d3409d5e |
1 | /* -*-c-*- |
2 | * |
a951033d |
3 | * $Id: mptext.c,v 1.10 2001/06/16 13:22:39 mdw Exp $ |
d3409d5e |
4 | * |
5 | * Textual representation of multiprecision numbers |
6 | * |
7 | * (c) 1999 Straylight/Edgeware |
8 | */ |
9 | |
10 | /*----- Licensing notice --------------------------------------------------* |
11 | * |
12 | * This file is part of Catacomb. |
13 | * |
14 | * Catacomb is free software; you can redistribute it and/or modify |
15 | * it under the terms of the GNU Library General Public License as |
16 | * published by the Free Software Foundation; either version 2 of the |
17 | * License, or (at your option) any later version. |
18 | * |
19 | * Catacomb is distributed in the hope that it will be useful, |
20 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
21 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
22 | * GNU Library General Public License for more details. |
23 | * |
24 | * You should have received a copy of the GNU Library General Public |
25 | * License along with Catacomb; if not, write to the Free |
26 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, |
27 | * MA 02111-1307, USA. |
28 | */ |
29 | |
30 | /*----- Revision history --------------------------------------------------* |
31 | * |
32 | * $Log: mptext.c,v $ |
a951033d |
33 | * Revision 1.10 2001/06/16 13:22:39 mdw |
34 | * Added fast-track code for binary output bases, and tests. |
35 | * |
3bc9cb53 |
36 | * Revision 1.9 2001/02/03 16:05:17 mdw |
37 | * Make flags be unsigned. Improve the write algorithm: recurse until the |
38 | * parts are one word long and use single-precision arithmetic from there. |
39 | * Fix off-by-one bug when breaking the number apart. |
40 | * |
9d3838a0 |
41 | * Revision 1.8 2000/12/06 20:32:42 mdw |
42 | * Reduce binary bytes (to allow marker bits to be ignored). Fix error |
43 | * message string a bit. Allow leading `+' signs. |
44 | * |
7d45ed6c |
45 | * Revision 1.7 2000/07/15 10:01:08 mdw |
46 | * Bug fix in binary input. |
47 | * |
dd9199f0 |
48 | * Revision 1.6 2000/06/25 12:58:23 mdw |
49 | * Fix the derivation of `depth' commentary. |
50 | * |
2b26f2d7 |
51 | * Revision 1.5 2000/06/17 11:46:19 mdw |
52 | * New and much faster stack-based algorithm for reading integers. Support |
53 | * reading and writing binary integers in bases between 2 and 256. |
54 | * |
e360a4f2 |
55 | * Revision 1.4 1999/12/22 15:56:56 mdw |
56 | * Use clever recursive algorithm for writing numbers out. |
57 | * |
9c3df6c0 |
58 | * Revision 1.3 1999/12/10 23:23:26 mdw |
59 | * Allocate slightly less memory. |
60 | * |
90b6f0be |
61 | * Revision 1.2 1999/11/20 22:24:15 mdw |
62 | * Use function versions of MPX_UMULN and MPX_UADDN. |
63 | * |
d3409d5e |
64 | * Revision 1.1 1999/11/17 18:02:16 mdw |
65 | * New multiprecision integer arithmetic suite. |
66 | * |
67 | */ |
68 | |
69 | /*----- Header files ------------------------------------------------------*/ |
70 | |
71 | #include <ctype.h> |
2b26f2d7 |
72 | #include <limits.h> |
d3409d5e |
73 | #include <stdio.h> |
74 | |
d3409d5e |
75 | #include "mp.h" |
76 | #include "mptext.h" |
e360a4f2 |
77 | #include "paranoia.h" |
d3409d5e |
78 | |
2b26f2d7 |
79 | /*----- Magical numbers ---------------------------------------------------*/ |
80 | |
