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
//! 回退算法的任意精度十进制类。
//!
//! 这仅在快速路径 (本地浮点数) 和 Eisel-Lemire 算法无法明确确定浮点数时使用。
//!
//! 使用的技术是简单的十进制转换,由 Nigel Tao 和 Ken Thompson 开发。
//! 该算法的详细描述可以在 "通过简单的十进制转换解析 NumberF64" 中找到,可在线获取: <https://nigeltao.github.io/blog/2020/parse-number-f64-simple.html>。
//!
//!
//!
//!

use crate::num::dec2flt::common::{is_8digits, ByteSlice};

#[derive(Clone)]
pub struct Decimal {
    /// 十进制中有效数字的位数。
    pub num_digits: usize,
    /// 有效数字中小数点的偏移量。
    pub decimal_point: i32,
    /// 如果十进制中存储的有效位数被截断。
    pub truncated: bool,
    /// 原始数字的缓冲区,范围为 [0, 9]。
    pub digits: [u8; Self::MAX_DIGITS],
}

impl Default for Decimal {
    fn default() -> Self {
        Self { num_digits: 0, decimal_point: 0, truncated: false, digits: [0; Self::MAX_DIGITS] }
    }
}

impl Decimal {
    /// 明确舍入浮点数所需的最大位数。
    ///
    /// 对于双精度 IEEE 754 浮点数,这需要 767 位数字,因此我们存储最大数字 + 1.
    ///
    /// 如果 `b` 可以被 2 整除,我们就可以从基数 2 准确地表示基数 `b` 中的浮点数。该函数计算精确表示该浮点数所需的确切位数。
    ///
    /// 根据浮点运算手册,对于 IEEE754,emin 为 min 指数,p2 为精度,b 为基数,位数如下:
    ///
    ///
    /// `−emin + p2 + ⌊(emin + 1) log(2, b) − log(1 − 2^(−p2), b)⌋`
    ///
    /// 对于 f32,如下所示:
    ///     emin = -126 p2 = 24
    ///
    /// 对于 f64,如下所示:
    ///     emin = -1022 p2 = 53
    ///
    /// 在 Python 中: `-emin + p2 + math.floor((emin+ 1)*math.log(2, b)-math.log(1-2**(-p2), b))`
    ///
    ///
    ///
    ///
    ///
    ///
    ///
    pub const MAX_DIGITS: usize = 768;
    /// 可以用 64 位整数精确表示的最大位数。
    pub const MAX_DIGITS_WITHOUT_OVERFLOW: usize = 19;
    pub const DECIMAL_POINT_RANGE: i32 = 2047;

    /// 将一个数字追加到缓冲区。
    pub fn try_add_digit(&mut self, digit: u8) {
        if self.num_digits < Self::MAX_DIGITS {
            self.digits[self.num_digits] = digit;
        }
        self.num_digits += 1;
    }

    /// 从缓冲区中修剪尾随零。
    pub fn trim(&mut self) {
        // 以下所有对 `Decimal::trim` 的调用都不能 panic,因为:
        //
        //  1. `parse_decimal` 将 `num_digits` 设置为最大 `Decimal::MAX_DIGITS`。
        //  2. `right_shift` 将 `num_digits` 设置为 `write_index`,它以 `num_digits` 为界。
        //  3. `left_shift` `num_digits` 到最大 `Decimal::MAX_DIGITS`。
        //
        // Trim 仅在 `right_shift` 和 `left_shift` 中调用。
        debug_assert!(self.num_digits <= Self::MAX_DIGITS);
        while self.num_digits != 0 && self.digits[self.num_digits - 1] == 0 {
            self.num_digits -= 1;
        }
    }

    pub fn round(&self) -> u64 {
        if self.num_digits == 0 || self.decimal_point < 0 {
            return 0;
        } else if self.decimal_point > 18 {
            return 0xFFFF_FFFF_FFFF_FFFF_u64;
        }
        let dp = self.decimal_point as usize;
        let mut n = 0_u64;
        for i in 0..dp {
            n *= 10;
            if i < self.num_digits {
                n += self.digits[i] as u64;
            }
        }
        let mut round_up = false;
        if dp < self.num_digits {
            round_up = self.digits[dp] >= 5;
            if self.digits[dp] == 5 && dp + 1 == self.num_digits {
                round_up = self.truncated || ((dp != 0) && (1 & self.digits[dp - 1] != 0))
            }
        }
        if round_up {
            n += 1;
        }
        n
    }

