PostgreSQL Source Code  git master
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros
regc_pg_locale.c
Go to the documentation of this file.
1 /*-------------------------------------------------------------------------
2  *
3  * regc_pg_locale.c
4  * ctype functions adapted to work on pg_wchar (a/k/a chr),
5  * and functions to cache the results of wholesale ctype probing.
6  *
7  * This file is #included by regcomp.c; it's not meant to compile standalone.
8  *
9  * Portions Copyright (c) 1996-2017, PostgreSQL Global Development Group
10  * Portions Copyright (c) 1994, Regents of the University of California
11  *
12  * IDENTIFICATION
13  * src/backend/regex/regc_pg_locale.c
14  *
15  *-------------------------------------------------------------------------
16  */
17 
18 #include "catalog/pg_collation.h"
19 #include "utils/pg_locale.h"
20 
21 /*
22  * To provide as much functionality as possible on a variety of platforms,
23  * without going so far as to implement everything from scratch, we use
24  * several implementation strategies depending on the situation:
25  *
26  * 1. In C/POSIX collations, we use hard-wired code. We can't depend on
27  * the <ctype.h> functions since those will obey LC_CTYPE. Note that these
28  * collations don't give a fig about multibyte characters.
29  *
30  * 2. In the "default" collation (which is supposed to obey LC_CTYPE):
31  *
32  * 2a. When working in UTF8 encoding, we use the <wctype.h> functions.
33  * This assumes that every platform uses Unicode codepoints directly
34  * as the wchar_t representation of Unicode. On some platforms
35  * wchar_t is only 16 bits wide, so we have to punt for codepoints > 0xFFFF.
36  *
37  * 2b. In all other encodings, we use the <ctype.h> functions for pg_wchar
38  * values up to 255, and punt for values above that. This is 100% correct
39  * only in single-byte encodings such as LATINn. However, non-Unicode
40  * multibyte encodings are mostly Far Eastern character sets for which the
41  * properties being tested here aren't very relevant for higher code values
42  * anyway. The difficulty with using the <wctype.h> functions with
43  * non-Unicode multibyte encodings is that we can have no certainty that
44  * the platform's wchar_t representation matches what we do in pg_wchar
45  * conversions.
46  *
47  * 3. Other collations are only supported on platforms that HAVE_LOCALE_T.
48  * Here, we use the locale_t-extended forms of the <wctype.h> and <ctype.h>
49  * functions, under exactly the same cases as #2.
50  *
51  * There is one notable difference between cases 2 and 3: in the "default"
52  * collation we force ASCII letters to follow ASCII upcase/downcase rules,
53  * while in a non-default collation we just let the library functions do what
54  * they will. The case where this matters is treatment of I/i in Turkish,
55  * and the behavior is meant to match the upper()/lower() SQL functions.
56  *
57  * We store the active collation setting in static variables. In principle
58  * it could be passed down to here via the regex library's "struct vars" data
59  * structure; but that would require somewhat invasive changes in the regex
60  * library, and right now there's no real benefit to be gained from that.
61  *
62  * NB: the coding here assumes pg_wchar is an unsigned type.