81 | /* --- Maximum recursion depth --- * |
82 | * |
83 | * This is the number of bits in a @size_t@ object. Why? |
84 | * |
dd9199f0 |
85 | * To see this, let %$b = \mathit{MPW\_MAX} + 1$% and let %$Z$% be the |
86 | * largest @size_t@ value. Then the largest possible @mp@ is %$M - 1$% where |
87 | * %$M = b^Z$%. Let %$r$% be a radix to read or write. Since the recursion |
88 | * squares the radix at each step, the highest number reached by the |
89 | * recursion is %$d$%, where: |
2b26f2d7 |
90 | * |
dd9199f0 |
91 | * %$r^{2^d} = b^Z$%. |
2b26f2d7 |
92 | * |
93 | * Solving gives that %$d = \lg \log_r b^Z$%. If %$r = 2$%, this is maximum, |
94 | * so choosing %$d = \lg \lg b^Z = \lg (Z \lg b) = \lg Z + \lg \lg b$%. |
95 | * |
96 | * Expressing %$\lg Z$% as @CHAR_BIT * sizeof(size_t)@ yields an |
97 | * overestimate, since a @size_t@ representation may contain `holes'. |
98 | * Choosing to represent %$\lg \lg b$% by 10 is almost certainly sufficient |
99 | * for `some time to come'. |
100 | */ |
101 | |
102 | #define DEPTH (CHAR_BIT * sizeof(size_t) + 10) |
103 | |
d3409d5e |
104 | /*----- Main code ---------------------------------------------------------*/ |
105 | |
106 | /* --- @mp_read@ --- * |
107 | * |
108 | * Arguments: @mp *m@ = destination multiprecision number |
109 | * @int radix@ = base to assume for data (or zero to guess) |
110 | * @const mptext_ops *ops@ = pointer to operations block |
111 | * @void *p@ = data for the operations block |
112 | * |
113 | * Returns: The integer read, or zero if it didn't work. |
114 | * |
115 | * Use: Reads an integer from some source. If the @radix@ is |
116 | * specified, the number is assumed to be given in that radix, |
117 | * with the letters `a' (either upper- or lower-case) upwards |
118 | * standing for digits greater than 9. Otherwise, base 10 is |
119 | * assumed unless the number starts with `0' (octal), `0x' (hex) |
120 | * or `nnn_' (base `nnn'). An arbitrary amount of whitespace |
121 | * before the number is ignored. |
122 | */ |
123 | |
2b26f2d7 |
124 | /* --- About the algorithm --- * |
125 | * |
126 | * The algorithm here is rather aggressive. I maintain an array of |
127 | * successive squarings of the radix, and a stack of partial results, each |
128 | * with a counter attached indicating which radix square to multiply by. |
129 | * Once the item at the top of the stack reaches the same counter level as |
130 | * the next item down, they are combined together and the result is given a |
131 | * counter level one higher than either of the results. |
132 | * |
133 | * Gluing the results together at the end is slightly tricky. Pay attention |
134 | * to the code. |
135 | * |
136 | * This is more complicated because of the need to handle the slightly |
137 | * bizarre syntax. |
138 | */ |
139 | |
d3409d5e |
140 | mp *mp_read(mp *m, int radix, const mptext_ops *ops, void *p) |
141 | { |
2b26f2d7 |
142 | int ch; /* Current char being considered */ |
143 | unsigned f = 0; /* Flags about the current number */ |
144 | int r; /* Radix to switch over to */ |
145 | mpw rd; /* Radix as an @mp@ digit */ |
146 | mp rr; /* The @mp@ for the radix */ |
147 | unsigned nf = m ? m->f & MP_BURN : 0; /* New @mp@ flags */ |
148 | |
149 | /* --- Stacks --- */ |
150 | |
151 | mp *pow[DEPTH]; /* List of powers */ |
