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
use core::cmp::Ordering;

use Iter;

pub trait Bits: Eq + PartialEq + PartialOrd + Ord + Copy {
    fn as_u64(self) -> u64;
}
impl Bits for u32 {
    fn as_u64(self) -> u64 { self as u64 }
}
impl Bits for u64 {
    fn as_u64(self) -> u64 { self }
}

/// Types that are IEEE754 floating point numbers.
pub trait Ieee754: Copy + PartialEq + PartialOrd {
    /// Iterate over each value of `Self` in `[self, lim]`.
    ///
    /// The returned iterator will include subnormal numbers, and will
    /// only include one of `-0.0` and `0.0`.
    ///
    /// # Panics
    ///
    /// Panics if `self > lim`, or if either are NaN.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// // there are 840 single-precision floats in between 1.0 and 1.0001
    /// // (inclusive).
    /// assert_eq!(1_f32.upto(1.0001).count(), 840);
    /// ```
    fn upto(self, lim: Self) -> Iter<Self>;

    /// A type that represents the raw bits of `Self`.
    type Bits: Bits;
    /// A type large enough to store the true exponent of `Self`.
    type Exponent;
    /// A type large enough to store the raw exponent (i.e. with the bias).
    type RawExponent;
    /// A type large enough to store the significand of `Self`.
    type Significand;

    /// Return the next value after `self`.
    ///
    /// Calling this on NaN or positive infinity will yield nonsense.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    /// let x: f32 = 1.0;
    /// assert_eq!(x.next(), 1.000000119209);
    /// ```
    fn next(self) -> Self;

    /// Return the previous value before `self`.
    ///
    /// Calling this on NaN or negative infinity will yield nonsense.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    /// let x: f32 = 1.0;
    /// assert_eq!(x.prev(), 0.99999995);
    /// ```
    fn prev(self) -> Self;

    /// Return the unit-in-the-last-place ulp of `self`. That is,
    /// `x.abs().next() - x.abs()`, but handling overflow properly.
    ///
    /// Returns `None` if `self` is not finite.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use std::f32;
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(0_f32.ulp(), Some(1.4e-45));
    ///
    /// assert_eq!(1_f32.ulp(), Some(1.1920928955078125e-07));
    /// assert_eq!((-1_f32).ulp(), Some(1.1920928955078125e-07));
    ///
    /// // 2^23
    /// assert_eq!(8_388_608_f32.ulp(), Some(1.0));
    /// // 2^24 - 1, the largest f32 with ULP 1
    /// assert_eq!(16_777_215_f32.ulp(), Some(1.0));
    /// // 2^24
    /// assert_eq!(16_777_216_f32.ulp(), Some(2.0));
    ///
    /// // non-finite
    /// assert_eq!(f32::INFINITY.ulp(), None);
    /// assert_eq!(f32::NAN.ulp(), None);
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use std::f64;
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(0_f64.ulp(), Some(4.9e-324));
    ///
    /// assert_eq!(1_f64.ulp(), Some(2.220446049250313e-16));
    /// assert_eq!((-1_f64).ulp(), Some(2.220446049250313e-16));
    ///
    /// // 2^52
    /// assert_eq!(4_503_599_627_370_496_f64.ulp(), Some(1.0));
    /// // 2^53 - 1, the largest f64 with ULP 1
    /// assert_eq!(9_007_199_254_740_991_f64.ulp(), Some(1.0));
    /// // 2^53
    /// assert_eq!(9_007_199_254_740_992_f64.ulp(), Some(2.0));
    ///
    /// // non-finite
    /// assert_eq!(f64::INFINITY.ulp(), None);
    /// assert_eq!(f64::NAN.ulp(), None);
    /// ```
    fn ulp(self) -> Option<Self>;

    /// View `self` as a collection of bits.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    /// let x: f32 = 1.0;
    /// assert_eq!(x.bits(), 0x3f80_0000);
    /// ```
    fn bits(self) -> Self::Bits;

