summaryrefslogtreecommitdiffstats
path: root/src/2geom/ellipse.cpp
blob: 42cb36dcb697e1f9c598e7d31f6fe24b0fe26840 (plain)
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
/** @file
 * @brief Ellipse shape
 *//*
 * Authors:
 *   Marco Cecchetti <mrcekets at gmail.com>
 *   Krzysztof Kosiński <tweenk.pl@gmail.com>
 *
 * Copyright 2008-2014 Authors
 *
 * This library is free software; you can redistribute it and/or
 * modify it either under the terms of the GNU Lesser General Public
 * License version 2.1 as published by the Free Software Foundation
 * (the "LGPL") or, at your option, under the terms of the Mozilla
 * Public License Version 1.1 (the "MPL"). If you do not alter this
 * notice, a recipient may use your version of this file under either
 * the MPL or the LGPL.
 *
 * You should have received a copy of the LGPL along with this library
 * in the file COPYING-LGPL-2.1; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 * You should have received a copy of the MPL along with this library
 * in the file COPYING-MPL-1.1
 *
 * The contents of this file are subject to the Mozilla Public License
 * Version 1.1 (the "License"); you may not use this file except in
 * compliance with the License. You may obtain a copy of the License at
 * http://www.mozilla.org/MPL/
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY
 * OF ANY KIND, either express or implied. See the LGPL or the MPL for
 * the specific language governing rights and limitations.
 */

#include <2geom/conicsec.h>
#include <2geom/ellipse.h>
#include <2geom/elliptical-arc.h>
#include <2geom/numeric/fitting-tool.h>
#include <2geom/numeric/fitting-model.h>

namespace Geom {

Ellipse::Ellipse(Geom::Circle const &c)
    : _center(c.center())
    , _rays(c.radius(), c.radius())
    , _angle(0)
{}

void Ellipse::setCoefficients(double A, double B, double C, double D, double E, double F)
{
    double den = 4*A*C - B*B;
    if (den == 0) {
        THROW_RANGEERROR("den == 0, while computing ellipse centre");
    }
    _center[X] = (B*E - 2*C*D) / den;
    _center[Y] = (B*D - 2*A*E) / den;

    // evaluate the a coefficient of the ellipse equation in normal form
    // E(x,y) = a*(x-cx)^2 + b*(x-cx)*(y-cy) + c*(y-cy)^2 = 1
    // where b = a*B , c = a*C, (cx,cy) == centre
    double num =   A * sqr(_center[X])
                 + B * _center[X] * _center[Y]
                 + C * sqr(_center[Y])
                 - F;


    //evaluate ellipse rotation angle
    _angle = std::atan2( -B, -(A - C) )/2;

    // evaluate the length of the ellipse rays
    double sinrot, cosrot;
    sincos(_angle, sinrot, cosrot);
    double cos2 = cosrot * cosrot;
    double sin2 = sinrot * sinrot;
    double cossin = cosrot * sinrot;

    den = A * cos2 + B * cossin + C * sin2;
    if (den == 0) {
        THROW_RANGEERROR("den == 0, while computing 'rx' coefficient");
    }
    double rx2 = num / den;
    if (rx2 < 0) {
        THROW_RANGEERROR("rx2 < 0, while computing 'rx' coefficient");
    }
    _rays[X] = std::sqrt(rx2);

    den = C * cos2 - B * cossin + A * sin2;
    if (den == 0) {
        THROW_RANGEERROR("den == 0, while computing 'ry' coefficient");
    }
    double ry2 = num / den;
    if (ry2 < 0) {
        THROW_RANGEERROR("ry2 < 0, while computing 'rx' coefficient");
    }
    _rays[Y] = std::sqrt(ry2);

    // the solution is not unique so we choose always the ellipse
    // with a rotation angle between 0 and PI/2
    makeCanonical();
}

Point Ellipse::initialPoint() const
{
    Coord sinrot, cosrot;
    sincos(_angle, sinrot, cosrot);
    Point p(ray(X) * cosrot + center(X), ray(X) * sinrot + center(Y));
    return p;
}


Affine Ellipse::unitCircleTransform() const
{
    Affine ret = Scale(ray(X), ray(Y)) * Rotate(_angle);
    ret.setTranslation(center());
    return ret;
}