    /// 计算十进制 * 2^shift。
    pub fn left_shift(&mut self, shift: usize) {
        if self.num_digits == 0 {
            return;
        }
        let num_new_digits = number_of_digits_decimal_left_shift(self, shift);
        let mut read_index = self.num_digits;
        let mut write_index = self.num_digits + num_new_digits;
        let mut n = 0_u64;
        while read_index != 0 {
            read_index -= 1;
            write_index -= 1;
            n += (self.digits[read_index] as u64) << shift;
            let quotient = n / 10;
            let remainder = n - (10 * quotient);
            if write_index < Self::MAX_DIGITS {
                self.digits[write_index] = remainder as u8;
            } else if remainder > 0 {
                self.truncated = true;
            }
            n = quotient;
        }
        while n > 0 {
            write_index -= 1;
            let quotient = n / 10;
            let remainder = n - (10 * quotient);
            if write_index < Self::MAX_DIGITS {
                self.digits[write_index] = remainder as u8;
            } else if remainder > 0 {
                self.truncated = true;
            }
            n = quotient;
        }
        self.num_digits += num_new_digits;
        if self.num_digits > Self::MAX_DIGITS {
            self.num_digits = Self::MAX_DIGITS;
        }
        self.decimal_point += num_new_digits as i32;
        self.trim();
    }

    /// 计算十进制 * 2^-shift。
    pub fn right_shift(&mut self, shift: usize) {
        let mut read_index = 0;
        let mut write_index = 0;
        let mut n = 0_u64;
        while (n >> shift) == 0 {
            if read_index < self.num_digits {
                n = (10 * n) + self.digits[read_index] as u64;
                read_index += 1;
            } else if n == 0 {
                return;
            } else {
                while (n >> shift) == 0 {
                    n *= 10;
                    read_index += 1;
                }
                break;
            }
        }
        self.decimal_point -= read_index as i32 - 1;
        if self.decimal_point < -Self::DECIMAL_POINT_RANGE {
            // `self = Self::Default()`,但没有清除 `digits` 的开销。
            self.num_digits = 0;
            self.decimal_point = 0;
            self.truncated = false;
            return;
        }
        let mask = (1_u64 << shift) - 1;
        while read_index < self.num_digits {
            let new_digit = (n >> shift) as u8;
            n = (10 * (n & mask)) + self.digits[read_index] as u64;
            read_index += 1;
            self.digits[write_index] = new_digit;
            write_index += 1;
        }
        while n > 0 {
            let new_digit = (n >> shift) as u8;
            n = 10 * (n & mask);
            if write_index < Self::MAX_DIGITS {
                self.digits[write_index] = new_digit;
                write_index += 1;
            } else if new_digit > 0 {
                self.truncated = true;
            }
        }
        self.num_digits = write_index;
        self.trim();
    }
}

/// 将浮点数的大整数表示解析为小数。
pub fn parse_decimal(mut s: &[u8]) -> Decimal {
    let mut d = Decimal::default();
    let start = s;

    while let Some((&b'0', s_next)) = s.split_first() {
        s = s_next;
    }

    s = s.parse_digits(|digit| d.try_add_digit(digit));

    if let Some((b'.', s_next)) = s.split_first() {
        s = s_next;
        let first = s;
        // 跳过前导零。
        if d.num_digits == 0 {
            while let Some((&b'0', s_next)) = s.split_first() {
                s = s_next;
            }
        }
        while s.len() >= 8 && d.num_digits + 8 < Decimal::MAX_DIGITS {
            let v = s.read_u64();
            if !is_8digits(v) {
                break;
            }
            d.digits[d.num_digits..].write_u64(v - 0x3030_3030_3030_3030);
            d.num_digits += 8;
            s = &s[8..];
        }
        s = s.parse_digits(|digit| d.try_add_digit(digit));
        d.decimal_point = s.len() as i32 - first.len() as i32;
    }
    if d.num_digits != 0 {
        // 如果有任何尾随零,请忽略
        let mut n_trailing_zeros = 0;
        for &c in start[..(start.len() - s.len())].iter().rev() {
            if c == b'0' {
                n_trailing_zeros += 1;
            } else if c != b'.' {
                break;
            }
        }
        d.decimal_point += n_trailing_zeros as i32;
        d.num_digits -= n_trailing_zeros;
        d.decimal_point += d.num_digits as i32;
        if d.num_digits > Decimal::MAX_DIGITS {
            d.truncated = true;
            d.num_digits = Decimal::MAX_DIGITS;
        }
    }
    if let Some((&ch, s_next)) = s.split_first() {
        if ch == b'e' || ch == b'E' {
            s = s_next;
            let mut neg_exp = false;
            if let Some((&ch, s_next)) = s.split_first() {
                neg_exp = ch == b'-';
                if ch == b'-' || ch == b'+' {
                    s = s_next;
                }
            }
            let mut exp_num = 0_i32;

