ViewVC logotype

Contents of /code/trunk/doc/pcre.3

Parent Directory Parent Directory | Revision Log Revision Log

Revision 41 - (show annotations)
Sat Feb 24 21:39:17 2007 UTC (14 years, 8 months ago) by nigel
File size: 69815 byte(s)
Error occurred while calculating annotation data.
Load pcre-2.08a into code/trunk.
1 .TH PCRE 3
3 pcre - Perl-compatible regular expressions.
5 .B #include <pcre.h>
6 .PP
7 .SM
8 .br
9 .B pcre *pcre_compile(const char *\fIpattern\fR, int \fIoptions\fR,
10 .ti +5n
11 .B const char **\fIerrptr\fR, int *\fIerroffset\fR,
12 .ti +5n
13 .B const unsigned char *\fItableptr\fR);
14 .PP
15 .br
16 .B pcre_extra *pcre_study(const pcre *\fIcode\fR, int \fIoptions\fR,
17 .ti +5n
18 .B const char **\fIerrptr\fR);
19 .PP
20 .br
21 .B int pcre_exec(const pcre *\fIcode\fR, "const pcre_extra *\fIextra\fR,"
22 .ti +5n
23 .B "const char *\fIsubject\fR," int \fIlength\fR, int \fIstartoffset\fR,
24 .ti +5n
25 .B int \fIoptions\fR, int *\fIovector\fR, int \fIovecsize\fR);
26 .PP
27 .br
28 .B int pcre_copy_substring(const char *\fIsubject\fR, int *\fIovector\fR,
29 .ti +5n
30 .B int \fIstringcount\fR, int \fIstringnumber\fR, char *\fIbuffer\fR,
31 .ti +5n
32 .B int \fIbuffersize\fR);
33 .PP
34 .br
35 .B int pcre_get_substring(const char *\fIsubject\fR, int *\fIovector\fR,
36 .ti +5n
37 .B int \fIstringcount\fR, int \fIstringnumber\fR,
38 .ti +5n
39 .B const char **\fIstringptr\fR);
40 .PP
41 .br
42 .B int pcre_get_substring_list(const char *\fIsubject\fR,
43 .ti +5n
44 .B int *\fIovector\fR, int \fIstringcount\fR, "const char ***\fIlistptr\fR);"
45 .PP
46 .br
47 .B const unsigned char *pcre_maketables(void);
48 .PP
49 .br
50 .B int pcre_info(const pcre *\fIcode\fR, int *\fIoptptr\fR, int
51 .B *\fIfirstcharptr\fR);
52 .PP
53 .br
54 .B char *pcre_version(void);
55 .PP
56 .br
57 .B void *(*pcre_malloc)(size_t);
58 .PP
59 .br
60 .B void (*pcre_free)(void *);
65 The PCRE library is a set of functions that implement regular expression
66 pattern matching using the same syntax and semantics as Perl 5, with just a few
67 differences (see below). The current implementation corresponds to Perl 5.005.
69 PCRE has its own native API, which is described in this document. There is also
70 a set of wrapper functions that correspond to the POSIX API. These are
71 described in the \fBpcreposix\fR documentation.
73 The native API function prototypes are defined in the header file \fBpcre.h\fR,
74 and on Unix systems the library itself is called \fBlibpcre.a\fR, so can be
75 accessed by adding \fB-lpcre\fR to the command for linking an application which
76 calls it.
78 The functions \fBpcre_compile()\fR, \fBpcre_study()\fR, and \fBpcre_exec()\fR
79 are used for compiling and matching regular expressions, while
80 \fBpcre_copy_substring()\fR, \fBpcre_get_substring()\fR, and
81 \fBpcre_get_substring_list()\fR are convenience functions for extracting
82 captured substrings from a matched subject string. The function
83 \fBpcre_maketables()\fR is used (optionally) to build a set of character tables
84 in the current locale for passing to \fBpcre_compile()\fR.
86 The function \fBpcre_info()\fR is used to find out information about a compiled
87 pattern, while the function \fBpcre_version()\fR returns a pointer to a string
88 containing the version of PCRE and its date of release.
90 The global variables \fBpcre_malloc\fR and \fBpcre_free\fR initially contain
91 the entry points of the standard \fBmalloc()\fR and \fBfree()\fR functions
92 respectively. PCRE calls the memory management functions via these variables,
93 so a calling program can replace them if it wishes to intercept the calls. This
94 should be done before calling any PCRE functions.
98 The PCRE functions can be used in multi-threading applications, with the
99 proviso that the memory management functions pointed to by \fBpcre_malloc\fR
100 and \fBpcre_free\fR are shared by all threads.
102 The compiled form of a regular expression is not altered during matching, so
103 the same compiled pattern can safely be used by several threads at once.
107 The function \fBpcre_compile()\fR is called to compile a pattern into an
108 internal form. The pattern is a C string terminated by a binary zero, and
109 is passed in the argument \fIpattern\fR. A pointer to a single block of memory
110 that is obtained via \fBpcre_malloc\fR is returned. This contains the
111 compiled code and related data. The \fBpcre\fR type is defined for this for
112 convenience, but in fact \fBpcre\fR is just a typedef for \fBvoid\fR, since the
113 contents of the block are not externally defined. It is up to the caller to
114 free the memory when it is no longer required.
115 .PP
116 The size of a compiled pattern is roughly proportional to the length of the
117 pattern string, except that each character class (other than those containing
118 just a single character, negated or not) requires 33 bytes, and repeat
119 quantifiers with a minimum greater than one or a bounded maximum cause the
120 relevant portions of the compiled pattern to be replicated.
121 .PP
122 The \fIoptions\fR argument contains independent bits that affect the
123 compilation. It should be zero if no options are required. Some of the options,
124 in particular, those that are compatible with Perl, can also be set and unset
125 from within the pattern (see the detailed description of regular expressions
126 below). For these options, the contents of the \fIoptions\fR argument specifies
127 their initial settings at the start of compilation and execution. The
128 PCRE_ANCHORED option can be set at the time of matching as well as at compile
129 time.
130 .PP
131 If \fIerrptr\fR is NULL, \fBpcre_compile()\fR returns NULL immediately.
132 Otherwise, if compilation of a pattern fails, \fBpcre_compile()\fR returns
133 NULL, and sets the variable pointed to by \fIerrptr\fR to point to a textual
134 error message. The offset from the start of the pattern to the character where
135 the error was discovered is placed in the variable pointed to by
136 \fIerroffset\fR, which must not be NULL. If it is, an immediate error is given.
137 .PP
138 If the final argument, \fItableptr\fR, is NULL, PCRE uses a default set of
139 character tables which are built when it is compiled, using the default C
140 locale. Otherwise, \fItableptr\fR must be the result of a call to
141 \fBpcre_maketables()\fR. See the section on locale support below.
142 .PP
143 The following option bits are defined in the header file:
147 If this bit is set, the pattern is forced to be "anchored", that is, it is
148 constrained to match only at the start of the string which is being searched
149 (the "subject string"). This effect can also be achieved by appropriate
150 constructs in the pattern itself, which is the only way to do it in Perl.
154 If this bit is set, letters in the pattern match both upper and lower case
155 letters. It is equivalent to Perl's /i option.
159 If this bit is set, a dollar metacharacter in the pattern matches only at the
160 end of the subject string. Without this option, a dollar also matches
161 immediately before the final character if it is a newline (but not before any
162 other newlines). The PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is
163 set. There is no equivalent to this option in Perl.
167 If this bit is set, a dot metacharater in the pattern matches all characters,
168 including newlines. Without it, newlines are excluded. This option is
169 equivalent to Perl's /s option. A negative class such as [^a] always matches a
170 newline character, independent of the setting of this option.
