Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * like.c
4 : * like expression handling code.
5 : *
6 : * NOTES
7 : * A big hack of the regexp.c code!! Contributed by
8 : * Keith Parks <emkxp01@mtcc.demon.co.uk> (7/95).
9 : *
10 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
11 : * Portions Copyright (c) 1994, Regents of the University of California
12 : *
13 : * IDENTIFICATION
14 : * src/backend/utils/adt/like.c
15 : *
16 : *-------------------------------------------------------------------------
17 : */
18 : #include "postgres.h"
19 :
20 : #include <ctype.h>
21 :
22 : #include "catalog/pg_collation.h"
23 : #include "mb/pg_wchar.h"
24 : #include "miscadmin.h"
25 : #include "utils/fmgrprotos.h"
26 : #include "utils/pg_locale.h"
27 : #include "varatt.h"
28 :
29 :
30 : #define LIKE_TRUE 1
31 : #define LIKE_FALSE 0
32 : #define LIKE_ABORT (-1)
33 :
34 :
35 : static int SB_MatchText(const char *t, int tlen, const char *p, int plen,
36 : pg_locale_t locale);
37 : static text *SB_do_like_escape(text *pat, text *esc);
38 :
39 : static int MB_MatchText(const char *t, int tlen, const char *p, int plen,
40 : pg_locale_t locale);
41 : static text *MB_do_like_escape(text *pat, text *esc);
42 :
43 : static int UTF8_MatchText(const char *t, int tlen, const char *p, int plen,
44 : pg_locale_t locale);
45 :
46 : static int SB_IMatchText(const char *t, int tlen, const char *p, int plen,
47 : pg_locale_t locale);
48 :
49 : static int GenericMatchText(const char *s, int slen, const char *p, int plen, Oid collation);
50 : static int Generic_Text_IC_like(text *str, text *pat, Oid collation);
51 :
52 : /*--------------------
53 : * Support routine for MatchText. Compares given multibyte streams
54 : * as wide characters. If they match, returns 1 otherwise returns 0.
55 : *--------------------
56 : */
57 : static inline int
58 924 : wchareq(const char *p1, const char *p2)
59 : {
60 : int p1_len;
61 :
62 : /* Optimization: quickly compare the first byte. */
63 924 : if (*p1 != *p2)
64 696 : return 0;
65 :
66 228 : p1_len = pg_mblen(p1);
67 228 : if (pg_mblen(p2) != p1_len)
68 0 : return 0;
69 :
70 : /* They are the same length */
71 456 : while (p1_len--)
72 : {
73 228 : if (*p1++ != *p2++)
74 0 : return 0;
75 : }
76 228 : return 1;
77 : }
78 :
79 : /*
80 : * Formerly we had a routine iwchareq() here that tried to do case-insensitive
81 : * comparison of multibyte characters. It did not work at all, however,
82 : * because it relied on tolower() which has a single-byte API ... and
83 : * towlower() wouldn't be much better since we have no suitably cheap way
84 : * of getting a single character transformed to the system's wchar_t format.
85 : * So now, we just downcase the strings using lower() and apply regular LIKE
86 : * comparison. This should be revisited when we install better locale support.
87 : */
88 :
89 : /*
90 : * We do handle case-insensitive matching for single-byte encodings using
91 : * fold-on-the-fly processing, however.