            s.parse_digits(|digit| {
                if exp_num < 0x10000 {
                    exp_num = 10 * exp_num + digit as i32;
                }
            });

            d.decimal_point += if neg_exp { -exp_num } else { exp_num };
        }
    }
    for i in d.num_digits..Decimal::MAX_DIGITS_WITHOUT_OVERFLOW {
        d.digits[i] = 0;
    }
    d
}

fn number_of_digits_decimal_left_shift(d: &Decimal, mut shift: usize) -> usize {
    #[rustfmt::skip]
    const TABLE: [u16; 65] = [
        0x0000, 0x0800, 0x0801, 0x0803, 0x1006, 0x1009, 0x100D, 0x1812, 0x1817, 0x181D, 0x2024,
        0x202B, 0x2033, 0x203C, 0x2846, 0x2850, 0x285B, 0x3067, 0x3073, 0x3080, 0x388E, 0x389C,
        0x38AB, 0x38BB, 0x40CC, 0x40DD, 0x40EF, 0x4902, 0x4915, 0x4929, 0x513E, 0x5153, 0x5169,
        0x5180, 0x5998, 0x59B0, 0x59C9, 0x61E3, 0x61FD, 0x6218, 0x6A34, 0x6A50, 0x6A6D, 0x6A8B,
        0x72AA, 0x72C9, 0x72E9, 0x7B0A, 0x7B2B, 0x7B4D, 0x8370, 0x8393, 0x83B7, 0x83DC, 0x8C02,
        0x8C28, 0x8C4F, 0x9477, 0x949F, 0x94C8, 0x9CF2, 0x051C, 0x051C, 0x051C, 0x051C,
    ];
    #[rustfmt::skip]
    const TABLE_POW5: [u8; 0x051C] = [
        5, 2, 5, 1, 2, 5, 6, 2, 5, 3, 1, 2, 5, 1, 5, 6, 2, 5, 7, 8, 1, 2, 5, 3, 9, 0, 6, 2, 5, 1,
        9, 5, 3, 1, 2, 5, 9, 7, 6, 5, 6, 2, 5, 4, 8, 8, 2, 8, 1, 2, 5, 2, 4, 4, 1, 4, 0, 6, 2, 5,
        1, 2, 2, 0, 7, 0, 3, 1, 2, 5, 6, 1, 0, 3, 5, 1, 5, 6, 2, 5, 3, 0, 5, 1, 7, 5, 7, 8, 1, 2,
        5, 1, 5, 2, 5, 8, 7, 8, 9, 0, 6, 2, 5, 7, 6, 2, 9, 3, 9, 4, 5, 3, 1, 2, 5, 3, 8, 1, 4, 6,
        9, 7, 2, 6, 5, 6, 2, 5, 1, 9, 0, 7, 3, 4, 8, 6, 3, 2, 8, 1, 2, 5, 9, 5, 3, 6, 7, 4, 3, 1,
        6, 4, 0, 6, 2, 5, 4, 7, 6, 8, 3, 7, 1, 5, 8, 2, 0, 3, 1, 2, 5, 2, 3, 8, 4, 1, 8, 5, 7, 9,
        1, 0, 1, 5, 6, 2, 5, 1, 1, 9, 2, 0, 9, 2, 8, 9, 5, 5, 0, 7, 8, 1, 2, 5, 5, 9, 6, 0, 4, 6,
        4, 4, 7, 7, 5, 3, 9, 0, 6, 2, 5, 2, 9, 8, 0, 2, 3, 2, 2, 3, 8, 7, 6, 9, 5, 3, 1, 2, 5, 1,
        4, 9, 0, 1, 1, 6, 1, 1, 9, 3, 8, 4, 7, 6, 5, 6, 2, 5, 7, 4, 5, 0, 5, 8, 0, 5, 9, 6, 9, 2,
