1 /*-------------------------------------------------------------------------
4 * ctype functions adapted to work on pg_wchar (a/k/a chr),
5 * and functions to cache the results of wholesale ctype probing.
7 * This file is #included by regcomp.c; it's not meant to compile standalone.
9 * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
10 * Portions Copyright (c) 1994, Regents of the University of California
13 * src/backend/regex/regc_pg_locale.c
15 *-------------------------------------------------------------------------
18 #include "catalog/pg_collation.h"
19 #include "common/unicode_case.h"
20 #include "common/unicode_category.h"
21 #include "utils/pg_locale.h"
24 * To provide as much functionality as possible on a variety of platforms,
25 * without going so far as to implement everything from scratch, we use
26 * several implementation strategies depending on the situation:
28 * 1. In C/POSIX collations, we use hard-wired code. We can't depend on
29 * the <ctype.h> functions since those will obey LC_CTYPE. Note that these
30 * collations don't give a fig about multibyte characters.
32 * 2. In the "default" collation (which is supposed to obey LC_CTYPE):
34 * 2a. When working in UTF8 encoding, we use the <wctype.h> functions.
35 * This assumes that every platform uses Unicode codepoints directly
36 * as the wchar_t representation of Unicode. On some platforms
37 * wchar_t is only 16 bits wide, so we have to punt for codepoints > 0xFFFF.
39 * 2b. In all other encodings, we use the <ctype.h> functions for pg_wchar
40 * values up to 255, and punt for values above that. This is 100% correct
41 * only in single-byte encodings such as LATINn. However, non-Unicode
42 * multibyte encodings are mostly Far Eastern character sets for which the
43 * properties being tested here aren't very relevant for higher code values
44 * anyway. The difficulty with using the <wctype.h> functions with
45 * non-Unicode multibyte encodings is that we can have no certainty that
46 * the platform's wchar_t representation matches what we do in pg_wchar
49 * 3. Here, we use the locale_t-extended forms of the <wctype.h> and <ctype.h>
50 * functions, under exactly the same cases as #2.
52 * There is one notable difference between cases 2 and 3: in the "default"
53 * collation we force ASCII letters to follow ASCII upcase/downcase rules,
54 * while in a non-default collation we just let the library functions do what
55 * they will. The case where this matters is treatment of I/i in Turkish,
56 * and the behavior is meant to match the upper()/lower() SQL functions.
58 * We store the active collation setting in static variables. In principle
59 * it could be passed down to here via the regex library's "struct vars" data
60 * structure; but that would require somewhat invasive changes in the regex
61 * library, and right now there's no real benefit to be gained from that.
63 * NB: the coding here assumes pg_wchar is an unsigned type.
68 PG_REGEX_STRATEGY_C
, /* C locale (encoding independent) */
69 PG_REGEX_STRATEGY_BUILTIN
, /* built-in Unicode semantics */
70 PG_REGEX_STRATEGY_LIBC_WIDE
, /* Use locale_t <wctype.h> functions */
71 PG_REGEX_STRATEGY_LIBC_1BYTE
, /* Use locale_t <ctype.h> functions */
72 PG_REGEX_STRATEGY_ICU
, /* Use ICU uchar.h functions */
75 static PG_Locale_Strategy pg_regex_strategy
;
76 static pg_locale_t pg_regex_locale
;
79 * Hard-wired character properties for C locale
81 #define PG_ISDIGIT 0x01
82 #define PG_ISALPHA 0x02
83 #define PG_ISALNUM (PG_ISDIGIT | PG_ISALPHA)
84 #define PG_ISUPPER 0x04
85 #define PG_ISLOWER 0x08
86 #define PG_ISGRAPH 0x10
87 #define PG_ISPRINT 0x20
88 #define PG_ISPUNCT 0x40
89 #define PG_ISSPACE 0x80
91 static const unsigned char pg_char_properties
[128] = {
124 /* */ PG_ISPRINT
| PG_ISSPACE
,
125 /* ! */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
126 /* " */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
127 /* # */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
128 /* $ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
129 /* % */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
130 /* & */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
131 /* ' */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
132 /* ( */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
133 /* ) */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
134 /* * */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
135 /* + */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
136 /* , */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
137 /* - */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
138 /* . */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
139 /* / */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
140 /* 0 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
141 /* 1 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
142 /* 2 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
143 /* 3 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
144 /* 4 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
145 /* 5 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
146 /* 6 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
147 /* 7 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
148 /* 8 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
149 /* 9 */ PG_ISDIGIT
| PG_ISGRAPH
| PG_ISPRINT
,
