d3409d5e |
1 | /* -*-c-*- |
2 | * |
6ea6fe51 |
3 | * $Id: mptext.c,v 1.13 2002/10/09 00:21:06 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 $ |
6ea6fe51 |
33 | * Revision 1.13 2002/10/09 00:21:06 mdw |
34 | * Allow user-specified `r_xx' bases to be up to 62. |
35 | * |
631673a1 |
36 | * Revision 1.12 2002/01/13 19:51:18 mdw |
37 | * Extend the textual format to bases up to 62 by distinguishing case. |
38 | * |
eaa515d8 |
39 | * Revision 1.11 2001/06/16 23:42:17 mdw |
40 | * Typesetting fixes. |
41 | * |
a951033d |
42 | * Revision 1.10 2001/06/16 13:22:39 mdw |
43 | * Added fast-track code for binary output bases, and tests. |
44 | * |
3bc9cb53 |
45 | * Revision 1.9 2001/02/03 16:05:17 mdw |
46 | * Make flags be unsigned. Improve the write algorithm: recurse until the |
47 | * parts are one word long and use single-precision arithmetic from there. |
48 | * Fix off-by-one bug when breaking the number apart. |
49 | * |
9d3838a0 |
50 | * Revision 1.8 2000/12/06 20:32:42 mdw |
51 | * Reduce binary bytes (to allow marker bits to be ignored). Fix error |
52 | * message string a bit. Allow leading `+' signs. |
53 | * |
7d45ed6c |
54 | * Revision 1.7 2000/07/15 10:01:08 mdw |
55 | * Bug fix in binary input. |
56 | * |
dd9199f0 |
57 | * Revision 1.6 2000/06/25 12:58:23 mdw |
58 | * Fix the derivation of `depth' commentary. |
59 | * |
2b26f2d7 |
60 | * Revision 1.5 2000/06/17 11:46:19 mdw |
61 | * New and much faster stack-based algorithm for reading integers. Support |
62 | * reading and writing binary integers in bases between 2 and 256. |
63 | * |
e360a4f2 |
64 | * Revision 1.4 1999/12/22 15:56:56 mdw |
65 | * Use clever recursive algorithm for writing numbers out. |
66 | * |
9c3df6c0 |
67 | * Revision 1.3 1999/12/10 23:23:26 mdw |
68 | * Allocate slightly less memory. |
69 | * |
90b6f0be |
70 | * Revision 1.2 1999/11/20 22:24:15 mdw |
71 | * Use function versions of MPX_UMULN and MPX_UADDN. |
72 | * |
d3409d5e |
73 | * Revision 1.1 1999/11/17 18:02:16 mdw |
74 | * New multiprecision integer arithmetic suite. |
75 | * |
76 | */ |
77 | |
78 | /*----- Header files ------------------------------------------------------*/ |
79 | |
80 | #include <ctype.h> |
2b26f2d7 |
81 | #include <limits.h> |
d3409d5e |
82 | #include <stdio.h> |
83 | |
d3409d5e |
84 | #include "mp.h" |
85 | #include "mptext.h" |
e360a4f2 |
86 | #include "paranoia.h" |
d3409d5e |
87 | |
2b26f2d7 |
88 | /*----- Magical numbers ---------------------------------------------------*/ |
89 | |
90 | /* --- Maximum recursion depth --- * |
91 | * |
92 | * This is the number of bits in a @size_t@ object. Why? |
93 | * |
eaa515d8 |
94 | * To see this, let %$b = \textit{MPW\_MAX} + 1$% and let %$Z$% be the |
dd9199f0 |
95 | * largest @size_t@ value. Then the largest possible @mp@ is %$M - 1$% where |
96 | * %$M = b^Z$%. Let %$r$% be a radix to read or write. Since the recursion |
97 | * squares the radix at each step, the highest number reached by the |
98 | * recursion is %$d$%, where: |
2b26f2d7 |
99 | * |
dd9199f0 |
100 | * %$r^{2^d} = b^Z$%. |
2b26f2d7 |
101 | * |
102 | * Solving gives that %$d = \lg \log_r b^Z$%. If %$r = 2$%, this is maximum, |
103 | * so choosing %$d = \lg \lg b^Z = \lg (Z \lg b) = \lg Z + \lg \lg b$%. |
104 | * |
