ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 512 - (show annotations)
Tue Mar 30 11:11:52 2010 UTC (11 years, 3 months ago) by ph10
File MIME type: text/plain
File size: 345721 byte(s)
Fix compile problems when heap is in use
1 -----------------------------------------------------------------------------
2 This file contains a concatenation of the PCRE man pages, converted to plain
3 text format for ease of searching with a text editor, or for use on systems
4 that do not have a man page processor. The small individual files that give
5 synopses of each function in the library have not been included. Neither has
6 the pcredemo program. There are separate text files for the pcregrep and
7 pcretest commands.
8 -----------------------------------------------------------------------------
11 PCRE(3) PCRE(3)
15 PCRE - Perl-compatible regular expressions
20 The PCRE library is a set of functions that implement regular expres-
21 sion pattern matching using the same syntax and semantics as Perl, with
22 just a few differences. Some features that appeared in Python and PCRE
23 before they appeared in Perl are also available using the Python syn-
24 tax, there is some support for one or two .NET and Oniguruma syntax
25 items, and there is an option for requesting some minor changes that
26 give better JavaScript compatibility.
28 The current implementation of PCRE corresponds approximately with Perl
29 5.10, including support for UTF-8 encoded strings and Unicode general
30 category properties. However, UTF-8 and Unicode support has to be
31 explicitly enabled; it is not the default. The Unicode tables corre-
32 spond to Unicode release 5.2.0.
34 In addition to the Perl-compatible matching function, PCRE contains an
35 alternative function that matches the same compiled patterns in a dif-
36 ferent way. In certain circumstances, the alternative function has some
37 advantages. For a discussion of the two matching algorithms, see the
38 pcrematching page.
40 PCRE is written in C and released as a C library. A number of people
41 have written wrappers and interfaces of various kinds. In particular,
42 Google Inc. have provided a comprehensive C++ wrapper. This is now
43 included as part of the PCRE distribution. The pcrecpp page has details
44 of this interface. Other people's contributions can be found in the
45 Contrib directory at the primary FTP site, which is:
47 ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre
49 Details of exactly which Perl regular expression features are and are
50 not supported by PCRE are given in separate documents. See the pcrepat-
51 tern and pcrecompat pages. There is a syntax summary in the pcresyntax
52 page.
54 Some features of PCRE can be included, excluded, or changed when the
55 library is built. The pcre_config() function makes it possible for a
56 client to discover which features are available. The features them-
57 selves are described in the pcrebuild page. Documentation about build-
58 ing PCRE for various operating systems can be found in the README and
59 NON-UNIX-USE files in the source distribution.
61 The library contains a number of undocumented internal functions and
62 data tables that are used by more than one of the exported external
63 functions, but which are not intended for use by external callers.
64 Their names all begin with "_pcre_", which hopefully will not provoke
65 any name clashes. In some environments, it is possible to control which
66 external symbols are exported when a shared library is built, and in
67 these cases the undocumented symbols are not exported.
72 The user documentation for PCRE comprises a number of different sec-
73 tions. In the "man" format, each of these is a separate "man page". In
74 the HTML format, each is a separate page, linked from the index page.
75 In the plain text format, all the sections, except the pcredemo sec-
76 tion, are concatenated, for ease of searching. The sections are as fol-
77 lows:
79 pcre this document
80 pcre-config show PCRE installation configuration information
81 pcreapi details of PCRE's native C API
82 pcrebuild options for building PCRE
83 pcrecallout details of the callout feature
84 pcrecompat discussion of Perl compatibility
85 pcrecpp details of the C++ wrapper
86 pcredemo a demonstration C program that uses PCRE
87 pcregrep description of the pcregrep command
88 pcrematching discussion of the two matching algorithms
89 pcrepartial details of the partial matching facility
90 pcrepattern syntax and semantics of supported
91 regular expressions
92 pcreperform discussion of performance issues
93 pcreposix the POSIX-compatible C API
94 pcreprecompile details of saving and re-using precompiled patterns
95 pcresample discussion of the pcredemo program
96 pcrestack discussion of stack usage
97 pcresyntax quick syntax reference
98 pcretest description of the pcretest testing command
100 In addition, in the "man" and HTML formats, there is a short page for
101 each C library function, listing its arguments and results.
106 There are some size limitations in PCRE but it is hoped that they will
107 never in practice be relevant.
109 The maximum length of a compiled pattern is 65539 (sic) bytes if PCRE
110 is compiled with the default internal linkage size of 2. If you want to
111 process regular expressions that are truly enormous, you can compile
112 PCRE with an internal linkage size of 3 or 4 (see the README file in
113 the source distribution and the pcrebuild documentation for details).
114 In these cases the limit is substantially larger. However, the speed
115 of execution is slower.
117 All values in repeating quantifiers must be less than 65536.
119 There is no limit to the number of parenthesized subpatterns, but there
120 can be no more than 65535 capturing subpatterns.
122 The maximum length of name for a named subpattern is 32 characters, and
123 the maximum number of named subpatterns is 10000.
125 The maximum length of a subject string is the largest positive number
126 that an integer variable can hold. However, when using the traditional
127 matching function, PCRE uses recursion to handle subpatterns and indef-
128 inite repetition. This means that the available stack space may limit
129 the size of a subject string that can be processed by certain patterns.
130 For a discussion of stack issues, see the pcrestack documentation.
135 From release 3.3, PCRE has had some support for character strings
136 encoded in the UTF-8 format. For release 4.0 this was greatly extended
137 to cover most common requirements, and in release 5.0 additional sup-
138 port for Unicode general category properties was added.
140 In order process UTF-8 strings, you must build PCRE to include UTF-8
141 support in the code, and, in addition, you must call pcre_compile()
142 with the PCRE_UTF8 option flag, or the pattern must start with the
143 sequence (*UTF8). When either of these is the case, both the pattern
144 and any subject strings that are matched against it are treated as
145 UTF-8 strings instead of strings of 1-byte characters.
147 If you compile PCRE with UTF-8 support, but do not use it at run time,
148 the library will be a bit bigger, but the additional run time overhead
149 is limited to testing the PCRE_UTF8 flag occasionally, so should not be
150 very big.
152 If PCRE is built with Unicode character property support (which implies
153 UTF-8 support), the escape sequences \p{..}, \P{..}, and \X are sup-
154 ported. The available properties that can be tested are limited to the
155 general category properties such as Lu for an upper case letter or Nd
156 for a decimal number, the Unicode script names such as Arabic or Han,
157 and the derived properties Any and L&. A full list is given in the
158 pcrepattern documentation. Only the short names for properties are sup-
159 ported. For example, \p{L} matches a letter. Its Perl synonym, \p{Let-
160 ter}, is not supported. Furthermore, in Perl, many properties may
161 optionally be prefixed by "Is", for compatibility with Perl 5.6. PCRE
162 does not support this.
164 Validity of UTF-8 strings
166 When you set the PCRE_UTF8 flag, the strings passed as patterns and
167 subjects are (by default) checked for validity on entry to the relevant
168 functions. From release 7.3 of PCRE, the check is according the rules
169 of RFC 3629, which are themselves derived from the Unicode specifica-
170 tion. Earlier releases of PCRE followed the rules of RFC 2279, which
171 allows the full range of 31-bit values (0 to 0x7FFFFFFF). The current
172 check allows only values in the range U+0 to U+10FFFF, excluding U+D800
173 to U+DFFF.
175 The excluded code points are the "Low Surrogate Area" of Unicode, of
176 which the Unicode Standard says this: "The Low Surrogate Area does not
177 contain any character assignments, consequently no character code
178 charts or namelists are provided for this area. Surrogates are reserved
179 for use with UTF-16 and then must be used in pairs." The code points
180 that are encoded by UTF-16 pairs are available as independent code
181 points in the UTF-8 encoding. (In other words, the whole surrogate
182 thing is a fudge for UTF-16 which unfortunately messes up UTF-8.)
184 If an invalid UTF-8 string is passed to PCRE, an error return
185 (PCRE_ERROR_BADUTF8) is given. In some situations, you may already know
186 that your strings are valid, and therefore want to skip these checks in
187 order to improve performance. If you set the PCRE_NO_UTF8_CHECK flag at
188 compile time or at run time, PCRE assumes that the pattern or subject
189 it is given (respectively) contains only valid UTF-8 codes. In this
190 case, it does not diagnose an invalid UTF-8 string.
192 If you pass an invalid UTF-8 string when PCRE_NO_UTF8_CHECK is set,
193 what happens depends on why the string is invalid. If the string con-
194 forms to the "old" definition of UTF-8 (RFC 2279), it is processed as a
195 string of characters in the range 0 to 0x7FFFFFFF. In other words,
196 apart from the initial validity test, PCRE (when in UTF-8 mode) handles
197 strings according to the more liberal rules of RFC 2279. However, if
198 the string does not even conform to RFC 2279, the result is undefined.
199 Your program may crash.
201 If you want to process strings of values in the full range 0 to
202 0x7FFFFFFF, encoded in a UTF-8-like manner as per the old RFC, you can
203 set PCRE_NO_UTF8_CHECK to bypass the more restrictive test. However, in
204 this situation, you will have to apply your own validity check.
206 General comments about UTF-8 mode
208 1. An unbraced hexadecimal escape sequence (such as \xb3) matches a
209 two-byte UTF-8 character if the value is greater than 127.
211 2. Octal numbers up to \777 are recognized, and match two-byte UTF-8
212 characters for values greater than \177.
214 3. Repeat quantifiers apply to complete UTF-8 characters, not to indi-
215 vidual bytes, for example: \x{100}{3}.
217 4. The dot metacharacter matches one UTF-8 character instead of a sin-
218 gle byte.
220 5. The escape sequence \C can be used to match a single byte in UTF-8
221 mode, but its use can lead to some strange effects. This facility is
222 not available in the alternative matching function, pcre_dfa_exec().
224 6. The character escapes \b, \B, \d, \D, \s, \S, \w, and \W correctly
225 test characters of any code value, but the characters that PCRE recog-
226 nizes as digits, spaces, or word characters remain the same set as
227 before, all with values less than 256. This remains true even when PCRE
228 includes Unicode property support, because to do otherwise would slow
229 down PCRE in many common cases. If you really want to test for a wider
230 sense of, say, "digit", you must use Unicode property tests such as
231 \p{Nd}. Note that this also applies to \b, because it is defined in
232 terms of \w and \W.
234 7. Similarly, characters that match the POSIX named character classes
235 are all low-valued characters.
237 8. However, the Perl 5.10 horizontal and vertical whitespace matching
238 escapes (\h, \H, \v, and \V) do match all the appropriate Unicode char-
239 acters.
241 9. Case-insensitive matching applies only to characters whose values
242 are less than 128, unless PCRE is built with Unicode property support.
243 Even when Unicode property support is available, PCRE still uses its
244 own character tables when checking the case of low-valued characters,
245 so as not to degrade performance. The Unicode property information is
246 used only for characters with higher values. Even when Unicode property
247 support is available, PCRE supports case-insensitive matching only when
248 there is a one-to-one mapping between a letter's cases. There are a
249 small number of many-to-one mappings in Unicode; these are not sup-
250 ported by PCRE.
255 Philip Hazel
256 University Computing Service
257 Cambridge CB2 3QH, England.
259 Putting an actual email address here seems to have been a spam magnet,
260 so I've taken it away. If you want to email me, use my two initials,
261 followed by the two digits 10, at the domain cam.ac.uk.
266 Last updated: 01 March 2010
267 Copyright (c) 1997-2010 University of Cambridge.
268 ------------------------------------------------------------------------------
274 NAME
275 PCRE - Perl-compatible regular expressions
280 This document describes the optional features of PCRE that can be
281 selected when the library is compiled. It assumes use of the configure
282 script, where the optional features are selected or deselected by pro-
283 viding options to configure before running the make command. However,
284 the same options can be selected in both Unix-like and non-Unix-like
285 environments using the GUI facility of cmake-gui if you are using CMake
286 instead of configure to build PCRE.
288 There is a lot more information about building PCRE in non-Unix-like
289 environments in the file called NON_UNIX_USE, which is part of the PCRE
290 distribution. You should consult this file as well as the README file
291 if you are building in a non-Unix-like environment.
293 The complete list of options for configure (which includes the standard
294 ones such as the selection of the installation directory) can be
295 obtained by running
297 ./configure --help
299 The following sections include descriptions of options whose names
300 begin with --enable or --disable. These settings specify changes to the
301 defaults for the configure command. Because of the way that configure
302 works, --enable and --disable always come in pairs, so the complemen-
303 tary option always exists as well, but as it specifies the default, it
304 is not described.
309 By default, the configure script will search for a C++ compiler and C++
310 header files. If it finds them, it automatically builds the C++ wrapper
311 library for PCRE. You can disable this by adding
313 --disable-cpp
315 to the configure command.
320 To build PCRE with support for UTF-8 Unicode character strings, add
322 --enable-utf8
324 to the configure command. Of itself, this does not make PCRE treat
325 strings as UTF-8. As well as compiling PCRE with this option, you also
326 have have to set the PCRE_UTF8 option when you call the pcre_compile()
327 or pcre_compile2() functions.
329 If you set --enable-utf8 when compiling in an EBCDIC environment, PCRE
330 expects its input to be either ASCII or UTF-8 (depending on the runtime
331 option). It is not possible to support both EBCDIC and UTF-8 codes in
332 the same version of the library. Consequently, --enable-utf8 and
333 --enable-ebcdic are mutually exclusive.
338 UTF-8 support allows PCRE to process character values greater than 255
339 in the strings that it handles. On its own, however, it does not pro-
340 vide any facilities for accessing the properties of such characters. If
341 you want to be able to use the pattern escapes \P, \p, and \X, which
342 refer to Unicode character properties, you must add
344 --enable-unicode-properties
346 to the configure command. This implies UTF-8 support, even if you have
347 not explicitly requested it.
349 Including Unicode property support adds around 30K of tables to the
350 PCRE library. Only the general category properties such as Lu and Nd
351 are supported. Details are given in the pcrepattern documentation.
356 By default, PCRE interprets the linefeed (LF) character as indicating
357 the end of a line. This is the normal newline character on Unix-like
358 systems. You can compile PCRE to use carriage return (CR) instead, by
359 adding
361 --enable-newline-is-cr
363 to the configure command. There is also a --enable-newline-is-lf
364 option, which explicitly specifies linefeed as the newline character.
366 Alternatively, you can specify that line endings are to be indicated by
367 the two character sequence CRLF. If you want this, add
369 --enable-newline-is-crlf
371 to the configure command. There is a fourth option, specified by
373 --enable-newline-is-anycrlf
375 which causes PCRE to recognize any of the three sequences CR, LF, or
376 CRLF as indicating a line ending. Finally, a fifth option, specified by
378 --enable-newline-is-any
380 causes PCRE to recognize any Unicode newline sequence.
382 Whatever line ending convention is selected when PCRE is built can be
383 overridden when the library functions are called. At build time it is
384 conventional to use the standard for your operating system.
389 By default, the sequence \R in a pattern matches any Unicode newline
390 sequence, whatever has been selected as the line ending sequence. If
391 you specify
393 --enable-bsr-anycrlf
395 the default is changed so that \R matches only CR, LF, or CRLF. What-
396 ever is selected when PCRE is built can be overridden when the library
397 functions are called.
402 The PCRE building process uses libtool to build both shared and static
403 Unix libraries by default. You can suppress one of these by adding one
404 of
406 --disable-shared
407 --disable-static
409 to the configure command, as required.
414 When PCRE is called through the POSIX interface (see the pcreposix doc-
415 umentation), additional working storage is required for holding the
416 pointers to capturing substrings, because PCRE requires three integers
417 per substring, whereas the POSIX interface provides only two. If the
418 number of expected substrings is small, the wrapper function uses space
419 on the stack, because this is faster than using malloc() for each call.
420 The default threshold above which the stack is no longer used is 10; it
421 can be changed by adding a setting such as
423 --with-posix-malloc-threshold=20
425 to the configure command.
430 Within a compiled pattern, offset values are used to point from one
431 part to another (for example, from an opening parenthesis to an alter-
432 nation metacharacter). By default, two-byte values are used for these
433 offsets, leading to a maximum size for a compiled pattern of around
434 64K. This is sufficient to handle all but the most gigantic patterns.
435 Nevertheless, some people do want to process truyl enormous patterns,
436 so it is possible to compile PCRE to use three-byte or four-byte off-
437 sets by adding a setting such as
439 --with-link-size=3
441 to the configure command. The value given must be 2, 3, or 4. Using
442 longer offsets slows down the operation of PCRE because it has to load
443 additional bytes when handling them.
448 When matching with the pcre_exec() function, PCRE implements backtrack-
449 ing by making recursive calls to an internal function called match().
450 In environments where the size of the stack is limited, this can se-
451 verely limit PCRE's operation. (The Unix environment does not usually
452 suffer from this problem, but it may sometimes be necessary to increase
453 the maximum stack size. There is a discussion in the pcrestack docu-
454 mentation.) An alternative approach to recursion that uses memory from
455 the heap to remember data, instead of using recursive function calls,
456 has been implemented to work round the problem of limited stack size.
457 If you want to build a version of PCRE that works this way, add
459 --disable-stack-for-recursion
461 to the configure command. With this configuration, PCRE will use the
462 pcre_stack_malloc and pcre_stack_free variables to call memory manage-
463 ment functions. By default these point to malloc() and free(), but you
464 can replace the pointers so that your own functions are used instead.
466 Separate functions are provided rather than using pcre_malloc and
467 pcre_free because the usage is very predictable: the block sizes
468 requested are always the same, and the blocks are always freed in
469 reverse order. A calling program might be able to implement optimized
470 functions that perform better than malloc() and free(). PCRE runs
471 noticeably more slowly when built in this way. This option affects only
472 the pcre_exec() function; it is not relevant for pcre_dfa_exec().
477 Internally, PCRE has a function called match(), which it calls repeat-
478 edly (sometimes recursively) when matching a pattern with the
479 pcre_exec() function. By controlling the maximum number of times this
480 function may be called during a single matching operation, a limit can
481 be placed on the resources used by a single call to pcre_exec(). The
482 limit can be changed at run time, as described in the pcreapi documen-
483 tation. The default is 10 million, but this can be changed by adding a
484 setting such as
486 --with-match-limit=500000
488 to the configure command. This setting has no effect on the
489 pcre_dfa_exec() matching function.
491 In some environments it is desirable to limit the depth of recursive
492 calls of match() more strictly than the total number of calls, in order
493 to restrict the maximum amount of stack (or heap, if --disable-stack-
494 for-recursion is specified) that is used. A second limit controls this;
495 it defaults to the value that is set for --with-match-limit, which
496 imposes no additional constraints. However, you can set a lower limit
497 by adding, for example,
499 --with-match-limit-recursion=10000
501 to the configure command. This value can also be overridden at run
502 time.
507 PCRE uses fixed tables for processing characters whose code values are
508 less than 256. By default, PCRE is built with a set of tables that are
509 distributed in the file pcre_chartables.c.dist. These tables are for
510 ASCII codes only. If you add
512 --enable-rebuild-chartables
514 to the configure command, the distributed tables are no longer used.
515 Instead, a program called dftables is compiled and run. This outputs
516 the source for new set of tables, created in the default locale of your
517 C runtime system. (This method of replacing the tables does not work if
518 you are cross compiling, because dftables is run on the local host. If
519 you need to create alternative tables when cross compiling, you will
520 have to do so "by hand".)
525 PCRE assumes by default that it will run in an environment where the
526 character code is ASCII (or Unicode, which is a superset of ASCII).
527 This is the case for most computer operating systems. PCRE can, how-
528 ever, be compiled to run in an EBCDIC environment by adding
530 --enable-ebcdic
532 to the configure command. This setting implies --enable-rebuild-charta-
533 bles. You should only use it if you know that you are in an EBCDIC
534 environment (for example, an IBM mainframe operating system). The
535 --enable-ebcdic option is incompatible with --enable-utf8.
540 By default, pcregrep reads all files as plain text. You can build it so
541 that it recognizes files whose names end in .gz or .bz2, and reads them
542 with libz or libbz2, respectively, by adding one or both of
544 --enable-pcregrep-libz
545 --enable-pcregrep-libbz2
547 to the configure command. These options naturally require that the rel-
548 evant libraries are installed on your system. Configuration will fail
549 if they are not.
554 If you add
556 --enable-pcretest-libreadline
558 to the configure command, pcretest is linked with the libreadline
559 library, and when its input is from a terminal, it reads it using the
560 readline() function. This provides line-editing and history facilities.
561 Note that libreadline is GPL-licensed, so if you distribute a binary of
562 pcretest linked in this way, there may be licensing issues.
564 Setting this option causes the -lreadline option to be added to the
565 pcretest build. In many operating environments with a sytem-installed
566 libreadline this is sufficient. However, in some environments (e.g. if
567 an unmodified distribution version of readline is in use), some extra
568 configuration may be necessary. The INSTALL file for libreadline says
569 this:
571 "Readline uses the termcap functions, but does not link with the
572 termcap or curses library itself, allowing applications which link
573 with readline the to choose an appropriate library."
575 If your environment has not been set up so that an appropriate library
576 is automatically included, you may need to add something like
578 LIBS="-ncurses"
580 immediately before the configure command.
585 pcreapi(3), pcre_config(3).
590 Philip Hazel
591 University Computing Service
592 Cambridge CB2 3QH, England.
597 Last updated: 29 September 2009
598 Copyright (c) 1997-2009 University of Cambridge.
599 ------------------------------------------------------------------------------
605 NAME
606 PCRE - Perl-compatible regular expressions
611 This document describes the two different algorithms that are available
612 in PCRE for matching a compiled regular expression against a given sub-
613 ject string. The "standard" algorithm is the one provided by the
614 pcre_exec() function. This works in the same was as Perl's matching
615 function, and provides a Perl-compatible matching operation.
617 An alternative algorithm is provided by the pcre_dfa_exec() function;
618 this operates in a different way, and is not Perl-compatible. It has
619 advantages and disadvantages compared with the standard algorithm, and
620 these are described below.
622 When there is only one possible way in which a given subject string can
623 match a pattern, the two algorithms give the same answer. A difference
624 arises, however, when there are multiple possibilities. For example, if
625 the pattern
627 ^<.*>
629 is matched against the string
631 <something> <something else> <something further>
633 there are three possible answers. The standard algorithm finds only one
634 of them, whereas the alternative algorithm finds all three.
639 The set of strings that are matched by a regular expression can be rep-
640 resented as a tree structure. An unlimited repetition in the pattern
641 makes the tree of infinite size, but it is still a tree. Matching the
642 pattern to a given subject string (from a given starting point) can be
643 thought of as a search of the tree. There are two ways to search a
644 tree: depth-first and breadth-first, and these correspond to the two
645 matching algorithms provided by PCRE.
650 In the terminology of Jeffrey Friedl's book "Mastering Regular Expres-
651 sions", the standard algorithm is an "NFA algorithm". It conducts a
652 depth-first search of the pattern tree. That is, it proceeds along a
653 single path through the tree, checking that the subject matches what is
654 required. When there is a mismatch, the algorithm tries any alterna-
655 tives at the current point, and if they all fail, it backs up to the
656 previous branch point in the tree, and tries the next alternative
657 branch at that level. This often involves backing up (moving to the
658 left) in the subject string as well. The order in which repetition
659 branches are tried is controlled by the greedy or ungreedy nature of
660 the quantifier.
662 If a leaf node is reached, a matching string has been found, and at
663 that point the algorithm stops. Thus, if there is more than one possi-
664 ble match, this algorithm returns the first one that it finds. Whether
665 this is the shortest, the longest, or some intermediate length depends
666 on the way the greedy and ungreedy repetition quantifiers are specified
667 in the pattern.
669 Because it ends up with a single path through the tree, it is rela-
670 tively straightforward for this algorithm to keep track of the sub-
671 strings that are matched by portions of the pattern in parentheses.
672 This provides support for capturing parentheses and back references.
677 This algorithm conducts a breadth-first search of the tree. Starting
678 from the first matching point in the subject, it scans the subject
679 string from left to right, once, character by character, and as it does
680 this, it remembers all the paths through the tree that represent valid
681 matches. In Friedl's terminology, this is a kind of "DFA algorithm",
682 though it is not implemented as a traditional finite state machine (it
683 keeps multiple states active simultaneously).
685 Although the general principle of this matching algorithm is that it
686 scans the subject string only once, without backtracking, there is one
687 exception: when a lookaround assertion is encountered, the characters
688 following or preceding the current point have to be independently
689 inspected.
691 The scan continues until either the end of the subject is reached, or
692 there are no more unterminated paths. At this point, terminated paths
693 represent the different matching possibilities (if there are none, the
694 match has failed). Thus, if there is more than one possible match,
695 this algorithm finds all of them, and in particular, it finds the long-
696 est. There is an option to stop the algorithm after the first match
697 (which is necessarily the shortest) is found.
699 Note that all the matches that are found start at the same point in the
700 subject. If the pattern
702 cat(er(pillar)?)
704 is matched against the string "the caterpillar catchment", the result
705 will be the three strings "cat", "cater", and "caterpillar" that start
706 at the fourth character of the subject. The algorithm does not automat-
707 ically move on to find matches that start at later positions.
709 There are a number of features of PCRE regular expressions that are not
710 supported by the alternative matching algorithm. They are as follows:
712 1. Because the algorithm finds all possible matches, the greedy or
713 ungreedy nature of repetition quantifiers is not relevant. Greedy and
714 ungreedy quantifiers are treated in exactly the same way. However, pos-
715 sessive quantifiers can make a difference when what follows could also
716 match what is quantified, for example in a pattern like this:
718 ^a++\w!
720 This pattern matches "aaab!" but not "aaa!", which would be matched by
721 a non-possessive quantifier. Similarly, if an atomic group is present,
722 it is matched as if it were a standalone pattern at the current point,
723 and the longest match is then "locked in" for the rest of the overall
724 pattern.
726 2. When dealing with multiple paths through the tree simultaneously, it
727 is not straightforward to keep track of captured substrings for the
728 different matching possibilities, and PCRE's implementation of this
729 algorithm does not attempt to do this. This means that no captured sub-
730 strings are available.
732 3. Because no substrings are captured, back references within the pat-
733 tern are not supported, and cause errors if encountered.
735 4. For the same reason, conditional expressions that use a backrefer-
736 ence as the condition or test for a specific group recursion are not
737 supported.
739 5. Because many paths through the tree may be active, the \K escape
740 sequence, which resets the start of the match when encountered (but may
741 be on some paths and not on others), is not supported. It causes an
742 error if encountered.
744 6. Callouts are supported, but the value of the capture_top field is
745 always 1, and the value of the capture_last field is always -1.
747 7. The \C escape sequence, which (in the standard algorithm) matches a
748 single byte, even in UTF-8 mode, is not supported because the alterna-
749 tive algorithm moves through the subject string one character at a
750 time, for all active paths through the tree.
752 8. Except for (*FAIL), the backtracking control verbs such as (*PRUNE)
753 are not supported. (*FAIL) is supported, and behaves like a failing
754 negative assertion.
759 Using the alternative matching algorithm provides the following advan-
760 tages:
762 1. All possible matches (at a single point in the subject) are automat-
763 ically found, and in particular, the longest match is found. To find
764 more than one match using the standard algorithm, you have to do kludgy
765 things with callouts.
767 2. Because the alternative algorithm scans the subject string just
768 once, and never needs to backtrack, it is possible to pass very long
769 subject strings to the matching function in several pieces, checking
770 for partial matching each time. The pcrepartial documentation gives
771 details of partial matching.
776 The alternative algorithm suffers from a number of disadvantages:
778 1. It is substantially slower than the standard algorithm. This is
779 partly because it has to search for all possible matches, but is also
780 because it is less susceptible to optimization.
782 2. Capturing parentheses and back references are not supported.
784 3. Although atomic groups are supported, their use does not provide the
785 performance advantage that it does for the standard algorithm.
