ucgendat.c revision 1.1.1.1.6.2 1 /* $OpenLDAP: pkg/ldap/libraries/liblunicode/ucdata/ucgendat.c,v 1.39.2.3 2008/02/11 23:26:42 kurt Exp $ */
2 /* This work is part of OpenLDAP Software <http://www.openldap.org/>.
3 *
4 * Copyright 1998-2008 The OpenLDAP Foundation.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted only as authorized by the OpenLDAP
9 * Public License.
10 *
11 * A copy of this license is available in file LICENSE in the
12 * top-level directory of the distribution or, alternatively, at
13 * <http://www.OpenLDAP.org/license.html>.
14 */
15 /* Copyright 2001 Computing Research Labs, New Mexico State University
16 *
17 * Permission is hereby granted, free of charge, to any person obtaining a
18 * copy of this software and associated documentation files (the "Software"),
19 * to deal in the Software without restriction, including without limitation
20 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
21 * and/or sell copies of the Software, and to permit persons to whom the
22 * Software is furnished to do so, subject to the following conditions:
23 *
24 * The above copyright notice and this permission notice shall be included in
25 * all copies or substantial portions of the Software.
26 *
27 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
28 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
29 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
30 * THE COMPUTING RESEARCH LAB OR NEW MEXICO STATE UNIVERSITY BE LIABLE FOR ANY
31 * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT
32 * OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
33 * THE USE OR OTHER DEALINGS IN THE SOFTWARE.
34 */
35 /* $Id: ucgendat.c,v 1.1.1.1.6.2 2008/09/18 05:15:03 wrstuden Exp $" */
36
37 #include "portable.h"
38 #include "ldap_config.h"
39
40 #include <stdio.h>
41 #include <ac/ctype.h>
42 #include <ac/stdlib.h>
43 #include <ac/string.h>
44 #include <ac/unistd.h>
45
46 #include <ac/bytes.h>
47
48 #include <lutil.h>
49
50 #ifndef HARDCODE_DATA
51 #define HARDCODE_DATA 1
52 #endif
53
54 #undef ishdigit
55 #define ishdigit(cc) (((cc) >= '0' && (cc) <= '9') ||\
56 ((cc) >= 'A' && (cc) <= 'F') ||\
57 ((cc) >= 'a' && (cc) <= 'f'))
58
59 /*
60 * A header written to the output file with the byte-order-mark and the number
61 * of property nodes.
62 */
63 static ac_uint2 hdr[2] = {0xfeff, 0};
64
65 #define NUMPROPS 50
66 #define NEEDPROPS (NUMPROPS + (4 - (NUMPROPS & 3)))
67
68 typedef struct {
69 char *name;
70 int len;
71 } _prop_t;
72
73 /*
74 * List of properties expected to be found in the Unicode Character Database
75 * including some implementation specific properties.
76 *
77 * The implementation specific properties are:
78 * Cm = Composed (can be decomposed)
79 * Nb = Non-breaking
80 * Sy = Symmetric (has left and right forms)
81 * Hd = Hex digit
82 * Qm = Quote marks
83 * Mr = Mirroring
84 * Ss = Space, other
85 * Cp = Defined character
86 */
87 static _prop_t props[NUMPROPS] = {
88 {"Mn", 2}, {"Mc", 2}, {"Me", 2}, {"Nd", 2}, {"Nl", 2}, {"No", 2},
89 {"Zs", 2}, {"Zl", 2}, {"Zp", 2}, {"Cc", 2}, {"Cf", 2}, {"Cs", 2},
90 {"Co", 2}, {"Cn", 2}, {"Lu", 2}, {"Ll", 2}, {"Lt", 2}, {"Lm", 2},
91 {"Lo", 2}, {"Pc", 2}, {"Pd", 2}, {"Ps", 2}, {"Pe", 2}, {"Po", 2},
92 {"Sm", 2}, {"Sc", 2}, {"Sk", 2}, {"So", 2}, {"L", 1}, {"R", 1},
93 {"EN", 2}, {"ES", 2}, {"ET", 2}, {"AN", 2}, {"CS", 2}, {"B", 1},
94 {"S", 1}, {"WS", 2}, {"ON", 2},
95 {"Cm", 2}, {"Nb", 2}, {"Sy", 2}, {"Hd", 2}, {"Qm", 2}, {"Mr", 2},
96 {"Ss", 2}, {"Cp", 2}, {"Pi", 2}, {"Pf", 2}, {"AL", 2}
97 };
98
99 typedef struct {
100 ac_uint4 *ranges;
101 ac_uint2 used;
102 ac_uint2 size;
103 } _ranges_t;
104
105 static _ranges_t proptbl[NUMPROPS];
106
107 /*
108 * Make sure this array is sized to be on a 4-byte boundary at compile time.
109 */
110 static ac_uint2 propcnt[NEEDPROPS];
111
112 /*
113 * Array used to collect a decomposition before adding it to the decomposition
114 * table.
115 */
116 static ac_uint4 dectmp[64];
117 static ac_uint4 dectmp_size;
118
119 typedef struct {
120 ac_uint4 code;
121 ac_uint2 size;
122 ac_uint2 used;
123 ac_uint4 *decomp;
124 } _decomp_t;
125
126 /*
127 * List of decomposition. Created and expanded in order as the characters are
128 * encountered. First list contains canonical mappings, second also includes
129 * compatibility mappings.
130 */
131 static _decomp_t *decomps;
132 static ac_uint4 decomps_used;
133 static ac_uint4 decomps_size;
134
135 static _decomp_t *kdecomps;
136 static ac_uint4 kdecomps_used;
137 static ac_uint4 kdecomps_size;
138
139 /*
140 * Composition exclusion table stuff.
141 */
142 #define COMPEX_SET(c) (compexs[(c) >> 5] |= (1 << ((c) & 31)))
143 #define COMPEX_TEST(c) (compexs[(c) >> 5] & (1 << ((c) & 31)))
144 static ac_uint4 compexs[8192];
145
146 /*
147 * Struct for holding a composition pair, and array of composition pairs
148 */
149 typedef struct {
150 ac_uint4 comp;
151 ac_uint4 count;
152 ac_uint4 code1;
153 ac_uint4 code2;
154 } _comp_t;
155
156 static _comp_t *comps;
157 static ac_uint4 comps_used;
158
159 /*
160 * Types and lists for handling lists of case mappings.
161 */
162 typedef struct {
163 ac_uint4 key;
164 ac_uint4 other1;
165 ac_uint4 other2;
166 } _case_t;
167
168 static _case_t *upper;
169 static _case_t *lower;
170 static _case_t *title;
171 static ac_uint4 upper_used;
172 static ac_uint4 upper_size;
173 static ac_uint4 lower_used;
174 static ac_uint4 lower_size;
175 static ac_uint4 title_used;
176 static ac_uint4 title_size;
177
178 /*
179 * Array used to collect case mappings before adding them to a list.
180 */
181 static ac_uint4 cases[3];
182
183 /*
184 * An array to hold ranges for combining classes.
185 */
186 static ac_uint4 *ccl;
187 static ac_uint4 ccl_used;
188 static ac_uint4 ccl_size;
189
190 /*
191 * Structures for handling numbers.
192 */
193 typedef struct {
194 ac_uint4 code;
195 ac_uint4 idx;
196 } _codeidx_t;
197
198 typedef struct {
199 short numerator;
200 short denominator;
201 } _num_t;
202
203 /*
204 * Arrays to hold the mapping of codes to numbers.
205 */
206 static _codeidx_t *ncodes;
207 static ac_uint4 ncodes_used;
208 static ac_uint4 ncodes_size;
209
210 static _num_t *nums;
211 static ac_uint4 nums_used;
212 static ac_uint4 nums_size;
213
214 /*
215 * Array for holding numbers.