105 | * Expressing %$\lg Z$% as @CHAR_BIT * sizeof(size_t)@ yields an |
106 | * overestimate, since a @size_t@ representation may contain `holes'. |
107 | * Choosing to represent %$\lg \lg b$% by 10 is almost certainly sufficient |
108 | * for `some time to come'. |
109 | */ |
110 | |
111 | #define DEPTH (CHAR_BIT * sizeof(size_t) + 10) |
112 | |
d3409d5e |
113 | /*----- Main code ---------------------------------------------------------*/ |
114 | |
115 | /* --- @mp_read@ --- * |
116 | * |
117 | * Arguments: @mp *m@ = destination multiprecision number |
118 | * @int radix@ = base to assume for data (or zero to guess) |
119 | * @const mptext_ops *ops@ = pointer to operations block |
120 | * @void *p@ = data for the operations block |
121 | * |
122 | * Returns: The integer read, or zero if it didn't work. |
123 | * |
124 | * Use: Reads an integer from some source. If the @radix@ is |
125 | * specified, the number is assumed to be given in that radix, |
126 | * with the letters `a' (either upper- or lower-case) upwards |
127 | * standing for digits greater than 9. Otherwise, base 10 is |
128 | * assumed unless the number starts with `0' (octal), `0x' (hex) |
129 | * or `nnn_' (base `nnn'). An arbitrary amount of whitespace |
130 | * before the number is ignored. |
131 | */ |
132 | |
2b26f2d7 |
133 | /* --- About the algorithm --- * |
134 | * |
135 | * The algorithm here is rather aggressive. I maintain an array of |
136 | * successive squarings of the radix, and a stack of partial results, each |
137 | * with a counter attached indicating which radix square to multiply by. |
138 | * Once the item at the top of the stack reaches the same counter level as |
139 | * the next item down, they are combined together and the result is given a |
140 | * counter level one higher than either of the results. |
141 | * |
142 | * Gluing the results together at the end is slightly tricky. Pay attention |
143 | * to the code. |
144 | * |
145 | * This is more complicated because of the need to handle the slightly |
146 | * bizarre syntax. |
147 | */ |
148 | |
d3409d5e |
149 | mp *mp_read(mp *m, int radix, const mptext_ops *ops, void *p) |
150 | { |
2b26f2d7 |
151 | int ch; /* Current char being considered */ |
152 | unsigned f = 0; /* Flags about the current number */ |
153 | int r; /* Radix to switch over to */ |
154 | mpw rd; /* Radix as an @mp@ digit */ |
155 | mp rr; /* The @mp@ for the radix */ |
156 | unsigned nf = m ? m->f & MP_BURN : 0; /* New @mp@ flags */ |
157 | |
158 | /* --- Stacks --- */ |
159 | |
160 | mp *pow[DEPTH]; /* List of powers */ |
161 | unsigned pows; /* Next index to fill */ |
162 | struct { unsigned i; mp *m; } s[DEPTH]; /* Main stack */ |
163 | unsigned sp; /* Current stack pointer */ |
164 | |
165 | /* --- Flags --- */ |
d3409d5e |
166 | |
3bc9cb53 |
167 | #define f_neg 1u |
168 | #define f_ok 2u |
a951033d |
169 | #define f_start 4u |
d3409d5e |
170 | |
2b26f2d7 |
171 | /* --- Initialize the stacks --- */ |
172 | |
173 | mp_build(&rr, &rd, &rd + 1); |
174 | pow[0] = &rr; |
175 | pows = 1; |
176 | |
177 | sp = 0; |
178 | |
d3409d5e |
179 | /* --- Initialize the destination number --- */ |
180 | |
2b26f2d7 |
181 | if (m) |
182 | MP_DROP(m); |
d3409d5e |
183 | |
184 | /* --- Read an initial character --- */ |
185 | |
186 | ch = ops->get(p); |