790 Philip Hazel
791 University Computing Service
792 Cambridge CB2 3QH, England.
797 Last updated: 29 September 2009
798 Copyright (c) 1997-2009 University of Cambridge.
799 ------------------------------------------------------------------------------
805 NAME
806 PCRE - Perl-compatible regular expressions
811 #include <pcre.h>
813 pcre *pcre_compile(const char *pattern, int options,
814 const char **errptr, int *erroffset,
815 const unsigned char *tableptr);
817 pcre *pcre_compile2(const char *pattern, int options,
818 int *errorcodeptr,
819 const char **errptr, int *erroffset,
820 const unsigned char *tableptr);
822 pcre_extra *pcre_study(const pcre *code, int options,
823 const char **errptr);
825 int pcre_exec(const pcre *code, const pcre_extra *extra,
826 const char *subject, int length, int startoffset,
827 int options, int *ovector, int ovecsize);
829 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
830 const char *subject, int length, int startoffset,
831 int options, int *ovector, int ovecsize,
832 int *workspace, int wscount);
834 int pcre_copy_named_substring(const pcre *code,
835 const char *subject, int *ovector,
836 int stringcount, const char *stringname,
837 char *buffer, int buffersize);
839 int pcre_copy_substring(const char *subject, int *ovector,
840 int stringcount, int stringnumber, char *buffer,
841 int buffersize);
843 int pcre_get_named_substring(const pcre *code,
844 const char *subject, int *ovector,
845 int stringcount, const char *stringname,
846 const char **stringptr);
848 int pcre_get_stringnumber(const pcre *code,
849 const char *name);
851 int pcre_get_stringtable_entries(const pcre *code,
852 const char *name, char **first, char **last);
854 int pcre_get_substring(const char *subject, int *ovector,
855 int stringcount, int stringnumber,
856 const char **stringptr);
858 int pcre_get_substring_list(const char *subject,
859 int *ovector, int stringcount, const char ***listptr);
861 void pcre_free_substring(const char *stringptr);
863 void pcre_free_substring_list(const char **stringptr);
865 const unsigned char *pcre_maketables(void);
867 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
868 int what, void *where);
870 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
872 int pcre_refcount(pcre *code, int adjust);
874 int pcre_config(int what, void *where);
876 char *pcre_version(void);
878 void *(*pcre_malloc)(size_t);
880 void (*pcre_free)(void *);
882 void *(*pcre_stack_malloc)(size_t);
884 void (*pcre_stack_free)(void *);
886 int (*pcre_callout)(pcre_callout_block *);
891 PCRE has its own native API, which is described in this document. There
892 are also some wrapper functions that correspond to the POSIX regular
893 expression API. These are described in the pcreposix documentation.
894 Both of these APIs define a set of C function calls. A C++ wrapper is
895 distributed with PCRE. It is documented in the pcrecpp page.
897 The native API C function prototypes are defined in the header file
898 pcre.h, and on Unix systems the library itself is called libpcre. It
899 can normally be accessed by adding -lpcre to the command for linking an
900 application that uses PCRE. The header file defines the macros
901 PCRE_MAJOR and PCRE_MINOR to contain the major and minor release num-
902 bers for the library. Applications can use these to include support
903 for different releases of PCRE.
905 The functions pcre_compile(), pcre_compile2(), pcre_study(), and
906 pcre_exec() are used for compiling and matching regular expressions in
907 a Perl-compatible manner. A sample program that demonstrates the sim-
908 plest way of using them is provided in the file called pcredemo.c in
909 the PCRE source distribution. A listing of this program is given in the
910 pcredemo documentation, and the pcresample documentation describes how
911 to compile and run it.
913 A second matching function, pcre_dfa_exec(), which is not Perl-compati-
914 ble, is also provided. This uses a different algorithm for the match-
915 ing. The alternative algorithm finds all possible matches (at a given
916 point in the subject), and scans the subject just once (unless there
917 are lookbehind assertions). However, this algorithm does not return
918 captured substrings. A description of the two matching algorithms and
919 their advantages and disadvantages is given in the pcrematching docu-
920 mentation.
922 In addition to the main compiling and matching functions, there are
923 convenience functions for extracting captured substrings from a subject
924 string that is matched by pcre_exec(). They are:
926 pcre_copy_substring()
927 pcre_copy_named_substring()
928 pcre_get_substring()
929 pcre_get_named_substring()
930 pcre_get_substring_list()
931 pcre_get_stringnumber()
932 pcre_get_stringtable_entries()
934 pcre_free_substring() and pcre_free_substring_list() are also provided,
935 to free the memory used for extracted strings.
937 The function pcre_maketables() is used to build a set of character
938 tables in the current locale for passing to pcre_compile(),
939 pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is
940 provided for specialist use. Most commonly, no special tables are
941 passed, in which case internal tables that are generated when PCRE is
942 built are used.
944 The function pcre_fullinfo() is used to find out information about a
945 compiled pattern; pcre_info() is an obsolete version that returns only
946 some of the available information, but is retained for backwards com-
947 patibility. The function pcre_version() returns a pointer to a string
948 containing the version of PCRE and its date of release.
950 The function pcre_refcount() maintains a reference count in a data
951 block containing a compiled pattern. This is provided for the benefit
952 of object-oriented applications.
954 The global variables pcre_malloc and pcre_free initially contain the
955 entry points of the standard malloc() and free() functions, respec-
956 tively. PCRE calls the memory management functions via these variables,
957 so a calling program can replace them if it wishes to intercept the
958 calls. This should be done before calling any PCRE functions.
960 The global variables pcre_stack_malloc and pcre_stack_free are also
961 indirections to memory management functions. These special functions
962 are used only when PCRE is compiled to use the heap for remembering
963 data, instead of recursive function calls, when running the pcre_exec()
964 function. See the pcrebuild documentation for details of how to do
965 this. It is a non-standard way of building PCRE, for use in environ-
966 ments that have limited stacks. Because of the greater use of memory
967 management, it runs more slowly. Separate functions are provided so
968 that special-purpose external code can be used for this case. When
969 used, these functions are always called in a stack-like manner (last
970 obtained, first freed), and always for memory blocks of the same size.
971 There is a discussion about PCRE's stack usage in the pcrestack docu-
972 mentation.
974 The global variable pcre_callout initially contains NULL. It can be set
975 by the caller to a "callout" function, which PCRE will then call at
976 specified points during a matching operation. Details are given in the
977 pcrecallout documentation.
982 PCRE supports five different conventions for indicating line breaks in
983 strings: a single CR (carriage return) character, a single LF (line-
984 feed) character, the two-character sequence CRLF, any of the three pre-
985 ceding, or any Unicode newline sequence. The Unicode newline sequences
986 are the three just mentioned, plus the single characters VT (vertical
987 tab, U+000B), FF (formfeed, U+000C), NEL (next line, U+0085), LS (line
988 separator, U+2028), and PS (paragraph separator, U+2029).
990 Each of the first three conventions is used by at least one operating
991 system as its standard newline sequence. When PCRE is built, a default
992 can be specified. The default default is LF, which is the Unix stan-
993 dard. When PCRE is run, the default can be overridden, either when a
994 pattern is compiled, or when it is matched.
996 At compile time, the newline convention can be specified by the options
997 argument of pcre_compile(), or it can be specified by special text at
998 the start of the pattern itself; this overrides any other settings. See
999 the pcrepattern page for details of the special character sequences.
1001 In the PCRE documentation the word "newline" is used to mean "the char-
1002 acter or pair of characters that indicate a line break". The choice of
1003 newline convention affects the handling of the dot, circumflex, and
1004 dollar metacharacters, the handling of #-comments in /x mode, and, when
1005 CRLF is a recognized line ending sequence, the match position advance-
1006 ment for a non-anchored pattern. There is more detail about this in the
1007 section on pcre_exec() options below.
1009 The choice of newline convention does not affect the interpretation of
1010 the \n or \r escape sequences, nor does it affect what \R matches,
1011 which is controlled in a similar way, but by separate options.
1016 The PCRE functions can be used in multi-threading applications, with
1017 the proviso that the memory management functions pointed to by
1018 pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the
1019 callout function pointed to by pcre_callout, are shared by all threads.
1021 The compiled form of a regular expression is not altered during match-
1022 ing, so the same compiled pattern can safely be used by several threads
1023 at once.
1028 The compiled form of a regular expression can be saved and re-used at a
1029 later time, possibly by a different program, and even on a host other
1030 than the one on which it was compiled. Details are given in the
1031 pcreprecompile documentation. However, compiling a regular expression
1032 with one version of PCRE for use with a different version is not guar-
1033 anteed to work and may cause crashes.
1038 int pcre_config(int what, void *where);
1040 The function pcre_config() makes it possible for a PCRE client to dis-
1041 cover which optional features have been compiled into the PCRE library.
1042 The pcrebuild documentation has more details about these optional fea-
1043 tures.
1045 The first argument for pcre_config() is an integer, specifying which
1046 information is required; the second argument is a pointer to a variable
1047 into which the information is placed. The following information is
1048 available:
1052 The output is an integer that is set to one if UTF-8 support is avail-
1053 able; otherwise it is set to zero.
1057 The output is an integer that is set to one if support for Unicode
1058 character properties is available; otherwise it is set to zero.
1062 The output is an integer whose value specifies the default character
1063 sequence that is recognized as meaning "newline". The four values that
1064 are supported are: 10 for LF, 13 for CR, 3338 for CRLF, -2 for ANYCRLF,
1065 and -1 for ANY. Though they are derived from ASCII, the same values
1066 are returned in EBCDIC environments. The default should normally corre-
1067 spond to the standard sequence for your operating system.
1071 The output is an integer whose value indicates what character sequences
1072 the \R escape sequence matches by default. A value of 0 means that \R
1073 matches any Unicode line ending sequence; a value of 1 means that \R
1074 matches only CR, LF, or CRLF. The default can be overridden when a pat-
1075 tern is compiled or matched.
1079 The output is an integer that contains the number of bytes used for
1080 internal linkage in compiled regular expressions. The value is 2, 3, or
1081 4. Larger values allow larger regular expressions to be compiled, at
1082 the expense of slower matching. The default value of 2 is sufficient
1083 for all but the most massive patterns, since it allows the compiled
1084 pattern to be up to 64K in size.
1088 The output is an integer that contains the threshold above which the
1089 POSIX interface uses malloc() for output vectors. Further details are
1090 given in the pcreposix documentation.
1094 The output is a long integer that gives the default limit for the num-
1095 ber of internal matching function calls in a pcre_exec() execution.
1096 Further details are given with pcre_exec() below.
1100 The output is a long integer that gives the default limit for the depth
1101 of recursion when calling the internal matching function in a
1102 pcre_exec() execution. Further details are given with pcre_exec()
1103 below.
1107 The output is an integer that is set to one if internal recursion when
1108 running pcre_exec() is implemented by recursive function calls that use
1109 the stack to remember their state. This is the usual way that PCRE is
1110 compiled. The output is zero if PCRE was compiled to use blocks of data
1111 on the heap instead of recursive function calls. In this case,
1112 pcre_stack_malloc and pcre_stack_free are called to manage memory
1113 blocks on the heap, thus avoiding the use of the stack.
1118 pcre *pcre_compile(const char *pattern, int options,
1119 const char **errptr, int *erroffset,
1120 const unsigned char *tableptr);
1122 pcre *pcre_compile2(const char *pattern, int options,
1123 int *errorcodeptr,
1124 const char **errptr, int *erroffset,
1125 const unsigned char *tableptr);
1127 Either of the functions pcre_compile() or pcre_compile2() can be called
1128 to compile a pattern into an internal form. The only difference between
1129 the two interfaces is that pcre_compile2() has an additional argument,
1130 errorcodeptr, via which a numerical error code can be returned. To
1131 avoid too much repetition, we refer just to pcre_compile() below, but
1132 the information applies equally to pcre_compile2().
1134 The pattern is a C string terminated by a binary zero, and is passed in
1135 the pattern argument. A pointer to a single block of memory that is
1136 obtained via pcre_malloc is returned. This contains the compiled code
1137 and related data. The pcre type is defined for the returned block; this
1138 is a typedef for a structure whose contents are not externally defined.
1139 It is up to the caller to free the memory (via pcre_free) when it is no
1140 longer required.
1142 Although the compiled code of a PCRE regex is relocatable, that is, it
1143 does not depend on memory location, the complete pcre data block is not
1144 fully relocatable, because it may contain a copy of the tableptr argu-
1145 ment, which is an address (see below).
1147 The options argument contains various bit settings that affect the com-
1148 pilation. It should be zero if no options are required. The available
1149 options are described below. Some of them (in particular, those that
1150 are compatible with Perl, but some others as well) can also be set and
1151 unset from within the pattern (see the detailed description in the
1152 pcrepattern documentation). For those options that can be different in
1153 different parts of the pattern, the contents of the options argument
1154 specifies their settings at the start of compilation and execution. The
1155 PCRE_ANCHORED, PCRE_BSR_xxx, and PCRE_NEWLINE_xxx options can be set at
1156 the time of matching as well as at compile time.
1158 If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise,
1159 if compilation of a pattern fails, pcre_compile() returns NULL, and
1160 sets the variable pointed to by errptr to point to a textual error mes-
1161 sage. This is a static string that is part of the library. You must not
1162 try to free it. The byte offset from the start of the pattern to the
1163 character that was being processed when the error was discovered is
1164 placed in the variable pointed to by erroffset, which must not be NULL.
1165 If it is, an immediate error is given. Some errors are not detected
1166 until checks are carried out when the whole pattern has been scanned;
1167 in this case the offset is set to the end of the pattern.
1169 If pcre_compile2() is used instead of pcre_compile(), and the error-
1170 codeptr argument is not NULL, a non-zero error code number is returned
1171 via this argument in the event of an error. This is in addition to the
1172 textual error message. Error codes and messages are listed below.
1174 If the final argument, tableptr, is NULL, PCRE uses a default set of
1175 character tables that are built when PCRE is compiled, using the
1176 default C locale. Otherwise, tableptr must be an address that is the
1177 result of a call to pcre_maketables(). This value is stored with the
1178 compiled pattern, and used again by pcre_exec(), unless another table
1179 pointer is passed to it. For more discussion, see the section on locale
1180 support below.
1182 This code fragment shows a typical straightforward call to pcre_com-
1183 pile():
1185 pcre *re;
1186 const char *error;
1187 int erroffset;
1188 re = pcre_compile(
1189 "^A.*Z", /* the pattern */
1190 0, /* default options */
1191 &error, /* for error message */
1192 &erroffset, /* for error offset */
1193 NULL); /* use default character tables */
1195 The following names for option bits are defined in the pcre.h header
1196 file:
1200 If this bit is set, the pattern is forced to be "anchored", that is, it
1201 is constrained to match only at the first matching point in the string
1202 that is being searched (the "subject string"). This effect can also be
1203 achieved by appropriate constructs in the pattern itself, which is the
1204 only way to do it in Perl.
1208 If this bit is set, pcre_compile() automatically inserts callout items,
1209 all with number 255, before each pattern item. For discussion of the
1210 callout facility, see the pcrecallout documentation.
1215 These options (which are mutually exclusive) control what the \R escape
1216 sequence matches. The choice is either to match only CR, LF, or CRLF,
1217 or to match any Unicode newline sequence. The default is specified when
1218 PCRE is built. It can be overridden from within the pattern, or by set-
1219 ting an option when a compiled pattern is matched.
1223 If this bit is set, letters in the pattern match both upper and lower
1224 case letters. It is equivalent to Perl's /i option, and it can be
1225 changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE
1226 always understands the concept of case for characters whose values are
1227 less than 128, so caseless matching is always possible. For characters
1228 with higher values, the concept of case is supported if PCRE is com-
1229 piled with Unicode property support, but not otherwise. If you want to
1230 use caseless matching for characters 128 and above, you must ensure
1231 that PCRE is compiled with Unicode property support as well as with
1232 UTF-8 support.
1236 If this bit is set, a dollar metacharacter in the pattern matches only
1237 at the end of the subject string. Without this option, a dollar also
1238 matches immediately before a newline at the end of the string (but not
1239 before any other newlines). The PCRE_DOLLAR_ENDONLY option is ignored
1240 if PCRE_MULTILINE is set. There is no equivalent to this option in
1241 Perl, and no way to set it within a pattern.
1245 If this bit is set, a dot metacharater in the pattern matches all char-
1246 acters, including those that indicate newline. Without it, a dot does
1247 not match when the current position is at a newline. This option is
1248 equivalent to Perl's /s option, and it can be changed within a pattern
1249 by a (?s) option setting. A negative class such as [^a] always matches
1250 newline characters, independent of the setting of this option.
1254 If this bit is set, names used to identify capturing subpatterns need
1255 not be unique. This can be helpful for certain types of pattern when it
1256 is known that only one instance of the named subpattern can ever be
1257 matched. There are more details of named subpatterns below; see also
1258 the pcrepattern documentation.
1262 If this bit is set, whitespace data characters in the pattern are
1263 totally ignored except when escaped or inside a character class. White-
1264 space does not include the VT character (code 11). In addition, charac-
1265 ters between an unescaped # outside a character class and the next new-
1266 line, inclusive, are also ignored. This is equivalent to Perl's /x
1267 option, and it can be changed within a pattern by a (?x) option set-
1268 ting.
1270 This option makes it possible to include comments inside complicated
1271 patterns. Note, however, that this applies only to data characters.
1272 Whitespace characters may never appear within special character
1273 sequences in a pattern, for example within the sequence (?( which
1274 introduces a conditional subpattern.
1278 This option was invented in order to turn on additional functionality
1279 of PCRE that is incompatible with Perl, but it is currently of very
1280 little use. When set, any backslash in a pattern that is followed by a
1281 letter that has no special meaning causes an error, thus reserving
1282 these combinations for future expansion. By default, as in Perl, a
1283 backslash followed by a letter with no special meaning is treated as a
1284 literal. (Perl can, however, be persuaded to give a warning for this.)
1285 There are at present no other features controlled by this option. It
1286 can also be set by a (?X) option setting within a pattern.
1290 If this option is set, an unanchored pattern is required to match
1291 before or at the first newline in the subject string, though the
1292 matched text may continue over the newline.
1296 If this option is set, PCRE's behaviour is changed in some ways so that
1297 it is compatible with JavaScript rather than Perl. The changes are as
1298 follows:
1300 (1) A lone closing square bracket in a pattern causes a compile-time
1301 error, because this is illegal in JavaScript (by default it is treated
1302 as a data character). Thus, the pattern AB]CD becomes illegal when this
1303 option is set.
1305 (2) At run time, a back reference to an unset subpattern group matches
1306 an empty string (by default this causes the current matching alterna-
1307 tive to fail). A pattern such as (\1)(a) succeeds when this option is
1308 set (assuming it can find an "a" in the subject), whereas it fails by
1309 default, for Perl compatibility.
1313 By default, PCRE treats the subject string as consisting of a single
1314 line of characters (even if it actually contains newlines). The "start
1315 of line" metacharacter (^) matches only at the start of the string,
1316 while the "end of line" metacharacter ($) matches only at the end of
1317 the string, or before a terminating newline (unless PCRE_DOLLAR_ENDONLY
1318 is set). This is the same as Perl.
1320 When PCRE_MULTILINE it is set, the "start of line" and "end of line"
1321 constructs match immediately following or immediately before internal
1322 newlines in the subject string, respectively, as well as at the very
1323 start and end. This is equivalent to Perl's /m option, and it can be
1324 changed within a pattern by a (?m) option setting. If there are no new-
1325 lines in a subject string, or no occurrences of ^ or $ in a pattern,
1326 setting PCRE_MULTILINE has no effect.
1334 These options override the default newline definition that was chosen
1335 when PCRE was built. Setting the first or the second specifies that a
1336 newline is indicated by a single character (CR or LF, respectively).
1337 Setting PCRE_NEWLINE_CRLF specifies that a newline is indicated by the
1338 two-character CRLF sequence. Setting PCRE_NEWLINE_ANYCRLF specifies
1339 that any of the three preceding sequences should be recognized. Setting
1340 PCRE_NEWLINE_ANY specifies that any Unicode newline sequence should be
1341 recognized. The Unicode newline sequences are the three just mentioned,
1342 plus the single characters VT (vertical tab, U+000B), FF (formfeed,
1343 U+000C), NEL (next line, U+0085), LS (line separator, U+2028), and PS
1344 (paragraph separator, U+2029). The last two are recognized only in
1345 UTF-8 mode.
1347 The newline setting in the options word uses three bits that are
1348 treated as a number, giving eight possibilities. Currently only six are
1349 used (default plus the five values above). This means that if you set
1350 more than one newline option, the combination may or may not be sensi-
1351 ble. For example, PCRE_NEWLINE_CR with PCRE_NEWLINE_LF is equivalent to
1352 PCRE_NEWLINE_CRLF, but other combinations may yield unused numbers and
1353 cause an error.
1355 The only time that a line break is specially recognized when compiling
1356 a pattern is if PCRE_EXTENDED is set, and an unescaped # outside a
1357 character class is encountered. This indicates a comment that lasts
1358 until after the next line break sequence. In other circumstances, line
1359 break sequences are treated as literal data, except that in
1360 PCRE_EXTENDED mode, both CR and LF are treated as whitespace characters
1361 and are therefore ignored.
1363 The newline option that is set at compile time becomes the default that
1364 is used for pcre_exec() and pcre_dfa_exec(), but it can be overridden.
1368 If this option is set, it disables the use of numbered capturing paren-
1369 theses in the pattern. Any opening parenthesis that is not followed by
1370 ? behaves as if it were followed by ?: but named parentheses can still
1371 be used for capturing (and they acquire numbers in the usual way).
1372 There is no equivalent of this option in Perl.
1376 This option inverts the "greediness" of the quantifiers so that they
1377 are not greedy by default, but become greedy if followed by "?". It is
1378 not compatible with Perl. It can also be set by a (?U) option setting
1379 within the pattern.
1381 PCRE_UTF8
1383 This option causes PCRE to regard both the pattern and the subject as
1384 strings of UTF-8 characters instead of single-byte character strings.
1385 However, it is available only when PCRE is built to include UTF-8 sup-
1386 port. If not, the use of this option provokes an error. Details of how
1387 this option changes the behaviour of PCRE are given in the section on
1388 UTF-8 support in the main pcre page.
1392 When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is
1393 automatically checked. There is a discussion about the validity of
1394 UTF-8 strings in the main pcre page. If an invalid UTF-8 sequence of
1395 bytes is found, pcre_compile() returns an error. If you already know
1396 that your pattern is valid, and you want to skip this check for perfor-
1397 mance reasons, you can set the PCRE_NO_UTF8_CHECK option. When it is
1398 set, the effect of passing an invalid UTF-8 string as a pattern is
1399 undefined. It may cause your program to crash. Note that this option
1400 can also be passed to pcre_exec() and pcre_dfa_exec(), to suppress the
1401 UTF-8 validity checking of subject strings.
1406 The following table lists the error codes than may be returned by
1407 pcre_compile2(), along with the error messages that may be returned by
1408 both compiling functions. As PCRE has developed, some error codes have
1409 fallen out of use. To avoid confusion, they have not been re-used.
1411 0 no error
1412 1 \ at end of pattern
1413 2 \c at end of pattern
1414 3 unrecognized character follows \
1415 4 numbers out of order in {} quantifier
1416 5 number too big in {} quantifier
1417 6 missing terminating ] for character class
1418 7 invalid escape sequence in character class
1419 8 range out of order in character class
1420 9 nothing to repeat
1421 10 [this code is not in use]
1422 11 internal error: unexpected repeat
1423 12 unrecognized character after (? or (?-
1424 13 POSIX named classes are supported only within a class
1425 14 missing )
1426 15 reference to non-existent subpattern
1427 16 erroffset passed as NULL
1428 17 unknown option bit(s) set
1429 18 missing ) after comment
1430 19 [this code is not in use]
1431 20 regular expression is too large
1432 21 failed to get memory
1433 22 unmatched parentheses
1434 23 internal error: code overflow
1435 24 unrecognized character after (?<
1436 25 lookbehind assertion is not fixed length
1437 26 malformed number or name after (?(
1438 27 conditional group contains more than two branches
1439 28 assertion expected after (?(
1440 29 (?R or (?[+-]digits must be followed by )
1441 30 unknown POSIX class name
1442 31 POSIX collating elements are not supported
1443 32 this version of PCRE is not compiled with PCRE_UTF8 support
1444 33 [this code is not in use]
1445 34 character value in \x{...} sequence is too large
1446 35 invalid condition (?(0)
1447 36 \C not allowed in lookbehind assertion
1448 37 PCRE does not support \L, \l, \N, \U, or \u
1449 38 number after (?C is > 255
1450 39 closing ) for (?C expected
1451 40 recursive call could loop indefinitely
1452 41 unrecognized character after (?P
1453 42 syntax error in subpattern name (missing terminator)
1454 43 two named subpatterns have the same name
1455 44 invalid UTF-8 string
1456 45 support for \P, \p, and \X has not been compiled
1457 46 malformed \P or \p sequence
1458 47 unknown property name after \P or \p
1459 48 subpattern name is too long (maximum 32 characters)
1460 49 too many named subpatterns (maximum 10000)
1461 50 [this code is not in use]
1462 51 octal value is greater than \377 (not in UTF-8 mode)
1463 52 internal error: overran compiling workspace
1464 53 internal error: previously-checked referenced subpattern not
1465 found
1466 54 DEFINE group contains more than one branch
1467 55 repeating a DEFINE group is not allowed
1468 56 inconsistent NEWLINE options
1469 57 \g is not followed by a braced, angle-bracketed, or quoted
1470 name/number or by a plain number
1471 58 a numbered reference must not be zero
1472 59 an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)
1473 60 (*VERB) not recognized
1474 61 number is too big
1475 62 subpattern name expected
1476 63 digit expected after (?+
1477 64 ] is an invalid data character in JavaScript compatibility mode
1478 65 different names for subpatterns of the same number are not
1479 allowed
1480 66 (*MARK) must have an argument
1482 The numbers 32 and 10000 in errors 48 and 49 are defaults; different
1483 values may be used if the limits were changed when PCRE was built.
1488 pcre_extra *pcre_study(const pcre *code, int options
1489 const char **errptr);
1491 If a compiled pattern is going to be used several times, it is worth
1492 spending more time analyzing it in order to speed up the time taken for
1493 matching. The function pcre_study() takes a pointer to a compiled pat-
1494 tern as its first argument. If studying the pattern produces additional
1495 information that will help speed up matching, pcre_study() returns a
1496 pointer to a pcre_extra block, in which the study_data field points to
1497 the results of the study.
1499 The returned value from pcre_study() can be passed directly to
1500 pcre_exec() or pcre_dfa_exec(). However, a pcre_extra block also con-
1501 tains other fields that can be set by the caller before the block is
1502 passed; these are described below in the section on matching a pattern.
1504 If studying the pattern does not produce any useful information,
1505 pcre_study() returns NULL. In that circumstance, if the calling program
1506 wants to pass any of the other fields to pcre_exec() or
1507 pcre_dfa_exec(), it must set up its own pcre_extra block.
1509 The second argument of pcre_study() contains option bits. At present,
1510 no options are defined, and this argument should always be zero.
1512 The third argument for pcre_study() is a pointer for an error message.
1513 If studying succeeds (even if no data is returned), the variable it
1514 points to is set to NULL. Otherwise it is set to point to a textual
1515 error message. This is a static string that is part of the library. You
1516 must not try to free it. You should test the error pointer for NULL
1517 after calling pcre_study(), to be sure that it has run successfully.
1519 This is a typical call to pcre_study():
1521 pcre_extra *pe;
1522 pe = pcre_study(
1523 re, /* result of pcre_compile() */
1524 0, /* no options exist */
1525 &error); /* set to NULL or points to a message */
1527 Studying a pattern does two things: first, a lower bound for the length
1528 of subject string that is needed to match the pattern is computed. This
1529 does not mean that there are any strings of that length that match, but
1530 it does guarantee that no shorter strings match. The value is used by
1531 pcre_exec() and pcre_dfa_exec() to avoid wasting time by trying to
1532 match strings that are shorter than the lower bound. You can find out
1533 the value in a calling program via the pcre_fullinfo() function.
1535 Studying a pattern is also useful for non-anchored patterns that do not
1536 have a single fixed starting character. A bitmap of possible starting
1537 bytes is created. This speeds up finding a position in the subject at
1538 which to start matching.
1543 PCRE handles caseless matching, and determines whether characters are
1544 letters, digits, or whatever, by reference to a set of tables, indexed
1545 by character value. When running in UTF-8 mode, this applies only to
1546 characters with codes less than 128. Higher-valued codes never match
1547 escapes such as \w or \d, but can be tested with \p if PCRE is built
1548 with Unicode character property support. The use of locales with Uni-
1549 code is discouraged. If you are handling characters with codes greater
1550 than 128, you should either use UTF-8 and Unicode, or use locales, but
1551 not try to mix the two.