216 */
217 static _num_t *nums;
218 static ac_uint4 nums_used;
219 static ac_uint4 nums_size;
220
221 static void
222 add_range(ac_uint4 start, ac_uint4 end, char *p1, char *p2)
223 {
224 int i, j, k, len;
225 _ranges_t *rlp;
226 char *name;
227
228 for (k = 0; k < 2; k++) {
229 if (k == 0) {
230 name = p1;
231 len = 2;
232 } else {
233 if (p2 == 0)
234 break;
235
236 name = p2;
237 len = 1;
238 }
239
240 for (i = 0; i < NUMPROPS; i++) {
241 if (props[i].len == len && memcmp(props[i].name, name, len) == 0)
242 break;
243 }
244
245 if (i == NUMPROPS)
246 continue;
247
248 rlp = &proptbl[i];
249
250 /*
251 * Resize the range list if necessary.
252 */
253 if (rlp->used == rlp->size) {
254 if (rlp->size == 0)
255 rlp->ranges = (ac_uint4 *)
256 malloc(sizeof(ac_uint4) << 3);
257 else
258 rlp->ranges = (ac_uint4 *)
259 realloc((char *) rlp->ranges,
260 sizeof(ac_uint4) * (rlp->size + 8));
261 rlp->size += 8;
262 }
263
264 /*
265 * If this is the first code for this property list, just add it
266 * and return.
267 */
268 if (rlp->used == 0) {
269 rlp->ranges[0] = start;
270 rlp->ranges[1] = end;
271 rlp->used += 2;
272 continue;
273 }
274
275 /*
276 * Optimize the case of adding the range to the end.
277 */
278 j = rlp->used - 1;
279 if (start > rlp->ranges[j]) {
280 j = rlp->used;
281 rlp->ranges[j++] = start;
282 rlp->ranges[j++] = end;
283 rlp->used = j;
284 continue;
285 }
286
287 /*
288 * Need to locate the insertion point.
289 */
290 for (i = 0;
291 i < rlp->used && start > rlp->ranges[i + 1] + 1; i += 2) ;
292
293 /*
294 * If the start value lies in the current range, then simply set the
295 * new end point of the range to the end value passed as a parameter.
296 */
297 if (rlp->ranges[i] <= start && start <= rlp->ranges[i + 1] + 1) {
298 rlp->ranges[i + 1] = end;
299 return;
300 }
301
302 /*
303 * Shift following values up by two.
304 */
305 for (j = rlp->used; j > i; j -= 2) {
306 rlp->ranges[j] = rlp->ranges[j - 2];
307 rlp->ranges[j + 1] = rlp->ranges[j - 1];
308 }
309
310 /*
311 * Add the new range at the insertion point.
312 */
313 rlp->ranges[i] = start;
314 rlp->ranges[i + 1] = end;
315 rlp->used += 2;
316 }
317 }
318
319 static void
320 ordered_range_insert(ac_uint4 c, char *name, int len)
321 {
322 int i, j;
323 ac_uint4 s, e;
324 _ranges_t *rlp;
325
326 if (len == 0)
327 return;
328
329 /*
330 * Deal with directionality codes introduced in Unicode 3.0.
331 */
332 if ((len == 2 && memcmp(name, "BN", 2) == 0) ||
333 (len == 3 &&
334 (memcmp(name, "NSM", 3) == 0 || memcmp(name, "PDF", 3) == 0 ||
335 memcmp(name, "LRE", 3) == 0 || memcmp(name, "LRO", 3) == 0 ||
336 memcmp(name, "RLE", 3) == 0 || memcmp(name, "RLO", 3) == 0))) {
337 /*
338 * Mark all of these as Other Neutral to preserve compatibility with
339 * older versions.
340 */
341 len = 2;
342 name = "ON";
343 }
344
345 for (i = 0; i < NUMPROPS; i++) {
346 if (props[i].len == len && memcmp(props[i].name, name, len) == 0)
347 break;
348 }
349
350 if (i == NUMPROPS)
351 return;
352
353 /*
354 * Have a match, so insert the code in order.
355 */
356 rlp = &proptbl[i];
357
358 /*
359 * Resize the range list if necessary.
360 */
361 if (rlp->used == rlp->size) {
362 if (rlp->size == 0)
363 rlp->ranges = (ac_uint4 *)
364 malloc(sizeof(ac_uint4) << 3);
365 else
366 rlp->ranges = (ac_uint4 *)
367 realloc((char *) rlp->ranges,
368 sizeof(ac_uint4) * (rlp->size + 8));
369 rlp->size += 8;
370 }
371
372 /*
373 * If this is the first code for this property list, just add it
374 * and return.
375 */
376 if (rlp->used == 0) {
377 rlp->ranges[0] = rlp->ranges[1] = c;
378 rlp->used += 2;
379 return;
380 }
381
382 /*
383 * Optimize the cases of extending the last range and adding new ranges to
384 * the end.
385 */
386 j = rlp->used - 1;
387 e = rlp->ranges[j];
388 s = rlp->ranges[j - 1];
389
390 if (c == e + 1) {
391 /*
392 * Extend the last range.
393 */
394 rlp->ranges[j] = c;
395 return;
396 }
397
398 if (c > e + 1) {
399 /*
400 * Start another range on the end.
401 */
402 j = rlp->used;
403 rlp->ranges[j] = rlp->ranges[j + 1] = c;
404 rlp->used += 2;
405 return;
406 }
407
408 if (c >= s)
409 /*
410 * The code is a duplicate of a code in the last range, so just return.
411 */
412 return;
413
414 /*
415 * The code should be inserted somewhere before the last range in the
416 * list. Locate the insertion point.
417 */
418 for (i = 0;
419 i < rlp->used && c > rlp->ranges[i + 1] + 1; i += 2) ;
420
421 s = rlp->ranges[i];
422 e = rlp->ranges[i + 1];
423
424 if (c == e + 1)
425 /*
426 * Simply extend the current range.
427 */
428 rlp->ranges[i + 1] = c;
429 else if (c < s) {
430 /*
431 * Add a new entry before the current location. Shift all entries
432 * before the current one up by one to make room.
433 */
434 for (j = rlp->used; j > i; j -= 2) {
435 rlp->ranges[j] = rlp->ranges[j - 2];
436 rlp->ranges[j + 1] = rlp->ranges[j - 1];
437 }
438 rlp->ranges[i] = rlp->ranges[i + 1] = c;
439
440 rlp->used += 2;
441 }
442 }
443
444 static void
445 add_decomp(ac_uint4 code, short compat)
446 {
447 ac_uint4 i, j, size;
448 _decomp_t **pdecomps;
449 ac_uint4 *pdecomps_used;
450 ac_uint4 *pdecomps_size;
451
452 if (compat) {
453 pdecomps = &kdecomps;
454 pdecomps_used = &kdecomps_used;
455 pdecomps_size = &kdecomps_size;
456 } else {
457 pdecomps = &decomps;
458 pdecomps_used = &decomps_used;
459 pdecomps_size = &decomps_size;
460 }
461
462 /*
463 * Add the code to the composite property.
464 */
465 if (!compat) {
466 ordered_range_insert(code, "Cm", 2);
467 }
468
469 /*
470 * Locate the insertion point for the code.
471 */
472 for (i = 0; i < *pdecomps_used && code > (*pdecomps)[i].code; i++) ;
473
474 /*
475 * Allocate space for a new decomposition.