152 | unsigned pows; /* Next index to fill */ |
153 | struct { unsigned i; mp *m; } s[DEPTH]; /* Main stack */ |
154 | unsigned sp; /* Current stack pointer */ |
155 | |
156 | /* --- Flags --- */ |
d3409d5e |
157 | |
3bc9cb53 |
158 | #define f_neg 1u |
159 | #define f_ok 2u |
a951033d |
160 | #define f_start 4u |
d3409d5e |
161 | |
2b26f2d7 |
162 | /* --- Initialize the stacks --- */ |
163 | |
164 | mp_build(&rr, &rd, &rd + 1); |
165 | pow[0] = &rr; |
166 | pows = 1; |
167 | |
168 | sp = 0; |
169 | |
d3409d5e |
170 | /* --- Initialize the destination number --- */ |
171 | |
2b26f2d7 |
172 | if (m) |
173 | MP_DROP(m); |
d3409d5e |
174 | |
175 | /* --- Read an initial character --- */ |
176 | |
177 | ch = ops->get(p); |
178 | while (isspace(ch)) |
179 | ch = ops->get(p); |
180 | |
181 | /* --- Handle an initial sign --- */ |
182 | |
9d3838a0 |
183 | if (radix >= 0 && (ch == '-' || ch == '+')) { |
184 | if (ch == '-') |
185 | f |= f_neg; |
186 | do ch = ops->get(p); while isspace(ch); |
d3409d5e |
187 | } |
188 | |
189 | /* --- If the radix is zero, look for leading zeros --- */ |
190 | |
2b26f2d7 |
191 | if (radix > 0) { |
192 | assert(((void)"ascii radix must be <= 36", radix <= 36)); |
193 | rd = radix; |
194 | r = -1; |
195 | } else if (radix < 0) { |
196 | rd = -radix; |
9d3838a0 |
197 | assert(((void)"binary radix must fit in a byte", rd < UCHAR_MAX)); |
d3409d5e |
198 | r = -1; |
2b26f2d7 |
199 | } else if (ch != '0') { |
200 | rd = 10; |
d3409d5e |
201 | r = 0; |
202 | } else { |
203 | ch = ops->get(p); |
204 | if (ch == 'x') { |
205 | ch = ops->get(p); |
2b26f2d7 |
206 | rd = 16; |
d3409d5e |
207 | } else { |
2b26f2d7 |
208 | rd = 8; |
d3409d5e |
209 | f |= f_ok; |
210 | } |
211 | r = -1; |
212 | } |
213 | |
a951033d |
214 | /* --- Use fast algorithm for binary radix --- * |
215 | * |
216 | * This is the restart point after having parsed a radix number from the |
217 | * input. We check whether the radix is binary, and if so use a fast |
218 | * algorithm which just stacks the bits up in the right order. |
219 | */ |
220 | |
221 | restart: |
222 | switch (rd) { |
223 | unsigned bit; |
224 | |
225 | case 2: bit = 1; goto bin; |
226 | case 4: bit = 2; goto bin; |
227 | case 8: bit = 3; goto bin; |
228 | case 16: bit = 4; goto bin; |
229 | case 32: bit = 5; goto bin; |
230 | case 64: bit = 6; goto bin; |
231 | case 128: bit = 7; goto bin; |
232 | default: |
233 | break; |
234 | |
235 | /* --- The fast binary algorithm --- * |
236 | * |
237 | * We stack bits up starting at the top end of a word. When one word is |
238 | * full, we write it to the integer, and start another with the left-over |
239 | * bits. When the array in the integer is full, we resize using low-level |
240 | * calls and copy the current data to the top end. Finally, we do a single |
241 | * bit-shift when we know where the end of the number is. |
242 | */ |
243 | |
244 | bin: { |
245 | mpw a = 0; |
246 | unsigned b = MPW_BITS; |
247 | size_t len, n; |
248 | mpw *v; |
249 | |
250 | m = mp_dest(MP_NEW, 1, nf); |
251 | len = n = m->sz; |
252 | n = len; |
253 | v = m->v + n; |
254 | for (;; ch = ops->get(p)) { |
255 | unsigned x; |
256 | |
257 | if (ch < 0) |
258 | break; |
259 | |
260 | /* --- Check that the character is a digit and in range --- */ |
261 | |
