ifc-lite-geometry 3.2.0

Geometry processing and mesh generation for IFC models
Documentation
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
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this
// file, You can obtain one at https://mozilla.org/MPL/2.0/.

//! Public predicate dispatch over `ImplicitPoint` configurations.
//!
//! Implements every explicit/implicit `orient3d`/`orient2d` configuration the
//! arrangement pipeline produces, each first through the fast interval and
//! fixed-width tiers and escalating to the exact (BigRational) tier on a
//! straddling filter — every fast tier verified `≡` the exact tier here.

use super::{fixed, interval, rational};
use super::{DropAxis, ImplicitPoint, Sign};

/// Exact `orient3d` over a mix of explicit + implicit points.
///
/// Cascade: explicit args go through the Shewchuk adaptive predicate (its own
/// semi-static→exact ladder). Indirect args try the interval tier first and
/// escalate to the exact (BigRational) tier only on a straddle. Every tier
/// returns the SAME sign — verified against the oracle in tests.
pub fn orient3d(a: &ImplicitPoint, b: &ImplicitPoint, c: &ImplicitPoint, d: &ImplicitPoint) -> Sign {
    use ImplicitPoint::{Explicit, Lpi, Tpi};
    match (a, b, c, d) {
        (Explicit(a), Explicit(b), Explicit(c), Explicit(d)) => {
            Sign::from_f64(geometry_predicates::orient3d(*a, *b, *c, *d))
        }
        (Lpi(l), Explicit(b), Explicit(c), Explicit(d)) => interval::lpi_orient3d(l, *b, *c, *d)
            .or_else(|| fixed::indirect_orient3d(a, *b, *c, *d))
            .unwrap_or_else(|| rational::lpi_orient3d(l, *b, *c, *d)),
        (Tpi(t), Explicit(b), Explicit(c), Explicit(d)) => interval::tpi_orient3d(t, *b, *c, *d)
            .or_else(|| fixed::indirect_orient3d(a, *b, *c, *d))
            .unwrap_or_else(|| rational::tpi_orient3d(t, *b, *c, *d)),
        // By-construction unreachable: kernel callers only ever build the configurations above.
        _ => unimplemented!(
            "kernel::orient3d: implicit-point configuration never produced by the arrangement pipeline"
        ),
    }
}

/// Exact `orient2d(a, b, c)` projected on the two axes remaining after dropping
/// `axis` (the in-plane predicate for re-triangulation). Same cascade as
/// `orient3d`; the indirect 1-implicit case shares the `sign(d)` flip.
pub fn orient2d(a: &ImplicitPoint, b: &ImplicitPoint, c: &ImplicitPoint, axis: DropAxis) -> Sign {
    use ImplicitPoint::{Explicit, Lpi, Tpi};
    let (i, j) = match axis {
        DropAxis::X => (1, 2),
        DropAxis::Y => (0, 2),
        DropAxis::Z => (0, 1),
    };
    match (a, b, c) {
        (Explicit(a), Explicit(b), Explicit(c)) => {
            Sign::from_f64(geometry_predicates::orient2d([a[i], a[j]], [b[i], b[j]], [c[i], c[j]]))
        }
        (Lpi(l), Explicit(b), Explicit(c)) => interval::lpi_orient2d(l, *b, *c, axis)
            .or_else(|| fixed::indirect_orient2d(a, *b, *c, axis))
            .unwrap_or_else(|| rational::lpi_orient2d(l, *b, *c, axis)),
        (Tpi(t), Explicit(b), Explicit(c)) => interval::tpi_orient2d(t, *b, *c, axis)
            .or_else(|| fixed::indirect_orient2d(a, *b, *c, axis))
            .unwrap_or_else(|| rational::tpi_orient2d(t, *b, *c, axis)),
        // By-construction unreachable: kernel callers only ever build the configurations above.
        _ => unimplemented!(
            "kernel::orient2d: implicit-point configuration never produced by the arrangement pipeline"
        ),
    }
}

/// orient2d with two implicit points (a,b) + one explicit (c) — cascade.
pub fn orient2d_2i(a: &ImplicitPoint, b: &ImplicitPoint, c: [f64; 3], axis: DropAxis) -> Sign {
    // cascade: interval filter → fixed-width exact (fast) → BigRational (off-grid / overflow)
    interval::orient2d_2i(a, b, c, axis)
        .or_else(|| fixed::orient2d_2i(a, b, c, axis))
        .unwrap_or_else(|| rational::orient2d_2i(a, b, c, axis))
}