150 /* : */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
151 /* ; */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
152 /* < */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
153 /* = */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
154 /* > */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
155 /* ? */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
156 /* @ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
157 /* A */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
158 /* B */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
159 /* C */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
160 /* D */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
161 /* E */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
162 /* F */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
163 /* G */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
164 /* H */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
165 /* I */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
166 /* J */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
167 /* K */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
168 /* L */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
169 /* M */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
170 /* N */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
171 /* O */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
172 /* P */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
173 /* Q */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
174 /* R */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
175 /* S */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
176 /* T */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
177 /* U */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
178 /* V */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
179 /* W */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
180 /* X */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
181 /* Y */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
182 /* Z */ PG_ISALPHA
| PG_ISUPPER
| PG_ISGRAPH
| PG_ISPRINT
,
183 /* [ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
184 /* \ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
185 /* ] */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
186 /* ^ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
187 /* _ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
188 /* ` */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
189 /* a */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
190 /* b */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
191 /* c */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
192 /* d */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
193 /* e */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
194 /* f */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
195 /* g */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
196 /* h */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
197 /* i */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
198 /* j */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
199 /* k */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
200 /* l */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
201 /* m */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
202 /* n */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
203 /* o */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
204 /* p */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
205 /* q */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
206 /* r */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
207 /* s */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
208 /* t */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
209 /* u */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
210 /* v */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
211 /* w */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
212 /* x */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
213 /* y */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
214 /* z */ PG_ISALPHA
| PG_ISLOWER
| PG_ISGRAPH
| PG_ISPRINT
,
215 /* { */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
216 /* | */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
217 /* } */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
218 /* ~ */ PG_ISGRAPH
| PG_ISPRINT
| PG_ISPUNCT
,
224 * pg_set_regex_collation: set collation for these functions to obey
226 * This is called when beginning compilation or execution of a regexp.
227 * Since there's no need for reentrancy of regexp operations, it's okay
228 * to store the results in static variables.
231 pg_set_regex_collation(Oid collation
)
233 pg_locale_t locale
= 0;
234 PG_Locale_Strategy strategy
;
236 if (!OidIsValid(collation
))
239 * This typically means that the parser could not resolve a conflict
240 * of implicit collations, so report it that way.
243 (errcode(ERRCODE_INDETERMINATE_COLLATION
),
244 errmsg("could not determine which collation to use for regular expression"),
245 errhint("Use the COLLATE clause to set the collation explicitly.")));
248 if (collation
== C_COLLATION_OID
)
251 * Some callers expect regexes to work for C_COLLATION_OID before
252 * catalog access is available, so we can't call
253 * pg_newlocale_from_collation().
