1/*-------------------------------------------------------------------------
4 * LIKE pattern matching internal code.
6 * This file is included by like.c four times, to provide matching code for
7 * (1) single-byte encodings, (2) UTF8, (3) other multi-byte encodings,
8 * and (4) case insensitive matches in single-byte encodings.
9 * (UTF8 is a special case because we can use a much more efficient version
10 * of NextChar than can be used for general multi-byte encodings.)
12 * Before the inclusion, we need to define the following macros:
15 * MatchText - to name of function wanted
16 * do_like_escape - name of function if wanted - needs CHAREQ and CopyAdvChar
17 * MATCH_LOWER - define for case (4) to specify case folding for 1-byte chars
19 * Copyright (c) 1996-2025, PostgreSQL Global Development Group
22 * src/backend/utils/adt/like_match.c
24 *-------------------------------------------------------------------------
28 * Originally written by Rich $alz, mirror!rs, Wed Nov 26 19:03:17 EST 1986.
29 * Rich $alz is now <rsalz@bbn.com>.
30 * Special thanks to Lars Mathiesen <thorinn@diku.dk> for the
33 * This code was shamelessly stolen from the "pql" code by myself and
34 * slightly modified :)
36 * All references to the word "star" were replaced by "percent"
37 * All references to the word "wild" were replaced by "like"
39 * All the nice shell RE matching stuff was replaced by just "_" and "%"
41 * As I don't have a copy of the SQL standard handy I wasn't sure whether
42 * to leave in the '\' escape character handling.
44 * Keith Parks. <keith@mtcc.demon.co.uk>
46 * SQL lets you specify the escape character by saying
47 * LIKE <pattern> ESCAPE <escape character>. We are a small operation
48 * so we force you to use '\'. - ay 7/95
50 * Now we have the like_escape() function that converts patterns with
51 * any specified escape character (or none at all) to the internal
52 * default escape character, which is still '\'. - tgl 9/2000
54 * The code is rewritten to avoid requiring null-terminated strings,
55 * which in turn allows us to leave out some memcpy() operations.
56 * This code should be faster and take less memory, but no promises...
57 * - thomas 2000年08月06日
62 * Match text and pattern, return LIKE_TRUE, LIKE_FALSE, or LIKE_ABORT.
64 * LIKE_TRUE: they match
65 * LIKE_FALSE: they don't match
66 * LIKE_ABORT: not only don't they match, but the text is too short.
68 * If LIKE_ABORT is returned, then no suffix of the text can match the
69 * pattern either, so an upper-level % scan can stop scanning now.
74#define GETCHAR(t, locale) MATCH_LOWER(t, locale)
76 #define GETCHAR(t, locale) (t)
82 /* Fast path for match-everything pattern */
83 if (plen == 1 && *p ==
'%')
86 /* Since this function recurses, it could be driven to stack overflow */
90 * In this loop, we advance by char when matching wildcards (and thus on
91 * recursive entry to this function we are properly char-synced). On other
92 * occasions it is safe to advance by byte, as the text and pattern will
93 * be in lockstep. This allows us to perform all comparisons between the
94 * text and pattern on a byte by byte basis, even for multi-byte
97 while (tlen > 0 && plen > 0)
101 /* Next pattern byte must match literally, whatever it is */
103 /* ... and there had better be one, per SQL standard */
106 (
errcode(ERRCODE_INVALID_ESCAPE_SEQUENCE),
107 errmsg(
"LIKE pattern must not end with escape character")));
116 * % processing is essentially a search for a text position at
117 * which the remainder of the text matches the remainder of the
118 * pattern, using a recursive call to check each potential match.
120 * If there are wildcards immediately following the %, we can skip
121 * over them first, using the idea that any sequence of N _'s and
122 * one or more %'s is equivalent to N _'s and one % (ie, it will
123 * match any sequence of at least N text characters). In this way
124 * we will always run the recursive search loop using a pattern
125 * fragment that begins with a literal character-to-match, thereby
126 * not recursing more than we have to.
136 /* If not enough text left to match the pattern, ABORT */
143 break;
/* Reached a non-wildcard pattern char */
147 * If we're at end of pattern, match: we have a trailing % which
148 * matches any remaining text string.
154 * Otherwise, scan for a text position at which we can match the
155 * rest of the pattern. The first remaining pattern char is known
156 * to be a regular or escaped literal character, so we can compare
157 * the first pattern byte to each text byte to avoid recursing
158 * more than we have to. This fact also guarantees that we don't
159 * have to consider a match to the zero-length substring at the
160 * end of the text. With a nondeterministic collation, we can't
161 * rely on the first bytes being equal, so we have to recurse in
168 (
errcode(ERRCODE_INVALID_ESCAPE_SEQUENCE),
169 errmsg(
"LIKE pattern must not end with escape character")));
182 return matched;
/* TRUE or ABORT */
189 * End of text with no match, so no point in trying later places
190 * to start matching this pattern.
