Line data Source code
1 : /*--------------------------------------------------------------------------
2 : *
3 : * test_regex.c
4 : * Test harness for the regular expression package.
5 : *
6 : * Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : * IDENTIFICATION
10 : * src/test/modules/test_regex/test_regex.c
11 : *
12 : * -------------------------------------------------------------------------
13 : */
14 :
15 : #include "postgres.h"
16 :
17 : #include "funcapi.h"
18 : #include "regex/regex.h"
19 : #include "utils/array.h"
20 : #include "utils/builtins.h"
21 :
22 4 : PG_MODULE_MAGIC;
23 :
24 :
25 : /* all the options of interest for regex functions */
26 : typedef struct test_re_flags
27 : {
28 : int cflags; /* compile flags for Spencer's regex code */
29 : int eflags; /* execute flags for Spencer's regex code */
30 : long info; /* expected re_info bits */
31 : bool glob; /* do it globally (for each occurrence) */
32 : bool indices; /* report indices not actual strings */
33 : bool partial; /* expect partial match */
34 : } test_re_flags;
35 :
36 : /* cross-call state for test_regex() */
37 : typedef struct test_regex_ctx
38 : {
39 : test_re_flags re_flags; /* flags */
40 : rm_detail_t details; /* "details" from execution */
41 : text *orig_str; /* data string in original TEXT form */
42 : int nmatches; /* number of places where pattern matched */
43 : int npatterns; /* number of capturing subpatterns */
44 : /* We store start char index and end+1 char index for each match */
45 : /* so the number of entries in match_locs is nmatches * npatterns * 2 */
46 : int *match_locs; /* 0-based character indexes */
47 : int next_match; /* 0-based index of next match to process */
48 : /* workspace for build_test_match_result() */
49 : Datum *elems; /* has npatterns+1 elements */
50 : bool *nulls; /* has npatterns+1 elements */
51 : pg_wchar *wide_str; /* wide-char version of original string */
52 : char *conv_buf; /* conversion buffer, if needed */
53 : int conv_bufsiz; /* size thereof */
54 : } test_regex_ctx;
55 :
56 : /* Local functions */
57 : static void test_re_compile(text *text_re, int cflags, Oid collation,
58 : regex_t *result_re);
59 : static void parse_test_flags(test_re_flags *flags, text *opts);
60 : static test_regex_ctx *setup_test_matches(text *orig_str,
61 : regex_t *cpattern,
62 : test_re_flags *re_flags,
63 : Oid collation,
64 : bool use_subpatterns);
65 : static ArrayType *build_test_info_result(regex_t *cpattern,
66 : test_re_flags *flags);
67 : static ArrayType *build_test_match_result(test_regex_ctx *matchctx);
68 :
69 :
70 : /*
71 : * test_regex(pattern text, string text, flags text) returns setof text[]
72 : *
73 : * This is largely based on regexp.c's regexp_matches, with additions
74 : * for debugging purposes.
75 : */
76 6 : PG_FUNCTION_INFO_V1(test_regex);
77 :
78 : Datum
79 3534 : test_regex(PG_FUNCTION_ARGS)
80 : {
81 : FuncCallContext *funcctx;
82 : test_regex_ctx *matchctx;
83 : ArrayType *result_ary;
84 :
85 3534 : if (SRF_IS_FIRSTCALL())
86 : {
87 1392 : text *pattern = PG_GETARG_TEXT_PP(0);
88 1392 : text *flags = PG_GETARG_TEXT_PP(2);
89 1392 : Oid collation = PG_GET_COLLATION();
90 : test_re_flags re_flags;
91 : regex_t cpattern;
92 : MemoryContext oldcontext;
93 :
94 1392 : funcctx = SRF_FIRSTCALL_INIT();
95 1392 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
96 :
97 : /* Determine options */
