Age Owner TLA Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * regexp.c
4 : * Postgres' interface to the regular expression package.
5 : *
6 : * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : *
10 : * IDENTIFICATION
11 : * src/backend/utils/adt/regexp.c
12 : *
13 : * Alistair Crooks added the code for the regex caching
14 : * agc - cached the regular expressions used - there's a good chance
15 : * that we'll get a hit, so this saves a compile step for every
16 : * attempted match. I haven't actually measured the speed improvement,
17 : * but it `looks' a lot quicker visually when watching regression
18 : * test output.
19 : *
20 : * agc - incorporated Keith Bostic's Berkeley regex code into
21 : * the tree for all ports. To distinguish this regex code from any that
22 : * is existent on a platform, I've prepended the string "pg_" to
23 : * the functions regcomp, regerror, regexec and regfree.
24 : * Fixed a bug that was originally a typo by me, where `i' was used
25 : * instead of `oldest' when compiling regular expressions - benign
26 : * results mostly, although occasionally it bit you...
27 : *
28 : *-------------------------------------------------------------------------
29 : */
30 : #include "postgres.h"
31 :
32 : #include "catalog/pg_type.h"
33 : #include "funcapi.h"
34 : #include "miscadmin.h"
35 : #include "regex/regex.h"
36 : #include "utils/array.h"
37 : #include "utils/builtins.h"
38 : #include "utils/memutils.h"
39 : #include "utils/varlena.h"
40 :
41 : #define PG_GETARG_TEXT_PP_IF_EXISTS(_n) \
42 : (PG_NARGS() > (_n) ? PG_GETARG_TEXT_PP(_n) : NULL)
43 :
44 :
45 : /* all the options of interest for regex functions */
46 : typedef struct pg_re_flags
47 : {
48 : int cflags; /* compile flags for Spencer's regex code */
49 : bool glob; /* do it globally (for each occurrence) */
50 : } pg_re_flags;
51 :
52 : /* cross-call state for regexp_match and regexp_split functions */
53 : typedef struct regexp_matches_ctx
54 : {
55 : text *orig_str; /* data string in original TEXT form */
56 : int nmatches; /* number of places where pattern matched */
57 : int npatterns; /* number of capturing subpatterns */
58 : /* We store start char index and end+1 char index for each match */
59 : /* so the number of entries in match_locs is nmatches * npatterns * 2 */
60 : int *match_locs; /* 0-based character indexes */
61 : int next_match; /* 0-based index of next match to process */
62 : /* workspace for build_regexp_match_result() */
63 : Datum *elems; /* has npatterns elements */
64 : bool *nulls; /* has npatterns elements */
65 : pg_wchar *wide_str; /* wide-char version of original string */
66 : char *conv_buf; /* conversion buffer, if needed */
67 : int conv_bufsiz; /* size thereof */
68 : } regexp_matches_ctx;
69 :
70 : /*
71 : * We cache precompiled regular expressions using a "self organizing list"
72 : * structure, in which recently-used items tend to be near the front.
73 : * Whenever we use an entry, it's moved up to the front of the list.
74 : * Over time, an item's average position corresponds to its frequency of use.
75 : *
76 : * When we first create an entry, it's inserted at the front of
77 : * the array, dropping the entry at the end of the array if necessary to
78 : * make room. (This might seem to be weighting the new entry too heavily,
79 : * but if we insert new entries further back, we'll be unable to adjust to
80 : * a sudden shift in the query mix where we are presented with MAX_CACHED_RES
81 : * never-before-seen items used circularly. We ought to be able to handle
82 : * that case, so we have to insert at the front.)
83 : *
84 : * Knuth mentions a variant strategy in which a used item is moved up just
85 : * one place in the list. Although he says this uses fewer comparisons on
86 : * average, it seems not to adapt very well to the situation where you have
87 : * both some reusable patterns and a steady stream of non-reusable patterns.
88 : * A reusable pattern that isn't used at least as often as non-reusable
89 : * patterns are seen will "fail to keep up" and will drop off the end of the
90 : * cache. With move-to-front, a reusable pattern is guaranteed to stay in
91 : * the cache as long as it's used at least once in every MAX_CACHED_RES uses.
92 : */
93 :
94 : /* this is the maximum number of cached regular expressions */
95 : #ifndef MAX_CACHED_RES
96 : #define MAX_CACHED_RES 32
97 : #endif
98 :
99 : /* A parent memory context for regular expressions. */
100 : static MemoryContext RegexpCacheMemoryContext;
101 :
102 : /* this structure describes one cached regular expression */
103 : typedef struct cached_re_str
104 : {
105 : MemoryContext cre_context; /* memory context for this regexp */
106 : char *cre_pat; /* original RE (not null terminated!) */
107 : int cre_pat_len; /* length of original RE, in bytes */
108 : int cre_flags; /* compile flags: extended,icase etc */
109 : Oid cre_collation; /* collation to use */
110 : regex_t cre_re; /* the compiled regular expression */
111 : } cached_re_str;
112 :
113 : static int num_res = 0; /* # of cached re's */
114 : static cached_re_str re_array[MAX_CACHED_RES]; /* cached re's */
115 :
116 :
117 : /* Local functions */
118 : static regexp_matches_ctx *setup_regexp_matches(text *orig_str, text *pattern,
119 : pg_re_flags *re_flags,
120 : int start_search,
121 : Oid collation,
122 : bool use_subpatterns,
123 : bool ignore_degenerate,
124 : bool fetching_unmatched);
125 : static ArrayType *build_regexp_match_result(regexp_matches_ctx *matchctx);
126 : static Datum build_regexp_split_result(regexp_matches_ctx *splitctx);
127 :
128 :
129 : /*
130 : * RE_compile_and_cache - compile a RE, caching if possible
131 : *
132 : * Returns regex_t *
133 : *
134 : * text_re --- the pattern, expressed as a TEXT object
135 : * cflags --- compile options for the pattern
136 : * collation --- collation to use for LC_CTYPE-dependent behavior
137 : *
138 : * Pattern is given in the database encoding. We internally convert to
139 : * an array of pg_wchar, which is what Spencer's regex package wants.
140 : */
141 : regex_t *
4382 tgl 142 GIC 459737 : RE_compile_and_cache(text *text_re, int cflags, Oid collation)
143 : {
5679 144 459737 : int text_re_len = VARSIZE_ANY_EXHDR(text_re);
145 459737 : char *text_re_val = VARDATA_ANY(text_re);
7368 tgl 146 ECB : pg_wchar *pattern;
147 : int pattern_len;
9344 bruce 148 : int i;
149 : int regcomp_result;
150 : cached_re_str re_temp;
151 : char errMsg[100];
152 : MemoryContext oldcontext;
153 :
154 : /*
155 : * Look for a match among previously compiled REs. Since the data
156 : * structure is self-organizing with most-used entries at the front, our
157 : * search strategy can just be to scan from the front.
158 : */
7368 tgl 159 GIC 728658 : for (i = 0; i < num_res; i++)
160 : {
5679 161 725642 : if (re_array[i].cre_pat_len == text_re_len &&
162 465084 : re_array[i].cre_flags == cflags &&
4382 163 464464 : re_array[i].cre_collation == collation &&
5679 tgl 164 CBC 464370 : memcmp(re_array[i].cre_pat, text_re_val, text_re_len) == 0)
165 : {
7368 tgl 166 ECB : /*
167 : * Found a match; move it to front if not there already.
168 : */
7368 tgl 169 CBC 456721 : if (i > 0)
170 : {
7368 tgl 171 GIC 224146 : re_temp = re_array[i];
172 224146 : memmove(&re_array[1], &re_array[0], i * sizeof(cached_re_str));
173 224146 : re_array[0] = re_temp;
7607 lockhart 174 ECB : }
175 :
6382 tgl 176 CBC 456721 : return &re_array[0].cre_re;
9345 bruce 177 ECB : }
9770 scrappy 178 : }
179 :
180 : /* Set up the cache memory on first go through. */
1 tmunro 181 GNC 3016 : if (unlikely(RegexpCacheMemoryContext == NULL))
182 795 : RegexpCacheMemoryContext =
183 795 : AllocSetContextCreate(TopMemoryContext,
184 : "RegexpCacheMemoryContext",
185 : ALLOCSET_SMALL_SIZES);
186 :
187 : /*
7368 tgl 188 ECB : * Couldn't find it, so try to compile the new RE. To avoid leaking
189 : * resources on failure, we build into the re_temp local.
190 : */
191 :
192 : /* Convert pattern string to wide characters */
5679 tgl 193 CBC 3016 : pattern = (pg_wchar *) palloc((text_re_len + 1) * sizeof(pg_wchar));
194 3016 : pattern_len = pg_mb2wchar_with_len(text_re_val,
7368 tgl 195 ECB : pattern,
196 : text_re_len);
197 :
198 : /*
199 : * Make a memory context for this compiled regexp. This is initially a
200 : * child of the current memory context, so it will be cleaned up
201 : * automatically if compilation is interrupted and throws an ERROR. We'll
202 : * re-parent it under the longer lived cache context if we make it to the
203 : * bottom of this function.
204 : */
1 tmunro 205 GNC 3016 : re_temp.cre_context = AllocSetContextCreate(CurrentMemoryContext,
206 : "RegexpMemoryContext",
207 : ALLOCSET_SMALL_SIZES);
208 3016 : oldcontext = MemoryContextSwitchTo(re_temp.cre_context);
209 :
7368 tgl 210 GIC 3016 : regcomp_result = pg_regcomp(&re_temp.cre_re,
211 : pattern,
212 : pattern_len,
213 : cflags,
214 : collation);
215 :
216 3004 : pfree(pattern);
7368 tgl 217 ECB :
6710 tgl 218 CBC 3004 : if (regcomp_result != REG_OKAY)
219 : {
220 : /* re didn't compile (no need for pg_regfree, if so) */
7368 tgl 221 GIC 18 : pg_regerror(regcomp_result, &re_temp.cre_re, errMsg, sizeof(errMsg));
7196 222 18 : ereport(ERROR,
7196 tgl 223 ECB : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
224 : errmsg("invalid regular expression: %s", errMsg)));
9770 scrappy 225 : }
226 :
227 : /* Copy the pattern into the per-regexp memory context. */
1 tmunro 228 GNC 2986 : re_temp.cre_pat = palloc(text_re_len + 1);
1 tmunro 229 GIC 2986 : memcpy(re_temp.cre_pat, text_re_val, text_re_len);
230 :
231 : /*
232 : * NUL-terminate it only for the benefit of the identifier used for the
233 : * memory context, visible in the pg_backend_memory_contexts view.
234 : */
1 tmunro 235 GNC 2986 : re_temp.cre_pat[text_re_len] = 0;
236 2986 : MemoryContextSetIdentifier(re_temp.cre_context, re_temp.cre_pat);
237 :
5679 tgl 238 GIC 2986 : re_temp.cre_pat_len = text_re_len;
7368 tgl 239 CBC 2986 : re_temp.cre_flags = cflags;
4382 240 2986 : re_temp.cre_collation = collation;
241 :
242 : /*
243 : * Okay, we have a valid new item in re_temp; insert it into the storage
244 : * array. Discard last entry if needed.
