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
//! Implementations for various factorization algorithms.
//!
//! Note general prime number field sieve is not planned to be implemented, since it's too complex
//!
//! See <https://web.archive.org/web/20110331180514/https://diamond.boisestate.edu/~liljanab/BOISECRYPTFall09/Jacobsen.pdf>
//! for a detailed comparison between different factorization algorithms


// XXX: make the factorization method resumable? Maybe let all of them returns a Future

use crate::traits::ExactRoots;
use num_integer::{Integer, Roots};
use num_modular::{ModularCoreOps, ModularUnaryOps};
use num_traits::{FromPrimitive, NumRef, RefNum, CheckedAdd};
use std::collections::BTreeMap;

/// Find factors by trial division, returns a tuple of the found factors and the residual.
///
/// The target is guaranteed fully factored only if bound * bound > target, where bound = max(primes).
/// The parameter limit additionally sets the maximum of primes to be tried.
/// The residual will be Ok(1) or Ok(p) if fully factored.
///
/// TODO: implement fast check for small primes with BigInts in the precomputed table, and skip them in this function
pub fn trial_division<
    I: Iterator<Item = u64>,
    T: Integer + Clone + Roots + NumRef + FromPrimitive,
>(
    primes: I,
    target: T,
    limit: Option<u64>,
) -> (BTreeMap<u64, usize>, Result<T, T>)
where
    for<'r> &'r T: RefNum<T>,
{
    let tsqrt: T = Roots::sqrt(&target) + T::one();
    let limit = if let Some(l) = limit {
        tsqrt.clone().min(T::from_u64(l).unwrap())
    } else {
        tsqrt.clone()
    };

    let mut residual = target;
    let mut result = BTreeMap::new();
    let mut factored = false;
    for (p, pt) in primes.map(|p| (p, T::from_u64(p).unwrap())) {
        if &pt > &tsqrt {
            factored = true;
        }
        if &pt > &limit {
            break;
        }

        while residual.is_multiple_of(&pt) {
            residual = residual / &pt;
            *result.entry(p).or_insert(0) += 1;
        }
        if residual.is_one() {
            factored = true;
            break;
        }
    }

    if factored {
        (result, Ok(residual))
    } else {
        (result, Err(residual))
    }
}

/// Find factors using Pollard's rho algorithm with Brent's loop detection algorithm
/// 
/// The returned values are the factor and the count of passed iterations.
pub fn pollard_rho<
    T: Integer
        + FromPrimitive
        + NumRef
        + Clone
        + for<'r> ModularCoreOps<&'r T, &'r T, Output = T>
        + for<'r> ModularUnaryOps<&'r T, Output = T>,
>(
    target: &T,
    start: T,
    offset: T,
    max_iter: usize,
) -> (Option<T>, usize)
where
    for<'r> &'r T: RefNum<T>,
{
    let mut a = start.clone();
    let mut b = start.clone();
    let mut z = T::one() % target; // accumulator for gcd

    // using Brent's loop detection, i = tortoise, j = hare
    let (mut i, mut j) = (0usize, 1usize);

    // backtracing states
    let mut s = start;
    let mut backtrace = false;

    while i < max_iter {
        i += 1;
        a = a.sqm(&target).addm(&offset, &target);
        if a == b {
            return (None, i);
        }

        // FIXME: optimize abs_diff for montgomery form if we are going to use the abs_diff in the std lib
        let diff = if b > a { &b - &a } else { &a - &b }; // abs_diff
        z = z.mulm(&diff, &target);
        if z.is_zero() {
            // the factor is missed by a combined GCD, do backtracing
            if backtrace {
                // ultimately failed
                return (None, i);
            } else {
                backtrace = true;
                a = std::mem::replace(&mut s, T::one()); // s is discarded
                z = T::one() % target; // clear gcd
                continue;
            }
        }

        // here we check gcd every 2^k steps or 128 steps
        // larger batch size leads to large overhead when backtracing.
        // reference: https://www.cnblogs.com/812-xiao-wen/p/10544546.html
        if i == j || i & 127 == 0 || backtrace {
            let d = z.gcd(target);
            if !d.is_one() && &d != target {
                return (Some(d), i);
            }

            // save state
            s = a.clone();
        }

        // when tortoise catches up with hare, let hare jump to the next stop
        if i == j {
            b = a.clone();
            j <<= 1;
        }
    }