Affine Ellipse::inverseUnitCircleTransform() const
{
    if (ray(X) == 0 || ray(Y) == 0) {
        THROW_RANGEERROR("a degenerate ellipse doesn't have an inverse unit circle transform");
    }
    Affine ret = Translate(-center()) * Rotate(-_angle) * Scale(1/ray(X), 1/ray(Y));
    return ret;
}


LineSegment Ellipse::axis(Dim2 d) const
{
    Point a(0, 0), b(0, 0);
    a[d] = -1;
    b[d] = 1;
    LineSegment ls(a, b);
    ls.transform(unitCircleTransform());
    return ls;
}

LineSegment Ellipse::semiaxis(Dim2 d, int sign) const
{
    Point a(0, 0), b(0, 0);
    b[d] = sgn(sign);
    LineSegment ls(a, b);
    ls.transform(unitCircleTransform());
    return ls;
}

Rect Ellipse::boundsExact() const
{
    auto const trans = unitCircleTransform();

    auto proj_bounds = [&] (Dim2 d) {
        // The dth coordinate function pulls back to trans[d] * x + trans[d + 2] * y + trans[d + 4]
        // in the coordinate system where the ellipse is a unit circle. We compute its range of
        // values on the unit circle.
        auto const r = std::hypot(trans[d], trans[d + 2]);
        auto const mid = trans[d + 4];
        return Interval(mid - r, mid + r);
    };

    return { proj_bounds(X), proj_bounds(Y) };
}

Rect Ellipse::boundsFast() const
{
    // Every ellipse is contained in the circle with the same center and radius
    // equal to the larger of the two rays. We return the bounding square
    // of this circle (this is really fast but only exact for circles).
    auto const larger_ray = (ray(X) > ray(Y) ? ray(X) : ray(Y));
    assert(larger_ray >= 0.0);
    auto const rr = Point(larger_ray, larger_ray);
    return Rect(_center - rr, _center + rr);
}

std::vector<double> Ellipse::coefficients() const
{
    std::vector<double> c(6);
    coefficients(c[0], c[1], c[2], c[3], c[4], c[5]);
    return c;
}

void Ellipse::coefficients(Coord &A, Coord &B, Coord &C, Coord &D, Coord &E, Coord &F) const
{
    if (ray(X) == 0 || ray(Y) == 0) {
        THROW_RANGEERROR("a degenerate ellipse doesn't have an implicit form");
    }

    double cosrot, sinrot;
    sincos(_angle, sinrot, cosrot);
    double cos2 = cosrot * cosrot;
    double sin2 = sinrot * sinrot;
    double cossin = cosrot * sinrot;
    double invrx2 = 1 / (ray(X) * ray(X));
    double invry2 = 1 / (ray(Y) * ray(Y));

    A = invrx2 * cos2 + invry2 * sin2;
    B = 2 * (invrx2 - invry2) * cossin;
    C = invrx2 * sin2 + invry2 * cos2;
    D = -2 * A * center(X) - B * center(Y);
    E = -2 * C * center(Y) - B * center(X);
    F =   A * center(X) * center(X)
        + B * center(X) * center(Y)
        + C * center(Y) * center(Y)
        - 1;
}


void Ellipse::fit(std::vector<Point> const &points)
{
    size_t sz = points.size();
    if (sz < 5) {
        THROW_RANGEERROR("fitting error: too few points passed");
    }
    NL::LFMEllipse model;
    NL::least_squeares_fitter<NL::LFMEllipse> fitter(model, sz);

    for (size_t i = 0; i < sz; ++i) {
        fitter.append(points[i]);
    }
    fitter.update();

    NL::Vector z(sz, 0.0);
    model.instance(*this, fitter.result(z));
}


EllipticalArc *
Ellipse::arc(Point const &ip, Point const &inner, Point const &fp)
{
    // This is resistant to degenerate ellipses:
    // both flags evaluate to false in that case.

    bool large_arc_flag = false;
    bool sweep_flag = false;