245 : */
7368 246 2986 : if (num_res >= MAX_CACHED_RES)
7368 tgl 247 ECB : {
7368 tgl 248 GIC 366 : --num_res;
7368 tgl 249 CBC 366 : Assert(num_res < MAX_CACHED_RES);
250 : /* Delete the memory context holding the regexp and pattern. */
1 tmunro 251 GNC 366 : MemoryContextDelete(re_array[num_res].cre_context);
252 : }
253 :
254 : /* Re-parent the memory context to our long-lived cache context. */
255 2986 : MemoryContextSetParent(re_temp.cre_context, RegexpCacheMemoryContext);
256 :
7368 tgl 257 GIC 2986 : if (num_res > 0)
258 2191 : memmove(&re_array[1], &re_array[0], num_res * sizeof(cached_re_str));
259 :
7368 tgl 260 CBC 2986 : re_array[0] = re_temp;
7368 tgl 261 GIC 2986 : num_res++;
9345 bruce 262 ECB :
1 tmunro 263 GNC 2986 : MemoryContextSwitchTo(oldcontext);
264 :
6382 tgl 265 CBC 2986 : return &re_array[0].cre_re;
266 : }
6482 bruce 267 ECB :
268 : /*
269 : * RE_wchar_execute - execute a RE on pg_wchar data
270 : *
2062 peter_e 271 : * Returns true on match, false on no match
272 : *
5864 neilc 273 : * re --- the compiled pattern as returned by RE_compile_and_cache
274 : * data --- the data to match against (need not be null-terminated)
275 : * data_len --- the length of the data string
276 : * start_search -- the offset in the data to start searching
6482 bruce 277 : * nmatch, pmatch --- optional return area for match details
278 : *
5864 neilc 279 : * Data is given as array of pg_wchar which is what Spencer's regex package
280 : * wants.
6482 bruce 281 : */
282 : static bool
5864 neilc 283 GIC 893861 : RE_wchar_execute(regex_t *re, pg_wchar *data, int data_len,
284 : int start_search, int nmatch, regmatch_t *pmatch)
285 : {
286 : int regexec_result;
287 : char errMsg[100];
288 :
289 : /* Perform RE match and return result */
6382 tgl 290 893861 : regexec_result = pg_regexec(re,
291 : data,
292 : data_len,
293 : start_search,
294 : NULL, /* no details */
295 : nmatch,
296 : pmatch,
297 : 0);
298 :
6710 tgl 299 CBC 893861 : if (regexec_result != REG_OKAY && regexec_result != REG_NOMATCH)
300 : {
301 : /* re failed??? */
6382 tgl 302 UIC 0 : pg_regerror(regexec_result, re, errMsg, sizeof(errMsg));
6710 303 0 : ereport(ERROR,
304 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
6710 tgl 305 ECB : errmsg("regular expression failed: %s", errMsg)));
306 : }
307 :
6710 tgl 308 GIC 893861 : return (regexec_result == REG_OKAY);
309 : }
310 :
311 : /*
312 : * RE_execute - execute a RE
313 : *
2062 peter_e 314 ECB : * Returns true on match, false on no match
315 : *
316 : * re --- the compiled pattern as returned by RE_compile_and_cache
5864 neilc 317 EUB : * dat --- the data to match against (need not be null-terminated)
318 : * dat_len --- the length of the data string
319 : * nmatch, pmatch --- optional return area for match details
320 : *
321 : * Data is given in the database encoding. We internally
322 : * convert to array of pg_wchar which is what Spencer's regex package wants.
5864 neilc 323 ECB : */
324 : static bool
5864 neilc 325 GIC 346274 : RE_execute(regex_t *re, char *dat, int dat_len,
326 : int nmatch, regmatch_t *pmatch)
327 : {
328 : pg_wchar *data;
329 : int data_len;
330 : bool match;
331 :
332 : /* Convert data string to wide characters */
333 346274 : data = (pg_wchar *) palloc((dat_len + 1) * sizeof(pg_wchar));
334 346274 : data_len = pg_mb2wchar_with_len(dat, data, dat_len);
335 :
336 : /* Perform RE match and return result */
337 346274 : match = RE_wchar_execute(re, data, data_len, 0, nmatch, pmatch);
338 :
339 346274 : pfree(data);
5864 neilc 340 CBC 346274 : return match;
341 : }
342 :
343 : /*
344 : * RE_compile_and_execute - compile and execute a RE
345 : *
346 : * Returns true on match, false on no match
347 : *
5679 tgl 348 ECB : * text_re --- the pattern, expressed as a TEXT object
5864 neilc 349 : * dat --- the data to match against (need not be null-terminated)
350 : * dat_len --- the length of the data string
351 : * cflags --- compile options for the pattern
4382 tgl 352 : * collation --- collation to use for LC_CTYPE-dependent behavior
353 : * nmatch, pmatch --- optional return area for match details
5864 neilc 354 : *
355 : * Both pattern and data are given in the database encoding. We internally
356 : * convert to array of pg_wchar which is what Spencer's regex package wants.
357 : */
358 : bool
5864 neilc 359 GIC 345212 : RE_compile_and_execute(text *text_re, char *dat, int dat_len,
360 : int cflags, Oid collation,
361 : int nmatch, regmatch_t *pmatch)
362 : {
363 : regex_t *re;
364 :
365 : /* Use REG_NOSUB if caller does not want sub-match details */
608 tgl 366 345212 : if (nmatch < 2)
367 345212 : cflags |= REG_NOSUB;
368 :
369 : /* Compile RE */
4382 370 345212 : re = RE_compile_and_cache(text_re, cflags, collation);
371 :
5864 neilc 372 345200 : return RE_execute(re, dat, dat_len, nmatch, pmatch);
373 : }
5864 neilc 374 ECB :
375 :
376 : /*
377 : * parse_re_flags - parse the options argument of regexp_match and friends
378 : *
379 : * flags --- output argument, filled with desired options
380 : * opts --- TEXT object, or NULL for defaults
5720 tgl 381 : *
382 : * This accepts all the options allowed by any of the callers; callers that
383 : * don't want some have to reject them after the fact.
384 : */
5864 neilc 385 : static void
5624 bruce 386 GIC 102964 : parse_re_flags(pg_re_flags *flags, text *opts)
5864 neilc 387 ECB : {
388 : /* regex flavor is always folded into the compile flags */
4918 tgl 389 GIC 102964 : flags->cflags = REG_ADVANCED;
5720 390 102964 : flags->glob = false;
391 :
5864 neilc 392 102964 : if (opts)
393 : {
5624 bruce 394 1370 : char *opt_p = VARDATA_ANY(opts);
395 1370 : int opt_len = VARSIZE_ANY_EXHDR(opts);
396 : int i;
397 :
5864 neilc 398 3389 : for (i = 0; i < opt_len; i++)
399 : {
400 2031 : switch (opt_p[i])
5864 neilc 401 ECB : {
5864 neilc 402 GIC 1209 : case 'g':
403 1209 : flags->glob = true;
5864 neilc 404 CBC 1209 : break;
5624 bruce 405 LBC 0 : case 'b': /* BREs (but why???) */
5720 tgl 406 UIC 0 : flags->cflags &= ~(REG_ADVANCED | REG_EXTENDED | REG_QUOTE);
5720 tgl 407 LBC 0 : break;
5624 bruce 408 GIC 5 : case 'c': /* case sensitive */
5720 tgl 409 CBC 5 : flags->cflags &= ~REG_ICASE;
410 5 : break;
5624 bruce 411 UIC 0 : case 'e': /* plain EREs */
5720 tgl 412 0 : flags->cflags |= REG_EXTENDED;
5720 tgl 413 LBC 0 : flags->cflags &= ~(REG_ADVANCED | REG_QUOTE);
5720 tgl 414 UIC 0 : break;
5624 bruce 415 CBC 146 : case 'i': /* case insensitive */
5864 neilc 416 GIC 146 : flags->cflags |= REG_ICASE;
5864 neilc 417 CBC 146 : break;
5624 bruce 418 650 : case 'm': /* Perloid synonym for n */
5624 bruce 419 ECB : case 'n': /* \n affects ^ $ . [^ */
5864 neilc 420 GBC 650 : flags->cflags |= REG_NEWLINE;
421 650 : break;
5624 bruce 422 UBC 0 : case 'p': /* ~Perl, \n affects . [^ */
5864 neilc 423 LBC 0 : flags->cflags |= REG_NLSTOP;
424 0 : flags->cflags &= ~REG_NLANCH;
425 0 : break;
5624 bruce 426 UBC 0 : case 'q': /* literal string */
5720 tgl 427 0 : flags->cflags |= REG_QUOTE;
428 0 : flags->cflags &= ~(REG_ADVANCED | REG_EXTENDED);
429 0 : break;
5624 bruce 430 CBC 6 : case 's': /* single line, \n ordinary */
5720 tgl 431 6 : flags->cflags &= ~REG_NEWLINE;
432 6 : break;
5624 bruce 433 LBC 0 : case 't': /* tight syntax */
5720 tgl 434 UIC 0 : flags->cflags &= ~REG_EXPANDED;
5720 tgl 435 LBC 0 : break;
5624 bruce 436 0 : case 'w': /* weird, \n affects ^ $ only */
5864 neilc 437 UBC 0 : flags->cflags &= ~REG_NLSTOP;
438 0 : flags->cflags |= REG_NLANCH;
439 0 : break;
5624 bruce 440 GBC 3 : case 'x': /* expanded syntax */
5864 neilc 441 3 : flags->cflags |= REG_EXPANDED;
442 3 : break;
443 12 : default:
444 12 : ereport(ERROR,
5864 neilc 445 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1014 tgl 446 : errmsg("invalid regular expression option: \"%.*s\"",
447 : pg_mblen(opt_p + i), opt_p + i)));
5864 neilc 448 EUB : break;
449 : }
450 : }
451 : }
5864 neilc 452 GBC 102952 : }
5864 neilc 453 EUB :
9770 scrappy 454 :
9770 scrappy 455 ECB : /*
8312 tgl 456 : * interface routines called by the function manager
9770 scrappy 457 : */
8312 tgl 458 :
459 : Datum
8312 tgl 460 GIC 127719 : nameregexeq(PG_FUNCTION_ARGS)
461 : {
462 127719 : Name n = PG_GETARG_NAME(0);
5679 463 127719 : text *p = PG_GETARG_TEXT_PP(1);
464 :
7368 465 127719 : PG_RETURN_BOOL(RE_compile_and_execute(p,
466 : NameStr(*n),
7368 tgl 467 ECB : strlen(NameStr(*n)),
468 : REG_ADVANCED,
469 : PG_GET_COLLATION(),
470 : 0, NULL));
471 : }
472 :
473 : Datum
8312 tgl 474 GIC 6782 : nameregexne(PG_FUNCTION_ARGS)
9770 scrappy 475 ECB : {
8312 tgl 476 GIC 6782 : Name n = PG_GETARG_NAME(0);
5679 tgl 477 CBC 6782 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 478 ECB :
7368 tgl 479 GIC 6782 : PG_RETURN_BOOL(!RE_compile_and_execute(p,
6406 tgl 480 ECB : NameStr(*n),
481 : strlen(NameStr(*n)),
482 : REG_ADVANCED,
483 : PG_GET_COLLATION(),
484 : 0, NULL));
485 : }
486 :
487 : Datum
8312 tgl 488 GIC 186505 : textregexeq(PG_FUNCTION_ARGS)
9770 scrappy 489 ECB : {
5679 tgl 490 GIC 186505 : text *s = PG_GETARG_TEXT_PP(0);
5679 tgl 491 CBC 186505 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 492 ECB :
7368 tgl 493 GIC 186505 : PG_RETURN_BOOL(RE_compile_and_execute(p,
5679 tgl 494 ECB : VARDATA_ANY(s),
495 : VARSIZE_ANY_EXHDR(s),
496 : REG_ADVANCED,
497 : PG_GET_COLLATION(),
498 : 0, NULL));
499 : }
500 :
501 : Datum
8312 tgl 502 GIC 17058 : textregexne(PG_FUNCTION_ARGS)
9770 scrappy 503 ECB : {
5679 tgl 504 GIC 17058 : text *s = PG_GETARG_TEXT_PP(0);
5679 tgl 505 CBC 17058 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 506 ECB :
7368 tgl 507 GIC 17058 : PG_RETURN_BOOL(!RE_compile_and_execute(p,
5679 tgl 508 ECB : VARDATA_ANY(s),
509 : VARSIZE_ANY_EXHDR(s),
510 : REG_ADVANCED,
511 : PG_GET_COLLATION(),
512 : 0, NULL));
513 : }
514 :
515 :
516 : /*
8053 bruce 517 : * routines that use the regexp stuff, but ignore the case.