476 */
477 if (*pdecomps_used == *pdecomps_size) {
478 if (*pdecomps_size == 0)
479 *pdecomps = (_decomp_t *) malloc(sizeof(_decomp_t) << 3);
480 else
481 *pdecomps = (_decomp_t *)
482 realloc((char *) *pdecomps,
483 sizeof(_decomp_t) * (*pdecomps_size + 8));
484 (void) memset((char *) (*pdecomps + *pdecomps_size), '\0',
485 sizeof(_decomp_t) << 3);
486 *pdecomps_size += 8;
487 }
488
489 if (i < *pdecomps_used && code != (*pdecomps)[i].code) {
490 /*
491 * Shift the decomps up by one if the codes don't match.
492 */
493 for (j = *pdecomps_used; j > i; j--)
494 (void) AC_MEMCPY((char *) &(*pdecomps)[j], (char *) &(*pdecomps)[j - 1],
495 sizeof(_decomp_t));
496 }
497
498 /*
499 * Insert or replace a decomposition.
500 */
501 size = dectmp_size + (4 - (dectmp_size & 3));
502 if ((*pdecomps)[i].size < size) {
503 if ((*pdecomps)[i].size == 0)
504 (*pdecomps)[i].decomp = (ac_uint4 *)
505 malloc(sizeof(ac_uint4) * size);
506 else
507 (*pdecomps)[i].decomp = (ac_uint4 *)
508 realloc((char *) (*pdecomps)[i].decomp,
509 sizeof(ac_uint4) * size);
510 (*pdecomps)[i].size = size;
511 }
512
513 if ((*pdecomps)[i].code != code)
514 (*pdecomps_used)++;
515
516 (*pdecomps)[i].code = code;
517 (*pdecomps)[i].used = dectmp_size;
518 (void) AC_MEMCPY((char *) (*pdecomps)[i].decomp, (char *) dectmp,
519 sizeof(ac_uint4) * dectmp_size);
520
521 /*
522 * NOTICE: This needs changing later so it is more general than simply
523 * pairs. This calculation is done here to simplify allocation elsewhere.
524 */
525 if (!compat && dectmp_size == 2)
526 comps_used++;
527 }
528
529 static void
530 add_title(ac_uint4 code)
531 {
532 ac_uint4 i, j;
533
534 /*
535 * Always map the code to itself.
536 */
537 cases[2] = code;
538
539 if (title_used == title_size) {
540 if (title_size == 0)
541 title = (_case_t *) malloc(sizeof(_case_t) << 3);
542 else
543 title = (_case_t *) realloc((char *) title,
544 sizeof(_case_t) * (title_size + 8));
545 title_size += 8;
546 }
547
548 /*
549 * Locate the insertion point.
550 */
551 for (i = 0; i < title_used && code > title[i].key; i++) ;
552
553 if (i < title_used) {
554 /*
555 * Shift the array up by one.
556 */
557 for (j = title_used; j > i; j--)
558 (void) AC_MEMCPY((char *) &title[j], (char *) &title[j - 1],
559 sizeof(_case_t));
560 }
561
562 title[i].key = cases[2]; /* Title */
563 title[i].other1 = cases[0]; /* Upper */
564 title[i].other2 = cases[1]; /* Lower */
565
566 title_used++;
567 }
568
569 static void
570 add_upper(ac_uint4 code)
571 {
572 ac_uint4 i, j;
573
574 /*
575 * Always map the code to itself.
576 */
577 cases[0] = code;
578
579 /*
580 * If the title case character is not present, then make it the same as
581 * the upper case.
582 */
583 if (cases[2] == 0)
584 cases[2] = code;
585
586 if (upper_used == upper_size) {
587 if (upper_size == 0)
588 upper = (_case_t *) malloc(sizeof(_case_t) << 3);
589 else
590 upper = (_case_t *) realloc((char *) upper,
591 sizeof(_case_t) * (upper_size + 8));
592 upper_size += 8;
593 }
594
595 /*
596 * Locate the insertion point.
597 */
598 for (i = 0; i < upper_used && code > upper[i].key; i++) ;
599
600 if (i < upper_used) {
601 /*
602 * Shift the array up by one.
603 */
604 for (j = upper_used; j > i; j--)
605 (void) AC_MEMCPY((char *) &upper[j], (char *) &upper[j - 1],
606 sizeof(_case_t));
607 }
608
609 upper[i].key = cases[0]; /* Upper */
610 upper[i].other1 = cases[1]; /* Lower */
611 upper[i].other2 = cases[2]; /* Title */
612
613 upper_used++;
614 }
615
616 static void
617 add_lower(ac_uint4 code)
618 {
619 ac_uint4 i, j;
620
621 /*
622 * Always map the code to itself.
623 */
624 cases[1] = code;
625
626 /*
627 * If the title case character is empty, then make it the same as the
628 * upper case.
629 */
630 if (cases[2] == 0)
631 cases[2] = cases[0];
632
633 if (lower_used == lower_size) {
634 if (lower_size == 0)
635 lower = (_case_t *) malloc(sizeof(_case_t) << 3);
636 else
637 lower = (_case_t *) realloc((char *) lower,
638 sizeof(_case_t) * (lower_size + 8));
639 lower_size += 8;
640 }
641
642 /*
643 * Locate the insertion point.
644 */
645 for (i = 0; i < lower_used && code > lower[i].key; i++) ;
646
647 if (i < lower_used) {
648 /*
649 * Shift the array up by one.
650 */
651 for (j = lower_used; j > i; j--)
652 (void) AC_MEMCPY((char *) &lower[j], (char *) &lower[j - 1],
653 sizeof(_case_t));
654 }
655
656 lower[i].key = cases[1]; /* Lower */
657 lower[i].other1 = cases[0]; /* Upper */
658 lower[i].other2 = cases[2]; /* Title */
659
660 lower_used++;
661 }
662
663 static void
664 ordered_ccl_insert(ac_uint4 c, ac_uint4 ccl_code)
665 {
666 ac_uint4 i, j;
667
668 if (ccl_used == ccl_size) {
669 if (ccl_size == 0)
670 ccl = (ac_uint4 *) malloc(sizeof(ac_uint4) * 24);
671 else
672 ccl = (ac_uint4 *)
673 realloc((char *) ccl, sizeof(ac_uint4) * (ccl_size + 24));
674 ccl_size += 24;
675 }
676
677 /*
678 * Optimize adding the first item.
679 */
680 if (ccl_used == 0) {
681 ccl[0] = ccl[1] = c;
682 ccl[2] = ccl_code;
683 ccl_used += 3;
684 return;
685 }
686
687 /*
688 * Handle the special case of extending the range on the end. This
689 * requires that the combining class codes are the same.
690 */
691 if (ccl_code == ccl[ccl_used - 1] && c == ccl[ccl_used - 2] + 1) {
692 ccl[ccl_used - 2] = c;
693 return;
694 }
695
696 /*
697 * Handle the special case of adding another range on the end.
698 */
699 if (c > ccl[ccl_used - 2] + 1 ||
700 (c == ccl[ccl_used - 2] + 1 && ccl_code != ccl[ccl_used - 1])) {
701 ccl[ccl_used++] = c;
702 ccl[ccl_used++] = c;
703 ccl[ccl_used++] = ccl_code;
704 return;
705 }
706
707 /*
708 * Locate either the insertion point or range for the code.
709 */
710 for (i = 0; i < ccl_used && c > ccl[i + 1] + 1; i += 3) ;
711
712 if (ccl_code == ccl[i + 2] && c == ccl[i + 1] + 1) {
713 /*
714 * Extend an existing range.
715 */
716 ccl[i + 1] = c;
717 return;
718 } else if (c < ccl[i]) {
719 /*
720 * Start a new range before the current location.