187 | while (isspace(ch)) |
188 | ch = ops->get(p); |
189 | |
190 | /* --- Handle an initial sign --- */ |
191 | |
9d3838a0 |
192 | if (radix >= 0 && (ch == '-' || ch == '+')) { |
193 | if (ch == '-') |
194 | f |= f_neg; |
195 | do ch = ops->get(p); while isspace(ch); |
d3409d5e |
196 | } |
197 | |
198 | /* --- If the radix is zero, look for leading zeros --- */ |
199 | |
2b26f2d7 |
200 | if (radix > 0) { |
631673a1 |
201 | assert(((void)"ascii radix must be <= 62", radix <= 62)); |
2b26f2d7 |
202 | rd = radix; |
203 | r = -1; |
204 | } else if (radix < 0) { |
205 | rd = -radix; |
9d3838a0 |
206 | assert(((void)"binary radix must fit in a byte", rd < UCHAR_MAX)); |
d3409d5e |
207 | r = -1; |
2b26f2d7 |
208 | } else if (ch != '0') { |
209 | rd = 10; |
d3409d5e |
210 | r = 0; |
211 | } else { |
212 | ch = ops->get(p); |
213 | if (ch == 'x') { |
214 | ch = ops->get(p); |
2b26f2d7 |
215 | rd = 16; |
d3409d5e |
216 | } else { |
2b26f2d7 |
217 | rd = 8; |
d3409d5e |
218 | f |= f_ok; |
219 | } |
220 | r = -1; |
221 | } |
222 | |
a951033d |
223 | /* --- Use fast algorithm for binary radix --- * |
224 | * |
225 | * This is the restart point after having parsed a radix number from the |
226 | * input. We check whether the radix is binary, and if so use a fast |
227 | * algorithm which just stacks the bits up in the right order. |
228 | */ |
229 | |
230 | restart: |
231 | switch (rd) { |
232 | unsigned bit; |
233 | |
234 | case 2: bit = 1; goto bin; |
235 | case 4: bit = 2; goto bin; |
236 | case 8: bit = 3; goto bin; |
237 | case 16: bit = 4; goto bin; |
238 | case 32: bit = 5; goto bin; |
239 | case 64: bit = 6; goto bin; |
240 | case 128: bit = 7; goto bin; |
241 | default: |
242 | break; |
243 | |
244 | /* --- The fast binary algorithm --- * |
245 | * |
246 | * We stack bits up starting at the top end of a word. When one word is |
247 | * full, we write it to the integer, and start another with the left-over |
248 | * bits. When the array in the integer is full, we resize using low-level |
249 | * calls and copy the current data to the top end. Finally, we do a single |
250 | * bit-shift when we know where the end of the number is. |
251 | */ |
252 | |
253 | bin: { |
254 | mpw a = 0; |
255 | unsigned b = MPW_BITS; |
256 | size_t len, n; |
257 | mpw *v; |
258 | |
259 | m = mp_dest(MP_NEW, 1, nf); |
260 | len = n = m->sz; |
261 | n = len; |
262 | v = m->v + n; |
263 | for (;; ch = ops->get(p)) { |
264 | unsigned x; |
265 | |
266 | if (ch < 0) |
267 | break; |
268 | |
269 | /* --- Check that the character is a digit and in range --- */ |
270 | |
271 | if (radix < 0) |
272 | x = ch % rd; |
273 | else { |
274 | if (!isalnum(ch)) |
275 | break; |
276 | if (ch >= '0' && ch <= '9') |
277 | x = ch - '0'; |
278 | else { |
631673a1 |
279 | if (rd <= 36) |
280 | ch = tolower(ch); |
a951033d |
281 | if (ch >= 'a' && ch <= 'z') /* ASCII dependent! */ |
282 | x = ch - 'a' + 10; |
631673a1 |
283 | else if (ch >= 'A' && ch <= 'Z') |
284 | x = ch - 'A' + 36; |
a951033d |
285 | else |
286 | break; |
287 | } |
288 | } |
289 | if (x >= rd) |
290 | break; |
291 | |
292 | /* --- Feed the digit into the accumulator --- */ |
293 | |
294 | f |= f_ok; |
295 | if (!x && !(f & f_start)) |
296 | continue; |
297 | f |= f_start; |
298 | if (b > bit) { |
