/[pcre]/code/trunk/doc/pcre.txt
ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 429 - (show annotations)
Tue Sep 1 16:10:16 2009 UTC (10 years, 1 month ago) by ph10
File MIME type: text/plain
File size: 321084 byte(s)
Add pcredemo man page, containing a listing of pcredemo.c.
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 -----------------------------------------------------------------------------
9
10
11 PCRE(3) PCRE(3)
12
13
14 NAME
15 PCRE - Perl-compatible regular expressions
16
17
18 INTRODUCTION
19
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. Certain features that appeared in Python and
23 PCRE before they appeared in Perl are also available using the Python
24 syntax. There is also some support for certain .NET and Oniguruma syn-
25 tax items, and there is an option for requesting some minor changes
26 that give better JavaScript compatibility.
27
28 The current implementation of PCRE (release 8.xx) corresponds approxi-
29 mately with Perl 5.10, including support for UTF-8 encoded strings and
30 Unicode general category properties. However, UTF-8 and Unicode support
31 has to be explicitly enabled; it is not the default. The Unicode tables
32 correspond to Unicode release 5.1.
33
34 In addition to the Perl-compatible matching function, PCRE contains an
35 alternative matching function that matches the same compiled patterns
36 in a different way. In certain circumstances, the alternative function
37 has some advantages. For a discussion of the two matching algorithms,
38 see the pcrematching page.
39
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:
46
47 ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre
48
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.
53
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 file
59 in the source distribution.
60
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.
68
69
70 USER DOCUMENTATION
71
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:
78
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 pcresyntax quick syntax reference
93 pcreperform discussion of performance issues
94 pcreposix the POSIX-compatible C API
95 pcreprecompile details of saving and re-using precompiled patterns
96 pcresample discussion of the pcredemo program
97 pcrestack discussion of stack usage
98 pcretest description of the pcretest testing command
99
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.
102
103
104 LIMITATIONS
105
106 There are some size limitations in PCRE but it is hoped that they will
107 never in practice be relevant.
108
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.
116
117 All values in repeating quantifiers must be less than 65536.
118
119 There is no limit to the number of parenthesized subpatterns, but there
120 can be no more than 65535 capturing subpatterns.
121
122 The maximum length of name for a named subpattern is 32 characters, and
123 the maximum number of named subpatterns is 10000.
124
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.
131
132
133 UTF-8 AND UNICODE PROPERTY SUPPORT
134
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.
139
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 just strings of bytes.
146
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.
151
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.
163
164 Validity of UTF-8 strings
165
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.
174
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.)
183
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.
191
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.
200
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.
205
206 General comments about UTF-8 mode
207
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.
210
211 2. Octal numbers up to \777 are recognized, and match two-byte UTF-8
212 characters for values greater than \177.
213
214 3. Repeat quantifiers apply to complete UTF-8 characters, not to indi-
215 vidual bytes, for example: \x{100}{3}.
216
217 4. The dot metacharacter matches one UTF-8 character instead of a sin-
218 gle byte.
219
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().
223
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.
233
234 7. Similarly, characters that match the POSIX named character classes
235 are all low-valued characters.
236
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.
240
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.
251
252
253 AUTHOR
254
255 Philip Hazel
256 University Computing Service
257 Cambridge CB2 3QH, England.
258
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.
262
263
264 REVISION
265
266 Last updated: 01 September 2009
267 Copyright (c) 1997-2009 University of Cambridge.
268 ------------------------------------------------------------------------------
269
270
271 PCREBUILD(3) PCREBUILD(3)
272
273
274 NAME
275 PCRE - Perl-compatible regular expressions
276
277
278 PCRE BUILD-TIME OPTIONS
279
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 CMakeSetup if you are using
286 CMake instead of configure to build PCRE.
287
288 The complete list of options for configure (which includes the standard
289 ones such as the selection of the installation directory) can be
290 obtained by running
291
292 ./configure --help
293
294 The following sections include descriptions of options whose names
295 begin with --enable or --disable. These settings specify changes to the
296 defaults for the configure command. Because of the way that configure
297 works, --enable and --disable always come in pairs, so the complemen-
298 tary option always exists as well, but as it specifies the default, it
299 is not described.
300
301
302 C++ SUPPORT
303
304 By default, the configure script will search for a C++ compiler and C++
305 header files. If it finds them, it automatically builds the C++ wrapper
306 library for PCRE. You can disable this by adding
307
308 --disable-cpp
309
310 to the configure command.
311
312
313 UTF-8 SUPPORT
314
315 To build PCRE with support for UTF-8 Unicode character strings, add
316
317 --enable-utf8
318
319 to the configure command. Of itself, this does not make PCRE treat
320 strings as UTF-8. As well as compiling PCRE with this option, you also
321 have have to set the PCRE_UTF8 option when you call the pcre_compile()
322 function.
323
324 If you set --enable-utf8 when compiling in an EBCDIC environment, PCRE
325 expects its input to be either ASCII or UTF-8 (depending on the runtime
326 option). It is not possible to support both EBCDIC and UTF-8 codes in
327 the same version of the library. Consequently, --enable-utf8 and
328 --enable-ebcdic are mutually exclusive.
329
330
331 UNICODE CHARACTER PROPERTY SUPPORT
332
333 UTF-8 support allows PCRE to process character values greater than 255
334 in the strings that it handles. On its own, however, it does not pro-
335 vide any facilities for accessing the properties of such characters. If
336 you want to be able to use the pattern escapes \P, \p, and \X, which
337 refer to Unicode character properties, you must add
338
339 --enable-unicode-properties
340
341 to the configure command. This implies UTF-8 support, even if you have
342 not explicitly requested it.
343
344 Including Unicode property support adds around 30K of tables to the
345 PCRE library. Only the general category properties such as Lu and Nd
346 are supported. Details are given in the pcrepattern documentation.
347
348
349 CODE VALUE OF NEWLINE
350
351 By default, PCRE interprets the linefeed (LF) character as indicating
352 the end of a line. This is the normal newline character on Unix-like
353 systems. You can compile PCRE to use carriage return (CR) instead, by
354 adding
355
356 --enable-newline-is-cr
357
358 to the configure command. There is also a --enable-newline-is-lf
359 option, which explicitly specifies linefeed as the newline character.
360
361 Alternatively, you can specify that line endings are to be indicated by
362 the two character sequence CRLF. If you want this, add
363
364 --enable-newline-is-crlf
365
366 to the configure command. There is a fourth option, specified by
367
368 --enable-newline-is-anycrlf
369
370 which causes PCRE to recognize any of the three sequences CR, LF, or
371 CRLF as indicating a line ending. Finally, a fifth option, specified by
372
373 --enable-newline-is-any
374
375 causes PCRE to recognize any Unicode newline sequence.
376
377 Whatever line ending convention is selected when PCRE is built can be
378 overridden when the library functions are called. At build time it is
379 conventional to use the standard for your operating system.
380
381
382 WHAT \R MATCHES
383
384 By default, the sequence \R in a pattern matches any Unicode newline
385 sequence, whatever has been selected as the line ending sequence. If
386 you specify
387
388 --enable-bsr-anycrlf
389
390 the default is changed so that \R matches only CR, LF, or CRLF. What-
391 ever is selected when PCRE is built can be overridden when the library
392 functions are called.
393
394
395 BUILDING SHARED AND STATIC LIBRARIES
396
397 The PCRE building process uses libtool to build both shared and static
398 Unix libraries by default. You can suppress one of these by adding one
399 of
400
401 --disable-shared
402 --disable-static
403
404 to the configure command, as required.
405
406
407 POSIX MALLOC USAGE
408
409 When PCRE is called through the POSIX interface (see the pcreposix doc-
410 umentation), additional working storage is required for holding the
411 pointers to capturing substrings, because PCRE requires three integers
412 per substring, whereas the POSIX interface provides only two. If the
413 number of expected substrings is small, the wrapper function uses space
414 on the stack, because this is faster than using malloc() for each call.
415 The default threshold above which the stack is no longer used is 10; it
416 can be changed by adding a setting such as
417
418 --with-posix-malloc-threshold=20
419
420 to the configure command.
421
422
423 HANDLING VERY LARGE PATTERNS
424
425 Within a compiled pattern, offset values are used to point from one
426 part to another (for example, from an opening parenthesis to an alter-
427 nation metacharacter). By default, two-byte values are used for these
428 offsets, leading to a maximum size for a compiled pattern of around
429 64K. This is sufficient to handle all but the most gigantic patterns.
430 Nevertheless, some people do want to process enormous patterns, so it
431 is possible to compile PCRE to use three-byte or four-byte offsets by
432 adding a setting such as
433
434 --with-link-size=3
435
436 to the configure command. The value given must be 2, 3, or 4. Using
437 longer offsets slows down the operation of PCRE because it has to load
438 additional bytes when handling them.
439
440
441 AVOIDING EXCESSIVE STACK USAGE
442
443 When matching with the pcre_exec() function, PCRE implements backtrack-
444 ing by making recursive calls to an internal function called match().
445 In environments where the size of the stack is limited, this can se-
446 verely limit PCRE's operation. (The Unix environment does not usually
447 suffer from this problem, but it may sometimes be necessary to increase
448 the maximum stack size. There is a discussion in the pcrestack docu-
449 mentation.) An alternative approach to recursion that uses memory from
450 the heap to remember data, instead of using recursive function calls,
451 has been implemented to work round the problem of limited stack size.
452 If you want to build a version of PCRE that works this way, add
453
454 --disable-stack-for-recursion
455
456 to the configure command. With this configuration, PCRE will use the
457 pcre_stack_malloc and pcre_stack_free variables to call memory manage-
458 ment functions. By default these point to malloc() and free(), but you
459 can replace the pointers so that your own functions are used.
460
461 Separate functions are provided rather than using pcre_malloc and
462 pcre_free because the usage is very predictable: the block sizes
463 requested are always the same, and the blocks are always freed in
464 reverse order. A calling program might be able to implement optimized
465 functions that perform better than malloc() and free(). PCRE runs
466 noticeably more slowly when built in this way. This option affects only
467 the pcre_exec() function; it is not relevant for the the
468 pcre_dfa_exec() function.
469
470
471 LIMITING PCRE RESOURCE USAGE
472
473 Internally, PCRE has a function called match(), which it calls repeat-
474 edly (sometimes recursively) when matching a pattern with the
475 pcre_exec() function. By controlling the maximum number of times this
476 function may be called during a single matching operation, a limit can
477 be placed on the resources used by a single call to pcre_exec(). The
478 limit can be changed at run time, as described in the pcreapi documen-
479 tation. The default is 10 million, but this can be changed by adding a
480 setting such as
481
482 --with-match-limit=500000
483
484 to the configure command. This setting has no effect on the
485 pcre_dfa_exec() matching function.
486
487 In some environments it is desirable to limit the depth of recursive
488 calls of match() more strictly than the total number of calls, in order
489 to restrict the maximum amount of stack (or heap, if --disable-stack-
490 for-recursion is specified) that is used. A second limit controls this;
491 it defaults to the value that is set for --with-match-limit, which
492 imposes no additional constraints. However, you can set a lower limit
493 by adding, for example,
494
495 --with-match-limit-recursion=10000
496
497 to the configure command. This value can also be overridden at run
498 time.
499
500
501 CREATING CHARACTER TABLES AT BUILD TIME
502
503 PCRE uses fixed tables for processing characters whose code values are
504 less than 256. By default, PCRE is built with a set of tables that are
505 distributed in the file pcre_chartables.c.dist. These tables are for
506 ASCII codes only. If you add
507
508 --enable-rebuild-chartables
509
510 to the configure command, the distributed tables are no longer used.