721 */
722 for (j = ccl_used; j > i; j -= 3) {
723 ccl[j] = ccl[j - 3];
724 ccl[j - 1] = ccl[j - 4];
725 ccl[j - 2] = ccl[j - 5];
726 }
727 ccl[i] = ccl[i + 1] = c;
728 ccl[i + 2] = ccl_code;
729 }
730 }
731
732 /*
733 * Adds a number if it does not already exist and returns an index value
734 * multiplied by 2.
735 */
736 static ac_uint4
737 make_number(short num, short denom)
738 {
739 ac_uint4 n;
740
741 /*
742 * Determine if the number already exists.
743 */
744 for (n = 0; n < nums_used; n++) {
745 if (nums[n].numerator == num && nums[n].denominator == denom)
746 return n << 1;
747 }
748
749 if (nums_used == nums_size) {
750 if (nums_size == 0)
751 nums = (_num_t *) malloc(sizeof(_num_t) << 3);
752 else
753 nums = (_num_t *) realloc((char *) nums,
754 sizeof(_num_t) * (nums_size + 8));
755 nums_size += 8;
756 }
757
758 n = nums_used++;
759 nums[n].numerator = num;
760 nums[n].denominator = denom;
761
762 return n << 1;
763 }
764
765 static void
766 add_number(ac_uint4 code, short num, short denom)
767 {
768 ac_uint4 i, j;
769
770 /*
771 * Insert the code in order.
772 */
773 for (i = 0; i < ncodes_used && code > ncodes[i].code; i++) ;
774
775 /*
776 * Handle the case of the codes matching and simply replace the number
777 * that was there before.
778 */
779 if (i < ncodes_used && code == ncodes[i].code) {
780 ncodes[i].idx = make_number(num, denom);
781 return;
782 }
783
784 /*
785 * Resize the array if necessary.
786 */
787 if (ncodes_used == ncodes_size) {
788 if (ncodes_size == 0)
789 ncodes = (_codeidx_t *) malloc(sizeof(_codeidx_t) << 3);
790 else
791 ncodes = (_codeidx_t *)
792 realloc((char *) ncodes, sizeof(_codeidx_t) * (ncodes_size + 8));
793
794 ncodes_size += 8;
795 }
796
797 /*
798 * Shift things around to insert the code if necessary.
799 */
800 if (i < ncodes_used) {
801 for (j = ncodes_used; j > i; j--) {
802 ncodes[j].code = ncodes[j - 1].code;
803 ncodes[j].idx = ncodes[j - 1].idx;
804 }
805 }
806 ncodes[i].code = code;
807 ncodes[i].idx = make_number(num, denom);
808
809 ncodes_used++;
810 }
811
812 /*
813 * This routine assumes that the line is a valid Unicode Character Database
814 * entry.
815 */
816 static void
817 read_cdata(FILE *in)
818 {
819 ac_uint4 i, lineno, skip, code, ccl_code;
820 short wnum, neg, number[2], compat;
821 char line[512], *s, *e;
822
823 lineno = skip = 0;
824 while (fgets(line, sizeof(line), in)) {
825 if( (s=strchr(line, '\n')) ) *s = '\0';
826 lineno++;
827
828 /*
829 * Skip blank lines and lines that start with a '#'.
830 */
831 if (line[0] == 0 || line[0] == '#')
832 continue;
833
834 /*
835 * If lines need to be skipped, do it here.
836 */
837 if (skip) {
838 skip--;
839 continue;
840 }
841
842 /*
843 * Collect the code. The code can be up to 6 hex digits in length to
844 * allow surrogates to be specified.
845 */
846 for (s = line, i = code = 0; *s != ';' && i < 6; i++, s++) {
847 code <<= 4;
848 if (*s >= '0' && *s <= '9')
849 code += *s - '0';
850 else if (*s >= 'A' && *s <= 'F')
851 code += (*s - 'A') + 10;
852 else if (*s >= 'a' && *s <= 'f')
853 code += (*s - 'a') + 10;
854 }
855
856 /*
857 * Handle the following special cases:
858 * 1. 4E00-9FA5 CJK Ideographs.
859 * 2. AC00-D7A3 Hangul Syllables.
860 * 3. D800-DFFF Surrogates.
861 * 4. E000-F8FF Private Use Area.
862 * 5. F900-FA2D Han compatibility.
863 * ...Plus additional ranges in newer Unicode versions...
864 */
865 switch (code) {
866 case 0x3400:
867 /* CJK Ideograph Extension A */
868 add_range(0x3400, 0x4db5, "Lo", "L");
869
870 add_range(0x3400, 0x4db5, "Cp", 0);
871
872 skip = 1;
873 break;
874 case 0x4e00:
875 /*
876 * The Han ideographs.
877 */
878 add_range(0x4e00, 0x9fff, "Lo", "L");
879
880 /*
881 * Add the characters to the defined category.
882 */
883 add_range(0x4e00, 0x9fa5, "Cp", 0);
884
885 skip = 1;
886 break;
887 case 0xac00:
888 /*
889 * The Hangul syllables.
890 */
891 add_range(0xac00, 0xd7a3, "Lo", "L");
892
893 /*
894 * Add the characters to the defined category.
895 */
896 add_range(0xac00, 0xd7a3, "Cp", 0);
897
898 skip = 1;
899 break;
900 case 0xd800:
901 /*
902 * Make a range of all surrogates and assume some default
903 * properties.
904 */
905 add_range(0x010000, 0x10ffff, "Cs", "L");
906 skip = 5;
907 break;
908 case 0xe000:
909 /*
910 * The Private Use area. Add with a default set of properties.
911 */
912 add_range(0xe000, 0xf8ff, "Co", "L");
913 skip = 1;
914 break;
915 case 0xf900:
916 /*
917 * The CJK compatibility area.
918 */
919 add_range(0xf900, 0xfaff, "Lo", "L");
920
921 /*
922 * Add the characters to the defined category.
923 */
924 add_range(0xf900, 0xfaff, "Cp", 0);
925
926 skip = 1;
927 break;
928 case 0x20000:
929 /* CJK Ideograph Extension B */
930 add_range(0x20000, 0x2a6d6, "Lo", "L");
931
932 add_range(0x20000, 0x2a6d6, "Cp", 0);
933
934 skip = 1;
935 break;
936 case 0xf0000:
937 /* Plane 15 private use */
938 add_range(0xf0000, 0xffffd, "Co", "L");
939 skip = 1;
940 break;
941
942 case 0x100000:
943 /* Plane 16 private use */
944 add_range(0x100000, 0x10fffd, "Co", "L");
945 skip = 1;
946 break;
947 }
948
949 if (skip)
950 continue;
951
952 /*
953 * Add the code to the defined category.
954 */
955 ordered_range_insert(code, "Cp", 2);
956
957 /*
958 * Locate the first character property field.
959 */
960 for (i = 0; *s != 0 && i < 2; s++) {
961 if (*s == ';')
962 i++;
963 }
964 for (e = s; *e && *e != ';'; e++) ;
965
966 ordered_range_insert(code, s, e - s);
967
968 /*
969 * Locate the combining class code.
970 */
971 for (s = e; *s != 0 && i < 3; s++) {
972 if (*s == ';')
973 i++;
974 }
975
976 /*
977 * Convert the combining class code from decimal.
978 */
979 for (ccl_code = 0, e = s; *e && *e != ';'; e++)
980 ccl_code = (ccl_code * 10) + (*e - '0');
981
982 /*
983 * Add the code if it not 0.
984 */
985 if (ccl_code != 0)
986 ordered_ccl_insert(code, ccl_code);
987
988 /*
989 * Locate the second character property field.