174 If this bit is set, whitespace data characters in the pattern are totally
175 ignored except when escaped or inside a character class, and characters between
176 an unescaped # outside a character class and the next newline character,
177 inclusive, are also ignored. This is equivalent to Perl's /x option, and makes
178 it possible to include comments inside complicated patterns. Note, however,
179 that this applies only to data characters. Whitespace characters may never
180 appear within special character sequences in a pattern, for example within the
181 sequence (?( which introduces a conditional subpattern.
185 This option turns on additional functionality of PCRE that is incompatible with
186 Perl. Any backslash in a pattern that is followed by a letter that has no
187 special meaning causes an error, thus reserving these combinations for future
188 expansion. By default, as in Perl, a backslash followed by a letter with no
189 special meaning is treated as a literal. There are at present no other features
190 controlled by this option.
194 By default, PCRE treats the subject string as consisting of a single "line" of
195 characters (even if it actually contains several newlines). The "start of line"
196 metacharacter (^) matches only at the start of the string, while the "end of
197 line" metacharacter ($) matches only at the end of the string, or before a
198 terminating newline (unless PCRE_DOLLAR_ENDONLY is set). This is the same as
199 Perl.
201 When PCRE_MULTILINE it is set, the "start of line" and "end of line" constructs
202 match immediately following or immediately before any newline in the subject
203 string, respectively, as well as at the very start and end. This is equivalent
204 to Perl's /m option. If there are no "\\n" characters in a subject string, or
205 no occurrences of ^ or $ in a pattern, setting PCRE_MULTILINE has no
206 effect.
210 This option inverts the "greediness" of the quantifiers so that they are not
211 greedy by default, but become greedy if followed by "?". It is not compatible
212 with Perl. It can also be set by a (?U) option setting within the pattern.
216 When a pattern is going to be used several times, it is worth spending more
217 time analyzing it in order to speed up the time taken for matching. The
218 function \fBpcre_study()\fR takes a pointer to a compiled pattern as its first
219 argument, and returns a pointer to a \fBpcre_extra\fR block (another \fBvoid\fR
220 typedef) containing additional information about the pattern; this can be
221 passed to \fBpcre_exec()\fR. If no additional information is available, NULL
222 is returned.
224 The second argument contains option bits. At present, no options are defined
225 for \fBpcre_study()\fR, and this argument should always be zero.
227 The third argument for \fBpcre_study()\fR is a pointer to an error message. If
228 studying succeeds (even if no data is returned), the variable it points to is
229 set to NULL. Otherwise it points to a textual error message.
231 At present, studying a pattern is useful only for non-anchored patterns that do
232 not have a single fixed starting character. A bitmap of possible starting
233 characters is created.
237 PCRE handles caseless matching, and determines whether characters are letters,
238 digits, or whatever, by reference to a set of tables. The library contains a
239 default set of tables which is created in the default C locale when PCRE is
240 compiled. This is used when the final argument of \fBpcre_compile()\fR is NULL,
241 and is sufficient for many applications.
243 An alternative set of tables can, however, be supplied. Such tables are built
244 by calling the \fBpcre_maketables()\fR function, which has no arguments, in the
245 relevant locale. The result can then be passed to \fBpcre_compile()\fR as often
246 as necessary. For example, to build and use tables that are appropriate for the
247 French locale (where accented characters with codes greater than 128 are
248 treated as letters), the following code could be used:
250 setlocale(LC_CTYPE, "fr");
251 tables = pcre_maketables();
252 re = pcre_compile(..., tables);
254 The tables are built in memory that is obtained via \fBpcre_malloc\fR. The
255 pointer that is passed to \fBpcre_compile\fR is saved with the compiled
256 pattern, and the same tables are used via this pointer by \fBpcre_study()\fR
257 and \fBpcre_exec()\fR. Thus for any single pattern, compilation, studying and
258 matching all happen in the same locale, but different patterns can be compiled
259 in different locales. It is the caller's responsibility to ensure that the
260 memory containing the tables remains available for as long as it is needed.
264 The \fBpcre_info()\fR function returns information about a compiled pattern.
265 Its yield is the number of capturing subpatterns, or one of the following
266 negative numbers:
268 PCRE_ERROR_NULL the argument \fIcode\fR was NULL
269 PCRE_ERROR_BADMAGIC the "magic number" was not found
271 If the \fIoptptr\fR argument is not NULL, a copy of the options with which the
272 pattern was compiled is placed in the integer it points to. These option bits
273 are those specified in the call to \fBpcre_compile()\fR, modified by any
274 top-level option settings within the pattern itself, and with the PCRE_ANCHORED
275 bit set if the form of the pattern implies that it can match only at the start
276 of a subject string.
278 If the pattern is not anchored and the \fIfirstcharptr\fR argument is not NULL,
279 it is used to pass back information about the first character of any matched
280 string. If there is a fixed first character, e.g. from a pattern such as
281 (cat|cow|coyote), then it is returned in the integer pointed to by
282 \fIfirstcharptr\fR. Otherwise, if either
284 (a) the pattern was compiled with the PCRE_MULTILINE option, and every branch
285 starts with "^", or
287 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not set
288 (if it were set, the pattern would be anchored),
290 then -1 is returned, indicating that the pattern matches only at the
291 start of a subject string or after any "\\n" within the string. Otherwise -2 is
292 returned.
296 The function \fBpcre_exec()\fR is called to match a subject string against a
297 pre-compiled pattern, which is passed in the \fIcode\fR argument. If the
298 pattern has been studied, the result of the study should be passed in the
299 \fIextra\fR argument. Otherwise this must be NULL.
301 The PCRE_ANCHORED option can be passed in the \fIoptions\fR argument, whose
302 unused bits must be zero. However, if a pattern was compiled with
303 PCRE_ANCHORED, or turned out to be anchored by virtue of its contents, it
304 cannot be made unachored at matching time.
306 There are also three further options that can be set only at matching time:
310 The first character of the string is not the beginning of a line, so the
311 circumflex metacharacter should not match before it. Setting this without
312 PCRE_MULTILINE (at compile time) causes circumflex never to match.
316 The end of the string is not the end of a line, so the dollar metacharacter
317 should not match it nor (except in multiline mode) a newline immediately before
318 it. Setting this without PCRE_MULTILINE (at compile time) causes dollar never
319 to match.
323 An empty string is not considered to be a valid match if this option is set. If
324 there are alternatives in the pattern, they are tried. If all the alternatives
325 match the empty string, the entire match fails. For example, if the pattern
327 a?b?
329 is applied to a string not beginning with "a" or "b", it matches the empty
330 string at the start of the subject. With PCRE_NOTEMPTY set, this match is not
331 valid, so PCRE searches further into the string for occurrences of "a" or "b".
333 Perl has no direct equivalent of PCRE_NOTEMPTY, but it does make a special case
334 of a pattern match of the empty string within its \fBsplit()\fR function, and
335 when using the /g modifier. It is possible to emulate Perl's behaviour after
336 matching a null string by first trying the match again at the same offset with
337 PCRE_NOTEMPTY set, and then if that fails by advancing the starting offset (see
338 below) and trying an ordinary match again.
340 The subject string is passed as a pointer in \fIsubject\fR, a length in
341 \fIlength\fR, and a starting offset in \fIstartoffset\fR. Unlike the pattern
342 string, it may contain binary zero characters. When the starting offset is
343 zero, the search for a match starts at the beginning of the subject, and this
344 is by far the most common case.
346 A non-zero starting offset is useful when searching for another match in the
347 same subject by calling \fBpcre_exec()\fR again after a previous success.
348 Setting \fIstartoffset\fR differs from just passing over a shortened string and
349 setting PCRE_NOTBOL in the case of a pattern that begins with any kind of
350 lookbehind. For example, consider the pattern
352 \\Biss\\B
354 which finds occurrences of "iss" in the middle of words. (\\B matches only if
355 the current position in the subject is not a word boundary.) When applied to
356 the string "Mississipi" the first call to \fBpcre_exec()\fR finds the first
357 occurrence. If \fBpcre_exec()\fR is called again with just the remainder of the
358 subject, namely "issipi", it does not match, because \\B is always false at the
359 start of the subject, which is deemed to be a word boundary. However, if
360 \fBpcre_exec()\fR is passed the entire string again, but with \fIstartoffset\fR
361 set to 4, it finds the second occurrence of "iss" because it is able to look
362 behind the starting point to discover that it is preceded by a letter.