1553 PCRE contains an internal set of tables that are used when the final
1554 argument of pcre_compile() is NULL. These are sufficient for many
1555 applications. Normally, the internal tables recognize only ASCII char-
1556 acters. However, when PCRE is built, it is possible to cause the inter-
1557 nal tables to be rebuilt in the default "C" locale of the local system,
1558 which may cause them to be different.
1560 The internal tables can always be overridden by tables supplied by the
1561 application that calls PCRE. These may be created in a different locale
1562 from the default. As more and more applications change to using Uni-
1563 code, the need for this locale support is expected to die away.
1565 External tables are built by calling the pcre_maketables() function,
1566 which has no arguments, in the relevant locale. The result can then be
1567 passed to pcre_compile() or pcre_exec() as often as necessary. For
1568 example, to build and use tables that are appropriate for the French
1569 locale (where accented characters with values greater than 128 are
1570 treated as letters), the following code could be used:
1572 setlocale(LC_CTYPE, "fr_FR");
1573 tables = pcre_maketables();
1574 re = pcre_compile(..., tables);
1576 The locale name "fr_FR" is used on Linux and other Unix-like systems;
1577 if you are using Windows, the name for the French locale is "french".
1579 When pcre_maketables() runs, the tables are built in memory that is
1580 obtained via pcre_malloc. It is the caller's responsibility to ensure
1581 that the memory containing the tables remains available for as long as
1582 it is needed.
1584 The pointer that is passed to pcre_compile() is saved with the compiled
1585 pattern, and the same tables are used via this pointer by pcre_study()
1586 and normally also by pcre_exec(). Thus, by default, for any single pat-
1587 tern, compilation, studying and matching all happen in the same locale,
1588 but different patterns can be compiled in different locales.
1590 It is possible to pass a table pointer or NULL (indicating the use of
1591 the internal tables) to pcre_exec(). Although not intended for this
1592 purpose, this facility could be used to match a pattern in a different
1593 locale from the one in which it was compiled. Passing table pointers at
1594 run time is discussed below in the section on matching a pattern.
1599 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
1600 int what, void *where);
1602 The pcre_fullinfo() function returns information about a compiled pat-
1603 tern. It replaces the obsolete pcre_info() function, which is neverthe-
1604 less retained for backwards compability (and is documented below).
1606 The first argument for pcre_fullinfo() is a pointer to the compiled
1607 pattern. The second argument is the result of pcre_study(), or NULL if
1608 the pattern was not studied. The third argument specifies which piece
1609 of information is required, and the fourth argument is a pointer to a
1610 variable to receive the data. The yield of the function is zero for
1611 success, or one of the following negative numbers:
1613 PCRE_ERROR_NULL the argument code was NULL
1614 the argument where was NULL
1615 PCRE_ERROR_BADMAGIC the "magic number" was not found
1616 PCRE_ERROR_BADOPTION the value of what was invalid
1618 The "magic number" is placed at the start of each compiled pattern as
1619 an simple check against passing an arbitrary memory pointer. Here is a
1620 typical call of pcre_fullinfo(), to obtain the length of the compiled
1621 pattern:
1623 int rc;
1624 size_t length;
1625 rc = pcre_fullinfo(
1626 re, /* result of pcre_compile() */
1627 pe, /* result of pcre_study(), or NULL */
1628 PCRE_INFO_SIZE, /* what is required */
1629 &length); /* where to put the data */
1631 The possible values for the third argument are defined in pcre.h, and
1632 are as follows:
1636 Return the number of the highest back reference in the pattern. The
1637 fourth argument should point to an int variable. Zero is returned if
1638 there are no back references.
1642 Return the number of capturing subpatterns in the pattern. The fourth
1643 argument should point to an int variable.
1647 Return a pointer to the internal default character tables within PCRE.
1648 The fourth argument should point to an unsigned char * variable. This
1649 information call is provided for internal use by the pcre_study() func-
1650 tion. External callers can cause PCRE to use its internal tables by
1651 passing a NULL table pointer.
1655 Return information about the first byte of any matched string, for a
1656 non-anchored pattern. The fourth argument should point to an int vari-
1657 able. (This option used to be called PCRE_INFO_FIRSTCHAR; the old name
1658 is still recognized for backwards compatibility.)
1660 If there is a fixed first byte, for example, from a pattern such as
1661 (cat|cow|coyote), its value is returned. Otherwise, if either
1663 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
1664 branch starts with "^", or
1666 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
1667 set (if it were set, the pattern would be anchored),
1669 -1 is returned, indicating that the pattern matches only at the start
1670 of a subject string or after any newline within the string. Otherwise
1671 -2 is returned. For anchored patterns, -2 is returned.
1675 If the pattern was studied, and this resulted in the construction of a
1676 256-bit table indicating a fixed set of bytes for the first byte in any
1677 matching string, a pointer to the table is returned. Otherwise NULL is
1678 returned. The fourth argument should point to an unsigned char * vari-
1679 able.
1683 Return 1 if the pattern contains any explicit matches for CR or LF
1684 characters, otherwise 0. The fourth argument should point to an int
1685 variable. An explicit match is either a literal CR or LF character, or
1686 \r or \n.
1690 Return 1 if the (?J) or (?-J) option setting is used in the pattern,
1691 otherwise 0. The fourth argument should point to an int variable. (?J)
1692 and (?-J) set and unset the local PCRE_DUPNAMES option, respectively.
1696 Return the value of the rightmost literal byte that must exist in any
1697 matched string, other than at its start, if such a byte has been
1698 recorded. The fourth argument should point to an int variable. If there
1699 is no such byte, -1 is returned. For anchored patterns, a last literal
1700 byte is recorded only if it follows something of variable length. For
1701 example, for the pattern /^a\d+z\d+/ the returned value is "z", but for
1702 /^a\dz\d/ the returned value is -1.
1706 If the pattern was studied and a minimum length for matching subject
1707 strings was computed, its value is returned. Otherwise the returned
1708 value is -1. The value is a number of characters, not bytes (this may
1709 be relevant in UTF-8 mode). The fourth argument should point to an int
1710 variable. A non-negative value is a lower bound to the length of any
1711 matching string. There may not be any strings of that length that do
1712 actually match, but every string that does match is at least that long.
1718 PCRE supports the use of named as well as numbered capturing parenthe-
1719 ses. The names are just an additional way of identifying the parenthe-
1720 ses, which still acquire numbers. Several convenience functions such as
1721 pcre_get_named_substring() are provided for extracting captured sub-
1722 strings by name. It is also possible to extract the data directly, by
1723 first converting the name to a number in order to access the correct
1724 pointers in the output vector (described with pcre_exec() below). To do
1725 the conversion, you need to use the name-to-number map, which is
1726 described by these three values.
1728 The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT
1729 gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size
1730 of each entry; both of these return an int value. The entry size
1731 depends on the length of the longest name. PCRE_INFO_NAMETABLE returns
1732 a pointer to the first entry of the table (a pointer to char). The
1733 first two bytes of each entry are the number of the capturing parenthe-
1734 sis, most significant byte first. The rest of the entry is the corre-
1735 sponding name, zero terminated.
1737 The names are in alphabetical order. Duplicate names may appear if (?|
1738 is used to create multiple groups with the same number, as described in
1739 the section on duplicate subpattern numbers in the pcrepattern page.
1740 Duplicate names for subpatterns with different numbers are permitted
1741 only if PCRE_DUPNAMES is set. In all cases of duplicate names, they
1742 appear in the table in the order in which they were found in the pat-
1743 tern. In the absence of (?| this is the order of increasing number;
1744 when (?| is used this is not necessarily the case because later subpat-
1745 terns may have lower numbers.
1747 As a simple example of the name/number table, consider the following
1748 pattern (assume PCRE_EXTENDED is set, so white space - including new-
1749 lines - is ignored):
1751 (?<date> (?<year>(\d\d)?\d\d) -
1752 (?<month>\d\d) - (?<day>\d\d) )
1754 There are four named subpatterns, so the table has four entries, and
1755 each entry in the table is eight bytes long. The table is as follows,
1756 with non-printing bytes shows in hexadecimal, and undefined bytes shown
1757 as ??:
1759 00 01 d a t e 00 ??
1760 00 05 d a y 00 ?? ??
1761 00 04 m o n t h 00
1762 00 02 y e a r 00 ??
1764 When writing code to extract data from named subpatterns using the
1765 name-to-number map, remember that the length of the entries is likely
1766 to be different for each compiled pattern.
1770 Return 1 if the pattern can be used for partial matching with
1771 pcre_exec(), otherwise 0. The fourth argument should point to an int
1772 variable. From release 8.00, this always returns 1, because the
1773 restrictions that previously applied to partial matching have been
1774 lifted. The pcrepartial documentation gives details of partial match-
1775 ing.
1779 Return a copy of the options with which the pattern was compiled. The
1780 fourth argument should point to an unsigned long int variable. These
1781 option bits are those specified in the call to pcre_compile(), modified
1782 by any top-level option settings at the start of the pattern itself. In
1783 other words, they are the options that will be in force when matching
1784 starts. For example, if the pattern /(?im)abc(?-i)d/ is compiled with
1785 the PCRE_EXTENDED option, the result is PCRE_CASELESS, PCRE_MULTILINE,
1788 A pattern is automatically anchored by PCRE if all of its top-level
1789 alternatives begin with one of the following:
1791 ^ unless PCRE_MULTILINE is set
1792 \A always
1793 \G always
1794 .* if PCRE_DOTALL is set and there are no back
1795 references to the subpattern in which .* appears
1797 For such patterns, the PCRE_ANCHORED bit is set in the options returned
1798 by pcre_fullinfo().
1802 Return the size of the compiled pattern, that is, the value that was
1803 passed as the argument to pcre_malloc() when PCRE was getting memory in
1804 which to place the compiled data. The fourth argument should point to a
1805 size_t variable.
1809 Return the size of the data block pointed to by the study_data field in
1810 a pcre_extra block. That is, it is the value that was passed to
1811 pcre_malloc() when PCRE was getting memory into which to place the data
1812 created by pcre_study(). If pcre_extra is NULL, or there is no study
1813 data, zero is returned. The fourth argument should point to a size_t
1814 variable.
1819 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
1821 The pcre_info() function is now obsolete because its interface is too
1822 restrictive to return all the available data about a compiled pattern.
1823 New programs should use pcre_fullinfo() instead. The yield of
1824 pcre_info() is the number of capturing subpatterns, or one of the fol-
1825 lowing negative numbers:
1827 PCRE_ERROR_NULL the argument code was NULL
1828 PCRE_ERROR_BADMAGIC the "magic number" was not found
1830 If the optptr argument is not NULL, a copy of the options with which
1831 the pattern was compiled is placed in the integer it points to (see
1832 PCRE_INFO_OPTIONS above).
1834 If the pattern is not anchored and the firstcharptr argument is not
1835 NULL, it is used to pass back information about the first character of
1836 any matched string (see PCRE_INFO_FIRSTBYTE above).
1841 int pcre_refcount(pcre *code, int adjust);
1843 The pcre_refcount() function is used to maintain a reference count in
1844 the data block that contains a compiled pattern. It is provided for the
1845 benefit of applications that operate in an object-oriented manner,
1846 where different parts of the application may be using the same compiled
1847 pattern, but you want to free the block when they are all done.
1849 When a pattern is compiled, the reference count field is initialized to
1850 zero. It is changed only by calling this function, whose action is to
1851 add the adjust value (which may be positive or negative) to it. The
1852 yield of the function is the new value. However, the value of the count
1853 is constrained to lie between 0 and 65535, inclusive. If the new value
1854 is outside these limits, it is forced to the appropriate limit value.
1856 Except when it is zero, the reference count is not correctly preserved
1857 if a pattern is compiled on one host and then transferred to a host
1858 whose byte-order is different. (This seems a highly unlikely scenario.)
1863 int pcre_exec(const pcre *code, const pcre_extra *extra,
1864 const char *subject, int length, int startoffset,
1865 int options, int *ovector, int ovecsize);
1867 The function pcre_exec() is called to match a subject string against a
1868 compiled pattern, which is passed in the code argument. If the pattern
1869 was studied, the result of the study should be passed in the extra
1870 argument. This function is the main matching facility of the library,
1871 and it operates in a Perl-like manner. For specialist use there is also
1872 an alternative matching function, which is described below in the sec-
1873 tion about the pcre_dfa_exec() function.
1875 In most applications, the pattern will have been compiled (and option-
1876 ally studied) in the same process that calls pcre_exec(). However, it
1877 is possible to save compiled patterns and study data, and then use them
1878 later in different processes, possibly even on different hosts. For a
1879 discussion about this, see the pcreprecompile documentation.
1881 Here is an example of a simple call to pcre_exec():
1883 int rc;
1884 int ovector[30];
1885 rc = pcre_exec(
1886 re, /* result of pcre_compile() */
1887 NULL, /* we didn't study the pattern */
1888 "some string", /* the subject string */
1889 11, /* the length of the subject string */
1890 0, /* start at offset 0 in the subject */
1891 0, /* default options */
1892 ovector, /* vector of integers for substring information */
1893 30); /* number of elements (NOT size in bytes) */
1895 Extra data for pcre_exec()
1897 If the extra argument is not NULL, it must point to a pcre_extra data
1898 block. The pcre_study() function returns such a block (when it doesn't
1899 return NULL), but you can also create one for yourself, and pass addi-
1900 tional information in it. The pcre_extra block contains the following
1901 fields (not necessarily in this order):
1903 unsigned long int flags;
1904 void *study_data;
1905 unsigned long int match_limit;
1906 unsigned long int match_limit_recursion;
1907 void *callout_data;
1908 const unsigned char *tables;
1909 unsigned char **mark;
1911 The flags field is a bitmap that specifies which of the other fields
1912 are set. The flag bits are:
1921 Other flag bits should be set to zero. The study_data field is set in
1922 the pcre_extra block that is returned by pcre_study(), together with
1923 the appropriate flag bit. You should not set this yourself, but you may
1924 add to the block by setting the other fields and their corresponding
1925 flag bits.
1927 The match_limit field provides a means of preventing PCRE from using up
1928 a vast amount of resources when running patterns that are not going to
1929 match, but which have a very large number of possibilities in their
1930 search trees. The classic example is a pattern that uses nested unlim-
1931 ited repeats.
1933 Internally, PCRE uses a function called match() which it calls repeat-
1934 edly (sometimes recursively). The limit set by match_limit is imposed
1935 on the number of times this function is called during a match, which
1936 has the effect of limiting the amount of backtracking that can take
1937 place. For patterns that are not anchored, the count restarts from zero
1938 for each position in the subject string.
1940 The default value for the limit can be set when PCRE is built; the
1941 default default is 10 million, which handles all but the most extreme
1942 cases. You can override the default by suppling pcre_exec() with a
1943 pcre_extra block in which match_limit is set, and
1944 PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is
1945 exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.
1947 The match_limit_recursion field is similar to match_limit, but instead
1948 of limiting the total number of times that match() is called, it limits
1949 the depth of recursion. The recursion depth is a smaller number than
1950 the total number of calls, because not all calls to match() are recur-
1951 sive. This limit is of use only if it is set smaller than match_limit.
1953 Limiting the recursion depth limits the amount of stack that can be
1954 used, or, when PCRE has been compiled to use memory on the heap instead
1955 of the stack, the amount of heap memory that can be used.
1957 The default value for match_limit_recursion can be set when PCRE is
1958 built; the default default is the same value as the default for
1959 match_limit. You can override the default by suppling pcre_exec() with
1960 a pcre_extra block in which match_limit_recursion is set, and
1961 PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the
1962 limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT.
1964 The callout_data field is used in conjunction with the "callout" fea-
1965 ture, and is described in the pcrecallout documentation.
1967 The tables field is used to pass a character tables pointer to
1968 pcre_exec(); this overrides the value that is stored with the compiled
1969 pattern. A non-NULL value is stored with the compiled pattern only if
1970 custom tables were supplied to pcre_compile() via its tableptr argu-
1971 ment. If NULL is passed to pcre_exec() using this mechanism, it forces
1972 PCRE's internal tables to be used. This facility is helpful when re-
1973 using patterns that have been saved after compiling with an external
1974 set of tables, because the external tables might be at a different
1975 address when pcre_exec() is called. See the pcreprecompile documenta-
1976 tion for a discussion of saving compiled patterns for later use.
1978 If PCRE_EXTRA_MARK is set in the flags field, the mark field must be
1979 set to point to a char * variable. If the pattern contains any back-
1980 tracking control verbs such as (*MARK:NAME), and the execution ends up
1981 with a name to pass back, a pointer to the name string (zero termi-
1982 nated) is placed in the variable pointed to by the mark field. The
1983 names are within the compiled pattern; if you wish to retain such a
1984 name you must copy it before freeing the memory of a compiled pattern.
1985 If there is no name to pass back, the variable pointed to by the mark
1986 field set to NULL. For details of the backtracking control verbs, see
1987 the section entitled "Backtracking control" in the pcrepattern documen-
1988 tation.
1990 Option bits for pcre_exec()
1992 The unused bits of the options argument for pcre_exec() must be zero.
1993 The only bits that may be set are PCRE_ANCHORED, PCRE_NEWLINE_xxx,
2000 The PCRE_ANCHORED option limits pcre_exec() to matching at the first
2001 matching position. If a pattern was compiled with PCRE_ANCHORED, or
2002 turned out to be anchored by virtue of its contents, it cannot be made
2003 unachored at matching time.
2008 These options (which are mutually exclusive) control what the \R escape
2009 sequence matches. The choice is either to match only CR, LF, or CRLF,
2010 or to match any Unicode newline sequence. These options override the
2011 choice that was made or defaulted when the pattern was compiled.
2019 These options override the newline definition that was chosen or
2020 defaulted when the pattern was compiled. For details, see the descrip-
2021 tion of pcre_compile() above. During matching, the newline choice
2022 affects the behaviour of the dot, circumflex, and dollar metacharac-
2023 ters. It may also alter the way the match position is advanced after a
2024 match failure for an unanchored pattern.
2027 set, and a match attempt for an unanchored pattern fails when the cur-
2028 rent position is at a CRLF sequence, and the pattern contains no
2029 explicit matches for CR or LF characters, the match position is
2030 advanced by two characters instead of one, in other words, to after the
2031 CRLF.
2033 The above rule is a compromise that makes the most common cases work as
2034 expected. For example, if the pattern is .+A (and the PCRE_DOTALL
2035 option is not set), it does not match the string "\r\nA" because, after
2036 failing at the start, it skips both the CR and the LF before retrying.
2037 However, the pattern [\r\n]A does match that string, because it con-
2038 tains an explicit CR or LF reference, and so advances only by one char-
2039 acter after the first failure.
2041 An explicit match for CR of LF is either a literal appearance of one of
2042 those characters, or one of the \r or \n escape sequences. Implicit
2043 matches such as [^X] do not count, nor does \s (which includes CR and
2044 LF in the characters that it matches).
2046 Notwithstanding the above, anomalous effects may still occur when CRLF
2047 is a valid newline sequence and explicit \r or \n escapes appear in the
2048 pattern.
2052 This option specifies that first character of the subject string is not
2053 the beginning of a line, so the circumflex metacharacter should not
2054 match before it. Setting this without PCRE_MULTILINE (at compile time)
2055 causes circumflex never to match. This option affects only the behav-
2056 iour of the circumflex metacharacter. It does not affect \A.
2060 This option specifies that the end of the subject string is not the end
2061 of a line, so the dollar metacharacter should not match it nor (except
2062 in multiline mode) a newline immediately before it. Setting this with-
2063 out PCRE_MULTILINE (at compile time) causes dollar never to match. This
2064 option affects only the behaviour of the dollar metacharacter. It does
2065 not affect \Z or \z.
2069 An empty string is not considered to be a valid match if this option is
2070 set. If there are alternatives in the pattern, they are tried. If all
2071 the alternatives match the empty string, the entire match fails. For
2072 example, if the pattern
2074 a?b?
2076 is applied to a string not beginning with "a" or "b", it matches an
2077 empty string at the start of the subject. With PCRE_NOTEMPTY set, this
2078 match is not valid, so PCRE searches further into the string for occur-
2079 rences of "a" or "b".
2083 This is like PCRE_NOTEMPTY, except that an empty string match that is
2084 not at the start of the subject is permitted. If the pattern is
2085 anchored, such a match can occur only if the pattern contains \K.
2087 Perl has no direct equivalent of PCRE_NOTEMPTY or
2088 PCRE_NOTEMPTY_ATSTART, but it does make a special case of a pattern
2089 match of the empty string within its split() function, and when using
2090 the /g modifier. It is possible to emulate Perl's behaviour after
2091 matching a null string by first trying the match again at the same off-
2092 set with PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED, and then if that
2093 fails, by advancing the starting offset (see below) and trying an ordi-
2094 nary match again. There is some code that demonstrates how to do this
2095 in the pcredemo sample program.
2099 There are a number of optimizations that pcre_exec() uses at the start
2100 of a match, in order to speed up the process. For example, if it is
2101 known that a match must start with a specific character, it searches
2102 the subject for that character, and fails immediately if it cannot find
2103 it, without actually running the main matching function. When callouts
2104 are in use, these optimizations can cause them to be skipped. This
2105 option disables the "start-up" optimizations, causing performance to
2106 suffer, but ensuring that the callouts do occur.
2110 When PCRE_UTF8 is set at compile time, the validity of the subject as a
2111 UTF-8 string is automatically checked when pcre_exec() is subsequently
2112 called. The value of startoffset is also checked to ensure that it
2113 points to the start of a UTF-8 character. There is a discussion about
2114 the validity of UTF-8 strings in the section on UTF-8 support in the
2115 main pcre page. If an invalid UTF-8 sequence of bytes is found,
2116 pcre_exec() returns the error PCRE_ERROR_BADUTF8. If startoffset con-
2117 tains an invalid value, PCRE_ERROR_BADUTF8_OFFSET is returned.
2119 If you already know that your subject is valid, and you want to skip
2120 these checks for performance reasons, you can set the
2121 PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to
2122 do this for the second and subsequent calls to pcre_exec() if you are
2123 making repeated calls to find all the matches in a single subject
2124 string. However, you should be sure that the value of startoffset
2125 points to the start of a UTF-8 character. When PCRE_NO_UTF8_CHECK is
2126 set, the effect of passing an invalid UTF-8 string as a subject, or a
2127 value of startoffset that does not point to the start of a UTF-8 char-
2128 acter, is undefined. Your program may crash.
2133 These options turn on the partial matching feature. For backwards com-
2134 patibility, PCRE_PARTIAL is a synonym for PCRE_PARTIAL_SOFT. A partial
2135 match occurs if the end of the subject string is reached successfully,
2136 but there are not enough subject characters to complete the match. If
2137 this happens when PCRE_PARTIAL_HARD is set, pcre_exec() immediately
2138 returns PCRE_ERROR_PARTIAL. Otherwise, if PCRE_PARTIAL_SOFT is set,
2139 matching continues by testing any other alternatives. Only if they all
2140 fail is PCRE_ERROR_PARTIAL returned (instead of PCRE_ERROR_NOMATCH).
2141 The portion of the string that was inspected when the partial match was
2142 found is set as the first matching string. There is a more detailed
2143 discussion in the pcrepartial documentation.
2145 The string to be matched by pcre_exec()
2147 The subject string is passed to pcre_exec() as a pointer in subject, a
2148 length (in bytes) in length, and a starting byte offset in startoffset.
2149 In UTF-8 mode, the byte offset must point to the start of a UTF-8 char-
2150 acter. Unlike the pattern string, the subject may contain binary zero
2151 bytes. When the starting offset is zero, the search for a match starts
2152 at the beginning of the subject, and this is by far the most common
2153 case.
2155 A non-zero starting offset is useful when searching for another match
2156 in the same subject by calling pcre_exec() again after a previous suc-
2157 cess. Setting startoffset differs from just passing over a shortened
2158 string and setting PCRE_NOTBOL in the case of a pattern that begins
2159 with any kind of lookbehind. For example, consider the pattern
2161 \Biss\B
2163 which finds occurrences of "iss" in the middle of words. (\B matches
2164 only if the current position in the subject is not a word boundary.)
2165 When applied to the string "Mississipi" the first call to pcre_exec()
2166 finds the first occurrence. If pcre_exec() is called again with just
2167 the remainder of the subject, namely "issipi", it does not match,
2168 because \B is always false at the start of the subject, which is deemed
2169 to be a word boundary. However, if pcre_exec() is passed the entire
2170 string again, but with startoffset set to 4, it finds the second occur-
2171 rence of "iss" because it is able to look behind the starting point to
2172 discover that it is preceded by a letter.
2174 If a non-zero starting offset is passed when the pattern is anchored,
2175 one attempt to match at the given offset is made. This can only succeed
2176 if the pattern does not require the match to be at the start of the
2177 subject.
2179 How pcre_exec() returns captured substrings
2181 In general, a pattern matches a certain portion of the subject, and in
2182 addition, further substrings from the subject may be picked out by
2183 parts of the pattern. Following the usage in Jeffrey Friedl's book,
2184 this is called "capturing" in what follows, and the phrase "capturing
2185 subpattern" is used for a fragment of a pattern that picks out a sub-
2186 string. PCRE supports several other kinds of parenthesized subpattern
2187 that do not cause substrings to be captured.
2189 Captured substrings are returned to the caller via a vector of integers
2190 whose address is passed in ovector. The number of elements in the vec-
2191 tor is passed in ovecsize, which must be a non-negative number. Note:
2192 this argument is NOT the size of ovector in bytes.
2194 The first two-thirds of the vector is used to pass back captured sub-
2195 strings, each substring using a pair of integers. The remaining third
2196 of the vector is used as workspace by pcre_exec() while matching cap-
2197 turing subpatterns, and is not available for passing back information.
2198 The number passed in ovecsize should always be a multiple of three. If
2199 it is not, it is rounded down.
2201 When a match is successful, information about captured substrings is
2202 returned in pairs of integers, starting at the beginning of ovector,
2203 and continuing up to two-thirds of its length at the most. The first
2204 element of each pair is set to the byte offset of the first character
2205 in a substring, and the second is set to the byte offset of the first
2206 character after the end of a substring. Note: these values are always
2207 byte offsets, even in UTF-8 mode. They are not character counts.
2209 The first pair of integers, ovector[0] and ovector[1], identify the
2210 portion of the subject string matched by the entire pattern. The next
2211 pair is used for the first capturing subpattern, and so on. The value
2212 returned by pcre_exec() is one more than the highest numbered pair that
2213 has been set. For example, if two substrings have been captured, the
2214 returned value is 3. If there are no capturing subpatterns, the return
2215 value from a successful match is 1, indicating that just the first pair
2216 of offsets has been set.
2218 If a capturing subpattern is matched repeatedly, it is the last portion
2219 of the string that it matched that is returned.
2221 If the vector is too small to hold all the captured substring offsets,
2222 it is used as far as possible (up to two-thirds of its length), and the
2223 function returns a value of zero. If the substring offsets are not of
2224 interest, pcre_exec() may be called with ovector passed as NULL and
2225 ovecsize as zero. However, if the pattern contains back references and
2226 the ovector is not big enough to remember the related substrings, PCRE
2227 has to get additional memory for use during matching. Thus it is usu-
2228 ally advisable to supply an ovector.
2230 The pcre_fullinfo() function can be used to find out how many capturing
2231 subpatterns there are in a compiled pattern. The smallest size for
2232 ovector that will allow for n captured substrings, in addition to the
2233 offsets of the substring matched by the whole pattern, is (n+1)*3.
2235 It is possible for capturing subpattern number n+1 to match some part
2236 of the subject when subpattern n has not been used at all. For example,
2237 if the string "abc" is matched against the pattern (a|(z))(bc) the
2238 return from the function is 4, and subpatterns 1 and 3 are matched, but
2239 2 is not. When this happens, both values in the offset pairs corre-
2240 sponding to unused subpatterns are set to -1.
2242 Offset values that correspond to unused subpatterns at the end of the
2243 expression are also set to -1. For example, if the string "abc" is
2244 matched against the pattern (abc)(x(yz)?)? subpatterns 2 and 3 are not
2245 matched. The return from the function is 2, because the highest used
2246 capturing subpattern number is 1. However, you can refer to the offsets
2247 for the second and third capturing subpatterns if you wish (assuming
2248 the vector is large enough, of course).