    (None, i)
}

/// This function implements Shanks's square forms factorization (SQUFOF).
///
/// The input is usually multiplied by a multiplier, and the multiplied integer should be put in
/// the `mul_target` argument. The multiplier can be choosen from SQUFOF_MULTIPLIERS, or other square-free odd numbers.
/// The returned values are the factor and the count of passed iterations.
/// 
/// The max iteration can be choosed as 2*n^(1/4), based on Theorem 4.22 from [1].
///
/// Reference: Gower, J., & Wagstaff Jr, S. (2008). Square form factorization.
/// In [1] [Mathematics of Computation](https://homes.cerias.purdue.edu/~ssw/gowerthesis804/wthe.pdf)
/// or [2] [his thesis](https://homes.cerias.purdue.edu/~ssw/gowerthesis804/wthe.pdf)
/// The code is from [3] [Rosetta code](https://rosettacode.org/wiki/Square_form_factorization)
pub fn squfof<T: Integer + NumRef + Clone + ExactRoots + std::fmt::Debug>(target: &T, mul_target: T, max_iter: usize) -> (Option<T>, usize)
where
    for<'r> &'r T: RefNum<T>,
{
    assert!(&mul_target.is_multiple_of(&target), "mul_target should be multiples of target");
    let rd = Roots::sqrt(&mul_target); // root of k*N

    /// Reduction operator for binary quadratic forms. It's equivalent to
    /// the one used in the `num-irrational` crate, in a little different form.
    /// 
    /// This function reduces (a, b, c) = (qm1, p, q), updates qm1 and q, returns new p.
    #[inline]
    fn rho<T: Integer + Clone + NumRef> (rd: &T, p: &T, q: &mut T, qm1: &mut T) -> T where
        for<'r> &'r T: RefNum<T>, {
        let b = (rd + p).div_floor(&*q);
        let new_p = &b * &*q - p;
        let new_q = if p > &new_p {
            &*qm1 + b * (p - &new_p)
        } else {
            &*qm1 - b * (&new_p - p)
        };

        *qm1 = std::mem::replace(q, new_q);
        new_p
    }

    // forward loop, search principal cycle
    let (mut p, mut q, mut qm1) = (rd.clone(), &mul_target - &rd * &rd, T::one());
    if q.is_zero() {
        // shortcut for perfect square
        return (Some(rd), 0);
    }

    for i in 1..max_iter {
        p = rho(&rd, &p, &mut q, &mut qm1);
        if i.is_odd() {
            if let Some(rq) = q.sqrt_exact() {
                let b = (&rd - &p) / &rq;
                let mut u = b * &rq + &p;
                let (mut v, mut vm1) = ((&mul_target - &u * &u) / &rq, rq);

                // backward loop, search ambiguous cycle
                loop {
                    let new_u = rho(&rd, &u, &mut v, &mut vm1);
                    if new_u == u {
                        break;
                    } else {
                        u = new_u
                    }
                }

                let d = target.gcd(&u);
                if d > T::one() && &d < target {
                   return (Some(d), i)
                }
            }
        }
    };
    (None, max_iter)
}

/// Good squfof multipliers sorted by efficiency descendingly, from Dana Jacobsen.
/// 
/// Note: square-free odd numbers are suitable as SQUFOF multipliers
pub const SQUFOF_MULTIPLIERS: [u16; 38] = [
    3 * 5 * 7 * 11, 3 * 5 * 7, 3 * 5 * 7 * 11 * 13, 3 * 5 * 7 * 13, 3 * 5 * 7 * 11 * 17, 3 * 5 * 11,
    3 * 5 * 7 * 17, 3 * 5, 3 * 5 * 7 * 11 * 19, 3 * 5 * 11 * 13, 3 * 5 * 7 * 19, 3 * 5 * 7 * 13 * 17,
    3 * 5 * 13, 3 * 7 * 11, 3 * 7, 5 * 7 * 11, 3 * 7 * 13, 5 * 7,
    3 * 5 * 17, 5 * 7 * 13, 3 * 5 * 19, 3 * 11, 3 * 7 * 17, 3,
    3 * 11 * 13, 5 * 11, 3 * 7 * 19, 3 * 13, 5, 5 * 11 * 13,
    5 * 7 * 19, 5 * 13, 7 * 11, 7, 3 * 17, 7 * 13,
    11, 1
];