63  */
64 
65 typedef enum
66 {
67  PG_REGEX_LOCALE_C, /* C locale (encoding independent) */
68  PG_REGEX_LOCALE_WIDE, /* Use <wctype.h> functions */
69  PG_REGEX_LOCALE_1BYTE, /* Use <ctype.h> functions */
70  PG_REGEX_LOCALE_WIDE_L, /* Use locale_t <wctype.h> functions */
71  PG_REGEX_LOCALE_1BYTE_L, /* Use locale_t <ctype.h> functions */
72  PG_REGEX_LOCALE_ICU /* Use ICU uchar.h functions */
74 
78 
79 /*
80  * Hard-wired character properties for C locale
81  */
82 #define PG_ISDIGIT 0x01
83 #define PG_ISALPHA 0x02
84 #define PG_ISALNUM (PG_ISDIGIT | PG_ISALPHA)
85 #define PG_ISUPPER 0x04
86 #define PG_ISLOWER 0x08
87 #define PG_ISGRAPH 0x10
88 #define PG_ISPRINT 0x20
89 #define PG_ISPUNCT 0x40
90 #define PG_ISSPACE 0x80
91 
92 static const unsigned char pg_char_properties[128] = {
93  /* NUL */ 0,
94  /* ^A */ 0,
95  /* ^B */ 0,
96  /* ^C */ 0,
97  /* ^D */ 0,
98  /* ^E */ 0,
99  /* ^F */ 0,
100  /* ^G */ 0,
101  /* ^H */ 0,
102  /* ^I */ PG_ISSPACE,
103  /* ^J */ PG_ISSPACE,
104  /* ^K */ PG_ISSPACE,
105  /* ^L */ PG_ISSPACE,
106  /* ^M */ PG_ISSPACE,
107  /* ^N */ 0,
108  /* ^O */ 0,
109  /* ^P */ 0,
110  /* ^Q */ 0,
111  /* ^R */ 0,
112  /* ^S */ 0,
113  /* ^T */ 0,
114  /* ^U */ 0,
115  /* ^V */ 0,
116  /* ^W */ 0,
117  /* ^X */ 0,
118  /* ^Y */ 0,
119  /* ^Z */ 0,
120  /* ^[ */ 0,
121  /* ^\ */ 0,
122  /* ^] */ 0,
123  /* ^^ */ 0,
124  /* ^_ */ 0,
125  /* */ PG_ISPRINT | PG_ISSPACE,
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  /* / */ PG_ISGRAPH | PG_ISPRINT | PG_ISPUNCT,
141  /* 0 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
142  /* 1 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
143  /* 2 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
144  /* 3 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
145  /* 4 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
146  /* 5 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
147  /* 6 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
148  /* 7 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
149  /* 8 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
150  /* 9 */ PG_ISDIGIT | PG_ISGRAPH | PG_ISPRINT,
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  /* @ */ 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  /* ` */ 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,
219  /* ~ */ PG_ISGRAPH | PG_ISPRINT | PG_ISPUNCT,
220  /* DEL */ 0
221 };
222 
223 
224 /*
225  * pg_set_regex_collation: set collation for these functions to obey
226  *
227  * This is called when beginning compilation or execution of a regexp.
228  * Since there's no need for reentrancy of regexp operations, it's okay
229  * to store the results in static variables.
230  */
231 void
233 {
234  if (lc_ctype_is_c(collation))
235  {
236  /* C/POSIX collations use this path regardless of database encoding */
238  pg_regex_locale = 0;
240  }
241  else
242  {
243  if (collation == DEFAULT_COLLATION_OID)
244  pg_regex_locale = 0;
245  else if (OidIsValid(collation))
246  {
247  /*
248  * NB: pg_newlocale_from_collation will fail if not HAVE_LOCALE_T;
249  * the case of pg_regex_locale != 0 but not HAVE_LOCALE_T does not
250  * have to be considered below.
251  */
252  pg_regex_locale = pg_newlocale_from_collation(collation);
253  }
254  else
255  {
256  /*
257  * This typically means that the parser could not resolve a
258  * conflict of implicit collations, so report it that way.