2250 Some convenience functions are provided for extracting the captured
2251 substrings as separate strings. These are described below.
2253 Error return values from pcre_exec()
2255 If pcre_exec() fails, it returns a negative number. The following are
2256 defined in the header file:
2260 The subject string did not match the pattern.
2264 Either code or subject was passed as NULL, or ovector was NULL and
2265 ovecsize was not zero.
2269 An unrecognized bit was set in the options argument.
2273 PCRE stores a 4-byte "magic number" at the start of the compiled code,
2274 to catch the case when it is passed a junk pointer and to detect when a
2275 pattern that was compiled in an environment of one endianness is run in
2276 an environment with the other endianness. This is the error that PCRE
2277 gives when the magic number is not present.
2281 While running the pattern match, an unknown item was encountered in the
2282 compiled pattern. This error could be caused by a bug in PCRE or by
2283 overwriting of the compiled pattern.
2287 If a pattern contains back references, but the ovector that is passed
2288 to pcre_exec() is not big enough to remember the referenced substrings,
2289 PCRE gets a block of memory at the start of matching to use for this
2290 purpose. If the call via pcre_malloc() fails, this error is given. The
2291 memory is automatically freed at the end of matching.
2295 This error is used by the pcre_copy_substring(), pcre_get_substring(),
2296 and pcre_get_substring_list() functions (see below). It is never
2297 returned by pcre_exec().
2301 The backtracking limit, as specified by the match_limit field in a
2302 pcre_extra structure (or defaulted) was reached. See the description
2303 above.
2307 This error is never generated by pcre_exec() itself. It is provided for
2308 use by callout functions that want to yield a distinctive error code.
2309 See the pcrecallout documentation for details.
2313 A string that contains an invalid UTF-8 byte sequence was passed as a
2314 subject.
2318 The UTF-8 byte sequence that was passed as a subject was valid, but the
2319 value of startoffset did not point to the beginning of a UTF-8 charac-
2320 ter.
2324 The subject string did not match, but it did match partially. See the
2325 pcrepartial documentation for details of partial matching.
2329 This code is no longer in use. It was formerly returned when the
2330 PCRE_PARTIAL option was used with a compiled pattern containing items
2331 that were not supported for partial matching. From release 8.00
2332 onwards, there are no restrictions on partial matching.
2336 An unexpected internal error has occurred. This error could be caused
2337 by a bug in PCRE or by overwriting of the compiled pattern.
2341 This error is given if the value of the ovecsize argument is negative.
2345 The internal recursion limit, as specified by the match_limit_recursion
2346 field in a pcre_extra structure (or defaulted) was reached. See the
2347 description above.
2351 An invalid combination of PCRE_NEWLINE_xxx options was given.
2353 Error numbers -16 to -20 and -22 are not used by pcre_exec().
2358 int pcre_copy_substring(const char *subject, int *ovector,
2359 int stringcount, int stringnumber, char *buffer,
2360 int buffersize);
2362 int pcre_get_substring(const char *subject, int *ovector,
2363 int stringcount, int stringnumber,
2364 const char **stringptr);
2366 int pcre_get_substring_list(const char *subject,
2367 int *ovector, int stringcount, const char ***listptr);
2369 Captured substrings can be accessed directly by using the offsets
2370 returned by pcre_exec() in ovector. For convenience, the functions
2371 pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub-
2372 string_list() are provided for extracting captured substrings as new,
2373 separate, zero-terminated strings. These functions identify substrings
2374 by number. The next section describes functions for extracting named
2375 substrings.
2377 A substring that contains a binary zero is correctly extracted and has
2378 a further zero added on the end, but the result is not, of course, a C
2379 string. However, you can process such a string by referring to the
2380 length that is returned by pcre_copy_substring() and pcre_get_sub-
2381 string(). Unfortunately, the interface to pcre_get_substring_list() is
2382 not adequate for handling strings containing binary zeros, because the
2383 end of the final string is not independently indicated.
2385 The first three arguments are the same for all three of these func-
2386 tions: subject is the subject string that has just been successfully
2387 matched, ovector is a pointer to the vector of integer offsets that was
2388 passed to pcre_exec(), and stringcount is the number of substrings that
2389 were captured by the match, including the substring that matched the
2390 entire regular expression. This is the value returned by pcre_exec() if
2391 it is greater than zero. If pcre_exec() returned zero, indicating that
2392 it ran out of space in ovector, the value passed as stringcount should
2393 be the number of elements in the vector divided by three.
2395 The functions pcre_copy_substring() and pcre_get_substring() extract a
2396 single substring, whose number is given as stringnumber. A value of
2397 zero extracts the substring that matched the entire pattern, whereas
2398 higher values extract the captured substrings. For pcre_copy_sub-
2399 string(), the string is placed in buffer, whose length is given by
2400 buffersize, while for pcre_get_substring() a new block of memory is
2401 obtained via pcre_malloc, and its address is returned via stringptr.
2402 The yield of the function is the length of the string, not including
2403 the terminating zero, or one of these error codes:
2407 The buffer was too small for pcre_copy_substring(), or the attempt to
2408 get memory failed for pcre_get_substring().
2412 There is no substring whose number is stringnumber.
2414 The pcre_get_substring_list() function extracts all available sub-
2415 strings and builds a list of pointers to them. All this is done in a
2416 single block of memory that is obtained via pcre_malloc. The address of
2417 the memory block is returned via listptr, which is also the start of
2418 the list of string pointers. The end of the list is marked by a NULL
2419 pointer. The yield of the function is zero if all went well, or the
2420 error code
2424 if the attempt to get the memory block failed.
2426 When any of these functions encounter a substring that is unset, which
2427 can happen when capturing subpattern number n+1 matches some part of
2428 the subject, but subpattern n has not been used at all, they return an
2429 empty string. This can be distinguished from a genuine zero-length sub-
2430 string by inspecting the appropriate offset in ovector, which is nega-
2431 tive for unset substrings.
2433 The two convenience functions pcre_free_substring() and pcre_free_sub-
2434 string_list() can be used to free the memory returned by a previous
2435 call of pcre_get_substring() or pcre_get_substring_list(), respec-
2436 tively. They do nothing more than call the function pointed to by
2437 pcre_free, which of course could be called directly from a C program.
2438 However, PCRE is used in some situations where it is linked via a spe-
2439 cial interface to another programming language that cannot use
2440 pcre_free directly; it is for these cases that the functions are pro-
2441 vided.
2446 int pcre_get_stringnumber(const pcre *code,
2447 const char *name);
2449 int pcre_copy_named_substring(const pcre *code,
2450 const char *subject, int *ovector,
2451 int stringcount, const char *stringname,
2452 char *buffer, int buffersize);
2454 int pcre_get_named_substring(const pcre *code,
2455 const char *subject, int *ovector,
2456 int stringcount, const char *stringname,
2457 const char **stringptr);
2459 To extract a substring by name, you first have to find associated num-
2460 ber. For example, for this pattern
2462 (a+)b(?<xxx>\d+)...
2464 the number of the subpattern called "xxx" is 2. If the name is known to
2465 be unique (PCRE_DUPNAMES was not set), you can find the number from the
2466 name by calling pcre_get_stringnumber(). The first argument is the com-
2467 piled pattern, and the second is the name. The yield of the function is
2468 the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if there is no
2469 subpattern of that name.
2471 Given the number, you can extract the substring directly, or use one of
2472 the functions described in the previous section. For convenience, there
2473 are also two functions that do the whole job.
2475 Most of the arguments of pcre_copy_named_substring() and
2476 pcre_get_named_substring() are the same as those for the similarly
2477 named functions that extract by number. As these are described in the
2478 previous section, they are not re-described here. There are just two
2479 differences:
2481 First, instead of a substring number, a substring name is given. Sec-
2482 ond, there is an extra argument, given at the start, which is a pointer
2483 to the compiled pattern. This is needed in order to gain access to the
2484 name-to-number translation table.
2486 These functions call pcre_get_stringnumber(), and if it succeeds, they
2487 then call pcre_copy_substring() or pcre_get_substring(), as appropri-
2488 ate. NOTE: If PCRE_DUPNAMES is set and there are duplicate names, the
2489 behaviour may not be what you want (see the next section).
2491 Warning: If the pattern uses the (?| feature to set up multiple subpat-
2492 terns with the same number, as described in the section on duplicate
2493 subpattern numbers in the pcrepattern page, you cannot use names to
2494 distinguish the different subpatterns, because names are not included
2495 in the compiled code. The matching process uses only numbers. For this
2496 reason, the use of different names for subpatterns of the same number
2497 causes an error at compile time.
2502 int pcre_get_stringtable_entries(const pcre *code,
2503 const char *name, char **first, char **last);
2505 When a pattern is compiled with the PCRE_DUPNAMES option, names for
2506 subpatterns are not required to be unique. (Duplicate names are always
2507 allowed for subpatterns with the same number, created by using the (?|
2508 feature. Indeed, if such subpatterns are named, they are required to
2509 use the same names.)
2511 Normally, patterns with duplicate names are such that in any one match,
2512 only one of the named subpatterns participates. An example is shown in
2513 the pcrepattern documentation.
2515 When duplicates are present, pcre_copy_named_substring() and
2516 pcre_get_named_substring() return the first substring corresponding to
2517 the given name that is set. If none are set, PCRE_ERROR_NOSUBSTRING
2518 (-7) is returned; no data is returned. The pcre_get_stringnumber()
2519 function returns one of the numbers that are associated with the name,
2520 but it is not defined which it is.
2522 If you want to get full details of all captured substrings for a given
2523 name, you must use the pcre_get_stringtable_entries() function. The
2524 first argument is the compiled pattern, and the second is the name. The
2525 third and fourth are pointers to variables which are updated by the
2526 function. After it has run, they point to the first and last entries in
2527 the name-to-number table for the given name. The function itself
2528 returns the length of each entry, or PCRE_ERROR_NOSUBSTRING (-7) if
2529 there are none. The format of the table is described above in the sec-
2530 tion entitled Information about a pattern. Given all the relevant
2531 entries for the name, you can extract each of their numbers, and hence
2532 the captured data, if any.
2537 The traditional matching function uses a similar algorithm to Perl,
2538 which stops when it finds the first match, starting at a given point in
2539 the subject. If you want to find all possible matches, or the longest
2540 possible match, consider using the alternative matching function (see
2541 below) instead. If you cannot use the alternative function, but still
2542 need to find all possible matches, you can kludge it up by making use
2543 of the callout facility, which is described in the pcrecallout documen-
2544 tation.
2546 What you have to do is to insert a callout right at the end of the pat-
2547 tern. When your callout function is called, extract and save the cur-
2548 rent matched substring. Then return 1, which forces pcre_exec() to
2549 backtrack and try other alternatives. Ultimately, when it runs out of
2550 matches, pcre_exec() will yield PCRE_ERROR_NOMATCH.
2555 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
2556 const char *subject, int length, int startoffset,
2557 int options, int *ovector, int ovecsize,
2558 int *workspace, int wscount);
2560 The function pcre_dfa_exec() is called to match a subject string
2561 against a compiled pattern, using a matching algorithm that scans the
2562 subject string just once, and does not backtrack. This has different
2563 characteristics to the normal algorithm, and is not compatible with
2564 Perl. Some of the features of PCRE patterns are not supported. Never-
2565 theless, there are times when this kind of matching can be useful. For
2566 a discussion of the two matching algorithms, and a list of features
2567 that pcre_dfa_exec() does not support, see the pcrematching documenta-
2568 tion.
2570 The arguments for the pcre_dfa_exec() function are the same as for
2571 pcre_exec(), plus two extras. The ovector argument is used in a differ-
2572 ent way, and this is described below. The other common arguments are
2573 used in the same way as for pcre_exec(), so their description is not
2574 repeated here.
2576 The two additional arguments provide workspace for the function. The
2577 workspace vector should contain at least 20 elements. It is used for
2578 keeping track of multiple paths through the pattern tree. More
2579 workspace will be needed for patterns and subjects where there are a
2580 lot of potential matches.
2582 Here is an example of a simple call to pcre_dfa_exec():
2584 int rc;
2585 int ovector[10];
2586 int wspace[20];
2587 rc = pcre_dfa_exec(
2588 re, /* result of pcre_compile() */
2589 NULL, /* we didn't study the pattern */
2590 "some string", /* the subject string */
2591 11, /* the length of the subject string */
2592 0, /* start at offset 0 in the subject */
2593 0, /* default options */
2594 ovector, /* vector of integers for substring information */
2595 10, /* number of elements (NOT size in bytes) */
2596 wspace, /* working space vector */
2597 20); /* number of elements (NOT size in bytes) */
2599 Option bits for pcre_dfa_exec()
2601 The unused bits of the options argument for pcre_dfa_exec() must be
2602 zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NEW-
2606 four of these are exactly the same as for pcre_exec(), so their
2607 description is not repeated here.
2612 These have the same general effect as they do for pcre_exec(), but the
2613 details are slightly different. When PCRE_PARTIAL_HARD is set for
2614 pcre_dfa_exec(), it returns PCRE_ERROR_PARTIAL if the end of the sub-
2615 ject is reached and there is still at least one matching possibility
2616 that requires additional characters. This happens even if some complete
2617 matches have also been found. When PCRE_PARTIAL_SOFT is set, the return
2618 code PCRE_ERROR_NOMATCH is converted into PCRE_ERROR_PARTIAL if the end
2619 of the subject is reached, there have been no complete matches, but
2620 there is still at least one matching possibility. The portion of the
2621 string that was inspected when the longest partial match was found is
2622 set as the first matching string in both cases.
2626 Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to
2627 stop as soon as it has found one match. Because of the way the alterna-
2628 tive algorithm works, this is necessarily the shortest possible match
2629 at the first possible matching point in the subject string.
2633 When pcre_dfa_exec() returns a partial match, it is possible to call it
2634 again, with additional subject characters, and have it continue with
2635 the same match. The PCRE_DFA_RESTART option requests this action; when
2636 it is set, the workspace and wscount options must reference the same
2637 vector as before because data about the match so far is left in them
2638 after a partial match. There is more discussion of this facility in the
2639 pcrepartial documentation.
2641 Successful returns from pcre_dfa_exec()
2643 When pcre_dfa_exec() succeeds, it may have matched more than one sub-
2644 string in the subject. Note, however, that all the matches from one run
2645 of the function start at the same point in the subject. The shorter
2646 matches are all initial substrings of the longer matches. For example,
2647 if the pattern
2649 <.*>
2651 is matched against the string
2653 This is <something> <something else> <something further> no more
2655 the three matched strings are
2657 <something>
2658 <something> <something else>
2659 <something> <something else> <something further>
2661 On success, the yield of the function is a number greater than zero,
2662 which is the number of matched substrings. The substrings themselves
2663 are returned in ovector. Each string uses two elements; the first is
2664 the offset to the start, and the second is the offset to the end. In
2665 fact, all the strings have the same start offset. (Space could have
2666 been saved by giving this only once, but it was decided to retain some
2667 compatibility with the way pcre_exec() returns data, even though the
2668 meaning of the strings is different.)
2670 The strings are returned in reverse order of length; that is, the long-
2671 est matching string is given first. If there were too many matches to
2672 fit into ovector, the yield of the function is zero, and the vector is
2673 filled with the longest matches.
2675 Error returns from pcre_dfa_exec()
2677 The pcre_dfa_exec() function returns a negative number when it fails.
2678 Many of the errors are the same as for pcre_exec(), and these are
2679 described above. There are in addition the following errors that are
2680 specific to pcre_dfa_exec():
2684 This return is given if pcre_dfa_exec() encounters an item in the pat-
2685 tern that it does not support, for instance, the use of \C or a back
2686 reference.
2690 This return is given if pcre_dfa_exec() encounters a condition item
2691 that uses a back reference for the condition, or a test for recursion
2692 in a specific group. These are not supported.
2696 This return is given if pcre_dfa_exec() is called with an extra block
2697 that contains a setting of the match_limit field. This is not supported
2698 (it is meaningless).
2702 This return is given if pcre_dfa_exec() runs out of space in the
2703 workspace vector.
2707 When a recursive subpattern is processed, the matching function calls
2708 itself recursively, using private vectors for ovector and workspace.
2709 This error is given if the output vector is not large enough. This
2710 should be extremely rare, as a vector of size 1000 is used.
2715 pcrebuild(3), pcrecallout(3), pcrecpp(3)(3), pcrematching(3), pcrepar-
2716 tial(3), pcreposix(3), pcreprecompile(3), pcresample(3), pcrestack(3).
2721 Philip Hazel
2722 University Computing Service
2723 Cambridge CB2 3QH, England.
2728 Last updated: 26 March 2010
2729 Copyright (c) 1997-2010 University of Cambridge.
2730 ------------------------------------------------------------------------------
2736 NAME
2737 PCRE - Perl-compatible regular expressions
2742 int (*pcre_callout)(pcre_callout_block *);
2744 PCRE provides a feature called "callout", which is a means of temporar-
2745 ily passing control to the caller of PCRE in the middle of pattern
2746 matching. The caller of PCRE provides an external function by putting
2747 its entry point in the global variable pcre_callout. By default, this
2748 variable contains NULL, which disables all calling out.
2750 Within a regular expression, (?C) indicates the points at which the
2751 external function is to be called. Different callout points can be
2752 identified by putting a number less than 256 after the letter C. The
2753 default value is zero. For example, this pattern has two callout
2754 points:
2756 (?C1)abc(?C2)def
2758 If the PCRE_AUTO_CALLOUT option bit is set when pcre_compile() or
2759 pcre_compile2() is called, PCRE automatically inserts callouts, all
2760 with number 255, before each item in the pattern. For example, if
2761 PCRE_AUTO_CALLOUT is used with the pattern
2763 A(\d{2}|--)
2765 it is processed as if it were
2767 (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255)
2769 Notice that there is a callout before and after each parenthesis and
2770 alternation bar. Automatic callouts can be used for tracking the
2771 progress of pattern matching. The pcretest command has an option that
2772 sets automatic callouts; when it is used, the output indicates how the
2773 pattern is matched. This is useful information when you are trying to
2774 optimize the performance of a particular pattern.
2779 You should be aware that, because of optimizations in the way PCRE
2780 matches patterns by default, callouts sometimes do not happen. For
2781 example, if the pattern is
2783 ab(?C4)cd
2785 PCRE knows that any matching string must contain the letter "d". If the
2786 subject string is "abyz", the lack of "d" means that matching doesn't
2787 ever start, and the callout is never reached. However, with "abyd",
2788 though the result is still no match, the callout is obeyed.
2790 If the pattern is studied, PCRE knows the minimum length of a matching
2791 string, and will immediately give a "no match" return without actually
2792 running a match if the subject is not long enough, or, for unanchored
2793 patterns, if it has been scanned far enough.
2795 You can disable these optimizations by passing the PCRE_NO_START_OPTI-
2796 MIZE option to pcre_exec() or pcre_dfa_exec(). This slows down the
2797 matching process, but does ensure that callouts such as the example
2798 above are obeyed.
2803 During matching, when PCRE reaches a callout point, the external func-
2804 tion defined by pcre_callout is called (if it is set). This applies to
2805 both the pcre_exec() and the pcre_dfa_exec() matching functions. The
2806 only argument to the callout function is a pointer to a pcre_callout
2807 block. This structure contains the following fields:
2809 int version;
2810 int callout_number;
2811 int *offset_vector;
2812 const char *subject;
2813 int subject_length;
2814 int start_match;
2815 int current_position;
2816 int capture_top;
2817 int capture_last;
2818 void *callout_data;
2819 int pattern_position;
2820 int next_item_length;
2822 The version field is an integer containing the version number of the
2823 block format. The initial version was 0; the current version is 1. The
2824 version number will change again in future if additional fields are
2825 added, but the intention is never to remove any of the existing fields.
2827 The callout_number field contains the number of the callout, as com-
2828 piled into the pattern (that is, the number after ?C for manual call-
2829 outs, and 255 for automatically generated callouts).
2831 The offset_vector field is a pointer to the vector of offsets that was
2832 passed by the caller to pcre_exec() or pcre_dfa_exec(). When
2833 pcre_exec() is used, the contents can be inspected in order to extract
2834 substrings that have been matched so far, in the same way as for
2835 extracting substrings after a match has completed. For pcre_dfa_exec()
2836 this field is not useful.
2838 The subject and subject_length fields contain copies of the values that
2839 were passed to pcre_exec().
2841 The start_match field normally contains the offset within the subject
2842 at which the current match attempt started. However, if the escape
2843 sequence \K has been encountered, this value is changed to reflect the
2844 modified starting point. If the pattern is not anchored, the callout
2845 function may be called several times from the same point in the pattern
2846 for different starting points in the subject.
2848 The current_position field contains the offset within the subject of
2849 the current match pointer.
2851 When the pcre_exec() function is used, the capture_top field contains
2852 one more than the number of the highest numbered captured substring so
2853 far. If no substrings have been captured, the value of capture_top is
2854 one. This is always the case when pcre_dfa_exec() is used, because it
2855 does not support captured substrings.
2857 The capture_last field contains the number of the most recently cap-
2858 tured substring. If no substrings have been captured, its value is -1.
2859 This is always the case when pcre_dfa_exec() is used.
2861 The callout_data field contains a value that is passed to pcre_exec()
2862 or pcre_dfa_exec() specifically so that it can be passed back in call-
2863 outs. It is passed in the pcre_callout field of the pcre_extra data
2864 structure. If no such data was passed, the value of callout_data in a
2865 pcre_callout block is NULL. There is a description of the pcre_extra
2866 structure in the pcreapi documentation.
2868 The pattern_position field is present from version 1 of the pcre_call-
2869 out structure. It contains the offset to the next item to be matched in
2870 the pattern string.
2872 The next_item_length field is present from version 1 of the pcre_call-
2873 out structure. It contains the length of the next item to be matched in
2874 the pattern string. When the callout immediately precedes an alterna-
2875 tion bar, a closing parenthesis, or the end of the pattern, the length
2876 is zero. When the callout precedes an opening parenthesis, the length
2877 is that of the entire subpattern.
2879 The pattern_position and next_item_length fields are intended to help
2880 in distinguishing between different automatic callouts, which all have
2881 the same callout number. However, they are set for all callouts.
2886 The external callout function returns an integer to PCRE. If the value
2887 is zero, matching proceeds as normal. If the value is greater than
2888 zero, matching fails at the current point, but the testing of other
2889 matching possibilities goes ahead, just as if a lookahead assertion had
2890 failed. If the value is less than zero, the match is abandoned, and
2891 pcre_exec() or pcre_dfa_exec() returns the negative value.
2893 Negative values should normally be chosen from the set of
2894 PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan-
2895 dard "no match" failure. The error number PCRE_ERROR_CALLOUT is
2896 reserved for use by callout functions; it will never be used by PCRE
2897 itself.
2902 Philip Hazel
2903 University Computing Service
2904 Cambridge CB2 3QH, England.
2909 Last updated: 29 September 2009
2910 Copyright (c) 1997-2009 University of Cambridge.
2911 ------------------------------------------------------------------------------
2917 NAME
2918 PCRE - Perl-compatible regular expressions
2923 This document describes the differences in the ways that PCRE and Perl
2924 handle regular expressions. The differences described here are with
2925 respect to Perl 5.10.
2927 1. PCRE has only a subset of Perl's UTF-8 and Unicode support. Details
2928 of what it does have are given in the section on UTF-8 support in the
2929 main pcre page.
2931 2. PCRE does not allow repeat quantifiers on lookahead assertions. Perl
2932 permits them, but they do not mean what you might think. For example,
2933 (?!a){3} does not assert that the next three characters are not "a". It
2934 just asserts that the next character is not "a" three times.
2936 3. Capturing subpatterns that occur inside negative lookahead asser-
2937 tions are counted, but their entries in the offsets vector are never
2938 set. Perl sets its numerical variables from any such patterns that are
2939 matched before the assertion fails to match something (thereby succeed-
2940 ing), but only if the negative lookahead assertion contains just one
2941 branch.
2943 4. Though binary zero characters are supported in the subject string,
2944 they are not allowed in a pattern string because it is passed as a nor-
2945 mal C string, terminated by zero. The escape sequence \0 can be used in
2946 the pattern to represent a binary zero.
2948 5. The following Perl escape sequences are not supported: \l, \u, \L,
2949 \U, and \N. In fact these are implemented by Perl's general string-han-
2950 dling and are not part of its pattern matching engine. If any of these
2951 are encountered by PCRE, an error is generated.
2953 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE
2954 is built with Unicode character property support. The properties that
2955 can be tested with \p and \P are limited to the general category prop-
2956 erties such as Lu and Nd, script names such as Greek or Han, and the
2957 derived properties Any and L&. PCRE does support the Cs (surrogate)
2958 property, which Perl does not; the Perl documentation says "Because
2959 Perl hides the need for the user to understand the internal representa-
2960 tion of Unicode characters, there is no need to implement the somewhat
2961 messy concept of surrogates."
2963 7. PCRE does support the \Q...\E escape for quoting substrings. Charac-
2964 ters in between are treated as literals. This is slightly different
2965 from Perl in that $ and @ are also handled as literals inside the
2966 quotes. In Perl, they cause variable interpolation (but of course PCRE
2967 does not have variables). Note the following examples:
2969 Pattern PCRE matches Perl matches
2971 \Qabc$xyz\E abc$xyz abc followed by the
2972 contents of $xyz
2973 \Qabc\$xyz\E abc\$xyz abc\$xyz
2974 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
2976 The \Q...\E sequence is recognized both inside and outside character
2977 classes.
2979 8. Fairly obviously, PCRE does not support the (?{code}) and (??{code})
2980 constructions. However, there is support for recursive patterns. This
2981 is not available in Perl 5.8, but it is in Perl 5.10. Also, the PCRE
2982 "callout" feature allows an external function to be called during pat-
2983 tern matching. See the pcrecallout documentation for details.
2985 9. Subpatterns that are called recursively or as "subroutines" are
2986 always treated as atomic groups in PCRE. This is like Python, but
2987 unlike Perl. There is a discussion of an example that explains this in
2988 more detail in the section on recursion differences from Perl in the
2989 pcrepattern page.
2991 10. There are some differences that are concerned with the settings of
2992 captured strings when part of a pattern is repeated. For example,
2993 matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2
2994 unset, but in PCRE it is set to "b".
2996 11. PCRE does support Perl 5.10's backtracking verbs (*ACCEPT),
2997 (*FAIL), (*F), (*COMMIT), (*PRUNE), (*SKIP), and (*THEN), but only in
2998 the forms without an argument. PCRE does not support (*MARK).
3000 12. PCRE's handling of duplicate subpattern numbers and duplicate sub-
3001 pattern names is not as general as Perl's. This is a consequence of the
3002 fact the PCRE works internally just with numbers, using an external ta-
3003 ble to translate between numbers and names. In particular, a pattern
3004 such as (?|(?<a>A)|(?<b)B), where the two capturing parentheses have
3005 the same number but different names, is not supported, and causes an
3006 error at compile time. If it were allowed, it would not be possible to
3007 distinguish which parentheses matched, because both names map to cap-
3008 turing subpattern number 1. To avoid this confusing situation, an error
3009 is given at compile time.
3011 13. PCRE provides some extensions to the Perl regular expression facil-
3012 ities. Perl 5.10 includes new features that are not in earlier ver-
3013 sions of Perl, some of which (such as named parentheses) have been in
3014 PCRE for some time. This list is with respect to Perl 5.10:
3016 (a) Although lookbehind assertions in PCRE must match fixed length
3017 strings, each alternative branch of a lookbehind assertion can match a
3018 different length of string. Perl requires them all to have the same
3019 length.
3021 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $
3022 meta-character matches only at the very end of the string.
3024 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe-
3025 cial meaning is faulted. Otherwise, like Perl, the backslash is quietly
3026 ignored. (Perl can be made to issue a warning.)
3028 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti-
3029 fiers is inverted, that is, by default they are not greedy, but if fol-
3030 lowed by a question mark they are.
3032 (e) PCRE_ANCHORED can be used at matching time to force a pattern to be
3033 tried only at the first matching position in the subject string.
3036 and PCRE_NO_AUTO_CAPTURE options for pcre_exec() have no Perl equiva-
3037 lents.
3039 (g) The \R escape sequence can be restricted to match only CR, LF, or
3040 CRLF by the PCRE_BSR_ANYCRLF option.
3042 (h) The callout facility is PCRE-specific.