    // Determination of large arc flag:
    // large_arc is false when the inner point is on the same side
    // of the center---initial point line as the final point, AND
    // is on the same side of the center---final point line as the
    // initial point.
    // Additionally, large_arc is always false when we have exactly
    // 1/2 of an arc, i.e. the cross product of the center -> initial point
    // and center -> final point vectors is zero.
    // Negating the above leads to the condition for large_arc being true.
    Point fv = fp - _center;
    Point iv = ip - _center;
    Point innerv = inner - _center;
    double ifcp = cross(fv, iv);

    if (ifcp != 0 && (sgn(cross(fv, innerv)) != sgn(ifcp) ||
                      sgn(cross(iv, innerv)) != sgn(-ifcp)))
    {
        large_arc_flag = true;
    }

    // Determination of sweep flag:
    // For clarity, let's assume that Y grows up. Then the cross product
    // is positive for points on the left side of a vector and negative
    // on the right side of a vector.
    //
    //     cross(?, v) > 0
    //  o------------------->
    //     cross(?, v) < 0
    //
    // If the arc is small (large_arc_flag is false) and the final point
    // is on the right side of the vector initial point -> center,
    // we have to go in the direction of increasing angles
    // (counter-clockwise) and the sweep flag is true.
    // If the arc is large, the opposite is true, since we have to reach
    // the final point going the long way - in the other direction.
    // We can express this observation as:
    // cross(_center - ip, fp - _center) < 0 xor large_arc flag
    // This is equal to:
    // cross(-iv, fv) < 0 xor large_arc flag
    // But cross(-iv, fv) is equal to cross(fv, iv) due to antisymmetry
    // of the cross product, so we end up with the condition below.
    if ((ifcp < 0) ^ large_arc_flag) {
        sweep_flag = true;
    }

    EllipticalArc *ret_arc = new EllipticalArc(ip, ray(X), ray(Y), rotationAngle(),
                                               large_arc_flag, sweep_flag, fp);
    return ret_arc;
}

Ellipse &Ellipse::operator*=(Rotate const &r)
{
    _angle += r.angle();
    _center *= r;
    return *this;
}

Ellipse &Ellipse::operator*=(Affine const& m)
{
    Affine a = Scale(ray(X), ray(Y)) * Rotate(_angle);
    Affine mwot = m.withoutTranslation();
    Affine am = a * mwot;
    Point new_center = _center * m;

    if (are_near(am.descrim(), 0)) {
        double angle;
        if (am[0] != 0) {
            angle = std::atan2(am[2], am[0]);
        } else if (am[1] != 0) {
            angle = std::atan2(am[3], am[1]);
        } else {
            angle = M_PI/2;
        }
        Point v = Point::polar(angle) * am;
        _center = new_center;
        _rays[X] = L2(v);
        _rays[Y] = 0;
        _angle = atan2(v);
        return *this;
    } else if (mwot.isScale(0) && _angle.radians() == 0) {
        _rays[X] *= std::abs(mwot[0]);
        _rays[Y] *= std::abs(mwot[3]);
        _center = new_center;
        return *this;
    }

    std::vector<double> coeff = coefficients();
    Affine q( coeff[0],   coeff[1]/2,
              coeff[1]/2, coeff[2],
              0,          0   );

    Affine invm = mwot.inverse();
    q = invm * q ;
    std::swap(invm[1], invm[2]);
    q *= invm;
    setCoefficients(q[0], 2*q[1], q[3], 0, 0, -1);
    _center = new_center;

    return *this;
}

Ellipse Ellipse::canonicalForm() const
{
    Ellipse result(*this);
    result.makeCanonical();
    return result;
}

void Ellipse::makeCanonical()
{
    if (_rays[X] == _rays[Y]) {
        _angle = 0;
        return;
    }

    if (_angle < 0) {
        _angle += M_PI;
    }
    if (_angle >= M_PI/2) {
        std::swap(_rays[X], _rays[Y]);
        _angle -= M_PI/2;
    }
}

Point Ellipse::pointAt(Coord t) const
{
    Point p = Point::polar(t);
    p *= unitCircleTransform();
    return p;
}

Coord Ellipse::valueAt(Coord t, Dim2 d) const
{
    Coord sinrot, cosrot, cost, sint;
    sincos(rotationAngle(), sinrot, cosrot);
    sincos(t, sint, cost);

    if ( d == X ) {
        return    ray(X) * cosrot * cost
                - ray(Y) * sinrot * sint
                + center(X);
    } else {
        return    ray(X) * sinrot * cost
                + ray(Y) * cosrot * sint
                + center(Y);
    }
}