255 strategy
= PG_REGEX_STRATEGY_C
;
260 locale
= pg_newlocale_from_collation(collation
);
262 if (!locale
->deterministic
)
264 (errcode(ERRCODE_FEATURE_NOT_SUPPORTED
),
265 errmsg("nondeterministic collations are not supported for regular expressions")));
267 if (locale
->ctype_is_c
)
270 * C/POSIX collations use this path regardless of database
273 strategy
= PG_REGEX_STRATEGY_C
;
276 else if (locale
->provider
== COLLPROVIDER_BUILTIN
)
278 Assert(GetDatabaseEncoding() == PG_UTF8
);
279 strategy
= PG_REGEX_STRATEGY_BUILTIN
;
282 else if (locale
->provider
== COLLPROVIDER_ICU
)
284 strategy
= PG_REGEX_STRATEGY_ICU
;
289 Assert(locale
->provider
== COLLPROVIDER_LIBC
);
290 if (GetDatabaseEncoding() == PG_UTF8
)
291 strategy
= PG_REGEX_STRATEGY_LIBC_WIDE
;
293 strategy
= PG_REGEX_STRATEGY_LIBC_1BYTE
;
297 pg_regex_strategy
= strategy
;
298 pg_regex_locale
= locale
;
302 pg_wc_isdigit(pg_wchar c
)
304 switch (pg_regex_strategy
)
306 case PG_REGEX_STRATEGY_C
:
307 return (c
<= (pg_wchar
) 127 &&
308 (pg_char_properties
[c
] & PG_ISDIGIT
));
309 case PG_REGEX_STRATEGY_BUILTIN
:
310 return pg_u_isdigit(c
, true);
311 case PG_REGEX_STRATEGY_LIBC_WIDE
:
312 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
313 return iswdigit_l((wint_t) c
, pg_regex_locale
->info
.lt
);
315 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
316 return (c
<= (pg_wchar
) UCHAR_MAX
&&
317 isdigit_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
319 case PG_REGEX_STRATEGY_ICU
:
325 return 0; /* can't get here, but keep compiler quiet */
329 pg_wc_isalpha(pg_wchar c
)
331 switch (pg_regex_strategy
)
333 case PG_REGEX_STRATEGY_C
:
334 return (c
<= (pg_wchar
) 127 &&
335 (pg_char_properties
[c
] & PG_ISALPHA
));
336 case PG_REGEX_STRATEGY_BUILTIN
:
337 return pg_u_isalpha(c
);
338 case PG_REGEX_STRATEGY_LIBC_WIDE
:
339 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
340 return iswalpha_l((wint_t) c
, pg_regex_locale
->info
.lt
);
342 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
343 return (c
<= (pg_wchar
) UCHAR_MAX
&&
344 isalpha_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
346 case PG_REGEX_STRATEGY_ICU
:
352 return 0; /* can't get here, but keep compiler quiet */
356 pg_wc_isalnum(pg_wchar c
)
358 switch (pg_regex_strategy
)
360 case PG_REGEX_STRATEGY_C
:
361 return (c
<= (pg_wchar
) 127 &&
362 (pg_char_properties
[c
] & PG_ISALNUM
));
363 case PG_REGEX_STRATEGY_BUILTIN
:
364 return pg_u_isalnum(c
, true);
365 case PG_REGEX_STRATEGY_LIBC_WIDE
:
366 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
367 return iswalnum_l((wint_t) c
, pg_regex_locale
->info
.lt
);
369 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
370 return (c
<= (pg_wchar
) UCHAR_MAX
&&
371 isalnum_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
373 case PG_REGEX_STRATEGY_ICU
:
379 return 0; /* can't get here, but keep compiler quiet */
383 pg_wc_isword(pg_wchar c
)
385 /* We define word characters as alnum class plus underscore */
388 return pg_wc_isalnum(c
);
392 pg_wc_isupper(pg_wchar c
)
394 switch (pg_regex_strategy
)
396 case PG_REGEX_STRATEGY_C
:
397 return (c
<= (pg_wchar
) 127 &&
398 (pg_char_properties
[c
] & PG_ISUPPER
));
399 case PG_REGEX_STRATEGY_BUILTIN
:
400 return pg_u_isupper(c
);
401 case PG_REGEX_STRATEGY_LIBC_WIDE
:
402 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
403 return iswupper_l((wint_t) c
, pg_regex_locale
->info
.lt
);
405 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
406 return (c
<= (pg_wchar
) UCHAR_MAX
&&
407 isupper_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
