malachite-float 0.9.2

The arbitrary-precision floating-point type Float, with efficient algorithms partially derived from MPFR.
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
// Copyright © 2026 Mikhail Hogrefe
//
// This file is part of Malachite.
//
// Malachite is free software: you can redistribute it and/or modify it under the terms of the GNU
// Lesser General Public License (LGPL) as published by the Free Software Foundation; either version
// 3 of the License, or (at your option) any later version. See <https://www.gnu.org/licenses/>.

use crate::arithmetic::log_base_2::extended_log_base_2_of_rational;
use crate::basic::extended::ExtendedFloat;
use crate::{Float, emulate_rational_to_float_fn};
use core::cmp::Ordering::{self, *};
use malachite_base::num::arithmetic::traits::{CeilingLogBase2, CheckedLogBase, Sign};
use malachite_base::num::basic::floats::PrimitiveFloat;
use malachite_base::num::basic::integers::PrimitiveInt;
use malachite_base::num::basic::traits::{NaN, NegativeInfinity, Zero};
use malachite_base::num::conversion::traits::{ExactFrom, RoundingFrom};
use malachite_base::num::factorization::traits::ExpressAsPower;
use malachite_base::num::logic::traits::SignificantBits;
use malachite_base::rounding_modes::RoundingMode::{self, *};
use malachite_nz::natural::arithmetic::float_extras::float_can_round;
use malachite_nz::platform::Limb;
use malachite_q::Rational;

// Returns `Some(log_base(x))` when it is rational, and `None` when it is irrational (or when the
// inputs are too large for the check to be worthwhile). `x` must be positive and not equal to 1,
// and `base` must be greater than 1.
//
// `log_base(x)` is rational exactly when `x` and `base` are both powers of a common rational `g`,
// say `x = g^a` and `base = g^e_base`; then `log_base(x) = a / e_base`. Taking `g` to be the
// primitive root of `base` (`base.express_as_power()`), this holds iff `x` is an integer power of
// `g`, found by `Rational::checked_log_base` (which also covers `x < 1`, giving a negative `a`).
//
// Detecting these rational results up front is essential, not just an optimization: when the result
// is exactly representable (for example `log_9(3) = 1/2`), the Ziv loop in
// `log_base_rational_rational_base_prec_round_normal` would never terminate, because the rounding
// test can never certify a value sitting exactly on a representable point or tie.
//
// `express_as_power` perfect-power-tests `base`'s numerator and denominator, which is infeasible
// for an astronomically large (for example near-1) base, so the whole check is skipped when either
// input exceeds `64 * prec` bits. Such an input cannot be a power of `g` with a representable
// exponent at this precision (the common cases like `log_4(8)` involve tiny operands), so it is
// left to the Ziv loop. The one residual gap is a representable *fractional* result with a
// perfect-power base larger than the bound -- which would require multi-hundred-megabyte
// commensurable `x` and `base` -- where the loop would not terminate; such inputs are far beyond
// any realistic or testable range.
pub(crate) fn rational_log_base_rational_rational_base(
    x: &Rational,
    base: &Rational,
    prec: u64,
) -> Option<Rational> {
    let bound = prec.saturating_mul(64);
    if x.significant_bits() > bound || base.significant_bits() > bound {
        return None;
    }
    // `express_as_power` returns `None` when `base` is not a perfect power, in which case `base`
    // itself is `g` (with exponent 1).
    let (root, e_base) = base.express_as_power().unwrap_or_else(|| (base.clone(), 1));
    let a = x.checked_log_base(&root)?;
    Some(Rational::from_signeds(a, i64::exact_from(e_base)))
}