259  */
260  ereport(ERROR,
261  (errcode(ERRCODE_INDETERMINATE_COLLATION),
262  errmsg("could not determine which collation to use for regular expression"),
263  errhint("Use the COLLATE clause to set the collation explicitly.")));
264  }
265 
266 #ifdef USE_ICU
267  if (pg_regex_locale && pg_regex_locale->provider == COLLPROVIDER_ICU)
269  else
270 #endif
271  if (GetDatabaseEncoding() == PG_UTF8)
272  {
273  if (pg_regex_locale)
275  else
277  }
278  else
279  {
280  if (pg_regex_locale)
282  else
284  }
285 
286  pg_regex_collation = collation;
287  }
288 }
289 
290 static int
292 {
293  switch (pg_regex_strategy)
294  {
295  case PG_REGEX_LOCALE_C:
296  return (c <= (pg_wchar) 127 &&
299  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
300  return iswdigit((wint_t) c);
301  /* FALL THRU */
303  return (c <= (pg_wchar) UCHAR_MAX &&
304  isdigit((unsigned char) c));
306 #ifdef HAVE_LOCALE_T
307  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
308  return iswdigit_l((wint_t) c, pg_regex_locale->info.lt);
309 #endif
310  /* FALL THRU */
312 #ifdef HAVE_LOCALE_T
313  return (c <= (pg_wchar) UCHAR_MAX &&
314  isdigit_l((unsigned char) c, pg_regex_locale->info.lt));
315 #endif
316  break;
317  case PG_REGEX_LOCALE_ICU:
318 #ifdef USE_ICU
319  return u_isdigit(c);
320 #endif
321  break;
322  }
323  return 0; /* can't get here, but keep compiler quiet */
324 }
325 
326 static int
328 {
329  switch (pg_regex_strategy)
330  {
331  case PG_REGEX_LOCALE_C:
332  return (c <= (pg_wchar) 127 &&
335  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
336  return iswalpha((wint_t) c);
337  /* FALL THRU */
339  return (c <= (pg_wchar) UCHAR_MAX &&
340  isalpha((unsigned char) c));
342 #ifdef HAVE_LOCALE_T
343  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
344  return iswalpha_l((wint_t) c, pg_regex_locale->info.lt);
345 #endif
346  /* FALL THRU */
348 #ifdef HAVE_LOCALE_T
349  return (c <= (pg_wchar) UCHAR_MAX &&
350  isalpha_l((unsigned char) c, pg_regex_locale->info.lt));
351 #endif
352  break;
353  case PG_REGEX_LOCALE_ICU:
354 #ifdef USE_ICU
355  return u_isalpha(c);
356 #endif
357  break;
358  }
359  return 0; /* can't get here, but keep compiler quiet */
360 }
361 
362 static int
364 {
365  switch (pg_regex_strategy)
366  {
367  case PG_REGEX_LOCALE_C:
368  return (c <= (pg_wchar) 127 &&
371  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
372  return iswalnum((wint_t) c);
373  /* FALL THRU */
375  return (c <= (pg_wchar) UCHAR_MAX &&
376  isalnum((unsigned char) c));
378 #ifdef HAVE_LOCALE_T
379  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
380  return iswalnum_l((wint_t) c, pg_regex_locale->info.lt);
381 #endif
382  /* FALL THRU */
384 #ifdef HAVE_LOCALE_T
385  return (c <= (pg_wchar) UCHAR_MAX &&
386  isalnum_l((unsigned char) c, pg_regex_locale->info.lt));
387 #endif
388  break;
389  case PG_REGEX_LOCALE_ICU:
390 #ifdef USE_ICU
391  return u_isalnum(c);
392 #endif
393  break;
394  }
395  return 0; /* can't get here, but keep compiler quiet */
396 }
397 
398 static int
400 {
401  switch (pg_regex_strategy)
402  {
403  case PG_REGEX_LOCALE_C:
404  return (c <= (pg_wchar) 127 &&
407  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
408  return iswupper((wint_t) c);
409  /* FALL THRU */
411  return (c <= (pg_wchar) UCHAR_MAX &&
412  isupper((unsigned char) c));
414 #ifdef HAVE_LOCALE_T
415  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
416  return iswupper_l((wint_t) c, pg_regex_locale->info.lt);
417 #endif
418  /* FALL THRU */
420 #ifdef HAVE_LOCALE_T
421  return (c <= (pg_wchar) UCHAR_MAX &&
422  isupper_l((unsigned char) c, pg_regex_locale->info.lt));
423 #endif
424  break;