3044 (i) The partial matching facility is PCRE-specific.
3046 (j) Patterns compiled by PCRE can be saved and re-used at a later time,
3047 even on different hosts that have the other endianness.
3049 (k) The alternative matching function (pcre_dfa_exec()) matches in a
3050 different way and is not Perl-compatible.
3052 (l) PCRE recognizes some special sequences such as (*CR) at the start
3053 of a pattern that set overall options that cannot be changed within the
3054 pattern.
3059 Philip Hazel
3060 University Computing Service
3061 Cambridge CB2 3QH, England.
3066 Last updated: 04 October 2009
3067 Copyright (c) 1997-2009 University of Cambridge.
3068 ------------------------------------------------------------------------------
3074 NAME
3075 PCRE - Perl-compatible regular expressions
3080 The syntax and semantics of the regular expressions that are supported
3081 by PCRE are described in detail below. There is a quick-reference syn-
3082 tax summary in the pcresyntax page. PCRE tries to match Perl syntax and
3083 semantics as closely as it can. PCRE also supports some alternative
3084 regular expression syntax (which does not conflict with the Perl syn-
3085 tax) in order to provide some compatibility with regular expressions in
3086 Python, .NET, and Oniguruma.
3088 Perl's regular expressions are described in its own documentation, and
3089 regular expressions in general are covered in a number of books, some
3090 of which have copious examples. Jeffrey Friedl's "Mastering Regular
3091 Expressions", published by O'Reilly, covers regular expressions in
3092 great detail. This description of PCRE's regular expressions is
3093 intended as reference material.
3095 The original operation of PCRE was on strings of one-byte characters.
3096 However, there is now also support for UTF-8 character strings. To use
3097 this, PCRE must be built to include UTF-8 support, and you must call
3098 pcre_compile() or pcre_compile2() with the PCRE_UTF8 option. There is
3099 also a special sequence that can be given at the start of a pattern:
3101 (*UTF8)
3103 Starting a pattern with this sequence is equivalent to setting the
3104 PCRE_UTF8 option. This feature is not Perl-compatible. How setting
3105 UTF-8 mode affects pattern matching is mentioned in several places
3106 below. There is also a summary of UTF-8 features in the section on
3107 UTF-8 support in the main pcre page.
3109 The remainder of this document discusses the patterns that are sup-
3110 ported by PCRE when its main matching function, pcre_exec(), is used.
3111 From release 6.0, PCRE offers a second matching function,
3112 pcre_dfa_exec(), which matches using a different algorithm that is not
3113 Perl-compatible. Some of the features discussed below are not available
3114 when pcre_dfa_exec() is used. The advantages and disadvantages of the
3115 alternative function, and how it differs from the normal function, are
3116 discussed in the pcrematching page.
3121 PCRE supports five different conventions for indicating line breaks in
3122 strings: a single CR (carriage return) character, a single LF (line-
3123 feed) character, the two-character sequence CRLF, any of the three pre-
3124 ceding, or any Unicode newline sequence. The pcreapi page has further
3125 discussion about newlines, and shows how to set the newline convention
3126 in the options arguments for the compiling and matching functions.
3128 It is also possible to specify a newline convention by starting a pat-
3129 tern string with one of the following five sequences:
3131 (*CR) carriage return
3132 (*LF) linefeed
3133 (*CRLF) carriage return, followed by linefeed
3134 (*ANYCRLF) any of the three above
3135 (*ANY) all Unicode newline sequences
3137 These override the default and the options given to pcre_compile() or
3138 pcre_compile2(). For example, on a Unix system where LF is the default
3139 newline sequence, the pattern
3141 (*CR)a.b
3143 changes the convention to CR. That pattern matches "a\nb" because LF is
3144 no longer a newline. Note that these special settings, which are not
3145 Perl-compatible, are recognized only at the very start of a pattern,
3146 and that they must be in upper case. If more than one of them is
3147 present, the last one is used.
3149 The newline convention does not affect what the \R escape sequence
3150 matches. By default, this is any Unicode newline sequence, for Perl
3151 compatibility. However, this can be changed; see the description of \R
3152 in the section entitled "Newline sequences" below. A change of \R set-
3153 ting can be combined with a change of newline convention.
3158 A regular expression is a pattern that is matched against a subject
3159 string from left to right. Most characters stand for themselves in a
3160 pattern, and match the corresponding characters in the subject. As a
3161 trivial example, the pattern
3163 The quick brown fox
3165 matches a portion of a subject string that is identical to itself. When
3166 caseless matching is specified (the PCRE_CASELESS option), letters are
3167 matched independently of case. In UTF-8 mode, PCRE always understands
3168 the concept of case for characters whose values are less than 128, so
3169 caseless matching is always possible. For characters with higher val-
3170 ues, the concept of case is supported if PCRE is compiled with Unicode
3171 property support, but not otherwise. If you want to use caseless
3172 matching for characters 128 and above, you must ensure that PCRE is
3173 compiled with Unicode property support as well as with UTF-8 support.
3175 The power of regular expressions comes from the ability to include
3176 alternatives and repetitions in the pattern. These are encoded in the
3177 pattern by the use of metacharacters, which do not stand for themselves
3178 but instead are interpreted in some special way.
3180 There are two different sets of metacharacters: those that are recog-
3181 nized anywhere in the pattern except within square brackets, and those
3182 that are recognized within square brackets. Outside square brackets,
3183 the metacharacters are as follows:
3185 \ general escape character with several uses
3186 ^ assert start of string (or line, in multiline mode)
3187 $ assert end of string (or line, in multiline mode)
3188 . match any character except newline (by default)
3189 [ start character class definition
3190 | start of alternative branch
3191 ( start subpattern
3192 ) end subpattern
3193 ? extends the meaning of (
3194 also 0 or 1 quantifier
3195 also quantifier minimizer
3196 * 0 or more quantifier
3197 + 1 or more quantifier
3198 also "possessive quantifier"
3199 { start min/max quantifier
3201 Part of a pattern that is in square brackets is called a "character
3202 class". In a character class the only metacharacters are:
3204 \ general escape character
3205 ^ negate the class, but only if the first character
3206 - indicates character range
3207 [ POSIX character class (only if followed by POSIX
3208 syntax)
3209 ] terminates the character class
3211 The following sections describe the use of each of the metacharacters.
3216 The backslash character has several uses. Firstly, if it is followed by
3217 a non-alphanumeric character, it takes away any special meaning that
3218 character may have. This use of backslash as an escape character
3219 applies both inside and outside character classes.
3221 For example, if you want to match a * character, you write \* in the
3222 pattern. This escaping action applies whether or not the following
3223 character would otherwise be interpreted as a metacharacter, so it is
3224 always safe to precede a non-alphanumeric with backslash to specify
3225 that it stands for itself. In particular, if you want to match a back-
3226 slash, you write \\.
3228 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in
3229 the pattern (other than in a character class) and characters between a
3230 # outside a character class and the next newline are ignored. An escap-
3231 ing backslash can be used to include a whitespace or # character as
3232 part of the pattern.
3234 If you want to remove the special meaning from a sequence of charac-
3235 ters, you can do so by putting them between \Q and \E. This is differ-
3236 ent from Perl in that $ and @ are handled as literals in \Q...\E
3237 sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
3238 tion. Note the following examples:
3240 Pattern PCRE matches Perl matches
3242 \Qabc$xyz\E abc$xyz abc followed by the
3243 contents of $xyz
3244 \Qabc\$xyz\E abc\$xyz abc\$xyz
3245 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
3247 The \Q...\E sequence is recognized both inside and outside character
3248 classes.
3250 Non-printing characters
3252 A second use of backslash provides a way of encoding non-printing char-
3253 acters in patterns in a visible manner. There is no restriction on the
3254 appearance of non-printing characters, apart from the binary zero that
3255 terminates a pattern, but when a pattern is being prepared by text
3256 editing, it is often easier to use one of the following escape
3257 sequences than the binary character it represents:
3259 \a alarm, that is, the BEL character (hex 07)
3260 \cx "control-x", where x is any character
3261 \e escape (hex 1B)
3262 \f formfeed (hex 0C)
3263 \n linefeed (hex 0A)
3264 \r carriage return (hex 0D)
3265 \t tab (hex 09)
3266 \ddd character with octal code ddd, or back reference
3267 \xhh character with hex code hh
3268 \x{hhh..} character with hex code hhh..
3270 The precise effect of \cx is as follows: if x is a lower case letter,
3271 it is converted to upper case. Then bit 6 of the character (hex 40) is
3272 inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c;
3273 becomes hex 7B.
3275 After \x, from zero to two hexadecimal digits are read (letters can be
3276 in upper or lower case). Any number of hexadecimal digits may appear
3277 between \x{ and }, but the value of the character code must be less
3278 than 256 in non-UTF-8 mode, and less than 2**31 in UTF-8 mode. That is,
3279 the maximum value in hexadecimal is 7FFFFFFF. Note that this is bigger
3280 than the largest Unicode code point, which is 10FFFF.
3282 If characters other than hexadecimal digits appear between \x{ and },
3283 or if there is no terminating }, this form of escape is not recognized.
3284 Instead, the initial \x will be interpreted as a basic hexadecimal
3285 escape, with no following digits, giving a character whose value is
3286 zero.
3288 Characters whose value is less than 256 can be defined by either of the
3289 two syntaxes for \x. There is no difference in the way they are han-
3290 dled. For example, \xdc is exactly the same as \x{dc}.
3292 After \0 up to two further octal digits are read. If there are fewer
3293 than two digits, just those that are present are used. Thus the
3294 sequence \0\x\07 specifies two binary zeros followed by a BEL character
3295 (code value 7). Make sure you supply two digits after the initial zero
3296 if the pattern character that follows is itself an octal digit.
3298 The handling of a backslash followed by a digit other than 0 is compli-
3299 cated. Outside a character class, PCRE reads it and any following dig-
3300 its as a decimal number. If the number is less than 10, or if there
3301 have been at least that many previous capturing left parentheses in the
3302 expression, the entire sequence is taken as a back reference. A
3303 description of how this works is given later, following the discussion
3304 of parenthesized subpatterns.
3306 Inside a character class, or if the decimal number is greater than 9
3307 and there have not been that many capturing subpatterns, PCRE re-reads
3308 up to three octal digits following the backslash, and uses them to gen-
3309 erate a data character. Any subsequent digits stand for themselves. In
3310 non-UTF-8 mode, the value of a character specified in octal must be
3311 less than \400. In UTF-8 mode, values up to \777 are permitted. For
3312 example:
3314 \040 is another way of writing a space
3315 \40 is the same, provided there are fewer than 40
3316 previous capturing subpatterns
3317 \7 is always a back reference
3318 \11 might be a back reference, or another way of
3319 writing a tab
3320 \011 is always a tab
3321 \0113 is a tab followed by the character "3"
3322 \113 might be a back reference, otherwise the
3323 character with octal code 113
3324 \377 might be a back reference, otherwise
3325 the byte consisting entirely of 1 bits
3326 \81 is either a back reference, or a binary zero
3327 followed by the two characters "8" and "1"
3329 Note that octal values of 100 or greater must not be introduced by a
3330 leading zero, because no more than three octal digits are ever read.
3332 All the sequences that define a single character value can be used both
3333 inside and outside character classes. In addition, inside a character
3334 class, the sequence \b is interpreted as the backspace character (hex
3335 08), and the sequences \R and \X are interpreted as the characters "R"
3336 and "X", respectively. Outside a character class, these sequences have
3337 different meanings (see below).
3339 Absolute and relative back references
3341 The sequence \g followed by an unsigned or a negative number, option-
3342 ally enclosed in braces, is an absolute or relative back reference. A
3343 named back reference can be coded as \g{name}. Back references are dis-
3344 cussed later, following the discussion of parenthesized subpatterns.
3346 Absolute and relative subroutine calls
3348 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
3349 name or a number enclosed either in angle brackets or single quotes, is
3350 an alternative syntax for referencing a subpattern as a "subroutine".
3351 Details are discussed later. Note that \g{...} (Perl syntax) and
3352 \g<...> (Oniguruma syntax) are not synonymous. The former is a back
3353 reference; the latter is a subroutine call.
3355 Generic character types
3357 Another use of backslash is for specifying generic character types. The
3358 following are always recognized:
3360 \d any decimal digit
3361 \D any character that is not a decimal digit
3362 \h any horizontal whitespace character
3363 \H any character that is not a horizontal whitespace character
3364 \s any whitespace character
3365 \S any character that is not a whitespace character
3366 \v any vertical whitespace character
3367 \V any character that is not a vertical whitespace character
3368 \w any "word" character
3369 \W any "non-word" character
3371 Each pair of escape sequences partitions the complete set of characters
3372 into two disjoint sets. Any given character matches one, and only one,
3373 of each pair.
3375 These character type sequences can appear both inside and outside char-
3376 acter classes. They each match one character of the appropriate type.
3377 If the current matching point is at the end of the subject string, all
3378 of them fail, since there is no character to match.
3380 For compatibility with Perl, \s does not match the VT character (code
3381 11). This makes it different from the the POSIX "space" class. The \s
3382 characters are HT (9), LF (10), FF (12), CR (13), and space (32). If
3383 "use locale;" is included in a Perl script, \s may match the VT charac-
3384 ter. In PCRE, it never does.
3386 In UTF-8 mode, characters with values greater than 128 never match \d,
3387 \s, or \w, and always match \D, \S, and \W. This is true even when Uni-
3388 code character property support is available. These sequences retain
3389 their original meanings from before UTF-8 support was available, mainly
3390 for efficiency reasons. Note that this also affects \b, because it is
3391 defined in terms of \w and \W.
3393 The sequences \h, \H, \v, and \V are Perl 5.10 features. In contrast to
3394 the other sequences, these do match certain high-valued codepoints in
3395 UTF-8 mode. The horizontal space characters are:
3397 U+0009 Horizontal tab
3398 U+0020 Space
3399 U+00A0 Non-break space
3400 U+1680 Ogham space mark
3401 U+180E Mongolian vowel separator
3402 U+2000 En quad
3403 U+2001 Em quad
3404 U+2002 En space
3405 U+2003 Em space
3406 U+2004 Three-per-em space
3407 U+2005 Four-per-em space
3408 U+2006 Six-per-em space
3409 U+2007 Figure space
3410 U+2008 Punctuation space
3411 U+2009 Thin space
3412 U+200A Hair space
3413 U+202F Narrow no-break space
3414 U+205F Medium mathematical space
3415 U+3000 Ideographic space
3417 The vertical space characters are:
3419 U+000A Linefeed
3420 U+000B Vertical tab
3421 U+000C Formfeed
3422 U+000D Carriage return
3423 U+0085 Next line
3424 U+2028 Line separator
3425 U+2029 Paragraph separator
3427 A "word" character is an underscore or any character less than 256 that
3428 is a letter or digit. The definition of letters and digits is con-
3429 trolled by PCRE's low-valued character tables, and may vary if locale-
3430 specific matching is taking place (see "Locale support" in the pcreapi
3431 page). For example, in a French locale such as "fr_FR" in Unix-like
3432 systems, or "french" in Windows, some character codes greater than 128
3433 are used for accented letters, and these are matched by \w. The use of
3434 locales with Unicode is discouraged.
3436 Newline sequences
3438 Outside a character class, by default, the escape sequence \R matches
3439 any Unicode newline sequence. This is a Perl 5.10 feature. In non-UTF-8
3440 mode \R is equivalent to the following:
3442 (?>\r\n|\n|\x0b|\f|\r|\x85)
3444 This is an example of an "atomic group", details of which are given
3445 below. This particular group matches either the two-character sequence
3446 CR followed by LF, or one of the single characters LF (linefeed,
3447 U+000A), VT (vertical tab, U+000B), FF (formfeed, U+000C), CR (carriage
3448 return, U+000D), or NEL (next line, U+0085). The two-character sequence
3449 is treated as a single unit that cannot be split.
3451 In UTF-8 mode, two additional characters whose codepoints are greater
3452 than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa-
3453 rator, U+2029). Unicode character property support is not needed for
3454 these characters to be recognized.
3456 It is possible to restrict \R to match only CR, LF, or CRLF (instead of
3457 the complete set of Unicode line endings) by setting the option
3458 PCRE_BSR_ANYCRLF either at compile time or when the pattern is matched.
3459 (BSR is an abbrevation for "backslash R".) This can be made the default
3460 when PCRE is built; if this is the case, the other behaviour can be
3461 requested via the PCRE_BSR_UNICODE option. It is also possible to
3462 specify these settings by starting a pattern string with one of the
3463 following sequences:
3465 (*BSR_ANYCRLF) CR, LF, or CRLF only
3466 (*BSR_UNICODE) any Unicode newline sequence
3468 These override the default and the options given to pcre_compile() or
3469 pcre_compile2(), but they can be overridden by options given to
3470 pcre_exec() or pcre_dfa_exec(). Note that these special settings, which
3471 are not Perl-compatible, are recognized only at the very start of a
3472 pattern, and that they must be in upper case. If more than one of them
3473 is present, the last one is used. They can be combined with a change of
3474 newline convention, for example, a pattern can start with:
3478 Inside a character class, \R matches the letter "R".
3480 Unicode character properties
3482 When PCRE is built with Unicode character property support, three addi-
3483 tional escape sequences that match characters with specific properties
3484 are available. When not in UTF-8 mode, these sequences are of course
3485 limited to testing characters whose codepoints are less than 256, but
3486 they do work in this mode. The extra escape sequences are:
3488 \p{xx} a character with the xx property
3489 \P{xx} a character without the xx property
3490 \X an extended Unicode sequence
3492 The property names represented by xx above are limited to the Unicode
3493 script names, the general category properties, and "Any", which matches
3494 any character (including newline). Other properties such as "InMusical-
3495 Symbols" are not currently supported by PCRE. Note that \P{Any} does
3496 not match any characters, so always causes a match failure.
3498 Sets of Unicode characters are defined as belonging to certain scripts.
3499 A character from one of these sets can be matched using a script name.
3500 For example:
3502 \p{Greek}
3503 \P{Han}
3505 Those that are not part of an identified script are lumped together as
3506 "Common". The current list of scripts is:
3508 Arabic, Armenian, Avestan, Balinese, Bamum, Bengali, Bopomofo, Braille,
3509 Buginese, Buhid, Canadian_Aboriginal, Carian, Cham, Cherokee, Common,
3510 Coptic, Cuneiform, Cypriot, Cyrillic, Deseret, Devanagari, Egyp-
3511 tian_Hieroglyphs, Ethiopic, Georgian, Glagolitic, Gothic, Greek,
3512 Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana, Impe-
3513 rial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscriptional_Parthian,
3514 Javanese, Kaithi, Kannada, Katakana, Kayah_Li, Kharoshthi, Khmer, Lao,
3515 Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian, Lydian, Malayalam,
3516 Meetei_Mayek, Mongolian, Myanmar, New_Tai_Lue, Nko, Ogham, Old_Italic,
3517 Old_Persian, Old_South_Arabian, Old_Turkic, Ol_Chiki, Oriya, Osmanya,
3518 Phags_Pa, Phoenician, Rejang, Runic, Samaritan, Saurashtra, Shavian,
3519 Sinhala, Sundanese, Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le,
3520 Tai_Tham, Tai_Viet, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh,
3521 Ugaritic, Vai, Yi.
3523 Each character has exactly one general category property, specified by
3524 a two-letter abbreviation. For compatibility with Perl, negation can be
3525 specified by including a circumflex between the opening brace and the
3526 property name. For example, \p{^Lu} is the same as \P{Lu}.
3528 If only one letter is specified with \p or \P, it includes all the gen-
3529 eral category properties that start with that letter. In this case, in
3530 the absence of negation, the curly brackets in the escape sequence are
3531 optional; these two examples have the same effect:
3533 \p{L}
3534 \pL
3536 The following general category property codes are supported:
3538 C Other
3539 Cc Control
3540 Cf Format
3541 Cn Unassigned
3542 Co Private use
3543 Cs Surrogate
3545 L Letter
3546 Ll Lower case letter
3547 Lm Modifier letter
3548 Lo Other letter
3549 Lt Title case letter
3550 Lu Upper case letter
3552 M Mark
3553 Mc Spacing mark
3554 Me Enclosing mark
3555 Mn Non-spacing mark
3557 N Number
3558 Nd Decimal number
3559 Nl Letter number
3560 No Other number
3562 P Punctuation
3563 Pc Connector punctuation
3564 Pd Dash punctuation
3565 Pe Close punctuation
3566 Pf Final punctuation
3567 Pi Initial punctuation
3568 Po Other punctuation
3569 Ps Open punctuation
3571 S Symbol
3572 Sc Currency symbol
3573 Sk Modifier symbol
3574 Sm Mathematical symbol
3575 So Other symbol
3577 Z Separator
3578 Zl Line separator
3579 Zp Paragraph separator
3580 Zs Space separator
3582 The special property L& is also supported: it matches a character that
3583 has the Lu, Ll, or Lt property, in other words, a letter that is not
3584 classified as a modifier or "other".
3586 The Cs (Surrogate) property applies only to characters in the range
3587 U+D800 to U+DFFF. Such characters are not valid in UTF-8 strings (see
3588 RFC 3629) and so cannot be tested by PCRE, unless UTF-8 validity check-
3589 ing has been turned off (see the discussion of PCRE_NO_UTF8_CHECK in
3590 the pcreapi page). Perl does not support the Cs property.
3592 The long synonyms for property names that Perl supports (such as
3593 \p{Letter}) are not supported by PCRE, nor is it permitted to prefix
3594 any of these properties with "Is".
3596 No character that is in the Unicode table has the Cn (unassigned) prop-
3597 erty. Instead, this property is assumed for any code point that is not
3598 in the Unicode table.
3600 Specifying caseless matching does not affect these escape sequences.
3601 For example, \p{Lu} always matches only upper case letters.
3603 The \X escape matches any number of Unicode characters that form an
3604 extended Unicode sequence. \X is equivalent to
3606 (?>\PM\pM*)
3608 That is, it matches a character without the "mark" property, followed
3609 by zero or more characters with the "mark" property, and treats the
3610 sequence as an atomic group (see below). Characters with the "mark"
3611 property are typically accents that affect the preceding character.
3612 None of them have codepoints less than 256, so in non-UTF-8 mode \X
3613 matches any one character.
3615 Matching characters by Unicode property is not fast, because PCRE has
3616 to search a structure that contains data for over fifteen thousand
3617 characters. That is why the traditional escape sequences such as \d and
3618 \w do not use Unicode properties in PCRE.
3620 Resetting the match start
3622 The escape sequence \K, which is a Perl 5.10 feature, causes any previ-
3623 ously matched characters not to be included in the final matched
3624 sequence. For example, the pattern:
3626 foo\Kbar
3628 matches "foobar", but reports that it has matched "bar". This feature
3629 is similar to a lookbehind assertion (described below). However, in
3630 this case, the part of the subject before the real match does not have
3631 to be of fixed length, as lookbehind assertions do. The use of \K does
3632 not interfere with the setting of captured substrings. For example,
3633 when the pattern
3635 (foo)\Kbar
3637 matches "foobar", the first substring is still set to "foo".
3639 Perl documents that the use of \K within assertions is "not well
3640 defined". In PCRE, \K is acted upon when it occurs inside positive
3641 assertions, but is ignored in negative assertions.
3643 Simple assertions
3645 The final use of backslash is for certain simple assertions. An asser-
3646 tion specifies a condition that has to be met at a particular point in
3647 a match, without consuming any characters from the subject string. The
3648 use of subpatterns for more complicated assertions is described below.
3649 The backslashed assertions are:
3651 \b matches at a word boundary
3652 \B matches when not at a word boundary
3653 \A matches at the start of the subject
3654 \Z matches at the end of the subject
3655 also matches before a newline at the end of the subject
3656 \z matches only at the end of the subject
3657 \G matches at the first matching position in the subject
3659 These assertions may not appear in character classes (but note that \b
3660 has a different meaning, namely the backspace character, inside a char-
3661 acter class).
3663 A word boundary is a position in the subject string where the current
3664 character and the previous character do not both match \w or \W (i.e.
3665 one matches \w and the other matches \W), or the start or end of the
3666 string if the first or last character matches \w, respectively. Neither
3667 PCRE nor Perl has a separte "start of word" or "end of word" metase-
3668 quence. However, whatever follows \b normally determines which it is.
3669 For example, the fragment \ba matches "a" at the start of a word.
3671 The \A, \Z, and \z assertions differ from the traditional circumflex
3672 and dollar (described in the next section) in that they only ever match
3673 at the very start and end of the subject string, whatever options are
3674 set. Thus, they are independent of multiline mode. These three asser-
3675 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
3676 affect only the behaviour of the circumflex and dollar metacharacters.
3677 However, if the startoffset argument of pcre_exec() is non-zero, indi-
3678 cating that matching is to start at a point other than the beginning of
3679 the subject, \A can never match. The difference between \Z and \z is
3680 that \Z matches before a newline at the end of the string as well as at
3681 the very end, whereas \z matches only at the end.
3683 The \G assertion is true only when the current matching position is at
3684 the start point of the match, as specified by the startoffset argument
3685 of pcre_exec(). It differs from \A when the value of startoffset is
3686 non-zero. By calling pcre_exec() multiple times with appropriate argu-
3687 ments, you can mimic Perl's /g option, and it is in this kind of imple-
3688 mentation where \G can be useful.
3690 Note, however, that PCRE's interpretation of \G, as the start of the
3691 current match, is subtly different from Perl's, which defines it as the
3692 end of the previous match. In Perl, these can be different when the
3693 previously matched string was empty. Because PCRE does just one match
3694 at a time, it cannot reproduce this behaviour.
3696 If all the alternatives of a pattern begin with \G, the expression is
3697 anchored to the starting match position, and the "anchored" flag is set
3698 in the compiled regular expression.
3703 Outside a character class, in the default matching mode, the circumflex
3704 character is an assertion that is true only if the current matching
3705 point is at the start of the subject string. If the startoffset argu-
3706 ment of pcre_exec() is non-zero, circumflex can never match if the
3707 PCRE_MULTILINE option is unset. Inside a character class, circumflex
3708 has an entirely different meaning (see below).
3710 Circumflex need not be the first character of the pattern if a number
3711 of alternatives are involved, but it should be the first thing in each
3712 alternative in which it appears if the pattern is ever to match that
3713 branch. If all possible alternatives start with a circumflex, that is,
3714 if the pattern is constrained to match only at the start of the sub-
3715 ject, it is said to be an "anchored" pattern. (There are also other
3716 constructs that can cause a pattern to be anchored.)
3718 A dollar character is an assertion that is true only if the current
3719 matching point is at the end of the subject string, or immediately
3720 before a newline at the end of the string (by default). Dollar need not
3721 be the last character of the pattern if a number of alternatives are
3722 involved, but it should be the last item in any branch in which it
3723 appears. Dollar has no special meaning in a character class.
3725 The meaning of dollar can be changed so that it matches only at the
3726 very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
3727 compile time. This does not affect the \Z assertion.
3729 The meanings of the circumflex and dollar characters are changed if the
3730 PCRE_MULTILINE option is set. When this is the case, a circumflex
3731 matches immediately after internal newlines as well as at the start of
3732 the subject string. It does not match after a newline that ends the
3733 string. A dollar matches before any newlines in the string, as well as
3734 at the very end, when PCRE_MULTILINE is set. When newline is specified
3735 as the two-character sequence CRLF, isolated CR and LF characters do
3736 not indicate newlines.
3738 For example, the pattern /^abc$/ matches the subject string "def\nabc"
3739 (where \n represents a newline) in multiline mode, but not otherwise.
3740 Consequently, patterns that are anchored in single line mode because
3741 all branches start with ^ are not anchored in multiline mode, and a
3742 match for circumflex is possible when the startoffset argument of
3743 pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
3744 PCRE_MULTILINE is set.
3746 Note that the sequences \A, \Z, and \z can be used to match the start
3747 and end of the subject in both modes, and if all branches of a pattern
3748 start with \A it is always anchored, whether or not PCRE_MULTILINE is
3749 set.
3754 Outside a character class, a dot in the pattern matches any one charac-
3755 ter in the subject string except (by default) a character that signi-
3756 fies the end of a line. In UTF-8 mode, the matched character may be
3757 more than one byte long.
3759 When a line ending is defined as a single character, dot never matches
3760 that character; when the two-character sequence CRLF is used, dot does
3761 not match CR if it is immediately followed by LF, but otherwise it
3762 matches all characters (including isolated CRs and LFs). When any Uni-
3763 code line endings are being recognized, dot does not match CR or LF or
3764 any of the other line ending characters.
