Line data Source code
1 : /*--------------------------------------------------------------------------
2 : *
3 : * test_regex.c
4 : * Test harness for 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 : * 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 "miscadmin.h"
19 : #include "regex/regex.h"
20 : #include "utils/array.h"
21 : #include "utils/builtins.h"
22 :
23 2 : PG_MODULE_MAGIC;
24 :
25 :
26 : /* all the options of interest for regex functions */
27 : typedef struct test_re_flags
28 : {
29 : int cflags; /* compile flags for Spencer's regex code */
30 : int eflags; /* execute flags for Spencer's regex code */
31 : long info; /* expected re_info bits */
32 : bool glob; /* do it globally (for each occurrence) */
33 : bool indices; /* report indices not actual strings */
34 : bool partial; /* expect partial match */
35 : } test_re_flags;
36 :
37 : /* cross-call state for test_regex() */
38 : typedef struct test_regex_ctx
39 : {
40 : test_re_flags re_flags; /* flags */
41 : rm_detail_t details; /* "details" from execution */
42 : text *orig_str; /* data string in original TEXT form */
43 : int nmatches; /* number of places where pattern matched */
44 : int npatterns; /* number of capturing subpatterns */
45 : /* We store start char index and end+1 char index for each match */
46 : /* so the number of entries in match_locs is nmatches * npatterns * 2 */
47 : int *match_locs; /* 0-based character indexes */
48 : int next_match; /* 0-based index of next match to process */
49 : /* workspace for build_test_match_result() */
50 : Datum *elems; /* has npatterns+1 elements */
51 : bool *nulls; /* has npatterns+1 elements */
52 : pg_wchar *wide_str; /* wide-char version of original string */
53 : char *conv_buf; /* conversion buffer, if needed */
54 : int conv_bufsiz; /* size thereof */
55 : } test_regex_ctx;
56 :
57 : /* Local functions */
58 : static void test_re_compile(text *text_re, int cflags, Oid collation,
59 : regex_t *result_re);
60 : static void parse_test_flags(test_re_flags *flags, text *opts);
61 : static test_regex_ctx *setup_test_matches(text *orig_str,
62 : regex_t *cpattern,
63 : test_re_flags *re_flags,
64 : Oid collation,
65 : bool use_subpatterns);
66 : static ArrayType *build_test_info_result(regex_t *cpattern,
67 : test_re_flags *flags);
68 : static ArrayType *build_test_match_result(test_regex_ctx *matchctx);
69 :
70 :
71 : /*
72 : * test_regex(pattern text, string text, flags text) returns setof text[]
73 : *
74 : * This is largely based on regexp.c's regexp_matches, with additions
75 : * for debugging purposes.
76 : */
77 4 : PG_FUNCTION_INFO_V1(test_regex);
78 :
79 : Datum
80 3382 : test_regex(PG_FUNCTION_ARGS)
81 : {
82 : FuncCallContext *funcctx;
83 : test_regex_ctx *matchctx;
84 : ArrayType *result_ary;
85 :
86 3382 : if (SRF_IS_FIRSTCALL())
87 : {
88 1334 : text *pattern = PG_GETARG_TEXT_PP(0);
89 1334 : text *flags = PG_GETARG_TEXT_PP(2);
90 1334 : Oid collation = PG_GET_COLLATION();
91 : test_re_flags re_flags;
92 : regex_t cpattern;
93 : MemoryContext oldcontext;
94 :
95 1334 : funcctx = SRF_FIRSTCALL_INIT();
96 1334 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
97 :
98 : /* Determine options */
99 1334 : parse_test_flags(&re_flags, flags);
100 :
101 : /* set up the compiled pattern */