92 : */
93 : static char
94 0 : SB_lower_char(unsigned char c, pg_locale_t locale)
95 : {
96 0 : if (locale->ctype_is_c)
97 0 : return pg_ascii_tolower(c);
98 0 : else if (locale->is_default)
99 0 : return pg_tolower(c);
100 : else
101 0 : return tolower_l(c, locale->info.lt);
102 : }
103 :
104 :
105 : #define NextByte(p, plen) ((p)++, (plen)--)
106 :
107 : /* Set up to compile like_match.c for multibyte characters */
108 : #define CHAREQ(p1, p2) wchareq((p1), (p2))
109 : #define NextChar(p, plen) \
110 : do { int __l = pg_mblen(p); (p) +=__l; (plen) -=__l; } while (0)
111 : #define CopyAdvChar(dst, src, srclen) \
112 : do { int __l = pg_mblen(src); \
113 : (srclen) -= __l; \
114 : while (__l-- > 0) \
115 : *(dst)++ = *(src)++; \
116 : } while (0)
117 :
118 : #define MatchText MB_MatchText
119 : #define do_like_escape MB_do_like_escape
120 :
121 : #include "like_match.c"
122 :
123 : /* Set up to compile like_match.c for single-byte characters */
124 : #define CHAREQ(p1, p2) (*(p1) == *(p2))
125 : #define NextChar(p, plen) NextByte((p), (plen))
126 : #define CopyAdvChar(dst, src, srclen) (*(dst)++ = *(src)++, (srclen)--)
127 :
128 : #define MatchText SB_MatchText
129 : #define do_like_escape SB_do_like_escape
130 :
131 : #include "like_match.c"
132 :
133 : /* setup to compile like_match.c for single byte case insensitive matches */
134 : #define MATCH_LOWER(t, locale) SB_lower_char((unsigned char) (t), locale)
135 : #define NextChar(p, plen) NextByte((p), (plen))
136 : #define MatchText SB_IMatchText
137 :
138 : #include "like_match.c"
139 :
140 : /* setup to compile like_match.c for UTF8 encoding, using fast NextChar */
141 :
142 : #define NextChar(p, plen) \
143 : do { (p)++; (plen)--; } while ((plen) > 0 && (*(p) & 0xC0) == 0x80 )
144 : #define MatchText UTF8_MatchText
145 :
146 : #include "like_match.c"
147 :
148 : /* Generic for all cases not requiring inline case-folding */
149 : static inline int
150 1075262 : GenericMatchText(const char *s, int slen, const char *p, int plen, Oid collation)
151 : {
152 : pg_locale_t locale;
153 :
154 1075262 : if (!OidIsValid(collation))
155 : {
156 : /*
157 : * This typically means that the parser could not resolve a conflict
158 : * of implicit collations, so report it that way.
159 : */
160 0 : ereport(ERROR,
161 : (errcode(ERRCODE_INDETERMINATE_COLLATION),
162 : errmsg("could not determine which collation to use for LIKE"),
163 : errhint("Use the COLLATE clause to set the collation explicitly.")));
164 : }
165 :
166 1075262 : locale = pg_newlocale_from_collation(collation);
167 :
168 1075262 : if (pg_database_encoding_max_length() == 1)
169 80884 : return SB_MatchText(s, slen, p, plen, locale);
170 994378 : else if (GetDatabaseEncoding() == PG_UTF8)
171 994378 : return UTF8_MatchText(s, slen, p, plen, locale);
172 : else
173 0 : return MB_MatchText(s, slen, p, plen, locale);
174 : }
175 :
176 : static inline int
177 85160 : Generic_Text_IC_like(text *str, text *pat, Oid collation)
178 : {
179 : char *s,
180 : *p;
181 : int slen,
182 : plen;
183 : pg_locale_t locale;
184 :
185 85160 : if (!OidIsValid(collation))
186 : {
187 : /*
188 : * This typically means that the parser could not resolve a conflict
189 : * of implicit collations, so report it that way.
190 : */
191 0 : ereport(ERROR,
192 : (errcode(ERRCODE_INDETERMINATE_COLLATION),
193 : errmsg("could not determine which collation to use for ILIKE"),
194 : errhint("Use the COLLATE clause to set the collation explicitly.")));
195 : }
196 :
197 85160 : locale = pg_newlocale_from_collation(collation);
198 :
199 85160 : if (!locale->deterministic)
200 12 : ereport(ERROR,
201 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
202 : errmsg("nondeterministic collations are not supported for ILIKE")));
203 :
204 : /*
205 : * For efficiency reasons, in the single byte case we don't call lower()
206 : * on the pattern and text, but instead call SB_lower_char on each
207 : * character. In the multi-byte case we don't have much choice :-(. Also,
208 : * ICU does not support single-character case folding, so we go the long
209 : * way.