3766 The behaviour of dot with regard to newlines can be changed. If the
3767 PCRE_DOTALL option is set, a dot matches any one character, without
3768 exception. If the two-character sequence CRLF is present in the subject
3769 string, it takes two dots to match it.
3771 The handling of dot is entirely independent of the handling of circum-
3772 flex and dollar, the only relationship being that they both involve
3773 newlines. Dot has no special meaning in a character class.
3778 Outside a character class, the escape sequence \C matches any one byte,
3779 both in and out of UTF-8 mode. Unlike a dot, it always matches any
3780 line-ending characters. The feature is provided in Perl in order to
3781 match individual bytes in UTF-8 mode. Because it breaks up UTF-8 char-
3782 acters into individual bytes, what remains in the string may be a mal-
3783 formed UTF-8 string. For this reason, the \C escape sequence is best
3784 avoided.
3786 PCRE does not allow \C to appear in lookbehind assertions (described
3787 below), because in UTF-8 mode this would make it impossible to calcu-
3788 late the length of the lookbehind.
3793 An opening square bracket introduces a character class, terminated by a
3794 closing square bracket. A closing square bracket on its own is not spe-
3795 cial by default. However, if the PCRE_JAVASCRIPT_COMPAT option is set,
3796 a lone closing square bracket causes a compile-time error. If a closing
3797 square bracket is required as a member of the class, it should be the
3798 first data character in the class (after an initial circumflex, if
3799 present) or escaped with a backslash.
3801 A character class matches a single character in the subject. In UTF-8
3802 mode, the character may be more than one byte long. A matched character
3803 must be in the set of characters defined by the class, unless the first
3804 character in the class definition is a circumflex, in which case the
3805 subject character must not be in the set defined by the class. If a
3806 circumflex is actually required as a member of the class, ensure it is
3807 not the first character, or escape it with a backslash.
3809 For example, the character class [aeiou] matches any lower case vowel,
3810 while [^aeiou] matches any character that is not a lower case vowel.
3811 Note that a circumflex is just a convenient notation for specifying the
3812 characters that are in the class by enumerating those that are not. A
3813 class that starts with a circumflex is not an assertion; it still con-
3814 sumes a character from the subject string, and therefore it fails if
3815 the current pointer is at the end of the string.
3817 In UTF-8 mode, characters with values greater than 255 can be included
3818 in a class as a literal string of bytes, or by using the \x{ escaping
3819 mechanism.
3821 When caseless matching is set, any letters in a class represent both
3822 their upper case and lower case versions, so for example, a caseless
3823 [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
3824 match "A", whereas a caseful version would. In UTF-8 mode, PCRE always
3825 understands the concept of case for characters whose values are less
3826 than 128, so caseless matching is always possible. For characters with
3827 higher values, the concept of case is supported if PCRE is compiled
3828 with Unicode property support, but not otherwise. If you want to use
3829 caseless matching in UTF8-mode for characters 128 and above, you must
3830 ensure that PCRE is compiled with Unicode property support as well as
3831 with UTF-8 support.
3833 Characters that might indicate line breaks are never treated in any
3834 special way when matching character classes, whatever line-ending
3835 sequence is in use, and whatever setting of the PCRE_DOTALL and
3836 PCRE_MULTILINE options is used. A class such as [^a] always matches one
3837 of these characters.
3839 The minus (hyphen) character can be used to specify a range of charac-
3840 ters in a character class. For example, [d-m] matches any letter
3841 between d and m, inclusive. If a minus character is required in a
3842 class, it must be escaped with a backslash or appear in a position
3843 where it cannot be interpreted as indicating a range, typically as the
3844 first or last character in the class.
3846 It is not possible to have the literal character "]" as the end charac-
3847 ter of a range. A pattern such as [W-]46] is interpreted as a class of
3848 two characters ("W" and "-") followed by a literal string "46]", so it
3849 would match "W46]" or "-46]". However, if the "]" is escaped with a
3850 backslash it is interpreted as the end of range, so [W-\]46] is inter-
3851 preted as a class containing a range followed by two other characters.
3852 The octal or hexadecimal representation of "]" can also be used to end
3853 a range.
3855 Ranges operate in the collating sequence of character values. They can
3856 also be used for characters specified numerically, for example
3857 [\000-\037]. In UTF-8 mode, ranges can include characters whose values
3858 are greater than 255, for example [\x{100}-\x{2ff}].
3860 If a range that includes letters is used when caseless matching is set,
3861 it matches the letters in either case. For example, [W-c] is equivalent
3862 to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if
3863 character tables for a French locale are in use, [\xc8-\xcb] matches
3864 accented E characters in both cases. In UTF-8 mode, PCRE supports the
3865 concept of case for characters with values greater than 128 only when
3866 it is compiled with Unicode property support.
3868 The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear
3869 in a character class, and add the characters that they match to the
3870 class. For example, [\dABCDEF] matches any hexadecimal digit. A circum-
3871 flex can conveniently be used with the upper case character types to
3872 specify a more restricted set of characters than the matching lower
3873 case type. For example, the class [^\W_] matches any letter or digit,
3874 but not underscore.
3876 The only metacharacters that are recognized in character classes are
3877 backslash, hyphen (only where it can be interpreted as specifying a
3878 range), circumflex (only at the start), opening square bracket (only
3879 when it can be interpreted as introducing a POSIX class name - see the
3880 next section), and the terminating closing square bracket. However,
3881 escaping other non-alphanumeric characters does no harm.
3886 Perl supports the POSIX notation for character classes. This uses names
3887 enclosed by [: and :] within the enclosing square brackets. PCRE also
3888 supports this notation. For example,
3890 [01[:alpha:]%]
3892 matches "0", "1", any alphabetic character, or "%". The supported class
3893 names are
3895 alnum letters and digits
3896 alpha letters
3897 ascii character codes 0 - 127
3898 blank space or tab only
3899 cntrl control characters
3900 digit decimal digits (same as \d)
3901 graph printing characters, excluding space
3902 lower lower case letters
3903 print printing characters, including space
3904 punct printing characters, excluding letters and digits
3905 space white space (not quite the same as \s)
3906 upper upper case letters
3907 word "word" characters (same as \w)
3908 xdigit hexadecimal digits
3910 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
3911 and space (32). Notice that this list includes the VT character (code
3912 11). This makes "space" different to \s, which does not include VT (for
3913 Perl compatibility).
3915 The name "word" is a Perl extension, and "blank" is a GNU extension
3916 from Perl 5.8. Another Perl extension is negation, which is indicated
3917 by a ^ character after the colon. For example,
3919 [12[:^digit:]]
3921 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
3922 POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
3923 these are not supported, and an error is given if they are encountered.
3925 In UTF-8 mode, characters with values greater than 128 do not match any
3926 of the POSIX character classes.
3931 Vertical bar characters are used to separate alternative patterns. For
3932 example, the pattern
3934 gilbert|sullivan
3936 matches either "gilbert" or "sullivan". Any number of alternatives may
3937 appear, and an empty alternative is permitted (matching the empty
3938 string). The matching process tries each alternative in turn, from left
3939 to right, and the first one that succeeds is used. If the alternatives
3940 are within a subpattern (defined below), "succeeds" means matching the
3941 rest of the main pattern as well as the alternative in the subpattern.
3946 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
3947 PCRE_EXTENDED options (which are Perl-compatible) can be changed from
3948 within the pattern by a sequence of Perl option letters enclosed
3949 between "(?" and ")". The option letters are
3951 i for PCRE_CASELESS
3953 s for PCRE_DOTALL
3954 x for PCRE_EXTENDED
3956 For example, (?im) sets caseless, multiline matching. It is also possi-
3957 ble to unset these options by preceding the letter with a hyphen, and a
3958 combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
3960 is also permitted. If a letter appears both before and after the
3961 hyphen, the option is unset.
3964 can be changed in the same way as the Perl-compatible options by using
3965 the characters J, U and X respectively.
3967 When one of these option changes occurs at top level (that is, not
3968 inside subpattern parentheses), the change applies to the remainder of
3969 the pattern that follows. If the change is placed right at the start of
3970 a pattern, PCRE extracts it into the global options (and it will there-
3971 fore show up in data extracted by the pcre_fullinfo() function).
3973 An option change within a subpattern (see below for a description of
3974 subpatterns) affects only that part of the current pattern that follows
3975 it, so
3977 (a(?i)b)c
3979 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
3980 used). By this means, options can be made to have different settings
3981 in different parts of the pattern. Any changes made in one alternative
3982 do carry on into subsequent branches within the same subpattern. For
3983 example,
3985 (a(?i)b|c)
3987 matches "ab", "aB", "c", and "C", even though when matching "C" the
3988 first branch is abandoned before the option setting. This is because
3989 the effects of option settings happen at compile time. There would be
3990 some very weird behaviour otherwise.
3992 Note: There are other PCRE-specific options that can be set by the
3993 application when the compile or match functions are called. In some
3994 cases the pattern can contain special leading sequences such as (*CRLF)
3995 to override what the application has set or what has been defaulted.
3996 Details are given in the section entitled "Newline sequences" above.
3997 There is also the (*UTF8) leading sequence that can be used to set
3998 UTF-8 mode; this is equivalent to setting the PCRE_UTF8 option.
4003 Subpatterns are delimited by parentheses (round brackets), which can be
4004 nested. Turning part of a pattern into a subpattern does two things:
4006 1. It localizes a set of alternatives. For example, the pattern
4008 cat(aract|erpillar|)
4010 matches one of the words "cat", "cataract", or "caterpillar". Without
4011 the parentheses, it would match "cataract", "erpillar" or an empty
4012 string.
4014 2. It sets up the subpattern as a capturing subpattern. This means
4015 that, when the whole pattern matches, that portion of the subject
4016 string that matched the subpattern is passed back to the caller via the
4017 ovector argument of pcre_exec(). Opening parentheses are counted from
4018 left to right (starting from 1) to obtain numbers for the capturing
4019 subpatterns.
4021 For example, if the string "the red king" is matched against the pat-
4022 tern
4024 the ((red|white) (king|queen))
4026 the captured substrings are "red king", "red", and "king", and are num-
4027 bered 1, 2, and 3, respectively.
4029 The fact that plain parentheses fulfil two functions is not always
4030 helpful. There are often times when a grouping subpattern is required
4031 without a capturing requirement. If an opening parenthesis is followed
4032 by a question mark and a colon, the subpattern does not do any captur-
4033 ing, and is not counted when computing the number of any subsequent
4034 capturing subpatterns. For example, if the string "the white queen" is
4035 matched against the pattern
4037 the ((?:red|white) (king|queen))
4039 the captured substrings are "white queen" and "queen", and are numbered
4040 1 and 2. The maximum number of capturing subpatterns is 65535.
4042 As a convenient shorthand, if any option settings are required at the
4043 start of a non-capturing subpattern, the option letters may appear
4044 between the "?" and the ":". Thus the two patterns
4046 (?i:saturday|sunday)
4047 (?:(?i)saturday|sunday)
4049 match exactly the same set of strings. Because alternative branches are
4050 tried from left to right, and options are not reset until the end of
4051 the subpattern is reached, an option setting in one branch does affect
4052 subsequent branches, so the above patterns match "SUNDAY" as well as
4053 "Saturday".
4058 Perl 5.10 introduced a feature whereby each alternative in a subpattern
4059 uses the same numbers for its capturing parentheses. Such a subpattern
4060 starts with (?| and is itself a non-capturing subpattern. For example,
4061 consider this pattern:
4063 (?|(Sat)ur|(Sun))day
4065 Because the two alternatives are inside a (?| group, both sets of cap-
4066 turing parentheses are numbered one. Thus, when the pattern matches,
4067 you can look at captured substring number one, whichever alternative
4068 matched. This construct is useful when you want to capture part, but
4069 not all, of one of a number of alternatives. Inside a (?| group, paren-
4070 theses are numbered as usual, but the number is reset at the start of
4071 each branch. The numbers of any capturing buffers that follow the sub-
4072 pattern start after the highest number used in any branch. The follow-
4073 ing example is taken from the Perl documentation. The numbers under-
4074 neath show in which buffer the captured content will be stored.
4076 # before ---------------branch-reset----------- after
4077 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
4078 # 1 2 2 3 2 3 4
4080 A back reference to a numbered subpattern uses the most recent value
4081 that is set for that number by any subpattern. The following pattern
4082 matches "abcabc" or "defdef":
4084 /(?|(abc)|(def))\1/
4086 In contrast, a recursive or "subroutine" call to a numbered subpattern
4087 always refers to the first one in the pattern with the given number.
4088 The following pattern matches "abcabc" or "defabc":
4090 /(?|(abc)|(def))(?1)/
4092 If a condition test for a subpattern's having matched refers to a non-
4093 unique number, the test is true if any of the subpatterns of that num-
4094 ber have matched.
4096 An alternative approach to using this "branch reset" feature is to use
4097 duplicate named subpatterns, as described in the next section.
4102 Identifying capturing parentheses by number is simple, but it can be
4103 very hard to keep track of the numbers in complicated regular expres-
4104 sions. Furthermore, if an expression is modified, the numbers may
4105 change. To help with this difficulty, PCRE supports the naming of sub-
4106 patterns. This feature was not added to Perl until release 5.10. Python
4107 had the feature earlier, and PCRE introduced it at release 4.0, using
4108 the Python syntax. PCRE now supports both the Perl and the Python syn-
4109 tax. Perl allows identically numbered subpatterns to have different
4110 names, but PCRE does not.
4112 In PCRE, a subpattern can be named in one of three ways: (?<name>...)
4113 or (?'name'...) as in Perl, or (?P<name>...) as in Python. References
4114 to capturing parentheses from other parts of the pattern, such as back
4115 references, recursion, and conditions, can be made by name as well as
4116 by number.
4118 Names consist of up to 32 alphanumeric characters and underscores.
4119 Named capturing parentheses are still allocated numbers as well as
4120 names, exactly as if the names were not present. The PCRE API provides
4121 function calls for extracting the name-to-number translation table from
4122 a compiled pattern. There is also a convenience function for extracting
4123 a captured substring by name.
4125 By default, a name must be unique within a pattern, but it is possible
4126 to relax this constraint by setting the PCRE_DUPNAMES option at compile
4127 time. (Duplicate names are also always permitted for subpatterns with
4128 the same number, set up as described in the previous section.) Dupli-
4129 cate names can be useful for patterns where only one instance of the
4130 named parentheses can match. Suppose you want to match the name of a
4131 weekday, either as a 3-letter abbreviation or as the full name, and in
4132 both cases you want to extract the abbreviation. This pattern (ignoring
4133 the line breaks) does the job:
4135 (?<DN>Mon|Fri|Sun)(?:day)?|
4136 (?<DN>Tue)(?:sday)?|
4137 (?<DN>Wed)(?:nesday)?|
4138 (?<DN>Thu)(?:rsday)?|
4139 (?<DN>Sat)(?:urday)?
4141 There are five capturing substrings, but only one is ever set after a
4142 match. (An alternative way of solving this problem is to use a "branch
4143 reset" subpattern, as described in the previous section.)
4145 The convenience function for extracting the data by name returns the
4146 substring for the first (and in this example, the only) subpattern of
4147 that name that matched. This saves searching to find which numbered
4148 subpattern it was.
4150 If you make a back reference to a non-unique named subpattern from
4151 elsewhere in the pattern, the one that corresponds to the first occur-
4152 rence of the name is used. In the absence of duplicate numbers (see the
4153 previous section) this is the one with the lowest number. If you use a
4154 named reference in a condition test (see the section about conditions
4155 below), either to check whether a subpattern has matched, or to check
4156 for recursion, all subpatterns with the same name are tested. If the
4157 condition is true for any one of them, the overall condition is true.
4158 This is the same behaviour as testing by number. For further details of
4159 the interfaces for handling named subpatterns, see the pcreapi documen-
4160 tation.
4162 Warning: You cannot use different names to distinguish between two sub-
4163 patterns with the same number because PCRE uses only the numbers when
4164 matching. For this reason, an error is given at compile time if differ-
4165 ent names are given to subpatterns with the same number. However, you
4166 can give the same name to subpatterns with the same number, even when
4167 PCRE_DUPNAMES is not set.
4172 Repetition is specified by quantifiers, which can follow any of the
4173 following items:
4175 a literal data character
4176 the dot metacharacter
4177 the \C escape sequence
4178 the \X escape sequence (in UTF-8 mode with Unicode properties)
4179 the \R escape sequence
4180 an escape such as \d that matches a single character
4181 a character class
4182 a back reference (see next section)
4183 a parenthesized subpattern (unless it is an assertion)
4184 a recursive or "subroutine" call to a subpattern
4186 The general repetition quantifier specifies a minimum and maximum num-
4187 ber of permitted matches, by giving the two numbers in curly brackets
4188 (braces), separated by a comma. The numbers must be less than 65536,
4189 and the first must be less than or equal to the second. For example:
4191 z{2,4}
4193 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
4194 special character. If the second number is omitted, but the comma is
4195 present, there is no upper limit; if the second number and the comma
4196 are both omitted, the quantifier specifies an exact number of required
4197 matches. Thus
4199 [aeiou]{3,}
4201 matches at least 3 successive vowels, but may match many more, while
4203 \d{8}
4205 matches exactly 8 digits. An opening curly bracket that appears in a
4206 position where a quantifier is not allowed, or one that does not match
4207 the syntax of a quantifier, is taken as a literal character. For exam-
4208 ple, {,6} is not a quantifier, but a literal string of four characters.
4210 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to
4211 individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char-
4212 acters, each of which is represented by a two-byte sequence. Similarly,
4213 when Unicode property support is available, \X{3} matches three Unicode
4214 extended sequences, each of which may be several bytes long (and they
4215 may be of different lengths).
4217 The quantifier {0} is permitted, causing the expression to behave as if
4218 the previous item and the quantifier were not present. This may be use-
4219 ful for subpatterns that are referenced as subroutines from elsewhere
4220 in the pattern. Items other than subpatterns that have a {0} quantifier
4221 are omitted from the compiled pattern.
4223 For convenience, the three most common quantifiers have single-charac-
4224 ter abbreviations:
4226 * is equivalent to {0,}
4227 + is equivalent to {1,}
4228 ? is equivalent to {0,1}
4230 It is possible to construct infinite loops by following a subpattern
4231 that can match no characters with a quantifier that has no upper limit,
4232 for example:
4234 (a?)*
4236 Earlier versions of Perl and PCRE used to give an error at compile time
4237 for such patterns. However, because there are cases where this can be
4238 useful, such patterns are now accepted, but if any repetition of the
4239 subpattern does in fact match no characters, the loop is forcibly bro-
4240 ken.
4242 By default, the quantifiers are "greedy", that is, they match as much
4243 as possible (up to the maximum number of permitted times), without
4244 causing the rest of the pattern to fail. The classic example of where
4245 this gives problems is in trying to match comments in C programs. These
4246 appear between /* and */ and within the comment, individual * and /
4247 characters may appear. An attempt to match C comments by applying the
4248 pattern
4250 /\*.*\*/
4252 to the string
4254 /* first comment */ not comment /* second comment */
4256 fails, because it matches the entire string owing to the greediness of
4257 the .* item.
4259 However, if a quantifier is followed by a question mark, it ceases to
4260 be greedy, and instead matches the minimum number of times possible, so
4261 the pattern
4263 /\*.*?\*/
4265 does the right thing with the C comments. The meaning of the various
4266 quantifiers is not otherwise changed, just the preferred number of
4267 matches. Do not confuse this use of question mark with its use as a
4268 quantifier in its own right. Because it has two uses, it can sometimes
4269 appear doubled, as in
4271 \d??\d
4273 which matches one digit by preference, but can match two if that is the
4274 only way the rest of the pattern matches.
4276 If the PCRE_UNGREEDY option is set (an option that is not available in
4277 Perl), the quantifiers are not greedy by default, but individual ones
4278 can be made greedy by following them with a question mark. In other
4279 words, it inverts the default behaviour.
4281 When a parenthesized subpattern is quantified with a minimum repeat
4282 count that is greater than 1 or with a limited maximum, more memory is
4283 required for the compiled pattern, in proportion to the size of the
4284 minimum or maximum.
4286 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
4287 alent to Perl's /s) is set, thus allowing the dot to match newlines,
4288 the pattern is implicitly anchored, because whatever follows will be
4289 tried against every character position in the subject string, so there
4290 is no point in retrying the overall match at any position after the
4291 first. PCRE normally treats such a pattern as though it were preceded
4292 by \A.
4294 In cases where it is known that the subject string contains no new-
4295 lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
4296 mization, or alternatively using ^ to indicate anchoring explicitly.
4298 However, there is one situation where the optimization cannot be used.
4299 When .* is inside capturing parentheses that are the subject of a back
4300 reference elsewhere in the pattern, a match at the start may fail where
4301 a later one succeeds. Consider, for example:
4303 (.*)abc\1
4305 If the subject is "xyz123abc123" the match point is the fourth charac-
4306 ter. For this reason, such a pattern is not implicitly anchored.
4308 When a capturing subpattern is repeated, the value captured is the sub-
4309 string that matched the final iteration. For example, after
4311 (tweedle[dume]{3}\s*)+
4313 has matched "tweedledum tweedledee" the value of the captured substring
4314 is "tweedledee". However, if there are nested capturing subpatterns,
4315 the corresponding captured values may have been set in previous itera-
4316 tions. For example, after
4318 /(a|(b))+/
4320 matches "aba" the value of the second captured substring is "b".
4325 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
4326 repetition, failure of what follows normally causes the repeated item
4327 to be re-evaluated to see if a different number of repeats allows the
4328 rest of the pattern to match. Sometimes it is useful to prevent this,
4329 either to change the nature of the match, or to cause it fail earlier
4330 than it otherwise might, when the author of the pattern knows there is
4331 no point in carrying on.
4333 Consider, for example, the pattern \d+foo when applied to the subject
4334 line
4336 123456bar
4338 After matching all 6 digits and then failing to match "foo", the normal
4339 action of the matcher is to try again with only 5 digits matching the
4340 \d+ item, and then with 4, and so on, before ultimately failing.
4341 "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
4342 the means for specifying that once a subpattern has matched, it is not
4343 to be re-evaluated in this way.
4345 If we use atomic grouping for the previous example, the matcher gives
4346 up immediately on failing to match "foo" the first time. The notation
4347 is a kind of special parenthesis, starting with (?> as in this example:
4349 (?>\d+)foo
4351 This kind of parenthesis "locks up" the part of the pattern it con-
4352 tains once it has matched, and a failure further into the pattern is
4353 prevented from backtracking into it. Backtracking past it to previous
4354 items, however, works as normal.
4356 An alternative description is that a subpattern of this type matches
4357 the string of characters that an identical standalone pattern would
4358 match, if anchored at the current point in the subject string.
4360 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
4361 such as the above example can be thought of as a maximizing repeat that
4362 must swallow everything it can. So, while both \d+ and \d+? are pre-
4363 pared to adjust the number of digits they match in order to make the
4364 rest of the pattern match, (?>\d+) can only match an entire sequence of
4365 digits.
4367 Atomic groups in general can of course contain arbitrarily complicated
4368 subpatterns, and can be nested. However, when the subpattern for an
4369 atomic group is just a single repeated item, as in the example above, a
4370 simpler notation, called a "possessive quantifier" can be used. This
4371 consists of an additional + character following a quantifier. Using
4372 this notation, the previous example can be rewritten as
4374 \d++foo
4376 Note that a possessive quantifier can be used with an entire group, for
4377 example:
4379 (abc|xyz){2,3}+
4381 Possessive quantifiers are always greedy; the setting of the
4382 PCRE_UNGREEDY option is ignored. They are a convenient notation for the
4383 simpler forms of atomic group. However, there is no difference in the
4384 meaning of a possessive quantifier and the equivalent atomic group,
4385 though there may be a performance difference; possessive quantifiers
4386 should be slightly faster.
4388 The possessive quantifier syntax is an extension to the Perl 5.8 syn-
4389 tax. Jeffrey Friedl originated the idea (and the name) in the first
4390 edition of his book. Mike McCloskey liked it, so implemented it when he
4391 built Sun's Java package, and PCRE copied it from there. It ultimately
4392 found its way into Perl at release 5.10.
4394 PCRE has an optimization that automatically "possessifies" certain sim-
4395 ple pattern constructs. For example, the sequence A+B is treated as
4396 A++B because there is no point in backtracking into a sequence of A's
4397 when B must follow.
4399 When a pattern contains an unlimited repeat inside a subpattern that
4400 can itself be repeated an unlimited number of times, the use of an
4401 atomic group is the only way to avoid some failing matches taking a
4402 very long time indeed. The pattern
4404 (\D+|<\d+>)*[!?]
4406 matches an unlimited number of substrings that either consist of non-
4407 digits, or digits enclosed in <>, followed by either ! or ?. When it
4408 matches, it runs quickly. However, if it is applied to
4410 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4412 it takes a long time before reporting failure. This is because the
4413 string can be divided between the internal \D+ repeat and the external
4414 * repeat in a large number of ways, and all have to be tried. (The
4415 example uses [!?] rather than a single character at the end, because
4416 both PCRE and Perl have an optimization that allows for fast failure
4417 when a single character is used. They remember the last single charac-
4418 ter that is required for a match, and fail early if it is not present
4419 in the string.) If the pattern is changed so that it uses an atomic
4420 group, like this:
4422 ((?>\D+)|<\d+>)*[!?]
4424 sequences of non-digits cannot be broken, and failure happens quickly.
4429 Outside a character class, a backslash followed by a digit greater than
4430 0 (and possibly further digits) is a back reference to a capturing sub-
4431 pattern earlier (that is, to its left) in the pattern, provided there
4432 have been that many previous capturing left parentheses.
4434 However, if the decimal number following the backslash is less than 10,
4435 it is always taken as a back reference, and causes an error only if
4436 there are not that many capturing left parentheses in the entire pat-
4437 tern. In other words, the parentheses that are referenced need not be
4438 to the left of the reference for numbers less than 10. A "forward back
4439 reference" of this type can make sense when a repetition is involved
4440 and the subpattern to the right has participated in an earlier itera-
4441 tion.
4443 It is not possible to have a numerical "forward back reference" to a
4444 subpattern whose number is 10 or more using this syntax because a
4445 sequence such as \50 is interpreted as a character defined in octal.
4446 See the subsection entitled "Non-printing characters" above for further
4447 details of the handling of digits following a backslash. There is no
4448 such problem when named parentheses are used. A back reference to any
4449 subpattern is possible using named parentheses (see below).
4451 Another way of avoiding the ambiguity inherent in the use of digits
4452 following a backslash is to use the \g escape sequence, which is a fea-
4453 ture introduced in Perl 5.10. This escape must be followed by an
4454 unsigned number or a negative number, optionally enclosed in braces.
4455 These examples are all identical:
4457 (ring), \1
4458 (ring), \g1
4459 (ring), \g{1}
4461 An unsigned number specifies an absolute reference without the ambigu-
4462 ity that is present in the older syntax. It is also useful when literal
4463 digits follow the reference. A negative number is a relative reference.
4464 Consider this example:
4466 (abc(def)ghi)\g{-1}
4468 The sequence \g{-1} is a reference to the most recently started captur-
4469 ing subpattern before \g, that is, is it equivalent to \2. Similarly,
4470 \g{-2} would be equivalent to \1. The use of relative references can be
4471 helpful in long patterns, and also in patterns that are created by
4472 joining together fragments that contain references within themselves.
4474 A back reference matches whatever actually matched the capturing sub-
4475 pattern in the current subject string, rather than anything matching
4476 the subpattern itself (see "Subpatterns as subroutines" below for a way
4477 of doing that). So the pattern
4479 (sens|respons)e and \1ibility
4481 matches "sense and sensibility" and "response and responsibility", but
4482 not "sense and responsibility". If caseful matching is in force at the
4483 time of the back reference, the case of letters is relevant. For exam-
4484 ple,
4486 ((?i)rah)\s+\1
4488 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
4489 original capturing subpattern is matched caselessly.