299 | b -= bit; |
300 | a |= MPW(x) << b; |
301 | } else { |
302 | a |= MPW(x) >> (bit - b); |
303 | b += MPW_BITS - bit; |
304 | *--v = MPW(a); |
305 | n--; |
306 | if (!n) { |
307 | n = len; |
308 | len <<= 1; |
309 | v = mpalloc(m->a, len); |
310 | memcpy(v + n, m->v, MPWS(n)); |
311 | mpfree(m->a, m->v); |
312 | m->v = v; |
313 | v = m->v + n; |
314 | } |
315 | a = (b < MPW_BITS) ? MPW(x) << b : 0; |
316 | } |
317 | } |
318 | |
319 | /* --- Finish up --- */ |
320 | |
321 | if (!(f & f_ok)) { |
322 | mp_drop(m); |
323 | m = 0; |
324 | } else { |
325 | *--v = MPW(a); |
326 | n--; |
327 | m->sz = len; |
328 | m->vl = m->v + len; |
329 | m->f &= ~MP_UNDEF; |
330 | m = mp_lsr(m, m, (unsigned long)n * MPW_BITS + b); |
331 | } |
332 | goto done; |
333 | }} |
334 | |
d3409d5e |
335 | /* --- Time to start --- */ |
336 | |
337 | for (;; ch = ops->get(p)) { |
a951033d |
338 | unsigned x; |
d3409d5e |
339 | |
7d45ed6c |
340 | if (ch < 0) |
341 | break; |
342 | |
d3409d5e |
343 | /* --- An underscore indicates a numbered base --- */ |
344 | |
6ea6fe51 |
345 | if (ch == '_' && r > 0 && r <= 62) { |
2b26f2d7 |
346 | unsigned i; |
347 | |
348 | /* --- Clear out the stacks --- */ |
349 | |
350 | for (i = 1; i < pows; i++) |
351 | MP_DROP(pow[i]); |
352 | pows = 1; |
353 | for (i = 0; i < sp; i++) |
354 | MP_DROP(s[i].m); |
355 | sp = 0; |
356 | |
357 | /* --- Restart the search --- */ |
358 | |
359 | rd = r; |
d3409d5e |
360 | r = -1; |
361 | f &= ~f_ok; |
a951033d |
362 | ch = ops->get(p); |
363 | goto restart; |
d3409d5e |
364 | } |
365 | |
366 | /* --- Check that the character is a digit and in range --- */ |
367 | |
2b26f2d7 |
368 | if (radix < 0) |
9d3838a0 |
369 | x = ch % rd; |
d3409d5e |
370 | else { |
2b26f2d7 |
371 | if (!isalnum(ch)) |
d3409d5e |
372 | break; |
2b26f2d7 |
373 | if (ch >= '0' && ch <= '9') |
374 | x = ch - '0'; |
375 | else { |
631673a1 |
376 | if (rd <= 36) |
377 | ch = tolower(ch); |
2b26f2d7 |
378 | if (ch >= 'a' && ch <= 'z') /* ASCII dependent! */ |
379 | x = ch - 'a' + 10; |
631673a1 |
380 | else if (ch >= 'A' && ch <= 'Z') |
381 | x = ch - 'A' + 36; |
2b26f2d7 |
382 | else |
383 | break; |
384 | } |
d3409d5e |
385 | } |
386 | |
387 | /* --- Sort out what to do with the character --- */ |
388 | |
389 | if (x >= 10 && r >= 0) |
390 | r = -1; |
2b26f2d7 |
391 | if (x >= rd) |
d3409d5e |
392 | break; |
393 | |
394 | if (r >= 0) |
395 | r = r * 10 + x; |
396 | |
397 | /* --- Stick the character on the end of my integer --- */ |
398 | |
2b26f2d7 |
399 | assert(((void)"Number is too unimaginably huge", sp < DEPTH)); |
400 | s[sp].m = m = mp_new(1, nf); |
401 | m->v[0] = x; |
402 | s[sp].i = 0; |
403 | |
404 | /* --- Now grind through the stack --- */ |
405 | |
406 | while (sp > 0 && s[sp - 1].i == s[sp].i) { |
407 | |
408 | /* --- Combine the top two items --- */ |
409 | |
410 | sp--; |
411 | m = s[sp].m; |
412 | m = mp_mul(m, m, pow[s[sp].i]); |
413 | m = mp_add(m, m, s[sp + 1].m); |
414 | s[sp].m = m; |
415 | MP_DROP(s[sp + 1].m); |
416 | s[sp].i++; |
417 | |
418 | /* --- Make a new radix power if necessary --- */ |
419 | |
420 | if (s[sp].i >= pows) { |
421 | assert(((void)"Number is too unimaginably huge", pows < DEPTH)); |
422 | pow[pows] = mp_sqr(MP_NEW, pow[pows - 1]); |
423 | pows++; |
424 | } |
425 | } |