425  case PG_REGEX_LOCALE_ICU:
426 #ifdef USE_ICU
427  return u_isupper(c);
428 #endif
429  break;
430  }
431  return 0; /* can't get here, but keep compiler quiet */
432 }
433 
434 static int
436 {
437  switch (pg_regex_strategy)
438  {
439  case PG_REGEX_LOCALE_C:
440  return (c <= (pg_wchar) 127 &&
443  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
444  return iswlower((wint_t) c);
445  /* FALL THRU */
447  return (c <= (pg_wchar) UCHAR_MAX &&
448  islower((unsigned char) c));
450 #ifdef HAVE_LOCALE_T
451  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
452  return iswlower_l((wint_t) c, pg_regex_locale->info.lt);
453 #endif
454  /* FALL THRU */
456 #ifdef HAVE_LOCALE_T
457  return (c <= (pg_wchar) UCHAR_MAX &&
458  islower_l((unsigned char) c, pg_regex_locale->info.lt));
459 #endif
460  break;
461  case PG_REGEX_LOCALE_ICU:
462 #ifdef USE_ICU
463  return u_islower(c);
464 #endif
465  break;
466  }
467  return 0; /* can't get here, but keep compiler quiet */
468 }
469 
470 static int
472 {
473  switch (pg_regex_strategy)
474  {
475  case PG_REGEX_LOCALE_C:
476  return (c <= (pg_wchar) 127 &&
479  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
480  return iswgraph((wint_t) c);
481  /* FALL THRU */
483  return (c <= (pg_wchar) UCHAR_MAX &&
484  isgraph((unsigned char) c));
486 #ifdef HAVE_LOCALE_T
487  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
488  return iswgraph_l((wint_t) c, pg_regex_locale->info.lt);
489 #endif
490  /* FALL THRU */
492 #ifdef HAVE_LOCALE_T
493  return (c <= (pg_wchar) UCHAR_MAX &&
494  isgraph_l((unsigned char) c, pg_regex_locale->info.lt));
495 #endif
496  break;
497  case PG_REGEX_LOCALE_ICU:
498 #ifdef USE_ICU
499  return u_isgraph(c);
500 #endif
501  break;
502  }
503  return 0; /* can't get here, but keep compiler quiet */
504 }
505 
506 static int
508 {
509  switch (pg_regex_strategy)
510  {
511  case PG_REGEX_LOCALE_C:
512  return (c <= (pg_wchar) 127 &&
515  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
516  return iswprint((wint_t) c);
517  /* FALL THRU */
519  return (c <= (pg_wchar) UCHAR_MAX &&
520  isprint((unsigned char) c));
522 #ifdef HAVE_LOCALE_T
523  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
524  return iswprint_l((wint_t) c, pg_regex_locale->info.lt);
525 #endif
526  /* FALL THRU */
528 #ifdef HAVE_LOCALE_T
529  return (c <= (pg_wchar) UCHAR_MAX &&
530  isprint_l((unsigned char) c, pg_regex_locale->info.lt));
531 #endif
532  break;
533  case PG_REGEX_LOCALE_ICU:
534 #ifdef USE_ICU
535  return u_isprint(c);
536 #endif
537  break;
538  }
539  return 0; /* can't get here, but keep compiler quiet */
540 }
541 
542 static int
544 {
545  switch (pg_regex_strategy)
546  {
547  case PG_REGEX_LOCALE_C:
548  return (c <= (pg_wchar) 127 &&
551  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
552  return iswpunct((wint_t) c);
553  /* FALL THRU */
555  return (c <= (pg_wchar) UCHAR_MAX &&
556  ispunct((unsigned char) c));
558 #ifdef HAVE_LOCALE_T
559  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
560  return iswpunct_l((wint_t) c, pg_regex_locale->info.lt);
561 #endif
562  /* FALL THRU */
564 #ifdef HAVE_LOCALE_T
565  return (c <= (pg_wchar) UCHAR_MAX &&
566  ispunct_l((unsigned char) c, pg_regex_locale->info.lt));
567 #endif
568  break;
569  case PG_REGEX_LOCALE_ICU:
570 #ifdef USE_ICU
571  return u_ispunct(c);
572 #endif
573  break;
574  }
575  return 0; /* can't get here, but keep compiler quiet */
576 }
577 
578 static int
580 {
581  switch (pg_regex_strategy)
582  {
583  case PG_REGEX_LOCALE_C:
584  return (c <= (pg_wchar) 127 &&
587  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
588  return iswspace((wint_t) c);