409 case PG_REGEX_STRATEGY_ICU
:
415 return 0; /* can't get here, but keep compiler quiet */
419 pg_wc_islower(pg_wchar c
)
421 switch (pg_regex_strategy
)
423 case PG_REGEX_STRATEGY_C
:
424 return (c
<= (pg_wchar
) 127 &&
425 (pg_char_properties
[c
] & PG_ISLOWER
));
426 case PG_REGEX_STRATEGY_BUILTIN
:
427 return pg_u_islower(c
);
428 case PG_REGEX_STRATEGY_LIBC_WIDE
:
429 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
430 return iswlower_l((wint_t) c
, pg_regex_locale
->info
.lt
);
432 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
433 return (c
<= (pg_wchar
) UCHAR_MAX
&&
434 islower_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
436 case PG_REGEX_STRATEGY_ICU
:
442 return 0; /* can't get here, but keep compiler quiet */
446 pg_wc_isgraph(pg_wchar c
)
448 switch (pg_regex_strategy
)
450 case PG_REGEX_STRATEGY_C
:
451 return (c
<= (pg_wchar
) 127 &&
452 (pg_char_properties
[c
] & PG_ISGRAPH
));
453 case PG_REGEX_STRATEGY_BUILTIN
:
454 return pg_u_isgraph(c
);
455 case PG_REGEX_STRATEGY_LIBC_WIDE
:
456 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
457 return iswgraph_l((wint_t) c
, pg_regex_locale
->info
.lt
);
459 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
460 return (c
<= (pg_wchar
) UCHAR_MAX
&&
461 isgraph_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
463 case PG_REGEX_STRATEGY_ICU
:
469 return 0; /* can't get here, but keep compiler quiet */
473 pg_wc_isprint(pg_wchar c
)
475 switch (pg_regex_strategy
)
477 case PG_REGEX_STRATEGY_C
:
478 return (c
<= (pg_wchar
) 127 &&
479 (pg_char_properties
[c
] & PG_ISPRINT
));
480 case PG_REGEX_STRATEGY_BUILTIN
:
481 return pg_u_isprint(c
);
482 case PG_REGEX_STRATEGY_LIBC_WIDE
:
483 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
484 return iswprint_l((wint_t) c
, pg_regex_locale
->info
.lt
);
486 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
487 return (c
<= (pg_wchar
) UCHAR_MAX
&&
488 isprint_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
490 case PG_REGEX_STRATEGY_ICU
:
496 return 0; /* can't get here, but keep compiler quiet */
500 pg_wc_ispunct(pg_wchar c
)
502 switch (pg_regex_strategy
)
504 case PG_REGEX_STRATEGY_C
:
505 return (c
<= (pg_wchar
) 127 &&
506 (pg_char_properties
[c
] & PG_ISPUNCT
));
507 case PG_REGEX_STRATEGY_BUILTIN
:
508 return pg_u_ispunct(c
, true);
509 case PG_REGEX_STRATEGY_LIBC_WIDE
:
510 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
511 return iswpunct_l((wint_t) c
, pg_regex_locale
->info
.lt
);
513 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
514 return (c
<= (pg_wchar
) UCHAR_MAX
&&
515 ispunct_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
517 case PG_REGEX_STRATEGY_ICU
:
523 return 0; /* can't get here, but keep compiler quiet */
527 pg_wc_isspace(pg_wchar c
)
529 switch (pg_regex_strategy
)
531 case PG_REGEX_STRATEGY_C
:
532 return (c
<= (pg_wchar
) 127 &&
533 (pg_char_properties
[c
] & PG_ISSPACE
));
534 case PG_REGEX_STRATEGY_BUILTIN
:
535 return pg_u_isspace(c
);
536 case PG_REGEX_STRATEGY_LIBC_WIDE
:
537 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
538 return iswspace_l((wint_t) c
, pg_regex_locale
->info
.lt
);
540 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
541 return (c
<= (pg_wchar
) UCHAR_MAX
&&
542 isspace_l((unsigned char) c
, pg_regex_locale
->info
.lt
));
544 case PG_REGEX_STRATEGY_ICU
:
550 return 0; /* can't get here, but keep compiler quiet */
554 pg_wc_toupper(pg_wchar c
)
556 switch (pg_regex_strategy
)
558 case PG_REGEX_STRATEGY_C
:
559 if (c
<= (pg_wchar
) 127)