262 | if (radix < 0) |
263 | x = ch % rd; |
264 | else { |
265 | if (!isalnum(ch)) |
266 | break; |
267 | if (ch >= '0' && ch <= '9') |
268 | x = ch - '0'; |
269 | else { |
270 | ch = tolower(ch); |
271 | if (ch >= 'a' && ch <= 'z') /* ASCII dependent! */ |
272 | x = ch - 'a' + 10; |
273 | else |
274 | break; |
275 | } |
276 | } |
277 | if (x >= rd) |
278 | break; |
279 | |
280 | /* --- Feed the digit into the accumulator --- */ |
281 | |
282 | f |= f_ok; |
283 | if (!x && !(f & f_start)) |
284 | continue; |
285 | f |= f_start; |
286 | if (b > bit) { |
287 | b -= bit; |
288 | a |= MPW(x) << b; |
289 | } else { |
290 | a |= MPW(x) >> (bit - b); |
291 | b += MPW_BITS - bit; |
292 | *--v = MPW(a); |
293 | n--; |
294 | if (!n) { |
295 | n = len; |
296 | len <<= 1; |
297 | v = mpalloc(m->a, len); |
298 | memcpy(v + n, m->v, MPWS(n)); |
299 | mpfree(m->a, m->v); |
300 | m->v = v; |
301 | v = m->v + n; |
302 | } |
303 | a = (b < MPW_BITS) ? MPW(x) << b : 0; |
304 | } |
305 | } |
306 | |
307 | /* --- Finish up --- */ |
308 | |
309 | if (!(f & f_ok)) { |
310 | mp_drop(m); |
311 | m = 0; |
312 | } else { |
313 | *--v = MPW(a); |
314 | n--; |
315 | m->sz = len; |
316 | m->vl = m->v + len; |
317 | m->f &= ~MP_UNDEF; |
318 | m = mp_lsr(m, m, (unsigned long)n * MPW_BITS + b); |
319 | } |
320 | goto done; |
321 | }} |
322 | |
d3409d5e |
323 | /* --- Time to start --- */ |
324 | |
325 | for (;; ch = ops->get(p)) { |
a951033d |
326 | unsigned x; |
d3409d5e |
327 | |
7d45ed6c |
328 | if (ch < 0) |
329 | break; |
330 | |
d3409d5e |
331 | /* --- An underscore indicates a numbered base --- */ |
332 | |
333 | if (ch == '_' && r > 0 && r <= 36) { |
2b26f2d7 |
334 | unsigned i; |
335 | |
336 | /* --- Clear out the stacks --- */ |
337 | |
338 | for (i = 1; i < pows; i++) |
339 | MP_DROP(pow[i]); |
340 | pows = 1; |
341 | for (i = 0; i < sp; i++) |
342 | MP_DROP(s[i].m); |
343 | sp = 0; |
344 | |
345 | /* --- Restart the search --- */ |
346 | |
347 | rd = r; |
d3409d5e |
348 | r = -1; |
349 | f &= ~f_ok; |
a951033d |
350 | ch = ops->get(p); |
351 | goto restart; |
d3409d5e |
352 | } |
353 | |
354 | /* --- Check that the character is a digit and in range --- */ |
355 | |
2b26f2d7 |
356 | if (radix < 0) |
9d3838a0 |
357 | x = ch % rd; |
d3409d5e |
358 | else { |
2b26f2d7 |
359 | if (!isalnum(ch)) |
d3409d5e |
360 | break; |
2b26f2d7 |
361 | if (ch >= '0' && ch <= '9') |
362 | x = ch - '0'; |
363 | else { |
364 | ch = tolower(ch); |
365 | if (ch >= 'a' && ch <= 'z') /* ASCII dependent! */ |
366 | x = ch - 'a' + 10; |
367 | else |
368 | break; |
369 | } |
d3409d5e |
370 | } |
371 | |
372 | /* --- Sort out what to do with the character --- */ |
373 | |
374 | if (x >= 10 && r >= 0) |
375 | r = -1; |
2b26f2d7 |
376 | if (x >= rd) |
d3409d5e |
377 | break; |
378 | |
379 | if (r >= 0) |
380 | r = r * 10 + x; |
381 | |
382 | /* --- Stick the character on the end of my integer --- */ |
383 | |
2b26f2d7 |
384 | assert(((void)"Number is too unimaginably huge", sp < DEPTH)); |
385 | s[sp].m = m = mp_new(1, nf); |
386 | m->v[0] = x; |
387 | s[sp].i = 0; |
388 | |
389 | /* --- Now grind through the stack --- */ |
390 | |
391 | while (sp > 0 && s[sp - 1].i == s[sp].i) { |
392 | |
393 | /* --- Combine the top two items --- */ |
394 | |
395 | sp--; |
396 | m = s[sp].m; |