/// William Hart's one line factorization algorithm for 64 bit integers.
///
/// The number to be factored could be multiplied by a smooth number (coprime to the target)
/// to speed up, put the multiplied number in the `mul_target` argument. A good multiplier given by Hart is 480.
/// `iters` determine the range for iterating the inner multiplier itself. The returned values are the factor
/// and the count of passed iterations.
/// 
/// 
/// The one line factorization algorithm is especially good at factoring semiprimes with form pq,
/// where p = next_prime(c^a+d1), p = next_prime(c^b+d2), a and b are close, and c, d1, d2 are small integers.
///
/// Reference: Hart, W. B. (2012). A one line factoring algorithm. Journal of the Australian Mathematical Society, 92(1), 61-69. doi:10.1017/S1446788712000146
// TODO: add multipliers preset for one_line method?
pub fn one_line<T: Integer + NumRef + FromPrimitive + ExactRoots + CheckedAdd>(target: &T, mul_target: T, max_iter: usize) -> (Option<T>, usize)
where
    for<'r> &'r T: RefNum<T>, {
    assert!(&mul_target.is_multiple_of(&target), "mul_target should be multiples of target");

    let mut ikn = mul_target.clone();
    for i in 1..max_iter {
        let s = ikn.sqrt() + T::one(); // assuming target is not perfect square
        let m = &s * &s - &ikn;
        if let Some(t) = m.sqrt_exact() {
            let g = target.gcd(&(s - t));
            if !g.is_one() && &g != target {
                return (Some(g), i);
            }
        }

        // prevent overflow
        ikn = if let Some(n) = (&ikn).checked_add(&mul_target) {
            n
        } else {
            return (None, i)
        }
    }
    return (None, max_iter);
}

// TODO: ECM, (self initialize) Quadratic sieve, Lehman's Fermat(https://en.wikipedia.org/wiki/Fermat%27s_factorization_method, n_factor_lehman)
// REF: https://pypi.org/project/primefac/
//      http://flintlib.org/doc/ulong_extras.html#factorisation
//      https://github.com/zademn/facto-rs/
//      https://github.com/elmomoilanen/prime-factorization
//      https://cseweb.ucsd.edu/~ethome/teaching/2022-cse-291-14/
fn pollard_pp1() {}
fn williams_pp1() {}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::mint::Mint;
    use num_modular::MontgomeryInt;
    use rand::random;

    #[test]
    fn pollard_rho_test() {
        assert_eq!(pollard_rho(&8051u16, 2, 1, 100).0, Some(97));
        assert!(matches!(pollard_rho(&8051u16, random(), 1, 100).0, Some(i) if i == 97 || i == 83));
        assert_eq!(pollard_rho(&455459u32, 2, 1, 100).0, Some(743));

        // Mint test
        for _ in 0..10 {
            let target = random::<u16>() | 1;
            let start = random::<u16>() % target;
            let offset = random::<u16>() % target;

            let expect = pollard_rho(&target, start, offset, 65536);
            let mint_result = pollard_rho(
                &Mint::from(target),
                MontgomeryInt::new(start, target).into(),
                MontgomeryInt::new(offset, target).into(),
                65536
            );
            assert_eq!(
                expect.0,
                mint_result.0.map(|v| v.value())
            );
        }
    }

    #[test]
    fn squfof_test() {
        // case from wikipedia
        assert_eq!(squfof(&11111u32, 11111u32, 100).0, Some(41));

        // cases from https://rosettacode.org/wiki/Square_form_factorization
        let cases: Vec<u64> = vec![
            2501,
            12851,
            13289,
            75301,
            120787,
            967009,
            997417,
            7091569,

            5214317,
            20834839,
            23515517,
            33409583,
            44524219,

            13290059,
            223553581,
            2027651281,
            11111111111,
            100895598169,
            60012462237239,
            287129523414791,
            9007199254740931,
            11111111111111111,
            314159265358979323,
            384307168202281507,
            419244183493398773,
            658812288346769681,
            922337203685477563,
            1000000000000000127,
            1152921505680588799,
            1537228672809128917,

            // this case should success at step 276, from https://rosettacode.org/wiki/Talk:Square_form_factorization
            4558849,
        ];
        for n in cases {
            let d = squfof(&n, n, 40000).0
            .or(squfof(&n, 3*n, 40000).0)
            .or(squfof(&n, 5*n, 40000).0)
            .or(squfof(&n, 7*n, 40000).0)
            .or(squfof(&n, 11*n, 40000).0);
            assert!(matches!(d, Some(_)), "{}", n);
        }
    }

    #[test]
    fn one_line_test() {
        assert_eq!(one_line(&11111u32, 11111u32, 100).0, Some(271));
    }
}