511 Instead, a program called dftables is compiled and run. This outputs
512 the source for new set of tables, created in the default locale of your
513 C runtime system. (This method of replacing the tables does not work if
514 you are cross compiling, because dftables is run on the local host. If
515 you need to create alternative tables when cross compiling, you will
516 have to do so "by hand".)
517
518
519 USING EBCDIC CODE
520
521 PCRE assumes by default that it will run in an environment where the
522 character code is ASCII (or Unicode, which is a superset of ASCII).
523 This is the case for most computer operating systems. PCRE can, how-
524 ever, be compiled to run in an EBCDIC environment by adding
525
526 --enable-ebcdic
527
528 to the configure command. This setting implies --enable-rebuild-charta-
529 bles. You should only use it if you know that you are in an EBCDIC
530 environment (for example, an IBM mainframe operating system). The
531 --enable-ebcdic option is incompatible with --enable-utf8.
532
533
534 PCREGREP OPTIONS FOR COMPRESSED FILE SUPPORT
535
536 By default, pcregrep reads all files as plain text. You can build it so
537 that it recognizes files whose names end in .gz or .bz2, and reads them
538 with libz or libbz2, respectively, by adding one or both of
539
540 --enable-pcregrep-libz
541 --enable-pcregrep-libbz2
542
543 to the configure command. These options naturally require that the rel-
544 evant libraries are installed on your system. Configuration will fail
545 if they are not.
546
547
548 PCRETEST OPTION FOR LIBREADLINE SUPPORT
549
550 If you add
551
552 --enable-pcretest-libreadline
553
554 to the configure command, pcretest is linked with the libreadline
555 library, and when its input is from a terminal, it reads it using the
556 readline() function. This provides line-editing and history facilities.
557 Note that libreadline is GPL-licenced, so if you distribute a binary of
558 pcretest linked in this way, there may be licensing issues.
559
560 Setting this option causes the -lreadline option to be added to the
561 pcretest build. In many operating environments with a sytem-installed
562 libreadline this is sufficient. However, in some environments (e.g. if
563 an unmodified distribution version of readline is in use), some extra
564 configuration may be necessary. The INSTALL file for libreadline says
565 this:
566
567 "Readline uses the termcap functions, but does not link with the
568 termcap or curses library itself, allowing applications which link
569 with readline the to choose an appropriate library."
570
571 If your environment has not been set up so that an appropriate library
572 is automatically included, you may need to add something like
573
574 LIBS="-ncurses"
575
576 immediately before the configure command.
577
578
579 SEE ALSO
580
581 pcreapi(3), pcre_config(3).
582
583
584 AUTHOR
585
586 Philip Hazel
587 University Computing Service
588 Cambridge CB2 3QH, England.
589
590
591 REVISION
592
593 Last updated: 17 March 2009
594 Copyright (c) 1997-2009 University of Cambridge.
595 ------------------------------------------------------------------------------
596
597
598 PCREMATCHING(3) PCREMATCHING(3)
599
600
601 NAME
602 PCRE - Perl-compatible regular expressions
603
604
605 PCRE MATCHING ALGORITHMS
606
607 This document describes the two different algorithms that are available
608 in PCRE for matching a compiled regular expression against a given sub-
609 ject string. The "standard" algorithm is the one provided by the
610 pcre_exec() function. This works in the same was as Perl's matching
611 function, and provides a Perl-compatible matching operation.
612
613 An alternative algorithm is provided by the pcre_dfa_exec() function;
614 this operates in a different way, and is not Perl-compatible. It has
615 advantages and disadvantages compared with the standard algorithm, and
616 these are described below.
617
618 When there is only one possible way in which a given subject string can
619 match a pattern, the two algorithms give the same answer. A difference
620 arises, however, when there are multiple possibilities. For example, if
621 the pattern
622
623 ^<.*>
624
625 is matched against the string
626
627 <something> <something else> <something further>
628
629 there are three possible answers. The standard algorithm finds only one
630 of them, whereas the alternative algorithm finds all three.
631
632
633 REGULAR EXPRESSIONS AS TREES
634
635 The set of strings that are matched by a regular expression can be rep-
636 resented as a tree structure. An unlimited repetition in the pattern
637 makes the tree of infinite size, but it is still a tree. Matching the
638 pattern to a given subject string (from a given starting point) can be
639 thought of as a search of the tree. There are two ways to search a
640 tree: depth-first and breadth-first, and these correspond to the two
641 matching algorithms provided by PCRE.
642
643
644 THE STANDARD MATCHING ALGORITHM
645
646 In the terminology of Jeffrey Friedl's book "Mastering Regular Expres-
647 sions", the standard algorithm is an "NFA algorithm". It conducts a
648 depth-first search of the pattern tree. That is, it proceeds along a
649 single path through the tree, checking that the subject matches what is
650 required. When there is a mismatch, the algorithm tries any alterna-
651 tives at the current point, and if they all fail, it backs up to the
652 previous branch point in the tree, and tries the next alternative
653 branch at that level. This often involves backing up (moving to the
654 left) in the subject string as well. The order in which repetition
655 branches are tried is controlled by the greedy or ungreedy nature of
656 the quantifier.
657
658 If a leaf node is reached, a matching string has been found, and at
659 that point the algorithm stops. Thus, if there is more than one possi-
660 ble match, this algorithm returns the first one that it finds. Whether
661 this is the shortest, the longest, or some intermediate length depends
662 on the way the greedy and ungreedy repetition quantifiers are specified
663 in the pattern.
664
665 Because it ends up with a single path through the tree, it is rela-
666 tively straightforward for this algorithm to keep track of the sub-
667 strings that are matched by portions of the pattern in parentheses.
668 This provides support for capturing parentheses and back references.
669
670
671 THE ALTERNATIVE MATCHING ALGORITHM
672
673 This algorithm conducts a breadth-first search of the tree. Starting
674 from the first matching point in the subject, it scans the subject
675 string from left to right, once, character by character, and as it does
676 this, it remembers all the paths through the tree that represent valid
677 matches. In Friedl's terminology, this is a kind of "DFA algorithm",
678 though it is not implemented as a traditional finite state machine (it
679 keeps multiple states active simultaneously).
680
681 The scan continues until either the end of the subject is reached, or
682 there are no more unterminated paths. At this point, terminated paths
683 represent the different matching possibilities (if there are none, the
684 match has failed). Thus, if there is more than one possible match,
685 this algorithm finds all of them, and in particular, it finds the long-
686 est. In PCRE, there is an option to stop the algorithm after the first
687 match (which is necessarily the shortest) has been found.
688
689 Note that all the matches that are found start at the same point in the
690 subject. If the pattern
691
692 cat(er(pillar)?)
693
694 is matched against the string "the caterpillar catchment", the result
695 will be the three strings "cat", "cater", and "caterpillar" that start
696 at the fourth character of the subject. The algorithm does not automat-
697 ically move on to find matches that start at later positions.
698
699 There are a number of features of PCRE regular expressions that are not
700 supported by the alternative matching algorithm. They are as follows:
701
702 1. Because the algorithm finds all possible matches, the greedy or
703 ungreedy nature of repetition quantifiers is not relevant. Greedy and
704 ungreedy quantifiers are treated in exactly the same way. However, pos-
705 sessive quantifiers can make a difference when what follows could also
706 match what is quantified, for example in a pattern like this:
707
708 ^a++\w!
709
710 This pattern matches "aaab!" but not "aaa!", which would be matched by
711 a non-possessive quantifier. Similarly, if an atomic group is present,
712 it is matched as if it were a standalone pattern at the current point,
713 and the longest match is then "locked in" for the rest of the overall
714 pattern.
715
716 2. When dealing with multiple paths through the tree simultaneously, it
717 is not straightforward to keep track of captured substrings for the
718 different matching possibilities, and PCRE's implementation of this
719 algorithm does not attempt to do this. This means that no captured sub-
720 strings are available.
721
722 3. Because no substrings are captured, back references within the pat-
723 tern are not supported, and cause errors if encountered.
724
725 4. For the same reason, conditional expressions that use a backrefer-
726 ence as the condition or test for a specific group recursion are not
727 supported.
728
729 5. Because many paths through the tree may be active, the \K escape
730 sequence, which resets the start of the match when encountered (but may
731 be on some paths and not on others), is not supported. It causes an
732 error if encountered.
733
734 6. Callouts are supported, but the value of the capture_top field is
735 always 1, and the value of the capture_last field is always -1.
736
737 7. The \C escape sequence, which (in the standard algorithm) matches a
738 single byte, even in UTF-8 mode, is not supported because the alterna-
739 tive algorithm moves through the subject string one character at a
740 time, for all active paths through the tree.
741
742 8. Except for (*FAIL), the backtracking control verbs such as (*PRUNE)
743 are not supported. (*FAIL) is supported, and behaves like a failing
744 negative assertion.
745
746
747 ADVANTAGES OF THE ALTERNATIVE ALGORITHM
748
749 Using the alternative matching algorithm provides the following advan-
750 tages:
751
752 1. All possible matches (at a single point in the subject) are automat-
753 ically found, and in particular, the longest match is found. To find
754 more than one match using the standard algorithm, you have to do kludgy
755 things with callouts.
756
757 2. Because the alternative algorithm scans the subject string just
758 once, and never needs to backtrack, it is possible to pass very long
759 subject strings to the matching function in several pieces, checking
760 for partial matching each time.
761
762
763 DISADVANTAGES OF THE ALTERNATIVE ALGORITHM
764
765 The alternative algorithm suffers from a number of disadvantages:
766
767 1. It is substantially slower than the standard algorithm. This is
768 partly because it has to search for all possible matches, but is also
769 because it is less susceptible to optimization.
770
771 2. Capturing parentheses and back references are not supported.
772
773 3. Although atomic groups are supported, their use does not provide the
774 performance advantage that it does for the standard algorithm.
775
776
777 AUTHOR
778
779 Philip Hazel
780 University Computing Service
781 Cambridge CB2 3QH, England.
782
783
784 REVISION
785
786 Last updated: 25 August 2009
787 Copyright (c) 1997-2009 University of Cambridge.
788 ------------------------------------------------------------------------------
789
790
791 PCREAPI(3) PCREAPI(3)
792
793
794 NAME
795 PCRE - Perl-compatible regular expressions
796
797
798 PCRE NATIVE API
799
800 #include <pcre.h>
801
802 pcre *pcre_compile(const char *pattern, int options,
803 const char **errptr, int *erroffset,
804 const unsigned char *tableptr);
805
806 pcre *pcre_compile2(const char *pattern, int options,
807 int *errorcodeptr,
808 const char **errptr, int *erroffset,
809 const unsigned char *tableptr);
810
811 pcre_extra *pcre_study(const pcre *code, int options,
812 const char **errptr);
813
814 int pcre_exec(const pcre *code, const pcre_extra *extra,
815 const char *subject, int length, int startoffset,
816 int options, int *ovector, int ovecsize);
817
818 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
819 const char *subject, int length, int startoffset,
820 int options, int *ovector, int ovecsize,
821 int *workspace, int wscount);
822
823 int pcre_copy_named_substring(const pcre *code,
824 const char *subject, int *ovector,
825 int stringcount, const char *stringname,
826 char *buffer, int buffersize);
827
828 int pcre_copy_substring(const char *subject, int *ovector,
829 int stringcount, int stringnumber, char *buffer,
830 int buffersize);
831
832 int pcre_get_named_substring(const pcre *code,
833 const char *subject, int *ovector,
834 int stringcount, const char *stringname,
835 const char **stringptr);
836
837 int pcre_get_stringnumber(const pcre *code,
838 const char *name);
839
840 int pcre_get_stringtable_entries(const pcre *code,
841 const char *name, char **first, char **last);
842
843 int pcre_get_substring(const char *subject, int *ovector,
844 int stringcount, int stringnumber,
845 const char **stringptr);
846
847 int pcre_get_substring_list(const char *subject,
848 int *ovector, int stringcount, const char ***listptr);
849
850 void pcre_free_substring(const char *stringptr);
851
852 void pcre_free_substring_list(const char **stringptr);
853
854 const unsigned char *pcre_maketables(void);
855
856 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
857 int what, void *where);
858
859 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
860
861 int pcre_refcount(pcre *code, int adjust);
862
863 int pcre_config(int what, void *where);
864
865 char *pcre_version(void);
866
867 void *(*pcre_malloc)(size_t);
868
869 void (*pcre_free)(void *);
870
871 void *(*pcre_stack_malloc)(size_t);
872
873 void (*pcre_stack_free)(void *);
874
875 int (*pcre_callout)(pcre_callout_block *);
876
877
878 PCRE API OVERVIEW
879
880 PCRE has its own native API, which is described in this document. There
881 are also some wrapper functions that correspond to the POSIX regular
882 expression API. These are described in the pcreposix documentation.