397 | m = mp_mul(m, m, pow[s[sp].i]); |
398 | m = mp_add(m, m, s[sp + 1].m); |
399 | s[sp].m = m; |
400 | MP_DROP(s[sp + 1].m); |
401 | s[sp].i++; |
402 | |
403 | /* --- Make a new radix power if necessary --- */ |
404 | |
405 | if (s[sp].i >= pows) { |
406 | assert(((void)"Number is too unimaginably huge", pows < DEPTH)); |
407 | pow[pows] = mp_sqr(MP_NEW, pow[pows - 1]); |
408 | pows++; |
409 | } |
410 | } |
d3409d5e |
411 | f |= f_ok; |
2b26f2d7 |
412 | sp++; |
d3409d5e |
413 | } |
414 | |
415 | ops->unget(ch, p); |
416 | |
2b26f2d7 |
417 | /* --- If we're done, compute the rest of the number --- */ |
418 | |
419 | if (f & f_ok) { |
420 | if (!sp) |
421 | return (MP_ZERO); |
422 | else { |
423 | mp *z = MP_ONE; |
424 | sp--; |
425 | |
426 | while (sp > 0) { |
427 | |
428 | /* --- Combine the top two items --- */ |
429 | |
430 | sp--; |
431 | m = s[sp].m; |
432 | z = mp_mul(z, z, pow[s[sp + 1].i]); |
433 | m = mp_mul(m, m, z); |
434 | m = mp_add(m, m, s[sp + 1].m); |
435 | s[sp].m = m; |
436 | MP_DROP(s[sp + 1].m); |
437 | |
438 | /* --- Make a new radix power if necessary --- */ |
439 | |
440 | if (s[sp].i >= pows) { |
441 | assert(((void)"Number is too unimaginably huge", pows < DEPTH)); |
442 | pow[pows] = mp_sqr(MP_NEW, pow[pows - 1]); |
443 | pows++; |
444 | } |
445 | } |
446 | MP_DROP(z); |
447 | m = s[0].m; |
448 | } |
449 | } else { |
450 | unsigned i; |
451 | for (i = 0; i < sp; i++) |
452 | MP_DROP(s[i].m); |
453 | } |
454 | |
455 | /* --- Clear the radix power list --- */ |
456 | |
457 | { |
458 | unsigned i; |
459 | for (i = 1; i < pows; i++) |
460 | MP_DROP(pow[i]); |
461 | } |
462 | |
d3409d5e |
463 | /* --- Bail out if the number was bad --- */ |
464 | |
a951033d |
465 | done: |
2b26f2d7 |
466 | if (!(f & f_ok)) |
d3409d5e |
467 | return (0); |
d3409d5e |
468 | |
469 | /* --- Set the sign and return --- */ |
470 | |
d3409d5e |
471 | if (f & f_neg) |
472 | m->f |= MP_NEG; |
473 | return (m); |
3bc9cb53 |
474 | |
a951033d |
475 | #undef f_start |
3bc9cb53 |
476 | #undef f_neg |
477 | #undef f_ok |
d3409d5e |
478 | } |
479 | |
480 | /* --- @mp_write@ --- * |
481 | * |
482 | * Arguments: @mp *m@ = pointer to a multi-precision integer |
483 | * @int radix@ = radix to use when writing the number out |
484 | * @const mptext_ops *ops@ = pointer to an operations block |
485 | * @void *p@ = data for the operations block |
486 | * |
487 | * Returns: Zero if it worked, nonzero otherwise. |
488 | * |
489 | * Use: Writes a large integer in textual form. |
490 | */ |
491 | |
e360a4f2 |
492 | /* --- Simple case --- * |
493 | * |
3bc9cb53 |
494 | * Use a fixed-sized buffer and single-precision arithmetic to pick off |
495 | * low-order digits. Put each digit in a buffer, working backwards from the |
496 | * end. If the buffer becomes full, recurse to get another one. Ensure that |
497 | * there are at least @z@ digits by writing leading zeroes if there aren't |
498 | * enough real digits. |
e360a4f2 |
499 | */ |
500 | |
3bc9cb53 |
501 | static int simple(mpw n, int radix, unsigned z, |
e360a4f2 |
502 | const mptext_ops *ops, void *p) |
503 | { |
504 | int rc = 0; |
505 | char buf[64]; |
506 | unsigned i = sizeof(buf); |
2b26f2d7 |
507 | int rd = radix > 0 ? radix : -radix; |
e360a4f2 |
508 | |
509 | do { |
510 | int ch; |
511 | mpw x; |