        3, 8, 2, 8, 1, 2, 5, 3, 7, 2, 5, 2, 9, 0, 2, 9, 8, 4, 6, 1, 9, 1, 4, 0, 6, 2, 5, 1, 8, 6,
        2, 6, 4, 5, 1, 4, 9, 2, 3, 0, 9, 5, 7, 0, 3, 1, 2, 5, 9, 3, 1, 3, 2, 2, 5, 7, 4, 6, 1, 5,
        4, 7, 8, 5, 1, 5, 6, 2, 5, 4, 6, 5, 6, 6, 1, 2, 8, 7, 3, 0, 7, 7, 3, 9, 2, 5, 7, 8, 1, 2,
        5, 2, 3, 2, 8, 3, 0, 6, 4, 3, 6, 5, 3, 8, 6, 9, 6, 2, 8, 9, 0, 6, 2, 5, 1, 1, 6, 4, 1, 5,
        3, 2, 1, 8, 2, 6, 9, 3, 4, 8, 1, 4, 4, 5, 3, 1, 2, 5, 5, 8, 2, 0, 7, 6, 6, 0, 9, 1, 3, 4,
        6, 7, 4, 0, 7, 2, 2, 6, 5, 6, 2, 5, 2, 9, 1, 0, 3, 8, 3, 0, 4, 5, 6, 7, 3, 3, 7, 0, 3, 6,
        1, 3, 2, 8, 1, 2, 5, 1, 4, 5, 5, 1, 9, 1, 5, 2, 2, 8, 3, 6, 6, 8, 5, 1, 8, 0, 6, 6, 4, 0,
        6, 2, 5, 7, 2, 7, 5, 9, 5, 7, 6, 1, 4, 1, 8, 3, 4, 2, 5, 9, 0, 3, 3, 2, 0, 3, 1, 2, 5, 3,
        6, 3, 7, 9, 7, 8, 8, 0, 7, 0, 9, 1, 7, 1, 2, 9, 5, 1, 6, 6, 0, 1, 5, 6, 2, 5, 1, 8, 1, 8,
        9, 8, 9, 4, 0, 3, 5, 4, 5, 8, 5, 6, 4, 7, 5, 8, 3, 0, 0, 7, 8, 1, 2, 5, 9, 0, 9, 4, 9, 4,
        7, 0, 1, 7, 7, 2, 9, 2, 8, 2, 3, 7, 9, 1, 5, 0, 3, 9, 0, 6, 2, 5, 4, 5, 4, 7, 4, 7, 3, 5,
        0, 8, 8, 6, 4, 6, 4, 1, 1, 8, 9, 5, 7, 5, 1, 9, 5, 3, 1, 2, 5, 2, 2, 7, 3, 7, 3, 6, 7, 5,
        4, 4, 3, 2, 3, 2, 0, 5, 9, 4, 7, 8, 7, 5, 9, 7, 6, 5, 6, 2, 5, 1, 1, 3, 6, 8, 6, 8, 3, 7,
        7, 2, 1, 6, 1, 6, 0, 2, 9, 7, 3, 9, 3, 7, 9, 8, 8, 2, 8, 1, 2, 5, 5, 6, 8, 4, 3, 4, 1, 8,
        8, 6, 0, 8, 0, 8, 0, 1, 4, 8, 6, 9, 6, 8, 9, 9, 4, 1, 4, 0, 6, 2, 5, 2, 8, 4, 2, 1, 7, 0,
        9, 4, 3, 0, 4, 0, 4, 0, 0, 7, 4, 3, 4, 8, 4, 4, 9, 7, 0, 7, 0, 3, 1, 2, 5, 1, 4, 2, 1, 0,
        8, 5, 4, 7, 1, 5, 2, 0, 2, 0, 0, 3, 7, 1, 7, 4, 2, 2, 4, 8, 5, 3, 5, 1, 5, 6, 2, 5, 7, 1,
        0, 5, 4, 2, 7, 3, 5, 7, 6, 0, 1, 0, 0, 1, 8, 5, 8, 7, 1, 1, 2, 4, 2, 6, 7, 5, 7, 8, 1, 2,
        5, 3, 5, 5, 2, 7, 1, 3, 6, 7, 8, 8, 0, 0, 5, 0, 0, 9, 2, 9, 3, 5, 5, 6, 2, 1, 3, 3, 7, 8,
        9, 0, 6, 2, 5, 1, 7, 7, 6, 3, 5, 6, 8, 3, 9, 4, 0, 0, 2, 5, 0, 4, 6, 4, 6, 7, 7, 8, 1, 0,