102 1334 : test_re_compile(pattern, re_flags.cflags, collation, &cpattern);
103 :
104 : /* be sure to copy the input string into the multi-call ctx */
105 1130 : matchctx = setup_test_matches(PG_GETARG_TEXT_P_COPY(1), &cpattern,
106 : &re_flags,
107 : collation,
108 : true);
109 :
110 : /* Pre-create workspace that build_test_match_result needs */
111 2260 : matchctx->elems = (Datum *) palloc(sizeof(Datum) *
112 1130 : (matchctx->npatterns + 1));
113 2260 : matchctx->nulls = (bool *) palloc(sizeof(bool) *
114 1130 : (matchctx->npatterns + 1));
115 :
116 1130 : MemoryContextSwitchTo(oldcontext);
117 1130 : funcctx->user_fctx = (void *) matchctx;
118 :
119 : /*
120 : * Return the first result row, which is info equivalent to Tcl's
121 : * "regexp -about" output
122 : */
123 1130 : result_ary = build_test_info_result(&cpattern, &re_flags);
124 :
125 1130 : pg_regfree(&cpattern);
126 :
127 1130 : SRF_RETURN_NEXT(funcctx, PointerGetDatum(result_ary));
128 : }
129 : else
130 : {
131 : /* Each subsequent row describes one match */
132 2048 : funcctx = SRF_PERCALL_SETUP();
133 2048 : matchctx = (test_regex_ctx *) funcctx->user_fctx;
134 :
135 2048 : if (matchctx->next_match < matchctx->nmatches)
136 : {
137 918 : result_ary = build_test_match_result(matchctx);
138 918 : matchctx->next_match++;
139 918 : SRF_RETURN_NEXT(funcctx, PointerGetDatum(result_ary));
140 : }
141 : }
142 :
143 1130 : SRF_RETURN_DONE(funcctx);
144 : }
145 :
146 :
147 : /*
148 : * test_re_compile - compile a RE
149 : *
150 : * text_re --- the pattern, expressed as a TEXT object
151 : * cflags --- compile options for the pattern
152 : * collation --- collation to use for LC_CTYPE-dependent behavior
153 : * result_re --- output, compiled RE is stored here
154 : *
155 : * Pattern is given in the database encoding. We internally convert to
156 : * an array of pg_wchar, which is what Spencer's regex package wants.
157 : *
158 : * Caller must eventually pg_regfree the resulting RE to avoid memory leaks.
159 : */
160 : static void
161 1334 : test_re_compile(text *text_re, int cflags, Oid collation,
162 : regex_t *result_re)
163 : {
164 1334 : int text_re_len = VARSIZE_ANY_EXHDR(text_re);
165 1334 : char *text_re_val = VARDATA_ANY(text_re);
166 : pg_wchar *pattern;
167 : int pattern_len;
168 : int regcomp_result;
169 : char errMsg[100];
170 :
171 : /* Convert pattern string to wide characters */
172 1334 : pattern = (pg_wchar *) palloc((text_re_len + 1) * sizeof(pg_wchar));
173 1334 : pattern_len = pg_mb2wchar_with_len(text_re_val,
174 : pattern,
175 : text_re_len);
176 :
177 1334 : regcomp_result = pg_regcomp(result_re,
178 : pattern,
179 : pattern_len,
180 : cflags,
181 : collation);
182 :
183 1334 : pfree(pattern);
184 :
185 1334 : if (regcomp_result != REG_OKAY)
186 : {
187 : /* re didn't compile (no need for pg_regfree, if so) */
188 204 : pg_regerror(regcomp_result, result_re, errMsg, sizeof(errMsg));
189 204 : ereport(ERROR,
190 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
191 : errmsg("invalid regular expression: %s", errMsg)));
192 : }
193 1130 : }
194 :
195 : /*
196 : * test_re_execute - execute a RE on pg_wchar data
197 : *
198 : * Returns true on match, false on no match
199 : * Arguments are as for pg_regexec
200 : */
201 : static bool
202 1130 : test_re_execute(regex_t *re, pg_wchar *data, int data_len,