    /// View a collections of bits as a floating point number.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    /// let float: f32 = Ieee754::from_bits(0xbf80_0000);
    /// assert_eq!(float, -1.0);
    /// ```
    fn from_bits(x: Self::Bits) -> Self;

    /// Get the bias of the stored exponent.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(f32::exponent_bias(), 127);
    /// assert_eq!(f64::exponent_bias(), 1023);
    /// ```
    fn exponent_bias() -> Self::Exponent;

    /// Break `self` into the three constituent parts of an IEEE754 float.
    ///
    /// The exponent returned is the raw bits, use `exponent_bias` to
    /// compute the offset required or use `decompose` to obtain this
    /// in precomputed form.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(1_f32.decompose_raw(), (false, 127, 0));
    /// assert_eq!(1234.567_f32.decompose_raw(), (false, 137, 0x1a5225));
    ///
    /// assert_eq!((-0.525_f32).decompose_raw(), (true, 126, 0x66666));
    ///
    /// assert_eq!(std::f32::INFINITY.decompose_raw(), (false, 255, 0));
    ///
    /// let (sign, expn, signif) = std::f32::NAN.decompose_raw();
    /// assert_eq!((sign, expn), (false, 255));
    /// assert!(signif != 0);
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(1_f64.decompose_raw(), (false, 1023, 0));
    /// assert_eq!(1234.567_f64.decompose_raw(), (false, 1033, 0x34a449ba5e354));
    ///
    /// assert_eq!((-0.525_f64).decompose_raw(), (true, 1022, 0xcccc_cccc_cccd));
    ///
    /// assert_eq!(std::f64::INFINITY.decompose_raw(), (false, 2047, 0));
    ///
    /// let (sign, expn, signif) = std::f64::NAN.decompose_raw();
    /// assert_eq!((sign, expn), (false, 2047));
    /// assert!(signif != 0);
    /// ```
    fn decompose_raw(self) -> (bool, Self::RawExponent, Self::Significand);

    /// Create a `Self` out of the three constituent parts of an IEEE754 float.
    ///
    /// This returns (-1)<sup><code>sign</code></sup> ×
    /// 1.<code>signif</code> × 2<sup><code>expn</code> - bias</sup>, where
    ///
    /// - `sign` is treated as if `true` == `1` (meaning `true` is
    ///   negative),
    /// - 1.<code>signif</code> refers to placing the bits of `signif`
    ///   as the fractional part of a number between 1 and 2, and
    /// - bias is the exponent bias for this float (see [`exponent_bias`]).
    ///
    /// The exponent should be the raw bits: use `exponent_bias` to
    /// compute the offset required, or use `recompose` to feed in an
    /// unbiased exponent.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(f32::recompose_raw(false, 127, 0), 1.0);
    /// assert_eq!(f32::recompose_raw(false, 137, 0x1a5225), 1234.567);
    /// assert_eq!(f32::recompose_raw(true, 126, 0x66666), -0.525);
    ///
    /// assert_eq!(f32::recompose_raw(false, 255, 0), std::f32::INFINITY);
    ///
    /// assert!(f32::recompose_raw(false, 255, 1).is_nan());
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(f64::recompose_raw(false, 1023, 0), 1.0);
    /// assert_eq!(f64::recompose_raw(false, 1033, 0x34a449ba5e354), 1234.567);
    /// assert_eq!(f64::recompose_raw(true, 1022, 0xcccc_cccc_cccd), -0.525);
    ///
    /// assert_eq!(f64::recompose_raw(false, 2047, 0), std::f64::INFINITY);
    ///
    /// assert!(f64::recompose_raw(false, 2047, 1).is_nan());
    /// ```
    fn recompose_raw(sign: bool, expn: Self::RawExponent, signif: Self::Significand) -> Self;