/// orient2d with three implicit points (a,b,c) — cascade.
pub fn orient2d_3i(a: &ImplicitPoint, b: &ImplicitPoint, c: &ImplicitPoint, axis: DropAxis) -> Sign {
    interval::orient2d_3i(a, b, c, axis)
        .or_else(|| fixed::orient2d_3i(a, b, c, axis))
        .unwrap_or_else(|| rational::orient2d_3i(a, b, c, axis))
}

/// Exact lexicographic total order on points — the interner's comparison (cascade).
pub fn cmp_lex(a: &ImplicitPoint, b: &ImplicitPoint) -> Sign {
    interval::cmp_lex(a, b).or_else(|| fixed::cmp_lex(a, b)).unwrap_or_else(|| rational::cmp_lex(a, b))
}

#[inline]
fn explicit_coord(p: &ImplicitPoint) -> [f64; 3] {
    match p {
        ImplicitPoint::Explicit(c) => *c,
        _ => unreachable!("explicit_coord on an implicit point"),
    }
}

/// `orient2d` over ANY mix of explicit/implicit points in ANY argument position
/// (the predicate the re-triangulation's point location needs). `orient2d` is
/// antisymmetric, so we canonicalise the args to implicit-first (stable, to keep
/// it a pure function), dispatch to the 0I/1I/2I/3I config, and flip the result
/// once per transposition (the permutation parity).

pub fn orient2d_any(a: &ImplicitPoint, b: &ImplicitPoint, c: &ImplicitPoint, axis: DropAxis) -> Sign {
    let pts = [a, b, c];
    let key = |p: &ImplicitPoint| u8::from(matches!(p, ImplicitPoint::Explicit(_))); // implicit=0
    let keys = [key(a), key(b), key(c)];
    let mut perm = [0usize, 1, 2];
    perm.sort_by_key(|&i| keys[i]); // stable → implicit first, original order kept
    let inversions = u8::from(perm[0] > perm[1]) + u8::from(perm[0] > perm[2]) + u8::from(perm[1] > perm[2]);
    let rp = [pts[perm[0]], pts[perm[1]], pts[perm[2]]];
    let n_implicit = 3 - (keys[0] + keys[1] + keys[2]) as usize;
    let canonical = match n_implicit {
        // (E,E,E) and (I,E,E) are handled by the position-specific dispatch.
        0 | 1 => orient2d(rp[0], rp[1], rp[2], axis),
        2 => orient2d_2i(rp[0], rp[1], explicit_coord(rp[2]), axis),
        _ => orient2d_3i(rp[0], rp[1], rp[2], axis),
    };
    if inversions % 2 == 1 {
        canonical.flip()
    } else {
        canonical
    }
}

#[cfg(test)]
mod tests {
    use super::super::{rational, DropAxis, Lpi, Tpi};
    use super::{
        cmp_lex, orient2d, orient2d_2i, orient2d_3i, orient2d_any, orient3d, ImplicitPoint, Sign,
    };

    fn e(p: [f64; 3]) -> ImplicitPoint {
        ImplicitPoint::Explicit(p)
    }

    /// Adversarial explicit-orient3d configurations (coplanar, building-scale
    /// off-plane, near-coincident large coords, sub-mm + mirrored tetra).
    fn battery() -> Vec<[[f64; 3]; 4]> {
        vec![
            [[0., 0., 0.], [1., 0., 0.], [0., 1., 0.], [1., 1., 0.]], // coplanar -> 0
            [[0., 0., 12.3456789], [10., 0., 12.3456789], [0., 7., 12.3456789], [3.3, 2.1, 12.3456789 + 1e-9]],
            [[0., 0., 12.3456789], [10., 0., 12.3456789], [0., 7., 12.3456789], [3.3, 2.1, 12.3456789 - 1e-9]],
            [[1e7, 1e7, 0.], [1e7 + 1., 1e7, 0.], [1e7, 1e7 + 1., 0.], [1e7 + 0.5, 1e7 + 0.5, 1e-7]],
            [[0., 0., 0.], [1., 2., 3.], [-2., 1., 0.5], [0.5, 0.5, 0.5]],
            [[0., 0., 0.], [1., 1., 1.], [2., 2., 2.], [5., 1., 9.]], // collinear base -> 0
            [[0., 0., 0.], [1e-4, 0., 0.], [0., 1e-4, 0.], [0., 0., 1e-4]],
            [[0., 0., 0.], [0., 1e-4, 0.], [1e-4, 0., 0.], [0., 0., 1e-4]],
            [[-3., 2., 5.], [7., -1., 2.], [4., 4., -6.], [1.5, 0.0, 0.25]],
        ]
    }

