xref: /freebsd/usr.bin/localedef/ctype.c (revision 6829dae12bb055451fa467da4589c43bd03b1e64)
1 /*-
2  * Copyright 2018 Nexenta Systems, Inc.
3  * Copyright 2012 Garrett D'Amore <garrett@damore.org>  All rights reserved.
4  * Copyright 2015 John Marino <draco@marino.st>
5  *
6  * This source code is derived from the illumos localedef command, and
7  * provided under BSD-style license terms by Nexenta Systems, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
20  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
23  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * LC_CTYPE database generation routines for localedef.
34  */
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include <sys/tree.h>
39 
40 #include <stdio.h>
41 #include <stdlib.h>
42 #include <stddef.h>
43 #include <string.h>
44 #include <sys/types.h>
45 #include <wchar.h>
46 #include <ctype.h>
47 #include <wctype.h>
48 #include <unistd.h>
49 #include "localedef.h"
50 #include "parser.h"
51 #include "runefile.h"
52 
53 
54 /* Needed for bootstrapping, _CTYPE_N */
55 #ifndef _CTYPE_N
56 #define _CTYPE_N       0x00400000L
57 #endif
58 
59 #define _ISUPPER	_CTYPE_U
60 #define _ISLOWER	_CTYPE_L
61 #define	_ISDIGIT	_CTYPE_D
62 #define	_ISXDIGIT	_CTYPE_X
63 #define	_ISSPACE	_CTYPE_S
64 #define	_ISBLANK	_CTYPE_B
65 #define	_ISALPHA	_CTYPE_A
66 #define	_ISPUNCT	_CTYPE_P
67 #define	_ISGRAPH	_CTYPE_G
68 #define	_ISPRINT	_CTYPE_R
69 #define	_ISCNTRL	_CTYPE_C
70 #define	_E1		_CTYPE_Q
71 #define	_E2		_CTYPE_I
72 #define	_E3		0
73 #define	_E4		_CTYPE_N
74 #define	_E5		_CTYPE_T
75 
76 static wchar_t		last_ctype;
77 static int ctype_compare(const void *n1, const void *n2);
78 
79 typedef struct ctype_node {
80 	wchar_t wc;
81 	int32_t	ctype;
82 	int32_t	toupper;
83 	int32_t	tolower;
84 	RB_ENTRY(ctype_node) entry;
85 } ctype_node_t;
86 
87 static RB_HEAD(ctypes, ctype_node) ctypes;
88 RB_GENERATE_STATIC(ctypes, ctype_node, entry, ctype_compare);
89 
90 static int
91 ctype_compare(const void *n1, const void *n2)
92 {
93 	const ctype_node_t *c1 = n1;
94 	const ctype_node_t *c2 = n2;
95 
96 	return (c1->wc < c2->wc ? -1 : c1->wc > c2->wc ? 1 : 0);
97 }
98 
99 void
100 init_ctype(void)
101 {
102 	RB_INIT(&ctypes);
103 }
104 
105 
106 static void
107 add_ctype_impl(ctype_node_t *ctn)
108 {
109 	switch (last_kw) {
110 	case T_ISUPPER:
111 		ctn->ctype |= (_ISUPPER | _ISALPHA | _ISGRAPH | _ISPRINT);
112 		break;
113 	case T_ISLOWER:
114 		ctn->ctype |= (_ISLOWER | _ISALPHA | _ISGRAPH | _ISPRINT);
115 		break;
116 	case T_ISALPHA:
117 		ctn->ctype |= (_ISALPHA | _ISGRAPH | _ISPRINT);
118 		break;
119 	case T_ISDIGIT:
120 		ctn->ctype |= (_ISDIGIT | _ISGRAPH | _ISPRINT | _ISXDIGIT | _E4);
121 		break;
122 	case T_ISSPACE:
123 		/*
124 		 * This can be troublesome as <form-feed>, <newline>,
125 		 * <carriage-return>, <tab>, and <vertical-tab> are defined both
126 		 * as space and cntrl, and POSIX doesn't allow cntrl/print
127 		 * combination.  We will take care of this in dump_ctype().