512 | |
3bc9cb53 |
513 | x = n % rd; |
514 | n /= rd; |
2b26f2d7 |
515 | if (radix < 0) |
516 | ch = x; |
3bc9cb53 |
517 | else if (x < 10) |
518 | ch = '0' + x; |
519 | else |
520 | ch = 'a' + x - 10; |
e360a4f2 |
521 | buf[--i] = ch; |
522 | if (z) |
523 | z--; |
3bc9cb53 |
524 | } while (i && n); |
e360a4f2 |
525 | |
3bc9cb53 |
526 | if (n) |
527 | rc = simple(n, radix, z, ops, p); |
e360a4f2 |
528 | else { |
a951033d |
529 | char zbuf[32]; |
530 | memset(zbuf, (radix < 0) ? 0 : '0', sizeof(zbuf)); |
531 | while (!rc && z >= sizeof(zbuf)) { |
532 | rc = ops->put(zbuf, sizeof(zbuf), p); |
533 | z -= sizeof(zbuf); |
e360a4f2 |
534 | } |
535 | if (!rc && z) |
a951033d |
536 | rc = ops->put(zbuf, z, p); |
e360a4f2 |
537 | } |
538 | if (!rc) |
3bc9cb53 |
539 | rc = ops->put(buf + i, sizeof(buf) - i, p); |
540 | BURN(buf); |
e360a4f2 |
541 | return (rc); |
542 | } |
543 | |
544 | /* --- Complicated case --- * |
545 | * |
546 | * If the number is small, fall back to the simple case above. Otherwise |
547 | * divide and take remainder by current large power of the radix, and emit |
548 | * each separately. Don't emit a zero quotient. Be very careful about |
549 | * leading zeroes on the remainder part, because they're deeply significant. |
550 | */ |
551 | |
552 | static int complicated(mp *m, int radix, mp **pr, unsigned i, unsigned z, |
553 | const mptext_ops *ops, void *p) |
554 | { |
555 | int rc = 0; |
556 | mp *q = MP_NEW; |
557 | unsigned d = 1 << i; |
558 | |
3bc9cb53 |
559 | if (MP_LEN(m) < 2) |
560 | return (simple(MP_LEN(m) ? m->v[0] : 0, radix, z, ops, p)); |
e360a4f2 |
561 | |
3bc9cb53 |
562 | assert(i); |
e360a4f2 |
563 | mp_div(&q, &m, m, pr[i]); |
564 | if (!MP_LEN(q)) |
565 | d = z; |
566 | else { |
567 | if (z > d) |
568 | z -= d; |
569 | else |
570 | z = 0; |
571 | rc = complicated(q, radix, pr, i - 1, z, ops, p); |
572 | } |
573 | if (!rc) |
574 | rc = complicated(m, radix, pr, i - 1, d, ops, p); |
575 | mp_drop(q); |
576 | return (rc); |
577 | } |
578 | |
a951033d |
579 | /* --- Binary case --- * |
580 | * |
581 | * Special case for binary output. Goes much faster. |
582 | */ |
583 | |
584 | static int binary(mp *m, int bit, int radix, const mptext_ops *ops, void *p) |
585 | { |
586 | mpw *v; |
587 | mpw a; |
588 | int rc = 0; |
589 | unsigned b; |
590 | unsigned mask; |
591 | unsigned long n; |
592 | unsigned f = 0; |
593 | char buf[8], *q; |
594 | unsigned x; |
595 | int ch; |
596 | |
597 | #define f_out 1u |
598 | |
599 | /* --- Work out where to start --- */ |
600 | |
601 | n = mp_bits(m); |
602 | n += bit - (n % bit); |
603 | b = n % MPW_BITS; |
604 | n /= MPW_BITS; |
605 | |
606 | if (n > MP_LEN(m)) { |
607 | n--; |
608 | b += MPW_BITS; |
609 | } |
610 | |
611 | v = m->v + n; |
612 | a = *v; |
613 | mask = (1 << bit) - 1; |
614 | q = buf; |
615 | |
616 | /* --- Main code --- */ |
617 | |
618 | for (;;) { |
619 | if (b > bit) { |
620 | b -= bit; |
621 | x = a >> b; |
622 | } else { |
623 | x = a << (bit - b); |
624 | b += MPW_BITS - bit; |
625 | if (v == m->v) |
626 | break; |
627 | a = *--v; |
628 | if (b < MPW_BITS) |
629 | x |= a >> b; |
630 | } |
631 | x &= mask; |
632 | if (!x && !(f & f_out)) |
633 | continue; |
634 | |
635 | if (radix < 0) |
636 | ch = x; |
637 | else if (x < 10) |
638 | ch = '0' + x; |