4491 There are several different ways of writing back references to named
4492 subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or
4493 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's
4494 unified back reference syntax, in which \g can be used for both numeric
4495 and named references, is also supported. We could rewrite the above
4496 example in any of the following ways:
4498 (?<p1>(?i)rah)\s+\k<p1>
4499 (?'p1'(?i)rah)\s+\k{p1}
4500 (?P<p1>(?i)rah)\s+(?P=p1)
4501 (?<p1>(?i)rah)\s+\g{p1}
4503 A subpattern that is referenced by name may appear in the pattern
4504 before or after the reference.
4506 There may be more than one back reference to the same subpattern. If a
4507 subpattern has not actually been used in a particular match, any back
4508 references to it always fail by default. For example, the pattern
4510 (a|(bc))\2
4512 always fails if it starts to match "a" rather than "bc". However, if
4513 the PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back refer-
4514 ence to an unset value matches an empty string.
4516 Because there may be many capturing parentheses in a pattern, all dig-
4517 its following a backslash are taken as part of a potential back refer-
4518 ence number. If the pattern continues with a digit character, some
4519 delimiter must be used to terminate the back reference. If the
4520 PCRE_EXTENDED option is set, this can be whitespace. Otherwise, the \g{
4521 syntax or an empty comment (see "Comments" below) can be used.
4523 Recursive back references
4525 A back reference that occurs inside the parentheses to which it refers
4526 fails when the subpattern is first used, so, for example, (a\1) never
4527 matches. However, such references can be useful inside repeated sub-
4528 patterns. For example, the pattern
4530 (a|b\1)+
4532 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
4533 ation of the subpattern, the back reference matches the character
4534 string corresponding to the previous iteration. In order for this to
4535 work, the pattern must be such that the first iteration does not need
4536 to match the back reference. This can be done using alternation, as in
4537 the example above, or by a quantifier with a minimum of zero.
4539 Back references of this type cause the group that they reference to be
4540 treated as an atomic group. Once the whole group has been matched, a
4541 subsequent matching failure cannot cause backtracking into the middle
4542 of the group.
4547 An assertion is a test on the characters following or preceding the
4548 current matching point that does not actually consume any characters.
4549 The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
4550 described above.
4552 More complicated assertions are coded as subpatterns. There are two
4553 kinds: those that look ahead of the current position in the subject
4554 string, and those that look behind it. An assertion subpattern is
4555 matched in the normal way, except that it does not cause the current
4556 matching position to be changed.
4558 Assertion subpatterns are not capturing subpatterns, and may not be
4559 repeated, because it makes no sense to assert the same thing several
4560 times. If any kind of assertion contains capturing subpatterns within
4561 it, these are counted for the purposes of numbering the capturing sub-
4562 patterns in the whole pattern. However, substring capturing is carried
4563 out only for positive assertions, because it does not make sense for
4564 negative assertions.
4566 Lookahead assertions
4568 Lookahead assertions start with (?= for positive assertions and (?! for
4569 negative assertions. For example,
4571 \w+(?=;)
4573 matches a word followed by a semicolon, but does not include the semi-
4574 colon in the match, and
4576 foo(?!bar)
4578 matches any occurrence of "foo" that is not followed by "bar". Note
4579 that the apparently similar pattern
4581 (?!foo)bar
4583 does not find an occurrence of "bar" that is preceded by something
4584 other than "foo"; it finds any occurrence of "bar" whatsoever, because
4585 the assertion (?!foo) is always true when the next three characters are
4586 "bar". A lookbehind assertion is needed to achieve the other effect.
4588 If you want to force a matching failure at some point in a pattern, the
4589 most convenient way to do it is with (?!) because an empty string
4590 always matches, so an assertion that requires there not to be an empty
4591 string must always fail. The Perl 5.10 backtracking control verb
4592 (*FAIL) or (*F) is essentially a synonym for (?!).
4594 Lookbehind assertions
4596 Lookbehind assertions start with (?<= for positive assertions and (?<!
4597 for negative assertions. For example,
4599 (?<!foo)bar
4601 does find an occurrence of "bar" that is not preceded by "foo". The
4602 contents of a lookbehind assertion are restricted such that all the
4603 strings it matches must have a fixed length. However, if there are sev-
4604 eral top-level alternatives, they do not all have to have the same
4605 fixed length. Thus
4607 (?<=bullock|donkey)
4609 is permitted, but
4611 (?<!dogs?|cats?)
4613 causes an error at compile time. Branches that match different length
4614 strings are permitted only at the top level of a lookbehind assertion.
4615 This is an extension compared with Perl (5.8 and 5.10), which requires
4616 all branches to match the same length of string. An assertion such as
4618 (?<=ab(c|de))
4620 is not permitted, because its single top-level branch can match two
4621 different lengths, but it is acceptable to PCRE if rewritten to use two
4622 top-level branches:
4624 (?<=abc|abde)
4626 In some cases, the Perl 5.10 escape sequence \K (see above) can be used
4627 instead of a lookbehind assertion to get round the fixed-length
4628 restriction.
4630 The implementation of lookbehind assertions is, for each alternative,
4631 to temporarily move the current position back by the fixed length and
4632 then try to match. If there are insufficient characters before the cur-
4633 rent position, the assertion fails.
4635 PCRE does not allow the \C escape (which matches a single byte in UTF-8
4636 mode) to appear in lookbehind assertions, because it makes it impossi-
4637 ble to calculate the length of the lookbehind. The \X and \R escapes,
4638 which can match different numbers of bytes, are also not permitted.
4640 "Subroutine" calls (see below) such as (?2) or (?&X) are permitted in
4641 lookbehinds, as long as the subpattern matches a fixed-length string.
4642 Recursion, however, is not supported.
4644 Possessive quantifiers can be used in conjunction with lookbehind
4645 assertions to specify efficient matching of fixed-length strings at the
4646 end of subject strings. Consider a simple pattern such as
4648 abcd$
4650 when applied to a long string that does not match. Because matching
4651 proceeds from left to right, PCRE will look for each "a" in the subject
4652 and then see if what follows matches the rest of the pattern. If the
4653 pattern is specified as
4655 ^.*abcd$
4657 the initial .* matches the entire string at first, but when this fails
4658 (because there is no following "a"), it backtracks to match all but the
4659 last character, then all but the last two characters, and so on. Once
4660 again the search for "a" covers the entire string, from right to left,
4661 so we are no better off. However, if the pattern is written as
4663 ^.*+(?<=abcd)
4665 there can be no backtracking for the .*+ item; it can match only the
4666 entire string. The subsequent lookbehind assertion does a single test
4667 on the last four characters. If it fails, the match fails immediately.
4668 For long strings, this approach makes a significant difference to the
4669 processing time.
4671 Using multiple assertions
4673 Several assertions (of any sort) may occur in succession. For example,
4675 (?<=\d{3})(?<!999)foo
4677 matches "foo" preceded by three digits that are not "999". Notice that
4678 each of the assertions is applied independently at the same point in
4679 the subject string. First there is a check that the previous three
4680 characters are all digits, and then there is a check that the same
4681 three characters are not "999". This pattern does not match "foo" pre-
4682 ceded by six characters, the first of which are digits and the last
4683 three of which are not "999". For example, it doesn't match "123abc-
4684 foo". A pattern to do that is
4686 (?<=\d{3}...)(?<!999)foo
4688 This time the first assertion looks at the preceding six characters,
4689 checking that the first three are digits, and then the second assertion
4690 checks that the preceding three characters are not "999".
4692 Assertions can be nested in any combination. For example,
4694 (?<=(?<!foo)bar)baz
4696 matches an occurrence of "baz" that is preceded by "bar" which in turn
4697 is not preceded by "foo", while
4699 (?<=\d{3}(?!999)...)foo
4701 is another pattern that matches "foo" preceded by three digits and any
4702 three characters that are not "999".
4707 It is possible to cause the matching process to obey a subpattern con-
4708 ditionally or to choose between two alternative subpatterns, depending
4709 on the result of an assertion, or whether a specific capturing subpat-
4710 tern has already been matched. The two possible forms of conditional
4711 subpattern are:
4713 (?(condition)yes-pattern)
4714 (?(condition)yes-pattern|no-pattern)
4716 If the condition is satisfied, the yes-pattern is used; otherwise the
4717 no-pattern (if present) is used. If there are more than two alterna-
4718 tives in the subpattern, a compile-time error occurs.
4720 There are four kinds of condition: references to subpatterns, refer-
4721 ences to recursion, a pseudo-condition called DEFINE, and assertions.
4723 Checking for a used subpattern by number
4725 If the text between the parentheses consists of a sequence of digits,
4726 the condition is true if a capturing subpattern of that number has pre-
4727 viously matched. If there is more than one capturing subpattern with
4728 the same number (see the earlier section about duplicate subpattern
4729 numbers), the condition is true if any of them have been set. An alter-
4730 native notation is to precede the digits with a plus or minus sign. In
4731 this case, the subpattern number is relative rather than absolute. The
4732 most recently opened parentheses can be referenced by (?(-1), the next
4733 most recent by (?(-2), and so on. In looping constructs it can also
4734 make sense to refer to subsequent groups with constructs such as
4735 (?(+2).
4737 Consider the following pattern, which contains non-significant white
4738 space to make it more readable (assume the PCRE_EXTENDED option) and to
4739 divide it into three parts for ease of discussion:
4741 ( \( )? [^()]+ (?(1) \) )
4743 The first part matches an optional opening parenthesis, and if that
4744 character is present, sets it as the first captured substring. The sec-
4745 ond part matches one or more characters that are not parentheses. The
4746 third part is a conditional subpattern that tests whether the first set
4747 of parentheses matched or not. If they did, that is, if subject started
4748 with an opening parenthesis, the condition is true, and so the yes-pat-
4749 tern is executed and a closing parenthesis is required. Otherwise,
4750 since no-pattern is not present, the subpattern matches nothing. In
4751 other words, this pattern matches a sequence of non-parentheses,
4752 optionally enclosed in parentheses.
4754 If you were embedding this pattern in a larger one, you could use a
4755 relative reference:
4757 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
4759 This makes the fragment independent of the parentheses in the larger
4760 pattern.
4762 Checking for a used subpattern by name
4764 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a
4765 used subpattern by name. For compatibility with earlier versions of
4766 PCRE, which had this facility before Perl, the syntax (?(name)...) is
4767 also recognized. However, there is a possible ambiguity with this syn-
4768 tax, because subpattern names may consist entirely of digits. PCRE
4769 looks first for a named subpattern; if it cannot find one and the name
4770 consists entirely of digits, PCRE looks for a subpattern of that num-
4771 ber, which must be greater than zero. Using subpattern names that con-
4772 sist entirely of digits is not recommended.
4774 Rewriting the above example to use a named subpattern gives this:
4776 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
4778 If the name used in a condition of this kind is a duplicate, the test
4779 is applied to all subpatterns of the same name, and is true if any one
4780 of them has matched.
4782 Checking for pattern recursion
4784 If the condition is the string (R), and there is no subpattern with the
4785 name R, the condition is true if a recursive call to the whole pattern
4786 or any subpattern has been made. If digits or a name preceded by amper-
4787 sand follow the letter R, for example:
4789 (?(R3)...) or (?(R&name)...)
4791 the condition is true if the most recent recursion is into a subpattern
4792 whose number or name is given. This condition does not check the entire
4793 recursion stack. If the name used in a condition of this kind is a
4794 duplicate, the test is applied to all subpatterns of the same name, and
4795 is true if any one of them is the most recent recursion.
4797 At "top level", all these recursion test conditions are false. The
4798 syntax for recursive patterns is described below.
4800 Defining subpatterns for use by reference only
4802 If the condition is the string (DEFINE), and there is no subpattern
4803 with the name DEFINE, the condition is always false. In this case,
4804 there may be only one alternative in the subpattern. It is always
4805 skipped if control reaches this point in the pattern; the idea of
4806 DEFINE is that it can be used to define "subroutines" that can be ref-
4807 erenced from elsewhere. (The use of "subroutines" is described below.)
4808 For example, a pattern to match an IPv4 address could be written like
4809 this (ignore whitespace and line breaks):
4811 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
4812 \b (?&byte) (\.(?&byte)){3} \b
4814 The first part of the pattern is a DEFINE group inside which a another
4815 group named "byte" is defined. This matches an individual component of
4816 an IPv4 address (a number less than 256). When matching takes place,
4817 this part of the pattern is skipped because DEFINE acts like a false
4818 condition. The rest of the pattern uses references to the named group
4819 to match the four dot-separated components of an IPv4 address, insist-
4820 ing on a word boundary at each end.
4822 Assertion conditions
4824 If the condition is not in any of the above formats, it must be an
4825 assertion. This may be a positive or negative lookahead or lookbehind
4826 assertion. Consider this pattern, again containing non-significant
4827 white space, and with the two alternatives on the second line:
4829 (?(?=[^a-z]*[a-z])
4830 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
4832 The condition is a positive lookahead assertion that matches an
4833 optional sequence of non-letters followed by a letter. In other words,
4834 it tests for the presence of at least one letter in the subject. If a
4835 letter is found, the subject is matched against the first alternative;
4836 otherwise it is matched against the second. This pattern matches
4837 strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are
4838 letters and dd are digits.
4843 The sequence (?# marks the start of a comment that continues up to the
4844 next closing parenthesis. Nested parentheses are not permitted. The
4845 characters that make up a comment play no part in the pattern matching
4846 at all.
4848 If the PCRE_EXTENDED option is set, an unescaped # character outside a
4849 character class introduces a comment that continues to immediately
4850 after the next newline in the pattern.
4855 Consider the problem of matching a string in parentheses, allowing for
4856 unlimited nested parentheses. Without the use of recursion, the best
4857 that can be done is to use a pattern that matches up to some fixed
4858 depth of nesting. It is not possible to handle an arbitrary nesting
4859 depth.
4861 For some time, Perl has provided a facility that allows regular expres-
4862 sions to recurse (amongst other things). It does this by interpolating
4863 Perl code in the expression at run time, and the code can refer to the
4864 expression itself. A Perl pattern using code interpolation to solve the
4865 parentheses problem can be created like this:
4867 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
4869 The (?p{...}) item interpolates Perl code at run time, and in this case
4870 refers recursively to the pattern in which it appears.
4872 Obviously, PCRE cannot support the interpolation of Perl code. Instead,
4873 it supports special syntax for recursion of the entire pattern, and
4874 also for individual subpattern recursion. After its introduction in
4875 PCRE and Python, this kind of recursion was subsequently introduced
4876 into Perl at release 5.10.
4878 A special item that consists of (? followed by a number greater than
4879 zero and a closing parenthesis is a recursive call of the subpattern of
4880 the given number, provided that it occurs inside that subpattern. (If
4881 not, it is a "subroutine" call, which is described in the next sec-
4882 tion.) The special item (?R) or (?0) is a recursive call of the entire
4883 regular expression.
4885 This PCRE pattern solves the nested parentheses problem (assume the
4886 PCRE_EXTENDED option is set so that white space is ignored):
4888 \( ( [^()]++ | (?R) )* \)
4890 First it matches an opening parenthesis. Then it matches any number of
4891 substrings which can either be a sequence of non-parentheses, or a
4892 recursive match of the pattern itself (that is, a correctly parenthe-
4893 sized substring). Finally there is a closing parenthesis. Note the use
4894 of a possessive quantifier to avoid backtracking into sequences of non-
4895 parentheses.
4897 If this were part of a larger pattern, you would not want to recurse
4898 the entire pattern, so instead you could use this:
4900 ( \( ( [^()]++ | (?1) )* \) )
4902 We have put the pattern into parentheses, and caused the recursion to
4903 refer to them instead of the whole pattern.
4905 In a larger pattern, keeping track of parenthesis numbers can be
4906 tricky. This is made easier by the use of relative references (a Perl
4907 5.10 feature). Instead of (?1) in the pattern above you can write
4908 (?-2) to refer to the second most recently opened parentheses preceding
4909 the recursion. In other words, a negative number counts capturing
4910 parentheses leftwards from the point at which it is encountered.
4912 It is also possible to refer to subsequently opened parentheses, by
4913 writing references such as (?+2). However, these cannot be recursive
4914 because the reference is not inside the parentheses that are refer-
4915 enced. They are always "subroutine" calls, as described in the next
4916 section.
4918 An alternative approach is to use named parentheses instead. The Perl
4919 syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also
4920 supported. We could rewrite the above example as follows:
4922 (?<pn> \( ( [^()]++ | (?&pn) )* \) )
4924 If there is more than one subpattern with the same name, the earliest
4925 one is used.
4927 This particular example pattern that we have been looking at contains
4928 nested unlimited repeats, and so the use of a possessive quantifier for
4929 matching strings of non-parentheses is important when applying the pat-
4930 tern to strings that do not match. For example, when this pattern is
4931 applied to
4933 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
4935 it yields "no match" quickly. However, if a possessive quantifier is
4936 not used, the match runs for a very long time indeed because there are
4937 so many different ways the + and * repeats can carve up the subject,
4938 and all have to be tested before failure can be reported.
4940 At the end of a match, the values of capturing parentheses are those
4941 from the outermost level. If you want to obtain intermediate values, a
4942 callout function can be used (see below and the pcrecallout documenta-
4943 tion). If the pattern above is matched against
4945 (ab(cd)ef)
4947 the value for the inner capturing parentheses (numbered 2) is "ef",
4948 which is the last value taken on at the top level. If a capturing sub-
4949 pattern is not matched at the top level, its final value is unset, even
4950 if it is (temporarily) set at a deeper level.
4952 If there are more than 15 capturing parentheses in a pattern, PCRE has
4953 to obtain extra memory to store data during a recursion, which it does
4954 by using pcre_malloc, freeing it via pcre_free afterwards. If no memory
4955 can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
4957 Do not confuse the (?R) item with the condition (R), which tests for
4958 recursion. Consider this pattern, which matches text in angle brack-
4959 ets, allowing for arbitrary nesting. Only digits are allowed in nested
4960 brackets (that is, when recursing), whereas any characters are permit-
4961 ted at the outer level.
4963 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
4965 In this pattern, (?(R) is the start of a conditional subpattern, with
4966 two different alternatives for the recursive and non-recursive cases.
4967 The (?R) item is the actual recursive call.
4969 Recursion difference from Perl
4971 In PCRE (like Python, but unlike Perl), a recursive subpattern call is
4972 always treated as an atomic group. That is, once it has matched some of
4973 the subject string, it is never re-entered, even if it contains untried
4974 alternatives and there is a subsequent matching failure. This can be
4975 illustrated by the following pattern, which purports to match a palin-
4976 dromic string that contains an odd number of characters (for example,
4977 "a", "aba", "abcba", "abcdcba"):
4979 ^(.|(.)(?1)\2)$
4981 The idea is that it either matches a single character, or two identical
4982 characters surrounding a sub-palindrome. In Perl, this pattern works;
4983 in PCRE it does not if the pattern is longer than three characters.
4984 Consider the subject string "abcba":
4986 At the top level, the first character is matched, but as it is not at
4987 the end of the string, the first alternative fails; the second alterna-
4988 tive is taken and the recursion kicks in. The recursive call to subpat-
4989 tern 1 successfully matches the next character ("b"). (Note that the
4990 beginning and end of line tests are not part of the recursion).
4992 Back at the top level, the next character ("c") is compared with what
4993 subpattern 2 matched, which was "a". This fails. Because the recursion
4994 is treated as an atomic group, there are now no backtracking points,
4995 and so the entire match fails. (Perl is able, at this point, to re-
4996 enter the recursion and try the second alternative.) However, if the
4997 pattern is written with the alternatives in the other order, things are
4998 different:
5000 ^((.)(?1)\2|.)$
5002 This time, the recursing alternative is tried first, and continues to
5003 recurse until it runs out of characters, at which point the recursion
5004 fails. But this time we do have another alternative to try at the
5005 higher level. That is the big difference: in the previous case the
5006 remaining alternative is at a deeper recursion level, which PCRE cannot
5007 use.
5009 To change the pattern so that matches all palindromic strings, not just
5010 those with an odd number of characters, it is tempting to change the
5011 pattern to this:
5013 ^((.)(?1)\2|.?)$
5015 Again, this works in Perl, but not in PCRE, and for the same reason.
5016 When a deeper recursion has matched a single character, it cannot be
5017 entered again in order to match an empty string. The solution is to
5018 separate the two cases, and write out the odd and even cases as alter-
5019 natives at the higher level:
5021 ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
5023 If you want to match typical palindromic phrases, the pattern has to
5024 ignore all non-word characters, which can be done like this:
5026 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
5028 If run with the PCRE_CASELESS option, this pattern matches phrases such
5029 as "A man, a plan, a canal: Panama!" and it works well in both PCRE and
5030 Perl. Note the use of the possessive quantifier *+ to avoid backtrack-
5031 ing into sequences of non-word characters. Without this, PCRE takes a
5032 great deal longer (ten times or more) to match typical phrases, and
5033 Perl takes so long that you think it has gone into a loop.
5035 WARNING: The palindrome-matching patterns above work only if the sub-
5036 ject string does not start with a palindrome that is shorter than the
5037 entire string. For example, although "abcba" is correctly matched, if
5038 the subject is "ababa", PCRE finds the palindrome "aba" at the start,
5039 then fails at top level because the end of the string does not follow.
5040 Once again, it cannot jump back into the recursion to try other alter-
5041 natives, so the entire match fails.
5046 If the syntax for a recursive subpattern reference (either by number or
5047 by name) is used outside the parentheses to which it refers, it oper-
5048 ates like a subroutine in a programming language. The "called" subpat-
5049 tern may be defined before or after the reference. A numbered reference
5050 can be absolute or relative, as in these examples:
5052 (...(absolute)...)...(?2)...
5053 (...(relative)...)...(?-1)...
5054 (...(?+1)...(relative)...
5056 An earlier example pointed out that the pattern
5058 (sens|respons)e and \1ibility
5060 matches "sense and sensibility" and "response and responsibility", but
5061 not "sense and responsibility". If instead the pattern
5063 (sens|respons)e and (?1)ibility
5065 is used, it does match "sense and responsibility" as well as the other
5066 two strings. Another example is given in the discussion of DEFINE
5067 above.
5069 Like recursive subpatterns, a subroutine call is always treated as an
5070 atomic group. That is, once it has matched some of the subject string,
5071 it is never re-entered, even if it contains untried alternatives and
5072 there is a subsequent matching failure. Any capturing parentheses that
5073 are set during the subroutine call revert to their previous values
5074 afterwards.
5076 When a subpattern is used as a subroutine, processing options such as
5077 case-independence are fixed when the subpattern is defined. They cannot
5078 be changed for different calls. For example, consider this pattern:
5080 (abc)(?i:(?-1))
5082 It matches "abcabc". It does not match "abcABC" because the change of
5083 processing option does not affect the called subpattern.
5088 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
5089 name or a number enclosed either in angle brackets or single quotes, is
5090 an alternative syntax for referencing a subpattern as a subroutine,
5091 possibly recursively. Here are two of the examples used above, rewrit-
5092 ten using this syntax:
5094 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
5095 (sens|respons)e and \g'1'ibility
5097 PCRE supports an extension to Oniguruma: if a number is preceded by a
5098 plus or a minus sign it is taken as a relative reference. For example:
5100 (abc)(?i:\g<-1>)
5102 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not
5103 synonymous. The former is a back reference; the latter is a subroutine
5104 call.
5109 Perl has a feature whereby using the sequence (?{...}) causes arbitrary
5110 Perl code to be obeyed in the middle of matching a regular expression.
5111 This makes it possible, amongst other things, to extract different sub-
5112 strings that match the same pair of parentheses when there is a repeti-
5113 tion.
5115 PCRE provides a similar feature, but of course it cannot obey arbitrary
5116 Perl code. The feature is called "callout". The caller of PCRE provides
5117 an external function by putting its entry point in the global variable
5118 pcre_callout. By default, this variable contains NULL, which disables
5119 all calling out.
5121 Within a regular expression, (?C) indicates the points at which the
5122 external function is to be called. If you want to identify different
5123 callout points, you can put a number less than 256 after the letter C.
5124 The default value is zero. For example, this pattern has two callout
5125 points:
5127 (?C1)abc(?C2)def
5129 If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are
5130 automatically installed before each item in the pattern. They are all
5131 numbered 255.
5133 During matching, when PCRE reaches a callout point (and pcre_callout is
5134 set), the external function is called. It is provided with the number
5135 of the callout, the position in the pattern, and, optionally, one item
5136 of data originally supplied by the caller of pcre_exec(). The callout
5137 function may cause matching to proceed, to backtrack, or to fail alto-
5138 gether. A complete description of the interface to the callout function
5139 is given in the pcrecallout documentation.
5144 Perl 5.10 introduced a number of "Special Backtracking Control Verbs",
5145 which are described in the Perl documentation as "experimental and sub-
5146 ject to change or removal in a future version of Perl". It goes on to
5147 say: "Their usage in production code should be noted to avoid problems
5148 during upgrades." The same remarks apply to the PCRE features described
5149 in this section.
5151 Since these verbs are specifically related to backtracking, most of
5152 them can be used only when the pattern is to be matched using
5153 pcre_exec(), which uses a backtracking algorithm. With the exception of
5154 (*FAIL), which behaves like a failing negative assertion, they cause an
5155 error if encountered by pcre_dfa_exec().
5157 If any of these verbs are used in an assertion or subroutine subpattern
5158 (including recursive subpatterns), their effect is confined to that
5159 subpattern; it does not extend to the surrounding pattern. Note that
5160 such subpatterns are processed as anchored at the point where they are
5161 tested.
5163 The new verbs make use of what was previously invalid syntax: an open-
5164 ing parenthesis followed by an asterisk. They are generally of the form
5165 (*VERB) or (*VERB:NAME). Some may take either form, with differing be-
5166 haviour, depending on whether or not an argument is present. An name is
5167 a sequence of letters, digits, and underscores. If the name is empty,
5168 that is, if the closing parenthesis immediately follows the colon, the
5169 effect is as if the colon were not there. Any number of these verbs may
5170 occur in a pattern.
5172 PCRE contains some optimizations that are used to speed up matching by
5173 running some checks at the start of each match attempt. For example, it
5174 may know the minimum length of matching subject, or that a particular
5175 character must be present. When one of these optimizations suppresses
5176 the running of a match, any included backtracking verbs will not, of
5177 course, be processed. You can suppress the start-of-match optimizations
5178 by setting the PCRE_NO_START_OPTIMIZE option when calling pcre_exec().
5180 Verbs that act immediately
5182 The following verbs act as soon as they are encountered. They may not
5183 be followed by a name.
5185 (*ACCEPT)
5187 This verb causes the match to end successfully, skipping the remainder
5188 of the pattern. When inside a recursion, only the innermost pattern is
5189 ended immediately. If (*ACCEPT) is inside capturing parentheses, the
5190 data so far is captured. (This feature was added to PCRE at release
5191 8.00.) For example:
5193 A((?:A|B(*ACCEPT)|C)D)
5195 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is cap-
5196 tured by the outer parentheses.
5198 (*FAIL) or (*F)
5200 This verb causes the match to fail, forcing backtracking to occur. It
5201 is equivalent to (?!) but easier to read. The Perl documentation notes
5202 that it is probably useful only when combined with (?{}) or (??{}).
5203 Those are, of course, Perl features that are not present in PCRE. The
5204 nearest equivalent is the callout feature, as for example in this pat-
5205 tern:
5207 a+(?C)(*FAIL)
5209 A match with the string "aaaa" always fails, but the callout is taken
5210 before each backtrack happens (in this example, 10 times).
5212 Recording which path was taken
5214 There is one verb whose main purpose is to track how a match was
5215 arrived at, though it also has a secondary use in conjunction with
5216 advancing the match starting point (see (*SKIP) below).
5218 (*MARK:NAME) or (*:NAME)
5220 A name is always required with this verb. There may be as many
5221 instances of (*MARK) as you like in a pattern, and their names do not
5222 have to be unique.
5224 When a match succeeds, the name of the last-encountered (*MARK) is
5225 passed back to the caller via the pcre_extra data structure, as
5226 described in the section on pcre_extra in the pcreapi documentation. No
5227 data is returned for a partial match. Here is an example of pcretest
5228 output, where the /K modifier requests the retrieval and outputting of
5229 (*MARK) data:
5231 /X(*MARK:A)Y|X(*MARK:B)Z/K
5232 XY
5233 0: XY
5234 MK: A
5235 XZ
5236 0: XZ
5237 MK: B
5239 The (*MARK) name is tagged with "MK:" in this output, and in this exam-
5240 ple it indicates which of the two alternatives matched. This is a more
5241 efficient way of obtaining this information than putting each alterna-
5242 tive in its own capturing parentheses.