518 : * for this, we use the REG_ICASE flag to pg_regcomp
9770 scrappy 519 : */
8312 tgl 520 :
521 :
522 : Datum
7368 tgl 523 GIC 6853 : nameicregexeq(PG_FUNCTION_ARGS)
524 : {
525 6853 : Name n = PG_GETARG_NAME(0);
5679 526 6853 : text *p = PG_GETARG_TEXT_PP(1);
527 :
7368 528 6853 : PG_RETURN_BOOL(RE_compile_and_execute(p,
529 : NameStr(*n),
530 : strlen(NameStr(*n)),
531 : REG_ADVANCED | REG_ICASE,
532 : PG_GET_COLLATION(),
533 : 0, NULL));
534 : }
535 :
536 : Datum
537 3 : nameicregexne(PG_FUNCTION_ARGS)
9770 scrappy 538 ECB : {
7368 tgl 539 GIC 3 : Name n = PG_GETARG_NAME(0);
5679 tgl 540 CBC 3 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 541 ECB :
7368 tgl 542 GIC 3 : PG_RETURN_BOOL(!RE_compile_and_execute(p,
6406 tgl 543 ECB : NameStr(*n),
544 : strlen(NameStr(*n)),
545 : REG_ADVANCED | REG_ICASE,
546 : PG_GET_COLLATION(),
547 : 0, NULL));
548 : }
549 :
550 : Datum
7368 tgl 551 GIC 134 : texticregexeq(PG_FUNCTION_ARGS)
9770 scrappy 552 ECB : {
5679 tgl 553 GIC 134 : text *s = PG_GETARG_TEXT_PP(0);
5679 tgl 554 CBC 134 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 555 ECB :
7368 tgl 556 GIC 134 : PG_RETURN_BOOL(RE_compile_and_execute(p,
5679 tgl 557 ECB : VARDATA_ANY(s),
558 : VARSIZE_ANY_EXHDR(s),
559 : REG_ADVANCED | REG_ICASE,
560 : PG_GET_COLLATION(),
561 : 0, NULL));
562 : }
563 :
564 : Datum
7368 tgl 565 GIC 8 : texticregexne(PG_FUNCTION_ARGS)
9770 scrappy 566 ECB : {
5679 tgl 567 GIC 8 : text *s = PG_GETARG_TEXT_PP(0);
5679 tgl 568 CBC 8 : text *p = PG_GETARG_TEXT_PP(1);
8312 tgl 569 ECB :
7368 tgl 570 GIC 8 : PG_RETURN_BOOL(!RE_compile_and_execute(p,
5679 tgl 571 ECB : VARDATA_ANY(s),
572 : VARSIZE_ANY_EXHDR(s),
573 : REG_ADVANCED | REG_ICASE,
574 : PG_GET_COLLATION(),
575 : 0, NULL));
576 : }
577 :
578 :
579 : /*
7368 580 : * textregexsubstr()
581 : * Return a substring matched by a regular expression.
7607 lockhart 582 : */
583 : Datum
7607 lockhart 584 GIC 1074 : textregexsubstr(PG_FUNCTION_ARGS)
7607 lockhart 585 ECB : {
5679 tgl 586 GIC 1074 : text *s = PG_GETARG_TEXT_PP(0);
587 1074 : text *p = PG_GETARG_TEXT_PP(1);
588 : regex_t *re;
589 : regmatch_t pmatch[2];
590 : int so,
591 : eo;
592 :
593 : /* Compile RE */
4382 594 1074 : re = RE_compile_and_cache(p, REG_ADVANCED, PG_GET_COLLATION());
595 :
596 : /*
597 : * We pass two regmatch_t structs to get info about the overall match and
598 : * the match for the first parenthesized subexpression (if any). If there
6385 bruce 599 ECB : * is a parenthesized subexpression, we return what it matched; else
600 : * return what the whole regexp matched.
7607 lockhart 601 : */
5499 tgl 602 CBC 2148 : if (!RE_execute(re,
5499 tgl 603 GIC 2148 : VARDATA_ANY(s), VARSIZE_ANY_EXHDR(s),
604 : 2, pmatch))
605 3 : PG_RETURN_NULL(); /* definitely no match */
606 :
607 1071 : if (re->re_nsub > 0)
608 : {
5499 tgl 609 ECB : /* has parenthesized subexpressions, use the first one */
7504 tgl 610 GIC 761 : so = pmatch[1].rm_so;
611 761 : eo = pmatch[1].rm_eo;
612 : }
613 : else
614 : {
615 : /* no parenthesized subexpression, use whole match */
5499 616 310 : so = pmatch[0].rm_so;
5499 tgl 617 CBC 310 : eo = pmatch[0].rm_eo;
5499 tgl 618 ECB : }
619 :
620 : /*
621 : * It is possible to have a match to the whole pattern but no match for a
5050 bruce 622 : * subexpression; for example 'foo(bar)?' is considered to match 'foo' but
623 : * there is no subexpression match. So this extra test for match failure
624 : * is not redundant.
5499 tgl 625 : */
5499 tgl 626 CBC 1071 : if (so < 0 || eo < 0)
5499 tgl 627 GIC 3 : PG_RETURN_NULL();
628 :
629 1068 : return DirectFunctionCall3(text_substr,
630 : PointerGetDatum(s),
5499 tgl 631 ECB : Int32GetDatum(so + 1),
632 : Int32GetDatum(eo - so));
633 : }
634 :
635 : /*
636 : * textregexreplace_noopt()
637 : * Return a string matched by a regular expression, with replacement.
638 : *
639 : * This version doesn't have an option argument: we default to case
640 : * sensitive match, replace the first instance only.
6482 bruce 641 : */
642 : Datum
6482 bruce 643 GIC 4069 : textregexreplace_noopt(PG_FUNCTION_ARGS)
6482 bruce 644 ECB : {
5679 tgl 645 GIC 4069 : text *s = PG_GETARG_TEXT_PP(0);
646 4069 : text *p = PG_GETARG_TEXT_PP(1);
647 4069 : text *r = PG_GETARG_TEXT_PP(2);
648 :
608 649 4069 : PG_RETURN_TEXT_P(replace_text_regexp(s, p, r,
650 : REG_ADVANCED, PG_GET_COLLATION(),
651 : 0, 1));
652 : }
653 :
654 : /*
655 : * textregexreplace()
656 : * Return a string matched by a regular expression, with replacement.
657 : */
6482 bruce 658 ECB : Datum
6482 bruce 659 GIC 1173 : textregexreplace(PG_FUNCTION_ARGS)
6482 bruce 660 ECB : {
5679 tgl 661 CBC 1173 : text *s = PG_GETARG_TEXT_PP(0);
662 1173 : text *p = PG_GETARG_TEXT_PP(1);
5679 tgl 663 GIC 1173 : text *r = PG_GETARG_TEXT_PP(2);
5679 tgl 664 CBC 1173 : text *opt = PG_GETARG_TEXT_PP(3);
665 : pg_re_flags flags;
666 :
667 : /*
668 : * regexp_replace() with four arguments will be preferentially resolved as
669 : * this form when the fourth argument is of type UNKNOWN. However, the
670 : * user might have intended to call textregexreplace_extended_no_n. If we
671 : * see flags that look like an integer, emit the same error that
672 : * parse_re_flags would, but add a HINT about how to fix it.
673 : */
614 674 1173 : if (VARSIZE_ANY_EXHDR(opt) > 0)
675 : {
676 1173 : char *opt_p = VARDATA_ANY(opt);
614 tgl 677 ECB :
614 tgl 678 CBC 1173 : if (*opt_p >= '0' && *opt_p <= '9')
679 3 : ereport(ERROR,
680 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
681 : errmsg("invalid regular expression option: \"%.*s\"",
682 : pg_mblen(opt_p), opt_p),
683 : errhint("If you meant to use regexp_replace() with a start parameter, cast the fourth argument to integer explicitly.")));
684 : }
685 :
5720 tgl 686 GIC 1170 : parse_re_flags(&flags, opt);
687 :
608 688 1167 : PG_RETURN_TEXT_P(replace_text_regexp(s, p, r,
608 tgl 689 ECB : flags.cflags, PG_GET_COLLATION(),
690 : 0, flags.glob ? 0 : 1));
614 691 : }
692 :
693 : /*
694 : * textregexreplace_extended()
695 : * Return a string matched by a regular expression, with replacement.
696 : * Extends textregexreplace by allowing a start position and the
697 : * choice of the occurrence to replace (0 means all occurrences).
698 : */
699 : Datum
614 tgl 700 GIC 33 : textregexreplace_extended(PG_FUNCTION_ARGS)
614 tgl 701 ECB : {
614 tgl 702 GIC 33 : text *s = PG_GETARG_TEXT_PP(0);
614 tgl 703 CBC 33 : text *p = PG_GETARG_TEXT_PP(1);
614 tgl 704 GIC 33 : text *r = PG_GETARG_TEXT_PP(2);
705 33 : int start = 1;
706 33 : int n = 1;
707 33 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(5);
708 : pg_re_flags re_flags;
709 :
710 : /* Collect optional parameters */
711 33 : if (PG_NARGS() > 3)
712 : {
713 33 : start = PG_GETARG_INT32(3);
714 33 : if (start <= 0)
614 tgl 715 CBC 3 : ereport(ERROR,
716 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 717 ECB : errmsg("invalid value for parameter \"%s\": %d",
718 : "start", start)));
719 : }
614 tgl 720 CBC 30 : if (PG_NARGS() > 4)
614 tgl 721 ECB : {
614 tgl 722 CBC 27 : n = PG_GETARG_INT32(4);
614 tgl 723 GIC 27 : if (n < 0)
724 3 : ereport(ERROR,
725 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 726 ECB : errmsg("invalid value for parameter \"%s\": %d",
727 : "n", n)));
728 : }
729 :
730 : /* Determine options */
614 tgl 731 GIC 27 : parse_re_flags(&re_flags, flags);
732 :
733 : /* If N was not specified, deduce it from the 'g' flag */
734 27 : if (PG_NARGS() <= 4)
614 tgl 735 CBC 3 : n = re_flags.glob ? 0 : 1;
736 :
614 tgl 737 ECB : /* Do the replacement(s) */
608 tgl 738 CBC 27 : PG_RETURN_TEXT_P(replace_text_regexp(s, p, r,
608 tgl 739 ECB : re_flags.cflags, PG_GET_COLLATION(),
740 : start - 1, n));
741 : }
742 :
743 : /* This is separate to keep the opr_sanity regression test from complaining */
744 : Datum
614 tgl 745 GIC 3 : textregexreplace_extended_no_n(PG_FUNCTION_ARGS)
614 tgl 746 ECB : {
614 tgl 747 GIC 3 : return textregexreplace_extended(fcinfo);
748 : }
614 tgl 749 ECB :
750 : /* This is separate to keep the opr_sanity regression test from complaining */
751 : Datum
614 tgl 752 GIC 3 : textregexreplace_extended_no_flags(PG_FUNCTION_ARGS)
614 tgl 753 ECB : {
614 tgl 754 GIC 3 : return textregexreplace_extended(fcinfo);
755 : }
756 :
757 : /*
758 : * similar_to_escape(), similar_escape()
759 : *
1310 tgl 760 ECB : * Convert a SQL "SIMILAR TO" regexp pattern to POSIX style, so it can be
761 : * used by our regexp engine.
762 : *
763 : * similar_escape_internal() is the common workhorse for three SQL-exposed
764 : * functions. esc_text can be passed as NULL to select the default escape
765 : * (which is '\'), or as an empty string to select no escape character.