Coord Ellipse::timeAt(Point const &p) const
{
    // degenerate ellipse is basically a reparametrized line segment
    if (ray(X) == 0 || ray(Y) == 0) {
        if (ray(X) != 0) {
            return asin(Line(axis(X)).timeAt(p));
        } else if (ray(Y) != 0) {
            return acos(Line(axis(Y)).timeAt(p));
        } else {
            return 0;
        }
    }
    Affine iuct = inverseUnitCircleTransform();
    return Angle(atan2(p * iuct)).radians0(); // return a value in [0, 2pi)
}

Point Ellipse::unitTangentAt(Coord t) const
{
    Point p = Point::polar(t + M_PI/2);
    p *= unitCircleTransform().withoutTranslation();
    p.normalize();
    return p;
}

bool Ellipse::contains(Point const &p) const
{
    Point tp = p * inverseUnitCircleTransform();
    return tp.length() <= 1;
}

/** @brief Convert curve time on the major axis to the corresponding angle
 *         parameters on a degenerate ellipse collapsed onto that axis.
 * @param t The curve time on the major axis of an ellipse.
 * @param vertical If true, the major axis goes from angle -π/2 to +π/2;
 *                 otherwise, the major axis connects angles π and 0.
 * @return The two angles at which the collapsed ellipse passes through the
 *         major axis point corresponding to the given time \f$t \in [0, 1]\f$.
 */
static std::array<Coord, 2> axis_time_to_angles(Coord t, bool vertical)
{
    Coord const to_unit = std::clamp(2.0 * t - 1.0, -1.0, 1.0);
    if (vertical) {
        double const arcsin = std::asin(to_unit);
        return {arcsin, M_PI - arcsin};
    } else {
        double const arccos = std::acos(to_unit);
        return {arccos, -arccos};
    }
}

/** @brief For each intersection of some shape with the major axis of an ellipse, produce one or two
 *         intersections of a degenerate ellipse (collapsed onto that axis) with the same shape.
 *
 * @param axis_intersections The intersections of some shape with the major axis.
 * @param vertical Whether this is the vertical major axis (in the ellipse's natural coordinates).
 * @return A vector with doubled intersections (corresponding to the two passages of the squashed
 *         ellipse through that point) and swapped order of the intersected shapes.
*/
static std::vector<ShapeIntersection> double_axis_intersections(std::vector<ShapeIntersection> &&axis_intersections,
                                                                bool vertical)
{
    if (axis_intersections.empty()) {
        return {};
    }
    std::vector<ShapeIntersection> result;
    result.reserve(2 * axis_intersections.size());

    for (auto const &x : axis_intersections) {
        for (auto a : axis_time_to_angles(x.second, vertical)) {
            result.emplace_back(a, x.first, x.point()); // Swap first <-> converted second.
            if (x.second == 0.0 || x.second == 1.0) {
                break; // Do not double up endpoint intersections.
            }
        }
    }
    return result;
}

std::vector<ShapeIntersection> Ellipse::intersect(Line const &line) const
{
    std::vector<ShapeIntersection> result;

    if (line.isDegenerate()) {
        return result;
    }
    if (ray(X) == 0 || ray(Y) == 0) {
        return double_axis_intersections(line.intersect(majorAxis()), ray(X) == 0);
    }

    // Ax^2 + Bxy + Cy^2 + Dx + Ey + F
    std::array<Coord, 6> coeffs;
    coefficients(coeffs[0], coeffs[1], coeffs[2], coeffs[3], coeffs[4], coeffs[5]);
    rescale_homogenous(coeffs);
    auto [A, B, C, D, E, F] = coeffs;
    Affine iuct = inverseUnitCircleTransform();

    // generic case
    std::array<Coord, 3> line_coeffs;
    line.coefficients(line_coeffs[0], line_coeffs[1], line_coeffs[2]);
    rescale_homogenous(line_coeffs);
    auto [a, b, c] = line_coeffs;
    Point lv = line.vector();

    if (fabs(lv[X]) > fabs(lv[Y])) {
        // y = -a/b x - c/b
        Coord q = -a/b;
        Coord r = -c/b;