203 : int start_search,
204 : rm_detail_t *details,
205 : int nmatch, regmatch_t *pmatch,
206 : int eflags)
207 : {
208 : int regexec_result;
209 : char errMsg[100];
210 :
211 : /* Initialize match locations in case engine doesn't */
212 1130 : details->rm_extend.rm_so = -1;
213 1130 : details->rm_extend.rm_eo = -1;
214 2832 : for (int i = 0; i < nmatch; i++)
215 : {
216 1702 : pmatch[i].rm_so = -1;
217 1702 : pmatch[i].rm_eo = -1;
218 : }
219 :
220 : /* Perform RE match and return result */
221 1130 : regexec_result = pg_regexec(re,
222 : data,
223 : data_len,
224 : start_search,
225 : details,
226 : nmatch,
227 : pmatch,
228 : eflags);
229 :
230 1130 : if (regexec_result != REG_OKAY && regexec_result != REG_NOMATCH)
231 : {
232 : /* re failed??? */
233 0 : pg_regerror(regexec_result, re, errMsg, sizeof(errMsg));
234 0 : ereport(ERROR,
235 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
236 : errmsg("regular expression failed: %s", errMsg)));
237 : }
238 :
239 1130 : return (regexec_result == REG_OKAY);
240 : }
241 :
242 :
243 : /*
244 : * parse_test_flags - parse the flags argument
245 : *
246 : * flags --- output argument, filled with desired options
247 : * opts --- TEXT object, or NULL for defaults
248 : */
249 : static void
250 1334 : parse_test_flags(test_re_flags *flags, text *opts)
251 : {
252 : /* these defaults must match Tcl's */
253 1334 : int cflags = REG_ADVANCED;
254 1334 : int eflags = 0;
255 1334 : long info = 0;
256 :
257 1334 : flags->glob = false;
258 1334 : flags->indices = false;
259 1334 : flags->partial = false;
260 :
261 1334 : if (opts)
262 : {
263 1334 : char *opt_p = VARDATA_ANY(opts);
264 1334 : int opt_len = VARSIZE_ANY_EXHDR(opts);
265 : int i;
266 :
267 3620 : for (i = 0; i < opt_len; i++)
268 : {
269 2286 : switch (opt_p[i])
270 : {
271 156 : case '-':
272 : /* allowed, no-op */
273 156 : break;
274 14 : case '!':
275 14 : flags->partial = true;
276 14 : break;
277 0 : case '*':
278 : /* test requires Unicode --- ignored here */
279 0 : break;
280 106 : case '0':
281 106 : flags->indices = true;
282 106 : break;
283 :
284 : /* These flags correspond to user-exposed RE options: */
285 0 : case 'g': /* global match */
286 0 : flags->glob = true;
287 0 : break;
288 38 : case 'i': /* case insensitive */
289 38 : cflags |= REG_ICASE;
290 38 : break;
291 70 : case 'n': /* \n affects ^ $ . [^ */
292 70 : cflags |= REG_NEWLINE;
293 70 : break;
294 4 : case 'p': /* ~Perl, \n affects . [^ */
295 4 : cflags |= REG_NLSTOP;
296 4 : cflags &= ~REG_NLANCH;
297 4 : break;
298 4 : case 'w': /* weird, \n affects ^ $ only */
299 4 : cflags &= ~REG_NLSTOP;
300 4 : cflags |= REG_NLANCH;
301 4 : break;
302 26 : case 'x': /* expanded syntax */
303 26 : cflags |= REG_EXPANDED;
304 26 : break;
305 :
306 : /* These flags correspond to Tcl's -xflags options: */
307 4 : case 'a':
308 4 : cflags |= REG_ADVF;
309 4 : break;
310 262 : case 'b':
311 262 : cflags &= ~REG_ADVANCED;
312 262 : break;
313 22 : case 'c':
314 :
315 : /*
316 : * Tcl calls this TCL_REG_CANMATCH, but it's really
317 : * REG_EXPECT. In this implementation we must also set
318 : * the partial and indices flags, so that
319 : * setup_test_matches and build_test_match_result will
320 : * emit the desired data. (They'll emit more fields than
321 : * Tcl would, but that's fine.)