766 : */
767 : static text *
1310 tgl 768 GIC 66 : similar_escape_internal(text *pat_text, text *esc_text)
7504 tgl 769 ECB : {
770 : text *result;
771 : char *p,
772 : *e,
773 : *r;
774 : int plen,
775 : elen;
7504 tgl 776 GIC 66 : bool afterescape = false;
4845 777 66 : bool incharclass = false;
7504 778 66 : int nquotes = 0;
779 :
5679 780 66 : p = VARDATA_ANY(pat_text);
781 66 : plen = VARSIZE_ANY_EXHDR(pat_text);
1310 782 66 : if (esc_text == NULL)
7504 tgl 783 ECB : {
784 : /* No ESCAPE clause provided; default to backslash as escape */
7504 tgl 785 GIC 20 : e = "\\";
786 20 : elen = 1;
787 : }
788 : else
789 : {
5679 790 46 : e = VARDATA_ANY(esc_text);
5679 tgl 791 CBC 46 : elen = VARSIZE_ANY_EXHDR(esc_text);
7504 792 46 : if (elen == 0)
793 3 : e = NULL; /* no escape character */
1310 tgl 794 GIC 43 : else if (elen > 1)
3147 jdavis 795 ECB : {
3147 jdavis 796 CBC 3 : int escape_mblen = pg_mbstrlen_with_len(e, elen);
3147 jdavis 797 ECB :
3147 jdavis 798 GIC 3 : if (escape_mblen > 1)
799 3 : ereport(ERROR,
3147 jdavis 800 ECB : (errcode(ERRCODE_INVALID_ESCAPE_SEQUENCE),
801 : errmsg("invalid escape string"),
802 : errhint("Escape string must be empty or one character.")));
803 : }
804 : }
7504 tgl 805 :
6205 806 : /*----------
807 : * We surround the transformed input string with
4918 808 : * ^(?: ... )$
809 : * which requires some explanation. We need "^" and "$" to force
810 : * the pattern to match the entire input string as per the SQL spec.
811 : * The "(?:" and ")" are a non-capturing set of parens; we have to have
812 : * parens in case the string contains "|", else the "^" and "$" will
6205 813 : * be bound into the first and last alternatives which is not what we
814 : * want, and the parens must be non capturing because we don't want them
815 : * to count when selecting output for SUBSTRING.
816 : *
817 : * When the pattern is divided into three parts by escape-double-quotes,
818 : * what we emit is
819 : * ^(?:part1){1,1}?(part2){1,1}(?:part3)$
820 : * which requires even more explanation. The "{1,1}?" on part1 makes it
821 : * non-greedy so that it will match the smallest possible amount of text
822 : * not the largest, as required by SQL. The plain parens around part2
823 : * are capturing parens so that that part is what controls the result of
824 : * SUBSTRING. The "{1,1}" forces part2 to be greedy, so that it matches
825 : * the largest possible amount of text; hence part3 must match the
826 : * smallest amount of text, as required by SQL. We don't need an explicit
827 : * greediness marker on part3. Note that this also confines the effects
828 : * of any "|" characters to the respective part, which is what we want.
829 : *
830 : * The SQL spec says that SUBSTRING's pattern must contain exactly two
831 : * escape-double-quotes, but we only complain if there's more than two.
832 : * With none, we act as though part1 and part3 are empty; with one, we
833 : * act as though part3 is empty. Both behaviors fall out of omitting
834 : * the relevant part separators in the above expansion. If the result
835 : * of this function is used in a plain regexp match (SIMILAR TO), the
836 : * escape-double-quotes have no effect on the match behavior.
837 : *----------
838 : */
839 :
840 : /*
841 : * We need room for the prefix/postfix and part separators, plus as many
842 : * as 3 output bytes per input byte; since the input is at most 1GB this
843 : * can't overflow size_t.
844 : */
1426 tgl 845 GIC 63 : result = (text *) palloc(VARHDRSZ + 23 + 3 * (size_t) plen);
7504 846 63 : r = VARDATA(result);
847 :
848 63 : *r++ = '^';
6205 849 63 : *r++ = '(';
850 63 : *r++ = '?';
851 63 : *r++ = ':';
852 :
7504 853 479 : while (plen > 0)
854 : {
6385 bruce 855 419 : char pchar = *p;
856 :
857 : /*
858 : * If both the escape character and the current character from the
859 : * pattern are multi-byte, we need to take the slow path.
3147 jdavis 860 ECB : *
861 : * But if one of them is single-byte, we can process the pattern one
862 : * byte at a time, ignoring multi-byte characters. (This works
863 : * because all server-encodings have the property that a valid
864 : * multi-byte character representation cannot contain the
865 : * representation of a valid single-byte character.)
866 : */
867 :
3147 jdavis 868 CBC 419 : if (elen > 1)
869 : {
2878 bruce 870 LBC 0 : int mblen = pg_mblen(p);
871 :
3147 jdavis 872 UIC 0 : if (mblen > 1)
873 : {
874 : /* slow, multi-byte path */
875 0 : if (afterescape)
876 : {
877 0 : *r++ = '\\';
878 0 : memcpy(r, p, mblen);
879 0 : r += mblen;
880 0 : afterescape = false;
881 : }
882 0 : else if (e && elen == mblen && memcmp(e, p, mblen) == 0)
3147 jdavis 883 ECB : {
884 : /* SQL escape character; do not send to output */
3147 jdavis 885 UBC 0 : afterescape = true;
886 : }
3147 jdavis 887 EUB : else
888 : {
889 : /*
890 : * We know it's a multi-byte character, so we don't need
891 : * to do all the comparisons to single-byte characters
892 : * that we do below.
893 : */
3147 jdavis 894 UBC 0 : memcpy(r, p, mblen);
895 0 : r += mblen;
896 : }
3147 jdavis 897 EUB :
3147 jdavis 898 UIC 0 : p += mblen;
899 0 : plen -= mblen;
3147 jdavis 900 EUB :
3147 jdavis 901 UIC 0 : continue;
902 : }
903 : }
904 :
905 : /* fast path */
7504 tgl 906 GIC 419 : if (afterescape)
907 : {
1426 908 71 : if (pchar == '"' && !incharclass) /* escape-double-quote? */
1426 tgl 909 EUB : {
910 : /* emit appropriate part separator, per notes above */
1426 tgl 911 GIC 62 : if (nquotes == 0)
912 : {
1426 tgl 913 GBC 31 : *r++ = ')';
914 31 : *r++ = '{';
1426 tgl 915 GIC 31 : *r++ = '1';
1426 tgl 916 GBC 31 : *r++ = ',';
1426 tgl 917 GIC 31 : *r++ = '1';
918 31 : *r++ = '}';
919 31 : *r++ = '?';
920 31 : *r++ = '(';
1426 tgl 921 ECB : }
1426 tgl 922 GIC 31 : else if (nquotes == 1)
1426 tgl 923 ECB : {
1426 tgl 924 GIC 28 : *r++ = ')';
925 28 : *r++ = '{';
1426 tgl 926 CBC 28 : *r++ = '1';
1426 tgl 927 GIC 28 : *r++ = ',';
1426 tgl 928 CBC 28 : *r++ = '1';
929 28 : *r++ = '}';
930 28 : *r++ = '(';
931 28 : *r++ = '?';
932 28 : *r++ = ':';
1426 tgl 933 ECB : }
934 : else
1426 tgl 935 CBC 3 : ereport(ERROR,
936 : (errcode(ERRCODE_INVALID_USE_OF_ESCAPE_CHARACTER),
1426 tgl 937 ECB : errmsg("SQL regular expression may not contain more than two escape-double-quote separators")));
1426 tgl 938 GIC 59 : nquotes++;
1426 tgl 939 ECB : }
7504 940 : else
941 : {
1426 942 : /*
943 : * We allow any character at all to be escaped; notably, this
944 : * allows access to POSIX character-class escapes such as
945 : * "\d". The SQL spec is considerably more restrictive.
946 : */
7504 tgl 947 CBC 9 : *r++ = '\\';
7504 tgl 948 GIC 9 : *r++ = pchar;
949 : }
7504 tgl 950 CBC 68 : afterescape = false;
951 : }
7504 tgl 952 GIC 348 : else if (e && pchar == *e)
7504 tgl 953 ECB : {
954 : /* SQL escape character; do not send to output */
7504 tgl 955 GIC 71 : afterescape = true;
956 : }
4845 957 277 : else if (incharclass)
958 : {
4845 tgl 959 UIC 0 : if (pchar == '\\')
960 0 : *r++ = '\\';
961 0 : *r++ = pchar;
4845 tgl 962 LBC 0 : if (pchar == ']')
963 0 : incharclass = false;
964 : }
4845 tgl 965 CBC 277 : else if (pchar == '[')
966 : {
4845 tgl 967 LBC 0 : *r++ = pchar;
4845 tgl 968 UIC 0 : incharclass = true;
969 : }
7504 tgl 970 CBC 277 : else if (pchar == '%')
971 : {
972 54 : *r++ = '.';
7504 tgl 973 GIC 54 : *r++ = '*';
7504 tgl 974 EUB : }
7504 tgl 975 GBC 223 : else if (pchar == '_')
976 26 : *r++ = '.';
4845 977 197 : else if (pchar == '(')
4845 tgl 978 EUB : {
979 : /* convert to non-capturing parenthesis */
4845 tgl 980 CBC 9 : *r++ = '(';
4845 tgl 981 GIC 9 : *r++ = '?';
4845 tgl 982 GBC 9 : *r++ = ':';
4845 tgl 983 EUB : }
4929 tgl 984 GIC 188 : else if (pchar == '\\' || pchar == '.' ||
4929 tgl 985 CBC 183 : pchar == '^' || pchar == '$')
986 : {
7504 987 5 : *r++ = '\\';
988 5 : *r++ = pchar;
989 : }
7504 tgl 990 ECB : else
7504 tgl 991 CBC 183 : *r++ = pchar;
992 416 : p++, plen--;
993 : }
994 :
6205 995 60 : *r++ = ')';
7504 996 60 : *r++ = '$';
7188 bruce 997 ECB :
5885 tgl 998 GIC 60 : SET_VARSIZE(result, r - ((char *) result));
7504 tgl 999 ECB :
1310 tgl 1000 CBC 60 : return result;
1001 : }
1310 tgl 1002 ECB :
1003 : /*
1004 : * similar_to_escape(pattern, escape)
1005 : */
1006 : Datum
1310 tgl 1007 CBC 46 : similar_to_escape_2(PG_FUNCTION_ARGS)
1008 : {
1310 tgl 1009 GIC 46 : text *pat_text = PG_GETARG_TEXT_PP(0);
1310 tgl 1010 CBC 46 : text *esc_text = PG_GETARG_TEXT_PP(1);
1310 tgl 1011 ECB : text *result;
1012 :
1310 tgl 1013 CBC 46 : result = similar_escape_internal(pat_text, esc_text);
1014 :
1015 40 : PG_RETURN_TEXT_P(result);
1016 : }
1017 :
1018 : /*
1019 : * similar_to_escape(pattern)
1020 : * Inserts a default escape character.
1021 : */
1310 tgl 1022 ECB : Datum
1310 tgl 1023 GIC 20 : similar_to_escape_1(PG_FUNCTION_ARGS)
1310 tgl 1024 ECB : {
1310 tgl 1025 CBC 20 : text *pat_text = PG_GETARG_TEXT_PP(0);
1026 : text *result;
1027 :
1028 20 : result = similar_escape_internal(pat_text, NULL);
1029 :
1030 20 : PG_RETURN_TEXT_P(result);
1031 : }
1032 :
1033 : /*
1034 : * similar_escape(pattern, escape)
1035 : *
1036 : * Legacy function for compatibility with views stored using the
1037 : * pre-v13 expansion of SIMILAR TO. Unlike the above functions, this
1310 tgl 1038 ECB : * is non-strict, which leads to not-per-spec handling of "ESCAPE NULL".
1039 : */
1040 : Datum
1310 tgl 1041 UIC 0 : similar_escape(PG_FUNCTION_ARGS)
1042 : {
1310 tgl 1043 ECB : text *pat_text;
1044 : text *esc_text;
1045 : text *result;
1046 :
1047 : /* This function is not strict, so must test explicitly */
1310 tgl 1048 UIC 0 : if (PG_ARGISNULL(0))
1049 0 : PG_RETURN_NULL();
1050 0 : pat_text = PG_GETARG_TEXT_PP(0);
1051 :
1052 0 : if (PG_ARGISNULL(1))
1053 0 : esc_text = NULL; /* use default escape character */
1054 : else
1055 0 : esc_text = PG_GETARG_TEXT_PP(1);
1310 tgl 1056 EUB :
1310 tgl 1057 UIC 0 : result = similar_escape_internal(pat_text, esc_text);
1058 :
7504 1059 0 : PG_RETURN_TEXT_P(result);
1060 : }
1061 :
1062 : /*
614 tgl 1063 EUB : * regexp_count()
1064 : * Return the number of matches of a pattern within a string.