883 Both of these APIs define a set of C function calls. A C++ wrapper is
884 distributed with PCRE. It is documented in the pcrecpp page.
885
886 The native API C function prototypes are defined in the header file
887 pcre.h, and on Unix systems the library itself is called libpcre. It
888 can normally be accessed by adding -lpcre to the command for linking an
889 application that uses PCRE. The header file defines the macros
890 PCRE_MAJOR and PCRE_MINOR to contain the major and minor release num-
891 bers for the library. Applications can use these to include support
892 for different releases of PCRE.
893
894 The functions pcre_compile(), pcre_compile2(), pcre_study(), and
895 pcre_exec() are used for compiling and matching regular expressions in
896 a Perl-compatible manner. A sample program that demonstrates the sim-
897 plest way of using them is provided in the file called pcredemo.c in
898 the PCRE source distribution. A listing of this program is given in the
899 pcredemo documentation, and the pcresample documentation describes how
900 to compile and run it.
901
902 A second matching function, pcre_dfa_exec(), which is not Perl-compati-
903 ble, is also provided. This uses a different algorithm for the match-
904 ing. The alternative algorithm finds all possible matches (at a given
905 point in the subject), and scans the subject just once. However, this
906 algorithm does not return captured substrings. A description of the two
907 matching algorithms and their advantages and disadvantages is given in
908 the pcrematching documentation.
909
910 In addition to the main compiling and matching functions, there are
911 convenience functions for extracting captured substrings from a subject
912 string that is matched by pcre_exec(). They are:
913
914 pcre_copy_substring()
915 pcre_copy_named_substring()
916 pcre_get_substring()
917 pcre_get_named_substring()
918 pcre_get_substring_list()
919 pcre_get_stringnumber()
920 pcre_get_stringtable_entries()
921
922 pcre_free_substring() and pcre_free_substring_list() are also provided,
923 to free the memory used for extracted strings.
924
925 The function pcre_maketables() is used to build a set of character
926 tables in the current locale for passing to pcre_compile(),
927 pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is
928 provided for specialist use. Most commonly, no special tables are
929 passed, in which case internal tables that are generated when PCRE is
930 built are used.
931
932 The function pcre_fullinfo() is used to find out information about a
933 compiled pattern; pcre_info() is an obsolete version that returns only
934 some of the available information, but is retained for backwards com-
935 patibility. The function pcre_version() returns a pointer to a string
936 containing the version of PCRE and its date of release.
937
938 The function pcre_refcount() maintains a reference count in a data
939 block containing a compiled pattern. This is provided for the benefit
940 of object-oriented applications.
941
942 The global variables pcre_malloc and pcre_free initially contain the
943 entry points of the standard malloc() and free() functions, respec-
944 tively. PCRE calls the memory management functions via these variables,
945 so a calling program can replace them if it wishes to intercept the
946 calls. This should be done before calling any PCRE functions.
947
948 The global variables pcre_stack_malloc and pcre_stack_free are also
949 indirections to memory management functions. These special functions
950 are used only when PCRE is compiled to use the heap for remembering
951 data, instead of recursive function calls, when running the pcre_exec()
952 function. See the pcrebuild documentation for details of how to do
953 this. It is a non-standard way of building PCRE, for use in environ-
954 ments that have limited stacks. Because of the greater use of memory
955 management, it runs more slowly. Separate functions are provided so
956 that special-purpose external code can be used for this case. When
957 used, these functions are always called in a stack-like manner (last
958 obtained, first freed), and always for memory blocks of the same size.
959 There is a discussion about PCRE's stack usage in the pcrestack docu-
960 mentation.
961
962 The global variable pcre_callout initially contains NULL. It can be set
963 by the caller to a "callout" function, which PCRE will then call at
964 specified points during a matching operation. Details are given in the
965 pcrecallout documentation.
966
967
968 NEWLINES
969
970 PCRE supports five different conventions for indicating line breaks in
971 strings: a single CR (carriage return) character, a single LF (line-
972 feed) character, the two-character sequence CRLF, any of the three pre-
973 ceding, or any Unicode newline sequence. The Unicode newline sequences
974 are the three just mentioned, plus the single characters VT (vertical
975 tab, U+000B), FF (formfeed, U+000C), NEL (next line, U+0085), LS (line
976 separator, U+2028), and PS (paragraph separator, U+2029).
977
978 Each of the first three conventions is used by at least one operating
979 system as its standard newline sequence. When PCRE is built, a default
980 can be specified. The default default is LF, which is the Unix stan-
981 dard. When PCRE is run, the default can be overridden, either when a
982 pattern is compiled, or when it is matched.
983
984 At compile time, the newline convention can be specified by the options
985 argument of pcre_compile(), or it can be specified by special text at
986 the start of the pattern itself; this overrides any other settings. See
987 the pcrepattern page for details of the special character sequences.
988
989 In the PCRE documentation the word "newline" is used to mean "the char-
990 acter or pair of characters that indicate a line break". The choice of
991 newline convention affects the handling of the dot, circumflex, and
992 dollar metacharacters, the handling of #-comments in /x mode, and, when
993 CRLF is a recognized line ending sequence, the match position advance-
994 ment for a non-anchored pattern. There is more detail about this in the
995 section on pcre_exec() options below.
996
997 The choice of newline convention does not affect the interpretation of
998 the \n or \r escape sequences, nor does it affect what \R matches,
999 which is controlled in a similar way, but by separate options.
1000
1001
1002 MULTITHREADING
1003
1004 The PCRE functions can be used in multi-threading applications, with
1005 the proviso that the memory management functions pointed to by
1006 pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the
1007 callout function pointed to by pcre_callout, are shared by all threads.
1008
1009 The compiled form of a regular expression is not altered during match-
1010 ing, so the same compiled pattern can safely be used by several threads
1011 at once.
1012
1013
1014 SAVING PRECOMPILED PATTERNS FOR LATER USE
1015
1016 The compiled form of a regular expression can be saved and re-used at a
1017 later time, possibly by a different program, and even on a host other
1018 than the one on which it was compiled. Details are given in the
1019 pcreprecompile documentation. However, compiling a regular expression
1020 with one version of PCRE for use with a different version is not guar-
1021 anteed to work and may cause crashes.
1022
1023
1024 CHECKING BUILD-TIME OPTIONS
1025
1026 int pcre_config(int what, void *where);
1027
1028 The function pcre_config() makes it possible for a PCRE client to dis-
1029 cover which optional features have been compiled into the PCRE library.
1030 The pcrebuild documentation has more details about these optional fea-
1031 tures.
1032
1033 The first argument for pcre_config() is an integer, specifying which
1034 information is required; the second argument is a pointer to a variable
1035 into which the information is placed. The following information is
1036 available:
1037
1038 PCRE_CONFIG_UTF8
1039
1040 The output is an integer that is set to one if UTF-8 support is avail-
1041 able; otherwise it is set to zero.
1042
1043 PCRE_CONFIG_UNICODE_PROPERTIES
1044
1045 The output is an integer that is set to one if support for Unicode
1046 character properties is available; otherwise it is set to zero.
1047
1048 PCRE_CONFIG_NEWLINE
1049
1050 The output is an integer whose value specifies the default character
1051 sequence that is recognized as meaning "newline". The four values that
1052 are supported are: 10 for LF, 13 for CR, 3338 for CRLF, -2 for ANYCRLF,
1053 and -1 for ANY. Though they are derived from ASCII, the same values
1054 are returned in EBCDIC environments. The default should normally corre-
1055 spond to the standard sequence for your operating system.
1056
1057 PCRE_CONFIG_BSR
1058
1059 The output is an integer whose value indicates what character sequences
1060 the \R escape sequence matches by default. A value of 0 means that \R
1061 matches any Unicode line ending sequence; a value of 1 means that \R
1062 matches only CR, LF, or CRLF. The default can be overridden when a pat-
1063 tern is compiled or matched.
1064
1065 PCRE_CONFIG_LINK_SIZE
1066
1067 The output is an integer that contains the number of bytes used for
1068 internal linkage in compiled regular expressions. The value is 2, 3, or
1069 4. Larger values allow larger regular expressions to be compiled, at
1070 the expense of slower matching. The default value of 2 is sufficient
1071 for all but the most massive patterns, since it allows the compiled
1072 pattern to be up to 64K in size.
1073
1074 PCRE_CONFIG_POSIX_MALLOC_THRESHOLD
1075
1076 The output is an integer that contains the threshold above which the
1077 POSIX interface uses malloc() for output vectors. Further details are
1078 given in the pcreposix documentation.
1079
1080 PCRE_CONFIG_MATCH_LIMIT
1081
1082 The output is a long integer that gives the default limit for the num-
1083 ber of internal matching function calls in a pcre_exec() execution.
1084 Further details are given with pcre_exec() below.
1085
1086 PCRE_CONFIG_MATCH_LIMIT_RECURSION
1087
1088 The output is a long integer that gives the default limit for the depth
1089 of recursion when calling the internal matching function in a
1090 pcre_exec() execution. Further details are given with pcre_exec()
1091 below.
1092
1093 PCRE_CONFIG_STACKRECURSE
1094
1095 The output is an integer that is set to one if internal recursion when
1096 running pcre_exec() is implemented by recursive function calls that use
1097 the stack to remember their state. This is the usual way that PCRE is
1098 compiled. The output is zero if PCRE was compiled to use blocks of data
1099 on the heap instead of recursive function calls. In this case,
1100 pcre_stack_malloc and pcre_stack_free are called to manage memory
1101 blocks on the heap, thus avoiding the use of the stack.
1102
1103
1104 COMPILING A PATTERN
1105
1106 pcre *pcre_compile(const char *pattern, int options,
1107 const char **errptr, int *erroffset,
1108 const unsigned char *tableptr);
1109
1110 pcre *pcre_compile2(const char *pattern, int options,
1111 int *errorcodeptr,
1112 const char **errptr, int *erroffset,
1113 const unsigned char *tableptr);
1114
1115 Either of the functions pcre_compile() or pcre_compile2() can be called
1116 to compile a pattern into an internal form. The only difference between
1117 the two interfaces is that pcre_compile2() has an additional argument,
1118 errorcodeptr, via which a numerical error code can be returned.
1119
1120 The pattern is a C string terminated by a binary zero, and is passed in
1121 the pattern argument. A pointer to a single block of memory that is
1122 obtained via pcre_malloc is returned. This contains the compiled code
1123 and related data. The pcre type is defined for the returned block; this
1124 is a typedef for a structure whose contents are not externally defined.
1125 It is up to the caller to free the memory (via pcre_free) when it is no
1126 longer required.
1127
1128 Although the compiled code of a PCRE regex is relocatable, that is, it
1129 does not depend on memory location, the complete pcre data block is not
1130 fully relocatable, because it may contain a copy of the tableptr argu-
1131 ment, which is an address (see below).