128 		 */
129 		ctn->ctype |= (_ISSPACE | _ISPRINT);
130 		break;
131 	case T_ISCNTRL:
132 		ctn->ctype |= _ISCNTRL;
133 		break;
134 	case T_ISGRAPH:
135 		ctn->ctype |= (_ISGRAPH | _ISPRINT);
136 		break;
137 	case T_ISPRINT:
138 		ctn->ctype |= _ISPRINT;
139 		break;
140 	case T_ISPUNCT:
141 		ctn->ctype |= (_ISPUNCT | _ISGRAPH | _ISPRINT);
142 		break;
143 	case T_ISXDIGIT:
144 		ctn->ctype |= (_ISXDIGIT | _ISPRINT);
145 		break;
146 	case T_ISBLANK:
147 		ctn->ctype |= (_ISBLANK | _ISSPACE);
148 		break;
149 	case T_ISPHONOGRAM:
150 		ctn->ctype |= (_E1 | _ISPRINT | _ISGRAPH);
151 		break;
152 	case T_ISIDEOGRAM:
153 		ctn->ctype |= (_E2 | _ISPRINT | _ISGRAPH);
154 		break;
155 	case T_ISENGLISH:
156 		ctn->ctype |= (_E3 | _ISPRINT | _ISGRAPH);
157 		break;
158 	case T_ISNUMBER:
159 		ctn->ctype |= (_E4 | _ISPRINT | _ISGRAPH);
160 		break;
161 	case T_ISSPECIAL:
162 		ctn->ctype |= (_E5 | _ISPRINT | _ISGRAPH);
163 		break;
164 	case T_ISALNUM:
165 		/*
166 		 * We can't do anything with this.  The character
167 		 * should already be specified as a digit or alpha.
168 		 */
169 		break;
170 	default:
171 		errf("not a valid character class");
172 	}
173 }
174 
175 static ctype_node_t *
176 get_ctype(wchar_t wc)
177 {
178 	ctype_node_t	srch;
179 	ctype_node_t	*ctn;
180 
181 	srch.wc = wc;
182 	if ((ctn = RB_FIND(ctypes, &ctypes, &srch)) == NULL) {
183 		if ((ctn = calloc(1, sizeof (*ctn))) == NULL) {
184 			errf("out of memory");
185 			return (NULL);
186 		}
187 		ctn->wc = wc;
188 
189 		RB_INSERT(ctypes, &ctypes, ctn);
190 	}
191 	return (ctn);
192 }
193 
194 void
195 add_ctype(int val)
196 {
197 	ctype_node_t	*ctn;
198 
199 	if ((ctn = get_ctype(val)) == NULL) {
200 		INTERR;
201 		return;
202 	}
203 	add_ctype_impl(ctn);
204 	last_ctype = ctn->wc;
205 }
206 
207 void
208 add_ctype_range(wchar_t end)
209 {
210 	ctype_node_t	*ctn;
211 	wchar_t		cur;
212 
213 	if (end < last_ctype) {
214 		errf("malformed character range (%u ... %u))",
215 		    last_ctype, end);
216 		return;
217 	}
218 	for (cur = last_ctype + 1; cur <= end; cur++) {
219 		if ((ctn = get_ctype(cur)) == NULL) {
220 			INTERR;
221 			return;
222 		}
223 		add_ctype_impl(ctn);
224 	}
225 	last_ctype = end;
226 
227 }
228 
229 /*
230  * A word about widths: if the width mask is specified, then libc
231  * unconditionally honors it.  Otherwise, it assumes printable
232  * characters have width 1, and non-printable characters have width
233  * -1 (except for NULL which is special with width 0).  Hence, we have
234  * no need to inject defaults here -- the "default" unset value of 0
235  * indicates that libc should use its own logic in wcwidth as described.