1065 : */
1066 : Datum
614 tgl 1067 GBC 24 : regexp_count(PG_FUNCTION_ARGS)
614 tgl 1068 EUB : {
614 tgl 1069 GIC 24 : text *str = PG_GETARG_TEXT_PP(0);
614 tgl 1070 GBC 24 : text *pattern = PG_GETARG_TEXT_PP(1);
614 tgl 1071 GIC 24 : int start = 1;
614 tgl 1072 GBC 24 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(3);
1073 : pg_re_flags re_flags;
614 tgl 1074 EUB : regexp_matches_ctx *matchctx;
1075 :
1076 : /* Collect optional parameters */
614 tgl 1077 GIC 24 : if (PG_NARGS() > 2)
1078 : {
1079 21 : start = PG_GETARG_INT32(2);
1080 21 : if (start <= 0)
1081 6 : ereport(ERROR,
614 tgl 1082 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1083 : errmsg("invalid value for parameter \"%s\": %d",
1084 : "start", start)));
1085 : }
1086 :
1087 : /* Determine options */
614 tgl 1088 GIC 18 : parse_re_flags(&re_flags, flags);
1089 : /* User mustn't specify 'g' */
1090 18 : if (re_flags.glob)
614 tgl 1091 UIC 0 : ereport(ERROR,
614 tgl 1092 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1093 : /* translator: %s is a SQL function name */
1094 : errmsg("%s does not support the \"global\" option",
1095 : "regexp_count()")));
1096 : /* But we find all the matches anyway */
614 tgl 1097 GIC 18 : re_flags.glob = true;
1098 :
1099 : /* Do the matching */
1100 18 : matchctx = setup_regexp_matches(str, pattern, &re_flags, start - 1,
1101 : PG_GET_COLLATION(),
1102 : false, /* can ignore subexprs */
614 tgl 1103 ECB : false, false);
1104 :
614 tgl 1105 CBC 18 : PG_RETURN_INT32(matchctx->nmatches);
614 tgl 1106 EUB : }
1107 :
1108 : /* This is separate to keep the opr_sanity regression test from complaining */
1109 : Datum
614 tgl 1110 GIC 3 : regexp_count_no_start(PG_FUNCTION_ARGS)
1111 : {
614 tgl 1112 CBC 3 : return regexp_count(fcinfo);
1113 : }
1114 :
614 tgl 1115 ECB : /* This is separate to keep the opr_sanity regression test from complaining */
1116 : Datum
614 tgl 1117 GIC 15 : regexp_count_no_flags(PG_FUNCTION_ARGS)
1118 : {
1119 15 : return regexp_count(fcinfo);
614 tgl 1120 ECB : }
1121 :
1122 : /*
1123 : * regexp_instr()
1124 : * Return the match's position within the string
1125 : */
1126 : Datum
614 tgl 1127 CBC 78 : regexp_instr(PG_FUNCTION_ARGS)
1128 : {
614 tgl 1129 GIC 78 : text *str = PG_GETARG_TEXT_PP(0);
1130 78 : text *pattern = PG_GETARG_TEXT_PP(1);
1131 78 : int start = 1;
614 tgl 1132 CBC 78 : int n = 1;
614 tgl 1133 GIC 78 : int endoption = 0;
614 tgl 1134 CBC 78 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(5);
614 tgl 1135 GIC 78 : int subexpr = 0;
1136 : int pos;
1137 : pg_re_flags re_flags;
1138 : regexp_matches_ctx *matchctx;
1139 :
1140 : /* Collect optional parameters */
1141 78 : if (PG_NARGS() > 2)
614 tgl 1142 ECB : {
614 tgl 1143 GIC 69 : start = PG_GETARG_INT32(2);
614 tgl 1144 CBC 69 : if (start <= 0)
1145 3 : ereport(ERROR,
614 tgl 1146 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1147 : errmsg("invalid value for parameter \"%s\": %d",
1148 : "start", start)));
1149 : }
614 tgl 1150 CBC 75 : if (PG_NARGS() > 3)
1151 : {
614 tgl 1152 GIC 63 : n = PG_GETARG_INT32(3);
1153 63 : if (n <= 0)
1154 3 : ereport(ERROR,
1155 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 1156 ECB : errmsg("invalid value for parameter \"%s\": %d",
1157 : "n", n)));
1158 : }
614 tgl 1159 CBC 72 : if (PG_NARGS() > 4)
614 tgl 1160 ECB : {
614 tgl 1161 GIC 54 : endoption = PG_GETARG_INT32(4);
1162 54 : if (endoption != 0 && endoption != 1)
1163 6 : ereport(ERROR,
1164 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 1165 ECB : errmsg("invalid value for parameter \"%s\": %d",
1166 : "endoption", endoption)));
1167 : }
614 tgl 1168 CBC 66 : if (PG_NARGS() > 6)
614 tgl 1169 ECB : {
614 tgl 1170 GIC 42 : subexpr = PG_GETARG_INT32(6);
1171 42 : if (subexpr < 0)
1172 3 : ereport(ERROR,
1173 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 1174 ECB : errmsg("invalid value for parameter \"%s\": %d",
1175 : "subexpr", subexpr)));
1176 : }
1177 :
1178 : /* Determine options */
614 tgl 1179 GIC 63 : parse_re_flags(&re_flags, flags);
1180 : /* User mustn't specify 'g' */
1181 63 : if (re_flags.glob)
1182 3 : ereport(ERROR,
614 tgl 1183 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1184 : /* translator: %s is a SQL function name */
1185 : errmsg("%s does not support the \"global\" option",
1186 : "regexp_instr()")));
1187 : /* But we find all the matches anyway */
614 tgl 1188 GIC 60 : re_flags.glob = true;
1189 :
1190 : /* Do the matching */
1191 60 : matchctx = setup_regexp_matches(str, pattern, &re_flags, start - 1,
1192 : PG_GET_COLLATION(),
1193 : (subexpr > 0), /* need submatches? */
614 tgl 1194 ECB : false, false);
1195 :
1196 : /* When n exceeds matches return 0 (includes case of no matches) */
614 tgl 1197 CBC 60 : if (n > matchctx->nmatches)
614 tgl 1198 GIC 6 : PG_RETURN_INT32(0);
1199 :
1200 : /* When subexpr exceeds number of subexpressions return 0 */
1201 54 : if (subexpr > matchctx->npatterns)
1202 6 : PG_RETURN_INT32(0);
614 tgl 1203 ECB :
1204 : /* Select the appropriate match position to return */
614 tgl 1205 GIC 48 : pos = (n - 1) * matchctx->npatterns;
614 tgl 1206 CBC 48 : if (subexpr > 0)
614 tgl 1207 GIC 27 : pos += subexpr - 1;
1208 48 : pos *= 2;
1209 48 : if (endoption == 1)
1210 15 : pos += 1;
1211 :
614 tgl 1212 CBC 48 : if (matchctx->match_locs[pos] >= 0)
1213 45 : PG_RETURN_INT32(matchctx->match_locs[pos] + 1);
1214 : else
614 tgl 1215 GIC 3 : PG_RETURN_INT32(0); /* position not identifiable */
614 tgl 1216 ECB : }
1217 :
1218 : /* This is separate to keep the opr_sanity regression test from complaining */
1219 : Datum
614 tgl 1220 CBC 9 : regexp_instr_no_start(PG_FUNCTION_ARGS)
614 tgl 1221 ECB : {
614 tgl 1222 CBC 9 : return regexp_instr(fcinfo);
614 tgl 1223 ECB : }
1224 :
1225 : /* This is separate to keep the opr_sanity regression test from complaining */
1226 : Datum
614 tgl 1227 CBC 3 : regexp_instr_no_n(PG_FUNCTION_ARGS)
614 tgl 1228 ECB : {
614 tgl 1229 GIC 3 : return regexp_instr(fcinfo);
614 tgl 1230 ECB : }
1231 :
1232 : /* This is separate to keep the opr_sanity regression test from complaining */
1233 : Datum
614 tgl 1234 GIC 12 : regexp_instr_no_endoption(PG_FUNCTION_ARGS)
614 tgl 1235 ECB : {
614 tgl 1236 GIC 12 : return regexp_instr(fcinfo);
614 tgl 1237 ECB : }
1238 :
1239 : /* This is separate to keep the opr_sanity regression test from complaining */
1240 : Datum
614 tgl 1241 GIC 6 : regexp_instr_no_flags(PG_FUNCTION_ARGS)
614 tgl 1242 ECB : {
614 tgl 1243 GIC 6 : return regexp_instr(fcinfo);
614 tgl 1244 ECB : }
1245 :
1246 : /* This is separate to keep the opr_sanity regression test from complaining */
1247 : Datum
614 tgl 1248 GIC 6 : regexp_instr_no_subexpr(PG_FUNCTION_ARGS)
614 tgl 1249 ECB : {
614 tgl 1250 GIC 6 : return regexp_instr(fcinfo);
614 tgl 1251 ECB : }
1252 :
1253 : /*
1254 : * regexp_like()
1255 : * Test for a pattern match within a string.
1256 : */
1257 : Datum
614 tgl 1258 CBC 15 : regexp_like(PG_FUNCTION_ARGS)
1259 : {
614 tgl 1260 GIC 15 : text *str = PG_GETARG_TEXT_PP(0);
1261 15 : text *pattern = PG_GETARG_TEXT_PP(1);
1262 15 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(2);
614 tgl 1263 ECB : pg_re_flags re_flags;
1264 :
1265 : /* Determine options */
614 tgl 1266 GIC 15 : parse_re_flags(&re_flags, flags);
1267 : /* User mustn't specify 'g' */
1268 15 : if (re_flags.glob)
1269 3 : ereport(ERROR,
1270 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1271 : /* translator: %s is a SQL function name */
1272 : errmsg("%s does not support the \"global\" option",
614 tgl 1273 ECB : "regexp_like()")));
1274 :
1275 : /* Otherwise it's like textregexeq/texticregexeq */
614 tgl 1276 CBC 12 : PG_RETURN_BOOL(RE_compile_and_execute(pattern,
614 tgl 1277 ECB : VARDATA_ANY(str),
1278 : VARSIZE_ANY_EXHDR(str),
1279 : re_flags.cflags,
1280 : PG_GET_COLLATION(),
1281 : 0, NULL));
1282 : }
1283 :
1284 : /* This is separate to keep the opr_sanity regression test from complaining */
1285 : Datum
614 tgl 1286 GIC 3 : regexp_like_no_flags(PG_FUNCTION_ARGS)
1287 : {
1288 3 : return regexp_like(fcinfo);
1289 : }
1290 :
2426 tgl 1291 ECB : /*
1292 : * regexp_match()
1293 : * Return the first substring(s) matching a pattern within a string.