        6, 6, 8, 9, 4, 5, 3, 1, 2, 5, 8, 8, 8, 1, 7, 8, 4, 1, 9, 7, 0, 0, 1, 2, 5, 2, 3, 2, 3, 3,
        8, 9, 0, 5, 3, 3, 4, 4, 7, 2, 6, 5, 6, 2, 5, 4, 4, 4, 0, 8, 9, 2, 0, 9, 8, 5, 0, 0, 6, 2,
        6, 1, 6, 1, 6, 9, 4, 5, 2, 6, 6, 7, 2, 3, 6, 3, 2, 8, 1, 2, 5, 2, 2, 2, 0, 4, 4, 6, 0, 4,
        9, 2, 5, 0, 3, 1, 3, 0, 8, 0, 8, 4, 7, 2, 6, 3, 3, 3, 6, 1, 8, 1, 6, 4, 0, 6, 2, 5, 1, 1,
        1, 0, 2, 2, 3, 0, 2, 4, 6, 2, 5, 1, 5, 6, 5, 4, 0, 4, 2, 3, 6, 3, 1, 6, 6, 8, 0, 9, 0, 8,
        2, 0, 3, 1, 2, 5, 5, 5, 5, 1, 1, 1, 5, 1, 2, 3, 1, 2, 5, 7, 8, 2, 7, 0, 2, 1, 1, 8, 1, 5,
        8, 3, 4, 0, 4, 5, 4, 1, 0, 1, 5, 6, 2, 5, 2, 7, 7, 5, 5, 5, 7, 5, 6, 1, 5, 6, 2, 8, 9, 1,
        3, 5, 1, 0, 5, 9, 0, 7, 9, 1, 7, 0, 2, 2, 7, 0, 5, 0, 7, 8, 1, 2, 5, 1, 3, 8, 7, 7, 7, 8,
        7, 8, 0, 7, 8, 1, 4, 4, 5, 6, 7, 5, 5, 2, 9, 5, 3, 9, 5, 8, 5, 1, 1, 3, 5, 2, 5, 3, 9, 0,
        6, 2, 5, 6, 9, 3, 8, 8, 9, 3, 9, 0, 3, 9, 0, 7, 2, 2, 8, 3, 7, 7, 6, 4, 7, 6, 9, 7, 9, 2,
        5, 5, 6, 7, 6, 2, 6, 9, 5, 3, 1, 2, 5, 3, 4, 6, 9, 4, 4, 6, 9, 5, 1, 9, 5, 3, 6, 1, 4, 1,
        8, 8, 8, 2, 3, 8, 4, 8, 9, 6, 2, 7, 8, 3, 8, 1, 3, 4, 7, 6, 5, 6, 2, 5, 1, 7, 3, 4, 7, 2,
        3, 4, 7, 5, 9, 7, 6, 8, 0, 7, 0, 9, 4, 4, 1, 1, 9, 2, 4, 4, 8, 1, 3, 9, 1, 9, 0, 6, 7, 3,
        8, 2, 8, 1, 2, 5, 8, 6, 7, 3, 6, 1, 7, 3, 7, 9, 8, 8, 4, 0, 3, 5, 4, 7, 2, 0, 5, 9, 6, 2,
        2, 4, 0, 6, 9, 5, 9, 5, 3, 3, 6, 9, 1, 4, 0, 6, 2, 5,
    ];

    shift &= 63;
    let x_a = TABLE[shift];
    let x_b = TABLE[shift + 1];
    let num_new_digits = (x_a >> 11) as _;
    let pow5_a = (0x7FF & x_a) as usize;
    let pow5_b = (0x7FF & x_b) as usize;
    let pow5 = &TABLE_POW5[pow5_a..];
    for (i, &p5) in pow5.iter().enumerate().take(pow5_b - pow5_a) {
        if i >= d.num_digits {
            return num_new_digits - 1;
        } else if d.digits[i] == p5 {
            continue;
        } else if d.digits[i] < p5 {
            return num_new_digits - 1;
        } else {
            return num_new_digits;
        }
    }
    num_new_digits
}