// Computes log_base(x) for `Rational` `x` and `base`, by log_base(x) = log_2(x) / log_2(base). The
// input `x` is positive, and `base` is greater than 1.
//
// Both logs are computed in the extended exponent range (see `extended_log_base_2_of_rational`) so
// that neither operand underflows when `x` or `base` is near 1, and so that their quotient may
// temporarily leave the representable range. The single conversion back to a `Float`, via
// `ExtendedFloat::into_float_helper`, performs the one correctly-rounded clamp to an infinity,
// maximum, zero, or minimum as dictated by the rounding mode. This handles both overflow (a base
// near 1, so `log_2(base)` is tiny and the quotient is huge) and underflow (an `x` near 1, so
// `log_2(x)` is tiny).
fn log_base_rational_rational_base_prec_round_normal(
    x: &Rational,
    base: &Rational,
    prec: u64,
    rm: RoundingMode,
) -> (Float, Ordering) {
    // If x is 1, the result is 0.
    if *x == 1u32 {
        return (Float::ZERO, Equal);
    }
    // If log_base(x) is rational -- x and base are both powers of a common rational -- compute it
    // directly. This includes exactly-representable results (which the Ziv loop could never
    // certify) as well as non-representable rationals (cheaper and exact this way).
    if let Some(q) = rational_log_base_rational_rational_base(x, base, prec) {
        return Float::from_rational_prec_round(q, prec, rm);
    }
    // The result is irrational, so it is never exactly representable.
    assert_ne!(rm, Exact, "Inexact log_base_rational_rational_base");
    // The initial slack keeps working_prec at least 7, so the working_prec - 6 below stays
    // positive.
    let mut working_prec = prec + 6 + prec.ceiling_log_base_2();
    let mut increment = Limb::WIDTH;
    loop {
        // log_2(x), extended (finite and nonzero, since x is positive and not 1).
        let num = extended_log_base_2_of_rational(x, working_prec);
        // log_2(base) > 0, extended.
        let den = extended_log_base_2_of_rational(base, working_prec);
        // log_2(x) / log_2(base) in the extended range; cannot overflow or underflow here.
        let (quotient, _) = num.div_prec_val_ref(&den, working_prec);
        // Each log is accurate to within 2 ulps and the division adds at most 1 more, for at most 5
        // ulps total; working_prec - 6 correct bits comfortably suffice for the rounding test.
        if float_can_round(
            quotient.x.significand_ref().unwrap(),
            working_prec - 6,
            prec,
            rm,
        ) {
            // Round the mantissa to prec, then place the extended exponent, clamping once to the
            // Float range as the rounding mode dictates.
            let (rounded, o) = Float::from_float_prec_round(quotient.x, prec, rm);
            let mut result = ExtendedFloat::from(rounded);
            result.exp = result.exp.checked_add(quotient.exp).unwrap();
            return result.into_float_helper(prec, rm, o);
        }
        // Increase the precision.
        working_prec += increment;
        increment = working_prec >> 1;
    }
}