236  */
237 void
238 add_width(int wc, int width)
239 {
240 	ctype_node_t	*ctn;
241 
242 	if ((ctn = get_ctype(wc)) == NULL) {
243 		INTERR;
244 		return;
245 	}
246 	ctn->ctype &= ~(_CTYPE_SWM);
247 	switch (width) {
248 	case 0:
249 		ctn->ctype |= _CTYPE_SW0;
250 		break;
251 	case 1:
252 		ctn->ctype |= _CTYPE_SW1;
253 		break;
254 	case 2:
255 		ctn->ctype |= _CTYPE_SW2;
256 		break;
257 	case 3:
258 		ctn->ctype |= _CTYPE_SW3;
259 		break;
260 	}
261 }
262 
263 void
264 add_width_range(int start, int end, int width)
265 {
266 	for (; start <= end; start++) {
267 		add_width(start, width);
268 	}
269 }
270 
271 void
272 add_caseconv(int val, int wc)
273 {
274 	ctype_node_t	*ctn;
275 
276 	ctn = get_ctype(val);
277 	if (ctn == NULL) {
278 		INTERR;
279 		return;
280 	}
281 
282 	switch (last_kw) {
283 	case T_TOUPPER:
284 		ctn->toupper = wc;
285 		break;
286 	case T_TOLOWER:
287 		ctn->tolower = wc;
288 		break;
289 	default:
290 		INTERR;
291 		break;
292 	}
293 }
294 
295 void
296 dump_ctype(void)
297 {
298 	FILE		*f;
299 	_FileRuneLocale	rl;
300 	ctype_node_t	*ctn, *last_ct, *last_lo, *last_up;
301 	_FileRuneEntry	*ct = NULL;
302 	_FileRuneEntry	*lo = NULL;
303 	_FileRuneEntry	*up = NULL;
304 	wchar_t		wc;
305 	uint32_t	runetype_ext_nranges;
306 	uint32_t	maplower_ext_nranges;
307 	uint32_t	mapupper_ext_nranges;
308 
309 	(void) memset(&rl, 0, sizeof (rl));
310 	runetype_ext_nranges = 0;
311 	last_ct = NULL;
312 	maplower_ext_nranges = 0;
313 	last_lo = NULL;
314 	mapupper_ext_nranges = 0;
315 	last_up = NULL;
316 
317 	if ((f = open_category()) == NULL)
318 		return;
319 
320 	(void) memcpy(rl.magic, _FILE_RUNE_MAGIC_1, 8);
321 	(void) strlcpy(rl.encoding, get_wide_encoding(), sizeof (rl.encoding));
322 
323 	/*
324 	 * Initialize the identity map.
325 	 */
326 	for (wc = 0; (unsigned)wc < _CACHED_RUNES; wc++) {
327 		rl.maplower[wc] = htote(wc);
328 		rl.mapupper[wc] = htote(wc);
329 	}
330 
331 	RB_FOREACH(ctn, ctypes, &ctypes) {
332 		int conflict = 0;
333 
334 		wc = ctn->wc;
335 
336 		/*
337 		 * POSIX requires certain portable characters have
338 		 * certain types.  Add them if they are missing.
339 		 */
340 		if ((wc >= 1) && (wc <= 127)) {
341 			if ((wc >= 'A') && (wc <= 'Z'))
342 				ctn->ctype |= _ISUPPER;
343 			if ((wc >= 'a') && (wc <= 'z'))
344 				ctn->ctype |= _ISLOWER;
345 			if ((wc >= '0') && (wc <= '9'))
346 				ctn->ctype |= _ISDIGIT;
347 			if (wc == ' ')
348 				ctn->ctype |= _ISPRINT;
349 			if (strchr(" \f\n\r\t\v", (char)wc) != NULL)
350 				ctn->ctype |= _ISSPACE;
351 			if (strchr("0123456789ABCDEFabcdef", (char)wc) != NULL)
352 				ctn->ctype |= _ISXDIGIT;
353 			if (strchr(" \t", (char)wc))
354 				ctn->ctype |= _ISBLANK;
355 
356 			/*
357 			 * Technically these settings are only
358 			 * required for the C locale.  However, it
359 			 * turns out that because of the historical
360 			 * version of isprint(), we need them for all
361 			 * locales as well.  Note that these are not
362 			 * necessarily valid punctation characters in
363 			 * the current language, but ispunct() needs
364 			 * to return TRUE for them.
365 			 */
366 			if (strchr("!\"'#$%&()*+,-./:;<=>?@[\\]^_`{|}~",
367 			    (char)wc))
368 				ctn->ctype |= _ISPUNCT;
369 		}
370 
371 		/*
372 		 * POSIX also requires that certain types imply
373 		 * others.  Add any inferred types here.
374 		 */
375 		if (ctn->ctype & (_ISUPPER |_ISLOWER))
376 			ctn->ctype |= _ISALPHA;
377 		if (ctn->ctype & _ISDIGIT)
378 			ctn->ctype |= _ISXDIGIT;
379 		if (ctn->ctype & _ISBLANK)
380 			ctn->ctype |= _ISSPACE;
381 		if (ctn->ctype & (_ISALPHA|_ISDIGIT|_ISXDIGIT))
382 			ctn->ctype |= _ISGRAPH;
383 		if (ctn->ctype & _ISGRAPH)
384 			ctn->ctype |= _ISPRINT;
385 
386 		/*
387 		 * POSIX requires that certain combinations are invalid.
388 		 * Try fixing the cases we know about (see add_ctype_impl()).