    /// LPI cases: (line PQ ∩ plane RST), plus a query triangle (p2,p3,p4).
    fn lpi_cases() -> Vec<(Lpi, [f64; 3], [f64; 3], [f64; 3])> {
        vec![
            // vertical line ∩ z=0 plane -> (0.3,0.3,0); query triangle at z=1 (LPI below)
            (
                Lpi { p: [0.3, 0.3, -1.], q: [0.3, 0.3, 1.], r: [0., 0., 0.], s: [2., 0., 0.], t: [0., 2., 0.] },
                [0., 0., 1.], [1., 0., 1.], [0., 1., 1.],
            ),
            // same LPI, query triangle at z=-1 (LPI above)
            (
                Lpi { p: [0.3, 0.3, -1.], q: [0.3, 0.3, 1.], r: [0., 0., 0.], s: [2., 0., 0.], t: [0., 2., 0.] },
                [0., 0., -1.], [1., 0., -1.], [0., 1., -1.],
            ),
            // tilted line ∩ tilted plane
            (
                Lpi { p: [1., 1., 0.], q: [2., 3., 4.], r: [0., 0., 1.], s: [3., 0., 2.], t: [0., 3., 2.] },
                [5., -2., 0.], [-1., 4., 3.], [2., 2., -3.],
            ),
            // building-scale
            (
                Lpi { p: [12.3, 4.5, -2.], q: [12.3, 4.5, 6.], r: [0., 0., 3.1], s: [20., 0., 3.1], t: [0., 9., 3.1] },
                [10., 10., 10.], [-5., 0., 0.], [0., -5., 8.],
            ),
        ]
    }

    #[test]
    fn explicit_orient3d_matches_rational_oracle() {
        for cfg in battery() {
            let [a, b, c, d] = cfg;
            let fast = orient3d(&e(a), &e(b), &e(c), &e(d));
            let oracle = rational::orient3d_exact(a, b, c, d);
            assert_eq!(fast, oracle, "explicit orient3d != rational oracle on {cfg:?}");
        }
    }

    #[test]
    fn lpi_orient3d_matches_materialised_point() {
        // The homogenised LPI-orient3d must equal the direct orient3d on the
        // exact materialised λ/d point — proving the Λ′ + sign(d)-flip.
        for (l, p2, p3, p4) in lpi_cases() {
            let homog = rational::lpi_orient3d(&l, p2, p3, p4);
            let direct = rational::orient3d_exact_pt(&rational::lpi_point(&l), p2, p3, p4);
            assert_eq!(homog, direct, "LPI homogenisation/flip wrong for {l:?}");
            // sanity: these are non-degenerate, so the sign is definite
            assert_ne!(homog, Sign::Zero, "test LPI case should be off-plane: {l:?}");
        }
    }

    #[test]
    fn lpi_orient3d_sign_invariant_to_plane_winding() {
        // Re-wind the plane (swap S,T): flips sign(d) but the point + geometry
        // are identical, so the per-config flip must yield the SAME sign. This
        // is the test that catches a missing/extra `sign(d)` flip.
        for (l, p2, p3, p4) in lpi_cases() {
            let l_rewound = Lpi { s: l.t, t: l.s, ..l };
            assert_eq!(
                rational::lpi_orient3d(&l, p2, p3, p4),
                rational::lpi_orient3d(&l_rewound, p2, p3, p4),
                "LPI-orient3d sign changed under plane re-winding — the sign(d) flip is wrong/missing"
            );
        }
    }

    #[test]
    fn assemble_sign_per_config_flip() {
        use super::super::assemble_sign;
        // odd #negatives -> flip; even -> no flip; any zero -> Zero.
        assert_eq!(assemble_sign(Sign::Positive, &[Sign::Negative]), Sign::Negative);
        assert_eq!(assemble_sign(Sign::Positive, &[Sign::Negative, Sign::Negative]), Sign::Positive);
        assert_eq!(assemble_sign(Sign::Negative, &[Sign::Positive]), Sign::Negative);
        assert_eq!(assemble_sign(Sign::Positive, &[Sign::Zero]), Sign::Zero);
        assert_eq!(assemble_sign(Sign::Positive, &[]), Sign::Positive);
    }