    /// Break `self` into the three constituent parts of an IEEE754 float.
    ///
    /// The exponent returned is the true exponent, after accounting
    /// for the bias it is stored with. The significand does not
    /// include the implicit highest bit (if it exists), e.g. the
    /// 24-bit for single precision.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(1_f32.decompose(), (false, 0, 0));
    /// assert_eq!(1234.567_f32.decompose(), (false, 10, 0x1a5225));
    ///
    /// assert_eq!((-0.525_f32).decompose(), (true, -1, 0x66666));
    ///
    /// assert_eq!(std::f32::INFINITY.decompose(), (false, 128, 0));
    ///
    /// let (sign, expn, signif) = std::f32::NAN.decompose();
    /// assert_eq!((sign, expn), (false, 128));
    /// assert!(signif != 0);
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(1_f64.decompose(), (false, 0, 0));
    /// assert_eq!(1234.567_f64.decompose(), (false, 10, 0x34a449ba5e354));
    ///
    /// assert_eq!((-0.525_f64).decompose(), (true, -1, 0xcccc_cccc_cccd));
    ///
    /// assert_eq!(std::f64::INFINITY.decompose(), (false, 1024, 0));
    ///
    /// let (sign, expn, signif) = std::f64::NAN.decompose();
    /// assert_eq!((sign, expn), (false, 1024));
    /// assert!(signif != 0);
    /// ```
    fn decompose(self) -> (bool, Self::Exponent, Self::Significand);

    /// Create a `Self` out of the three constituent parts of an IEEE754 float.
    ///
    /// This returns (-1)<sup><code>sign</code></sup> ×
    /// 1.<code>signif</code> × 2<sup><code>expn</code></sup>, where
    ///
    /// - `sign` is treated as if `true` == `1` (meaning `true` is
    ///   negative), and
    /// - 1.<code>signif</code> refers to placing the bits of `signif`
    ///   as the fractional part of a number between 1 and 2.
    ///
    /// The exponent should be the true exponent, not accounting for any
    /// bias. The significand should not include the implicit highest
    /// bit (if it exists), e.g. the 24-th bit for single precision.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// // normal numbers
    /// assert_eq!(f32::recompose(false, 0, 0), 1.0);
    /// assert_eq!(f32::recompose(false, 10, 0x1a5225), 1234.567);
    /// assert_eq!(f32::recompose(true, -1, 0x66666), -0.525);
    ///
    /// // infinity
    /// assert_eq!(f32::recompose(false, 128, 0), std::f32::INFINITY);
    ///
    /// // NaN
    /// assert!(f32::recompose(false, 128, 1).is_nan());
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use ieee754::Ieee754;
    ///
    /// // normal numbers
    /// assert_eq!(f64::recompose(false, 0, 0), 1.0);
    /// assert_eq!(f64::recompose(false, 10, 0x34a449ba5e354), 1234.567);
    /// assert_eq!(f64::recompose(true, -1, 0xcccc_cccc_cccd), -0.525);
    ///
    /// // infinity
    /// assert_eq!(f64::recompose(false, 1024, 0), std::f64::INFINITY);
    ///
    /// // NaN
    /// assert!(f64::recompose(false, 1024, 1).is_nan());
    /// ```
    fn recompose(sign: bool, expn: Self::Exponent, signif: Self::Significand) -> Self;