210 : */
211 :
212 85148 : if (pg_database_encoding_max_length() > 1 || (locale->provider == COLLPROVIDER_ICU))
213 : {
214 85148 : pat = DatumGetTextPP(DirectFunctionCall1Coll(lower, collation,
215 : PointerGetDatum(pat)));
216 85148 : p = VARDATA_ANY(pat);
217 85148 : plen = VARSIZE_ANY_EXHDR(pat);
218 85148 : str = DatumGetTextPP(DirectFunctionCall1Coll(lower, collation,
219 : PointerGetDatum(str)));
220 85148 : s = VARDATA_ANY(str);
221 85148 : slen = VARSIZE_ANY_EXHDR(str);
222 85148 : if (GetDatabaseEncoding() == PG_UTF8)
223 85148 : return UTF8_MatchText(s, slen, p, plen, 0);
224 : else
225 0 : return MB_MatchText(s, slen, p, plen, 0);
226 : }
227 : else
228 : {
229 0 : p = VARDATA_ANY(pat);
230 0 : plen = VARSIZE_ANY_EXHDR(pat);
231 0 : s = VARDATA_ANY(str);
232 0 : slen = VARSIZE_ANY_EXHDR(str);
233 0 : return SB_IMatchText(s, slen, p, plen, locale);
234 : }
235 : }
236 :
237 : /*
238 : * interface routines called by the function manager
239 : */
240 :
241 : Datum
242 173510 : namelike(PG_FUNCTION_ARGS)
243 : {
244 173510 : Name str = PG_GETARG_NAME(0);
245 173510 : text *pat = PG_GETARG_TEXT_PP(1);
246 : bool result;
247 : char *s,
248 : *p;
249 : int slen,
250 : plen;
251 :
252 173510 : s = NameStr(*str);
253 173510 : slen = strlen(s);
254 173510 : p = VARDATA_ANY(pat);
255 173510 : plen = VARSIZE_ANY_EXHDR(pat);
256 :
257 173510 : result = (GenericMatchText(s, slen, p, plen, PG_GET_COLLATION()) == LIKE_TRUE);
258 :
259 173510 : PG_RETURN_BOOL(result);
260 : }
261 :
262 : Datum
263 5358 : namenlike(PG_FUNCTION_ARGS)
264 : {
265 5358 : Name str = PG_GETARG_NAME(0);
266 5358 : text *pat = PG_GETARG_TEXT_PP(1);
267 : bool result;
268 : char *s,
269 : *p;
270 : int slen,
271 : plen;
272 :
273 5358 : s = NameStr(*str);
274 5358 : slen = strlen(s);
275 5358 : p = VARDATA_ANY(pat);
276 5358 : plen = VARSIZE_ANY_EXHDR(pat);
277 :
278 5358 : result = (GenericMatchText(s, slen, p, plen, PG_GET_COLLATION()) != LIKE_TRUE);
279 :
280 5358 : PG_RETURN_BOOL(result);
281 : }
282 :
283 : Datum
284 575616 : textlike(PG_FUNCTION_ARGS)
285 : {
286 575616 : text *str = PG_GETARG_TEXT_PP(0);
287 575616 : text *pat = PG_GETARG_TEXT_PP(1);
288 : bool result;
289 : char *s,
290 : *p;
291 : int slen,
292 : plen;
293 :
294 575616 : s = VARDATA_ANY(str);
295 575616 : slen = VARSIZE_ANY_EXHDR(str);
296 575616 : p = VARDATA_ANY(pat);
297 575616 : plen = VARSIZE_ANY_EXHDR(pat);
298 :
299 575616 : result = (GenericMatchText(s, slen, p, plen, PG_GET_COLLATION()) == LIKE_TRUE);
300 :
301 575610 : PG_RETURN_BOOL(result);
302 : }
303 :
304 : Datum
305 320778 : textnlike(PG_FUNCTION_ARGS)
306 : {
307 320778 : text *str = PG_GETARG_TEXT_PP(0);
308 320778 : text *pat = PG_GETARG_TEXT_PP(1);
309 : bool result;
310 : char *s,
311 : *p;
312 : int slen,
313 : plen;
314 :
315 320778 : s = VARDATA_ANY(str);
316 320778 : slen = VARSIZE_ANY_EXHDR(str);
317 320778 : p = VARDATA_ANY(pat);
318 320778 : plen = VARSIZE_ANY_EXHDR(pat);
319 :
320 320778 : result = (GenericMatchText(s, slen, p, plen, PG_GET_COLLATION()) != LIKE_TRUE);
321 :
322 320778 : PG_RETURN_BOOL(result);
323 : }
324 :
325 : Datum
326 12 : bytealike(PG_FUNCTION_ARGS)
327 : {
328 12 : bytea *str = PG_GETARG_BYTEA_PP(0);
329 12 : bytea *pat = PG_GETARG_BYTEA_PP(1);
330 : bool result;
331 : char *s,
332 : *p;
333 : int slen,
334 : plen;