98 1392 : parse_test_flags(&re_flags, flags);
99 :
100 : /* set up the compiled pattern */
101 1392 : test_re_compile(pattern, re_flags.cflags, collation, &cpattern);
102 :
103 : /* be sure to copy the input string into the multi-call ctx */
104 1180 : matchctx = setup_test_matches(PG_GETARG_TEXT_P_COPY(1), &cpattern,
105 : &re_flags,
106 : collation,
107 : true);
108 :
109 : /* Pre-create workspace that build_test_match_result needs */
110 2360 : matchctx->elems = (Datum *) palloc(sizeof(Datum) *
111 1180 : (matchctx->npatterns + 1));
112 2360 : matchctx->nulls = (bool *) palloc(sizeof(bool) *
113 1180 : (matchctx->npatterns + 1));
114 :
115 1180 : MemoryContextSwitchTo(oldcontext);
116 1180 : funcctx->user_fctx = (void *) matchctx;
117 :
118 : /*
119 : * Return the first result row, which is info equivalent to Tcl's
120 : * "regexp -about" output
121 : */
122 1180 : result_ary = build_test_info_result(&cpattern, &re_flags);
123 :
124 1180 : pg_regfree(&cpattern);
125 :
126 1180 : SRF_RETURN_NEXT(funcctx, PointerGetDatum(result_ary));
127 : }
128 : else
129 : {
130 : /* Each subsequent row describes one match */
131 2142 : funcctx = SRF_PERCALL_SETUP();
132 2142 : matchctx = (test_regex_ctx *) funcctx->user_fctx;
133 :
134 2142 : if (matchctx->next_match < matchctx->nmatches)
135 : {
136 962 : result_ary = build_test_match_result(matchctx);
137 962 : matchctx->next_match++;
138 962 : SRF_RETURN_NEXT(funcctx, PointerGetDatum(result_ary));
139 : }
140 : }
141 :
142 1180 : SRF_RETURN_DONE(funcctx);
143 : }
144 :
145 :
146 : /*
147 : * test_re_compile - compile a RE
148 : *
149 : * text_re --- the pattern, expressed as a TEXT object
150 : * cflags --- compile options for the pattern
151 : * collation --- collation to use for LC_CTYPE-dependent behavior
152 : * result_re --- output, compiled RE is stored here
153 : *
154 : * Pattern is given in the database encoding. We internally convert to
155 : * an array of pg_wchar, which is what Spencer's regex package wants.
156 : *
157 : * Caller must eventually pg_regfree the resulting RE to avoid memory leaks.
158 : */
159 : static void
160 1392 : test_re_compile(text *text_re, int cflags, Oid collation,
161 : regex_t *result_re)
162 : {
163 1392 : int text_re_len = VARSIZE_ANY_EXHDR(text_re);
164 1392 : char *text_re_val = VARDATA_ANY(text_re);
165 : pg_wchar *pattern;
166 : int pattern_len;
167 : int regcomp_result;
168 : char errMsg[100];
169 :
170 : /* Convert pattern string to wide characters */
171 1392 : pattern = (pg_wchar *) palloc((text_re_len + 1) * sizeof(pg_wchar));
172 1392 : pattern_len = pg_mb2wchar_with_len(text_re_val,
173 : pattern,
174 : text_re_len);
175 :
176 1392 : regcomp_result = pg_regcomp(result_re,
177 : pattern,
178 : pattern_len,
179 : cflags,
180 : collation);
181 :
182 1392 : pfree(pattern);
183 :
184 1392 : if (regcomp_result != REG_OKAY)
185 : {
186 : /* re didn't compile (no need for pg_regfree, if so) */
187 212 : pg_regerror(regcomp_result, result_re, errMsg, sizeof(errMsg));
188 212 : ereport(ERROR,
189 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
190 : errmsg("invalid regular expression: %s", errMsg)));
191 : }
192 1180 : }
193 :
194 : /*
195 : * test_re_execute - execute a RE on pg_wchar data
196 : *
197 : * Returns true on match, false on no match
198 : * Arguments are as for pg_regexec
199 : */
200 : static bool