d3409d5e |
426 | f |= f_ok; |
2b26f2d7 |
427 | sp++; |
d3409d5e |
428 | } |
429 | |
430 | ops->unget(ch, p); |
431 | |
2b26f2d7 |
432 | /* --- If we're done, compute the rest of the number --- */ |
433 | |
434 | if (f & f_ok) { |
435 | if (!sp) |
436 | return (MP_ZERO); |
437 | else { |
438 | mp *z = MP_ONE; |
439 | sp--; |
440 | |
441 | while (sp > 0) { |
442 | |
443 | /* --- Combine the top two items --- */ |
444 | |
445 | sp--; |
446 | m = s[sp].m; |
447 | z = mp_mul(z, z, pow[s[sp + 1].i]); |
448 | m = mp_mul(m, m, z); |
449 | m = mp_add(m, m, s[sp + 1].m); |
450 | s[sp].m = m; |
451 | MP_DROP(s[sp + 1].m); |
452 | |
453 | /* --- Make a new radix power if necessary --- */ |
454 | |
455 | if (s[sp].i >= pows) { |
456 | assert(((void)"Number is too unimaginably huge", pows < DEPTH)); |
457 | pow[pows] = mp_sqr(MP_NEW, pow[pows - 1]); |
458 | pows++; |
459 | } |
460 | } |
461 | MP_DROP(z); |
462 | m = s[0].m; |
463 | } |
464 | } else { |
465 | unsigned i; |
466 | for (i = 0; i < sp; i++) |
467 | MP_DROP(s[i].m); |
468 | } |
469 | |
470 | /* --- Clear the radix power list --- */ |
471 | |
472 | { |
473 | unsigned i; |
474 | for (i = 1; i < pows; i++) |
475 | MP_DROP(pow[i]); |
476 | } |
477 | |
d3409d5e |
478 | /* --- Bail out if the number was bad --- */ |
479 | |
a951033d |
480 | done: |
2b26f2d7 |
481 | if (!(f & f_ok)) |
d3409d5e |
482 | return (0); |
d3409d5e |
483 | |
484 | /* --- Set the sign and return --- */ |
485 | |
d3409d5e |
486 | if (f & f_neg) |
487 | m->f |= MP_NEG; |
488 | return (m); |
3bc9cb53 |
489 | |
a951033d |
490 | #undef f_start |
3bc9cb53 |
491 | #undef f_neg |
492 | #undef f_ok |
d3409d5e |
493 | } |
494 | |
495 | /* --- @mp_write@ --- * |
496 | * |
497 | * Arguments: @mp *m@ = pointer to a multi-precision integer |
498 | * @int radix@ = radix to use when writing the number out |
499 | * @const mptext_ops *ops@ = pointer to an operations block |
500 | * @void *p@ = data for the operations block |
501 | * |
502 | * Returns: Zero if it worked, nonzero otherwise. |
503 | * |
504 | * Use: Writes a large integer in textual form. |
505 | */ |
506 | |
e360a4f2 |
507 | /* --- Simple case --- * |
508 | * |
3bc9cb53 |
509 | * Use a fixed-sized buffer and single-precision arithmetic to pick off |
510 | * low-order digits. Put each digit in a buffer, working backwards from the |
511 | * end. If the buffer becomes full, recurse to get another one. Ensure that |
512 | * there are at least @z@ digits by writing leading zeroes if there aren't |
513 | * enough real digits. |
e360a4f2 |
514 | */ |
515 | |
3bc9cb53 |
516 | static int simple(mpw n, int radix, unsigned z, |
e360a4f2 |
517 | const mptext_ops *ops, void *p) |
518 | { |
519 | int rc = 0; |
520 | char buf[64]; |
521 | unsigned i = sizeof(buf); |
2b26f2d7 |
522 | int rd = radix > 0 ? radix : -radix; |
e360a4f2 |
523 | |
524 | do { |
525 | int ch; |
526 | mpw x; |
527 | |
3bc9cb53 |
528 | x = n % rd; |
529 | n /= rd; |
2b26f2d7 |
530 | if (radix < 0) |
531 | ch = x; |
3bc9cb53 |
532 | else if (x < 10) |
533 | ch = '0' + x; |
631673a1 |
534 | else if (x < 36) /* Ascii specific */ |
3bc9cb53 |
535 | ch = 'a' + x - 10; |
631673a1 |
536 | else |
537 | ch = 'A' + x - 36; |
e360a4f2 |
538 | buf[--i] = ch; |
539 | if (z) |
540 | z--; |
3bc9cb53 |