990 */
991 for (s = e; *s != 0 && i < 4; s++) {
992 if (*s == ';')
993 i++;
994 }
995 for (e = s; *e && *e != ';'; e++) ;
996
997 ordered_range_insert(code, s, e - s);
998
999 /*
1000 * Check for a decomposition.
1001 */
1002 s = ++e;
1003 if (*s != ';') {
1004 compat = *s == '<';
1005 if (compat) {
1006 /*
1007 * Skip compatibility formatting tag.
1008 */
1009 while (*s++ != '>');
1010 }
1011 /*
1012 * Collect the codes of the decomposition.
1013 */
1014 for (dectmp_size = 0; *s != ';'; ) {
1015 /*
1016 * Skip all leading non-hex digits.
1017 */
1018 while (!ishdigit(*s))
1019 s++;
1020
1021 for (dectmp[dectmp_size] = 0; ishdigit(*s); s++) {
1022 dectmp[dectmp_size] <<= 4;
1023 if (*s >= '0' && *s <= '9')
1024 dectmp[dectmp_size] += *s - '0';
1025 else if (*s >= 'A' && *s <= 'F')
1026 dectmp[dectmp_size] += (*s - 'A') + 10;
1027 else if (*s >= 'a' && *s <= 'f')
1028 dectmp[dectmp_size] += (*s - 'a') + 10;
1029 }
1030 dectmp_size++;
1031 }
1032
1033 /*
1034 * If there are any codes in the temporary decomposition array,
1035 * then add the character with its decomposition.
1036 */
1037 if (dectmp_size > 0) {
1038 if (!compat) {
1039 add_decomp(code, 0);
1040 }
1041 add_decomp(code, 1);
1042 }
1043 }
1044
1045 /*
1046 * Skip to the number field.
1047 */
1048 for (i = 0; i < 3 && *s; s++) {
1049 if (*s == ';')
1050 i++;
1051 }
1052
1053 /*
1054 * Scan the number in.
1055 */
1056 number[0] = number[1] = 0;
1057 for (e = s, neg = wnum = 0; *e && *e != ';'; e++) {
1058 if (*e == '-') {
1059 neg = 1;
1060 continue;
1061 }
1062
1063 if (*e == '/') {
1064 /*
1065 * Move the the denominator of the fraction.
1066 */
1067 if (neg)
1068 number[wnum] *= -1;
1069 neg = 0;
1070 e++;
1071 wnum++;
1072 }
1073 number[wnum] = (number[wnum] * 10) + (*e - '0');
1074 }
1075
1076 if (e > s) {
1077 /*
1078 * Adjust the denominator in case of integers and add the number.
1079 */
1080 if (wnum == 0)
1081 number[1] = 1;
1082
1083 add_number(code, number[0], number[1]);
1084 }
1085
1086 /*
1087 * Skip to the start of the possible case mappings.
1088 */
1089 for (s = e, i = 0; i < 4 && *s; s++) {
1090 if (*s == ';')
1091 i++;
1092 }
1093
1094 /*
1095 * Collect the case mappings.
1096 */
1097 cases[0] = cases[1] = cases[2] = 0;
1098 for (i = 0; i < 3; i++) {
1099 while (ishdigit(*s)) {
1100 cases[i] <<= 4;
1101 if (*s >= '0' && *s <= '9')
1102 cases[i] += *s - '0';
1103 else if (*s >= 'A' && *s <= 'F')
1104 cases[i] += (*s - 'A') + 10;
1105 else if (*s >= 'a' && *s <= 'f')
1106 cases[i] += (*s - 'a') + 10;
1107 s++;
1108 }
1109 if (*s == ';')
1110 s++;
1111 }
1112 if (cases[0] && cases[1])
1113 /*
1114 * Add the upper and lower mappings for a title case character.
1115 */
1116 add_title(code);
1117 else if (cases[1])
1118 /*
1119 * Add the lower and title case mappings for the upper case
1120 * character.
1121 */
1122 add_upper(code);
1123 else if (cases[0])
1124 /*
1125 * Add the upper and title case mappings for the lower case
1126 * character.
1127 */
1128 add_lower(code);
1129 }
1130 }
1131
1132 static _decomp_t *
1133 find_decomp(ac_uint4 code, short compat)
1134 {
1135 long l, r, m;
1136 _decomp_t *decs;
1137
1138 l = 0;
1139 r = (compat ? kdecomps_used : decomps_used) - 1;
1140 decs = compat ? kdecomps : decomps;
1141 while (l <= r) {
1142 m = (l + r) >> 1;
1143 if (code > decs[m].code)
1144 l = m + 1;
1145 else if (code < decs[m].code)
1146 r = m - 1;
1147 else
1148 return &decs[m];
1149 }
1150 return 0;
1151 }
1152
1153 static void
1154 decomp_it(_decomp_t *d, short compat)
1155 {
1156 ac_uint4 i;
1157 _decomp_t *dp;
1158
1159 for (i = 0; i < d->used; i++) {
1160 if ((dp = find_decomp(d->decomp[i], compat)) != 0)
1161 decomp_it(dp, compat);
1162 else
1163 dectmp[dectmp_size++] = d->decomp[i];
1164 }
1165 }
1166
1167 /*
1168 * Expand all decompositions by recursively decomposing each character
1169 * in the decomposition.
1170 */
1171 static void
1172 expand_decomp(void)
1173 {
1174 ac_uint4 i;
1175
1176 for (i = 0; i < decomps_used; i++) {
1177 dectmp_size = 0;
1178 decomp_it(&decomps[i], 0);
1179 if (dectmp_size > 0)
1180 add_decomp(decomps[i].code, 0);
1181 }
1182
1183 for (i = 0; i < kdecomps_used; i++) {
1184 dectmp_size = 0;
1185 decomp_it(&kdecomps[i], 1);
1186 if (dectmp_size > 0)
1187 add_decomp(kdecomps[i].code, 1);
1188 }
1189 }
1190
1191 static int
1192 cmpcomps(const void *v_comp1, const void *v_comp2)
1193 {
1194 const _comp_t *comp1 = v_comp1, *comp2 = v_comp2;
1195 long diff = comp1->code1 - comp2->code1;
1196
1197 if (!diff)
1198 diff = comp1->code2 - comp2->code2;
1199 return (int) diff;
1200 }
1201
1202 /*
1203 * Load composition exclusion data
1204 */
1205 static void
1206 read_compexdata(FILE *in)
1207 {
1208 ac_uint2 i;
1209 ac_uint4 code;
1210 char line[512], *s;
1211
1212 (void) memset((char *) compexs, 0, sizeof(compexs));
1213
1214 while (fgets(line, sizeof(line), in)) {
1215 if( (s=strchr(line, '\n')) ) *s = '\0';
1216 /*
1217 * Skip blank lines and lines that start with a '#'.