364 If a non-zero starting offset is passed when the pattern is anchored, one
365 attempt to match at the given offset is tried. This can only succeed if the
366 pattern does not require the match to be at the start of the subject.
368 In general, a pattern matches a certain portion of the subject, and in
369 addition, further substrings from the subject may be picked out by parts of the
370 pattern. Following the usage in Jeffrey Friedl's book, this is called
371 "capturing" in what follows, and the phrase "capturing subpattern" is used for
372 a fragment of a pattern that picks out a substring. PCRE supports several other
373 kinds of parenthesized subpattern that do not cause substrings to be captured.
375 Captured substrings are returned to the caller via a vector of integer offsets
376 whose address is passed in \fIovector\fR. The number of elements in the vector
377 is passed in \fIovecsize\fR. The first two-thirds of the vector is used to pass
378 back captured substrings, each substring using a pair of integers. The
379 remaining third of the vector is used as workspace by \fBpcre_exec()\fR while
380 matching capturing subpatterns, and is not available for passing back
381 information. The length passed in \fIovecsize\fR should always be a multiple of
382 three. If it is not, it is rounded down.
384 When a match has been successful, information about captured substrings is
385 returned in pairs of integers, starting at the beginning of \fIovector\fR, and
386 continuing up to two-thirds of its length at the most. The first element of a
387 pair is set to the offset of the first character in a substring, and the second
388 is set to the offset of the first character after the end of a substring. The
389 first pair, \fIovector[0]\fR and \fIovector[1]\fR, identify the portion of the
390 subject string matched by the entire pattern. The next pair is used for the
391 first capturing subpattern, and so on. The value returned by \fBpcre_exec()\fR
392 is the number of pairs that have been set. If there are no capturing
393 subpatterns, the return value from a successful match is 1, indicating that
394 just the first pair of offsets has been set.
396 Some convenience functions are provided for extracting the captured substrings
397 as separate strings. These are described in the following section.
399 It is possible for an capturing subpattern number \fIn+1\fR to match some
400 part of the subject when subpattern \fIn\fR has not been used at all. For
401 example, if the string "abc" is matched against the pattern (a|(z))(bc)
402 subpatterns 1 and 3 are matched, but 2 is not. When this happens, both offset
403 values corresponding to the unused subpattern are set to -1.
405 If a capturing subpattern is matched repeatedly, it is the last portion of the
406 string that it matched that gets returned.
408 If the vector is too small to hold all the captured substrings, it is used as
409 far as possible (up to two-thirds of its length), and the function returns a
410 value of zero. In particular, if the substring offsets are not of interest,
411 \fBpcre_exec()\fR may be called with \fIovector\fR passed as NULL and
412 \fIovecsize\fR as zero. However, if the pattern contains back references and
413 the \fIovector\fR isn't big enough to remember the related substrings, PCRE has
414 to get additional memory for use during matching. Thus it is usually advisable
415 to supply an \fIovector\fR.
417 Note that \fBpcre_info()\fR can be used to find out how many capturing
418 subpatterns there are in a compiled pattern. The smallest size for
419 \fIovector\fR that will allow for \fIn\fR captured substrings in addition to
420 the offsets of the substring matched by the whole pattern is (\fIn\fR+1)*3.
422 If \fBpcre_exec()\fR fails, it returns a negative number. The following are
423 defined in the header file:
427 The subject string did not match the pattern.
431 Either \fIcode\fR or \fIsubject\fR was passed as NULL, or \fIovector\fR was
432 NULL and \fIovecsize\fR was not zero.
436 An unrecognized bit was set in the \fIoptions\fR argument.
440 PCRE stores a 4-byte "magic number" at the start of the compiled code, to catch
441 the case when it is passed a junk pointer. This is the error it gives when the
442 magic number isn't present.
446 While running the pattern match, an unknown item was encountered in the
447 compiled pattern. This error could be caused by a bug in PCRE or by overwriting
448 of the compiled pattern.
452 If a pattern contains back references, but the \fIovector\fR that is passed to
453 \fBpcre_exec()\fR is not big enough to remember the referenced substrings, PCRE
454 gets a block of memory at the start of matching to use for this purpose. If the
455 call via \fBpcre_malloc()\fR fails, this error is given. The memory is freed at
456 the end of matching.
460 Captured substrings can be accessed directly by using the offsets returned by
461 \fBpcre_exec()\fR in \fIovector\fR. For convenience, the functions
462 \fBpcre_copy_substring()\fR, \fBpcre_get_substring()\fR, and
463 \fBpcre_get_substring_list()\fR are provided for extracting captured substrings
464 as new, separate, zero-terminated strings. A substring that contains a binary
465 zero is correctly extracted and has a further zero added on the end, but the
466 result does not, of course, function as a C string.
468 The first three arguments are the same for all three functions: \fIsubject\fR
469 is the subject string which has just been successfully matched, \fIovector\fR
470 is a pointer to the vector of integer offsets that was passed to
471 \fBpcre_exec()\fR, and \fIstringcount\fR is the number of substrings that
472 were captured by the match, including the substring that matched the entire
473 regular expression. This is the value returned by \fBpcre_exec\fR if it
474 is greater than zero. If \fBpcre_exec()\fR returned zero, indicating that it
475 ran out of space in \fIovector\fR, then the value passed as
476 \fIstringcount\fR should be the size of the vector divided by three.
478 The functions \fBpcre_copy_substring()\fR and \fBpcre_get_substring()\fR
479 extract a single substring, whose number is given as \fIstringnumber\fR. A
480 value of zero extracts the substring that matched the entire pattern, while
481 higher values extract the captured substrings. For \fBpcre_copy_substring()\fR,
482 the string is placed in \fIbuffer\fR, whose length is given by
483 \fIbuffersize\fR, while for \fBpcre_get_substring()\fR a new block of store is
484 obtained via \fBpcre_malloc\fR, and its address is returned via
485 \fIstringptr\fR. The yield of the function is the length of the string, not
486 including the terminating zero, or one of
490 The buffer was too small for \fBpcre_copy_substring()\fR, or the attempt to get
491 memory failed for \fBpcre_get_substring()\fR.
495 There is no substring whose number is \fIstringnumber\fR.
497 The \fBpcre_get_substring_list()\fR function extracts all available substrings
498 and builds a list of pointers to them. All this is done in a single block of
499 memory which is obtained via \fBpcre_malloc\fR. The address of the memory block
500 is returned via \fIlistptr\fR, which is also the start of the list of string
501 pointers. The end of the list is marked by a NULL pointer. The yield of the
502 function is zero if all went well, or
506 if the attempt to get the memory block failed.
508 When any of these functions encounter a substring that is unset, which can
509 happen when capturing subpattern number \fIn+1\fR matches some part of the
510 subject, but subpattern \fIn\fR has not been used at all, they return an empty
511 string. This can be distinguished from a genuine zero-length substring by
512 inspecting the appropriate offset in \fIovector\fR, which is negative for unset
513 substrings.
518 There are some size limitations in PCRE but it is hoped that they will never in
519 practice be relevant.
520 The maximum length of a compiled pattern is 65539 (sic) bytes.
521 All values in repeating quantifiers must be less than 65536.
522 The maximum number of capturing subpatterns is 99.
523 The maximum number of all parenthesized subpatterns, including capturing
524 subpatterns, assertions, and other types of subpattern, is 200.
526 The maximum length of a subject string is the largest positive number that an
527 integer variable can hold. However, PCRE uses recursion to handle subpatterns
528 and indefinite repetition. This means that the available stack space may limit
529 the size of a subject string that can be processed by certain patterns.