        // substitute that into the ellipse equation, making it quadratic in x
        Coord I = A + B*q + C*q*q;          // x^2 terms
        Coord J = B*r + C*2*q*r + D + E*q;  // x^1 terms
        Coord K = C*r*r + E*r + F;          // x^0 terms
        std::vector<Coord> xs = solve_quadratic(I, J, K);

        for (double x : xs) {
            Point p(x, q*x + r);
            result.emplace_back(atan2(p * iuct), line.timeAt(p), p);
        }
    } else {
        Coord q = -b/a;
        Coord r = -c/a;

        Coord I = A*q*q + B*q + C;
        Coord J = A*2*q*r + B*r + D*q + E;
        Coord K = A*r*r + D*r + F;
        std::vector<Coord> xs = solve_quadratic(I, J, K);

        for (double x : xs) {
            Point p(q*x + r, x);
            result.emplace_back(atan2(p * iuct), line.timeAt(p), p);
        }
    }
    return result;
}

std::vector<ShapeIntersection> Ellipse::intersect(LineSegment const &seg) const
{
    if (!boundsFast().intersects(seg.boundsFast())) {
        return {};
    }

    // We simply reuse the procedure for lines and filter out
    // results where the line time value is outside of the unit interval,
    // but we apply a small tolerance to account for numerical errors.
    double const param_prec = EPSILON / seg.length(0.0);
    // TODO: accept a precision setting instead of always using EPSILON
    // (requires an ABI break).

    auto xings = intersect(Line(seg));
    if (xings.empty()) {
        return xings;
    }
    decltype(xings) result;
    result.reserve(xings.size());

    for (auto const &x : xings) {
        if (x.second < -param_prec || x.second > 1.0 + param_prec) {
            continue;
        }
        result.emplace_back(x.first, std::clamp(x.second, 0.0, 1.0), x.point());
    }
    return result;
}

std::vector<ShapeIntersection> Ellipse::intersect(Ellipse const &other) const
{
    // Handle degenerate cases first.
    if (ray(X) == 0 || ray(Y) == 0) { // Degenerate ellipse, collapsed to the major axis.
        return double_axis_intersections(other.intersect(majorAxis()), ray(X) == 0);
    }
    if (*this == other) { // Two identical ellipses.
        THROW_INFINITELY_MANY_SOLUTIONS("The two ellipses are identical.");
    }
    if (!boundsFast().intersects(other.boundsFast())) {
        return {};
    }

    // Find coefficients of the implicit equations of the two ellipses and rescale
    // them (losslessly) for better numerical conditioning.
    std::array<double, 6> coeffs;
    coefficients(coeffs[0], coeffs[1], coeffs[2], coeffs[3], coeffs[4], coeffs[5]);
    rescale_homogenous(coeffs);
    auto [A, B, C, D, E, F] = coeffs;

    std::array<double, 6> otheffs;
    other.coefficients(otheffs[0], otheffs[1], otheffs[2], otheffs[3], otheffs[4], otheffs[5]);
    rescale_homogenous(otheffs);
    auto [a, b, c, d, e, f] = otheffs;

    // Assume that Q(x, y) = 0 is the ellipse equation given by uppercase letters
    // and R(x, y) = 0 is the equation given by lowercase ones.
    // In other words, Q is the quadratic function describing this ellipse and
    // R is the quadratic function for the other ellipse.
    //
    // A point (x, y) is common to both ellipses if and only if it solves the system
    // { Q(x, y) = 0,
    // { R(x, y) = 0.
    //
    // If µ is any real number, we can multiply the first equation by µ and add that
    // to the first equation, obtaining the new system of equations:
    // {            Q(x, y) = 0,
    // { µQ(x, y) + R(x, y) = 0.
    //
    // The first equation still says that (x, y) is a point on this ellipse, but the
    // second equation uses the new expression (µQ + R) instead of the original R.
    //
    // Why do we do this? The reason is that the set of functions {µQ + R : µ real}
    // is a "real system of conics" and there's a theorem which guarantees that such a system
    // always contains a "degenerate conic" [proof below].
    // In practice, the degenerate conic will describe a line or a pair of lines, and intersecting
    // a line with an ellipse is much easier than intersecting two ellipses directly.
    //
    // But in order to be able to do this, we must find a value of µ for which µQ + R is degenerate.
    // We can write the expression (µQ + R)(x, y) in the following way:
    //
    //                          |  aa  bb/2 dd/2 | |x|
    // (µQ + R)(x, y) = [x y 1] | bb/2  cc  ee/2 | |y|
    //                          | dd/2 ee/2  ff  | |1|
    //
    // where aa = µA + a and so on. The determinant can be explicitly written out,
    // giving an equation which is cubic in µ and can be solved analytically.
    // The conic µQ + R is degenerate if and only if this determinant is 0.
    //
    // Proof that there's always a degenerate conic: a cubic real polynomial always has a root,
    // and if the polynomial in µ isn't cubic (coefficient of µ^3 is zero), then the starting
    // conic is already degenerate.