389 		 */
390 		if ((ctn->ctype & (_ISSPACE|_ISCNTRL)) == (_ISSPACE|_ISCNTRL))
391 			ctn->ctype &= ~_ISPRINT;
392 
393 		/*
394 		 * Finally, don't flag remaining cases as a fatal error,
395 		 * and just warn about them.
396 		 */
397 		if ((ctn->ctype & _ISALPHA) &&
398 		    (ctn->ctype & (_ISPUNCT|_ISDIGIT)))
399 			conflict++;
400 		if ((ctn->ctype & _ISPUNCT) &&
401 		    (ctn->ctype & (_ISDIGIT|_ISALPHA|_ISXDIGIT)))
402 			conflict++;
403 		if ((ctn->ctype & _ISSPACE) && (ctn->ctype & _ISGRAPH))
404 			conflict++;
405 		if ((ctn->ctype & _ISCNTRL) && (ctn->ctype & _ISPRINT))
406 			conflict++;
407 		if ((wc == ' ') && (ctn->ctype & (_ISPUNCT|_ISGRAPH)))
408 			conflict++;
409 
410 		if (conflict) {
411 			warn("conflicting classes for character 0x%x (%x)",
412 			    wc, ctn->ctype);
413 		}
414 		/*
415 		 * Handle the lower 256 characters using the simple
416 		 * optimization.  Note that if we have not defined the
417 		 * upper/lower case, then we identity map it.
418 		 */
419 		if ((unsigned)wc < _CACHED_RUNES) {
420 			rl.runetype[wc] = htote(ctn->ctype);
421 			if (ctn->tolower)
422 				rl.maplower[wc] = htote(ctn->tolower);
423 			if (ctn->toupper)
424 				rl.mapupper[wc] = htote(ctn->toupper);
425 			continue;
426 		}
427 
428 		if ((last_ct != NULL) && (last_ct->ctype == ctn->ctype) &&
429 		    (last_ct->wc + 1 == wc)) {
430 			ct[runetype_ext_nranges - 1].max = htote(wc);
431 		} else {
432 			runetype_ext_nranges++;
433 			ct = realloc(ct, sizeof (*ct) * runetype_ext_nranges);
434 			ct[runetype_ext_nranges - 1].min = htote(wc);
435 			ct[runetype_ext_nranges - 1].max = htote(wc);
436 			ct[runetype_ext_nranges - 1].map =
437 			    htote(ctn->ctype);
438 		}
439 		last_ct = ctn;
440 		if (ctn->tolower == 0) {
441 			last_lo = NULL;
442 		} else if ((last_lo != NULL) &&
443 		    (last_lo->tolower + 1 == ctn->tolower)) {
444 			lo[maplower_ext_nranges - 1].max = htote(wc);
445 			last_lo = ctn;
446 		} else {
447 			maplower_ext_nranges++;
448 			lo = realloc(lo, sizeof (*lo) * maplower_ext_nranges);
449 			lo[maplower_ext_nranges - 1].min = htote(wc);
450 			lo[maplower_ext_nranges - 1].max = htote(wc);
451 			lo[maplower_ext_nranges - 1].map =
452 			    htote(ctn->tolower);
453 			last_lo = ctn;
454 		}
455 
456 		if (ctn->toupper == 0) {
457 			last_up = NULL;
458 		} else if ((last_up != NULL) &&
459 		    (last_up->toupper + 1 == ctn->toupper)) {
460 			up[mapupper_ext_nranges-1].max = htote(wc);
461 			last_up = ctn;
462 		} else {
463 			mapupper_ext_nranges++;
464 			up = realloc(up, sizeof (*up) * mapupper_ext_nranges);
465 			up[mapupper_ext_nranges - 1].min = htote(wc);
466 			up[mapupper_ext_nranges - 1].max = htote(wc);
467 			up[mapupper_ext_nranges - 1].map =
468 			    htote(ctn->toupper);
469 			last_up = ctn;
470 		}
471 	}
472 
473 	rl.runetype_ext_nranges = htote(runetype_ext_nranges);
474 	rl.maplower_ext_nranges = htote(maplower_ext_nranges);
475 	rl.mapupper_ext_nranges = htote(mapupper_ext_nranges);
476 	if ((wr_category(&rl, sizeof (rl), f) < 0) ||
477 	    (wr_category(ct, sizeof (*ct) * runetype_ext_nranges, f) < 0) ||
478 	    (wr_category(lo, sizeof (*lo) * maplower_ext_nranges, f) < 0) ||
479 	    (wr_category(up, sizeof (*up) * mapupper_ext_nranges, f) < 0)) {
480 		return;
481 	}
482 
483 	close_category(f);
484 }
485