196 /* _ matches any single character, and we know there is one */
204 * For nondeterministic locales, we find the next substring of the
205 * pattern that does not contain wildcards and try to find a
206 * matching substring in the text. Crucially, we cannot do this
207 * character by character, as in the normal case, but must do it
208 * substring by substring, partitioned by the wildcard characters.
209 * (This is per SQL standard.)
221 * Determine next substring of pattern without wildcards. p is
222 * the start of the subpattern, p1 is one past the last byte. Also
223 * track if we found an escape character.
227 found_escape =
false;
236 (
errcode(ERRCODE_INVALID_ESCAPE_SEQUENCE),
237 errmsg(
"LIKE pattern must not end with escape character")));
239 else if (*p1 ==
'_' || *p1 ==
'%')
245 * If we found an escape character, then make an unescaped copy of
253 for (
const char *
c = p;
c < p1;
c++)
271 * Shortcut: If this is the end of the pattern, then the rest of
272 * the text has to match the rest of the pattern.
289 * Now build a substring of the text and try to match it against
290 * the subpattern. t is the start of the text, t1 is one past the
291 * last byte. We start with a zero-length string.
304 * If we found a match, we have to test if the rest of pattern
305 * can match against the rest of the string. Otherwise we
306 * have to continue here try matching with a longer substring.
307 * (This is similar to the recursion for the '%' wildcard
310 * Note that we can't just wind forward p and t and continue
311 * with the main loop. This would fail for example with
313 * U&'0061円0308円bc' LIKE U&'00円E4_c' COLLATE ignore_accents
315 * You'd find that t=0061円 matches p=00円E4, but then the rest
316 * won't match; but t=0061円0308円 also matches p=00円E4, and
317 * then the rest will match.
332 * Didn't match. If we used up the whole text, then the match
333 * fails. Otherwise, try again with a longer substring.
347 /* non-wildcard pattern char fails to match text char */
352 * Pattern and text match, so advance.
354 * It is safe to use NextByte instead of NextChar here, even for
355 * multi-byte character sets, because we are not following immediately
356 * after a wildcard character. If we are in the middle of a multibyte
357 * character, we must already have matched at least one byte of the
358 * character from both text and pattern; so we cannot get out-of-sync
359 * on character boundaries. And we know that no backend-legal
360 * encoding allows ASCII characters such as '%' to appear as non-first
361 * bytes of characters, so we won't mistakenly detect a new wildcard.
368 return LIKE_FALSE;
/* end of pattern, but not of text */
371 * End of text, but perhaps not of pattern. Match iff the remaining
372 * pattern can match a zero-length string, ie, it's zero or more %'s.
374 while (plen > 0 && *p ==
'%')
380 * End of text with no match, so no point in trying later places to start
381 * matching this pattern.
387 * like_escape() --- given a pattern and an ESCAPE string,
388 * convert the pattern to use Postgres' standard backslash escape convention.
409 * Worst-case pattern growth is 2x --- unlikely, but it's hardly worth
410 * trying to calculate the size more accurately than that.
418 * No escape character is wanted. Double any backslashes in the
419 * pattern to make them act like ordinary characters.
431 * The specified escape must be only a single character.
436 (
errcode(ERRCODE_INVALID_ESCAPE_SEQUENCE),
437 errmsg(
"invalid escape string"),
438 errhint(
"Escape string must be empty or one character.")));
443 * If specified escape is '\', just copy the pattern as-is.
452 * Otherwise, convert occurrences of the specified escape character to
453 * '\', and double occurrences of '\' --- unless they immediately
454 * follow an escape character!
459 if (
CHAREQ(p,
e) && !afterescape)
485#endif /* do_like_escape */
int errhint(const char *fmt,...)
int errcode(int sqlerrcode)
int errmsg(const char *fmt,...)
#define ereport(elevel,...)
#define CopyAdvChar(dst, src, srclen)
#define NextByte(p, plen)
#define NextChar(p, plen)
static int MatchText(const char *t, int tlen, const char *p, int plen, pg_locale_t locale)
#define GETCHAR(t, locale)
void pfree(void *pointer)
#define CHECK_FOR_INTERRUPTS()
int pg_strncoll(const char *arg1, ssize_t len1, const char *arg2, ssize_t len2, pg_locale_t locale)
static int cmp(const chr *x, const chr *y, size_t len)
void check_stack_depth(void)
static Size VARSIZE_ANY(const void *PTR)
static Size VARSIZE_ANY_EXHDR(const void *PTR)
static char * VARDATA(const void *PTR)
static char * VARDATA_ANY(const void *PTR)
static void SET_VARSIZE(void *PTR, Size len)