533 The differences described here are with respect to Perl 5.005.
535 1. By default, a whitespace character is any character that the C library
536 function \fBisspace()\fR recognizes, though it is possible to compile PCRE with
537 alternative character type tables. Normally \fBisspace()\fR matches space,
538 formfeed, newline, carriage return, horizontal tab, and vertical tab. Perl 5
539 no longer includes vertical tab in its set of whitespace characters. The \\v
540 escape that was in the Perl documentation for a long time was never in fact
541 recognized. However, the character itself was treated as whitespace at least
542 up to 5.002. In 5.004 and 5.005 it does not match \\s.
544 2. PCRE does not allow repeat quantifiers on lookahead assertions. Perl permits
545 them, but they do not mean what you might think. For example, (?!a){3} does
546 not assert that the next three characters are not "a". It just asserts that the
547 next character is not "a" three times.
549 3. Capturing subpatterns that occur inside negative lookahead assertions are
550 counted, but their entries in the offsets vector are never set. Perl sets its
551 numerical variables from any such patterns that are matched before the
552 assertion fails to match something (thereby succeeding), but only if the
553 negative lookahead assertion contains just one branch.
555 4. Though binary zero characters are supported in the subject string, they are
556 not allowed in a pattern string because it is passed as a normal C string,
557 terminated by zero. The escape sequence "\\0" can be used in the pattern to
558 represent a binary zero.
560 5. The following Perl escape sequences are not supported: \\l, \\u, \\L, \\U,
561 \\E, \\Q. In fact these are implemented by Perl's general string-handling and
562 are not part of its pattern matching engine.
564 6. The Perl \\G assertion is not supported as it is not relevant to single
565 pattern matches.
567 7. Fairly obviously, PCRE does not support the (?{code}) construction.
569 8. There are at the time of writing some oddities in Perl 5.005_02 concerned
570 with the settings of captured strings when part of a pattern is repeated. For
571 example, matching "aba" against the pattern /^(a(b)?)+$/ sets $2 to the value
572 "b", but matching "aabbaa" against /^(aa(bb)?)+$/ leaves $2 unset. However, if
573 the pattern is changed to /^(aa(b(b))?)+$/ then $2 (and $3) get set.
575 In Perl 5.004 $2 is set in both cases, and that is also true of PCRE. If in the
576 future Perl changes to a consistent state that is different, PCRE may change to
577 follow.
579 9. Another as yet unresolved discrepancy is that in Perl 5.005_02 the pattern
580 /^(a)?(?(1)a|b)+$/ matches the string "a", whereas in PCRE it does not.
581 However, in both Perl and PCRE /^(a)?a/ matched against "a" leaves $1 unset.
583 10. PCRE provides some extensions to the Perl regular expression facilities:
585 (a) Although lookbehind assertions must match fixed length strings, each
586 alternative branch of a lookbehind assertion can match a different length of
587 string. Perl 5.005 requires them all to have the same length.
589 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $ meta-
590 character matches only at the very end of the string.
592 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no special
593 meaning is faulted.
595 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quantifiers is
596 inverted, that is, by default they are not greedy, but if followed by a
597 question mark they are.
599 (e) PCRE_ANCHORED can be used to force a pattern to be tried only at the start
600 of the subject.
602 (f) The PCRE_NOTBOL, PCRE_NOTEOL, and PCRE_NOTEMPTY options for
603 \fBpcre_exec()\fR have no Perl equivalents.
607 The syntax and semantics of the regular expressions supported by PCRE are
608 described below. Regular expressions are also described in the Perl
609 documentation and in a number of other books, some of which have copious
610 examples. Jeffrey Friedl's "Mastering Regular Expressions", published by
611 O'Reilly (ISBN 1-56592-257-3), covers them in great detail. The description
612 here is intended as reference documentation.
614 A regular expression is a pattern that is matched against a subject string from
615 left to right. Most characters stand for themselves in a pattern, and match the
616 corresponding characters in the subject. As a trivial example, the pattern
618 The quick brown fox
620 matches a portion of a subject string that is identical to itself. The power of
621 regular expressions comes from the ability to include alternatives and
622 repetitions in the pattern. These are encoded in the pattern by the use of
623 \fImeta-characters\fR, which do not stand for themselves but instead are
624 interpreted in some special way.
626 There are two different sets of meta-characters: those that are recognized
627 anywhere in the pattern except within square brackets, and those that are
628 recognized in square brackets. Outside square brackets, the meta-characters are
629 as follows:
631 \\ general escape character with several uses
632 ^ assert start of subject (or line, in multiline mode)
633 $ assert end of subject (or line, in multiline mode)
634 . match any character except newline (by default)
635 [ start character class definition
636 | start of alternative branch
637 ( start subpattern
638 ) end subpattern
639 ? extends the meaning of (
640 also 0 or 1 quantifier
641 also quantifier minimizer
642 * 0 or more quantifier
643 + 1 or more quantifier
644 { start min/max quantifier
646 Part of a pattern that is in square brackets is called a "character class". In
647 a character class the only meta-characters are:
649 \\ general escape character
650 ^ negate the class, but only if the first character
651 - indicates character range
652 ] terminates the character class
654 The following sections describe the use of each of the meta-characters.
658 The backslash character has several uses. Firstly, if it is followed by a
659 non-alphameric character, it takes away any special meaning that character may
660 have. This use of backslash as an escape character applies both inside and
661 outside character classes.
663 For example, if you want to match a "*" character, you write "\\*" in the
664 pattern. This applies whether or not the following character would otherwise be
665 interpreted as a meta-character, so it is always safe to precede a
666 non-alphameric with "\\" to specify that it stands for itself. In particular,
667 if you want to match a backslash, you write "\\\\".
669 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
670 pattern (other than in a character class) and characters between a "#" outside
671 a character class and the next newline character are ignored. An escaping
672 backslash can be used to include a whitespace or "#" character as part of the
673 pattern.
675 A second use of backslash provides a way of encoding non-printing characters
676 in patterns in a visible manner. There is no restriction on the appearance of
677 non-printing characters, apart from the binary zero that terminates a pattern,
678 but when a pattern is being prepared by text editing, it is usually easier to
679 use one of the following escape sequences than the binary character it
680 represents:
682 \\a alarm, that is, the BEL character (hex 07)
683 \\cx "control-x", where x is any character
684 \\e escape (hex 1B)
685 \\f formfeed (hex 0C)
686 \\n newline (hex 0A)
687 \\r carriage return (hex 0D)
688 \\t tab (hex 09)
689 \\xhh character with hex code hh
690 \\ddd character with octal code ddd, or backreference
692 The precise effect of "\\cx" is as follows: if "x" is a lower case letter, it
693 is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
694 Thus "\\cz" becomes hex 1A, but "\\c{" becomes hex 3B, while "\\c;" becomes hex
695 7B.
697 After "\\x", up to two hexadecimal digits are read (letters can be in upper or
698 lower case).
700 After "\\0" up to two further octal digits are read. In both cases, if there
701 are fewer than two digits, just those that are present are used. Thus the
702 sequence "\\0\\x\\07" specifies two binary zeros followed by a BEL character.
703 Make sure you supply two digits after the initial zero if the character that
704 follows is itself an octal digit.
706 The handling of a backslash followed by a digit other than 0 is complicated.
707 Outside a character class, PCRE reads it and any following digits as a decimal
708 number. If the number is less than 10, or if there have been at least that many
709 previous capturing left parentheses in the expression, the entire sequence is
710 taken as a \fIback reference\fR. A description of how this works is given
711 later, following the discussion of parenthesized subpatterns.
713 Inside a character class, or if the decimal number is greater than 9 and there
714 have not been that many capturing subpatterns, PCRE re-reads up to three octal
715 digits following the backslash, and generates a single byte from the least
716 significant 8 bits of the value. Any subsequent digits stand for themselves.