1132
1133 The options argument contains various bit settings that affect the com-
1134 pilation. It should be zero if no options are required. The available
1135 options are described below. Some of them (in particular, those that
1136 are compatible with Perl, but also some others) can also be set and
1137 unset from within the pattern (see the detailed description in the
1138 pcrepattern documentation). For those options that can be different in
1139 different parts of the pattern, the contents of the options argument
1140 specifies their initial settings at the start of compilation and execu-
1141 tion. The PCRE_ANCHORED and PCRE_NEWLINE_xxx options can be set at the
1142 time of matching as well as at compile time.
1143
1144 If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise,
1145 if compilation of a pattern fails, pcre_compile() returns NULL, and
1146 sets the variable pointed to by errptr to point to a textual error mes-
1147 sage. This is a static string that is part of the library. You must not
1148 try to free it. The offset from the start of the pattern to the charac-
1149 ter where the error was discovered is placed in the variable pointed to
1150 by erroffset, which must not be NULL. If it is, an immediate error is
1151 given.
1152
1153 If pcre_compile2() is used instead of pcre_compile(), and the error-
1154 codeptr argument is not NULL, a non-zero error code number is returned
1155 via this argument in the event of an error. This is in addition to the
1156 textual error message. Error codes and messages are listed below.
1157
1158 If the final argument, tableptr, is NULL, PCRE uses a default set of
1159 character tables that are built when PCRE is compiled, using the
1160 default C locale. Otherwise, tableptr must be an address that is the
1161 result of a call to pcre_maketables(). This value is stored with the
1162 compiled pattern, and used again by pcre_exec(), unless another table
1163 pointer is passed to it. For more discussion, see the section on locale
1164 support below.
1165
1166 This code fragment shows a typical straightforward call to pcre_com-
1167 pile():
1168
1169 pcre *re;
1170 const char *error;
1171 int erroffset;
1172 re = pcre_compile(
1173 "^A.*Z", /* the pattern */
1174 0, /* default options */
1175 &error, /* for error message */
1176 &erroffset, /* for error offset */
1177 NULL); /* use default character tables */
1178
1179 The following names for option bits are defined in the pcre.h header
1180 file:
1181
1182 PCRE_ANCHORED
1183
1184 If this bit is set, the pattern is forced to be "anchored", that is, it
1185 is constrained to match only at the first matching point in the string
1186 that is being searched (the "subject string"). This effect can also be
1187 achieved by appropriate constructs in the pattern itself, which is the
1188 only way to do it in Perl.
1189
1190 PCRE_AUTO_CALLOUT
1191
1192 If this bit is set, pcre_compile() automatically inserts callout items,
1193 all with number 255, before each pattern item. For discussion of the
1194 callout facility, see the pcrecallout documentation.
1195
1196 PCRE_BSR_ANYCRLF
1197 PCRE_BSR_UNICODE
1198
1199 These options (which are mutually exclusive) control what the \R escape
1200 sequence matches. The choice is either to match only CR, LF, or CRLF,
1201 or to match any Unicode newline sequence. The default is specified when
1202 PCRE is built. It can be overridden from within the pattern, or by set-
1203 ting an option when a compiled pattern is matched.
1204
1205 PCRE_CASELESS
1206
1207 If this bit is set, letters in the pattern match both upper and lower
1208 case letters. It is equivalent to Perl's /i option, and it can be
1209 changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE
1210 always understands the concept of case for characters whose values are
1211 less than 128, so caseless matching is always possible. For characters
1212 with higher values, the concept of case is supported if PCRE is com-
1213 piled with Unicode property support, but not otherwise. If you want to
1214 use caseless matching for characters 128 and above, you must ensure
1215 that PCRE is compiled with Unicode property support as well as with
1216 UTF-8 support.
1217
1218 PCRE_DOLLAR_ENDONLY
1219
1220 If this bit is set, a dollar metacharacter in the pattern matches only
1221 at the end of the subject string. Without this option, a dollar also
1222 matches immediately before a newline at the end of the string (but not
1223 before any other newlines). The PCRE_DOLLAR_ENDONLY option is ignored
1224 if PCRE_MULTILINE is set. There is no equivalent to this option in
1225 Perl, and no way to set it within a pattern.
1226
1227 PCRE_DOTALL
1228
1229 If this bit is set, a dot metacharater in the pattern matches all char-
1230 acters, including those that indicate newline. Without it, a dot does
1231 not match when the current position is at a newline. This option is
1232 equivalent to Perl's /s option, and it can be changed within a pattern
1233 by a (?s) option setting. A negative class such as [^a] always matches
1234 newline characters, independent of the setting of this option.
1235
1236 PCRE_DUPNAMES
1237
1238 If this bit is set, names used to identify capturing subpatterns need
1239 not be unique. This can be helpful for certain types of pattern when it
1240 is known that only one instance of the named subpattern can ever be
1241 matched. There are more details of named subpatterns below; see also
1242 the pcrepattern documentation.
1243
1244 PCRE_EXTENDED
1245
1246 If this bit is set, whitespace data characters in the pattern are
1247 totally ignored except when escaped or inside a character class. White-
1248 space does not include the VT character (code 11). In addition, charac-
1249 ters between an unescaped # outside a character class and the next new-
1250 line, inclusive, are also ignored. This is equivalent to Perl's /x
1251 option, and it can be changed within a pattern by a (?x) option set-
1252 ting.
1253
1254 This option makes it possible to include comments inside complicated
1255 patterns. Note, however, that this applies only to data characters.
1256 Whitespace characters may never appear within special character
1257 sequences in a pattern, for example within the sequence (?( which
1258 introduces a conditional subpattern.
1259
1260 PCRE_EXTRA
1261
1262 This option was invented in order to turn on additional functionality
1263 of PCRE that is incompatible with Perl, but it is currently of very
1264 little use. When set, any backslash in a pattern that is followed by a
1265 letter that has no special meaning causes an error, thus reserving
1266 these combinations for future expansion. By default, as in Perl, a
1267 backslash followed by a letter with no special meaning is treated as a
1268 literal. (Perl can, however, be persuaded to give a warning for this.)
1269 There are at present no other features controlled by this option. It
1270 can also be set by a (?X) option setting within a pattern.
1271
1272 PCRE_FIRSTLINE
1273
1274 If this option is set, an unanchored pattern is required to match
1275 before or at the first newline in the subject string, though the
1276 matched text may continue over the newline.
1277
1278 PCRE_JAVASCRIPT_COMPAT
1279
1280 If this option is set, PCRE's behaviour is changed in some ways so that
1281 it is compatible with JavaScript rather than Perl. The changes are as
1282 follows:
1283
1284 (1) A lone closing square bracket in a pattern causes a compile-time
1285 error, because this is illegal in JavaScript (by default it is treated
1286 as a data character). Thus, the pattern AB]CD becomes illegal when this
1287 option is set.
1288
1289 (2) At run time, a back reference to an unset subpattern group matches
1290 an empty string (by default this causes the current matching alterna-
1291 tive to fail). A pattern such as (\1)(a) succeeds when this option is
1292 set (assuming it can find an "a" in the subject), whereas it fails by
1293 default, for Perl compatibility.
1294
1295 PCRE_MULTILINE
1296
1297 By default, PCRE treats the subject string as consisting of a single
1298 line of characters (even if it actually contains newlines). The "start
1299 of line" metacharacter (^) matches only at the start of the string,
1300 while the "end of line" metacharacter ($) matches only at the end of
1301 the string, or before a terminating newline (unless PCRE_DOLLAR_ENDONLY
1302 is set). This is the same as Perl.
1303
1304 When PCRE_MULTILINE it is set, the "start of line" and "end of line"
1305 constructs match immediately following or immediately before internal
1306 newlines in the subject string, respectively, as well as at the very
1307 start and end. This is equivalent to Perl's /m option, and it can be
1308 changed within a pattern by a (?m) option setting. If there are no new-
1309 lines in a subject string, or no occurrences of ^ or $ in a pattern,
1310 setting PCRE_MULTILINE has no effect.
1311
1312 PCRE_NEWLINE_CR
1313 PCRE_NEWLINE_LF
1314 PCRE_NEWLINE_CRLF
1315 PCRE_NEWLINE_ANYCRLF
1316 PCRE_NEWLINE_ANY
1317
1318 These options override the default newline definition that was chosen
1319 when PCRE was built. Setting the first or the second specifies that a
1320 newline is indicated by a single character (CR or LF, respectively).
1321 Setting PCRE_NEWLINE_CRLF specifies that a newline is indicated by the
1322 two-character CRLF sequence. Setting PCRE_NEWLINE_ANYCRLF specifies
1323 that any of the three preceding sequences should be recognized. Setting
1324 PCRE_NEWLINE_ANY specifies that any Unicode newline sequence should be
1325 recognized. The Unicode newline sequences are the three just mentioned,
1326 plus the single characters VT (vertical tab, U+000B), FF (formfeed,
1327 U+000C), NEL (next line, U+0085), LS (line separator, U+2028), and PS
1328 (paragraph separator, U+2029). The last two are recognized only in
1329 UTF-8 mode.
1330
1331 The newline setting in the options word uses three bits that are
1332 treated as a number, giving eight possibilities. Currently only six are
1333 used (default plus the five values above). This means that if you set
1334 more than one newline option, the combination may or may not be sensi-
1335 ble. For example, PCRE_NEWLINE_CR with PCRE_NEWLINE_LF is equivalent to
1336 PCRE_NEWLINE_CRLF, but other combinations may yield unused numbers and
1337 cause an error.
1338
1339 The only time that a line break is specially recognized when compiling
1340 a pattern is if PCRE_EXTENDED is set, and an unescaped # outside a
1341 character class is encountered. This indicates a comment that lasts
1342 until after the next line break sequence. In other circumstances, line
1343 break sequences are treated as literal data, except that in
1344 PCRE_EXTENDED mode, both CR and LF are treated as whitespace characters
1345 and are therefore ignored.
1346
1347 The newline option that is set at compile time becomes the default that
1348 is used for pcre_exec() and pcre_dfa_exec(), but it can be overridden.
1349
1350 PCRE_NO_AUTO_CAPTURE
1351
1352 If this option is set, it disables the use of numbered capturing paren-
1353 theses in the pattern. Any opening parenthesis that is not followed by
1354 ? behaves as if it were followed by ?: but named parentheses can still
1355 be used for capturing (and they acquire numbers in the usual way).
1356 There is no equivalent of this option in Perl.
1357
1358 PCRE_UNGREEDY
1359
1360 This option inverts the "greediness" of the quantifiers so that they
1361 are not greedy by default, but become greedy if followed by "?". It is
1362 not compatible with Perl. It can also be set by a (?U) option setting
1363 within the pattern.
1364
1365 PCRE_UTF8
1366
1367 This option causes PCRE to regard both the pattern and the subject as
1368 strings of UTF-8 characters instead of single-byte character strings.
1369 However, it is available only when PCRE is built to include UTF-8 sup-
1370 port. If not, the use of this option provokes an error. Details of how
1371 this option changes the behaviour of PCRE are given in the section on
1372 UTF-8 support in the main pcre page.
1373
1374 PCRE_NO_UTF8_CHECK
1375
1376 When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is
1377 automatically checked. There is a discussion about the validity of
1378 UTF-8 strings in the main pcre page. If an invalid UTF-8 sequence of
1379 bytes is found, pcre_compile() returns an error. If you already know
1380 that your pattern is valid, and you want to skip this check for perfor-
1381 mance reasons, you can set the PCRE_NO_UTF8_CHECK option. When it is
1382 set, the effect of passing an invalid UTF-8 string as a pattern is
1383 undefined. It may cause your program to crash. Note that this option
1384 can also be passed to pcre_exec() and pcre_dfa_exec(), to suppress the
1385 UTF-8 validity checking of subject strings.
1386
1387
1388 COMPILATION ERROR CODES
1389
1390 The following table lists the error codes than may be returned by
1391 pcre_compile2(), along with the error messages that may be returned by
1392 both compiling functions. As PCRE has developed, some error codes have
1393 fallen out of use. To avoid confusion, they have not been re-used.