322 : */
323 22 : cflags |= REG_EXPECT;
324 22 : flags->partial = true;
325 22 : flags->indices = true;
326 22 : break;
327 20 : case 'e':
328 20 : cflags &= ~REG_ADVANCED;
329 20 : cflags |= REG_EXTENDED;
330 20 : break;
331 12 : case 'q':
332 12 : cflags &= ~REG_ADVANCED;
333 12 : cflags |= REG_QUOTE;
334 12 : break;
335 4 : case 'o': /* o for opaque */
336 4 : cflags |= REG_NOSUB;
337 4 : break;
338 4 : case 's': /* s for start */
339 4 : cflags |= REG_BOSONLY;
340 4 : break;
341 12 : case '+':
342 12 : cflags |= REG_FAKE;
343 12 : break;
344 0 : case ',':
345 0 : cflags |= REG_PROGRESS;
346 0 : break;
347 0 : case '.':
348 0 : cflags |= REG_DUMP;
349 0 : break;
350 0 : case ':':
351 0 : eflags |= REG_MTRACE;
352 0 : break;
353 0 : case ';':
354 0 : eflags |= REG_FTRACE;
355 0 : break;
356 12 : case '^':
357 12 : eflags |= REG_NOTBOL;
358 12 : break;
359 8 : case '$':
360 8 : eflags |= REG_NOTEOL;
361 8 : break;
362 34 : case 't':
363 34 : cflags |= REG_EXPECT;
364 34 : break;
365 10 : case '%':
366 10 : eflags |= REG_SMALL;
367 10 : break;
368 :
369 : /* These flags define expected info bits: */
370 10 : case 'A':
371 10 : info |= REG_UBSALNUM;
372 10 : break;
373 8 : case 'B':
374 8 : info |= REG_UBRACES;
375 8 : break;
376 70 : case 'E':
377 70 : info |= REG_UBBS;
378 70 : break;
379 68 : case 'H':
380 68 : info |= REG_ULOOKAROUND;
381 68 : break;
382 20 : case 'I':
383 20 : info |= REG_UIMPOSSIBLE;
384 20 : break;
385 290 : case 'L':
386 290 : info |= REG_ULOCALE;
387 290 : break;
388 76 : case 'M':
389 76 : info |= REG_UUNPORT;
390 76 : break;
391 92 : case 'N':
392 92 : info |= REG_UEMPTYMATCH;
393 92 : break;
394 584 : case 'P':
395 584 : info |= REG_UNONPOSIX;
396 584 : break;
397 70 : case 'Q':
398 70 : info |= REG_UBOUNDS;
399 70 : break;
400 84 : case 'R':
401 84 : info |= REG_UBACKREF;
402 84 : break;
403 50 : case 'S':
404 50 : info |= REG_UUNSPEC;
405 50 : break;
406 40 : case 'T':
407 40 : info |= REG_USHORTEST;
408 40 : break;
409 2 : case 'U':
410 2 : info |= REG_UPBOTCH;
411 2 : break;
412 :
413 0 : default:
414 0 : ereport(ERROR,
415 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
416 : errmsg("invalid regular expression test option: \"%.*s\"",
417 : pg_mblen(opt_p + i), opt_p + i)));
418 : break;
419 : }
420 : }
421 : }
422 1334 : flags->cflags = cflags;
423 1334 : flags->eflags = eflags;
424 1334 : flags->info = info;
425 1334 : }
426 :
427 : /*
428 : * setup_test_matches --- do the initial matching
429 : *
430 : * To simplify memory management, we do all the matching in one swoop.
431 : * The returned test_regex_ctx contains the locations of all the substrings
432 : * matching the pattern.