impl Float {
    /// Computes $\log_b x$, where $x$ and the base $b$ are both [`Rational`]s with $b>1$, returning
    /// a [`Float`] rounded to the specified precision and with the specified rounding mode. Both
    /// are taken by value. An [`Ordering`] is also returned, indicating whether the rounded value
    /// is less than, equal to, or greater than the exact value. Although `NaN`s are not comparable
    /// to any [`Float`], whenever this function returns a `NaN` it also returns `Equal`.
    ///
    /// This computes $\log_2 x / \log_2 b$. Both logarithms are evaluated in an extended exponent
    /// range, so that an $x$ or $b$ extremely close to 1 (where the logarithm is tiny) does not
    /// lose accuracy, and the single conversion of the quotient back to a [`Float`] performs the
    /// one correctly-rounded clamp.
    ///
    /// See [`RoundingMode`] for a description of the possible rounding modes.
    ///
    /// $$
    /// f(x,b,p,m) = \log_b x+\varepsilon.
    /// $$
    /// - If $\log_b x$ is infinite, zero, or `NaN`, $\varepsilon$ may be ignored or assumed to be
    ///   0.
    /// - If $\log_b x$ is finite and nonzero, and $m$ is not `Nearest`, then $|\varepsilon| <
    ///   2^{\lfloor\log_2 |\log_b x|\rfloor-p+1}$.
    /// - If $\log_b x$ is finite and nonzero, and $m$ is `Nearest`, then $|\varepsilon| \leq
    ///   2^{\lfloor\log_2 |\log_b x|\rfloor-p}$.
    ///
    /// If the output has a precision, it is `prec`.
    ///
    /// Special cases:
    /// - $f(0,b,p,m)=-\infty$
    /// - $f(x,b,p,m)=\text{NaN}$ for $x<0$
    /// - $f(1,b,p,m)=0$
    /// - $f(x,b,p,m)=a/e$ when $x=g^a$, where $g$ is the primitive root of $b$ and $b=g^e$, rounded
    ///   to precision $p$; the result is exact if and only if $a/e$ is representable with precision
    ///   $p$ (for example $\log_4 8=3/2$ is exact)
    ///
    /// Like a logarithm of a [`Float`] with a [`Rational`] base, this can both overflow (for a base
    /// near 1) and underflow (for an $x$ near 1).
    ///
    /// # Worst-case complexity
    /// $T(n) = O(n (\log n)^2 \log\log n)$
    ///
    /// $M(n) = O(n (\log n)^2)$
    ///
    /// where $T$ is time, $M$ is additional memory, and $n$ is `prec`.
    ///
    /// # Panics
    /// Panics if `prec` is zero, if `base` is less than or equal to 1, or if `rm` is `Exact` but
    /// the result cannot be represented exactly with the given precision.
    ///
    /// # Examples
    /// ```
    /// use malachite_base::rounding_modes::RoundingMode::*;
    /// use malachite_float::Float;
    /// use malachite_q::Rational;
    /// use std::cmp::Ordering::*;
    ///
    /// let (log, o) = Float::log_base_rational_rational_base_prec_round(
    ///     Rational::from(8),
    ///     Rational::from(4),
    ///     10,
    ///     Exact,
    /// );
    /// assert_eq!(log.to_string(), "1.5"); // log_4(8) = 3/2
    /// assert_eq!(o, Equal);
    ///
    /// let (log, o) = Float::log_base_rational_rational_base_prec_round(
    ///     Rational::from(2),
    ///     Rational::from(3),
    ///     10,
    ///     Floor,
    /// );
    /// assert_eq!(log.to_string(), "0.631"); // log_3(2) = 0.6309...
    /// assert_eq!(o, Less);
    /// ```
    #[allow(clippy::needless_pass_by_value)]
    #[inline]
    pub fn log_base_rational_rational_base_prec_round(
        x: Rational,
        base: Rational,
        prec: u64,
        rm: RoundingMode,
    ) -> (Self, Ordering) {
        Self::log_base_rational_rational_base_prec_round_ref(&x, &base, prec, rm)
    }

    /// Computes $\log_b x$, where $x$ and the base $b$ are both [`Rational`]s with $b>1$, returning
    /// a [`Float`] rounded to the specified precision and with the specified rounding mode. Both
    /// are taken by reference. An [`Ordering`] is also returned, indicating whether the rounded
    /// value is less than, equal to, or greater than the exact value.
    ///
    /// See [`Float::log_base_rational_rational_base_prec_round`] for details, special cases, and a
    /// description of the rounding behavior.
    ///
    /// # Worst-case complexity
    /// $T(n) = O(n (\log n)^2 \log\log n)$
    ///
    /// $M(n) = O(n (\log n)^2)$
    ///
    /// where $T$ is time, $M$ is additional memory, and $n$ is `prec`.
    ///
    /// # Panics
    /// Panics if `prec` is zero, if `base` is less than or equal to 1, or if `rm` is `Exact` but
    /// the result cannot be represented exactly with the given precision.
    ///
    /// # Examples
    /// ```
    /// use malachite_base::rounding_modes::RoundingMode::*;
    /// use malachite_float::Float;
    /// use malachite_q::Rational;
    /// use std::cmp::Ordering::*;
    ///
    /// let (log, o) = Float::log_base_rational_rational_base_prec_round_ref(
    ///     &Rational::from(9),
    ///     &Rational::from(3),
    ///     10,
    ///     Exact,
    /// );
    /// assert_eq!(log.to_string(), "2.0"); // log_3(9) = 2
    /// assert_eq!(o, Equal);
    ///
    /// let (log, o) = Float::log_base_rational_rational_base_prec_round_ref(
    ///     &Rational::from_signeds(1, 8),
    ///     &Rational::from(2),
    ///     10,
    ///     Exact,
    /// );
    /// assert_eq!(log.to_string(), "-3.0"); // log_2(1/8) = -3
    /// assert_eq!(o, Equal);
    /// ```
    pub fn log_base_rational_rational_base_prec_round_ref(
        x: &Rational,
        base: &Rational,
        prec: u64,
        rm: RoundingMode,
    ) -> (Self, Ordering) {
        assert_ne!(prec, 0);
        assert!(*base > 1u32, "Logarithm base must be greater than 1");
        match x.sign() {
            Less => (Self::NAN, Equal),
            Equal => (Self::NEGATIVE_INFINITY, Equal),
            Greater => log_base_rational_rational_base_prec_round_normal(x, base, prec, rm),
        }
    }