    #[test]
    fn next_up_down_are_adjacent() {
        use super::super::interval::{next_down, next_up};
        for &x in &[1.0, -1.0, 0.0, 1e7, -1e-9, 12.3456789, f64::MIN_POSITIVE] {
            assert!(next_up(x) > x, "next_up({x}) not strictly greater");
            assert!(next_down(x) < x, "next_down({x}) not strictly less");
            // Round-trip = adjacency: nothing representable strictly between.
            assert_eq!(next_down(next_up(x)), x, "next_up/next_down not adjacent at {x}");
        }
    }

    /// Deterministic LCG for the soundness fuzz (no Math::random; fixed seed).
    struct Lcg(u64);
    impl Lcg {
        fn u(&mut self) -> u64 {
            self.0 = self.0.wrapping_mul(6364136223846793005).wrapping_add(1442695040888963407);
            self.0
        }
        fn f(&mut self, lo: f64, hi: f64) -> f64 {
            let unit = (self.u() >> 11) as f64 / (1u64 << 53) as f64; // [0,1)
            lo + (hi - lo) * unit
        }
        fn p(&mut self) -> [f64; 3] {
            [self.f(-10., 10.), self.f(-10., 10.), self.f(-10., 10.)]
        }
    }

    #[test]
    fn interval_tier_is_sound_and_the_cascade_equals_exact() {
        use super::super::{interval, rational, Lpi};
        let mut rng = Lcg(0x1234_5678_9abc_def0);
        let (mut definite, mut escalated) = (0u32, 0u32);
        for _ in 0..3000 {
            let l = Lpi { p: rng.p(), q: rng.p(), r: rng.p(), s: rng.p(), t: rng.p() };
            let (p2, p3, p4) = (rng.p(), rng.p(), rng.p());
            let exact = rational::lpi_orient3d(&l, p2, p3, p4);
            // Soundness: a definite interval sign must equal the exact sign.
            match interval::lpi_orient3d(&l, p2, p3, p4) {
                Some(s) => {
                    assert_eq!(s, exact, "interval returned a WRONG definite sign for {l:?}");
                    definite += 1;
                }
                None => escalated += 1,
            }
            // The public cascade (interval → escalate) must always equal exact.
            let cascade = orient3d(&ImplicitPoint::Lpi(l), &e(p2), &e(p3), &e(p4));
            assert_eq!(cascade, exact, "cascade != exact for {l:?}");
        }
        // The interval fast path must carry the overwhelming majority (perf gate).
        assert!(
            definite as f64 / (definite + escalated) as f64 > 0.95,
            "interval resolved only {definite}/{} — fast path too cold",
            definite + escalated
        );
        eprintln!("interval tier: {definite} definite, {escalated} escalated to exact");
    }

    /// TPI cases: three planes (each a triangle) + a query triangle.
    fn tpi_cases() -> Vec<(Tpi, [f64; 3], [f64; 3], [f64; 3])> {
        // planes x=0.3, y=0.4, z=0 -> point (0.3,0.4,0)
        let axis_aligned = Tpi {
            planes: [
                [[0., 0., 0.], [1., 0., 0.], [0., 1., 0.]],     // z=0
                [[0.3, 0., 0.], [0.3, 1., 0.], [0.3, 0., 1.]],  // x=0.3
                [[0., 0.4, 0.], [1., 0.4, 0.], [0., 0.4, 1.]],  // y=0.4
            ],
        };
        // three tilted planes meeting at a general point
        let tilted = Tpi {
            planes: [
                [[0., 0., 1.], [3., 0., 2.], [0., 3., 2.]],
                [[1., 0., 0.], [1., 2., 1.], [2., 0., 3.]],
                [[-1., -1., 0.], [2., -1., 1.], [-1., 2., 2.]],
            ],
        };
        vec![
            (axis_aligned, [0., 0., 1.], [1., 0., 1.], [0., 1., 1.]),   // query above
            (axis_aligned, [0., 0., -1.], [1., 0., -1.], [0., 1., -1.]), // query below
            (tilted, [5., -2., 0.], [-1., 4., 3.], [2., 2., -3.]),
            (tilted, [10., 10., 10.], [-5., 0., 0.], [0., -5., 8.]),
        ]
    }