335 :
336 12 : s = VARDATA_ANY(str);
337 12 : slen = VARSIZE_ANY_EXHDR(str);
338 12 : p = VARDATA_ANY(pat);
339 12 : plen = VARSIZE_ANY_EXHDR(pat);
340 :
341 12 : result = (SB_MatchText(s, slen, p, plen, 0) == LIKE_TRUE);
342 :
343 12 : PG_RETURN_BOOL(result);
344 : }
345 :
346 : Datum
347 12 : byteanlike(PG_FUNCTION_ARGS)
348 : {
349 12 : bytea *str = PG_GETARG_BYTEA_PP(0);
350 12 : bytea *pat = PG_GETARG_BYTEA_PP(1);
351 : bool result;
352 : char *s,
353 : *p;
354 : int slen,
355 : plen;
356 :
357 12 : s = VARDATA_ANY(str);
358 12 : slen = VARSIZE_ANY_EXHDR(str);
359 12 : p = VARDATA_ANY(pat);
360 12 : plen = VARSIZE_ANY_EXHDR(pat);
361 :
362 12 : result = (SB_MatchText(s, slen, p, plen, 0) != LIKE_TRUE);
363 :
364 12 : PG_RETURN_BOOL(result);
365 : }
366 :
367 : /*
368 : * Case-insensitive versions
369 : */
370 :
371 : Datum
372 16642 : nameiclike(PG_FUNCTION_ARGS)
373 : {
374 16642 : Name str = PG_GETARG_NAME(0);
375 16642 : text *pat = PG_GETARG_TEXT_PP(1);
376 : bool result;
377 : text *strtext;
378 :
379 16642 : strtext = DatumGetTextPP(DirectFunctionCall1(name_text,
380 : NameGetDatum(str)));
381 16642 : result = (Generic_Text_IC_like(strtext, pat, PG_GET_COLLATION()) == LIKE_TRUE);
382 :
383 16642 : PG_RETURN_BOOL(result);
384 : }
385 :
386 : Datum
387 6 : nameicnlike(PG_FUNCTION_ARGS)
388 : {
389 6 : Name str = PG_GETARG_NAME(0);
390 6 : text *pat = PG_GETARG_TEXT_PP(1);
391 : bool result;
392 : text *strtext;
393 :
394 6 : strtext = DatumGetTextPP(DirectFunctionCall1(name_text,
395 : NameGetDatum(str)));
396 6 : result = (Generic_Text_IC_like(strtext, pat, PG_GET_COLLATION()) != LIKE_TRUE);
397 :
398 6 : PG_RETURN_BOOL(result);
399 : }
400 :
401 : Datum
402 68456 : texticlike(PG_FUNCTION_ARGS)
403 : {
404 68456 : text *str = PG_GETARG_TEXT_PP(0);
405 68456 : text *pat = PG_GETARG_TEXT_PP(1);
406 : bool result;
407 :
408 68456 : result = (Generic_Text_IC_like(str, pat, PG_GET_COLLATION()) == LIKE_TRUE);
409 :
410 68444 : PG_RETURN_BOOL(result);
411 : }
412 :
413 : Datum
414 56 : texticnlike(PG_FUNCTION_ARGS)
415 : {
416 56 : text *str = PG_GETARG_TEXT_PP(0);
417 56 : text *pat = PG_GETARG_TEXT_PP(1);
418 : bool result;
419 :
420 56 : result = (Generic_Text_IC_like(str, pat, PG_GET_COLLATION()) != LIKE_TRUE);
421 :
422 56 : PG_RETURN_BOOL(result);
423 : }
424 :
425 : /*
426 : * like_escape() --- given a pattern and an ESCAPE string,
427 : * convert the pattern to use Postgres' standard backslash escape convention.
428 : */
429 : Datum
430 212 : like_escape(PG_FUNCTION_ARGS)
431 : {
432 212 : text *pat = PG_GETARG_TEXT_PP(0);
433 212 : text *esc = PG_GETARG_TEXT_PP(1);
434 : text *result;
435 :
436 212 : if (pg_database_encoding_max_length() == 1)
437 0 : result = SB_do_like_escape(pat, esc);
438 : else
439 212 : result = MB_do_like_escape(pat, esc);
440 :
441 212 : PG_RETURN_TEXT_P(result);
442 : }
443 :
444 : /*
445 : * like_escape_bytea() --- given a pattern and an ESCAPE string,
446 : * convert the pattern to use Postgres' standard backslash escape convention.
447 : */
448 : Datum
449 12 : like_escape_bytea(PG_FUNCTION_ARGS)
450 : {
451 12 : bytea *pat = PG_GETARG_BYTEA_PP(0);
452 12 : bytea *esc = PG_GETARG_BYTEA_PP(1);
453 12 : bytea *result = SB_do_like_escape((text *) pat, (text *) esc);
454 :
455 12 : PG_RETURN_BYTEA_P((bytea *) result);
456 : }
|