639 | else |
640 | ch = 'a' + x - 10; |
641 | *q++ = ch; |
642 | if (q >= buf + sizeof(buf)) { |
643 | if ((rc = ops->put(buf, sizeof(buf), p)) != 0) |
644 | goto done; |
645 | q = buf; |
646 | } |
647 | f |= f_out; |
648 | } |
649 | |
650 | x &= mask; |
651 | if (radix < 0) |
652 | ch = x; |
653 | else if (x < 10) |
654 | ch = '0' + x; |
655 | else |
656 | ch = 'a' + x - 10; |
657 | *q++ = ch; |
658 | rc = ops->put(buf, q - buf, p); |
659 | |
660 | done: |
661 | mp_drop(m); |
662 | return (rc); |
663 | |
664 | #undef f_out |
665 | } |
666 | |
e360a4f2 |
667 | /* --- Main driver code --- */ |
668 | |
d3409d5e |
669 | int mp_write(mp *m, int radix, const mptext_ops *ops, void *p) |
670 | { |
e360a4f2 |
671 | int rc; |
d3409d5e |
672 | |
673 | /* --- Set various things up --- */ |
674 | |
675 | m = MP_COPY(m); |
e360a4f2 |
676 | MP_SPLIT(m); |
d3409d5e |
677 | |
2b26f2d7 |
678 | /* --- Check the radix for sensibleness --- */ |
679 | |
680 | if (radix > 0) |
681 | assert(((void)"ascii radix must be <= 36", radix <= 36)); |
682 | else if (radix < 0) |
683 | assert(((void)"binary radix must fit in a byte", -radix < UCHAR_MAX)); |
684 | else |
685 | assert(((void)"radix can't be zero in mp_write", 0)); |
686 | |
d3409d5e |
687 | /* --- If the number is negative, sort that out --- */ |
688 | |
689 | if (m->f & MP_NEG) { |
690 | if (ops->put("-", 1, p)) |
691 | return (EOF); |
2b26f2d7 |
692 | m->f &= ~MP_NEG; |
d3409d5e |
693 | } |
694 | |
a951033d |
695 | /* --- Handle binary radix --- */ |
696 | |
697 | switch (radix) { |
698 | case 2: case -2: return (binary(m, 1, radix, ops, p)); |
699 | case 4: case -4: return (binary(m, 2, radix, ops, p)); |
700 | case 8: case -8: return (binary(m, 3, radix, ops, p)); |
701 | case 16: case -16: return (binary(m, 4, radix, ops, p)); |
702 | case 32: case -32: return (binary(m, 5, radix, ops, p)); |
703 | case -64: return (binary(m, 6, radix, ops, p)); |
704 | case -128: return (binary(m, 7, radix, ops, p)); |
705 | } |
706 | |
e360a4f2 |
707 | /* --- If the number is small, do it the easy way --- */ |
708 | |
3bc9cb53 |
709 | if (MP_LEN(m) < 2) |
710 | rc = simple(MP_LEN(m) ? m->v[0] : 0, radix, 0, ops, p); |
e360a4f2 |
711 | |
712 | /* --- Use a clever algorithm --- * |
713 | * |
714 | * Square the radix repeatedly, remembering old results, until I get |
715 | * something more than half the size of the number @m@. Use this to divide |
716 | * the number: the quotient and remainder will be approximately the same |
717 | * size, and I'll have split them on a digit boundary, so I can just emit |
718 | * the quotient and remainder recursively, in order. |
e360a4f2 |
719 | */ |
720 | |
721 | else { |
2b26f2d7 |
722 | mp *pr[DEPTH]; |
3bc9cb53 |
723 | size_t target = (MP_LEN(m) + 1) / 2; |
e360a4f2 |
724 | unsigned i = 0; |
2b26f2d7 |
725 | mp *z = mp_new(1, 0); |
e360a4f2 |
726 | |
727 | /* --- Set up the exponent table --- */ |
728 | |
2b26f2d7 |
729 | z->v[0] = (radix > 0 ? radix : -radix); |
e360a4f2 |
730 | z->f = 0; |
731 | for (;;) { |
2b26f2d7 |
732 | assert(((void)"Number is too unimaginably huge", i < DEPTH)); |
e360a4f2 |
733 | pr[i++] = z; |
734 | if (MP_LEN(z) > target) |
735 | break; |
736 | z = mp_sqr(MP_NEW, z); |
737 | } |
d3409d5e |
738 | |
e360a4f2 |
739 | /* --- Write out the answer --- */ |