201 1180 : test_re_execute(regex_t *re, pg_wchar *data, int data_len,
202 : int start_search,
203 : rm_detail_t *details,
204 : int nmatch, regmatch_t *pmatch,
205 : int eflags)
206 : {
207 : int regexec_result;
208 : char errMsg[100];
209 :
210 : /* Initialize match locations in case engine doesn't */
211 1180 : details->rm_extend.rm_so = -1;
212 1180 : details->rm_extend.rm_eo = -1;
213 2932 : for (int i = 0; i < nmatch; i++)
214 : {
215 1752 : pmatch[i].rm_so = -1;
216 1752 : pmatch[i].rm_eo = -1;
217 : }
218 :
219 : /* Perform RE match and return result */
220 1180 : regexec_result = pg_regexec(re,
221 : data,
222 : data_len,
223 : start_search,
224 : details,
225 : nmatch,
226 : pmatch,
227 : eflags);
228 :
229 1180 : if (regexec_result != REG_OKAY && regexec_result != REG_NOMATCH)
230 : {
231 : /* re failed??? */
232 0 : pg_regerror(regexec_result, re, errMsg, sizeof(errMsg));
233 0 : ereport(ERROR,
234 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
235 : errmsg("regular expression failed: %s", errMsg)));
236 : }
237 :
238 1180 : return (regexec_result == REG_OKAY);
239 : }
240 :
241 :
242 : /*
243 : * parse_test_flags - parse the flags argument
244 : *
245 : * flags --- output argument, filled with desired options
246 : * opts --- TEXT object, or NULL for defaults
247 : */
248 : static void
249 1392 : parse_test_flags(test_re_flags *flags, text *opts)
250 : {
251 : /* these defaults must match Tcl's */
252 1392 : int cflags = REG_ADVANCED;
253 1392 : int eflags = 0;
254 1392 : long info = 0;
255 :
256 1392 : flags->glob = false;
257 1392 : flags->indices = false;
258 1392 : flags->partial = false;
259 :
260 1392 : if (opts)
261 : {
262 1392 : char *opt_p = VARDATA_ANY(opts);
263 1392 : int opt_len = VARSIZE_ANY_EXHDR(opts);
264 : int i;
265 :
266 3782 : for (i = 0; i < opt_len; i++)
267 : {
268 2390 : switch (opt_p[i])
269 : {
270 156 : case '-':
271 : /* allowed, no-op */
272 156 : break;
273 14 : case '!':
274 14 : flags->partial = true;
275 14 : break;
276 2 : case '*':
277 : /* test requires Unicode --- ignored here */
278 2 : break;
279 106 : case '0':
280 106 : flags->indices = true;
281 106 : break;
282 :
283 : /* These flags correspond to user-exposed RE options: */
284 0 : case 'g': /* global match */
285 0 : flags->glob = true;
286 0 : break;
287 40 : case 'i': /* case insensitive */
288 40 : cflags |= REG_ICASE;
289 40 : break;
290 70 : case 'n': /* \n affects ^ $ . [^ */
291 70 : cflags |= REG_NEWLINE;
292 70 : break;
293 4 : case 'p': /* ~Perl, \n affects . [^ */
294 4 : cflags |= REG_NLSTOP;
295 4 : cflags &= ~REG_NLANCH;
296 4 : break;
297 4 : case 'w': /* weird, \n affects ^ $ only */
298 4 : cflags &= ~REG_NLSTOP;
299 4 : cflags |= REG_NLANCH;
300 4 : break;
301 28 : case 'x': /* expanded syntax */
302 28 : cflags |= REG_EXPANDED;
303 28 : break;
304 :
305 : /* These flags correspond to Tcl's -xflags options: */
306 4 : case 'a':
307 4 : cflags |= REG_ADVF;
308 4 : break;
309 262 : case 'b':
310 262 : cflags &= ~REG_ADVANCED;
311 262 : break;
312 22 : case 'c':
313 :
314 : /*
315 : * Tcl calls this TCL_REG_CANMATCH, but it's really
316 : * REG_EXPECT. In this implementation we must also set
317 : * the partial and indices flags, so that
318 : * setup_test_matches and build_test_match_result will
319 : * emit the desired data. (They'll emit more fields than
320 : * Tcl would, but that's fine.)