433 : */
434 : static test_regex_ctx *
435 1130 : setup_test_matches(text *orig_str,
436 : regex_t *cpattern, test_re_flags *re_flags,
437 : Oid collation,
438 : bool use_subpatterns)
439 : {
440 1130 : test_regex_ctx *matchctx = palloc0(sizeof(test_regex_ctx));
441 1130 : int eml = pg_database_encoding_max_length();
442 : int orig_len;
443 : pg_wchar *wide_str;
444 : int wide_len;
445 : regmatch_t *pmatch;
446 : int pmatch_len;
447 : int array_len;
448 : int array_idx;
449 : int prev_match_end;
450 : int start_search;
451 1130 : int maxlen = 0; /* largest fetch length in characters */
452 :
453 : /* save flags */
454 1130 : matchctx->re_flags = *re_flags;
455 :
456 : /* save original string --- we'll extract result substrings from it */
457 1130 : matchctx->orig_str = orig_str;
458 :
459 : /* convert string to pg_wchar form for matching */
460 1130 : orig_len = VARSIZE_ANY_EXHDR(orig_str);
461 1130 : wide_str = (pg_wchar *) palloc(sizeof(pg_wchar) * (orig_len + 1));
462 1130 : wide_len = pg_mb2wchar_with_len(VARDATA_ANY(orig_str), wide_str, orig_len);
463 :
464 : /* do we want to remember subpatterns? */
465 1130 : if (use_subpatterns && cpattern->re_nsub > 0)
466 : {
467 254 : matchctx->npatterns = cpattern->re_nsub + 1;
468 254 : pmatch_len = cpattern->re_nsub + 1;
469 : }
470 : else
471 : {
472 876 : use_subpatterns = false;
473 876 : matchctx->npatterns = 1;
474 876 : pmatch_len = 1;
475 : }
476 :
477 : /* temporary output space for RE package */
478 1130 : pmatch = palloc(sizeof(regmatch_t) * pmatch_len);
479 :
480 : /*
481 : * the real output space (grown dynamically if needed)
482 : *
483 : * use values 2^n-1, not 2^n, so that we hit the limit at 2^28-1 rather
484 : * than at 2^27
485 : */
486 1130 : array_len = re_flags->glob ? 255 : 31;
487 1130 : matchctx->match_locs = (int *) palloc(sizeof(int) * array_len);
488 1130 : array_idx = 0;
489 :
490 : /* search for the pattern, perhaps repeatedly */
491 1130 : prev_match_end = 0;
492 1130 : start_search = 0;
493 1130 : while (test_re_execute(cpattern, wide_str, wide_len,
494 : start_search,
495 : &matchctx->details,
496 : pmatch_len, pmatch,
497 : re_flags->eflags))
498 : {
499 : /* enlarge output space if needed */
500 882 : while (array_idx + matchctx->npatterns * 2 + 1 > array_len)
501 : {
502 0 : array_len += array_len + 1; /* 2^n-1 => 2^(n+1)-1 */
503 0 : if (array_len > MaxAllocSize / sizeof(int))
504 0 : ereport(ERROR,
505 : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
506 : errmsg("too many regular expression matches")));
507 0 : matchctx->match_locs = (int *) repalloc(matchctx->match_locs,
508 : sizeof(int) * array_len);
509 : }
510 :
511 : /* save this match's locations */
512 2100 : for (int i = 0; i < matchctx->npatterns; i++)
513 : {
514 1218 : int so = pmatch[i].rm_so;
515 1218 : int eo = pmatch[i].rm_eo;
516 :
517 1218 : matchctx->match_locs[array_idx++] = so;
518 1218 : matchctx->match_locs[array_idx++] = eo;
519 1218 : if (so >= 0 && eo >= 0 && (eo - so) > maxlen)
520 834 : maxlen = (eo - so);
521 : }
522 882 : matchctx->nmatches++;
523 882 : prev_match_end = pmatch[0].rm_eo;
524 :
525 : /* if not glob, stop after one match */
526 882 : if (!re_flags->glob)
527 882 : break;
528 :
529 : /*
530 : * Advance search position. Normally we start the next search at the
531 : * end of the previous match; but if the match was of zero length, we
532 : * have to advance by one character, or we'd just find the same match
533 : * again.
534 : */
535 0 : start_search = prev_match_end;
536 0 : if (pmatch[0].rm_so == pmatch[0].rm_eo)
537 0 : start_search++;
538 0 : if (start_search > wide_len)
539 0 : break;
540 : }
541 :
542 : /*
543 : * If we had no match, but "partial" and "indices" are set, emit the
544 : * details.