1394
1395 0 no error
1396 1 \ at end of pattern
1397 2 \c at end of pattern
1398 3 unrecognized character follows \
1399 4 numbers out of order in {} quantifier
1400 5 number too big in {} quantifier
1401 6 missing terminating ] for character class
1402 7 invalid escape sequence in character class
1403 8 range out of order in character class
1404 9 nothing to repeat
1405 10 [this code is not in use]
1406 11 internal error: unexpected repeat
1407 12 unrecognized character after (? or (?-
1408 13 POSIX named classes are supported only within a class
1409 14 missing )
1410 15 reference to non-existent subpattern
1411 16 erroffset passed as NULL
1412 17 unknown option bit(s) set
1413 18 missing ) after comment
1414 19 [this code is not in use]
1415 20 regular expression is too large
1416 21 failed to get memory
1417 22 unmatched parentheses
1418 23 internal error: code overflow
1419 24 unrecognized character after (?<
1420 25 lookbehind assertion is not fixed length
1421 26 malformed number or name after (?(
1422 27 conditional group contains more than two branches
1423 28 assertion expected after (?(
1424 29 (?R or (?[+-]digits must be followed by )
1425 30 unknown POSIX class name
1426 31 POSIX collating elements are not supported
1427 32 this version of PCRE is not compiled with PCRE_UTF8 support
1428 33 [this code is not in use]
1429 34 character value in \x{...} sequence is too large
1430 35 invalid condition (?(0)
1431 36 \C not allowed in lookbehind assertion
1432 37 PCRE does not support \L, \l, \N, \U, or \u
1433 38 number after (?C is > 255
1434 39 closing ) for (?C expected
1435 40 recursive call could loop indefinitely
1436 41 unrecognized character after (?P
1437 42 syntax error in subpattern name (missing terminator)
1438 43 two named subpatterns have the same name
1439 44 invalid UTF-8 string
1440 45 support for \P, \p, and \X has not been compiled
1441 46 malformed \P or \p sequence
1442 47 unknown property name after \P or \p
1443 48 subpattern name is too long (maximum 32 characters)
1444 49 too many named subpatterns (maximum 10000)
1445 50 [this code is not in use]
1446 51 octal value is greater than \377 (not in UTF-8 mode)
1447 52 internal error: overran compiling workspace
1448 53 internal error: previously-checked referenced subpattern not
1449 found
1450 54 DEFINE group contains more than one branch
1451 55 repeating a DEFINE group is not allowed
1452 56 inconsistent NEWLINE options
1453 57 \g is not followed by a braced, angle-bracketed, or quoted
1454 name/number or by a plain number
1455 58 a numbered reference must not be zero
1456 59 (*VERB) with an argument is not supported
1457 60 (*VERB) not recognized
1458 61 number is too big
1459 62 subpattern name expected
1460 63 digit expected after (?+
1461 64 ] is an invalid data character in JavaScript compatibility mode
1462
1463 The numbers 32 and 10000 in errors 48 and 49 are defaults; different
1464 values may be used if the limits were changed when PCRE was built.
1465
1466
1467 STUDYING A PATTERN
1468
1469 pcre_extra *pcre_study(const pcre *code, int options
1470 const char **errptr);
1471
1472 If a compiled pattern is going to be used several times, it is worth
1473 spending more time analyzing it in order to speed up the time taken for
1474 matching. The function pcre_study() takes a pointer to a compiled pat-
1475 tern as its first argument. If studying the pattern produces additional
1476 information that will help speed up matching, pcre_study() returns a
1477 pointer to a pcre_extra block, in which the study_data field points to
1478 the results of the study.
1479
1480 The returned value from pcre_study() can be passed directly to
1481 pcre_exec(). However, a pcre_extra block also contains other fields
1482 that can be set by the caller before the block is passed; these are
1483 described below in the section on matching a pattern.
1484
1485 If studying the pattern does not produce any additional information
1486 pcre_study() returns NULL. In that circumstance, if the calling program
1487 wants to pass any of the other fields to pcre_exec(), it must set up
1488 its own pcre_extra block.
1489
1490 The second argument of pcre_study() contains option bits. At present,
1491 no options are defined, and this argument should always be zero.
1492
1493 The third argument for pcre_study() is a pointer for an error message.
1494 If studying succeeds (even if no data is returned), the variable it
1495 points to is set to NULL. Otherwise it is set to point to a textual
1496 error message. This is a static string that is part of the library. You
1497 must not try to free it. You should test the error pointer for NULL
1498 after calling pcre_study(), to be sure that it has run successfully.
1499
1500 This is a typical call to pcre_study():
1501
1502 pcre_extra *pe;
1503 pe = pcre_study(
1504 re, /* result of pcre_compile() */
1505 0, /* no options exist */
1506 &error); /* set to NULL or points to a message */
1507
1508 At present, studying a pattern is useful only for non-anchored patterns
1509 that do not have a single fixed starting character. A bitmap of possi-
1510 ble starting bytes is created.
1511
1512
1513 LOCALE SUPPORT
1514
1515 PCRE handles caseless matching, and determines whether characters are
1516 letters, digits, or whatever, by reference to a set of tables, indexed
1517 by character value. When running in UTF-8 mode, this applies only to
1518 characters with codes less than 128. Higher-valued codes never match
1519 escapes such as \w or \d, but can be tested with \p if PCRE is built
1520 with Unicode character property support. The use of locales with Uni-
1521 code is discouraged. If you are handling characters with codes greater
1522 than 128, you should either use UTF-8 and Unicode, or use locales, but
1523 not try to mix the two.
1524
1525 PCRE contains an internal set of tables that are used when the final
1526 argument of pcre_compile() is NULL. These are sufficient for many
1527 applications. Normally, the internal tables recognize only ASCII char-
1528 acters. However, when PCRE is built, it is possible to cause the inter-
1529 nal tables to be rebuilt in the default "C" locale of the local system,
1530 which may cause them to be different.
1531
1532 The internal tables can always be overridden by tables supplied by the
1533 application that calls PCRE. These may be created in a different locale
1534 from the default. As more and more applications change to using Uni-
1535 code, the need for this locale support is expected to die away.
1536
1537 External tables are built by calling the pcre_maketables() function,
1538 which has no arguments, in the relevant locale. The result can then be
1539 passed to pcre_compile() or pcre_exec() as often as necessary. For
1540 example, to build and use tables that are appropriate for the French
1541 locale (where accented characters with values greater than 128 are
1542 treated as letters), the following code could be used:
1543
1544 setlocale(LC_CTYPE, "fr_FR");
1545 tables = pcre_maketables();
1546 re = pcre_compile(..., tables);
1547
1548 The locale name "fr_FR" is used on Linux and other Unix-like systems;
1549 if you are using Windows, the name for the French locale is "french".
1550
1551 When pcre_maketables() runs, the tables are built in memory that is
1552 obtained via pcre_malloc. It is the caller's responsibility to ensure
1553 that the memory containing the tables remains available for as long as
1554 it is needed.
1555
1556 The pointer that is passed to pcre_compile() is saved with the compiled
1557 pattern, and the same tables are used via this pointer by pcre_study()
1558 and normally also by pcre_exec(). Thus, by default, for any single pat-
1559 tern, compilation, studying and matching all happen in the same locale,
1560 but different patterns can be compiled in different locales.
1561
1562 It is possible to pass a table pointer or NULL (indicating the use of
1563 the internal tables) to pcre_exec(). Although not intended for this
1564 purpose, this facility could be used to match a pattern in a different
1565 locale from the one in which it was compiled. Passing table pointers at
1566 run time is discussed below in the section on matching a pattern.
1567
1568
1569 INFORMATION ABOUT A PATTERN
1570
1571 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
1572 int what, void *where);
1573
1574 The pcre_fullinfo() function returns information about a compiled pat-
1575 tern. It replaces the obsolete pcre_info() function, which is neverthe-
1576 less retained for backwards compability (and is documented below).
1577
1578 The first argument for pcre_fullinfo() is a pointer to the compiled
1579 pattern. The second argument is the result of pcre_study(), or NULL if
1580 the pattern was not studied. The third argument specifies which piece
1581 of information is required, and the fourth argument is a pointer to a
1582 variable to receive the data. The yield of the function is zero for
1583 success, or one of the following negative numbers:
1584
1585 PCRE_ERROR_NULL the argument code was NULL
1586 the argument where was NULL
1587 PCRE_ERROR_BADMAGIC the "magic number" was not found
1588 PCRE_ERROR_BADOPTION the value of what was invalid
1589
1590 The "magic number" is placed at the start of each compiled pattern as
1591 an simple check against passing an arbitrary memory pointer. Here is a
1592 typical call of pcre_fullinfo(), to obtain the length of the compiled
1593 pattern:
1594
1595 int rc;
1596 size_t length;
1597 rc = pcre_fullinfo(
1598 re, /* result of pcre_compile() */
1599 pe, /* result of pcre_study(), or NULL */
1600 PCRE_INFO_SIZE, /* what is required */
1601 &length); /* where to put the data */
1602
1603 The possible values for the third argument are defined in pcre.h, and
1604 are as follows:
1605
1606 PCRE_INFO_BACKREFMAX
1607
1608 Return the number of the highest back reference in the pattern. The
1609 fourth argument should point to an int variable. Zero is returned if
1610 there are no back references.
1611
1612 PCRE_INFO_CAPTURECOUNT
1613
1614 Return the number of capturing subpatterns in the pattern. The fourth
1615 argument should point to an int variable.
1616
1617 PCRE_INFO_DEFAULT_TABLES
1618
1619 Return a pointer to the internal default character tables within PCRE.
1620 The fourth argument should point to an unsigned char * variable. This
1621 information call is provided for internal use by the pcre_study() func-
1622 tion. External callers can cause PCRE to use its internal tables by
1623 passing a NULL table pointer.
1624
1625 PCRE_INFO_FIRSTBYTE
1626
1627 Return information about the first byte of any matched string, for a
1628 non-anchored pattern. The fourth argument should point to an int vari-
1629 able. (This option used to be called PCRE_INFO_FIRSTCHAR; the old name
1630 is still recognized for backwards compatibility.)
1631
1632 If there is a fixed first byte, for example, from a pattern such as
1633 (cat|cow|coyote), its value is returned. Otherwise, if either
1634
1635 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
1636 branch starts with "^", or
1637
1638 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
1639 set (if it were set, the pattern would be anchored),
1640
1641 -1 is returned, indicating that the pattern matches only at the start
1642 of a subject string or after any newline within the string. Otherwise
1643 -2 is returned. For anchored patterns, -2 is returned.
1644
1645 PCRE_INFO_FIRSTTABLE
1646
1647 If the pattern was studied, and this resulted in the construction of a
1648 256-bit table indicating a fixed set of bytes for the first byte in any
1649 matching string, a pointer to the table is returned. Otherwise NULL is
1650 returned. The fourth argument should point to an unsigned char * vari-
1651 able.
1652
1653 PCRE_INFO_HASCRORLF
1654
1655 Return 1 if the pattern contains any explicit matches for CR or LF
1656 characters, otherwise 0. The fourth argument should point to an int
1657 variable. An explicit match is either a literal CR or LF character, or
1658 \r or \n.
1659
1660 PCRE_INFO_JCHANGED
1661
1662 Return 1 if the (?J) or (?-J) option setting is used in the pattern,
1663 otherwise 0. The fourth argument should point to an int variable. (?J)
1664 and (?-J) set and unset the local PCRE_DUPNAMES option, respectively.