    /// Compare `x` and `y` using the IEEE-754 `totalOrder` predicate
    /// (Section 5.10).
    ///
    /// This orders NaNs before or after all non-NaN floats, depending
    /// on the sign bit. Using -qNaN to represent a quiet NaN with
    /// negative sign bit and similarly for a signalling NaN (sNaN),
    /// the order is:
    ///
    /// ```txt
    /// -qNaN < -sNaN < -∞ < -12.34 < -0.0 < +0.0 < +12.34 < +∞ < +sNaN < +qNaN
    /// ```
    ///
    /// (NaNs are ordered according to their payload.)
    ///
    /// # Examples
    ///
    /// Sorting:
    ///
    /// ```rust
    /// use std::f32;
    ///
    /// use ieee754::Ieee754;
    ///
    /// let mut data = vec![0.0, f32::NEG_INFINITY, -1.0, f32::INFINITY,
    ///                     f32::NAN, -0.0, 12.34e5, -f32::NAN];
    /// data.sort_by(|a, b| a.total_cmp(b));
    ///
    /// assert_eq!(format!("{:.0?}", data),
    ///            "[NaN, -inf, -1, -0, 0, 1234000, inf, NaN]");
    /// ```
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use std::cmp::Ordering;
    /// use std::f32;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // normal comparison
    /// assert_eq!(0_f32.total_cmp(&0_f32), Ordering::Equal);
    /// assert_eq!(0_f32.total_cmp(&1_f32), Ordering::Less);
    /// assert_eq!(1e10_f32.total_cmp(&f32::NEG_INFINITY), Ordering::Greater);
    ///
    /// // signed zero
    /// assert_eq!(0_f32.total_cmp(&-0_f32), Ordering::Greater);
    ///
    /// // NaNs
    /// assert_eq!(f32::NAN.total_cmp(&0_f32), Ordering::Greater);
    /// assert_eq!(f32::NAN.total_cmp(&f32::INFINITY), Ordering::Greater);
    /// assert_eq!((-f32::NAN).total_cmp(&f32::NEG_INFINITY), Ordering::Less);
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use std::cmp::Ordering;
    /// use std::f64;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // normal comparison
    /// assert_eq!(0_f64.total_cmp(&0_f64), Ordering::Equal);
    /// assert_eq!(0_f64.total_cmp(&1_f64), Ordering::Less);
    /// assert_eq!(1e10_f64.total_cmp(&f64::NEG_INFINITY), Ordering::Greater);
    ///
    /// // signed zero
    /// assert_eq!(0_f64.total_cmp(&-0_f64), Ordering::Greater);
    ///
    /// // NaNs
    /// assert_eq!(f64::NAN.total_cmp(&0_f64), Ordering::Greater);
    /// assert_eq!(f64::NAN.total_cmp(&f64::INFINITY), Ordering::Greater);
    /// assert_eq!((-f64::NAN).total_cmp(&f64::NEG_INFINITY), Ordering::Less);
    /// ```
    fn total_cmp(&self, other: &Self) -> Ordering;

    /// Return the absolute value of `x`.
    ///
    /// This provides a no_std/core-only version of the built-in `abs` in
    /// `std`, until
    /// [#50145](https://github.com/rust-lang/rust/issues/50145) is
    /// addressed.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// #![no_std]
    /// # extern crate std; // this makes this "test" a lie, unfortunately
    /// # extern crate ieee754;
    /// use core::f32;
    ///
    /// use ieee754::Ieee754;
    ///
    /// # fn main() {
    /// assert_eq!((0_f32).abs(), 0.0);
    ///
    /// assert_eq!((12.34_f32).abs(), 12.34);
    /// assert_eq!((-12.34_f32).abs(), 12.34);
    ///
    /// assert_eq!(f32::INFINITY.abs(), f32::INFINITY);
    /// assert_eq!(f32::NEG_INFINITY.abs(), f32::INFINITY);
    /// assert!(f32::NAN.abs().is_nan());
    /// # }
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// #![no_std]
    /// # extern crate std; // this makes this "test" a lie, unfortunately
    /// # extern crate ieee754;
    /// use core::f64;
    ///
    /// use ieee754::Ieee754;
    ///
    /// # fn main() {
    /// assert_eq!((0_f64).abs(), 0.0);
    ///
    /// assert_eq!((12.34_f64).abs(), 12.34);
    /// assert_eq!((-12.34_f64).abs(), 12.34);
    ///
    /// assert_eq!(f64::INFINITY.abs(), f64::INFINITY);
    /// assert_eq!(f64::NEG_INFINITY.abs(), f64::INFINITY);
    /// assert!(f64::NAN.abs().is_nan());
    /// # }
    /// ```
    fn abs(self) -> Self;