321 : */
322 22 : cflags |= REG_EXPECT;
323 22 : flags->partial = true;
324 22 : flags->indices = true;
325 22 : break;
326 20 : case 'e':
327 20 : cflags &= ~REG_ADVANCED;
328 20 : cflags |= REG_EXTENDED;
329 20 : break;
330 12 : case 'q':
331 12 : cflags &= ~REG_ADVANCED;
332 12 : cflags |= REG_QUOTE;
333 12 : break;
334 4 : case 'o': /* o for opaque */
335 4 : cflags |= REG_NOSUB;
336 4 : break;
337 4 : case 's': /* s for start */
338 4 : cflags |= REG_BOSONLY;
339 4 : break;
340 12 : case '+':
341 12 : cflags |= REG_FAKE;
342 12 : break;
343 0 : case ',':
344 0 : cflags |= REG_PROGRESS;
345 0 : break;
346 0 : case '.':
347 0 : cflags |= REG_DUMP;
348 0 : break;
349 0 : case ':':
350 0 : eflags |= REG_MTRACE;
351 0 : break;
352 0 : case ';':
353 0 : eflags |= REG_FTRACE;
354 0 : break;
355 12 : case '^':
356 12 : eflags |= REG_NOTBOL;
357 12 : break;
358 8 : case '$':
359 8 : eflags |= REG_NOTEOL;
360 8 : break;
361 34 : case 't':
362 34 : cflags |= REG_EXPECT;
363 34 : break;
364 10 : case '%':
365 10 : eflags |= REG_SMALL;
366 10 : break;
367 :
368 : /* These flags define expected info bits: */
369 10 : case 'A':
370 10 : info |= REG_UBSALNUM;
371 10 : break;
372 8 : case 'B':
373 8 : info |= REG_UBRACES;
374 8 : break;
375 84 : case 'E':
376 84 : info |= REG_UBBS;
377 84 : break;
378 68 : case 'H':
379 68 : info |= REG_ULOOKAROUND;
380 68 : break;
381 22 : case 'I':
382 22 : info |= REG_UIMPOSSIBLE;
383 22 : break;
384 328 : case 'L':
385 328 : info |= REG_ULOCALE;
386 328 : break;
387 86 : case 'M':
388 86 : info |= REG_UUNPORT;
389 86 : break;
390 94 : case 'N':
391 94 : info |= REG_UEMPTYMATCH;
392 94 : break;
393 614 : case 'P':
394 614 : info |= REG_UNONPOSIX;
395 614 : break;
396 72 : case 'Q':
397 72 : info |= REG_UBOUNDS;
398 72 : break;
399 84 : case 'R':
400 84 : info |= REG_UBACKREF;
401 84 : break;
402 50 : case 'S':
403 50 : info |= REG_UUNSPEC;
404 50 : break;
405 40 : case 'T':
406 40 : info |= REG_USHORTEST;
407 40 : break;
408 2 : case 'U':
409 2 : info |= REG_UPBOTCH;
410 2 : break;
411 :
412 0 : default:
413 0 : ereport(ERROR,
414 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
415 : errmsg("invalid regular expression test option: \"%.*s\"",
416 : pg_mblen(opt_p + i), opt_p + i)));
417 : break;
418 : }
419 : }
420 : }
421 1392 : flags->cflags = cflags;
422 1392 : flags->eflags = eflags;
423 1392 : flags->info = info;
424 1392 : }
425 :
426 : /*
427 : * setup_test_matches --- do the initial matching
428 : *
429 : * To simplify memory management, we do all the matching in one swoop.
430 : * The returned test_regex_ctx contains the locations of all the substrings
431 : * matching the pattern.