    #[test]
    fn tpi_orient3d_matches_materialised_point() {
        // The homogenised TPI-orient3d must equal the direct orient3d on the
        // exact materialised λ/d point — proving the TPI Cramer λ + the flip.
        for (t, p2, p3, p4) in tpi_cases() {
            let homog = rational::tpi_orient3d(&t, p2, p3, p4);
            let direct = rational::orient3d_exact_pt(&rational::tpi_point(&t), p2, p3, p4);
            assert_eq!(homog, direct, "TPI homogenisation/flip wrong for {t:?}");
            assert_ne!(homog, Sign::Zero, "test TPI case should be off-plane: {t:?}");
        }
    }

    #[test]
    fn tpi_orient3d_sign_invariant_to_plane_winding() {
        // Re-wind plane 0 (swap its 2nd/3rd points): flips that plane's normal
        // and hence sign(d), but the meeting point is identical → the sign(d)
        // flip must yield the SAME geometric sign.
        for (t, p2, p3, p4) in tpi_cases() {
            let mut rewound = t;
            rewound.planes[0].swap(1, 2);
            assert_eq!(
                rational::tpi_orient3d(&t, p2, p3, p4),
                rational::tpi_orient3d(&rewound, p2, p3, p4),
                "TPI-orient3d sign changed under plane re-winding — the sign(d) flip is wrong/missing"
            );
        }
    }

    #[test]
    fn tpi_interval_is_sound_and_the_cascade_equals_exact() {
        use super::super::interval;
        let mut rng = Lcg(0xfeed_face_cafe_d00d);
        let (mut definite, mut escalated) = (0u32, 0u32);
        for _ in 0..2000 {
            // a random TPI = three random planes (generically meet at a point)
            let plane = |rng: &mut Lcg| [rng.p(), rng.p(), rng.p()];
            let t = Tpi { planes: [plane(&mut rng), plane(&mut rng), plane(&mut rng)] };
            let (p2, p3, p4) = (rng.p(), rng.p(), rng.p());
            let exact = rational::tpi_orient3d(&t, p2, p3, p4);
            match interval::tpi_orient3d(&t, p2, p3, p4) {
                Some(s) => {
                    assert_eq!(s, exact, "TPI interval returned a WRONG definite sign for {t:?}");
                    definite += 1;
                }
                None => escalated += 1,
            }
            let cascade = orient3d(&ImplicitPoint::Tpi(t), &e(p2), &e(p3), &e(p4));
            assert_eq!(cascade, exact, "TPI cascade != exact for {t:?}");
        }
        // TPI Λ′ is degree-3 in the planes (heavier than LPI) so the interval is
        // wider — still expect a healthy majority to resolve in f64.
        assert!(
            definite as f64 / (definite + escalated) as f64 > 0.80,
            "TPI interval resolved only {definite}/{}",
            definite + escalated
        );
        eprintln!("TPI interval tier: {definite} definite, {escalated} escalated");
    }

    const AXES: [DropAxis; 3] = [DropAxis::X, DropAxis::Y, DropAxis::Z];

    #[test]
    fn explicit_orient2d_matches_oracle() {
        for axis in AXES {
            for cfg in battery() {
                let [a, b, c, _d] = cfg;
                let fast = orient2d(&e(a), &e(b), &e(c), axis);
                let oracle = rational::orient2d_exact(a, b, c, axis);
                assert_eq!(fast, oracle, "explicit orient2d != oracle on {cfg:?} axis {axis:?}");
            }
        }
    }

    #[test]
    fn indirect_orient2d_matches_materialised_point() {
        // Homogenised LPI/TPI orient2d == the direct orient2d on the exact
        // materialised λ/d point, for every projection axis; cascade == exact.
        for axis in AXES {
            for (l, p2, p3, _p4) in lpi_cases() {
                let homog = rational::lpi_orient2d(&l, p2, p3, axis);
                let direct = rational::orient2d_exact_pt(&rational::lpi_point(&l), p2, p3, axis);
                assert_eq!(homog, direct, "LPI orient2d homog/flip wrong, axis {axis:?}");
                let cascade = orient2d(&ImplicitPoint::Lpi(l), &e(p2), &e(p3), axis);
                assert_eq!(cascade, direct, "LPI orient2d cascade != exact, axis {axis:?}");
            }
            for (t, p2, p3, _p4) in tpi_cases() {
                let homog = rational::tpi_orient2d(&t, p2, p3, axis);
                let direct = rational::orient2d_exact_pt(&rational::tpi_point(&t), p2, p3, axis);
                assert_eq!(homog, direct, "TPI orient2d homog/flip wrong, axis {axis:?}");
                let cascade = orient2d(&ImplicitPoint::Tpi(t), &e(p2), &e(p3), axis);
                assert_eq!(cascade, direct, "TPI orient2d cascade != exact, axis {axis:?}");
            }
        }
    }