589  /* FALL THRU */
591  return (c <= (pg_wchar) UCHAR_MAX &&
592  isspace((unsigned char) c));
594 #ifdef HAVE_LOCALE_T
595  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
596  return iswspace_l((wint_t) c, pg_regex_locale->info.lt);
597 #endif
598  /* FALL THRU */
600 #ifdef HAVE_LOCALE_T
601  return (c <= (pg_wchar) UCHAR_MAX &&
602  isspace_l((unsigned char) c, pg_regex_locale->info.lt));
603 #endif
604  break;
605  case PG_REGEX_LOCALE_ICU:
606 #ifdef USE_ICU
607  return u_isspace(c);
608 #endif
609  break;
610  }
611  return 0; /* can't get here, but keep compiler quiet */
612 }
613 
614 static pg_wchar
616 {
617  switch (pg_regex_strategy)
618  {
619  case PG_REGEX_LOCALE_C:
620  if (c <= (pg_wchar) 127)
621  return pg_ascii_toupper((unsigned char) c);
622  return c;
624  /* force C behavior for ASCII characters, per comments above */
625  if (c <= (pg_wchar) 127)
626  return pg_ascii_toupper((unsigned char) c);
627  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
628  return towupper((wint_t) c);
629  /* FALL THRU */
631  /* force C behavior for ASCII characters, per comments above */
632  if (c <= (pg_wchar) 127)
633  return pg_ascii_toupper((unsigned char) c);
634  if (c <= (pg_wchar) UCHAR_MAX)
635  return toupper((unsigned char) c);
636  return c;
638 #ifdef HAVE_LOCALE_T
639  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
640  return towupper_l((wint_t) c, pg_regex_locale->info.lt);
641 #endif
642  /* FALL THRU */
644 #ifdef HAVE_LOCALE_T
645  if (c <= (pg_wchar) UCHAR_MAX)
646  return toupper_l((unsigned char) c, pg_regex_locale->info.lt);
647 #endif
648  return c;
649  case PG_REGEX_LOCALE_ICU:
650 #ifdef USE_ICU
651  return u_toupper(c);
652 #endif
653  break;
654  }
655  return 0; /* can't get here, but keep compiler quiet */
656 }
657 
658 static pg_wchar
660 {
661  switch (pg_regex_strategy)
662  {
663  case PG_REGEX_LOCALE_C:
664  if (c <= (pg_wchar) 127)
665  return pg_ascii_tolower((unsigned char) c);
666  return c;
668  /* force C behavior for ASCII characters, per comments above */
669  if (c <= (pg_wchar) 127)
670  return pg_ascii_tolower((unsigned char) c);
671  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
672  return towlower((wint_t) c);
673  /* FALL THRU */
675  /* force C behavior for ASCII characters, per comments above */
676  if (c <= (pg_wchar) 127)
677  return pg_ascii_tolower((unsigned char) c);
678  if (c <= (pg_wchar) UCHAR_MAX)
679  return tolower((unsigned char) c);
680  return c;
682 #ifdef HAVE_LOCALE_T
683  if (sizeof(wchar_t) >= 4 || c <= (pg_wchar) 0xFFFF)
684  return towlower_l((wint_t) c, pg_regex_locale->info.lt);
685 #endif
686  /* FALL THRU */
688 #ifdef HAVE_LOCALE_T
689  if (c <= (pg_wchar) UCHAR_MAX)
690  return tolower_l((unsigned char) c, pg_regex_locale->info.lt);
691 #endif
692  return c;
693  case PG_REGEX_LOCALE_ICU:
694 #ifdef USE_ICU
695  return u_tolower(c);
696 #endif
697  break;
698  }
699  return 0; /* can't get here, but keep compiler quiet */
700 }
701 
702 
703 /*
704  * These functions cache the results of probing libc's ctype behavior for
705  * all character codes of interest in a given encoding/collation. The
706  * result is provided as a "struct cvec", but notice that the representation
707  * is a touch different from a cvec created by regc_cvec.c: we allocate the
708  * chrs[] and ranges[] arrays separately from the struct so that we can
709  * realloc them larger at need. This is okay since the cvecs made here
710  * should never be freed by freecvec().
711  *
712  * We use malloc not palloc since we mustn't lose control on out-of-memory;
713  * the main regex code expects us to return a failure indication instead.