560 return pg_ascii_toupper((unsigned char) c
);
562 case PG_REGEX_STRATEGY_BUILTIN
:
563 return unicode_uppercase_simple(c
);
564 case PG_REGEX_STRATEGY_LIBC_WIDE
:
565 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
566 return towupper_l((wint_t) c
, pg_regex_locale
->info
.lt
);
568 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
569 if (c
<= (pg_wchar
) UCHAR_MAX
)
570 return toupper_l((unsigned char) c
, pg_regex_locale
->info
.lt
);
572 case PG_REGEX_STRATEGY_ICU
:
578 return 0; /* can't get here, but keep compiler quiet */
582 pg_wc_tolower(pg_wchar c
)
584 switch (pg_regex_strategy
)
586 case PG_REGEX_STRATEGY_C
:
587 if (c
<= (pg_wchar
) 127)
588 return pg_ascii_tolower((unsigned char) c
);
590 case PG_REGEX_STRATEGY_BUILTIN
:
591 return unicode_lowercase_simple(c
);
592 case PG_REGEX_STRATEGY_LIBC_WIDE
:
593 if (sizeof(wchar_t) >= 4 || c
<= (pg_wchar
) 0xFFFF)
594 return towlower_l((wint_t) c
, pg_regex_locale
->info
.lt
);
596 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
597 if (c
<= (pg_wchar
) UCHAR_MAX
)
598 return tolower_l((unsigned char) c
, pg_regex_locale
->info
.lt
);
600 case PG_REGEX_STRATEGY_ICU
:
606 return 0; /* can't get here, but keep compiler quiet */
611 * These functions cache the results of probing libc's ctype behavior for
612 * all character codes of interest in a given encoding/collation. The
613 * result is provided as a "struct cvec", but notice that the representation
614 * is a touch different from a cvec created by regc_cvec.c: we allocate the
615 * chrs[] and ranges[] arrays separately from the struct so that we can
616 * realloc them larger at need. This is okay since the cvecs made here
617 * should never be freed by freecvec().
619 * We use malloc not palloc since we mustn't lose control on out-of-memory;
620 * the main regex code expects us to return a failure indication instead.
623 typedef int (*pg_wc_probefunc
) (pg_wchar c
);
625 typedef struct pg_ctype_cache
627 pg_wc_probefunc probefunc
; /* pg_wc_isalpha or a sibling */
628 pg_locale_t locale
; /* locale this entry is for */
629 struct cvec cv
; /* cache entry contents */
630 struct pg_ctype_cache
*next
; /* chain link */
633 static pg_ctype_cache
*pg_ctype_cache_list
= NULL
;
636 * Add a chr or range to pcc->cv; return false if run out of memory
639 store_match(pg_ctype_cache
*pcc
, pg_wchar chr1
, int nchrs
)
645 if (pcc
->cv
.nranges
>= pcc
->cv
.rangespace
)
647 pcc
->cv
.rangespace
*= 2;
648 newchrs
= (chr
*) realloc(pcc
->cv
.ranges
,
649 pcc
->cv
.rangespace
* sizeof(chr
) * 2);
652 pcc
->cv
.ranges
= newchrs
;
654 pcc
->cv
.ranges
[pcc
->cv
.nranges
* 2] = chr1
;
655 pcc
->cv
.ranges
[pcc
->cv
.nranges
* 2 + 1] = chr1
+ nchrs
- 1;
661 if (pcc
->cv
.nchrs
>= pcc
->cv
.chrspace
)
663 pcc
->cv
.chrspace
*= 2;
664 newchrs
= (chr
*) realloc(pcc
->cv
.chrs
,
665 pcc
->cv
.chrspace
* sizeof(chr
));
668 pcc
->cv
.chrs
= newchrs
;
670 pcc
->cv
.chrs
[pcc
->cv
.nchrs
++] = chr1
;
676 * Given a probe function (e.g., pg_wc_isalpha) get a struct cvec for all
677 * chrs satisfying the probe function. The active collation is the one
678 * previously set by pg_set_regex_collation. Return NULL if out of memory.
680 * Note that the result must not be freed or modified by caller.
683 pg_ctype_get_cache(pg_wc_probefunc probefunc
, int cclasscode
)
692 * Do we already have the answer cached?
694 for (pcc
= pg_ctype_cache_list
; pcc
!= NULL
; pcc
= pcc
->next
)
696 if (pcc
->probefunc
== probefunc
&&
697 pcc
->locale
== pg_regex_locale
)
702 * Nope, so initialize some workspace ...