1665
1666 PCRE_INFO_LASTLITERAL
1667
1668 Return the value of the rightmost literal byte that must exist in any
1669 matched string, other than at its start, if such a byte has been
1670 recorded. The fourth argument should point to an int variable. If there
1671 is no such byte, -1 is returned. For anchored patterns, a last literal
1672 byte is recorded only if it follows something of variable length. For
1673 example, for the pattern /^a\d+z\d+/ the returned value is "z", but for
1674 /^a\dz\d/ the returned value is -1.
1675
1676 PCRE_INFO_NAMECOUNT
1677 PCRE_INFO_NAMEENTRYSIZE
1678 PCRE_INFO_NAMETABLE
1679
1680 PCRE supports the use of named as well as numbered capturing parenthe-
1681 ses. The names are just an additional way of identifying the parenthe-
1682 ses, which still acquire numbers. Several convenience functions such as
1683 pcre_get_named_substring() are provided for extracting captured sub-
1684 strings by name. It is also possible to extract the data directly, by
1685 first converting the name to a number in order to access the correct
1686 pointers in the output vector (described with pcre_exec() below). To do
1687 the conversion, you need to use the name-to-number map, which is
1688 described by these three values.
1689
1690 The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT
1691 gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size
1692 of each entry; both of these return an int value. The entry size
1693 depends on the length of the longest name. PCRE_INFO_NAMETABLE returns
1694 a pointer to the first entry of the table (a pointer to char). The
1695 first two bytes of each entry are the number of the capturing parenthe-
1696 sis, most significant byte first. The rest of the entry is the corre-
1697 sponding name, zero terminated. The names are in alphabetical order.
1698 When PCRE_DUPNAMES is set, duplicate names are in order of their paren-
1699 theses numbers. For example, consider the following pattern (assume
1700 PCRE_EXTENDED is set, so white space - including newlines - is
1701 ignored):
1702
1703 (?<date> (?<year>(\d\d)?\d\d) -
1704 (?<month>\d\d) - (?<day>\d\d) )
1705
1706 There are four named subpatterns, so the table has four entries, and
1707 each entry in the table is eight bytes long. The table is as follows,
1708 with non-printing bytes shows in hexadecimal, and undefined bytes shown
1709 as ??:
1710
1711 00 01 d a t e 00 ??
1712 00 05 d a y 00 ?? ??
1713 00 04 m o n t h 00
1714 00 02 y e a r 00 ??
1715
1716 When writing code to extract data from named subpatterns using the
1717 name-to-number map, remember that the length of the entries is likely
1718 to be different for each compiled pattern.
1719
1720 PCRE_INFO_OKPARTIAL
1721
1722 Return 1 if the pattern can be used for partial matching, otherwise 0.
1723 The fourth argument should point to an int variable. From release 8.00,
1724 this always returns 1, because the restrictions that previously applied
1725 to partial matching have been lifted. The pcrepartial documentation
1726 gives details of partial matching.
1727
1728 PCRE_INFO_OPTIONS
1729
1730 Return a copy of the options with which the pattern was compiled. The
1731 fourth argument should point to an unsigned long int variable. These
1732 option bits are those specified in the call to pcre_compile(), modified
1733 by any top-level option settings at the start of the pattern itself. In
1734 other words, they are the options that will be in force when matching
1735 starts. For example, if the pattern /(?im)abc(?-i)d/ is compiled with
1736 the PCRE_EXTENDED option, the result is PCRE_CASELESS, PCRE_MULTILINE,
1737 and PCRE_EXTENDED.
1738
1739 A pattern is automatically anchored by PCRE if all of its top-level
1740 alternatives begin with one of the following:
1741
1742 ^ unless PCRE_MULTILINE is set
1743 \A always
1744 \G always
1745 .* if PCRE_DOTALL is set and there are no back
1746 references to the subpattern in which .* appears
1747
1748 For such patterns, the PCRE_ANCHORED bit is set in the options returned
1749 by pcre_fullinfo().
1750
1751 PCRE_INFO_SIZE
1752
1753 Return the size of the compiled pattern, that is, the value that was
1754 passed as the argument to pcre_malloc() when PCRE was getting memory in
1755 which to place the compiled data. The fourth argument should point to a
1756 size_t variable.
1757
1758 PCRE_INFO_STUDYSIZE
1759
1760 Return the size of the data block pointed to by the study_data field in
1761 a pcre_extra block. That is, it is the value that was passed to
1762 pcre_malloc() when PCRE was getting memory into which to place the data
1763 created by pcre_study(). The fourth argument should point to a size_t
1764 variable.
1765
1766
1767 OBSOLETE INFO FUNCTION
1768
1769 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
1770
1771 The pcre_info() function is now obsolete because its interface is too
1772 restrictive to return all the available data about a compiled pattern.
1773 New programs should use pcre_fullinfo() instead. The yield of
1774 pcre_info() is the number of capturing subpatterns, or one of the fol-
1775 lowing negative numbers:
1776
1777 PCRE_ERROR_NULL the argument code was NULL
1778 PCRE_ERROR_BADMAGIC the "magic number" was not found
1779
1780 If the optptr argument is not NULL, a copy of the options with which
1781 the pattern was compiled is placed in the integer it points to (see
1782 PCRE_INFO_OPTIONS above).
1783
1784 If the pattern is not anchored and the firstcharptr argument is not
1785 NULL, it is used to pass back information about the first character of
1786 any matched string (see PCRE_INFO_FIRSTBYTE above).
1787
1788
1789 REFERENCE COUNTS
1790
1791 int pcre_refcount(pcre *code, int adjust);
1792
1793 The pcre_refcount() function is used to maintain a reference count in
1794 the data block that contains a compiled pattern. It is provided for the
1795 benefit of applications that operate in an object-oriented manner,
1796 where different parts of the application may be using the same compiled
1797 pattern, but you want to free the block when they are all done.
1798
1799 When a pattern is compiled, the reference count field is initialized to
1800 zero. It is changed only by calling this function, whose action is to
1801 add the adjust value (which may be positive or negative) to it. The
1802 yield of the function is the new value. However, the value of the count
1803 is constrained to lie between 0 and 65535, inclusive. If the new value
1804 is outside these limits, it is forced to the appropriate limit value.
1805
1806 Except when it is zero, the reference count is not correctly preserved
1807 if a pattern is compiled on one host and then transferred to a host
1808 whose byte-order is different. (This seems a highly unlikely scenario.)
1809
1810
1811 MATCHING A PATTERN: THE TRADITIONAL FUNCTION
1812
1813 int pcre_exec(const pcre *code, const pcre_extra *extra,
1814 const char *subject, int length, int startoffset,
1815 int options, int *ovector, int ovecsize);
1816
1817 The function pcre_exec() is called to match a subject string against a
1818 compiled pattern, which is passed in the code argument. If the pattern
1819 has been studied, the result of the study should be passed in the extra
1820 argument. This function is the main matching facility of the library,
1821 and it operates in a Perl-like manner. For specialist use there is also
1822 an alternative matching function, which is described below in the sec-
1823 tion about the pcre_dfa_exec() function.
1824
1825 In most applications, the pattern will have been compiled (and option-
1826 ally studied) in the same process that calls pcre_exec(). However, it
1827 is possible to save compiled patterns and study data, and then use them
1828 later in different processes, possibly even on different hosts. For a
1829 discussion about this, see the pcreprecompile documentation.
1830
1831 Here is an example of a simple call to pcre_exec():
1832
1833 int rc;
1834 int ovector[30];
1835 rc = pcre_exec(
1836 re, /* result of pcre_compile() */
1837 NULL, /* we didn't study the pattern */
1838 "some string", /* the subject string */
1839 11, /* the length of the subject string */
1840 0, /* start at offset 0 in the subject */
1841 0, /* default options */
1842 ovector, /* vector of integers for substring information */
1843 30); /* number of elements (NOT size in bytes) */
1844
1845 Extra data for pcre_exec()
1846
1847 If the extra argument is not NULL, it must point to a pcre_extra data
1848 block. The pcre_study() function returns such a block (when it doesn't
1849 return NULL), but you can also create one for yourself, and pass addi-
1850 tional information in it. The pcre_extra block contains the following
1851 fields (not necessarily in this order):
1852
1853 unsigned long int flags;
1854 void *study_data;
1855 unsigned long int match_limit;
1856 unsigned long int match_limit_recursion;
1857 void *callout_data;
1858 const unsigned char *tables;
1859
1860 The flags field is a bitmap that specifies which of the other fields
1861 are set. The flag bits are:
1862
1863 PCRE_EXTRA_STUDY_DATA
1864 PCRE_EXTRA_MATCH_LIMIT
1865 PCRE_EXTRA_MATCH_LIMIT_RECURSION
1866 PCRE_EXTRA_CALLOUT_DATA
1867 PCRE_EXTRA_TABLES
1868
1869 Other flag bits should be set to zero. The study_data field is set in
1870 the pcre_extra block that is returned by pcre_study(), together with
1871 the appropriate flag bit. You should not set this yourself, but you may
1872 add to the block by setting the other fields and their corresponding
1873 flag bits.
1874
1875 The match_limit field provides a means of preventing PCRE from using up
1876 a vast amount of resources when running patterns that are not going to
1877 match, but which have a very large number of possibilities in their
1878 search trees. The classic example is the use of nested unlimited
1879 repeats.
1880
1881 Internally, PCRE uses a function called match() which it calls repeat-
1882 edly (sometimes recursively). The limit set by match_limit is imposed
1883 on the number of times this function is called during a match, which
1884 has the effect of limiting the amount of backtracking that can take
1885 place. For patterns that are not anchored, the count restarts from zero
1886 for each position in the subject string.
1887
1888 The default value for the limit can be set when PCRE is built; the
1889 default default is 10 million, which handles all but the most extreme
1890 cases. You can override the default by suppling pcre_exec() with a
1891 pcre_extra block in which match_limit is set, and
1892 PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is
1893 exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.
1894
1895 The match_limit_recursion field is similar to match_limit, but instead
1896 of limiting the total number of times that match() is called, it limits
1897 the depth of recursion. The recursion depth is a smaller number than
1898 the total number of calls, because not all calls to match() are recur-
1899 sive. This limit is of use only if it is set smaller than match_limit.
1900
1901 Limiting the recursion depth limits the amount of stack that can be
1902 used, or, when PCRE has been compiled to use memory on the heap instead
1903 of the stack, the amount of heap memory that can be used.
1904
1905 The default value for match_limit_recursion can be set when PCRE is
1906 built; the default default is the same value as the default for
1907 match_limit. You can override the default by suppling pcre_exec() with
1908 a pcre_extra block in which match_limit_recursion is set, and
1909 PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the
1910 limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT.
1911
1912 The pcre_callout field is used in conjunction with the "callout" fea-
1913 ture, which is described in the pcrecallout documentation.
1914
1915 The tables field is used to pass a character tables pointer to
1916 pcre_exec(); this overrides the value that is stored with the compiled
1917 pattern. A non-NULL value is stored with the compiled pattern only if
1918 custom tables were supplied to pcre_compile() via its tableptr argu-
1919 ment. If NULL is passed to pcre_exec() using this mechanism, it forces
1920 PCRE's internal tables to be used. This facility is helpful when re-
1921 using patterns that have been saved after compiling with an external
1922 set of tables, because the external tables might be at a different
1923 address when pcre_exec() is called. See the pcreprecompile documenta-
1924 tion for a discussion of saving compiled patterns for later use.