432 : */
433 : static test_regex_ctx *
434 1180 : setup_test_matches(text *orig_str,
435 : regex_t *cpattern, test_re_flags *re_flags,
436 : Oid collation,
437 : bool use_subpatterns)
438 : {
439 1180 : test_regex_ctx *matchctx = palloc0(sizeof(test_regex_ctx));
440 1180 : int eml = pg_database_encoding_max_length();
441 : int orig_len;
442 : pg_wchar *wide_str;
443 : int wide_len;
444 : regmatch_t *pmatch;
445 : int pmatch_len;
446 : int array_len;
447 : int array_idx;
448 : int prev_match_end;
449 : int start_search;
450 1180 : int maxlen = 0; /* largest fetch length in characters */
451 :
452 : /* save flags */
453 1180 : matchctx->re_flags = *re_flags;
454 :
455 : /* save original string --- we'll extract result substrings from it */
456 1180 : matchctx->orig_str = orig_str;
457 :
458 : /* convert string to pg_wchar form for matching */
459 1180 : orig_len = VARSIZE_ANY_EXHDR(orig_str);
460 1180 : wide_str = (pg_wchar *) palloc(sizeof(pg_wchar) * (orig_len + 1));
461 1180 : wide_len = pg_mb2wchar_with_len(VARDATA_ANY(orig_str), wide_str, orig_len);
462 :
463 : /* do we want to remember subpatterns? */
464 1180 : if (use_subpatterns && cpattern->re_nsub > 0)
465 : {
466 254 : matchctx->npatterns = cpattern->re_nsub + 1;
467 254 : pmatch_len = cpattern->re_nsub + 1;
468 : }
469 : else
470 : {
471 926 : use_subpatterns = false;
472 926 : matchctx->npatterns = 1;
473 926 : pmatch_len = 1;
474 : }
475 :
476 : /* temporary output space for RE package */
477 1180 : pmatch = palloc(sizeof(regmatch_t) * pmatch_len);
478 :
479 : /*
480 : * the real output space (grown dynamically if needed)
481 : *
482 : * use values 2^n-1, not 2^n, so that we hit the limit at 2^28-1 rather
483 : * than at 2^27
484 : */
485 1180 : array_len = re_flags->glob ? 255 : 31;
486 1180 : matchctx->match_locs = (int *) palloc(sizeof(int) * array_len);
487 1180 : array_idx = 0;
488 :
489 : /* search for the pattern, perhaps repeatedly */
490 1180 : prev_match_end = 0;
491 1180 : start_search = 0;
492 1180 : while (test_re_execute(cpattern, wide_str, wide_len,
493 : start_search,
494 : &matchctx->details,
495 : pmatch_len, pmatch,
496 : re_flags->eflags))
497 : {
498 : /* enlarge output space if needed */
499 926 : while (array_idx + matchctx->npatterns * 2 + 1 > array_len)
500 : {
501 0 : array_len += array_len + 1; /* 2^n-1 => 2^(n+1)-1 */
502 0 : if (array_len > MaxAllocSize / sizeof(int))
503 0 : ereport(ERROR,
504 : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
505 : errmsg("too many regular expression matches")));
506 0 : matchctx->match_locs = (int *) repalloc(matchctx->match_locs,
507 : sizeof(int) * array_len);
508 : }
509 :
510 : /* save this match's locations */
511 2188 : for (int i = 0; i < matchctx->npatterns; i++)
512 : {
513 1262 : int so = pmatch[i].rm_so;
514 1262 : int eo = pmatch[i].rm_eo;
515 :
516 1262 : matchctx->match_locs[array_idx++] = so;
517 1262 : matchctx->match_locs[array_idx++] = eo;
518 1262 : if (so >= 0 && eo >= 0 && (eo - so) > maxlen)
519 876 : maxlen = (eo - so);
520 : }
521 926 : matchctx->nmatches++;
522 926 : prev_match_end = pmatch[0].rm_eo;
523 :
524 : /* if not glob, stop after one match */
525 926 : if (!re_flags->glob)
526 926 : break;
527 :
528 : /*
529 : * Advance search position. Normally we start the next search at the
530 : * end of the previous match; but if the match was of zero length, we
531 : * have to advance by one character, or we'd just find the same match
532 : * again.
533 : */
534 0 : start_search = prev_match_end;
535 0 : if (pmatch[0].rm_so == pmatch[0].rm_eo)
536 0 : start_search++;
537 0 : if (start_search > wide_len)
538 0 : break;
539 : }
540 :
541 : /*
542 : * If we had no match, but "partial" and "indices" are set, emit the
543 : * details.