5244 A name may also be returned after a failed match if the final path
5245 through the pattern involves (*MARK). However, unless (*MARK) used in
5246 conjunction with (*COMMIT), this is unlikely to happen for an unan-
5247 chored pattern because, as the starting point for matching is advanced,
5248 the final check is often with an empty string, causing a failure before
5249 (*MARK) is reached. For example:
5251 /X(*MARK:A)Y|X(*MARK:B)Z/K
5252 XP
5253 No match
5255 There are three potential starting points for this match (starting with
5256 X, starting with P, and with an empty string). If the pattern is
5257 anchored, the result is different:
5259 /^X(*MARK:A)Y|^X(*MARK:B)Z/K
5260 XP
5261 No match, mark = B
5263 PCRE's start-of-match optimizations can also interfere with this. For
5264 example, if, as a result of a call to pcre_study(), it knows the mini-
5265 mum subject length for a match, a shorter subject will not be scanned
5266 at all.
5268 Note that similar anomalies (though different in detail) exist in Perl,
5269 no doubt for the same reasons. The use of (*MARK) data after a failed
5270 match of an unanchored pattern is not recommended, unless (*COMMIT) is
5271 involved.
5273 Verbs that act after backtracking
5275 The following verbs do nothing when they are encountered. Matching con-
5276 tinues with what follows, but if there is no subsequent match, causing
5277 a backtrack to the verb, a failure is forced. That is, backtracking
5278 cannot pass to the left of the verb. However, when one of these verbs
5279 appears inside an atomic group, its effect is confined to that group,
5280 because once the group has been matched, there is never any backtrack-
5281 ing into it. In this situation, backtracking can "jump back" to the
5282 left of the entire atomic group. (Remember also, as stated above, that
5283 this localization also applies in subroutine calls and assertions.)
5285 These verbs differ in exactly what kind of failure occurs when back-
5286 tracking reaches them.
5288 (*COMMIT)
5290 This verb, which may not be followed by a name, causes the whole match
5291 to fail outright if the rest of the pattern does not match. Even if the
5292 pattern is unanchored, no further attempts to find a match by advancing
5293 the starting point take place. Once (*COMMIT) has been passed,
5294 pcre_exec() is committed to finding a match at the current starting
5295 point, or not at all. For example:
5297 a+(*COMMIT)b
5299 This matches "xxaab" but not "aacaab". It can be thought of as a kind
5300 of dynamic anchor, or "I've started, so I must finish." The name of the
5301 most recently passed (*MARK) in the path is passed back when (*COMMIT)
5302 forces a match failure.
5304 Note that (*COMMIT) at the start of a pattern is not the same as an
5305 anchor, unless PCRE's start-of-match optimizations are turned off, as
5306 shown in this pcretest example:
5308 /(*COMMIT)abc/
5309 xyzabc
5310 0: abc
5311 xyzabc\Y
5312 No match
5314 PCRE knows that any match must start with "a", so the optimization
5315 skips along the subject to "a" before running the first match attempt,
5316 which succeeds. When the optimization is disabled by the \Y escape in
5317 the second subject, the match starts at "x" and so the (*COMMIT) causes
5318 it to fail without trying any other starting points.
5320 (*PRUNE) or (*PRUNE:NAME)
5322 This verb causes the match to fail at the current starting position in
5323 the subject if the rest of the pattern does not match. If the pattern
5324 is unanchored, the normal "bumpalong" advance to the next starting
5325 character then happens. Backtracking can occur as usual to the left of
5326 (*PRUNE), before it is reached, or when matching to the right of
5327 (*PRUNE), but if there is no match to the right, backtracking cannot
5328 cross (*PRUNE). In simple cases, the use of (*PRUNE) is just an alter-
5329 native to an atomic group or possessive quantifier, but there are some
5330 uses of (*PRUNE) that cannot be expressed in any other way. The behav-
5331 iour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE) when the
5332 match fails completely; the name is passed back if this is the final
5333 attempt. (*PRUNE:NAME) does not pass back a name if the match suc-
5334 ceeds. In an anchored pattern (*PRUNE) has the same effect as (*COM-
5335 MIT).
5337 (*SKIP)
5339 This verb, when given without a name, is like (*PRUNE), except that if
5340 the pattern is unanchored, the "bumpalong" advance is not to the next
5341 character, but to the position in the subject where (*SKIP) was encoun-
5342 tered. (*SKIP) signifies that whatever text was matched leading up to
5343 it cannot be part of a successful match. Consider:
5345 a+(*SKIP)b
5347 If the subject is "aaaac...", after the first match attempt fails
5348 (starting at the first character in the string), the starting point
5349 skips on to start the next attempt at "c". Note that a possessive quan-
5350 tifer does not have the same effect as this example; although it would
5351 suppress backtracking during the first match attempt, the second
5352 attempt would start at the second character instead of skipping on to
5353 "c".
5355 (*SKIP:NAME)
5357 When (*SKIP) has an associated name, its behaviour is modified. If the
5358 following pattern fails to match, the previous path through the pattern
5359 is searched for the most recent (*MARK) that has the same name. If one
5360 is found, the "bumpalong" advance is to the subject position that cor-
5361 responds to that (*MARK) instead of to where (*SKIP) was encountered.
5362 If no (*MARK) with a matching name is found, normal "bumpalong" of one
5363 character happens (the (*SKIP) is ignored).
5365 (*THEN) or (*THEN:NAME)
5367 This verb causes a skip to the next alternation if the rest of the pat-
5368 tern does not match. That is, it cancels pending backtracking, but only
5369 within the current alternation. Its name comes from the observation
5370 that it can be used for a pattern-based if-then-else block:
5372 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
5374 If the COND1 pattern matches, FOO is tried (and possibly further items
5375 after the end of the group if FOO succeeds); on failure the matcher
5376 skips to the second alternative and tries COND2, without backtracking
5377 into COND1. The behaviour of (*THEN:NAME) is exactly the same as
5378 (*MARK:NAME)(*THEN) if the overall match fails. If (*THEN) is not
5379 directly inside an alternation, it acts like (*PRUNE).
5384 pcreapi(3), pcrecallout(3), pcrematching(3), pcresyntax(3), pcre(3).
5389 Philip Hazel
5390 University Computing Service
5391 Cambridge CB2 3QH, England.
5396 Last updated: 27 March 2010
5397 Copyright (c) 1997-2010 University of Cambridge.
5398 ------------------------------------------------------------------------------
5404 NAME
5405 PCRE - Perl-compatible regular expressions
5410 The full syntax and semantics of the regular expressions that are sup-
5411 ported by PCRE are described in the pcrepattern documentation. This
5412 document contains just a quick-reference summary of the syntax.
5417 \x where x is non-alphanumeric is a literal x
5418 \Q...\E treat enclosed characters as literal
5423 \a alarm, that is, the BEL character (hex 07)
5424 \cx "control-x", where x is any character
5425 \e escape (hex 1B)
5426 \f formfeed (hex 0C)
5427 \n newline (hex 0A)
5428 \r carriage return (hex 0D)
5429 \t tab (hex 09)
5430 \ddd character with octal code ddd, or backreference
5431 \xhh character with hex code hh
5432 \x{hhh..} character with hex code hhh..
5437 . any character except newline;
5438 in dotall mode, any character whatsoever
5439 \C one byte, even in UTF-8 mode (best avoided)
5440 \d a decimal digit
5441 \D a character that is not a decimal digit
5442 \h a horizontal whitespace character
5443 \H a character that is not a horizontal whitespace character
5444 \p{xx} a character with the xx property
5445 \P{xx} a character without the xx property
5446 \R a newline sequence
5447 \s a whitespace character
5448 \S a character that is not a whitespace character
5449 \v a vertical whitespace character
5450 \V a character that is not a vertical whitespace character
5451 \w a "word" character
5452 \W a "non-word" character
5453 \X an extended Unicode sequence
5455 In PCRE, \d, \D, \s, \S, \w, and \W recognize only ASCII characters.
5460 C Other
5461 Cc Control
5462 Cf Format
5463 Cn Unassigned
5464 Co Private use
5465 Cs Surrogate
5467 L Letter
5468 Ll Lower case letter
5469 Lm Modifier letter
5470 Lo Other letter
5471 Lt Title case letter
5472 Lu Upper case letter
5473 L& Ll, Lu, or Lt
5475 M Mark
5476 Mc Spacing mark
5477 Me Enclosing mark
5478 Mn Non-spacing mark
5480 N Number
5481 Nd Decimal number
5482 Nl Letter number
5483 No Other number
5485 P Punctuation
5486 Pc Connector punctuation
5487 Pd Dash punctuation
5488 Pe Close punctuation
5489 Pf Final punctuation
5490 Pi Initial punctuation
5491 Po Other punctuation
5492 Ps Open punctuation
5494 S Symbol
5495 Sc Currency symbol
5496 Sk Modifier symbol
5497 Sm Mathematical symbol
5498 So Other symbol
5500 Z Separator
5501 Zl Line separator
5502 Zp Paragraph separator
5503 Zs Space separator
5508 Arabic, Armenian, Avestan, Balinese, Bamum, Bengali, Bopomofo, Braille,
5509 Buginese, Buhid, Canadian_Aboriginal, Carian, Cham, Cherokee, Common,
5510 Coptic, Cuneiform, Cypriot, Cyrillic, Deseret, Devanagari, Egyp-
5511 tian_Hieroglyphs, Ethiopic, Georgian, Glagolitic, Gothic, Greek,
5512 Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana, Impe-
5513 rial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscriptional_Parthian,
5514 Javanese, Kaithi, Kannada, Katakana, Kayah_Li, Kharoshthi, Khmer, Lao,
5515 Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian, Lydian, Malayalam,
5516 Meetei_Mayek, Mongolian, Myanmar, New_Tai_Lue, Nko, Ogham, Old_Italic,
5517 Old_Persian, Old_South_Arabian, Old_Turkic, Ol_Chiki, Oriya, Osmanya,
5518 Phags_Pa, Phoenician, Rejang, Runic, Samaritan, Saurashtra, Shavian,
5519 Sinhala, Sundanese, Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le,
5520 Tai_Tham, Tai_Viet, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh,
5521 Ugaritic, Vai, Yi.
5526 [...] positive character class
5527 [^...] negative character class
5528 [x-y] range (can be used for hex characters)
5529 [[:xxx:]] positive POSIX named set
5530 [[:^xxx:]] negative POSIX named set
5532 alnum alphanumeric
5533 alpha alphabetic
5534 ascii 0-127
5535 blank space or tab
5536 cntrl control character
5537 digit decimal digit
5538 graph printing, excluding space
5539 lower lower case letter
5540 print printing, including space
5541 punct printing, excluding alphanumeric
5542 space whitespace
5543 upper upper case letter
5544 word same as \w
5545 xdigit hexadecimal digit
5547 In PCRE, POSIX character set names recognize only ASCII characters. You
5548 can use \Q...\E inside a character class.
5553 ? 0 or 1, greedy
5554 ?+ 0 or 1, possessive
5555 ?? 0 or 1, lazy
5556 * 0 or more, greedy
5557 *+ 0 or more, possessive
5558 *? 0 or more, lazy
5559 + 1 or more, greedy
5560 ++ 1 or more, possessive
5561 +? 1 or more, lazy
5562 {n} exactly n
5563 {n,m} at least n, no more than m, greedy
5564 {n,m}+ at least n, no more than m, possessive
5565 {n,m}? at least n, no more than m, lazy
5566 {n,} n or more, greedy
5567 {n,}+ n or more, possessive
5568 {n,}? n or more, lazy
5573 \b word boundary (only ASCII letters recognized)
5574 \B not a word boundary
5575 ^ start of subject
5576 also after internal newline in multiline mode
5577 \A start of subject
5578 $ end of subject
5579 also before newline at end of subject
5580 also before internal newline in multiline mode
5581 \Z end of subject
5582 also before newline at end of subject
5583 \z end of subject
5584 \G first matching position in subject
5589 \K reset start of match
5594 expr|expr|expr...
5599 (...) capturing group
5600 (?<name>...) named capturing group (Perl)
5601 (?'name'...) named capturing group (Perl)
5602 (?P<name>...) named capturing group (Python)
5603 (?:...) non-capturing group
5604 (?|...) non-capturing group; reset group numbers for
5605 capturing groups in each alternative
5610 (?>...) atomic, non-capturing group
5615 (?#....) comment (not nestable)
5620 (?i) caseless
5621 (?J) allow duplicate names
5622 (?m) multiline
5623 (?s) single line (dotall)
5624 (?U) default ungreedy (lazy)
5625 (?x) extended (ignore white space)
5626 (?-...) unset option(s)
5628 The following is recognized only at the start of a pattern or after one
5629 of the newline-setting options with similar syntax:
5631 (*UTF8) set UTF-8 mode
5636 (?=...) positive look ahead
5637 (?!...) negative look ahead
5638 (?<=...) positive look behind
5639 (?<!...) negative look behind
5641 Each top-level branch of a look behind must be of a fixed length.
5646 \n reference by number (can be ambiguous)
5647 \gn reference by number
5648 \g{n} reference by number
5649 \g{-n} relative reference by number
5650 \k<name> reference by name (Perl)
5651 \k'name' reference by name (Perl)
5652 \g{name} reference by name (Perl)
5653 \k{name} reference by name (.NET)
5654 (?P=name) reference by name (Python)
5659 (?R) recurse whole pattern
5660 (?n) call subpattern by absolute number
5661 (?+n) call subpattern by relative number
5662 (?-n) call subpattern by relative number
5663 (?&name) call subpattern by name (Perl)
5664 (?P>name) call subpattern by name (Python)
5665 \g<name> call subpattern by name (Oniguruma)
5666 \g'name' call subpattern by name (Oniguruma)
5667 \g<n> call subpattern by absolute number (Oniguruma)
5668 \g'n' call subpattern by absolute number (Oniguruma)
5669 \g<+n> call subpattern by relative number (PCRE extension)
5670 \g'+n' call subpattern by relative number (PCRE extension)
5671 \g<-n> call subpattern by relative number (PCRE extension)
5672 \g'-n' call subpattern by relative number (PCRE extension)
5677 (?(condition)yes-pattern)
5678 (?(condition)yes-pattern|no-pattern)
5680 (?(n)... absolute reference condition
5681 (?(+n)... relative reference condition
5682 (?(-n)... relative reference condition
5683 (?(<name>)... named reference condition (Perl)
5684 (?('name')... named reference condition (Perl)
5685 (?(name)... named reference condition (PCRE)
5686 (?(R)... overall recursion condition
5687 (?(Rn)... specific group recursion condition
5688 (?(R&name)... specific recursion condition
5689 (?(DEFINE)... define subpattern for reference
5690 (?(assert)... assertion condition
5695 The following act immediately they are reached:
5697 (*ACCEPT) force successful match
5698 (*FAIL) force backtrack; synonym (*F)
5700 The following act only when a subsequent match failure causes a back-
5701 track to reach them. They all force a match failure, but they differ in
5702 what happens afterwards. Those that advance the start-of-match point do
5703 so only if the pattern is not anchored.
5705 (*COMMIT) overall failure, no advance of starting point
5706 (*PRUNE) advance to next starting character
5707 (*SKIP) advance start to current matching position
5708 (*THEN) local failure, backtrack to next alternation
5713 These are recognized only at the very start of the pattern or after a
5714 (*BSR_...) or (*UTF8) option.
5716 (*CR) carriage return only
5717 (*LF) linefeed only
5718 (*CRLF) carriage return followed by linefeed
5719 (*ANYCRLF) all three of the above
5720 (*ANY) any Unicode newline sequence
5725 These are recognized only at the very start of the pattern or after a
5726 (*...) option that sets the newline convention or UTF-8 mode.
5729 (*BSR_UNICODE) any Unicode newline sequence
5734 (?C) callout
5735 (?Cn) callout with data n
5740 pcrepattern(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
5745 Philip Hazel
5746 University Computing Service
5747 Cambridge CB2 3QH, England.
5752 Last updated: 01 March 2010
5753 Copyright (c) 1997-2010 University of Cambridge.
5754 ------------------------------------------------------------------------------
5760 NAME
5761 PCRE - Perl-compatible regular expressions
5766 In normal use of PCRE, if the subject string that is passed to
5767 pcre_exec() or pcre_dfa_exec() matches as far as it goes, but is too
5768 short to match the entire pattern, PCRE_ERROR_NOMATCH is returned.
5769 There are circumstances where it might be helpful to distinguish this
5770 case from other cases in which there is no match.
5772 Consider, for example, an application where a human is required to type
5773 in data for a field with specific formatting requirements. An example
5774 might be a date in the form ddmmmyy, defined by this pattern:
5776 ^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$
5778 If the application sees the user's keystrokes one by one, and can check
5779 that what has been typed so far is potentially valid, it is able to
5780 raise an error as soon as a mistake is made, by beeping and not
5781 reflecting the character that has been typed, for example. This immedi-
5782 ate feedback is likely to be a better user interface than a check that
5783 is delayed until the entire string has been entered. Partial matching
5784 can also sometimes be useful when the subject string is very long and
5785 is not all available at once.
5787 PCRE supports partial matching by means of the PCRE_PARTIAL_SOFT and
5788 PCRE_PARTIAL_HARD options, which can be set when calling pcre_exec() or
5789 pcre_dfa_exec(). For backwards compatibility, PCRE_PARTIAL is a synonym
5790 for PCRE_PARTIAL_SOFT. The essential difference between the two options
5791 is whether or not a partial match is preferred to an alternative com-
5792 plete match, though the details differ between the two matching func-
5793 tions. If both options are set, PCRE_PARTIAL_HARD takes precedence.
5795 Setting a partial matching option disables two of PCRE's optimizations.
5796 PCRE remembers the last literal byte in a pattern, and abandons match-
5797 ing immediately if such a byte is not present in the subject string.
5798 This optimization cannot be used for a subject string that might match
5799 only partially. If the pattern was studied, PCRE knows the minimum
5800 length of a matching string, and does not bother to run the matching
5801 function on shorter strings. This optimization is also disabled for
5802 partial matching.
5807 A partial match occurs during a call to pcre_exec() whenever the end of
5808 the subject string is reached successfully, but matching cannot con-
5809 tinue because more characters are needed. However, at least one charac-
5810 ter must have been matched. (In other words, a partial match can never
5811 be an empty string.)
5813 If PCRE_PARTIAL_SOFT is set, the partial match is remembered, but
5814 matching continues as normal, and other alternatives in the pattern are
5815 tried. If no complete match can be found, pcre_exec() returns
5816 PCRE_ERROR_PARTIAL instead of PCRE_ERROR_NOMATCH. If there are at least
5817 two slots in the offsets vector, the first of them is set to the offset
5818 of the earliest character that was inspected when the partial match was
5819 found. For convenience, the second offset points to the end of the
5820 string so that a substring can easily be identified.
5822 For the majority of patterns, the first offset identifies the start of
5823 the partially matched string. However, for patterns that contain look-
5824 behind assertions, or \K, or begin with \b or \B, earlier characters
5825 have been inspected while carrying out the match. For example:
5827 /(?<=abc)123/
5829 This pattern matches "123", but only if it is preceded by "abc". If the
5830 subject string is "xyzabc12", the offsets after a partial match are for
5831 the substring "abc12", because all these characters are needed if
5832 another match is tried with extra characters added.
5834 If there is more than one partial match, the first one that was found
5835 provides the data that is returned. Consider this pattern:
5837 /123\w+X|dogY/
5839 If this is matched against the subject string "abc123dog", both alter-
5840 natives fail to match, but the end of the subject is reached during
5841 matching, so PCRE_ERROR_PARTIAL is returned instead of
5842 PCRE_ERROR_NOMATCH. The offsets are set to 3 and 9, identifying
5843 "123dog" as the first partial match that was found. (In this example,
5844 there are two partial matches, because "dog" on its own partially
5845 matches the second alternative.)
5847 If PCRE_PARTIAL_HARD is set for pcre_exec(), it returns PCRE_ERROR_PAR-
5848 TIAL as soon as a partial match is found, without continuing to search
5849 for possible complete matches. The difference between the two options
5850 can be illustrated by a pattern such as:
5852 /dog(sbody)?/
5854 This matches either "dog" or "dogsbody", greedily (that is, it prefers
5855 the longer string if possible). If it is matched against the string
5856 "dog" with PCRE_PARTIAL_SOFT, it yields a complete match for "dog".
5857 However, if PCRE_PARTIAL_HARD is set, the result is PCRE_ERROR_PARTIAL.
5858 On the other hand, if the pattern is made ungreedy the result is dif-
5859 ferent:
5861 /dog(sbody)??/
5863 In this case the result is always a complete match because pcre_exec()
5864 finds that first, and it never continues after finding a match. It
5865 might be easier to follow this explanation by thinking of the two pat-
5866 terns like this:
5868 /dog(sbody)?/ is the same as /dogsbody|dog/
5869 /dog(sbody)??/ is the same as /dog|dogsbody/
5871 The second pattern will never match "dogsbody" when pcre_exec() is
5872 used, because it will always find the shorter match first.
5875 PARTIAL MATCHING USING pcre_dfa_exec()
5877 The pcre_dfa_exec() function moves along the subject string character
5878 by character, without backtracking, searching for all possible matches
5879 simultaneously. If the end of the subject is reached before the end of
5880 the pattern, there is the possibility of a partial match, again pro-
5881 vided that at least one character has matched.
5883 When PCRE_PARTIAL_SOFT is set, PCRE_ERROR_PARTIAL is returned only if
5884 there have been no complete matches. Otherwise, the complete matches
5885 are returned. However, if PCRE_PARTIAL_HARD is set, a partial match
5886 takes precedence over any complete matches. The portion of the string
5887 that was inspected when the longest partial match was found is set as
5888 the first matching string, provided there are at least two slots in the
5889 offsets vector.
5891 Because pcre_dfa_exec() always searches for all possible matches, and
5892 there is no difference between greedy and ungreedy repetition, its be-
5893 haviour is different from pcre_exec when PCRE_PARTIAL_HARD is set. Con-
5894 sider the string "dog" matched against the ungreedy pattern shown
5895 above:
5897 /dog(sbody)??/
5899 Whereas pcre_exec() stops as soon as it finds the complete match for
5900 "dog", pcre_dfa_exec() also finds the partial match for "dogsbody", and
5901 so returns that when PCRE_PARTIAL_HARD is set.
5906 If a pattern ends with one of sequences \b or \B, which test for word
5907 boundaries, partial matching with PCRE_PARTIAL_SOFT can give counter-
5908 intuitive results. Consider this pattern:
5910 /\bcat\b/
5912 This matches "cat", provided there is a word boundary at either end. If
5913 the subject string is "the cat", the comparison of the final "t" with a
5914 following character cannot take place, so a partial match is found.
5915 However, pcre_exec() carries on with normal matching, which matches \b
5916 at the end of the subject when the last character is a letter, thus
5917 finding a complete match. The result, therefore, is not PCRE_ERROR_PAR-
5918 TIAL. The same thing happens with pcre_dfa_exec(), because it also
5919 finds the complete match.
5921 Using PCRE_PARTIAL_HARD in this case does yield PCRE_ERROR_PARTIAL,
5922 because then the partial match takes precedence.
5927 For releases of PCRE prior to 8.00, because of the way certain internal
5928 optimizations were implemented in the pcre_exec() function, the
5929 PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be
5930 used with all patterns. From release 8.00 onwards, the restrictions no
5931 longer apply, and partial matching with pcre_exec() can be requested
5932 for any pattern.
5934 Items that were formerly restricted were repeated single characters and
5935 repeated metasequences. If PCRE_PARTIAL was set for a pattern that did
5936 not conform to the restrictions, pcre_exec() returned the error code
5937 PCRE_ERROR_BADPARTIAL (-13). This error code is no longer in use. The
5938 PCRE_INFO_OKPARTIAL call to pcre_fullinfo() to find out if a compiled
5939 pattern can be used for partial matching now always returns 1.
5944 If the escape sequence \P is present in a pcretest data line, the
5945 PCRE_PARTIAL_SOFT option is used for the match. Here is a run of
5946 pcretest that uses the date example quoted above:
5948 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
5949 data> 25jun04\P
5950 0: 25jun04
5951 1: jun
5952 data> 25dec3\P
5953 Partial match: 23dec3
5954 data> 3ju\P
5955 Partial match: 3ju
5956 data> 3juj\P
5957 No match
5958 data> j\P
5959 No match
5961 The first data string is matched completely, so pcretest shows the
5962 matched substrings. The remaining four strings do not match the com-
5963 plete pattern, but the first two are partial matches. Similar output is
5964 obtained when pcre_dfa_exec() is used.
5966 If the escape sequence \P is present more than once in a pcretest data
5967 line, the PCRE_PARTIAL_HARD option is set for the match.
5970 MULTI-SEGMENT MATCHING WITH pcre_dfa_exec()
5972 When a partial match has been found using pcre_dfa_exec(), it is possi-
5973 ble to continue the match by providing additional subject data and
5974 calling pcre_dfa_exec() again with the same compiled regular expres-
5975 sion, this time setting the PCRE_DFA_RESTART option. You must pass the
5976 same working space as before, because this is where details of the pre-
5977 vious partial match are stored. Here is an example using pcretest,
5978 using the \R escape sequence to set the PCRE_DFA_RESTART option (\D
5979 specifies the use of pcre_dfa_exec()):
5981 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
5982 data> 23ja\P\D
5983 Partial match: 23ja
5984 data> n05\R\D
5985 0: n05
5987 The first call has "23ja" as the subject, and requests partial match-
5988 ing; the second call has "n05" as the subject for the continued
5989 (restarted) match. Notice that when the match is complete, only the
5990 last part is shown; PCRE does not retain the previously partially-
5991 matched string. It is up to the calling program to do that if it needs
5992 to.
5994 You can set the PCRE_PARTIAL_SOFT or PCRE_PARTIAL_HARD options with
5995 PCRE_DFA_RESTART to continue partial matching over multiple segments.
5996 This facility can be used to pass very long subject strings to
5997 pcre_dfa_exec().
6002 From release 8.00, pcre_exec() can also be used to do multi-segment
6003 matching. Unlike pcre_dfa_exec(), it is not possible to restart the
6004 previous match with a new segment of data. Instead, new data must be
6005 added to the previous subject string, and the entire match re-run,
6006 starting from the point where the partial match occurred. Earlier data
6007 can be discarded. Consider an unanchored pattern that matches dates:
6009 re> /\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d/
6010 data> The date is 23ja\P
6011 Partial match: 23ja
6013 At this stage, an application could discard the text preceding "23ja",
6014 add on text from the next segment, and call pcre_exec() again. Unlike
6015 pcre_dfa_exec(), the entire matching string must always be available,
6016 and the complete matching process occurs for each call, so more memory
6017 and more processing time is needed.
6019 Note: If the pattern contains lookbehind assertions, or \K, or starts
6020 with \b or \B, the string that is returned for a partial match will
6021 include characters that precede the partially matched string itself,
6022 because these must be retained when adding on more characters for a
6023 subsequent matching attempt.
6028 Certain types of pattern may give problems with multi-segment matching,
6029 whichever matching function is used.
6031 1. If the pattern contains tests for the beginning or end of a line,
6032 you need to pass the PCRE_NOTBOL or PCRE_NOTEOL options, as appropri-
6033 ate, when the subject string for any call does not contain the begin-
6034 ning or end of a line.
6036 2. Lookbehind assertions at the start of a pattern are catered for in
6037 the offsets that are returned for a partial match. However, in theory,
6038 a lookbehind assertion later in the pattern could require even earlier
6039 characters to be inspected, and it might not have been reached when a
6040 partial match occurs. This is probably an extremely unlikely case; you
6041 could guard against it to a certain extent by always including extra
6042 characters at the start.
6044 3. Matching a subject string that is split into multiple segments may
6045 not always produce exactly the same result as matching over one single
6046 long string, especially when PCRE_PARTIAL_SOFT is used. The section
6047 "Partial Matching and Word Boundaries" above describes an issue that
6048 arises if the pattern ends with \b or \B. Another kind of difference
6049 may occur when there are multiple matching possibilities, because a
6050 partial match result is given only when there are no completed matches.
6051 This means that as soon as the shortest match has been found, continua-
6052 tion to a new subject segment is no longer possible. Consider again
6053 this pcretest example:
6055 re> /dog(sbody)?/
6056 data> dogsb\P
6057 0: dog
6058 data>