );
    }

    #[test]
    fn find_cubic_max_curvature_1() {
        let src = [
            Point::from_xy(20.0, 160.0),
            Point::from_xy(20.0001, 160.0),
            Point::from_xy(160.0, 20.0),
            Point::from_xy(160.0001, 20.0),
        ];

        let mut t_values = [NormalizedF32::ZERO; 3];
        let t_values = find_cubic_max_curvature(&src, &mut t_values);

        assert_eq!(
            &t_values,
            &[
                NormalizedF32::ZERO,
                NormalizedF32::new_clamped(0.5),
                NormalizedF32::ONE,
            ]
        );
    }
}