714  */
715 
716 typedef int (*pg_wc_probefunc) (pg_wchar c);
717 
718 typedef struct pg_ctype_cache
719 {
720  pg_wc_probefunc probefunc; /* pg_wc_isalpha or a sibling */
721  Oid collation; /* collation this entry is for */
722  struct cvec cv; /* cache entry contents */
723  struct pg_ctype_cache *next; /* chain link */
725 
727 
728 /*
729  * Add a chr or range to pcc->cv; return false if run out of memory
730  */
731 static bool
732 store_match(pg_ctype_cache *pcc, pg_wchar chr1, int nchrs)
733 {
734  chr *newchrs;
735 
736  if (nchrs > 1)
737  {
738  if (pcc->cv.nranges >= pcc->cv.rangespace)
739  {
740  pcc->cv.rangespace *= 2;
741  newchrs = (chr *) realloc(pcc->cv.ranges,
742  pcc->cv.rangespace * sizeof(chr) * 2);
743  if (newchrs == NULL)
744  return false;
745  pcc->cv.ranges = newchrs;
746  }
747  pcc->cv.ranges[pcc->cv.nranges * 2] = chr1;
748  pcc->cv.ranges[pcc->cv.nranges * 2 + 1] = chr1 + nchrs - 1;
749  pcc->cv.nranges++;
750  }
751  else
752  {
753  assert(nchrs == 1);
754  if (pcc->cv.nchrs >= pcc->cv.chrspace)
755  {
756  pcc->cv.chrspace *= 2;
757  newchrs = (chr *) realloc(pcc->cv.chrs,
758  pcc->cv.chrspace * sizeof(chr));
759  if (newchrs == NULL)
760  return false;
761  pcc->cv.chrs = newchrs;
762  }
763  pcc->cv.chrs[pcc->cv.nchrs++] = chr1;
764  }
765  return true;
766 }
767 
768 /*
769  * Given a probe function (e.g., pg_wc_isalpha) get a struct cvec for all
770  * chrs satisfying the probe function. The active collation is the one
771  * previously set by pg_set_regex_collation. Return NULL if out of memory.
772  *
773  * Note that the result must not be freed or modified by caller.
774  */
775 static struct cvec *
777 {
778  pg_ctype_cache *pcc;
779  pg_wchar max_chr;
780  pg_wchar cur_chr;
781  int nmatches;
782  chr *newchrs;
783 
784  /*
785  * Do we already have the answer cached?
786  */
787  for (pcc = pg_ctype_cache_list; pcc != NULL; pcc = pcc->next)
788  {
789  if (pcc->probefunc == probefunc &&
791  return &pcc->cv;
792  }
793 
794  /*
795  * Nope, so initialize some workspace ...
796  */
797  pcc = (pg_ctype_cache *) malloc(sizeof(pg_ctype_cache));
798  if (pcc == NULL)
799  return NULL;
800  pcc->probefunc = probefunc;
802  pcc->cv.nchrs = 0;
803  pcc->cv.chrspace = 128;
804  pcc->cv.chrs = (chr *) malloc(pcc->cv.chrspace * sizeof(chr));
805  pcc->cv.nranges = 0;
806  pcc->cv.rangespace = 64;
807  pcc->cv.ranges = (chr *) malloc(pcc->cv.rangespace * sizeof(chr) * 2);
808  if (pcc->cv.chrs == NULL || pcc->cv.ranges == NULL)
809  goto out_of_memory;
810  pcc->cv.cclasscode = cclasscode;
811 
812  /*
813  * Decide how many character codes we ought to look through. In general
814  * we don't go past MAX_SIMPLE_CHR; chr codes above that are handled at
815  * runtime using the "high colormap" mechanism. However, in C locale
816  * there's no need to go further than 127, and if we only have a 1-byte
817  * <ctype.h> API there's no need to go further than that can handle.
818  *
819  * If it's not MAX_SIMPLE_CHR that's constraining the search, mark the
820  * output cvec as not having any locale-dependent behavior, since there
821  * will be no need to do any run-time locale checks. (The #if's here
822  * would always be true for production values of MAX_SIMPLE_CHR, but it's
823  * useful to allow it to be small for testing purposes.)