1925
1926 Option bits for pcre_exec()
1927
1928 The unused bits of the options argument for pcre_exec() must be zero.
1929 The only bits that may be set are PCRE_ANCHORED, PCRE_NEWLINE_xxx,
1930 PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NO_START_OPTIMIZE,
1931 PCRE_NO_UTF8_CHECK, PCRE_PARTIAL_SOFT, and PCRE_PARTIAL_HARD.
1932
1933 PCRE_ANCHORED
1934
1935 The PCRE_ANCHORED option limits pcre_exec() to matching at the first
1936 matching position. If a pattern was compiled with PCRE_ANCHORED, or
1937 turned out to be anchored by virtue of its contents, it cannot be made
1938 unachored at matching time.
1939
1940 PCRE_BSR_ANYCRLF
1941 PCRE_BSR_UNICODE
1942
1943 These options (which are mutually exclusive) control what the \R escape
1944 sequence matches. The choice is either to match only CR, LF, or CRLF,
1945 or to match any Unicode newline sequence. These options override the
1946 choice that was made or defaulted when the pattern was compiled.
1947
1948 PCRE_NEWLINE_CR
1949 PCRE_NEWLINE_LF
1950 PCRE_NEWLINE_CRLF
1951 PCRE_NEWLINE_ANYCRLF
1952 PCRE_NEWLINE_ANY
1953
1954 These options override the newline definition that was chosen or
1955 defaulted when the pattern was compiled. For details, see the descrip-
1956 tion of pcre_compile() above. During matching, the newline choice
1957 affects the behaviour of the dot, circumflex, and dollar metacharac-
1958 ters. It may also alter the way the match position is advanced after a
1959 match failure for an unanchored pattern.
1960
1961 When PCRE_NEWLINE_CRLF, PCRE_NEWLINE_ANYCRLF, or PCRE_NEWLINE_ANY is
1962 set, and a match attempt for an unanchored pattern fails when the cur-
1963 rent position is at a CRLF sequence, and the pattern contains no
1964 explicit matches for CR or LF characters, the match position is
1965 advanced by two characters instead of one, in other words, to after the
1966 CRLF.
1967
1968 The above rule is a compromise that makes the most common cases work as
1969 expected. For example, if the pattern is .+A (and the PCRE_DOTALL
1970 option is not set), it does not match the string "\r\nA" because, after
1971 failing at the start, it skips both the CR and the LF before retrying.
1972 However, the pattern [\r\n]A does match that string, because it con-
1973 tains an explicit CR or LF reference, and so advances only by one char-
1974 acter after the first failure.
1975
1976 An explicit match for CR of LF is either a literal appearance of one of
1977 those characters, or one of the \r or \n escape sequences. Implicit
1978 matches such as [^X] do not count, nor does \s (which includes CR and
1979 LF in the characters that it matches).
1980
1981 Notwithstanding the above, anomalous effects may still occur when CRLF
1982 is a valid newline sequence and explicit \r or \n escapes appear in the
1983 pattern.
1984
1985 PCRE_NOTBOL
1986
1987 This option specifies that first character of the subject string is not
1988 the beginning of a line, so the circumflex metacharacter should not
1989 match before it. Setting this without PCRE_MULTILINE (at compile time)
1990 causes circumflex never to match. This option affects only the behav-
1991 iour of the circumflex metacharacter. It does not affect \A.
1992
1993 PCRE_NOTEOL
1994
1995 This option specifies that the end of the subject string is not the end
1996 of a line, so the dollar metacharacter should not match it nor (except
1997 in multiline mode) a newline immediately before it. Setting this with-
1998 out PCRE_MULTILINE (at compile time) causes dollar never to match. This
1999 option affects only the behaviour of the dollar metacharacter. It does
2000 not affect \Z or \z.
2001
2002 PCRE_NOTEMPTY
2003
2004 An empty string is not considered to be a valid match if this option is
2005 set. If there are alternatives in the pattern, they are tried. If all
2006 the alternatives match the empty string, the entire match fails. For
2007 example, if the pattern
2008
2009 a?b?
2010
2011 is applied to a string not beginning with "a" or "b", it matches the
2012 empty string at the start of the subject. With PCRE_NOTEMPTY set, this
2013 match is not valid, so PCRE searches further into the string for occur-
2014 rences of "a" or "b".
2015
2016 Perl has no direct equivalent of PCRE_NOTEMPTY, but it does make a spe-
2017 cial case of a pattern match of the empty string within its split()
2018 function, and when using the /g modifier. It is possible to emulate
2019 Perl's behaviour after matching a null string by first trying the match
2020 again at the same offset with PCRE_NOTEMPTY and PCRE_ANCHORED, and then
2021 if that fails by advancing the starting offset (see below) and trying
2022 an ordinary match again. There is some code that demonstrates how to do
2023 this in the pcredemo sample program.
2024
2025 PCRE_NO_START_OPTIMIZE
2026
2027 There are a number of optimizations that pcre_exec() uses at the start
2028 of a match, in order to speed up the process. For example, if it is
2029 known that a match must start with a specific character, it searches
2030 the subject for that character, and fails immediately if it cannot find
2031 it, without actually running the main matching function. When callouts
2032 are in use, these optimizations can cause them to be skipped. This
2033 option disables the "start-up" optimizations, causing performance to
2034 suffer, but ensuring that the callouts do occur.
2035
2036 PCRE_NO_UTF8_CHECK
2037
2038 When PCRE_UTF8 is set at compile time, the validity of the subject as a
2039 UTF-8 string is automatically checked when pcre_exec() is subsequently
2040 called. The value of startoffset is also checked to ensure that it
2041 points to the start of a UTF-8 character. There is a discussion about
2042 the validity of UTF-8 strings in the section on UTF-8 support in the
2043 main pcre page. If an invalid UTF-8 sequence of bytes is found,
2044 pcre_exec() returns the error PCRE_ERROR_BADUTF8. If startoffset con-
2045 tains an invalid value, PCRE_ERROR_BADUTF8_OFFSET is returned.
2046
2047 If you already know that your subject is valid, and you want to skip
2048 these checks for performance reasons, you can set the
2049 PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to
2050 do this for the second and subsequent calls to pcre_exec() if you are
2051 making repeated calls to find all the matches in a single subject
2052 string. However, you should be sure that the value of startoffset
2053 points to the start of a UTF-8 character. When PCRE_NO_UTF8_CHECK is
2054 set, the effect of passing an invalid UTF-8 string as a subject, or a
2055 value of startoffset that does not point to the start of a UTF-8 char-
2056 acter, is undefined. Your program may crash.
2057
2058 PCRE_PARTIAL_HARD
2059 PCRE_PARTIAL_SOFT
2060
2061 These options turn on the partial matching feature. For backwards com-
2062 patibility, PCRE_PARTIAL is a synonym for PCRE_PARTIAL_SOFT. A partial
2063 match occurs if the end of the subject string is reached successfully,
2064 but there are not enough subject characters to complete the match. If
2065 this happens when PCRE_PARTIAL_HARD is set, pcre_exec() immediately
2066 returns PCRE_ERROR_PARTIAL. Otherwise, if PCRE_PARTIAL_SOFT is set,
2067 matching continues by testing any other alternatives. Only if they all
2068 fail is PCRE_ERROR_PARTIAL returned (instead of PCRE_ERROR_NOMATCH).
2069 The portion of the string that provided the partial match is set as the
2070 first matching string. There is a more detailed discussion in the
2071 pcrepartial documentation.
2072
2073 The string to be matched by pcre_exec()
2074
2075 The subject string is passed to pcre_exec() as a pointer in subject, a
2076 length (in bytes) in length, and a starting byte offset in startoffset.
2077 In UTF-8 mode, the byte offset must point to the start of a UTF-8 char-
2078 acter. Unlike the pattern string, the subject may contain binary zero
2079 bytes. When the starting offset is zero, the search for a match starts
2080 at the beginning of the subject, and this is by far the most common
2081 case.
2082
2083 A non-zero starting offset is useful when searching for another match
2084 in the same subject by calling pcre_exec() again after a previous suc-
2085 cess. Setting startoffset differs from just passing over a shortened
2086 string and setting PCRE_NOTBOL in the case of a pattern that begins
2087 with any kind of lookbehind. For example, consider the pattern
2088
2089 \Biss\B
2090
2091 which finds occurrences of "iss" in the middle of words. (\B matches
2092 only if the current position in the subject is not a word boundary.)
2093 When applied to the string "Mississipi" the first call to pcre_exec()
2094 finds the first occurrence. If pcre_exec() is called again with just
2095 the remainder of the subject, namely "issipi", it does not match,
2096 because \B is always false at the start of the subject, which is deemed
2097 to be a word boundary. However, if pcre_exec() is passed the entire
2098 string again, but with startoffset set to 4, it finds the second occur-
2099 rence of "iss" because it is able to look behind the starting point to
2100 discover that it is preceded by a letter.
2101
2102 If a non-zero starting offset is passed when the pattern is anchored,
2103 one attempt to match at the given offset is made. This can only succeed
2104 if the pattern does not require the match to be at the start of the
2105 subject.
2106
2107 How pcre_exec() returns captured substrings
2108
2109 In general, a pattern matches a certain portion of the subject, and in
2110 addition, further substrings from the subject may be picked out by
2111 parts of the pattern. Following the usage in Jeffrey Friedl's book,
2112 this is called "capturing" in what follows, and the phrase "capturing
2113 subpattern" is used for a fragment of a pattern that picks out a sub-
2114 string. PCRE supports several other kinds of parenthesized subpattern
2115 that do not cause substrings to be captured.
2116
2117 Captured substrings are returned to the caller via a vector of integers
2118 whose address is passed in ovector. The number of elements in the vec-
2119 tor is passed in ovecsize, which must be a non-negative number. Note:
2120 this argument is NOT the size of ovector in bytes.
2121
2122 The first two-thirds of the vector is used to pass back captured sub-
2123 strings, each substring using a pair of integers. The remaining third
2124 of the vector is used as workspace by pcre_exec() while matching cap-
2125 turing subpatterns, and is not available for passing back information.
2126 The number passed in ovecsize should always be a multiple of three. If
2127 it is not, it is rounded down.
2128
2129 When a match is successful, information about captured substrings is
2130 returned in pairs of integers, starting at the beginning of ovector,
2131 and continuing up to two-thirds of its length at the most. The first
2132 element of each pair is set to the byte offset of the first character
2133 in a substring, and the second is set to the byte offset of the first
2134 character after the end of a substring. Note: these values are always
2135 byte offsets, even in UTF-8 mode. They are not character counts.
2136
2137 The first pair of integers, ovector[0] and ovector[1], identify the
2138 portion of the subject string matched by the entire pattern. The next
2139 pair is used for the first capturing subpattern, and so on. The value
2140 returned by pcre_exec() is one more than the highest numbered pair that
2141 has been set. For example, if two substrings have been captured, the
2142 returned value is 3. If there are no capturing subpatterns, the return
2143 value from a successful match is 1, indicating that just the first pair
2144 of offsets has been set.
2145
2146 If a capturing subpattern is matched repeatedly, it is the last portion
2147 of the string that it matched that is returned.
2148
2149 If the vector is too small to hold all the captured substring offsets,
2150 it is used as far as possible (up to two-thirds of its length), and the
2151 function returns a value of zero. If the substring offsets are not of
2152 interest, pcre_exec() may be called with ovector passed as NULL and
2153 ovecsize as zero. However, if the pattern contains back references and
2154 the ovector is not big enough to remember the related substrings, PCRE
2155 has to get additional memory for use during matching. Thus it is usu-
2156 ally advisable to supply an ovector.
2157
2158 The pcre_info() function can be used to find out how many capturing
2159 subpatterns there are in a compiled pattern. The smallest size for
2160 ovector that will allow for n captured substrings, in addition to the
2161 offsets of the substring matched by the whole pattern, is (n+1)*3.