545 : */
546 1130 : if (matchctx->nmatches == 0 && re_flags->partial && re_flags->indices)
547 : {
548 : /* enlarge output space if needed */
549 36 : while (array_idx + matchctx->npatterns * 2 + 1 > array_len)
550 : {
551 0 : array_len += array_len + 1; /* 2^n-1 => 2^(n+1)-1 */
552 0 : if (array_len > MaxAllocSize / sizeof(int))
553 0 : ereport(ERROR,
554 : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
555 : errmsg("too many regular expression matches")));
556 0 : matchctx->match_locs = (int *) repalloc(matchctx->match_locs,
557 : sizeof(int) * array_len);
558 : }
559 :
560 36 : matchctx->match_locs[array_idx++] = matchctx->details.rm_extend.rm_so;
561 36 : matchctx->match_locs[array_idx++] = matchctx->details.rm_extend.rm_eo;
562 : /* we don't have pmatch data, so emit -1 */
563 40 : for (int i = 1; i < matchctx->npatterns; i++)
564 : {
565 4 : matchctx->match_locs[array_idx++] = -1;
566 4 : matchctx->match_locs[array_idx++] = -1;
567 : }
568 36 : matchctx->nmatches++;
569 : }
570 :
571 : Assert(array_idx <= array_len);
572 :
573 1130 : if (eml > 1)
574 : {
575 0 : int64 maxsiz = eml * (int64) maxlen;
576 : int conv_bufsiz;
577 :
578 : /*
579 : * Make the conversion buffer large enough for any substring of
580 : * interest.
581 : *
582 : * Worst case: assume we need the maximum size (maxlen*eml), but take
583 : * advantage of the fact that the original string length in bytes is
584 : * an upper bound on the byte length of any fetched substring (and we
585 : * know that len+1 is safe to allocate because the varlena header is
586 : * longer than 1 byte).
587 : */
588 0 : if (maxsiz > orig_len)
589 0 : conv_bufsiz = orig_len + 1;
590 : else
591 0 : conv_bufsiz = maxsiz + 1; /* safe since maxsiz < 2^30 */
592 :
593 0 : matchctx->conv_buf = palloc(conv_bufsiz);
594 0 : matchctx->conv_bufsiz = conv_bufsiz;
595 0 : matchctx->wide_str = wide_str;
596 : }
597 : else
598 : {
599 : /* No need to keep the wide string if we're in a single-byte charset. */
600 1130 : pfree(wide_str);
601 1130 : matchctx->wide_str = NULL;
602 1130 : matchctx->conv_buf = NULL;
603 1130 : matchctx->conv_bufsiz = 0;
604 : }
605 :
606 : /* Clean up temp storage */
607 1130 : pfree(pmatch);
608 :
609 1130 : return matchctx;
610 : }
611 :
612 : /*
613 : * build_test_info_result - build output array describing compiled regexp
614 : *
615 : * This borrows some code from Tcl's TclRegAbout().
616 : */
617 : static ArrayType *
618 1130 : build_test_info_result(regex_t *cpattern, test_re_flags *flags)
619 : {
620 : /* Translation data for flag bits in regex_t.re_info */
621 : struct infoname
622 : {
623 : int bit;
624 : const char *text;
625 : };
626 : static const struct infoname infonames[] = {
627 : {REG_UBACKREF, "REG_UBACKREF"},
628 : {REG_ULOOKAROUND, "REG_ULOOKAROUND"},
629 : {REG_UBOUNDS, "REG_UBOUNDS"},
630 : {REG_UBRACES, "REG_UBRACES"},
631 : {REG_UBSALNUM, "REG_UBSALNUM"},
632 : {REG_UPBOTCH, "REG_UPBOTCH"},
633 : {REG_UBBS, "REG_UBBS"},
634 : {REG_UNONPOSIX, "REG_UNONPOSIX"},
635 : {REG_UUNSPEC, "REG_UUNSPEC"},
636 : {REG_UUNPORT, "REG_UUNPORT"},
637 : {REG_ULOCALE, "REG_ULOCALE"},
638 : {REG_UEMPTYMATCH, "REG_UEMPTYMATCH"},
639 : {REG_UIMPOSSIBLE, "REG_UIMPOSSIBLE"},
640 : {REG_USHORTEST, "REG_USHORTEST"},
641 : {0, NULL}
642 : };
643 : const struct infoname *inf;
644 : Datum elems[lengthof(infonames) + 1];
645 1130 : int nresults = 0;
646 : char buf[80];
647 : int dims[1];
648 : int lbs[1];
649 :
650 : /* Set up results: first, the number of subexpressions */
651 1130 : snprintf(buf, sizeof(buf), "%d", (int) cpattern->re_nsub);
652 1130 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
653 :
654 : /* Report individual info bit states */
655 16950 : for (inf = infonames; inf->bit != 0; inf++)
656 : {
657 15820 : if (cpattern->re_info & inf->bit)
658 : {
659 1426 : if (flags->info & inf->bit)
660 1426 : elems[nresults++] = PointerGetDatum(cstring_to_text(inf->text));
661 : else
662 : {
663 0 : snprintf(buf, sizeof(buf), "unexpected %s!", inf->text);
664 0 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
665 : }
666 : }
667 : else
668 : {
669 14394 : if (flags->info & inf->bit)
670 : {
671 0 : snprintf(buf, sizeof(buf), "missing %s!", inf->text);
672 0 : elems[nresults++] = PointerGetDatum(cstring_to_text(buf));
673 : }
674 : }
675 : }
676 :
677 : /* And form an array */
678 1130 : dims[0] = nresults;
679 1130 : lbs[0] = 1;
680 : /* XXX: this hardcodes assumptions about the text type */
681 1130 : return construct_md_array(elems, NULL, 1, dims, lbs,
682 : TEXTOID, -1, false, TYPALIGN_INT);
683 : }
684 :
685 : /*
686 : * build_test_match_result - build output array for current match
687 : *
688 : * Note that if the indices flag is set, we don't need any strings,
689 : * just the location data.