824  */
825  switch (pg_regex_strategy)
826  {
827  case PG_REGEX_LOCALE_C:
828 #if MAX_SIMPLE_CHR >= 127
829  max_chr = (pg_wchar) 127;
830  pcc->cv.cclasscode = -1;
831 #else
832  max_chr = (pg_wchar) MAX_SIMPLE_CHR;
833 #endif
834  break;
837  max_chr = (pg_wchar) MAX_SIMPLE_CHR;
838  break;
841 #if MAX_SIMPLE_CHR >= UCHAR_MAX
842  max_chr = (pg_wchar) UCHAR_MAX;
843  pcc->cv.cclasscode = -1;
844 #else
845  max_chr = (pg_wchar) MAX_SIMPLE_CHR;
846 #endif
847  break;
848  case PG_REGEX_LOCALE_ICU:
849  max_chr = (pg_wchar) MAX_SIMPLE_CHR;
850  break;
851  default:
852  max_chr = 0; /* can't get here, but keep compiler quiet */
853  break;
854  }
855 
856  /*
857  * And scan 'em ...
858  */
859  nmatches = 0; /* number of consecutive matches */
860 
861  for (cur_chr = 0; cur_chr <= max_chr; cur_chr++)
862  {
863  if ((*probefunc) (cur_chr))
864  nmatches++;
865  else if (nmatches > 0)
866  {
867  if (!store_match(pcc, cur_chr - nmatches, nmatches))
868  goto out_of_memory;
869  nmatches = 0;
870  }
871  }
872 
873  if (nmatches > 0)
874  if (!store_match(pcc, cur_chr - nmatches, nmatches))
875  goto out_of_memory;
876 
877  /*
878  * We might have allocated more memory than needed, if so free it
879  */
880  if (pcc->cv.nchrs == 0)
881  {
882  free(pcc->cv.chrs);
883  pcc->cv.chrs = NULL;
884  pcc->cv.chrspace = 0;
885  }
886  else if (pcc->cv.nchrs < pcc->cv.chrspace)
887  {
888  newchrs = (chr *) realloc(pcc->cv.chrs,
889  pcc->cv.nchrs * sizeof(chr));
890  if (newchrs == NULL)
891  goto out_of_memory;
892  pcc->cv.chrs = newchrs;
893  pcc->cv.chrspace = pcc->cv.nchrs;
894  }
895  if (pcc->cv.nranges == 0)
896  {
897  free(pcc->cv.ranges);
898  pcc->cv.ranges = NULL;
899  pcc->cv.rangespace = 0;
900  }
901  else if (pcc->cv.nranges < pcc->cv.rangespace)
902  {
903  newchrs = (chr *) realloc(pcc->cv.ranges,
904  pcc->cv.nranges * sizeof(chr) * 2);
905  if (newchrs == NULL)
906  goto out_of_memory;
907  pcc->cv.ranges = newchrs;
908  pcc->cv.rangespace = pcc->cv.nranges;
909  }
910 
911  /*
912  * Success, link it into cache chain
913  */
914  pcc->next = pg_ctype_cache_list;
915  pg_ctype_cache_list = pcc;
916 
917  return &pcc->cv;
918 
919  /*
920  * Failure, clean up
921  */
922 out_of_memory:
923  if (pcc->cv.chrs)
924  free(pcc->cv.chrs);
925  if (pcc->cv.ranges)
926  free(pcc->cv.ranges);
927  free(pcc);
928 
929  return NULL;
930 }
#define ispunct_l
Definition: win32.h:342
#define COLLPROVIDER_ICU
Definition: pg_collation.h:85
static int pg_wc_ispunct(pg_wchar c)
#define PG_ISALPHA
#define isspace_l
Definition: win32.h:344
#define PG_ISPRINT
int errhint(const char *fmt,...)