    #[test]
    fn orient2d_interval_is_sound_vs_oracle() {
        use super::super::interval;
        let mut rng = Lcg(0xabcd_1234_5678_9999);
        for _ in 0..2000 {
            let l = Lpi { p: rng.p(), q: rng.p(), r: rng.p(), s: rng.p(), t: rng.p() };
            let (b, c) = (rng.p(), rng.p());
            let axis = AXES[(rng.u() % 3) as usize];
            let exact = rational::lpi_orient2d(&l, b, c, axis);
            if let Some(s) = interval::lpi_orient2d(&l, b, c, axis) {
                assert_eq!(s, exact, "orient2d interval returned a WRONG definite sign for {l:?}");
            }
            // cascade always equals exact
            assert_eq!(orient2d(&ImplicitPoint::Lpi(l), &e(b), &e(c), axis), exact);
        }
    }

    #[test]
    fn lpi_point_lies_on_its_defining_plane() {
        // GEOMETRIC correctness (not just self-consistency): orient3d(LPI, R,S,T)
        // must be 0 — the LPI point is on plane RST by definition. This guard is
        // what the consistency tests miss; it caught the λ = d·P ± n·qp sign bug.
        for (l, _, _, _) in lpi_cases() {
            assert_eq!(
                orient3d(&ImplicitPoint::Lpi(l), &e(l.r), &e(l.s), &e(l.t)),
                Sign::Zero,
                "LPI point is not on its defining plane R,S,T: {l:?}"
            );
        }
    }

    #[test]
    fn tpi_point_lies_on_all_three_defining_planes() {
        for (t, _, _, _) in tpi_cases() {
            for plane in &t.planes {
                assert_eq!(
                    orient3d(&ImplicitPoint::Tpi(t), &e(plane[0]), &e(plane[1]), &e(plane[2])),
                    Sign::Zero,
                    "TPI point is not on one of its defining planes: {t:?}"
                );
            }
        }
    }

    #[test]
    fn orient2d_1i_sign_invariant_to_plane_winding() {
        // Guards Risk #1 (the doc-vs-code sign-table conflict): rewinding the
        // implicit point's defining plane flips sign(d) while leaving the point +
        // 2D query geometrically identical, so the TRUE orient2d sign is
        // unchanged. The shipped d¹ flip preserves it; the doc's old d²/no-flip
        // would invert it. The orient3d analogue is tested at line ~137 — this
        // closes the matching gap for orient2d (the gap the LPI λ-sign bug used).
        for axis in AXES {
            for (l, p2, p3, _p4) in lpi_cases() {
                let rewound = Lpi { s: l.t, t: l.s, ..l };
                assert_eq!(
                    rational::lpi_orient2d(&l, p2, p3, axis),
                    rational::lpi_orient2d(&rewound, p2, p3, axis),
                    "orient2d 1I sign changed under plane re-winding (axis {axis:?})"
                );
            }
        }
    }

    #[test]
    fn multi_implicit_orient2d_matches_materialised_oracle() {
        // orient2d_2i / orient2d_3i over all {Lpi,Tpi} mixtures must equal
        // the direct orient2d on the materialised λ/d points, for every drop axis.
        let mut pts: Vec<ImplicitPoint> =
            lpi_cases().into_iter().map(|(l, ..)| ImplicitPoint::Lpi(l)).collect();
        pts.extend(tpi_cases().into_iter().map(|(t, ..)| ImplicitPoint::Tpi(t)));
        let c = [1.3, -0.7, 2.1];
        let cpt = rational::point_of(&ImplicitPoint::Explicit(c));
        for axis in AXES {
            for a in &pts {
                for b in &pts {
                    let oracle = rational::orient2d_pts(
                        &rational::point_of(a), &rational::point_of(b), &cpt, axis);
                    assert_eq!(rational::orient2d_2i(a, b, c, axis), oracle, "orient2d_2i (axis {axis:?})");
                    for cc in &pts {
                        let oracle3 = rational::orient2d_pts(
                            &rational::point_of(a), &rational::point_of(b), &rational::point_of(cc), axis);
                        assert_eq!(rational::orient2d_3i(a, b, cc, axis), oracle3, "orient2d_3i (axis {axis:?})");
                    }
                }
            }
        }
    }