541 | } while (i && n); |
e360a4f2 |
542 | |
3bc9cb53 |
543 | if (n) |
544 | rc = simple(n, radix, z, ops, p); |
e360a4f2 |
545 | else { |
a951033d |
546 | char zbuf[32]; |
547 | memset(zbuf, (radix < 0) ? 0 : '0', sizeof(zbuf)); |
548 | while (!rc && z >= sizeof(zbuf)) { |
549 | rc = ops->put(zbuf, sizeof(zbuf), p); |
550 | z -= sizeof(zbuf); |
e360a4f2 |
551 | } |
552 | if (!rc && z) |
a951033d |
553 | rc = ops->put(zbuf, z, p); |
e360a4f2 |
554 | } |
555 | if (!rc) |
3bc9cb53 |
556 | rc = ops->put(buf + i, sizeof(buf) - i, p); |
557 | BURN(buf); |
e360a4f2 |
558 | return (rc); |
559 | } |
560 | |
561 | /* --- Complicated case --- * |
562 | * |
563 | * If the number is small, fall back to the simple case above. Otherwise |
564 | * divide and take remainder by current large power of the radix, and emit |
565 | * each separately. Don't emit a zero quotient. Be very careful about |
566 | * leading zeroes on the remainder part, because they're deeply significant. |
567 | */ |
568 | |
569 | static int complicated(mp *m, int radix, mp **pr, unsigned i, unsigned z, |
570 | const mptext_ops *ops, void *p) |
571 | { |
572 | int rc = 0; |
573 | mp *q = MP_NEW; |
574 | unsigned d = 1 << i; |
575 | |
3bc9cb53 |
576 | if (MP_LEN(m) < 2) |
577 | return (simple(MP_LEN(m) ? m->v[0] : 0, radix, z, ops, p)); |
e360a4f2 |
578 | |
3bc9cb53 |
579 | assert(i); |
e360a4f2 |
580 | mp_div(&q, &m, m, pr[i]); |
581 | if (!MP_LEN(q)) |
582 | d = z; |
583 | else { |
584 | if (z > d) |
585 | z -= d; |
586 | else |
587 | z = 0; |
588 | rc = complicated(q, radix, pr, i - 1, z, ops, p); |
589 | } |
590 | if (!rc) |
591 | rc = complicated(m, radix, pr, i - 1, d, ops, p); |
592 | mp_drop(q); |
593 | return (rc); |
594 | } |
595 | |
a951033d |
596 | /* --- Binary case --- * |
597 | * |
598 | * Special case for binary output. Goes much faster. |
599 | */ |
600 | |
601 | static int binary(mp *m, int bit, int radix, const mptext_ops *ops, void *p) |
602 | { |
603 | mpw *v; |
604 | mpw a; |
605 | int rc = 0; |
606 | unsigned b; |
607 | unsigned mask; |
608 | unsigned long n; |
609 | unsigned f = 0; |
610 | char buf[8], *q; |
611 | unsigned x; |
612 | int ch; |
613 | |
614 | #define f_out 1u |
615 | |
616 | /* --- Work out where to start --- */ |
617 | |
618 | n = mp_bits(m); |
619 | n += bit - (n % bit); |
620 | b = n % MPW_BITS; |
621 | n /= MPW_BITS; |
622 | |
623 | if (n > MP_LEN(m)) { |
624 | n--; |
625 | b += MPW_BITS; |
626 | } |
627 | |
628 | v = m->v + n; |
629 | a = *v; |
630 | mask = (1 << bit) - 1; |
631 | q = buf; |
632 | |
633 | /* --- Main code --- */ |
634 | |
635 | for (;;) { |
636 | if (b > bit) { |
637 | b -= bit; |
638 | x = a >> b; |
639 | } else { |
640 | x = a << (bit - b); |
641 | b += MPW_BITS - bit; |
642 | if (v == m->v) |
643 | break; |
644 | a = *--v; |
645 | if (b < MPW_BITS) |
646 | x |= a >> b; |
647 | } |
648 | x &= mask; |
649 | if (!x && !(f & f_out)) |
650 | continue; |
651 | |
652 | if (radix < 0) |
653 | ch = x; |
654 | else if (x < 10) |
655 | ch = '0' + x; |
631673a1 |
656 | else if (x < 36) |
657 | ch = 'a' + x - 10; /* Ascii specific */ |
a951033d |
658 | else |
631673a1 |
659 | ch = 'A' + x - 36; |
a951033d |
660 | *q++ = ch; |
661 | if (q >= buf + sizeof(buf)) { |
662 | if ((rc = ops->put(buf, sizeof(buf), p)) != 0) |