544 : */
545 1180 : if (matchctx->nmatches == 0 && re_flags->partial && re_flags->indices)
546 : {
547 : /* enlarge output space if needed */
548 36 : while (array_idx + matchctx->npatterns * 2 + 1 > array_len)
549 : {
550 0 : array_len += array_len + 1; /* 2^n-1 => 2^(n+1)-1 */
551 0 : if (array_len > MaxAllocSize / sizeof(int))
552 0 : ereport(ERROR,
553 : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
554 : errmsg("too many regular expression matches")));
555 0 : matchctx->match_locs = (int *) repalloc(matchctx->match_locs,
556 : sizeof(int) * array_len);
557 : }
558 :
559 36 : matchctx->match_locs[array_idx++] = matchctx->details.rm_extend.rm_so;
560 36 : matchctx->match_locs[array_idx++] = matchctx->details.rm_extend.rm_eo;
561 : /* we don't have pmatch data, so emit -1 */
562 40 : for (int i = 1; i < matchctx->npatterns; i++)
563 : {
564 4 : matchctx->match_locs[array_idx++] = -1;
565 4 : matchctx->match_locs[array_idx++] = -1;
566 : }
567 36 : matchctx->nmatches++;
568 : }
569 :
570 : Assert(array_idx <= array_len);
571 :
572 1180 : if (eml > 1)
573 : {
574 1180 : int64 maxsiz = eml * (int64) maxlen;
575 : int conv_bufsiz;
576 :
577 : /*
578 : * Make the conversion buffer large enough for any substring of
579 : * interest.
580 : *
581 : * Worst case: assume we need the maximum size (maxlen*eml), but take
582 : * advantage of the fact that the original string length in bytes is
583 : * an upper bound on the byte length of any fetched substring (and we
584 : * know that len+1 is safe to allocate because the varlena header is
585 : * longer than 1 byte).
586 : */
587 1180 : if (maxsiz > orig_len)
588 830 : conv_bufsiz = orig_len + 1;
589 : else
590 350 : conv_bufsiz = maxsiz + 1; /* safe since maxsiz < 2^30 */
591 :
592 1180 : matchctx->conv_buf = palloc(conv_bufsiz);
593 1180 : matchctx->conv_bufsiz = conv_bufsiz;
594 1180 : matchctx->wide_str = wide_str;
595 : }
596 : else
597 : {
598 : /* No need to keep the wide string if we're in a single-byte charset. */
599 0 : pfree(wide_str);
600 0 : matchctx->wide_str = NULL;
601 0 : matchctx->conv_buf = NULL;
602 0 : matchctx->conv_bufsiz = 0;
603 : }
604 :
605 : /* Clean up temp storage */
606 1180 : pfree(pmatch);
607 :
608 1180 : return matchctx;
609 : }
610 :
611 : /*
612 : * build_test_info_result - build output array describing compiled regexp
613 : *
614 : * This borrows some code from Tcl's TclRegAbout().
615 : */
616 : static ArrayType *
617 1180 : build_test_info_result(regex_t *cpattern, test_re_flags *flags)
618 : {
619 : /* Translation data for flag bits in regex_t.re_info */
620 : struct infoname
621 : {
622 : int bit;
623 : const char *text;
624 : };
625 : static const struct infoname infonames[] = {
626 : {REG_UBACKREF, "REG_UBACKREF"},
627 : {REG_ULOOKAROUND, "REG_ULOOKAROUND"},
628 : {REG_UBOUNDS, "REG_UBOUNDS"},
629 : {REG_UBRACES, "REG_UBRACES"},
630 : {REG_UBSALNUM, "REG_UBSALNUM"},
631 : {REG_UPBOTCH, "REG_UPBOTCH"},
632 : {REG_UBBS, "REG_UBBS"},
633 : {REG_UNONPOSIX, "REG_UNONPOSIX"},
634 : {REG_UUNSPEC, "REG_UUNSPEC"},
635 : {REG_UUNPORT, "REG_UUNPORT"},
636 : {REG_ULOCALE, "REG_ULOCALE"},
637 : {REG_UEMPTYMATCH, "REG_UEMPTYMATCH"},
638 : {REG_UIMPOSSIBLE, "REG_UIMPOSSIBLE"},
639 : {REG_USHORTEST, "REG_USHORTEST"},
640 : {0, NULL}
641 : };
642 : const struct infoname *inf;
643 : Datum elems[lengthof(infonames) + 1];
644 1180 : int nresults = 0;
645 : char buf[80];
646 : int dims[1];
647 : int lbs[1];
648 :
649 : /* Set up results: first, the number of subexpressions */
650 1180 : snprintf(buf, sizeof(buf), "%d", (int) cpattern->re_nsub);
651 1180 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
652 :
653 : /* Report individual info bit states */
654 17700 : for (inf = infonames; inf->bit != 0; inf++)
655 : {
656 16520 : if (cpattern->re_info & inf->bit)
657 : {
658 1516 : if (flags->info & inf->bit)
659 1516 : elems[nresults++] = PointerGetDatum(cstring_to_text(inf->text));
660 : else
661 : {
662 0 : snprintf(buf, sizeof(buf), "unexpected %s!", inf->text);
663 0 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
664 : }
665 : }
666 : else
667 : {
668 15004 : if (flags->info & inf->bit)
669 : {
670 0 : snprintf(buf, sizeof(buf), "missing %s!", inf->text);
671 0 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
672 : }
673 : }
674 : }
675 :
676 : /* And form an array */
677 1180 : dims[0] = nresults;
678 1180 : lbs[0] = 1;
679 : /* XXX: this hardcodes assumptions about the text type */
680 1180 : return construct_md_array(elems, NULL, 1, dims, lbs,
681 : TEXTOID, -1, false, TYPALIGN_INT);
682 : }
683 :
684 : /*
685 : * build_test_match_result - build output array for current match
686 : *
687 : * Note that if the indices flag is set, we don't need any strings,
688 : * just the location data.