    #[test]
    fn multi_implicit_orient2d_winding_invariant() {
        // Rewinding a's defining plane flips sign(d); 2I/3I must keep the sign.
        let l = lpi_cases()[2].0;
        let a = ImplicitPoint::Lpi(l);
        let a_rw = ImplicitPoint::Lpi(Lpi { s: l.t, t: l.s, ..l });
        let b = ImplicitPoint::Tpi(tpi_cases()[0].0);
        let cc = ImplicitPoint::Lpi(lpi_cases()[0].0);
        let c = [0.4, 1.1, -0.3];
        for axis in AXES {
            assert_eq!(
                rational::orient2d_2i(&a, &b, c, axis),
                rational::orient2d_2i(&a_rw, &b, c, axis),
                "2I sign changed under plane re-winding"
            );
            assert_eq!(
                rational::orient2d_3i(&a, &b, &cc, axis),
                rational::orient2d_3i(&a_rw, &b, &cc, axis),
                "3I sign changed under plane re-winding"
            );
        }
    }

    #[test]
    fn orient2d_any_matches_oracle_for_every_permutation_and_mix() {
        // The general dispatch must equal the direct orient2d on the materialised
        // points for EVERY ordered triple (covers all positions + permutation
        // parities) and every drop axis, across explicit/LPI/TPI mixes.
        let mut pts: Vec<ImplicitPoint> = vec![e([0.0, 0.0, 0.0]), e([3.0, 1.0, -2.0])];
        pts.extend(lpi_cases().into_iter().take(2).map(|(l, ..)| ImplicitPoint::Lpi(l)));
        pts.extend(tpi_cases().into_iter().take(2).map(|(t, ..)| ImplicitPoint::Tpi(t)));
        for axis in AXES {
            for a in &pts {
                for b in &pts {
                    for c in &pts {
                        let got = orient2d_any(a, b, c, axis);
                        let want = rational::orient2d_pts(
                            &rational::point_of(a),
                            &rational::point_of(b),
                            &rational::point_of(c),
                            axis,
                        );
                        assert_eq!(got, want, "orient2d_any mismatch (axis {axis:?})");
                    }
                }
            }
        }
    }

    #[test]
    fn cmp_lex_matches_materialised_order_and_is_a_total_order() {
        use std::cmp::Ordering;
        let mut pts: Vec<ImplicitPoint> =
            lpi_cases().into_iter().map(|(l, ..)| ImplicitPoint::Lpi(l)).collect();
        pts.extend(tpi_cases().into_iter().map(|(t, ..)| ImplicitPoint::Tpi(t)));
        pts.push(e([1.5, -2.0, 0.25]));
        pts.push(e([0.0, 0.0, 0.0]));
        let oracle = |a: &ImplicitPoint, b: &ImplicitPoint| -> Sign {
            let (pa, pb) = (rational::point_of(a), rational::point_of(b));
            for k in 0..3 {
                match pa[k].cmp(&pb[k]) {
                    Ordering::Less => return Sign::Negative,
                    Ordering::Greater => return Sign::Positive,
                    Ordering::Equal => {}
                }
            }
            Sign::Zero
        };
        for a in &pts {
            assert_eq!(rational::cmp_lex(a, a), Sign::Zero, "cmp_lex not reflexive-zero");
            for b in &pts {
                assert_eq!(rational::cmp_lex(a, b), oracle(a, b), "cmp_lex != materialised lex");
                assert_eq!(
                    rational::cmp_lex(a, b),
                    rational::cmp_lex(b, a).flip(),
                    "cmp_lex not antisymmetric"
                );
            }
        }
        // transitivity: a<b<c ⇒ a<c (no ordering cycles).
        for a in &pts {
            for b in &pts {
                for c in &pts {
                    if rational::cmp_lex(a, b) == Sign::Negative
                        && rational::cmp_lex(b, c) == Sign::Negative
                    {
                        assert_eq!(rational::cmp_lex(a, c), Sign::Negative, "cmp_lex not transitive");
                    }
                }
            }
        }
    }