d3409d5e |
740 | |
e360a4f2 |
741 | rc = complicated(m, radix, pr, i - 1, 0, ops, p); |
d3409d5e |
742 | |
e360a4f2 |
743 | /* --- Tidy away the array --- */ |
d3409d5e |
744 | |
e360a4f2 |
745 | while (i > 0) |
746 | mp_drop(pr[--i]); |
d3409d5e |
747 | } |
e360a4f2 |
748 | |
749 | /* --- Tidying up code --- */ |
750 | |
751 | MP_DROP(m); |
752 | return (rc); |
d3409d5e |
753 | } |
754 | |
755 | /*----- Test rig ----------------------------------------------------------*/ |
756 | |
757 | #ifdef TEST_RIG |
758 | |
759 | #include <mLib/testrig.h> |
760 | |
761 | static int verify(dstr *v) |
762 | { |
763 | int ok = 1; |
764 | int ib = *(int *)v[0].buf, ob = *(int *)v[2].buf; |
765 | dstr d = DSTR_INIT; |
766 | mp *m = mp_readdstr(MP_NEW, &v[1], 0, ib); |
767 | if (m) { |
768 | if (!ob) { |
769 | fprintf(stderr, "*** unexpected successful parse\n" |
a951033d |
770 | "*** input [%2i] = ", ib); |
2b26f2d7 |
771 | if (ib < 0) |
772 | type_hex.dump(&v[1], stderr); |
773 | else |
774 | fputs(v[1].buf, stderr); |
d3409d5e |
775 | mp_writedstr(m, &d, 10); |
2b26f2d7 |
776 | fprintf(stderr, "\n*** (value = %s)\n", d.buf); |
d3409d5e |
777 | ok = 0; |
778 | } else { |
779 | mp_writedstr(m, &d, ob); |
780 | if (d.len != v[3].len || memcmp(d.buf, v[3].buf, d.len) != 0) { |
781 | fprintf(stderr, "*** failed read or write\n" |
a951033d |
782 | "*** input [%2i] = ", ib); |
2b26f2d7 |
783 | if (ib < 0) |
784 | type_hex.dump(&v[1], stderr); |
785 | else |
786 | fputs(v[1].buf, stderr); |
a951033d |
787 | fprintf(stderr, "\n*** output [%2i] = ", ob); |
2b26f2d7 |
788 | if (ob < 0) |
789 | type_hex.dump(&d, stderr); |
790 | else |
791 | fputs(d.buf, stderr); |
a951033d |
792 | fprintf(stderr, "\n*** expected [%2i] = ", ob); |
2b26f2d7 |
793 | if (ob < 0) |
794 | type_hex.dump(&v[3], stderr); |
795 | else |
796 | fputs(v[3].buf, stderr); |
797 | fputc('\n', stderr); |
d3409d5e |
798 | ok = 0; |
799 | } |
800 | } |
801 | mp_drop(m); |
802 | } else { |
803 | if (ob) { |
804 | fprintf(stderr, "*** unexpected parse failure\n" |
2b26f2d7 |
805 | "*** input [%i] = ", ib); |
806 | if (ib < 0) |
807 | type_hex.dump(&v[1], stderr); |
808 | else |
809 | fputs(v[1].buf, stderr); |
810 | fprintf(stderr, "\n*** expected [%i] = ", ob); |
811 | if (ob < 0) |
812 | type_hex.dump(&v[3], stderr); |
813 | else |
814 | fputs(v[3].buf, stderr); |
815 | fputc('\n', stderr); |
d3409d5e |
816 | ok = 0; |
817 | } |
818 | } |
819 | |
820 | dstr_destroy(&d); |
9c3df6c0 |
821 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
d3409d5e |
822 | return (ok); |
823 | } |
824 | |
825 | static test_chunk tests[] = { |
2b26f2d7 |
826 | { "mptext-ascii", verify, |
d3409d5e |
827 | { &type_int, &type_string, &type_int, &type_string, 0 } }, |
2b26f2d7 |
828 | { "mptext-bin-in", verify, |
829 | { &type_int, &type_hex, &type_int, &type_string, 0 } }, |
830 | { "mptext-bin-out", verify, |
831 | { &type_int, &type_string, &type_int, &type_hex, 0 } }, |
d3409d5e |
832 | { 0, 0, { 0 } } |
833 | }; |
834 | |
835 | int main(int argc, char *argv[]) |
836 | { |
837 | sub_init(); |
838 | test_run(argc, argv, tests, SRCDIR "/tests/mptext"); |
839 | return (0); |
840 | } |
841 | |
842 | #endif |
843 | |
844 | /*----- That's all, folks -------------------------------------------------*/ |