    Coord I, J, K, L; // Coefficients of µ in the expression for the determinant.
    I = (-B*B*F + 4*A*C*F + D*E*B - A*E*E - C*D*D) / 4;
    J = -((B*B - 4*A*C) * f + (2*B*F - D*E) * b + (2*A*E - D*B) * e +
          (2*C*D - E*B) * d + (D*D - 4*A*F) * c + (E*E - 4*C*F) * a) / 4;
    K = -((b*b - 4*a*c) * F + (2*b*f - d*e) * B + (2*a*e - d*b) * E +
          (2*c*d - e*b) * D + (d*d - 4*a*f) * C + (e*e - 4*c*f) * A) / 4;
    L = (-b*b*f + 4*a*c*f + d*e*b - a*e*e - c*d*d) / 4;

    std::vector<Coord> mus = solve_cubic(I, J, K, L);
    Coord mu = infinity();

    // Now that we have solved for µ, we need to check whether the conic
    // determined by µQ + R is reducible to a product of two lines. If it's not,
    // it means that there are no intersections. If it is, the intersections of these
    // lines with the original ellipses (if there are any) give the coordinates
    // of intersections.

    // Prefer middle root if there are three.
    // Out of three possible pairs of lines that go through four points of intersection
    // of two ellipses, this corresponds to cross-lines. These intersect the ellipses
    // at less shallow angles than the other two options.
    if (mus.size() == 3) {
        std::swap(mus[1], mus[0]);
    }
    /// Discriminant within this radius of 0 will be considered zero.
    static Coord const discriminant_precision = 1e-10;

    for (Coord candidate_mu : mus) {
        Coord const aa = std::fma(candidate_mu, A, a);
        Coord const bb = std::fma(candidate_mu, B, b);
        Coord const cc = std::fma(candidate_mu, C, c);
        Coord const delta = sqr(bb) - 4*aa*cc;
        if (delta < -discriminant_precision) {
            continue;
        }
        mu = candidate_mu;
        break;
    }

    // if no suitable mu was found, there are no intersections
    if (mu == infinity()) {
        return {};
    }

    // Create the degenerate conic and decompose it into lines.
    std::array<double, 6> degen = {std::fma(mu, A, a), std::fma(mu, B, b), std::fma(mu, C, c),
                                   std::fma(mu, D, d), std::fma(mu, E, e), std::fma(mu, F, f)};
    rescale_homogenous(degen);
    auto const lines = xAx(degen[0], degen[1], degen[2],
                           degen[3], degen[4], degen[5]).decompose_df(discriminant_precision);