    #[test]
    fn new_tier_interval_is_sound_and_cascade_equals_exact() {
        // The interval fast tiers for 2I/3I orient2d + cmp_lex never
        // return a wrong definite sign, and the public cascade always == exact.
        use super::super::interval;
        let mut rng = Lcg(0x0bad_c0de_1234_5678);
        for _ in 0..500 {
            let mk = |rng: &mut Lcg| Lpi { p: rng.p(), q: rng.p(), r: rng.p(), s: rng.p(), t: rng.p() };
            let a = ImplicitPoint::Lpi(mk(&mut rng));
            let b = ImplicitPoint::Lpi(mk(&mut rng));
            let cc = ImplicitPoint::Lpi(mk(&mut rng));
            let c = rng.p();
            for axis in AXES {
                let ex2 = rational::orient2d_2i(&a, &b, c, axis);
                if let Some(s) = interval::orient2d_2i(&a, &b, c, axis) {
                    assert_eq!(s, ex2, "orient2d_2i interval wrong sign");
                }
                assert_eq!(orient2d_2i(&a, &b, c, axis), ex2, "2i cascade != exact");
                let ex3 = rational::orient2d_3i(&a, &b, &cc, axis);
                if let Some(s) = interval::orient2d_3i(&a, &b, &cc, axis) {
                    assert_eq!(s, ex3, "orient2d_3i interval wrong sign");
                }
                assert_eq!(orient2d_3i(&a, &b, &cc, axis), ex3, "3i cascade != exact");
            }
            let exl = rational::cmp_lex(&a, &b);
            if let Some(s) = interval::cmp_lex(&a, &b) {
                assert_eq!(s, exl, "cmp_lex interval wrong sign");
            }
            assert_eq!(cmp_lex(&a, &b), exl, "cmp_lex cascade != exact");
        }
    }

    #[test]
    fn cmp_lex_welds_coincident_lpi_and_tpi() {
        // An LPI and a TPI built from DIFFERENT constructions at the SAME physical
        // point must compare equal (Zero) so the interner welds them to one VID.
        let lpi = ImplicitPoint::Lpi(Lpi {
            p: [0.3, 0.4, -1.],
            q: [0.3, 0.4, 1.],
            r: [0., 0., 0.],
            s: [1., 0., 0.],
            t: [0., 1., 0.],
        });
        let tpi = ImplicitPoint::Tpi(Tpi {
            planes: [
                [[0., 0., 0.], [1., 0., 0.], [0., 1., 0.]],    // z=0
                [[0.3, 0., 0.], [0.3, 1., 0.], [0.3, 0., 1.]], // x=0.3
                [[0., 0.4, 0.], [1., 0.4, 0.], [0., 0.4, 1.]], // y=0.4
            ],
        });
        assert_eq!(rational::point_of(&lpi), rational::point_of(&tpi), "test points not coincident");
        assert_eq!(
            rational::cmp_lex(&lpi, &tpi),
            Sign::Zero,
            "coincident LPI/TPI not welded by cmp_lex"
        );
    }

    #[test]
    fn cmp_along_matches_materialised_oracle() {
        use num_rational::BigRational;
        use num_traits::Signed;
        let mut pts: Vec<ImplicitPoint> =
            vec![e([0., 0., 0.]), e([3., 1., -2.]), e([1.5, -2., 0.25])];
        pts.extend(lpi_cases().into_iter().take(2).map(|(l, ..)| ImplicitPoint::Lpi(l)));
        pts.extend(tpi_cases().into_iter().take(2).map(|(t, ..)| ImplicitPoint::Tpi(t)));
        let dirs = [[1., 0., 0.], [0., 1., 0.], [0., 0., 1.], [1., 2., -1.], [-0.3, 1.7, 0.5]];
        for u in dirs {
            let ur = [
                BigRational::from_float(u[0]).unwrap(),
                BigRational::from_float(u[1]).unwrap(),
                BigRational::from_float(u[2]).unwrap(),
            ];
            for a in &pts {
                for b in &pts {
                    let pa = rational::point_of(a);
                    let pb = rational::point_of(b);
                    let dot = (&pa[0] - &pb[0]) * &ur[0]
                        + (&pa[1] - &pb[1]) * &ur[1]
                        + (&pa[2] - &pb[2]) * &ur[2];
                    let want = if dot.is_negative() {
                        Sign::Negative
                    } else if dot.is_positive() {
                        Sign::Positive
                    } else {
                        Sign::Zero
                    };
                    assert_eq!(rational::cmp_along(a, b, u), want, "cmp_along != oracle (u={u:?})");
                }
            }
        }
    }
}