    /// Return a float with the magnitude of `self` but the sign of
    /// `sign`.
    ///
    /// If `sign` is NaN, this still uses its sign bit, and does not
    /// (necessarily) return NaN.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use std::f32;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // normal numbers
    /// assert_eq!(1_f32.copy_sign(1.0), 1.0);
    /// assert_eq!(2_f32.copy_sign(-1.0), -2.0);
    /// assert_eq!((-3_f32).copy_sign(1.0), 3.0);
    /// assert_eq!((-4_f32).copy_sign(-1.0), -4.0);
    ///
    /// // infinities
    /// assert_eq!(5_f32.copy_sign(f32::NEG_INFINITY), -5.0);
    /// assert_eq!(f32::NEG_INFINITY.copy_sign(1.0), f32::INFINITY);
    ///
    /// // signs of zeros matter
    /// assert_eq!((-6_f32).copy_sign(0.0), 6.0);
    /// assert_eq!(7_f32.copy_sign(-0.0), -7.0);
    ///
    /// // NaNs only propagate on the self argument
    /// assert!(f32::NAN.copy_sign(1.0).is_nan());
    /// assert_eq!(8_f32.copy_sign(-f32::NAN), -8.0);
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use std::f64;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // normal numbers
    /// assert_eq!(1_f64.copy_sign(1.0), 1.0);
    /// assert_eq!(2_f64.copy_sign(-1.0), -2.0);
    /// assert_eq!((-3_f64).copy_sign(1.0), 3.0);
    /// assert_eq!((-4_f64).copy_sign(-1.0), -4.0);
    ///
    /// // infinities
    /// assert_eq!(5_f64.copy_sign(f64::NEG_INFINITY), -5.0);
    /// assert_eq!(f64::NEG_INFINITY.copy_sign(1.0), f64::INFINITY);
    ///
    /// // signs of zeros matter
    /// assert_eq!((-6_f64).copy_sign(0.0), 6.0);
    /// assert_eq!(7_f64.copy_sign(-0.0), -7.0);
    ///
    /// // NaNs only propagate on the self argument
    /// assert!(f64::NAN.copy_sign(1.0).is_nan());
    /// assert_eq!(8_f64.copy_sign(-f64::NAN), -8.0);
    /// ```
    fn copy_sign(self, sign: Self) -> Self;

    /// Return the sign of `x`.
    ///
    /// This provides a no_std/core-only function similar to the
    /// built-in `signum` in `std` (until
    /// [#50145](https://github.com/rust-lang/rust/issues/50145) is
    /// addressed). This `sign` function differs at two values; it
    /// matches the mathematical definitions when `self == 0.0` :
    ///
    /// | `x` | `x.signum()` (`std`) | `x.sign()` (`ieee754`) |
    /// |--:|--:|--:|
    /// |< 0.0|−1.0|−1.0|
    /// |−0.0|−1.0|**−0.0**|
    /// |+0.0|+1.0|**+0.0**|
    /// |> 0.0|+1.0|+1.0|
    /// |NaN|NaN|NaN|
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use std::f32;
    /// use std::cmp::Ordering;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // zeros
    /// assert_eq!(0_f32.sign().total_cmp(&0.0), Ordering::Equal);
    /// assert_eq!((-0_f32).sign().total_cmp(&-0.0), Ordering::Equal);
    ///
    /// // normal numbers
    /// assert_eq!((12.34_f32).sign(), 1.0);
    /// assert_eq!((-12.34_f32).sign(), -1.0);
    ///
    /// // extremes
    /// assert_eq!(f32::INFINITY.sign(), 1.0);
    /// assert_eq!(f32::NEG_INFINITY.sign(), -1.0);
    /// assert!(f32::NAN.sign().is_nan());
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use std::f64;
    /// use std::cmp::Ordering;
    ///
    /// use ieee754::Ieee754;
    ///
    /// // zeros
    /// assert_eq!(0_f64.sign().total_cmp(&0.0), Ordering::Equal);
    /// assert_eq!((-0_f64).sign().total_cmp(&-0.0), Ordering::Equal);
    ///
    /// // normal numbers
    /// assert_eq!((12.34_f64).sign(), 1.0);
    /// assert_eq!((-12.34_f64).sign(), -1.0);
    ///
    /// // extremes
    /// assert_eq!(f64::INFINITY.sign(), 1.0);
    /// assert_eq!(f64::NEG_INFINITY.sign(), -1.0);
    /// assert!(f64::NAN.sign().is_nan());
    /// ```
    fn sign(self) -> Self;