689 : */
690 : static ArrayType *
691 962 : build_test_match_result(test_regex_ctx *matchctx)
692 : {
693 962 : char *buf = matchctx->conv_buf;
694 962 : Datum *elems = matchctx->elems;
695 962 : bool *nulls = matchctx->nulls;
696 962 : bool indices = matchctx->re_flags.indices;
697 : char bufstr[80];
698 : int dims[1];
699 : int lbs[1];
700 : int loc;
701 : int i;
702 :
703 : /* Extract matching substrings from the original string */
704 962 : loc = matchctx->next_match * matchctx->npatterns * 2;
705 2264 : for (i = 0; i < matchctx->npatterns; i++)
706 : {
707 1302 : int so = matchctx->match_locs[loc++];
708 1302 : int eo = matchctx->match_locs[loc++];
709 :
710 1302 : if (indices)
711 : {
712 : /* Report eo this way for consistency with Tcl */
713 168 : snprintf(bufstr, sizeof(bufstr), "%d %d",
714 : so, so < 0 ? eo : eo - 1);
715 168 : elems[i] = PointerGetDatum(cstring_to_text(bufstr));
716 168 : nulls[i] = false;
717 : }
718 1134 : else if (so < 0 || eo < 0)
719 : {
720 24 : elems[i] = (Datum) 0;
721 24 : nulls[i] = true;
722 : }
723 1110 : else if (buf)
724 : {
725 1110 : int len = pg_wchar2mb_with_len(matchctx->wide_str + so,
726 : buf,
727 : eo - so);
728 :
729 : Assert(len < matchctx->conv_bufsiz);
730 1110 : elems[i] = PointerGetDatum(cstring_to_text_with_len(buf, len));
731 1110 : nulls[i] = false;
732 : }
733 : else
734 : {
735 0 : elems[i] = DirectFunctionCall3(text_substr,
736 : PointerGetDatum(matchctx->orig_str),
737 : Int32GetDatum(so + 1),
738 : Int32GetDatum(eo - so));
739 0 : nulls[i] = false;
740 : }
741 : }
742 :
743 : /* In EXPECT indices mode, also report the "details" */
744 962 : if (indices && (matchctx->re_flags.cflags & REG_EXPECT))
745 : {
746 56 : int so = matchctx->details.rm_extend.rm_so;
747 56 : int eo = matchctx->details.rm_extend.rm_eo;
748 :
749 56 : snprintf(bufstr, sizeof(bufstr), "%d %d",
750 : so, so < 0 ? eo : eo - 1);
751 56 : elems[i] = PointerGetDatum(cstring_to_text(bufstr));
752 56 : nulls[i] = false;
753 56 : i++;
754 : }
755 :
756 : /* And form an array */
757 962 : dims[0] = i;
758 962 : lbs[0] = 1;
759 : /* XXX: this hardcodes assumptions about the text type */
760 962 : return construct_md_array(elems, nulls, 1, dims, lbs,
761 : TEXTOID, -1, false, TYPALIGN_INT);
762 : }
|