1294 : */
1295 : Datum
2426 tgl 1296 GIC 1278 : regexp_match(PG_FUNCTION_ARGS)
1297 : {
1298 1278 : text *orig_str = PG_GETARG_TEXT_PP(0);
1299 1278 : text *pattern = PG_GETARG_TEXT_PP(1);
1300 1278 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(2);
2426 tgl 1301 ECB : pg_re_flags re_flags;
1302 : regexp_matches_ctx *matchctx;
1303 :
1304 : /* Determine options */
2426 tgl 1305 GIC 1278 : parse_re_flags(&re_flags, flags);
1306 : /* User mustn't specify 'g' */
1307 1278 : if (re_flags.glob)
1308 4 : ereport(ERROR,
1309 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1310 : /* translator: %s is a SQL function name */
1424 alvherre 1311 ECB : errmsg("%s does not support the \"global\" option",
1312 : "regexp_match()"),
2426 tgl 1313 : errhint("Use the regexp_matches function instead.")));
1314 :
614 tgl 1315 CBC 1274 : matchctx = setup_regexp_matches(orig_str, pattern, &re_flags, 0,
1316 : PG_GET_COLLATION(), true, false, false);
1317 :
2426 tgl 1318 GIC 1274 : if (matchctx->nmatches == 0)
1319 85 : PG_RETURN_NULL();
2426 tgl 1320 ECB :
2426 tgl 1321 GIC 1189 : Assert(matchctx->nmatches == 1);
2426 tgl 1322 ECB :
1323 : /* Create workspace that build_regexp_match_result needs */
2426 tgl 1324 GIC 1189 : matchctx->elems = (Datum *) palloc(sizeof(Datum) * matchctx->npatterns);
1325 1189 : matchctx->nulls = (bool *) palloc(sizeof(bool) * matchctx->npatterns);
1326 :
1327 1189 : PG_RETURN_DATUM(PointerGetDatum(build_regexp_match_result(matchctx)));
1328 : }
1329 :
2426 tgl 1330 ECB : /* This is separate to keep the opr_sanity regression test from complaining */
1331 : Datum
2426 tgl 1332 GIC 1263 : regexp_match_no_flags(PG_FUNCTION_ARGS)
2426 tgl 1333 ECB : {
2426 tgl 1334 CBC 1263 : return regexp_match(fcinfo);
1335 : }
2426 tgl 1336 ECB :
1337 : /*
1338 : * regexp_matches()
1339 : * Return a table of all matches of a pattern within a string.
5720 1340 : */
1341 : Datum
5864 neilc 1342 CBC 339 : regexp_matches(PG_FUNCTION_ARGS)
1343 : {
1344 : FuncCallContext *funcctx;
1345 : regexp_matches_ctx *matchctx;
1346 :
1347 339 : if (SRF_IS_FIRSTCALL())
1348 : {
5624 bruce 1349 144 : text *pattern = PG_GETARG_TEXT_PP(1);
5624 bruce 1350 GIC 144 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(2);
1351 : pg_re_flags re_flags;
1352 : MemoryContext oldcontext;
1353 :
5864 neilc 1354 144 : funcctx = SRF_FIRSTCALL_INIT();
1355 144 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
1356 :
2426 tgl 1357 ECB : /* Determine options */
2426 tgl 1358 GIC 144 : parse_re_flags(&re_flags, flags);
1359 :
1360 : /* be sure to copy the input string into the multi-call ctx */
5856 neilc 1361 141 : matchctx = setup_regexp_matches(PG_GETARG_TEXT_P_COPY(0), pattern,
614 tgl 1362 ECB : &re_flags, 0,
1363 : PG_GET_COLLATION(),
1685 rhodiumtoad 1364 : true, false, false);
5720 tgl 1365 :
1366 : /* Pre-create workspace that build_regexp_match_result needs */
5720 tgl 1367 GIC 135 : matchctx->elems = (Datum *) palloc(sizeof(Datum) * matchctx->npatterns);
1368 135 : matchctx->nulls = (bool *) palloc(sizeof(bool) * matchctx->npatterns);
5864 neilc 1369 ECB :
5864 neilc 1370 CBC 135 : MemoryContextSwitchTo(oldcontext);
5864 neilc 1371 GIC 135 : funcctx->user_fctx = (void *) matchctx;
1372 : }
5864 neilc 1373 ECB :
5864 neilc 1374 GIC 330 : funcctx = SRF_PERCALL_SETUP();
1375 330 : matchctx = (regexp_matches_ctx *) funcctx->user_fctx;
5864 neilc 1376 ECB :
5720 tgl 1377 GIC 330 : if (matchctx->next_match < matchctx->nmatches)
1378 : {
1379 : ArrayType *result_ary;
1380 :
2426 1381 195 : result_ary = build_regexp_match_result(matchctx);
5720 tgl 1382 CBC 195 : matchctx->next_match++;
1383 195 : SRF_RETURN_NEXT(funcctx, PointerGetDatum(result_ary));
1384 : }
5864 neilc 1385 ECB :
5864 neilc 1386 CBC 135 : SRF_RETURN_DONE(funcctx);
1387 : }
1388 :
5720 tgl 1389 ECB : /* This is separate to keep the opr_sanity regression test from complaining */
5864 neilc 1390 : Datum
5864 neilc 1391 GIC 177 : regexp_matches_no_flags(PG_FUNCTION_ARGS)
5864 neilc 1392 ECB : {
5864 neilc 1393 GIC 177 : return regexp_matches(fcinfo);
1394 : }
1395 :
5720 tgl 1396 ECB : /*
614 1397 : * setup_regexp_matches --- do the initial matching for regexp_match,
1398 : * regexp_split, and related functions
1399 : *
1400 : * To avoid having to re-find the compiled pattern on each call, we do
5720 1401 : * all the matching in one swoop. The returned regexp_matches_ctx contains
1402 : * the locations of all the substrings matching the pattern.
1403 : *
1404 : * start_search: the character (not byte) offset in orig_str at which to
1405 : * begin the search. Returned positions are relative to orig_str anyway.
614 1406 : * use_subpatterns: collect data about matches to parenthesized subexpressions.
1407 : * ignore_degenerate: ignore zero-length matches.
1408 : * fetching_unmatched: caller wants to fetch unmatched substrings.
1409 : *
1410 : * We don't currently assume that fetching_unmatched is exclusive of fetching
1411 : * the matched text too; if it's set, the conversion buffer is large enough to
1412 : * fetch any single matched or unmatched string, but not any larger
1413 : * substring. (In practice, when splitting the matches are usually small
1414 : * anyway, and it didn't seem worth complicating the code further.)
1415 : */
1416 : static regexp_matches_ctx *
2426 tgl 1417 GIC 101727 : setup_regexp_matches(text *orig_str, text *pattern, pg_re_flags *re_flags,
1418 : int start_search,
1419 : Oid collation,
1420 : bool use_subpatterns,
1421 : bool ignore_degenerate,
1422 : bool fetching_unmatched)
1423 : {
5720 1424 101727 : regexp_matches_ctx *matchctx = palloc0(sizeof(regexp_matches_ctx));
1685 rhodiumtoad 1425 101727 : int eml = pg_database_encoding_max_length();
1426 : int orig_len;
1427 : pg_wchar *wide_str;
1428 : int wide_len;
1429 : int cflags;
1430 : regex_t *cpattern;
1431 : regmatch_t *pmatch;
5720 tgl 1432 ECB : int pmatch_len;
1433 : int array_len;
1434 : int array_idx;
1435 : int prev_match_end;
1436 : int prev_valid_match_end;
1685 rhodiumtoad 1437 GIC 101727 : int maxlen = 0; /* largest fetch length in characters */
1438 :
5720 tgl 1439 ECB : /* save original string --- we'll extract result substrings from it */
5864 neilc 1440 CBC 101727 : matchctx->orig_str = orig_str;
1441 :
1442 : /* convert string to pg_wchar form for matching */
5679 tgl 1443 GIC 101727 : orig_len = VARSIZE_ANY_EXHDR(orig_str);
5720 1444 101727 : wide_str = (pg_wchar *) palloc(sizeof(pg_wchar) * (orig_len + 1));
5679 1445 101727 : wide_len = pg_mb2wchar_with_len(VARDATA_ANY(orig_str), wide_str, orig_len);
1446 :
1447 : /* set up the compiled pattern */
608 1448 101727 : cflags = re_flags->cflags;
1449 101727 : if (!use_subpatterns)
1450 100261 : cflags |= REG_NOSUB;
1451 101727 : cpattern = RE_compile_and_cache(pattern, cflags, collation);
5864 neilc 1452 ECB :
1453 : /* do we want to remember subpatterns? */
5720 tgl 1454 GIC 101721 : if (use_subpatterns && cpattern->re_nsub > 0)
5864 neilc 1455 ECB : {
5720 tgl 1456 GIC 1355 : matchctx->npatterns = cpattern->re_nsub;
1457 1355 : pmatch_len = cpattern->re_nsub + 1;
5720 tgl 1458 ECB : }
1459 : else
1460 : {
5720 tgl 1461 GIC 100366 : use_subpatterns = false;
1462 100366 : matchctx->npatterns = 1;
5720 tgl 1463 CBC 100366 : pmatch_len = 1;
5720 tgl 1464 ECB : }
5864 neilc 1465 :
5720 tgl 1466 : /* temporary output space for RE package */
5720 tgl 1467 GIC 101721 : pmatch = palloc(sizeof(regmatch_t) * pmatch_len);
1468 :
1685 rhodiumtoad 1469 ECB : /*
1470 : * the real output space (grown dynamically if needed)
1471 : *
1472 : * use values 2^n-1, not 2^n, so that we hit the limit at 2^28-1 rather
1473 : * than at 2^27
1474 : */
1685 rhodiumtoad 1475 GIC 101721 : array_len = re_flags->glob ? 255 : 31;
5720 tgl 1476 CBC 101721 : matchctx->match_locs = (int *) palloc(sizeof(int) * array_len);
1477 101721 : array_idx = 0;
5720 tgl 1478 ECB :
1479 : /* search for the pattern, perhaps repeatedly */
5720 tgl 1480 GIC 101721 : prev_match_end = 0;
1670 rhodiumtoad 1481 101721 : prev_valid_match_end = 0;
5720 tgl 1482 CBC 547587 : while (RE_wchar_execute(cpattern, wide_str, wide_len, start_search,
1483 : pmatch_len, pmatch))
1484 : {
1485 : /*
1486 : * If requested, ignore degenerate matches, which are zero-length
1487 : * matches occurring at the start or end of a string or just after a
1488 : * previous match.
1489 : */
1490 447173 : if (!ignore_degenerate ||
1491 445591 : (pmatch[0].rm_so < wide_len &&
1492 445570 : pmatch[0].rm_eo > prev_match_end))
1493 : {
1494 : /* enlarge output space if needed */
1685 rhodiumtoad 1495 447263 : while (array_idx + matchctx->npatterns * 2 + 1 > array_len)
5720 tgl 1496 ECB : {
1418 tgl 1497 CBC 180 : array_len += array_len + 1; /* 2^n-1 => 2^(n+1)-1 */
1418 tgl 1498 GIC 180 : if (array_len > MaxAllocSize / sizeof(int))
1685 rhodiumtoad 1499 UIC 0 : ereport(ERROR,
1500 : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
1501 : errmsg("too many regular expression matches")));
5720 tgl 1502 GIC 180 : matchctx->match_locs = (int *) repalloc(matchctx->match_locs,
1503 : sizeof(int) * array_len);
1504 : }
5864 neilc 1505 ECB :
5720 tgl 1506 : /* save this match's locations */
5720 tgl 1507 CBC 447083 : if (use_subpatterns)
1508 : {
1509 : int i;
5720 tgl 1510 ECB :
5720 tgl 1511 GIC 3936 : for (i = 1; i <= matchctx->npatterns; i++)
5720 tgl 1512 ECB : {
1418 tgl 1513 CBC 2657 : int so = pmatch[i].rm_so;
1418 tgl 1514 GBC 2657 : int eo = pmatch[i].rm_eo;
1515 :
1685 rhodiumtoad 1516 GIC 2657 : matchctx->match_locs[array_idx++] = so;
1685 rhodiumtoad 1517 CBC 2657 : matchctx->match_locs[array_idx++] = eo;
1685 rhodiumtoad 1518 GIC 2657 : if (so >= 0 && eo >= 0 && (eo - so) > maxlen)
1519 1710 : maxlen = (eo - so);
1520 : }
1521 : }
5864 neilc 1522 ECB : else
1523 : {
1418 tgl 1524 GIC 445804 : int so = pmatch[0].rm_so;
1525 445804 : int eo = pmatch[0].rm_eo;
1418 tgl 1526 ECB :
1685 rhodiumtoad 1527 GIC 445804 : matchctx->match_locs[array_idx++] = so;
1685 rhodiumtoad 1528 CBC 445804 : matchctx->match_locs[array_idx++] = eo;
1529 445804 : if (so >= 0 && eo >= 0 && (eo - so) > maxlen)
1685 rhodiumtoad 1530 GIC 100283 : maxlen = (eo - so);
5864 neilc 1531 ECB : }
5720 tgl 1532 CBC 447083 : matchctx->nmatches++;
1685 rhodiumtoad 1533 ECB :
1534 : /*
1535 : * check length of unmatched portion between end of previous valid
1536 : * (nondegenerate, or degenerate but not ignored) match and start
1537 : * of current one
1538 : */
1685 rhodiumtoad 1539 CBC 447083 : if (fetching_unmatched &&
1540 445501 : pmatch[0].rm_so >= 0 &&
1670 rhodiumtoad 1541 GIC 445501 : (pmatch[0].rm_so - prev_valid_match_end) > maxlen)
1670 rhodiumtoad 1542 CBC 190379 : maxlen = (pmatch[0].rm_so - prev_valid_match_end);
1543 447083 : prev_valid_match_end = pmatch[0].rm_eo;
5864 neilc 1544 ECB : }
5720 tgl 1545 CBC 447173 : prev_match_end = pmatch[0].rm_eo;
1546 :
5720 tgl 1547 ECB : /* if not glob, stop after one match */
2426 tgl 1548 GIC 447173 : if (!re_flags->glob)
5720 1549 1274 : break;
1550 :
1551 : /*
1552 : * Advance search position. Normally we start the next search at the
1553 : * end of the previous match; but if the match was of zero length, we
3539 tgl 1554 ECB : * have to advance by one character, or we'd just find the same match
1555 : * again.