    /// Computes $\log_b x$, where $x$ and the base $b$ are both [`Rational`]s with $b>1$, returning
    /// a [`Float`] rounded to the nearest value of the specified precision. Both are taken by
    /// value. An [`Ordering`] is also returned.
    ///
    /// See [`Float::log_base_rational_rational_base_prec_round`] for details and special cases.
    ///
    /// # Worst-case complexity
    /// $T(n) = O(n (\log n)^2 \log\log n)$
    ///
    /// $M(n) = O(n (\log n)^2)$
    ///
    /// where $T$ is time, $M$ is additional memory, and $n$ is `prec`.
    ///
    /// # Panics
    /// Panics if `prec` is zero or if `base` is less than or equal to 1.
    ///
    /// # Examples
    /// ```
    /// use malachite_float::Float;
    /// use malachite_q::Rational;
    /// use std::cmp::Ordering::*;
    ///
    /// let (log, o) =
    ///     Float::log_base_rational_rational_base_prec(Rational::from(8), Rational::from(4), 10);
    /// assert_eq!(log.to_string(), "1.5"); // log_4(8) = 3/2
    /// assert_eq!(o, Equal);
    /// ```
    #[allow(clippy::needless_pass_by_value)]
    #[inline]
    pub fn log_base_rational_rational_base_prec(
        x: Rational,
        base: Rational,
        prec: u64,
    ) -> (Self, Ordering) {
        Self::log_base_rational_rational_base_prec_round_ref(&x, &base, prec, Nearest)
    }

    /// Computes $\log_b x$, where $x$ and the base $b$ are both [`Rational`]s with $b>1$, returning
    /// a [`Float`] rounded to the nearest value of the specified precision. Both are taken by
    /// reference. An [`Ordering`] is also returned.
    ///
    /// See [`Float::log_base_rational_rational_base_prec_round`] for details and special cases.
    ///
    /// # Worst-case complexity
    /// $T(n) = O(n (\log n)^2 \log\log n)$
    ///
    /// $M(n) = O(n (\log n)^2)$
    ///
    /// where $T$ is time, $M$ is additional memory, and $n$ is `prec`.
    ///
    /// # Panics
    /// Panics if `prec` is zero or if `base` is less than or equal to 1.
    ///
    /// # Examples
    /// ```
    /// use malachite_float::Float;
    /// use malachite_q::Rational;
    /// use std::cmp::Ordering::*;
    ///
    /// let (log, o) = Float::log_base_rational_rational_base_prec_ref(
    ///     &Rational::from(9),
    ///     &Rational::from(3),
    ///     10,
    /// );
    /// assert_eq!(log.to_string(), "2.0"); // log_3(9) = 2
    /// assert_eq!(o, Equal);
    /// ```
    #[inline]
    pub fn log_base_rational_rational_base_prec_ref(
        x: &Rational,
        base: &Rational,
        prec: u64,
    ) -> (Self, Ordering) {
        Self::log_base_rational_rational_base_prec_round_ref(x, base, prec, Nearest)
    }
}