Definition: elog.c:987
struct cvec cv
#define toupper_l
Definition: win32.h:325
#define iswalnum_l
Definition: win32.h:333
static const unsigned char pg_char_properties[128]
int(* pg_wc_probefunc)(pg_wchar c)
static pg_locale_t pg_regex_locale
#define iswprint_l
Definition: win32.h:341
pg_wc_probefunc probefunc
static Oid pg_regex_collation
static PG_Locale_Strategy pg_regex_strategy
unsigned char pg_ascii_tolower(unsigned char ch)
Definition: pgstrcasecmp.c:146
#define isprint_l
Definition: win32.h:340
union pg_locale_struct::@120 info
#define PG_ISGRAPH
int errcode(int sqlerrcode)
Definition: elog.c:575
unsigned char pg_ascii_toupper(unsigned char ch)
Definition: pgstrcasecmp.c:135
static pg_wchar pg_wc_toupper(pg_wchar c)
#define isdigit_l
Definition: win32.h:328
unsigned int Oid
Definition: postgres_ext.h:31
int chrspace
Definition: regguts.h:259
#define OidIsValid(objectId)
Definition: c.h:532
#define iswupper_l
Definition: win32.h:335
#define malloc(a)
Definition: header.h:50
#define iswalpha_l
Definition: win32.h:331
#define isupper_l
Definition: win32.h:334
#define iswpunct_l
Definition: win32.h:343
pg_wchar chr
Definition: regcustom.h:68
chr * ranges
Definition: regguts.h:263
#define PG_ISSPACE
#define ERROR
Definition: elog.h:43
static int pg_wc_isalpha(pg_wchar c)
int nchrs
Definition: regguts.h:258
int cclasscode
Definition: regguts.h:264
#define PG_ISUPPER
static int pg_wc_isspace(pg_wchar c)
#define PG_ISALNUM
PG_Locale_Strategy
#define iswdigit_l
Definition: win32.h:329
char * c
#define DEFAULT_COLLATION_OID
Definition: pg_collation.h:75
#define assert(TEST)
Definition: imath.c:37
struct pg_ctype_cache * next
#define isalnum_l
Definition: win32.h:332
static struct cvec * pg_ctype_get_cache(pg_wc_probefunc probefunc, int cclasscode)
#define ereport(elevel, rest)
Definition: elog.h:122
unsigned int pg_wchar
Definition: mbprint.c:31
static int pg_wc_islower(pg_wchar c)
#define iswgraph_l
Definition: win32.h:339
Definition: regguts.h:256
static pg_wchar pg_wc_tolower(pg_wchar c)
static int pg_wc_isalnum(pg_wchar c)
#define iswlower_l
Definition: win32.h:337
static pg_ctype_cache * pg_ctype_cache_list
pg_locale_t pg_newlocale_from_collation(Oid collid)
Definition: pg_locale.c:1275
#define PG_ISPUNCT
#define iswspace_l
Definition: win32.h:345
#define islower_l
Definition: win32.h:336
int GetDatabaseEncoding(void)
Definition: mbutils.c:1004
static int pg_wc_isdigit(pg_wchar c)
#define free(a)
Definition: header.h:65
chr * chrs
Definition: regguts.h:260
void pg_set_regex_collation(Oid collation)
#define PG_ISLOWER
int rangespace
Definition: regguts.h:262
static bool store_match(pg_ctype_cache *pcc, pg_wchar chr1, int nchrs)
#define C_COLLATION_OID
Definition: pg_collation.h:78
#define realloc(a, b)
Definition: header.h:60
#define tolower_l
Definition: win32.h:324
#define isalpha_l
Definition: win32.h:330
#define towlower_l
Definition: win32.h:326
static int pg_wc_isprint(pg_wchar c)
int errmsg(const char *fmt,...)
Definition: elog.c:797
#define isgraph_l
Definition: win32.h:338
#define PG_ISDIGIT
bool lc_ctype_is_c(Oid collation)
Definition: pg_locale.c:1178
struct pg_ctype_cache pg_ctype_cache
static int pg_wc_isupper(pg_wchar c)
static int pg_wc_isgraph(pg_wchar c)
int nranges
Definition: regguts.h:261
#define towupper_l
Definition: win32.h:327
#define MAX_SIMPLE_CHR
Definition: regcustom.h:97