1218 */
1219 if (line[0] == 0 || line[0] == '#')
1220 continue;
1221
1222 /*
1223 * Collect the code. Assume max 6 digits
1224 */
1225
1226 for (s = line, i = code = 0; *s != '#' && i < 6; i++, s++) {
1227 if (isspace((unsigned char)*s)) break;
1228 code <<= 4;
1229 if (*s >= '0' && *s <= '9')
1230 code += *s - '0';
1231 else if (*s >= 'A' && *s <= 'F')
1232 code += (*s - 'A') + 10;
1233 else if (*s >= 'a' && *s <= 'f')
1234 code += (*s - 'a') + 10;
1235 }
1236 COMPEX_SET(code);
1237 }
1238 }
1239
1240 /*
1241 * Creates array of compositions from decomposition array
1242 */
1243 static void
1244 create_comps(void)
1245 {
1246 ac_uint4 i, cu;
1247
1248 comps = (_comp_t *) malloc(comps_used * sizeof(_comp_t));
1249
1250 for (i = cu = 0; i < decomps_used; i++) {
1251 if (decomps[i].used != 2 || COMPEX_TEST(decomps[i].code))
1252 continue;
1253 comps[cu].comp = decomps[i].code;
1254 comps[cu].count = 2;
1255 comps[cu].code1 = decomps[i].decomp[0];
1256 comps[cu].code2 = decomps[i].decomp[1];
1257 cu++;
1258 }
1259 comps_used = cu;
1260 qsort(comps, comps_used, sizeof(_comp_t), cmpcomps);
1261 }
1262
1263 #if HARDCODE_DATA
1264 static void
1265 write_case(FILE *out, _case_t *tab, int num, int first)
1266 {
1267 int i;
1268
1269 for (i=0; i<num; i++) {
1270 if (first) first = 0;
1271 else fprintf(out, ",");
1272 fprintf(out, "\n\t0x%08lx, 0x%08lx, 0x%08lx",
1273 (unsigned long) tab[i].key, (unsigned long) tab[i].other1,
1274 (unsigned long) tab[i].other2);
1275 }
1276 }
1277
1278 #define PREF "static const "
1279
1280 #endif
1281
1282 static void
1283 write_cdata(char *opath)
1284 {
1285 FILE *out;
1286 ac_uint4 bytes;
1287 ac_uint4 i, idx, nprops;
1288 #if !(HARDCODE_DATA)
1289 ac_uint2 casecnt[2];
1290 #endif
1291 char path[BUFSIZ];
1292 #if HARDCODE_DATA
1293 int j, k;
1294
1295 /*****************************************************************
1296 *
1297 * Generate the ctype data.
1298 *
1299 *****************************************************************/
1300
1301 /*
1302 * Open the output file.
1303 */
1304 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "uctable.h", opath);
1305 if ((out = fopen(path, "w")) == 0)
1306 return;
1307 #else
1308 /*
1309 * Open the ctype.dat file.
1310 */
1311 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "ctype.dat", opath);
1312 if ((out = fopen(path, "wb")) == 0)
1313 return;
1314 #endif
1315
1316 /*
1317 * Collect the offsets for the properties. The offsets array is
1318 * on a 4-byte boundary to keep things efficient for architectures
1319 * that need such a thing.
1320 */
1321 for (i = idx = 0; i < NUMPROPS; i++) {
1322 propcnt[i] = (proptbl[i].used != 0) ? idx : 0xffff;
1323 idx += proptbl[i].used;
1324 }
1325
1326 /*
1327 * Add the sentinel index which is used by the binary search as the upper
1328 * bound for a search.
1329 */
1330 propcnt[i] = idx;
1331
1332 /*
1333 * Record the actual number of property lists. This may be different than
1334 * the number of offsets actually written because of aligning on a 4-byte
1335 * boundary.
1336 */
1337 hdr[1] = NUMPROPS;
1338
1339 /*
1340 * Calculate the byte count needed and pad the property counts array to a
1341 * 4-byte boundary.
1342 */
1343 if ((bytes = sizeof(ac_uint2) * (NUMPROPS + 1)) & 3)
1344 bytes += 4 - (bytes & 3);
1345 nprops = bytes / sizeof(ac_uint2);
1346 bytes += sizeof(ac_uint4) * idx;
1347
1348 #if HARDCODE_DATA
1349 fprintf(out, PREF "ac_uint4 _ucprop_size = %d;\n\n", NUMPROPS);
1350
1351 fprintf(out, PREF "ac_uint2 _ucprop_offsets[] = {");
1352
1353 for (i = 0; i<nprops; i++) {
1354 if (i) fprintf(out, ",");
1355 if (!(i&7)) fprintf(out, "\n\t");
1356 else fprintf(out, " ");
1357 fprintf(out, "0x%04x", propcnt[i]);
1358 }
1359 fprintf(out, "\n};\n\n");
1360
1361 fprintf(out, PREF "ac_uint4 _ucprop_ranges[] = {");
1362
1363 k = 0;
1364 for (i = 0; i < NUMPROPS; i++) {
1365 if (proptbl[i].used > 0) {
1366 for (j=0; j<proptbl[i].used; j++) {
1367 if (k) fprintf(out, ",");
1368 if (!(k&3)) fprintf(out,"\n\t");
1369 else fprintf(out, " ");
1370 k++;
1371 fprintf(out, "0x%08lx", (unsigned long) proptbl[i].ranges[j]);
1372 }
1373 }
1374 }
1375 fprintf(out, "\n};\n\n");
1376 #else
1377 /*
1378 * Write the header.
1379 */
1380 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1381
1382 /*
1383 * Write the byte count.
1384 */
1385 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1386
1387 /*
1388 * Write the property list counts.
1389 */
1390 fwrite((char *) propcnt, sizeof(ac_uint2), nprops, out);
1391
1392 /*
1393 * Write the property lists.
1394 */
1395 for (i = 0; i < NUMPROPS; i++) {
1396 if (proptbl[i].used > 0)
1397 fwrite((char *) proptbl[i].ranges, sizeof(ac_uint4),
1398 proptbl[i].used, out);
1399 }
1400
1401 fclose(out);
1402 #endif
1403
1404 /*****************************************************************
1405 *
1406 * Generate the case mapping data.
1407 *
1408 *****************************************************************/
1409
1410 #if HARDCODE_DATA
1411 fprintf(out, PREF "ac_uint4 _uccase_size = %ld;\n\n",
1412 (long) (upper_used + lower_used + title_used));
1413
1414 fprintf(out, PREF "ac_uint2 _uccase_len[2] = {%ld, %ld};\n\n",
1415 (long) upper_used, (long) lower_used);
1416 fprintf(out, PREF "ac_uint4 _uccase_map[] = {");
1417
1418 if (upper_used > 0)
1419 /*
1420 * Write the upper case table.
1421 */
1422 write_case(out, upper, upper_used, 1);
1423
1424 if (lower_used > 0)
1425 /*
1426 * Write the lower case table.
1427 */
1428 write_case(out, lower, lower_used, !upper_used);
1429
1430 if (title_used > 0)
1431 /*
1432 * Write the title case table.
1433 */
1434 write_case(out, title, title_used, !(upper_used||lower_used));
1435
1436 if (!(upper_used || lower_used || title_used))
1437 fprintf(out, "\t0");
1438
1439 fprintf(out, "\n};\n\n");
1440 #else
1441 /*
1442 * Open the case.dat file.
1443 */
1444 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "case.dat", opath);
1445 if ((out = fopen(path, "wb")) == 0)
1446 return;
1447
1448 /*
1449 * Write the case mapping tables.
1450 */
1451 hdr[1] = upper_used + lower_used + title_used;
1452 casecnt[0] = upper_used;
1453 casecnt[1] = lower_used;
1454
1455 /*
1456 * Write the header.
1457 */
1458 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1459
1460 /*
1461 * Write the upper and lower case table sizes.
1462 */
1463 fwrite((char *) casecnt, sizeof(ac_uint2), 2, out);
1464
1465 if (upper_used > 0)
1466 /*
1467 * Write the upper case table.
1468 */
1469 fwrite((char *) upper, sizeof(_case_t), upper_used, out);
1470
1471 if (lower_used > 0)
1472 /*
1473 * Write the lower case table.
1474 */
1475 fwrite((char *) lower, sizeof(_case_t), lower_used, out);
1476
1477 if (title_used > 0)
1478 /*
1479 * Write the title case table.
1480 */
1481 fwrite((char *) title, sizeof(_case_t), title_used, out);
1482
1483 fclose(out);
1484 #endif
1485
1486 /*****************************************************************
1487 *
1488 * Generate the composition data.