663 | goto done; |
664 | q = buf; |
665 | } |
666 | f |= f_out; |
667 | } |
668 | |
669 | x &= mask; |
670 | if (radix < 0) |
671 | ch = x; |
672 | else if (x < 10) |
673 | ch = '0' + x; |
631673a1 |
674 | else if (x < 36) |
675 | ch = 'a' + x - 10; /* Ascii specific */ |
a951033d |
676 | else |
631673a1 |
677 | ch = 'A' + x - 36; |
a951033d |
678 | *q++ = ch; |
679 | rc = ops->put(buf, q - buf, p); |
680 | |
681 | done: |
682 | mp_drop(m); |
683 | return (rc); |
684 | |
685 | #undef f_out |
686 | } |
687 | |
e360a4f2 |
688 | /* --- Main driver code --- */ |
689 | |
d3409d5e |
690 | int mp_write(mp *m, int radix, const mptext_ops *ops, void *p) |
691 | { |
e360a4f2 |
692 | int rc; |
d3409d5e |
693 | |
694 | /* --- Set various things up --- */ |
695 | |
696 | m = MP_COPY(m); |
e360a4f2 |
697 | MP_SPLIT(m); |
d3409d5e |
698 | |
2b26f2d7 |
699 | /* --- Check the radix for sensibleness --- */ |
700 | |
701 | if (radix > 0) |
631673a1 |
702 | assert(((void)"ascii radix must be <= 62", radix <= 62)); |
2b26f2d7 |
703 | else if (radix < 0) |
704 | assert(((void)"binary radix must fit in a byte", -radix < UCHAR_MAX)); |
705 | else |
706 | assert(((void)"radix can't be zero in mp_write", 0)); |
707 | |
d3409d5e |
708 | /* --- If the number is negative, sort that out --- */ |
709 | |
710 | if (m->f & MP_NEG) { |
711 | if (ops->put("-", 1, p)) |
712 | return (EOF); |
2b26f2d7 |
713 | m->f &= ~MP_NEG; |
d3409d5e |
714 | } |
715 | |
a951033d |
716 | /* --- Handle binary radix --- */ |
717 | |
718 | switch (radix) { |
719 | case 2: case -2: return (binary(m, 1, radix, ops, p)); |
720 | case 4: case -4: return (binary(m, 2, radix, ops, p)); |
721 | case 8: case -8: return (binary(m, 3, radix, ops, p)); |
722 | case 16: case -16: return (binary(m, 4, radix, ops, p)); |
723 | case 32: case -32: return (binary(m, 5, radix, ops, p)); |
724 | case -64: return (binary(m, 6, radix, ops, p)); |
725 | case -128: return (binary(m, 7, radix, ops, p)); |
726 | } |
727 | |
e360a4f2 |
728 | /* --- If the number is small, do it the easy way --- */ |
729 | |
3bc9cb53 |
730 | if (MP_LEN(m) < 2) |
731 | rc = simple(MP_LEN(m) ? m->v[0] : 0, radix, 0, ops, p); |
e360a4f2 |
732 | |
733 | /* --- Use a clever algorithm --- * |
734 | * |
735 | * Square the radix repeatedly, remembering old results, until I get |
736 | * something more than half the size of the number @m@. Use this to divide |
737 | * the number: the quotient and remainder will be approximately the same |
738 | * size, and I'll have split them on a digit boundary, so I can just emit |
739 | * the quotient and remainder recursively, in order. |
e360a4f2 |
740 | */ |
741 | |
742 | else { |
2b26f2d7 |
743 | mp *pr[DEPTH]; |
3bc9cb53 |
744 | size_t target = (MP_LEN(m) + 1) / 2; |
e360a4f2 |
745 | unsigned i = 0; |
2b26f2d7 |
746 | mp *z = mp_new(1, 0); |
e360a4f2 |
747 | |
748 | /* --- Set up the exponent table --- */ |
749 | |
2b26f2d7 |
750 | z->v[0] = (radix > 0 ? radix : -radix); |
e360a4f2 |
751 | z->f = 0; |
752 | for (;;) { |
2b26f2d7 |
753 | assert(((void)"Number is too unimaginably huge", i < DEPTH)); |
e360a4f2 |
754 | pr[i++] = z; |
755 | if (MP_LEN(z) > target) |
756 | break; |
757 | z = mp_sqr(MP_NEW, z); |
758 | } |
d3409d5e |
759 | |
e360a4f2 |
760 | /* --- Write out the answer --- */ |