690 : */
691 : static ArrayType *
692 918 : build_test_match_result(test_regex_ctx *matchctx)
693 : {
694 918 : char *buf = matchctx->conv_buf;
695 918 : Datum *elems = matchctx->elems;
696 918 : bool *nulls = matchctx->nulls;
697 918 : bool indices = matchctx->re_flags.indices;
698 : char bufstr[80];
699 : int dims[1];
700 : int lbs[1];
701 : int loc;
702 : int i;
703 :
704 : /* Extract matching substrings from the original string */
705 918 : loc = matchctx->next_match * matchctx->npatterns * 2;
706 2176 : for (i = 0; i < matchctx->npatterns; i++)
707 : {
708 1258 : int so = matchctx->match_locs[loc++];
709 1258 : int eo = matchctx->match_locs[loc++];
710 :
711 1258 : if (indices)
712 : {
713 : /* Report eo this way for consistency with Tcl */
714 168 : snprintf(bufstr, sizeof(bufstr), "%d %d",
715 : so, so < 0 ? eo : eo - 1);
716 168 : elems[i] = PointerGetDatum(cstring_to_text(bufstr));
717 168 : nulls[i] = false;
718 : }
719 1090 : else if (so < 0 || eo < 0)
720 : {
721 24 : elems[i] = (Datum) 0;
722 24 : nulls[i] = true;
723 : }
724 1066 : else if (buf)
725 : {
726 0 : int len = pg_wchar2mb_with_len(matchctx->wide_str + so,
727 : buf,
728 : eo - so);
729 :
730 : Assert(len < matchctx->conv_bufsiz);
731 0 : elems[i] = PointerGetDatum(cstring_to_text_with_len(buf, len));
732 0 : nulls[i] = false;
733 : }
734 : else
735 : {
736 1066 : elems[i] = DirectFunctionCall3(text_substr,
737 : PointerGetDatum(matchctx->orig_str),
738 : Int32GetDatum(so + 1),
739 : Int32GetDatum(eo - so));
740 1066 : nulls[i] = false;
741 : }
742 : }
743 :
744 : /* In EXPECT indices mode, also report the "details" */
745 918 : if (indices && (matchctx->re_flags.cflags & REG_EXPECT))
746 : {
747 56 : int so = matchctx->details.rm_extend.rm_so;
748 56 : int eo = matchctx->details.rm_extend.rm_eo;
749 :
750 56 : snprintf(bufstr, sizeof(bufstr), "%d %d",
751 : so, so < 0 ? eo : eo - 1);
752 56 : elems[i] = PointerGetDatum(cstring_to_text(bufstr));
753 56 : nulls[i] = false;
754 56 : i++;
755 : }
756 :
757 : /* And form an array */
758 918 : dims[0] = i;
759 918 : lbs[0] = 1;
760 : /* XXX: this hardcodes assumptions about the text type */
761 918 : return construct_md_array(elems, nulls, 1, dims, lbs,
762 : TEXTOID, -1, false, TYPALIGN_INT);
763 : }
|