717 For example:
719 \\040 is another way of writing a space
720 \\40 is the same, provided there are fewer than 40
721 previous capturing subpatterns
722 \\7 is always a back reference
723 \\11 might be a back reference, or another way of
724 writing a tab
725 \\011 is always a tab
726 \\0113 is a tab followed by the character "3"
727 \\113 is the character with octal code 113 (since there
728 can be no more than 99 back references)
729 \\377 is a byte consisting entirely of 1 bits
730 \\81 is either a back reference, or a binary zero
731 followed by the two characters "8" and "1"
733 Note that octal values of 100 or greater must not be introduced by a leading
734 zero, because no more than three octal digits are ever read.
736 All the sequences that define a single byte value can be used both inside and
737 outside character classes. In addition, inside a character class, the sequence
738 "\\b" is interpreted as the backspace character (hex 08). Outside a character
739 class it has a different meaning (see below).
741 The third use of backslash is for specifying generic character types:
743 \\d any decimal digit
744 \\D any character that is not a decimal digit
745 \\s any whitespace character
746 \\S any character that is not a whitespace character
747 \\w any "word" character
748 \\W any "non-word" character
750 Each pair of escape sequences partitions the complete set of characters into
751 two disjoint sets. Any given character matches one, and only one, of each pair.
753 A "word" character is any letter or digit or the underscore character, that is,
754 any character which can be part of a Perl "word". The definition of letters and
755 digits is controlled by PCRE's character tables, and may vary if locale-
756 specific matching is taking place (see "Locale support" above). For example, in
757 the "fr" (French) locale, some character codes greater than 128 are used for
758 accented letters, and these are matched by \\w.
760 These character type sequences can appear both inside and outside character
761 classes. They each match one character of the appropriate type. If the current
762 matching point is at the end of the subject string, all of them fail, since
763 there is no character to match.
765 The fourth use of backslash is for certain simple assertions. An assertion
766 specifies a condition that has to be met at a particular point in a match,
767 without consuming any characters from the subject string. The use of
768 subpatterns for more complicated assertions is described below. The backslashed
769 assertions are
771 \\b word boundary
772 \\B not a word boundary
773 \\A start of subject (independent of multiline mode)
774 \\Z end of subject or newline at end (independent of multiline mode)
775 \\z end of subject (independent of multiline mode)
777 These assertions may not appear in character classes (but note that "\\b" has a
778 different meaning, namely the backspace character, inside a character class).
780 A word boundary is a position in the subject string where the current character
781 and the previous character do not both match \\w or \\W (i.e. one matches
782 \\w and the other matches \\W), or the start or end of the string if the
783 first or last character matches \\w, respectively.
785 The \\A, \\Z, and \\z assertions differ from the traditional circumflex and
786 dollar (described below) in that they only ever match at the very start and end
787 of the subject string, whatever options are set. They are not affected by the
788 PCRE_NOTBOL or PCRE_NOTEOL options. If the \fIstartoffset\fR argument of
789 \fBpcre_exec()\fR is non-zero, \\A can never match. The difference between \\Z
790 and \\z is that \\Z matches before a newline that is the last character of the
791 string as well as at the end of the string, whereas \\z matches only at the
792 end.
796 Outside a character class, in the default matching mode, the circumflex
797 character is an assertion which is true only if the current matching point is
798 at the start of the subject string. If the \fIstartoffset\fR argument of
799 \fBpcre_exec()\fR is non-zero, circumflex can never match. Inside a character
800 class, circumflex has an entirely different meaning (see below).
802 Circumflex need not be the first character of the pattern if a number of
803 alternatives are involved, but it should be the first thing in each alternative
804 in which it appears if the pattern is ever to match that branch. If all
805 possible alternatives start with a circumflex, that is, if the pattern is
806 constrained to match only at the start of the subject, it is said to be an
807 "anchored" pattern. (There are also other constructs that can cause a pattern
808 to be anchored.)
810 A dollar character is an assertion which is true only if the current matching
811 point is at the end of the subject string, or immediately before a newline
812 character that is the last character in the string (by default). Dollar need
813 not be the last character of the pattern if a number of alternatives are
814 involved, but it should be the last item in any branch in which it appears.
815 Dollar has no special meaning in a character class.
817 The meaning of dollar can be changed so that it matches only at the very end of
818 the string, by setting the PCRE_DOLLAR_ENDONLY option at compile or matching
819 time. This does not affect the \\Z assertion.
821 The meanings of the circumflex and dollar characters are changed if the
822 PCRE_MULTILINE option is set. When this is the case, they match immediately
823 after and immediately before an internal "\\n" character, respectively, in
824 addition to matching at the start and end of the subject string. For example,
825 the pattern /^abc$/ matches the subject string "def\\nabc" in multiline mode,
826 but not otherwise. Consequently, patterns that are anchored in single line mode
827 because all branches start with "^" are not anchored in multiline mode, and a
828 match for circumflex is possible when the \fIstartoffset\fR argument of
829 \fBpcre_exec()\fR is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
830 PCRE_MULTILINE is set.
832 Note that the sequences \\A, \\Z, and \\z can be used to match the start and
833 end of the subject in both modes, and if all branches of a pattern start with
834 \\A is it always anchored, whether PCRE_MULTILINE is set or not.
838 Outside a character class, a dot in the pattern matches any one character in
839 the subject, including a non-printing character, but not (by default) newline.
840 If the PCRE_DOTALL option is set, then dots match newlines as well. The
841 handling of dot is entirely independent of the handling of circumflex and
842 dollar, the only relationship being that they both involve newline characters.
843 Dot has no special meaning in a character class.
847 An opening square bracket introduces a character class, terminated by a closing
848 square bracket. A closing square bracket on its own is not special. If a
849 closing square bracket is required as a member of the class, it should be the
850 first data character in the class (after an initial circumflex, if present) or
851 escaped with a backslash.
853 A character class matches a single character in the subject; the character must
854 be in the set of characters defined by the class, unless the first character in
855 the class is a circumflex, in which case the subject character must not be in
856 the set defined by the class. If a circumflex is actually required as a member
857 of the class, ensure it is not the first character, or escape it with a
858 backslash.
860 For example, the character class [aeiou] matches any lower case vowel, while
861 [^aeiou] matches any character that is not a lower case vowel. Note that a
862 circumflex is just a convenient notation for specifying the characters which
863 are in the class by enumerating those that are not. It is not an assertion: it
864 still consumes a character from the subject string, and fails if the current
865 pointer is at the end of the string.
867 When caseless matching is set, any letters in a class represent both their
868 upper case and lower case versions, so for example, a caseless [aeiou] matches
869 "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
870 caseful version would.
872 The newline character is never treated in any special way in character classes,
873 whatever the setting of the PCRE_DOTALL or PCRE_MULTILINE options is. A class
874 such as [^a] will always match a newline.
876 The minus (hyphen) character can be used to specify a range of characters in a
877 character class. For example, [d-m] matches any letter between d and m,
878 inclusive. If a minus character is required in a class, it must be escaped with
879 a backslash or appear in a position where it cannot be interpreted as
880 indicating a range, typically as the first or last character in the class.
882 It is not possible to have the literal character "]" as the end character of a
883 range. A pattern such as [W-]46] is interpreted as a class of two characters
884 ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
885 "-46]". However, if the "]" is escaped with a backslash it is interpreted as
886 the end of range, so [W-\\]46] is interpreted as a single class containing a
887 range followed by two separate characters. The octal or hexadecimal
888 representation of "]" can also be used to end a range.
890 Ranges operate in ASCII collating sequence. They can also be used for
891 characters specified numerically, for example [\\000-\\037]. If a range that
892 includes letters is used when caseless matching is set, it matches the letters
893 in either case. For example, [W-c] is equivalent to [][\\^_`wxyzabc], matched
894 caselessly, and if character tables for the "fr" locale are in use,
895 [\\xc8-\\xcb] matches accented E characters in both cases.