d3409d5e |
761 | |
e360a4f2 |
762 | rc = complicated(m, radix, pr, i - 1, 0, ops, p); |
d3409d5e |
763 | |
e360a4f2 |
764 | /* --- Tidy away the array --- */ |
d3409d5e |
765 | |
e360a4f2 |
766 | while (i > 0) |
767 | mp_drop(pr[--i]); |
d3409d5e |
768 | } |
e360a4f2 |
769 | |
770 | /* --- Tidying up code --- */ |
771 | |
772 | MP_DROP(m); |
773 | return (rc); |
d3409d5e |
774 | } |
775 | |
776 | /*----- Test rig ----------------------------------------------------------*/ |
777 | |
778 | #ifdef TEST_RIG |
779 | |
780 | #include <mLib/testrig.h> |
781 | |
782 | static int verify(dstr *v) |
783 | { |
784 | int ok = 1; |
785 | int ib = *(int *)v[0].buf, ob = *(int *)v[2].buf; |
786 | dstr d = DSTR_INIT; |
787 | mp *m = mp_readdstr(MP_NEW, &v[1], 0, ib); |
788 | if (m) { |
789 | if (!ob) { |
790 | fprintf(stderr, "*** unexpected successful parse\n" |
a951033d |
791 | "*** input [%2i] = ", ib); |
2b26f2d7 |
792 | if (ib < 0) |
793 | type_hex.dump(&v[1], stderr); |
794 | else |
795 | fputs(v[1].buf, stderr); |
d3409d5e |
796 | mp_writedstr(m, &d, 10); |
2b26f2d7 |
797 | fprintf(stderr, "\n*** (value = %s)\n", d.buf); |
d3409d5e |
798 | ok = 0; |
799 | } else { |
800 | mp_writedstr(m, &d, ob); |
801 | if (d.len != v[3].len || memcmp(d.buf, v[3].buf, d.len) != 0) { |
802 | fprintf(stderr, "*** failed read or write\n" |
a951033d |
803 | "*** input [%2i] = ", ib); |
2b26f2d7 |
804 | if (ib < 0) |
805 | type_hex.dump(&v[1], stderr); |
806 | else |
807 | fputs(v[1].buf, stderr); |
a951033d |
808 | fprintf(stderr, "\n*** output [%2i] = ", ob); |
2b26f2d7 |
809 | if (ob < 0) |
810 | type_hex.dump(&d, stderr); |
811 | else |
812 | fputs(d.buf, stderr); |
a951033d |
813 | fprintf(stderr, "\n*** expected [%2i] = ", ob); |
2b26f2d7 |
814 | if (ob < 0) |
815 | type_hex.dump(&v[3], stderr); |
816 | else |
817 | fputs(v[3].buf, stderr); |
818 | fputc('\n', stderr); |
d3409d5e |
819 | ok = 0; |
820 | } |
821 | } |
822 | mp_drop(m); |
823 | } else { |
824 | if (ob) { |
825 | fprintf(stderr, "*** unexpected parse failure\n" |
2b26f2d7 |
826 | "*** input [%i] = ", ib); |
827 | if (ib < 0) |
828 | type_hex.dump(&v[1], stderr); |
829 | else |
830 | fputs(v[1].buf, stderr); |
831 | fprintf(stderr, "\n*** expected [%i] = ", ob); |
832 | if (ob < 0) |
833 | type_hex.dump(&v[3], stderr); |
834 | else |
835 | fputs(v[3].buf, stderr); |
836 | fputc('\n', stderr); |
d3409d5e |
837 | ok = 0; |
838 | } |
839 | } |
840 | |
841 | dstr_destroy(&d); |
9c3df6c0 |
842 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
d3409d5e |
843 | return (ok); |
844 | } |
845 | |
846 | static test_chunk tests[] = { |
2b26f2d7 |
847 | { "mptext-ascii", verify, |
d3409d5e |
848 | { &type_int, &type_string, &type_int, &type_string, 0 } }, |
2b26f2d7 |
849 | { "mptext-bin-in", verify, |
850 | { &type_int, &type_hex, &type_int, &type_string, 0 } }, |
851 | { "mptext-bin-out", verify, |
852 | { &type_int, &type_string, &type_int, &type_hex, 0 } }, |
d3409d5e |
853 | { 0, 0, { 0 } } |
854 | }; |
855 | |
856 | int main(int argc, char *argv[]) |
857 | { |
858 | sub_init(); |
859 | test_run(argc, argv, tests, SRCDIR "/tests/mptext"); |
860 | return (0); |
861 | } |
862 | |
863 | #endif |
864 | |
865 | /*----- That's all, folks -------------------------------------------------*/ |