    /// Compute the (generalized) **signed** relative error of `self`
    /// as an approximation to `exact`.
    ///
    /// This computes the signed value: positive indicates `self` in
    /// the opposite direction to 0 from `exact`; negative indicates
    /// `self` is in the same direction as 0 from `exact`. Use
    /// `x.rel_error(exact).abs()` to get the non-signed relative
    /// error.
    ///
    /// The "generalized" refers to `exact` being 0 or ±∞ the handling
    /// of which is designed to indicate a "failure" (infinite error),
    /// if `self` doesn't precisely equal `exact`. This behaviour is
    /// designed for checking output of algorithms on floats when it
    /// is often desirable to match 0.0 and ±∞ perfectly.
    ///
    /// The values of this function are:
    ///
    /// |`exact`|`x`|`x.rel_error(exact)`|
    /// |--:|--:|--:|
    /// |NaN|any value|NaN|
    /// |any value|NaN|NaN|
    /// |0|equal to `exact`|0|
    /// |0|not equal to `exact`|signum(`x`) × ∞|
    /// |±∞|equal to `exact`|0|
    /// |±∞|not equal to `exact`|-∞|
    /// |any other value|any value|`(x - exact) / exact`|
    ///
    /// The sign of a zero-valued argument has no effect on the result
    /// of this function.
    ///
    /// # Examples
    ///
    /// Single precision:
    ///
    /// ```rust
    /// use std::f32;
    ///
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(4_f32.rel_error(4.0), 0.0);
    /// assert_eq!(3_f32.rel_error(4.0), -0.25);
    /// assert_eq!(5_f32.rel_error(4.0), 0.25);
    ///
    /// // zero
    /// assert_eq!(0_f32.rel_error(0.0), 0.0);
    /// assert_eq!(1_f32.rel_error(0.0), f32::INFINITY);
    /// assert_eq!((-1_f32).rel_error(0.0), f32::NEG_INFINITY);
    ///
    /// // infinities
    /// assert_eq!(f32::INFINITY.rel_error(f32::INFINITY), 0.0);
    /// assert_eq!(0_f32.rel_error(f32::INFINITY), f32::NEG_INFINITY);
    ///
    /// assert_eq!(f32::NEG_INFINITY.rel_error(f32::NEG_INFINITY), 0.0);
    /// assert_eq!(0_f32.rel_error(f32::NEG_INFINITY), f32::NEG_INFINITY);
    ///
    /// // NaNs
    /// assert!(f32::NAN.rel_error(4.0).is_nan());
    /// assert!(4_f32.rel_error(f32::NAN).is_nan());
    /// ```
    ///
    /// Double precision:
    ///
    /// ```rust
    /// use std::f64;
    /// use ieee754::Ieee754;
    ///
    /// assert_eq!(4_f64.rel_error(4.0), 0.0);
    /// assert_eq!(3_f64.rel_error(4.0), -0.25);
    /// assert_eq!(5_f64.rel_error(4.0), 0.25);
    ///
    /// // zero
    /// assert_eq!(0_f64.rel_error(0.0), 0.0);
    /// assert_eq!(1_f64.rel_error(0.0), f64::INFINITY);
    /// assert_eq!((-1_f64).rel_error(0.0), f64::NEG_INFINITY);
    ///
    /// // infinities
    /// assert_eq!(f64::INFINITY.rel_error(f64::INFINITY), 0.0);
    /// assert_eq!(0_f64.rel_error(f64::INFINITY), f64::NEG_INFINITY);
    ///
    /// assert_eq!(f64::NEG_INFINITY.rel_error(f64::NEG_INFINITY), 0.0);
    /// assert_eq!(0_f64.rel_error(f64::NEG_INFINITY), f64::NEG_INFINITY);
    ///
    /// // NaNs
    /// assert!(f64::NAN.rel_error(4.0).is_nan());
    /// assert!(4_f64.rel_error(f64::NAN).is_nan());
    /// ```
    fn rel_error(self, exact: Self) -> Self;
}