897 The character types \\d, \\D, \\s, \\S, \\w, and \\W may also appear in a
898 character class, and add the characters that they match to the class. For
899 example, [\\dABCDEF] matches any hexadecimal digit. A circumflex can
900 conveniently be used with the upper case character types to specify a more
901 restricted set of characters than the matching lower case type. For example,
902 the class [^\\W_] matches any letter or digit, but not underscore.
904 All non-alphameric characters other than \\, -, ^ (at the start) and the
905 terminating ] are non-special in character classes, but it does no harm if they
906 are escaped.
910 Vertical bar characters are used to separate alternative patterns. For example,
911 the pattern
913 gilbert|sullivan
915 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
916 and an empty alternative is permitted (matching the empty string).
917 The matching process tries each alternative in turn, from left to right,
918 and the first one that succeeds is used. If the alternatives are within a
919 subpattern (defined below), "succeeds" means matching the rest of the main
920 pattern as well as the alternative in the subpattern.
925 can be changed from within the pattern by a sequence of Perl option letters
926 enclosed between "(?" and ")". The option letters are
930 s for PCRE_DOTALL
933 For example, (?im) sets caseless, multiline matching. It is also possible to
934 unset these options by preceding the letter with a hyphen, and a combined
935 setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
936 PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
937 permitted. If a letter appears both before and after the hyphen, the option is
938 unset.
940 The scope of these option changes depends on where in the pattern the setting
941 occurs. For settings that are outside any subpattern (defined below), the
942 effect is the same as if the options were set or unset at the start of
943 matching. The following patterns all behave in exactly the same way:
945 (?i)abc
946 a(?i)bc
947 ab(?i)c
948 abc(?i)
950 which in turn is the same as compiling the pattern abc with PCRE_CASELESS set.
951 In other words, such "top level" settings apply to the whole pattern (unless
952 there are other changes inside subpatterns). If there is more than one setting
953 of the same option at top level, the rightmost setting is used.
955 If an option change occurs inside a subpattern, the effect is different. This
956 is a change of behaviour in Perl 5.005. An option change inside a subpattern
957 affects only that part of the subpattern that follows it, so
959 (a(?i)b)c
961 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
962 By this means, options can be made to have different settings in different
963 parts of the pattern. Any changes made in one alternative do carry on
964 into subsequent branches within the same subpattern. For example,
966 (a(?i)b|c)
968 matches "ab", "aB", "c", and "C", even though when matching "C" the first
969 branch is abandoned before the option setting. This is because the effects of
970 option settings happen at compile time. There would be some very weird
971 behaviour otherwise.
973 The PCRE-specific options PCRE_UNGREEDY and PCRE_EXTRA can be changed in the
974 same way as the Perl-compatible options by using the characters U and X
975 respectively. The (?X) flag setting is special in that it must always occur
976 earlier in the pattern than any of the additional features it turns on, even
977 when it is at top level. It is best put at the start.
981 Subpatterns are delimited by parentheses (round brackets), which can be nested.
982 Marking part of a pattern as a subpattern does two things:
984 1. It localizes a set of alternatives. For example, the pattern
986 cat(aract|erpillar|)
988 matches one of the words "cat", "cataract", or "caterpillar". Without the
989 parentheses, it would match "cataract", "erpillar" or the empty string.
991 2. It sets up the subpattern as a capturing subpattern (as defined above).
992 When the whole pattern matches, that portion of the subject string that matched
993 the subpattern is passed back to the caller via the \fIovector\fR argument of
994 \fBpcre_exec()\fR. Opening parentheses are counted from left to right (starting
995 from 1) to obtain the numbers of the capturing subpatterns.
997 For example, if the string "the red king" is matched against the pattern
999 the ((red|white) (king|queen))
1001 the captured substrings are "red king", "red", and "king", and are numbered 1,
1002 2, and 3.
1004 The fact that plain parentheses fulfil two functions is not always helpful.
1005 There are often times when a grouping subpattern is required without a
1006 capturing requirement. If an opening parenthesis is followed by "?:", the
1007 subpattern does not do any capturing, and is not counted when computing the
1008 number of any subsequent capturing subpatterns. For example, if the string "the
1009 white queen" is matched against the pattern
1011 the ((?:red|white) (king|queen))
1013 the captured substrings are "white queen" and "queen", and are numbered 1 and
1014 2. The maximum number of captured substrings is 99, and the maximum number of
1015 all subpatterns, both capturing and non-capturing, is 200.
1017 As a convenient shorthand, if any option settings are required at the start of
1018 a non-capturing subpattern, the option letters may appear between the "?" and
1019 the ":". Thus the two patterns
1021 (?i:saturday|sunday)
1022 (?:(?i)saturday|sunday)
1024 match exactly the same set of strings. Because alternative branches are tried
1025 from left to right, and options are not reset until the end of the subpattern
1026 is reached, an option setting in one branch does affect subsequent branches, so
1027 the above patterns match "SUNDAY" as well as "Saturday".
1031 Repetition is specified by quantifiers, which can follow any of the following
1032 items:
1034 a single character, possibly escaped
1035 the . metacharacter
1036 a character class
1037 a back reference (see next section)
1038 a parenthesized subpattern (unless it is an assertion - see below)
1040 The general repetition quantifier specifies a minimum and maximum number of
1041 permitted matches, by giving the two numbers in curly brackets (braces),
1042 separated by a comma. The numbers must be less than 65536, and the first must
1043 be less than or equal to the second. For example:
1045 z{2,4}
1047 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1048 character. If the second number is omitted, but the comma is present, there is
1049 no upper limit; if the second number and the comma are both omitted, the
1050 quantifier specifies an exact number of required matches. Thus
1052 [aeiou]{3,}
1054 matches at least 3 successive vowels, but may match many more, while
1056 \\d{8}
1058 matches exactly 8 digits. An opening curly bracket that appears in a position
1059 where a quantifier is not allowed, or one that does not match the syntax of a
1060 quantifier, is taken as a literal character. For example, {,6} is not a
1061 quantifier, but a literal string of four characters.
1063 The quantifier {0} is permitted, causing the expression to behave as if the
1064 previous item and the quantifier were not present.
1066 For convenience (and historical compatibility) the three most common
1067 quantifiers have single-character abbreviations:
1069 * is equivalent to {0,}
1070 + is equivalent to {1,}
1071 ? is equivalent to {0,1}
1073 It is possible to construct infinite loops by following a subpattern that can
1074 match no characters with a quantifier that has no upper limit, for example:
1076 (a?)*
1078 Earlier versions of Perl and PCRE used to give an error at compile time for
1079 such patterns. However, because there are cases where this can be useful, such
1080 patterns are now accepted, but if any repetition of the subpattern does in fact
1081 match no characters, the loop is forcibly broken.
1083 By default, the quantifiers are "greedy", that is, they match as much as
1084 possible (up to the maximum number of permitted times), without causing the
1085 rest of the pattern to fail. The classic example of where this gives problems
1086 is in trying to match comments in C programs. These appear between the
1087 sequences /* and */ and within the sequence, individual * and / characters may
1088 appear. An attempt to match C comments by applying the pattern
1090 /\\*.*\\*/
1092 to the string
1094 /* first command */ not comment /* second comment */
1096 fails, because it matches the entire string due to the greediness of the .*
1097 item.
1099 However, if a quantifier is followed by a question mark, then it ceases to be
1100 greedy, and instead matches the minimum number of times possible, so the
1101 pattern
1103 /\\*.*?\\*/
1105 does the right thing with the C comments. The meaning of the various
1106 quantifiers is not otherwise changed, just the preferred number of matches.
1107 Do not confuse this use of question mark with its use as a quantifier in its
1108 own right. Because it has two uses, it can sometimes appear doubled, as in
1110 \\d??\\d
1112 which matches one digit by preference, but can match two if that is the only
1113 way the rest of the pattern matches.