1489 *
1490 *****************************************************************/
1491
1492 /*
1493 * Create compositions from decomposition data
1494 */
1495 create_comps();
1496
1497 #if HARDCODE_DATA
1498 fprintf(out, PREF "ac_uint4 _uccomp_size = %ld;\n\n",
1499 comps_used * 4L);
1500
1501 fprintf(out, PREF "ac_uint4 _uccomp_data[] = {");
1502
1503 /*
1504 * Now, if comps exist, write them out.
1505 */
1506 if (comps_used > 0) {
1507 for (i=0; i<comps_used; i++) {
1508 if (i) fprintf(out, ",");
1509 fprintf(out, "\n\t0x%08lx, 0x%08lx, 0x%08lx, 0x%08lx",
1510 (unsigned long) comps[i].comp, (unsigned long) comps[i].count,
1511 (unsigned long) comps[i].code1, (unsigned long) comps[i].code2);
1512 }
1513 } else {
1514 fprintf(out, "\t0");
1515 }
1516 fprintf(out, "\n};\n\n");
1517 #else
1518 /*
1519 * Open the comp.dat file.
1520 */
1521 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "comp.dat", opath);
1522 if ((out = fopen(path, "wb")) == 0)
1523 return;
1524
1525 /*
1526 * Write the header.
1527 */
1528 hdr[1] = (ac_uint2) comps_used * 4;
1529 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1530
1531 /*
1532 * Write out the byte count to maintain header size.
1533 */
1534 bytes = comps_used * sizeof(_comp_t);
1535 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1536
1537 /*
1538 * Now, if comps exist, write them out.
1539 */
1540 if (comps_used > 0)
1541 fwrite((char *) comps, sizeof(_comp_t), comps_used, out);
1542
1543 fclose(out);
1544 #endif
1545
1546 /*****************************************************************
1547 *
1548 * Generate the decomposition data.
1549 *
1550 *****************************************************************/
1551
1552 /*
1553 * Fully expand all decompositions before generating the output file.
1554 */
1555 expand_decomp();
1556
1557 #if HARDCODE_DATA
1558 fprintf(out, PREF "ac_uint4 _ucdcmp_size = %ld;\n\n",
1559 decomps_used * 2L);
1560
1561 fprintf(out, PREF "ac_uint4 _ucdcmp_nodes[] = {");
1562
1563 if (decomps_used) {
1564 /*
1565 * Write the list of decomp nodes.
1566 */
1567 for (i = idx = 0; i < decomps_used; i++) {
1568 fprintf(out, "\n\t0x%08lx, 0x%08lx,",
1569 (unsigned long) decomps[i].code, (unsigned long) idx);
1570 idx += decomps[i].used;
1571 }
1572
1573 /*
1574 * Write the sentinel index as the last decomp node.
1575 */
1576 fprintf(out, "\n\t0x%08lx\n};\n\n", (unsigned long) idx);
1577
1578 fprintf(out, PREF "ac_uint4 _ucdcmp_decomp[] = {");
1579 /*
1580 * Write the decompositions themselves.
1581 */
1582 k = 0;
1583 for (i = 0; i < decomps_used; i++)
1584 for (j=0; j<decomps[i].used; j++) {
1585 if (k) fprintf(out, ",");
1586 if (!(k&3)) fprintf(out,"\n\t");
1587 else fprintf(out, " ");
1588 k++;
1589 fprintf(out, "0x%08lx", (unsigned long) decomps[i].decomp[j]);
1590 }
1591 fprintf(out, "\n};\n\n");
1592 }
1593 #else
1594 /*
1595 * Open the decomp.dat file.
1596 */
1597 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "decomp.dat", opath);
1598 if ((out = fopen(path, "wb")) == 0)
1599 return;
1600
1601 hdr[1] = decomps_used;
1602
1603 /*
1604 * Write the header.
1605 */
1606 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1607
1608 /*
1609 * Write a temporary byte count which will be calculated as the
1610 * decompositions are written out.
1611 */
1612 bytes = 0;
1613 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1614
1615 if (decomps_used) {
1616 /*
1617 * Write the list of decomp nodes.
1618 */
1619 for (i = idx = 0; i < decomps_used; i++) {
1620 fwrite((char *) &decomps[i].code, sizeof(ac_uint4), 1, out);
1621 fwrite((char *) &idx, sizeof(ac_uint4), 1, out);
1622 idx += decomps[i].used;
1623 }
1624
1625 /*
1626 * Write the sentinel index as the last decomp node.
1627 */
1628 fwrite((char *) &idx, sizeof(ac_uint4), 1, out);
1629
1630 /*
1631 * Write the decompositions themselves.
1632 */
1633 for (i = 0; i < decomps_used; i++)
1634 fwrite((char *) decomps[i].decomp, sizeof(ac_uint4),
1635 decomps[i].used, out);
1636
1637 /*
1638 * Seek back to the beginning and write the byte count.
1639 */
1640 bytes = (sizeof(ac_uint4) * idx) +
1641 (sizeof(ac_uint4) * ((hdr[1] << 1) + 1));
1642 fseek(out, sizeof(ac_uint2) << 1, 0L);
1643 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1644
1645 fclose(out);
1646 }
1647 #endif
1648
1649 #ifdef HARDCODE_DATA
1650 fprintf(out, PREF "ac_uint4 _uckdcmp_size = %ld;\n\n",
1651 kdecomps_used * 2L);
1652
1653 fprintf(out, PREF "ac_uint4 _uckdcmp_nodes[] = {");
1654
1655 if (kdecomps_used) {
1656 /*
1657 * Write the list of kdecomp nodes.
1658 */
1659 for (i = idx = 0; i < kdecomps_used; i++) {
1660 fprintf(out, "\n\t0x%08lx, 0x%08lx,",
1661 (unsigned long) kdecomps[i].code, (unsigned long) idx);
1662 idx += kdecomps[i].used;
1663 }
1664
1665 /*
1666 * Write the sentinel index as the last decomp node.
1667 */
1668 fprintf(out, "\n\t0x%08lx\n};\n\n", (unsigned long) idx);
1669
1670 fprintf(out, PREF "ac_uint4 _uckdcmp_decomp[] = {");
1671
1672 /*
1673 * Write the decompositions themselves.
1674 */
1675 k = 0;
1676 for (i = 0; i < kdecomps_used; i++)
1677 for (j=0; j<kdecomps[i].used; j++) {
1678 if (k) fprintf(out, ",");
1679 if (!(k&3)) fprintf(out,"\n\t");
1680 else fprintf(out, " ");
1681 k++;
1682 fprintf(out, "0x%08lx", (unsigned long) kdecomps[i].decomp[j]);
1683 }
1684 fprintf(out, "\n};\n\n");
1685 }
1686 #else
1687 /*
1688 * Open the kdecomp.dat file.
1689 */
1690 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "kdecomp.dat", opath);
1691 if ((out = fopen(path, "wb")) == 0)
1692 return;
1693
1694 hdr[1] = kdecomps_used;
1695
1696 /*
1697 * Write the header.
1698 */
1699 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1700
1701 /*
1702 * Write a temporary byte count which will be calculated as the
1703 * decompositions are written out.
1704 */
1705 bytes = 0;
1706 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1707
1708 if (kdecomps_used) {
1709 /*
1710 * Write the list of kdecomp nodes.
1711 */
1712 for (i = idx = 0; i < kdecomps_used; i++) {
1713 fwrite((char *) &kdecomps[i].code, sizeof(ac_uint4), 1, out);
1714 fwrite((char *) &idx, sizeof(ac_uint4), 1, out);
1715 idx += kdecomps[i].used;
1716 }
1717
1718 /*
1719 * Write the sentinel index as the last decomp node.