    // intersect with the obtained lines and report intersections
    std::vector<ShapeIntersection> result;
    for (auto const &line : lines) {
        if (line.isDegenerate()) {
            continue;
        }
        auto as = intersect(line);
        // NOTE: If we only cared about the intersection points, we could simply
        // intersect this ellipse with the lines and ignore the other ellipse.
        // But we need the time coordinates on the other ellipse as well.
        auto bs = other.intersect(line);
        if (as.empty() || bs.empty()) {
            continue;
        }
        // Due to numerical errors, a tangency may sometimes be found as 1 intersection
        // on one ellipse and 2 intersections on the other. If this happens, we average
        // the points of the two intersections.
        auto const intersection_average = [](ShapeIntersection const &i,
                                             ShapeIntersection const &j) -> ShapeIntersection
        {
            return ShapeIntersection(i.first, j.first, middle_point(i.point(), j.point()));
        };
        auto const synthesize_intersection = [&](ShapeIntersection const &i,
                                                 ShapeIntersection const &j) -> void
        {
            result.emplace_back(i.first, j.first, middle_point(i.point(), j.point()));
        };
        if (as.size() == 2) {
            if (bs.size() == 2) {
                synthesize_intersection(as[0], bs[0]);
                synthesize_intersection(as[1], bs[1]);
            } else if (bs.size() == 1) {
                synthesize_intersection(intersection_average(as[0], as[1]), bs[0]);
            }
        } else if (as.size() == 1) {
            if (bs.size() == 2) {
                synthesize_intersection(as[0], intersection_average(bs[0], bs[1]));
            } else if (bs.size() == 1) {
                synthesize_intersection(as[0], bs[0]);
            }
        }
    }
    return result;
}

std::vector<ShapeIntersection> Ellipse::intersect(D2<Bezier> const &b) const
{
    Coord A, B, C, D, E, F;
    coefficients(A, B, C, D, E, F);

    // We plug the X and Y curves into the implicit equation and solve for t.
    Bezier x = A*b[X]*b[X] + B*b[X]*b[Y] + C*b[Y]*b[Y] + D*b[X] + E*b[Y] + F;
    std::vector<Coord> r = x.roots();

    std::vector<ShapeIntersection> result;
    for (double & i : r) {
        Point p = b.valueAt(i);
        result.emplace_back(timeAt(p), i, p);
    }
    return result;
}

bool Ellipse::operator==(Ellipse const &other) const
{
    if (_center != other._center) return false;

    Ellipse a = this->canonicalForm();
    Ellipse b = other.canonicalForm();

    if (a._rays != b._rays) return false;
    if (a._angle != b._angle) return false;

    return true;
}


bool are_near(Ellipse const &a, Ellipse const &b, Coord precision)
{
    // We want to know whether no point on ellipse a is further than precision
    // from the corresponding point on ellipse b. To check this, we compute
    // the four extreme points at the end of each ray for each ellipse
    // and check whether they are sufficiently close.

    // First, we need to correct the angles on the ellipses, so that they are
    // no further than M_PI/4 apart. This can always be done by rotating
    // and exchanging axes.
    Ellipse ac = a, bc = b;
    if (distance(ac.rotationAngle(), bc.rotationAngle()).radians0() >= M_PI/2) {
        ac.setRotationAngle(ac.rotationAngle() + M_PI);
    }
    if (distance(ac.rotationAngle(), bc.rotationAngle()) >= M_PI/4) {
        Angle d1 = distance(ac.rotationAngle() + M_PI/2, bc.rotationAngle());
        Angle d2 = distance(ac.rotationAngle() - M_PI/2, bc.rotationAngle());
        Coord adj = d1.radians0() < d2.radians0() ? M_PI/2 : -M_PI/2;
        ac.setRotationAngle(ac.rotationAngle() + adj);
        ac.setRays(ac.ray(Y), ac.ray(X));
    }

    // Do the actual comparison by computing four points on each ellipse.
    Point tps[] = {Point(1,0), Point(0,1), Point(-1,0), Point(0,-1)};
    for (auto & tp : tps) {
        if (!are_near(tp * ac.unitCircleTransform(),
                      tp * bc.unitCircleTransform(),
                      precision))
            return false;
    }
    return true;
}

std::ostream &operator<<(std::ostream &out, Ellipse const &e)
{
    out << "Ellipse(" << e.center() << ", " << e.rays()
        << ", " << format_coord_nice(e.rotationAngle()) << ")";
    return out;
}

}  // end namespace Geom


/*
  Local Variables:
  mode:c++
  c-file-style:"stroustrup"
  c-file-offsets:((innamespace . 0)(inline-open . 0)(case-label . +))
  indent-tabs-mode:nil
  fill-column:99
  End:
*/
// vim: filetype=cpp:expandtab:shiftwidth=4:tabstop=8:softtabstop=4:fileencoding=utf-8:textwidth=99 :