1115 If the PCRE_UNGREEDY option is set (an option which is not available in Perl)
1116 then the quantifiers are not greedy by default, but individual ones can be made
1117 greedy by following them with a question mark. In other words, it inverts the
1118 default behaviour.
1120 When a parenthesized subpattern is quantified with a minimum repeat count that
1121 is greater than 1 or with a limited maximum, more store is required for the
1122 compiled pattern, in proportion to the size of the minimum or maximum.
1124 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1125 to Perl's /s) is set, thus allowing the . to match newlines, then the pattern
1126 is implicitly anchored, because whatever follows will be tried against every
1127 character position in the subject string, so there is no point in retrying the
1128 overall match at any position after the first. PCRE treats such a pattern as
1129 though it were preceded by \\A. In cases where it is known that the subject
1130 string contains no newlines, it is worth setting PCRE_DOTALL when the pattern
1131 begins with .* in order to obtain this optimization, or alternatively using ^
1132 to indicate anchoring explicitly.
1134 When a capturing subpattern is repeated, the value captured is the substring
1135 that matched the final iteration. For example, after
1137 (tweedle[dume]{3}\\s*)+
1139 has matched "tweedledum tweedledee" the value of the captured substring is
1140 "tweedledee". However, if there are nested capturing subpatterns, the
1141 corresponding captured values may have been set in previous iterations. For
1142 example, after
1144 /(a|(b))+/
1146 matches "aba" the value of the second captured substring is "b".
1150 Outside a character class, a backslash followed by a digit greater than 0 (and
1151 possibly further digits) is a back reference to a capturing subpattern earlier
1152 (i.e. to its left) in the pattern, provided there have been that many previous
1153 capturing left parentheses.
1155 However, if the decimal number following the backslash is less than 10, it is
1156 always taken as a back reference, and causes an error only if there are not
1157 that many capturing left parentheses in the entire pattern. In other words, the
1158 parentheses that are referenced need not be to the left of the reference for
1159 numbers less than 10. See the section entitled "Backslash" above for further
1160 details of the handling of digits following a backslash.
1162 A back reference matches whatever actually matched the capturing subpattern in
1163 the current subject string, rather than anything matching the subpattern
1164 itself. So the pattern
1166 (sens|respons)e and \\1ibility
1168 matches "sense and sensibility" and "response and responsibility", but not
1169 "sense and responsibility". If caseful matching is in force at the time of the
1170 back reference, then the case of letters is relevant. For example,
1172 ((?i)rah)\\s+\\1
1174 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1175 capturing subpattern is matched caselessly.
1177 There may be more than one back reference to the same subpattern. If a
1178 subpattern has not actually been used in a particular match, then any back
1179 references to it always fail. For example, the pattern
1181 (a|(bc))\\2
1183 always fails if it starts to match "a" rather than "bc". Because there may be
1184 up to 99 back references, all digits following the backslash are taken
1185 as part of a potential back reference number. If the pattern continues with a
1186 digit character, then some delimiter must be used to terminate the back
1187 reference. If the PCRE_EXTENDED option is set, this can be whitespace.
1188 Otherwise an empty comment can be used.
1190 A back reference that occurs inside the parentheses to which it refers fails
1191 when the subpattern is first used, so, for example, (a\\1) never matches.
1192 However, such references can be useful inside repeated subpatterns. For
1193 example, the pattern
1195 (a|b\\1)+
1197 matches any number of "a"s and also "aba", "ababaa" etc. At each iteration of
1198 the subpattern, the back reference matches the character string corresponding
1199 to the previous iteration. In order for this to work, the pattern must be such
1200 that the first iteration does not need to match the back reference. This can be
1201 done using alternation, as in the example above, or by a quantifier with a
1202 minimum of zero.
1206 An assertion is a test on the characters following or preceding the current
1207 matching point that does not actually consume any characters. The simple
1208 assertions coded as \\b, \\B, \\A, \\Z, \\z, ^ and $ are described above. More
1209 complicated assertions are coded as subpatterns. There are two kinds: those
1210 that look ahead of the current position in the subject string, and those that
1211 look behind it.
1213 An assertion subpattern is matched in the normal way, except that it does not
1214 cause the current matching position to be changed. Lookahead assertions start
1215 with (?= for positive assertions and (?! for negative assertions. For example,
1217 \\w+(?=;)
1219 matches a word followed by a semicolon, but does not include the semicolon in
1220 the match, and
1222 foo(?!bar)
1224 matches any occurrence of "foo" that is not followed by "bar". Note that the
1225 apparently similar pattern
1227 (?!foo)bar
1229 does not find an occurrence of "bar" that is preceded by something other than
1230 "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1231 (?!foo) is always true when the next three characters are "bar". A
1232 lookbehind assertion is needed to achieve this effect.
1234 Lookbehind assertions start with (?<= for positive assertions and (?<! for
1235 negative assertions. For example,
1237 (?<!foo)bar
1239 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1240 a lookbehind assertion are restricted such that all the strings it matches must
1241 have a fixed length. However, if there are several alternatives, they do not
1242 all have to have the same fixed length. Thus
1244 (?<=bullock|donkey)
1246 is permitted, but
1248 (?<!dogs?|cats?)
1250 causes an error at compile time. Branches that match different length strings
1251 are permitted only at the top level of a lookbehind assertion. This is an
1252 extension compared with Perl 5.005, which requires all branches to match the
1253 same length of string. An assertion such as
1255 (?<=ab(c|de))
1257 is not permitted, because its single top-level branch can match two different
1258 lengths, but it is acceptable if rewritten to use two top-level branches:
1260 (?<=abc|abde)
1262 The implementation of lookbehind assertions is, for each alternative, to
1263 temporarily move the current position back by the fixed width and then try to
1264 match. If there are insufficient characters before the current position, the
1265 match is deemed to fail. Lookbehinds in conjunction with once-only subpatterns
1266 can be particularly useful for matching at the ends of strings; an example is
1267 given at the end of the section on once-only subpatterns.
1269 Several assertions (of any sort) may occur in succession. For example,
1271 (?<=\\d{3})(?<!999)foo
1273 matches "foo" preceded by three digits that are not "999". Notice that each of
1274 the assertions is applied independently at the same point in the subject
1275 string. First there is a check that the previous three characters are all
1276 digits, then there is a check that the same three characters are not "999".
1277 This pattern does \fInot\fR match "foo" preceded by six characters, the first
1278 of which are digits and the last three of which are not "999". For example, it
1279 doesn't match "123abcfoo". A pattern to do that is
1281 (?<=\\d{3}...)(?<!999)foo
1283 This time the first assertion looks at the preceding six characters, checking
1284 that the first three are digits, and then the second assertion checks that the
1285 preceding three characters are not "999".
1287 Assertions can be nested in any combination. For example,
1289 (?<=(?<!foo)bar)baz
1291 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
1292 preceded by "foo", while
1294 (?<=\\d{3}(?!999)...)foo
1296 is another pattern which matches "foo" preceded by three digits and any three
1297 characters that are not "999".
1299 Assertion subpatterns are not capturing subpatterns, and may not be repeated,
1300 because it makes no sense to assert the same thing several times. If any kind
1301 of assertion contains capturing subpatterns within it, these are counted for
1302 the purposes of numbering the capturing subpatterns in the whole pattern.
1303 However, substring capturing is carried out only for positive assertions,
1304 because it does not make sense for negative assertions.
1306 Assertions count towards the maximum of 200 parenthesized subpatterns.
1310 With both maximizing and minimizing repetition, failure of what follows
1311 normally causes the repeated item to be re-evaluated to see if a different
1312 number of repeats allows the rest of the pattern to match. Sometimes it is
1313 useful to prevent this, either to change the nature of the match, or to cause
1314 it fail earlier than it otherwise might, when the author of the pattern knows
1315 there is no point in carrying on.