1720 */
1721 fwrite((char *) &idx, sizeof(ac_uint4), 1, out);
1722
1723 /*
1724 * Write the decompositions themselves.
1725 */
1726 for (i = 0; i < kdecomps_used; i++)
1727 fwrite((char *) kdecomps[i].decomp, sizeof(ac_uint4),
1728 kdecomps[i].used, out);
1729
1730 /*
1731 * Seek back to the beginning and write the byte count.
1732 */
1733 bytes = (sizeof(ac_uint4) * idx) +
1734 (sizeof(ac_uint4) * ((hdr[1] << 1) + 1));
1735 fseek(out, sizeof(ac_uint2) << 1, 0L);
1736 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1737
1738 fclose(out);
1739 }
1740 #endif
1741
1742 /*****************************************************************
1743 *
1744 * Generate the combining class data.
1745 *
1746 *****************************************************************/
1747 #ifdef HARDCODE_DATA
1748 fprintf(out, PREF "ac_uint4 _uccmcl_size = %ld;\n\n", (long) ccl_used);
1749
1750 fprintf(out, PREF "ac_uint4 _uccmcl_nodes[] = {");
1751
1752 if (ccl_used > 0) {
1753 /*
1754 * Write the combining class ranges out.
1755 */
1756 for (i = 0; i<ccl_used; i++) {
1757 if (i) fprintf(out, ",");
1758 if (!(i&3)) fprintf(out, "\n\t");
1759 else fprintf(out, " ");
1760 fprintf(out, "0x%08lx", (unsigned long) ccl[i]);
1761 }
1762 } else {
1763 fprintf(out, "\t0");
1764 }
1765 fprintf(out, "\n};\n\n");
1766 #else
1767 /*
1768 * Open the cmbcl.dat file.
1769 */
1770 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "cmbcl.dat", opath);
1771 if ((out = fopen(path, "wb")) == 0)
1772 return;
1773
1774 /*
1775 * Set the number of ranges used. Each range has a combining class which
1776 * means each entry is a 3-tuple.
1777 */
1778 hdr[1] = ccl_used / 3;
1779
1780 /*
1781 * Write the header.
1782 */
1783 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1784
1785 /*
1786 * Write out the byte count to maintain header size.
1787 */
1788 bytes = ccl_used * sizeof(ac_uint4);
1789 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1790
1791 if (ccl_used > 0)
1792 /*
1793 * Write the combining class ranges out.
1794 */
1795 fwrite((char *) ccl, sizeof(ac_uint4), ccl_used, out);
1796
1797 fclose(out);
1798 #endif
1799
1800 /*****************************************************************
1801 *
1802 * Generate the number data.
1803 *
1804 *****************************************************************/
1805
1806 #if HARDCODE_DATA
1807 fprintf(out, PREF "ac_uint4 _ucnum_size = %lu;\n\n",
1808 (unsigned long)ncodes_used<<1);
1809
1810 fprintf(out, PREF "ac_uint4 _ucnum_nodes[] = {");
1811
1812 /*
1813 * Now, if number mappings exist, write them out.
1814 */
1815 if (ncodes_used > 0) {
1816 for (i = 0; i<ncodes_used; i++) {
1817 if (i) fprintf(out, ",");
1818 if (!(i&1)) fprintf(out, "\n\t");
1819 else fprintf(out, " ");
1820 fprintf(out, "0x%08lx, 0x%08lx",
1821 (unsigned long) ncodes[i].code, (unsigned long) ncodes[i].idx);
1822 }
1823 fprintf(out, "\n};\n\n");
1824
1825 fprintf(out, PREF "short _ucnum_vals[] = {");
1826 for (i = 0; i<nums_used; i++) {
1827 if (i) fprintf(out, ",");
1828 if (!(i&3)) fprintf(out, "\n\t");
1829 else fprintf(out, " ");
1830 if (nums[i].numerator < 0) {
1831 fprintf(out, "%6d, 0x%04x",
1832 nums[i].numerator, nums[i].denominator);
1833 } else {
1834 fprintf(out, "0x%04x, 0x%04x",
1835 nums[i].numerator, nums[i].denominator);
1836 }
1837 }
1838 fprintf(out, "\n};\n\n");
1839 }
1840 #else
1841 /*
1842 * Open the num.dat file.
1843 */
1844 snprintf(path, sizeof path, "%s" LDAP_DIRSEP "num.dat", opath);
1845 if ((out = fopen(path, "wb")) == 0)
1846 return;
1847
1848 /*
1849 * The count part of the header will be the total number of codes that
1850 * have numbers.
1851 */
1852 hdr[1] = (ac_uint2) (ncodes_used << 1);
1853 bytes = (ncodes_used * sizeof(_codeidx_t)) + (nums_used * sizeof(_num_t));
1854
1855 /*
1856 * Write the header.
1857 */
1858 fwrite((char *) hdr, sizeof(ac_uint2), 2, out);
1859
1860 /*
1861 * Write out the byte count to maintain header size.
1862 */
1863 fwrite((char *) &bytes, sizeof(ac_uint4), 1, out);
1864
1865 /*
1866 * Now, if number mappings exist, write them out.
1867 */
1868 if (ncodes_used > 0) {
1869 fwrite((char *) ncodes, sizeof(_codeidx_t), ncodes_used, out);
1870 fwrite((char *) nums, sizeof(_num_t), nums_used, out);
1871 }
1872 #endif
1873
1874 fclose(out);
1875 }
1876
1877 static void
1878 usage(char *prog)
1879 {
1880 fprintf(stderr,
1881 "Usage: %s [-o output-directory|-x composition-exclusions]", prog);
1882 fprintf(stderr, " datafile1 datafile2 ...\n\n");
1883 fprintf(stderr,
1884 "-o output-directory\n\t\tWrite the output files to a different");
1885 fprintf(stderr, " directory (default: .).\n");
1886 fprintf(stderr,
1887 "-x composition-exclusion\n\t\tFile of composition codes");
1888 fprintf(stderr, " that should be excluded.\n");
1889 exit(1);
1890 }
1891
1892 int
1893 main(int argc, char *argv[])
1894 {
1895 FILE *in;
1896 char *prog, *opath;
1897
1898 prog = lutil_progname( "ucgendat", argc, argv );
1899
1900 opath = 0;
1901 in = stdin;
1902
1903 argc--;
1904 argv++;
1905
1906 while (argc > 0) {
1907 if (argv[0][0] == '-') {
1908 switch (argv[0][1]) {
1909 case 'o':
1910 argc--;
1911 argv++;
1912 opath = argv[0];
1913 break;
1914 case 'x':
1915 argc--;
1916 argv++;
1917 if ((in = fopen(argv[0], "r")) == 0)
1918 fprintf(stderr,
1919 "%s: unable to open composition exclusion file %s\n",
1920 prog, argv[0]);
1921 else {
1922 read_compexdata(in);
1923 fclose(in);
1924 in = 0;
1925 }
1926 break;
1927 default:
1928 usage(prog);
1929 }
1930 } else {
1931 if (in != stdin && in != NULL)
1932 fclose(in);
1933 if ((in = fopen(argv[0], "r")) == 0)
1934 fprintf(stderr, "%s: unable to open ctype file %s\n",
1935 prog, argv[0]);
1936 else {
1937 read_cdata(in);
1938 fclose(in);
1939 in = 0;
1940 }
1941 }
1942 argc--;
1943 argv++;
1944 }
1945
1946 if (opath == 0)
1947 opath = ".";
1948 write_cdata(opath);
1949
1950 return 0;
1951 }
1952