/// Computes $\log_b x$, the base-$b$ logarithm of a [`Rational`], where the base $b$ is also a
/// [`Rational`] greater than 1, returning a primitive float result. Using this function is more
/// accurate than computing the logarithm using the standard library, whose logarithm functions are
/// not always correctly rounded.
///
/// If the logarithm is equidistant from two primitive floats, the primitive float with fewer 1s in
/// its binary expansion is chosen. See [`RoundingMode`] for a description of the `Nearest` rounding
/// mode.
///
/// The base-$b$ logarithm of any negative number is `NaN`.
///
/// $$
/// f(x,b) = \log_b x+\varepsilon.
/// $$
/// - If $\log_b x$ is infinite, zero, or `NaN`, $\varepsilon$ may be ignored or assumed to be 0.
/// - If $\log_b x$ is finite and nonzero, then $|\varepsilon| < 2^{\lfloor\log_2 |\log_b
///   x|\rfloor-p}$, where $p$ is precision of the output (typically 24 if `T` is a [`f32`] and 53
///   if `T` is a [`f64`], but less if the output is subnormal).
///
/// Special cases:
/// - $f(0,b)=-\infty$
/// - $f(x,b)=\text{NaN}$ for $x<0$
/// - $f(1,b)=0.0$
///
/// Unlike a logarithm with an integer base, this function can both overflow (for a base near 1) and
/// underflow (for an $x$ near 1).
///
/// # Worst-case complexity
/// Constant time and additional memory.
///
/// # Panics
/// Panics if `base` is less than or equal to 1.
///
/// # Examples
/// ```
/// use malachite_base::num::basic::traits::{NegativeInfinity, Zero};
/// use malachite_base::num::float::NiceFloat;
/// use malachite_float::arithmetic::log_base_rational_rational_base::*;
/// use malachite_q::Rational;
///
/// assert_eq!(
///     NiceFloat(primitive_float_log_base_rational_rational_base::<f32>(
///         &Rational::ZERO,
///         &Rational::from(10)
///     )),
///     NiceFloat(f32::NEGATIVE_INFINITY)
/// );
/// // log_4(8) = 3/2
/// assert_eq!(
///     NiceFloat(primitive_float_log_base_rational_rational_base::<f32>(
///         &Rational::from(8),
///         &Rational::from(4)
///     )),
///     NiceFloat(1.5)
/// );
/// // log_(3/2)(9/4) = 2
/// assert_eq!(
///     NiceFloat(primitive_float_log_base_rational_rational_base::<f32>(
///         &Rational::from_unsigneds(9u8, 4),
///         &Rational::from_unsigneds(3u8, 2)
///     )),
///     NiceFloat(2.0)
/// );
/// // log_10(50)
/// assert_eq!(
///     NiceFloat(primitive_float_log_base_rational_rational_base::<f32>(
///         &Rational::from(50),
///         &Rational::from(10)
///     )),
///     NiceFloat(1.69897)
/// );
/// assert_eq!(
///     NiceFloat(primitive_float_log_base_rational_rational_base::<f32>(
///         &Rational::from(-1000),
///         &Rational::from(10)
///     )),
///     NiceFloat(f32::NAN)
/// );
/// ```
#[inline]
#[allow(clippy::type_repetition_in_bounds)]
pub fn primitive_float_log_base_rational_rational_base<T: PrimitiveFloat>(
    x: &Rational,
    base: &Rational,
) -> T
where
    Float: PartialOrd<T>,
    for<'a> T: ExactFrom<&'a Float> + RoundingFrom<&'a Float>,
{
    emulate_rational_to_float_fn(
        |x, prec| Float::log_base_rational_rational_base_prec_ref(x, base, prec),
        x,
    )
}