1317 Consider, for example, the pattern \\d+foo when applied to the subject line
1319 123456bar
1321 After matching all 6 digits and then failing to match "foo", the normal
1322 action of the matcher is to try again with only 5 digits matching the \\d+
1323 item, and then with 4, and so on, before ultimately failing. Once-only
1324 subpatterns provide the means for specifying that once a portion of the pattern
1325 has matched, it is not to be re-evaluated in this way, so the matcher would
1326 give up immediately on failing to match "foo" the first time. The notation is
1327 another kind of special parenthesis, starting with (?> as in this example:
1329 (?>\\d+)bar
1331 This kind of parenthesis "locks up" the part of the pattern it contains once
1332 it has matched, and a failure further into the pattern is prevented from
1333 backtracking into it. Backtracking past it to previous items, however, works as
1334 normal.
1336 An alternative description is that a subpattern of this type matches the string
1337 of characters that an identical standalone pattern would match, if anchored at
1338 the current point in the subject string.
1340 Once-only subpatterns are not capturing subpatterns. Simple cases such as the
1341 above example can be thought of as a maximizing repeat that must swallow
1342 everything it can. So, while both \\d+ and \\d+? are prepared to adjust the
1343 number of digits they match in order to make the rest of the pattern match,
1344 (?>\\d+) can only match an entire sequence of digits.
1346 This construction can of course contain arbitrarily complicated subpatterns,
1347 and it can be nested.
1349 Once-only subpatterns can be used in conjunction with lookbehind assertions to
1350 specify efficient matching at the end of the subject string. Consider a simple
1351 pattern such as
1353 abcd$
1355 when applied to a long string which does not match it. Because matching
1356 proceeds from left to right, PCRE will look for each "a" in the subject and
1357 then see if what follows matches the rest of the pattern. If the pattern is
1358 specified as
1360 ^.*abcd$
1362 then the initial .* matches the entire string at first, but when this fails, it
1363 backtracks to match all but the last character, then all but the last two
1364 characters, and so on. Once again the search for "a" covers the entire string,
1365 from right to left, so we are no better off. However, if the pattern is written
1366 as
1368 ^(?>.*)(?<=abcd)
1370 then there can be no backtracking for the .* item; it can match only the entire
1371 string. The subsequent lookbehind assertion does a single test on the last four
1372 characters. If it fails, the match fails immediately. For long strings, this
1373 approach makes a significant difference to the processing time.
1377 It is possible to cause the matching process to obey a subpattern
1378 conditionally or to choose between two alternative subpatterns, depending on
1379 the result of an assertion, or whether a previous capturing subpattern matched
1380 or not. The two possible forms of conditional subpattern are
1382 (?(condition)yes-pattern)
1383 (?(condition)yes-pattern|no-pattern)
1385 If the condition is satisfied, the yes-pattern is used; otherwise the
1386 no-pattern (if present) is used. If there are more than two alternatives in the
1387 subpattern, a compile-time error occurs.
1389 There are two kinds of condition. If the text between the parentheses consists
1390 of a sequence of digits, then the condition is satisfied if the capturing
1391 subpattern of that number has previously matched. Consider the following
1392 pattern, which contains non-significant white space to make it more readable
1393 (assume the PCRE_EXTENDED option) and to divide it into three parts for ease
1394 of discussion:
1396 ( \\( )? [^()]+ (?(1) \\) )
1398 The first part matches an optional opening parenthesis, and if that
1399 character is present, sets it as the first captured substring. The second part
1400 matches one or more characters that are not parentheses. The third part is a
1401 conditional subpattern that tests whether the first set of parentheses matched
1402 or not. If they did, that is, if subject started with an opening parenthesis,
1403 the condition is true, and so the yes-pattern is executed and a closing
1404 parenthesis is required. Otherwise, since no-pattern is not present, the
1405 subpattern matches nothing. In other words, this pattern matches a sequence of
1406 non-parentheses, optionally enclosed in parentheses.
1408 If the condition is not a sequence of digits, it must be an assertion. This may
1409 be a positive or negative lookahead or lookbehind assertion. Consider this
1410 pattern, again containing non-significant white space, and with the two
1411 alternatives on the second line:
1413 (?(?=[^a-z]*[a-z])
1414 \\d{2}-[a-z]{3}-\\d{2} | \\d{2}-\\d{2}-\\d{2} )
1416 The condition is a positive lookahead assertion that matches an optional
1417 sequence of non-letters followed by a letter. In other words, it tests for the
1418 presence of at least one letter in the subject. If a letter is found, the
1419 subject is matched against the first alternative; otherwise it is matched
1420 against the second. This pattern matches strings in one of the two forms
1421 dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
1425 The sequence (?# marks the start of a comment which continues up to the next
1426 closing parenthesis. Nested parentheses are not permitted. The characters
1427 that make up a comment play no part in the pattern matching at all.
1429 If the PCRE_EXTENDED option is set, an unescaped # character outside a
1430 character class introduces a comment that continues up to the next newline
1431 character in the pattern.
1435 Certain items that may appear in patterns are more efficient than others. It is
1436 more efficient to use a character class like [aeiou] than a set of alternatives
1437 such as (a|e|i|o|u). In general, the simplest construction that provides the
1438 required behaviour is usually the most efficient. Jeffrey Friedl's book
1439 contains a lot of discussion about optimizing regular expressions for efficient
1440 performance.
1442 When a pattern begins with .* and the PCRE_DOTALL option is set, the pattern is
1443 implicitly anchored by PCRE, since it can match only at the start of a subject
1444 string. However, if PCRE_DOTALL is not set, PCRE cannot make this optimization,
1445 because the . metacharacter does not then match a newline, and if the subject
1446 string contains newlines, the pattern may match from the character immediately
1447 following one of them instead of from the very start. For example, the pattern
1449 (.*) second
1451 matches the subject "first\\nand second" (where \\n stands for a newline
1452 character) with the first captured substring being "and". In order to do this,
1453 PCRE has to retry the match starting after every newline in the subject.
1455 If you are using such a pattern with subject strings that do not contain
1456 newlines, the best performance is obtained by setting PCRE_DOTALL, or starting
1457 the pattern with ^.* to indicate explicit anchoring. That saves PCRE from
1458 having to scan along the subject looking for a newline to restart at.
1460 Beware of patterns that contain nested indefinite repeats. These can take a
1461 long time to run when applied to a string that does not match. Consider the
1462 pattern fragment
1464 (a+)*
1466 This can match "aaaa" in 33 different ways, and this number increases very
1467 rapidly as the string gets longer. (The * repeat can match 0, 1, 2, 3, or 4
1468 times, and for each of those cases other than 0, the + repeats can match
1469 different numbers of times.) When the remainder of the pattern is such that the
1470 entire match is going to fail, PCRE has in principle to try every possible
1471 variation, and this can take an extremely long time.
1473 An optimization catches some of the more simple cases such as
1475 (a+)*b
1477 where a literal character follows. Before embarking on the standard matching
1478 procedure, PCRE checks that there is a "b" later in the subject string, and if
1479 there is not, it fails the match immediately. However, when there is no
1480 following literal this optimization cannot be used. You can see the difference
1481 by comparing the behaviour of
1483 (a+)*\\d
1485 with the pattern above. The former gives a failure almost instantly when
1486 applied to a whole line of "a" characters, whereas the latter takes an
1487 appreciable time with strings longer than about 20 characters.
1490 Philip Hazel <ph10@cam.ac.uk>
1491 .br
1492 University Computing Service,
1493 .br
1494 New Museums Site,
1495 .br
1496 Cambridge CB2 3QG, England.
1497 .br
1498 Phone: +44 1223 334714
1500 Last updated: 29 July 1999
1501 .br
1502 Copyright (c) 1997-1999 University of Cambridge.

  ViewVC Help
Powered by ViewVC 1.1.5