704 pcc
= (pg_ctype_cache
*) malloc(sizeof(pg_ctype_cache
));
707 pcc
->probefunc
= probefunc
;
708 pcc
->locale
= pg_regex_locale
;
710 pcc
->cv
.chrspace
= 128;
711 pcc
->cv
.chrs
= (chr
*) malloc(pcc
->cv
.chrspace
* sizeof(chr
));
713 pcc
->cv
.rangespace
= 64;
714 pcc
->cv
.ranges
= (chr
*) malloc(pcc
->cv
.rangespace
* sizeof(chr
) * 2);
715 if (pcc
->cv
.chrs
== NULL
|| pcc
->cv
.ranges
== NULL
)
717 pcc
->cv
.cclasscode
= cclasscode
;
720 * Decide how many character codes we ought to look through. In general
721 * we don't go past MAX_SIMPLE_CHR; chr codes above that are handled at
722 * runtime using the "high colormap" mechanism. However, in C locale
723 * there's no need to go further than 127, and if we only have a 1-byte
724 * <ctype.h> API there's no need to go further than that can handle.
726 * If it's not MAX_SIMPLE_CHR that's constraining the search, mark the
727 * output cvec as not having any locale-dependent behavior, since there
728 * will be no need to do any run-time locale checks. (The #if's here
729 * would always be true for production values of MAX_SIMPLE_CHR, but it's
730 * useful to allow it to be small for testing purposes.)
732 switch (pg_regex_strategy
)
734 case PG_REGEX_STRATEGY_C
:
735 #if MAX_SIMPLE_CHR >= 127
736 max_chr
= (pg_wchar
) 127;
737 pcc
->cv
.cclasscode
= -1;
739 max_chr
= (pg_wchar
) MAX_SIMPLE_CHR
;
742 case PG_REGEX_STRATEGY_BUILTIN
:
743 max_chr
= (pg_wchar
) MAX_SIMPLE_CHR
;
745 case PG_REGEX_STRATEGY_LIBC_WIDE
:
746 max_chr
= (pg_wchar
) MAX_SIMPLE_CHR
;
748 case PG_REGEX_STRATEGY_LIBC_1BYTE
:
749 #if MAX_SIMPLE_CHR >= UCHAR_MAX
750 max_chr
= (pg_wchar
) UCHAR_MAX
;
751 pcc
->cv
.cclasscode
= -1;
753 max_chr
= (pg_wchar
) MAX_SIMPLE_CHR
;
756 case PG_REGEX_STRATEGY_ICU
:
757 max_chr
= (pg_wchar
) MAX_SIMPLE_CHR
;
761 max_chr
= 0; /* can't get here, but keep compiler quiet */
768 nmatches
= 0; /* number of consecutive matches */
770 for (cur_chr
= 0; cur_chr
<= max_chr
; cur_chr
++)
772 if ((*probefunc
) (cur_chr
))
774 else if (nmatches
> 0)
776 if (!store_match(pcc
, cur_chr
- nmatches
, nmatches
))
783 if (!store_match(pcc
, cur_chr
- nmatches
, nmatches
))
787 * We might have allocated more memory than needed, if so free it
789 if (pcc
->cv
.nchrs
== 0)
793 pcc
->cv
.chrspace
= 0;
795 else if (pcc
->cv
.nchrs
< pcc
->cv
.chrspace
)
797 newchrs
= (chr
*) realloc(pcc
->cv
.chrs
,
798 pcc
->cv
.nchrs
* sizeof(chr
));
801 pcc
->cv
.chrs
= newchrs
;
802 pcc
->cv
.chrspace
= pcc
->cv
.nchrs
;
804 if (pcc
->cv
.nranges
== 0)
806 free(pcc
->cv
.ranges
);
807 pcc
->cv
.ranges
= NULL
;
808 pcc
->cv
.rangespace
= 0;
810 else if (pcc
->cv
.nranges
< pcc
->cv
.rangespace
)
812 newchrs
= (chr
*) realloc(pcc
->cv
.ranges
,
813 pcc
->cv
.nranges
* sizeof(chr
) * 2);
816 pcc
->cv
.ranges
= newchrs
;
817 pcc
->cv
.rangespace
= pcc
->cv
.nranges
;
821 * Success, link it into cache chain
823 pcc
->next
= pg_ctype_cache_list
;
824 pg_ctype_cache_list
= pcc
;
833 free(pcc
->cv
.ranges
);