5720 1556 : */
3539 tgl 1557 CBC 445899 : start_search = prev_match_end;
1558 445899 : if (pmatch[0].rm_so == pmatch[0].rm_eo)
5720 tgl 1559 GIC 588 : start_search++;
5720 tgl 1560 CBC 445899 : if (start_search > wide_len)
5720 tgl 1561 GIC 33 : break;
1562 : }
5864 neilc 1563 ECB :
1685 rhodiumtoad 1564 : /*
1565 : * check length of unmatched portion between end of last match and end of
1566 : * input string
1567 : */
1685 rhodiumtoad 1568 GIC 101721 : if (fetching_unmatched &&
1670 1569 100192 : (wide_len - prev_valid_match_end) > maxlen)
1570 14 : maxlen = (wide_len - prev_valid_match_end);
1571 :
1685 rhodiumtoad 1572 ECB : /*
1573 : * Keep a note of the end position of the string for the benefit of
1574 : * splitting code.
1575 : */
1685 rhodiumtoad 1576 CBC 101721 : matchctx->match_locs[array_idx] = wide_len;
1577 :
1685 rhodiumtoad 1578 GIC 101721 : if (eml > 1)
1579 : {
1580 101721 : int64 maxsiz = eml * (int64) maxlen;
1581 : int conv_bufsiz;
1582 :
1685 rhodiumtoad 1583 ECB : /*
1584 : * Make the conversion buffer large enough for any substring of
1585 : * interest.
1586 : *
1587 : * Worst case: assume we need the maximum size (maxlen*eml), but take
1588 : * advantage of the fact that the original string length in bytes is
1589 : * an upper bound on the byte length of any fetched substring (and we
1590 : * know that len+1 is safe to allocate because the varlena header is
1418 tgl 1591 : * longer than 1 byte).
1592 : */
1685 rhodiumtoad 1593 CBC 101721 : if (maxsiz > orig_len)
1685 rhodiumtoad 1594 GIC 100240 : conv_bufsiz = orig_len + 1;
1685 rhodiumtoad 1595 ECB : else
1685 rhodiumtoad 1596 GIC 1481 : conv_bufsiz = maxsiz + 1; /* safe since maxsiz < 2^30 */
1597 :
1598 101721 : matchctx->conv_buf = palloc(conv_bufsiz);
1599 101721 : matchctx->conv_bufsiz = conv_bufsiz;
1600 101721 : matchctx->wide_str = wide_str;
1601 : }
1602 : else
1603 : {
1604 : /* No need to keep the wide string if we're in a single-byte charset. */
1685 rhodiumtoad 1605 UIC 0 : pfree(wide_str);
1606 0 : matchctx->wide_str = NULL;
1607 0 : matchctx->conv_buf = NULL;
1685 rhodiumtoad 1608 LBC 0 : matchctx->conv_bufsiz = 0;
1685 rhodiumtoad 1609 ECB : }
1610 :
5720 tgl 1611 : /* Clean up temp storage */
5720 tgl 1612 GIC 101721 : pfree(pmatch);
5864 neilc 1613 ECB :
5720 tgl 1614 CBC 101721 : return matchctx;
5720 tgl 1615 ECB : }
1616 :
1617 : /*
1618 : * build_regexp_match_result - build output array for current match
1619 : */
5720 tgl 1620 EUB : static ArrayType *
2426 tgl 1621 GBC 1384 : build_regexp_match_result(regexp_matches_ctx *matchctx)
5720 tgl 1622 EUB : {
1685 rhodiumtoad 1623 GBC 1384 : char *buf = matchctx->conv_buf;
5720 tgl 1624 GIC 1384 : Datum *elems = matchctx->elems;
1625 1384 : bool *nulls = matchctx->nulls;
1626 : int dims[1];
5624 bruce 1627 ECB : int lbs[1];
1628 : int loc;
5720 tgl 1629 : int i;
1630 :
1631 : /* Extract matching substrings from the original string */
5720 tgl 1632 GIC 1384 : loc = matchctx->next_match * matchctx->npatterns * 2;
1633 4011 : for (i = 0; i < matchctx->npatterns; i++)
1634 : {
5624 bruce 1635 2627 : int so = matchctx->match_locs[loc++];
5624 bruce 1636 CBC 2627 : int eo = matchctx->match_locs[loc++];
1637 :
5720 tgl 1638 2627 : if (so < 0 || eo < 0)
5720 tgl 1639 ECB : {
5720 tgl 1640 CBC 3 : elems[i] = (Datum) 0;
5720 tgl 1641 GIC 3 : nulls[i] = true;
1642 : }
1685 rhodiumtoad 1643 2624 : else if (buf)
1644 : {
1418 tgl 1645 2624 : int len = pg_wchar2mb_with_len(matchctx->wide_str + so,
1646 : buf,
1418 tgl 1647 ECB : eo - so);
1648 :
1179 tgl 1649 GIC 2624 : Assert(len < matchctx->conv_bufsiz);
1685 rhodiumtoad 1650 CBC 2624 : elems[i] = PointerGetDatum(cstring_to_text_with_len(buf, len));
1651 2624 : nulls[i] = false;
1652 : }
5720 tgl 1653 ECB : else
1654 : {
5720 tgl 1655 LBC 0 : elems[i] = DirectFunctionCall3(text_substr,
2118 tgl 1656 ECB : PointerGetDatum(matchctx->orig_str),
1657 : Int32GetDatum(so + 1),
5720 1658 : Int32GetDatum(eo - so));
5720 tgl 1659 UIC 0 : nulls[i] = false;
5720 tgl 1660 ECB : }
1661 : }
1662 :
1663 : /* And form an array */
5720 tgl 1664 CBC 1384 : dims[0] = matchctx->npatterns;
1665 1384 : lbs[0] = 1;
5856 neilc 1666 ECB : /* XXX: this hardcodes assumptions about the text type */
5720 tgl 1667 GIC 1384 : return construct_md_array(elems, nulls, 1, dims, lbs,
1668 : TEXTOID, -1, false, TYPALIGN_INT);
1669 : }
5864 neilc 1670 EUB :
1671 : /*
1672 : * regexp_split_to_table()
1673 : * Split the string at matches of the pattern, returning the
5720 tgl 1674 : * split-out substrings as a table.
1675 : */
1676 : Datum
5864 neilc 1677 GIC 311 : regexp_split_to_table(PG_FUNCTION_ARGS)
1678 : {
5624 bruce 1679 ECB : FuncCallContext *funcctx;
5720 tgl 1680 : regexp_matches_ctx *splitctx;
1681 :
5864 neilc 1682 CBC 311 : if (SRF_IS_FIRSTCALL())
1683 : {
5624 bruce 1684 GIC 26 : text *pattern = PG_GETARG_TEXT_PP(1);
1685 26 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(2);
1686 : pg_re_flags re_flags;
1687 : MemoryContext oldcontext;
1688 :
5864 neilc 1689 26 : funcctx = SRF_FIRSTCALL_INIT();
1690 26 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
1691 :
2426 tgl 1692 ECB : /* Determine options */
2426 tgl 1693 GIC 26 : parse_re_flags(&re_flags, flags);
1694 : /* User mustn't specify 'g' */
1695 23 : if (re_flags.glob)
1696 3 : ereport(ERROR,
2426 tgl 1697 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1698 : /* translator: %s is a SQL function name */
1424 alvherre 1699 : errmsg("%s does not support the \"global\" option",
1700 : "regexp_split_to_table()")));
1701 : /* But we find all the matches anyway */
2426 tgl 1702 GIC 20 : re_flags.glob = true;
1703 :
5720 tgl 1704 ECB : /* be sure to copy the input string into the multi-call ctx */
5720 tgl 1705 CBC 20 : splitctx = setup_regexp_matches(PG_GETARG_TEXT_P_COPY(0), pattern,
1706 : &re_flags, 0,
1707 : PG_GET_COLLATION(),
1685 rhodiumtoad 1708 ECB : false, true, true);
1709 :
5864 neilc 1710 CBC 20 : MemoryContextSwitchTo(oldcontext);
1711 20 : funcctx->user_fctx = (void *) splitctx;
1712 : }
1713 :
5864 neilc 1714 GIC 305 : funcctx = SRF_PERCALL_SETUP();
5720 tgl 1715 305 : splitctx = (regexp_matches_ctx *) funcctx->user_fctx;
1716 :
5720 tgl 1717 CBC 305 : if (splitctx->next_match <= splitctx->nmatches)
1718 : {
5624 bruce 1719 GIC 285 : Datum result = build_regexp_split_result(splitctx);
5720 tgl 1720 ECB :
5720 tgl 1721 GIC 285 : splitctx->next_match++;
1722 285 : SRF_RETURN_NEXT(funcctx, result);
1723 : }
1724 :
5720 tgl 1725 CBC 20 : SRF_RETURN_DONE(funcctx);
5864 neilc 1726 ECB : }
1727 :
1728 : /* This is separate to keep the opr_sanity regression test from complaining */
5624 bruce 1729 : Datum
5624 bruce 1730 CBC 276 : regexp_split_to_table_no_flags(PG_FUNCTION_ARGS)
1731 : {
5864 neilc 1732 276 : return regexp_split_to_table(fcinfo);
1733 : }
5864 neilc 1734 ECB :
1735 : /*
5720 tgl 1736 : * regexp_split_to_array()
1737 : * Split the string at matches of the pattern, returning the
1738 : * split-out substrings as an array.
1739 : */
5624 bruce 1740 : Datum
5624 bruce 1741 GIC 100178 : regexp_split_to_array(PG_FUNCTION_ARGS)
1742 : {
1743 100178 : ArrayBuildState *astate = NULL;
1744 : pg_re_flags re_flags;
5624 bruce 1745 ECB : regexp_matches_ctx *splitctx;
1746 :
2426 tgl 1747 : /* Determine options */
2426 tgl 1748 GIC 100178 : parse_re_flags(&re_flags, PG_GETARG_TEXT_PP_IF_EXISTS(2));
1749 : /* User mustn't specify 'g' */
1750 100175 : if (re_flags.glob)
1751 3 : ereport(ERROR,
1752 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1753 : /* translator: %s is a SQL function name */
1754 : errmsg("%s does not support the \"global\" option",
1755 : "regexp_split_to_array()")));
2426 tgl 1756 ECB : /* But we find all the matches anyway */
2426 tgl 1757 GIC 100172 : re_flags.glob = true;
2426 tgl 1758 ECB :
5679 tgl 1759 GIC 100172 : splitctx = setup_regexp_matches(PG_GETARG_TEXT_PP(0),
1760 100172 : PG_GETARG_TEXT_PP(1),
1761 : &re_flags, 0,
1762 : PG_GET_COLLATION(),
1685 rhodiumtoad 1763 ECB : false, true, true);
1764 :
5720 tgl 1765 CBC 645580 : while (splitctx->next_match <= splitctx->nmatches)
5864 neilc 1766 ECB : {
5864 neilc 1767 GIC 545408 : astate = accumArrayResult(astate,
1768 : build_regexp_split_result(splitctx),
1769 : false,
1770 : TEXTOID,
1771 : CurrentMemoryContext);
5720 tgl 1772 CBC 545408 : splitctx->next_match++;
1773 : }
5864 neilc 1774 ECB :
224 peter 1775 GNC 100172 : PG_RETURN_DATUM(makeArrayResult(astate, CurrentMemoryContext));
1776 : }
1777 :
1778 : /* This is separate to keep the opr_sanity regression test from complaining */
1779 : Datum
5624 bruce 1780 CBC 100157 : regexp_split_to_array_no_flags(PG_FUNCTION_ARGS)
1781 : {
5864 neilc 1782 100157 : return regexp_split_to_array(fcinfo);
1783 : }
1784 :
1785 : /*
1786 : * build_regexp_split_result - build output string for current match
5720 tgl 1787 ECB : *
1788 : * We return the string between the current match and the previous one,
1789 : * or the string after the last match when next_match == nmatches.
1790 : */
1791 : static Datum
5624 bruce 1792 GIC 545693 : build_regexp_split_result(regexp_matches_ctx *splitctx)
1793 : {
1685 rhodiumtoad 1794 545693 : char *buf = splitctx->conv_buf;
5624 bruce 1795 ECB : int startpos;
1796 : int endpos;
5864 neilc 1797 :
5720 tgl 1798 GIC 545693 : if (splitctx->next_match > 0)
1799 445501 : startpos = splitctx->match_locs[splitctx->next_match * 2 - 1];
1800 : else
1801 100192 : startpos = 0;
1802 545693 : if (startpos < 0)
5720 tgl 1803 UIC 0 : elog(ERROR, "invalid match ending position");
1804 :
1179 tgl 1805 GIC 545693 : endpos = splitctx->match_locs[splitctx->next_match * 2];
1806 545693 : if (endpos < startpos)
1179 tgl 1807 LBC 0 : elog(ERROR, "invalid match starting position");
1808 :
1685 rhodiumtoad 1809 CBC 545693 : if (buf)
1810 : {
1811 : int len;
1812 :
1813 545693 : len = pg_wchar2mb_with_len(splitctx->wide_str + startpos,
1685 rhodiumtoad 1814 ECB : buf,
1815 : endpos - startpos);
1179 tgl 1816 CBC 545693 : Assert(len < splitctx->conv_bufsiz);
1685 rhodiumtoad 1817 545693 : return PointerGetDatum(cstring_to_text_with_len(buf, len));
5720 tgl 1818 EUB : }
1819 : else
5720 tgl 1820 ECB : {
1685 rhodiumtoad 1821 LBC 0 : return DirectFunctionCall3(text_substr,
5720 tgl 1822 EUB : PointerGetDatum(splitctx->orig_str),
1823 : Int32GetDatum(startpos + 1),
1685 rhodiumtoad 1824 ECB : Int32GetDatum(endpos - startpos));
1825 : }
1826 : }
1827 :
614 tgl 1828 : /*
1829 : * regexp_substr()
1830 : * Return the substring that matches a regular expression pattern
1831 : */
1832 : Datum
614 tgl 1833 GIC 54 : regexp_substr(PG_FUNCTION_ARGS)
1834 : {
1835 54 : text *str = PG_GETARG_TEXT_PP(0);
614 tgl 1836 GBC 54 : text *pattern = PG_GETARG_TEXT_PP(1);
614 tgl 1837 GIC 54 : int start = 1;
1838 54 : int n = 1;
1839 54 : text *flags = PG_GETARG_TEXT_PP_IF_EXISTS(4);
1840 54 : int subexpr = 0;
1841 : int so,
1842 : eo,
1843 : pos;
1844 : pg_re_flags re_flags;
1845 : regexp_matches_ctx *matchctx;
1846 :
1847 : /* Collect optional parameters */
614 tgl 1848 CBC 54 : if (PG_NARGS() > 2)
1849 : {
1850 45 : start = PG_GETARG_INT32(2);
1851 45 : if (start <= 0)
1852 3 : ereport(ERROR,
614 tgl 1853 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1854 : errmsg("invalid value for parameter \"%s\": %d",
1855 : "start", start)));
1856 : }
614 tgl 1857 GIC 51 : if (PG_NARGS() > 3)
1858 : {
1859 39 : n = PG_GETARG_INT32(3);
1860 39 : if (n <= 0)
1861 3 : ereport(ERROR,
1862 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 1863 ECB : errmsg("invalid value for parameter \"%s\": %d",
1864 : "n", n)));
1865 : }
614 tgl 1866 CBC 48 : if (PG_NARGS() > 5)
614 tgl 1867 ECB : {
614 tgl 1868 GIC 24 : subexpr = PG_GETARG_INT32(5);
1869 24 : if (subexpr < 0)
1870 3 : ereport(ERROR,
1871 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
614 tgl 1872 ECB : errmsg("invalid value for parameter \"%s\": %d",
1873 : "subexpr", subexpr)));
1874 : }
1875 :
1876 : /* Determine options */
614 tgl 1877 GIC 45 : parse_re_flags(&re_flags, flags);
1878 : /* User mustn't specify 'g' */
1879 45 : if (re_flags.glob)
1880 3 : ereport(ERROR,
614 tgl 1881 ECB : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1882 : /* translator: %s is a SQL function name */
1883 : errmsg("%s does not support the \"global\" option",
1884 : "regexp_substr()")));
1885 : /* But we find all the matches anyway */
614 tgl 1886 GIC 42 : re_flags.glob = true;
1887 :
1888 : /* Do the matching */
1889 42 : matchctx = setup_regexp_matches(str, pattern, &re_flags, start - 1,
1890 : PG_GET_COLLATION(),
1891 : (subexpr > 0), /* need submatches? */
614 tgl 1892 ECB : false, false);
1893 :
1894 : /* When n exceeds matches return NULL (includes case of no matches) */
614 tgl 1895 CBC 42 : if (n > matchctx->nmatches)
614 tgl 1896 GIC 6 : PG_RETURN_NULL();
1897 :
1898 : /* When subexpr exceeds number of subexpressions return NULL */
1899 36 : if (subexpr > matchctx->npatterns)
1900 3 : PG_RETURN_NULL();
614 tgl 1901 ECB :
1902 : /* Select the appropriate match position to return */
614 tgl 1903 GIC 33 : pos = (n - 1) * matchctx->npatterns;
614 tgl 1904 CBC 33 : if (subexpr > 0)
614 tgl 1905 GIC 15 : pos += subexpr - 1;
1906 33 : pos *= 2;
1907 33 : so = matchctx->match_locs[pos];
1908 33 : eo = matchctx->match_locs[pos + 1];
1909 :
614 tgl 1910 CBC 33 : if (so < 0 || eo < 0)
1911 3 : PG_RETURN_NULL(); /* unidentifiable location */
1912 :
614 tgl 1913 GIC 30 : PG_RETURN_DATUM(DirectFunctionCall3(text_substr,
614 tgl 1914 ECB : PointerGetDatum(matchctx->orig_str),
1915 : Int32GetDatum(so + 1),
1916 : Int32GetDatum(eo - so)));
1917 : }
1918 :
1919 : /* This is separate to keep the opr_sanity regression test from complaining */
1920 : Datum
614 tgl 1921 CBC 9 : regexp_substr_no_start(PG_FUNCTION_ARGS)
614 tgl 1922 ECB : {
614 tgl 1923 CBC 9 : return regexp_substr(fcinfo);
1924 : }
614 tgl 1925 ECB :
1926 : /* This is separate to keep the opr_sanity regression test from complaining */
1927 : Datum
614 tgl 1928 CBC 3 : regexp_substr_no_n(PG_FUNCTION_ARGS)
1929 : {
614 tgl 1930 GIC 3 : return regexp_substr(fcinfo);
1931 : }
1932 :
1933 : /* This is separate to keep the opr_sanity regression test from complaining */
1934 : Datum
1935 12 : regexp_substr_no_flags(PG_FUNCTION_ARGS)
614 tgl 1936 ECB : {
614 tgl 1937 GIC 12 : return regexp_substr(fcinfo);
614 tgl 1938 ECB : }
1939 :
1940 : /* This is separate to keep the opr_sanity regression test from complaining */
1941 : Datum
614 tgl 1942 GIC 6 : regexp_substr_no_subexpr(PG_FUNCTION_ARGS)
614 tgl 1943 ECB : {
614 tgl 1944 GIC 6 : return regexp_substr(fcinfo);
614 tgl 1945 ECB : }
1946 :
1947 : /*
1948 : * regexp_fixed_prefix - extract fixed prefix, if any, for a regexp
1949 : *
3925 1950 : * The result is NULL if there is no fixed prefix, else a palloc'd string.
1951 : * If it is an exact match, not just a prefix, *exact is returned as true.
1952 : */
1953 : char *
3925 tgl 1954 GIC 6461 : regexp_fixed_prefix(text *text_re, bool case_insensitive, Oid collation,
1955 : bool *exact)
1956 : {
3925 tgl 1957 ECB : char *result;
1958 : regex_t *re;
1959 : int cflags;
1960 : int re_result;
1961 : pg_wchar *str;
1962 : size_t slen;
1963 : size_t maxlen;
1964 : char errMsg[100];
1965 :
3925 tgl 1966 GIC 6461 : *exact = false; /* default result */
1967 :
1968 : /* Compile RE */
3925 tgl 1969 CBC 6461 : cflags = REG_ADVANCED;
3925 tgl 1970 GIC 6461 : if (case_insensitive)
1971 46 : cflags |= REG_ICASE;
1972 :
608 1973 6461 : re = RE_compile_and_cache(text_re, cflags | REG_NOSUB, collation);
1974 :
1975 : /* Examine it to see if there's a fixed prefix */
3925 1976 6449 : re_result = pg_regprefix(re, &str, &slen);
1977 :
1978 6449 : switch (re_result)
1979 : {
1980 364 : case REG_NOMATCH:
3925 tgl 1981 CBC 364 : return NULL;
1982 :
3925 tgl 1983 GIC 626 : case REG_PREFIX:
3925 tgl 1984 ECB : /* continue with wchar conversion */
3925 tgl 1985 CBC 626 : break;
3925 tgl 1986 ECB :
3925 tgl 1987 GIC 5459 : case REG_EXACT:
3925 tgl 1988 CBC 5459 : *exact = true;
1989 : /* continue with wchar conversion */
3925 tgl 1990 GIC 5459 : break;
3925 tgl 1991 ECB :
3925 tgl 1992 UIC 0 : default:
3925 tgl 1993 ECB : /* re failed??? */
3925 tgl 1994 LBC 0 : pg_regerror(re_result, re, errMsg, sizeof(errMsg));
1995 0 : ereport(ERROR,
1996 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
3925 tgl 1997 ECB : errmsg("regular expression failed: %s", errMsg)));
1998 : break;
1999 : }
2000 :
2001 : /* Convert pg_wchar result back to database encoding */
3925 tgl 2002 CBC 6085 : maxlen = pg_database_encoding_max_length() * slen + 1;
3925 tgl 2003 GIC 6085 : result = (char *) palloc(maxlen);
3925 tgl 2004 CBC 6085 : slen = pg_wchar2mb_with_len(str, result, slen);
3925 tgl 2005 GIC 6085 : Assert(slen < maxlen);
3925 tgl 2006 EUB :
1 tmunro 2007 GNC 6085 : pfree(str);
3925 tgl 2008 EUB :
3925 tgl 2009 GBC 6085 : return result;
2010 : }
|