2162
2163 It is possible for capturing subpattern number n+1 to match some part
2164 of the subject when subpattern n has not been used at all. For example,
2165 if the string "abc" is matched against the pattern (a|(z))(bc) the
2166 return from the function is 4, and subpatterns 1 and 3 are matched, but
2167 2 is not. When this happens, both values in the offset pairs corre-
2168 sponding to unused subpatterns are set to -1.
2169
2170 Offset values that correspond to unused subpatterns at the end of the
2171 expression are also set to -1. For example, if the string "abc" is
2172 matched against the pattern (abc)(x(yz)?)? subpatterns 2 and 3 are not
2173 matched. The return from the function is 2, because the highest used
2174 capturing subpattern number is 1. However, you can refer to the offsets
2175 for the second and third capturing subpatterns if you wish (assuming
2176 the vector is large enough, of course).
2177
2178 Some convenience functions are provided for extracting the captured
2179 substrings as separate strings. These are described below.
2180
2181 Error return values from pcre_exec()
2182
2183 If pcre_exec() fails, it returns a negative number. The following are
2184 defined in the header file:
2185
2186 PCRE_ERROR_NOMATCH (-1)
2187
2188 The subject string did not match the pattern.
2189
2190 PCRE_ERROR_NULL (-2)
2191
2192 Either code or subject was passed as NULL, or ovector was NULL and
2193 ovecsize was not zero.
2194
2195 PCRE_ERROR_BADOPTION (-3)
2196
2197 An unrecognized bit was set in the options argument.
2198
2199 PCRE_ERROR_BADMAGIC (-4)
2200
2201 PCRE stores a 4-byte "magic number" at the start of the compiled code,
2202 to catch the case when it is passed a junk pointer and to detect when a
2203 pattern that was compiled in an environment of one endianness is run in
2204 an environment with the other endianness. This is the error that PCRE
2205 gives when the magic number is not present.
2206
2207 PCRE_ERROR_UNKNOWN_OPCODE (-5)
2208
2209 While running the pattern match, an unknown item was encountered in the
2210 compiled pattern. This error could be caused by a bug in PCRE or by
2211 overwriting of the compiled pattern.
2212
2213 PCRE_ERROR_NOMEMORY (-6)
2214
2215 If a pattern contains back references, but the ovector that is passed
2216 to pcre_exec() is not big enough to remember the referenced substrings,
2217 PCRE gets a block of memory at the start of matching to use for this
2218 purpose. If the call via pcre_malloc() fails, this error is given. The
2219 memory is automatically freed at the end of matching.
2220
2221 PCRE_ERROR_NOSUBSTRING (-7)
2222
2223 This error is used by the pcre_copy_substring(), pcre_get_substring(),
2224 and pcre_get_substring_list() functions (see below). It is never
2225 returned by pcre_exec().
2226
2227 PCRE_ERROR_MATCHLIMIT (-8)
2228
2229 The backtracking limit, as specified by the match_limit field in a
2230 pcre_extra structure (or defaulted) was reached. See the description
2231 above.
2232
2233 PCRE_ERROR_CALLOUT (-9)
2234
2235 This error is never generated by pcre_exec() itself. It is provided for
2236 use by callout functions that want to yield a distinctive error code.
2237 See the pcrecallout documentation for details.
2238
2239 PCRE_ERROR_BADUTF8 (-10)
2240
2241 A string that contains an invalid UTF-8 byte sequence was passed as a
2242 subject.
2243
2244 PCRE_ERROR_BADUTF8_OFFSET (-11)
2245
2246 The UTF-8 byte sequence that was passed as a subject was valid, but the
2247 value of startoffset did not point to the beginning of a UTF-8 charac-
2248 ter.
2249
2250 PCRE_ERROR_PARTIAL (-12)
2251
2252 The subject string did not match, but it did match partially. See the
2253 pcrepartial documentation for details of partial matching.
2254
2255 PCRE_ERROR_BADPARTIAL (-13)
2256
2257 This code is no longer in use. It was formerly returned when the
2258 PCRE_PARTIAL option was used with a compiled pattern containing items
2259 that were not supported for partial matching. From release 8.00
2260 onwards, there are no restrictions on partial matching.
2261
2262 PCRE_ERROR_INTERNAL (-14)
2263
2264 An unexpected internal error has occurred. This error could be caused
2265 by a bug in PCRE or by overwriting of the compiled pattern.
2266
2267 PCRE_ERROR_BADCOUNT (-15)
2268
2269 This error is given if the value of the ovecsize argument is negative.
2270
2271 PCRE_ERROR_RECURSIONLIMIT (-21)
2272
2273 The internal recursion limit, as specified by the match_limit_recursion
2274 field in a pcre_extra structure (or defaulted) was reached. See the
2275 description above.
2276
2277 PCRE_ERROR_BADNEWLINE (-23)
2278
2279 An invalid combination of PCRE_NEWLINE_xxx options was given.
2280
2281 Error numbers -16 to -20 and -22 are not used by pcre_exec().
2282
2283
2284 EXTRACTING CAPTURED SUBSTRINGS BY NUMBER
2285
2286 int pcre_copy_substring(const char *subject, int *ovector,
2287 int stringcount, int stringnumber, char *buffer,
2288 int buffersize);
2289
2290 int pcre_get_substring(const char *subject, int *ovector,
2291 int stringcount, int stringnumber,
2292 const char **stringptr);
2293
2294 int pcre_get_substring_list(const char *subject,
2295 int *ovector, int stringcount, const char ***listptr);
2296
2297 Captured substrings can be accessed directly by using the offsets
2298 returned by pcre_exec() in ovector. For convenience, the functions
2299 pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub-
2300 string_list() are provided for extracting captured substrings as new,
2301 separate, zero-terminated strings. These functions identify substrings
2302 by number. The next section describes functions for extracting named
2303 substrings.
2304
2305 A substring that contains a binary zero is correctly extracted and has
2306 a further zero added on the end, but the result is not, of course, a C
2307 string. However, you can process such a string by referring to the
2308 length that is returned by pcre_copy_substring() and pcre_get_sub-
2309 string(). Unfortunately, the interface to pcre_get_substring_list() is
2310 not adequate for handling strings containing binary zeros, because the
2311 end of the final string is not independently indicated.
2312
2313 The first three arguments are the same for all three of these func-
2314 tions: subject is the subject string that has just been successfully
2315 matched, ovector is a pointer to the vector of integer offsets that was
2316 passed to pcre_exec(), and stringcount is the number of substrings that
2317 were captured by the match, including the substring that matched the
2318 entire regular expression. This is the value returned by pcre_exec() if
2319 it is greater than zero. If pcre_exec() returned zero, indicating that
2320 it ran out of space in ovector, the value passed as stringcount should
2321 be the number of elements in the vector divided by three.
2322
2323 The functions pcre_copy_substring() and pcre_get_substring() extract a
2324 single substring, whose number is given as stringnumber. A value of
2325 zero extracts the substring that matched the entire pattern, whereas
2326 higher values extract the captured substrings. For pcre_copy_sub-
2327 string(), the string is placed in buffer, whose length is given by
2328 buffersize, while for pcre_get_substring() a new block of memory is
2329 obtained via pcre_malloc, and its address is returned via stringptr.
2330 The yield of the function is the length of the string, not including
2331 the terminating zero, or one of these error codes:
2332
2333 PCRE_ERROR_NOMEMORY (-6)
2334
2335 The buffer was too small for pcre_copy_substring(), or the attempt to
2336 get memory failed for pcre_get_substring().
2337
2338 PCRE_ERROR_NOSUBSTRING (-7)
2339
2340 There is no substring whose number is stringnumber.
2341
2342 The pcre_get_substring_list() function extracts all available sub-
2343 strings and builds a list of pointers to them. All this is done in a
2344 single block of memory that is obtained via pcre_malloc. The address of
2345 the memory block is returned via listptr, which is also the start of
2346 the list of string pointers. The end of the list is marked by a NULL
2347 pointer. The yield of the function is zero if all went well, or the
2348 error code
2349
2350 PCRE_ERROR_NOMEMORY (-6)
2351
2352 if the attempt to get the memory block failed.
2353
2354 When any of these functions encounter a substring that is unset, which
2355 can happen when capturing subpattern number n+1 matches some part of
2356 the subject, but subpattern n has not been used at all, they return an
2357 empty string. This can be distinguished from a genuine zero-length sub-
2358 string by inspecting the appropriate offset in ovector, which is nega-
2359 tive for unset substrings.
2360
2361 The two convenience functions pcre_free_substring() and pcre_free_sub-
2362 string_list() can be used to free the memory returned by a previous
2363 call of pcre_get_substring() or pcre_get_substring_list(), respec-
2364 tively. They do nothing more than call the function pointed to by
2365 pcre_free, which of course could be called directly from a C program.
2366 However, PCRE is used in some situations where it is linked via a spe-
2367 cial interface to another programming language that cannot use
2368 pcre_free directly; it is for these cases that the functions are pro-
2369 vided.
2370
2371
2372 EXTRACTING CAPTURED SUBSTRINGS BY NAME
2373
2374 int pcre_get_stringnumber(const pcre *code,
2375 const char *name);
2376
2377 int pcre_copy_named_substring(const pcre *code,
2378 const char *subject, int *ovector,
2379 int stringcount, const char *stringname,
2380 char *buffer, int buffersize);
2381
2382 int pcre_get_named_substring(const pcre *code,
2383 const char *subject, int *ovector,
2384 int stringcount, const char *stringname,
2385 const char **stringptr);
2386
2387 To extract a substring by name, you first have to find associated num-
2388 ber. For example, for this pattern
2389
2390 (a+)b(?<xxx>\d+)...
2391
2392 the number of the subpattern called "xxx" is 2. If the name is known to
2393 be unique (PCRE_DUPNAMES was not set), you can find the number from the
2394 name by calling pcre_get_stringnumber(). The first argument is the com-
2395 piled pattern, and the second is the name. The yield of the function is
2396 the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if there is no
2397 subpattern of that name.
2398
2399 Given the number, you can extract the substring directly, or use one of
2400 the functions described in the previous section. For convenience, there
2401 are also two functions that do the whole job.
2402
2403 Most of the arguments of pcre_copy_named_substring() and
2404 pcre_get_named_substring() are the same as those for the similarly
2405 named functions that extract by number. As these are described in the
2406 previous section, they are not re-described here. There are just two
2407 differences:
2408
2409 First, instead of a substring number, a substring name is given. Sec-
2410 ond, there is an extra argument, given at the start, which is a pointer
2411 to the compiled pattern. This is needed in order to gain access to the
2412 name-to-number translation table.
2413
2414 These functions call pcre_get_stringnumber(), and if it succeeds, they
2415 then call pcre_copy_substring() or pcre_get_substring(), as appropri-
2416 ate. NOTE: If PCRE_DUPNAMES is set and there are duplicate names, the
2417 behaviour may not be what you want (see the next section).
2418
2419 Warning: If the pattern uses the "(?|" feature to set up multiple sub-
2420 patterns with the same number, you cannot use names to distinguish
2421 them, because names are not included in the compiled code. The matching
2422 process uses only numbers.
2423
2424
2425 DUPLICATE SUBPATTERN NAMES
2426
2427 int pcre_get_stringtable_entries(const pcre *code,
2428 const char *name, char **first, char **last);
2429
2430 When a pattern is compiled with the PCRE_DUPNAMES option, names for
2431 subpatterns are not required to be unique. Normally, patterns with
2432 duplicate names are such that in any one match, only one of the named
2433 subpatterns participates. An example is shown in the pcrepattern docu-
2434 mentation.
2435
2436 When duplicates are present, pcre_copy_named_substring() and
2437 pcre_get_named_substring() return the first substring corresponding to
2438 the given name that is set. If none are set, PCRE_ERROR_NOSUBSTRING
2439 (-7) is returned; no data is returned. The pcre_get_stringnumber()
2440 function returns one of the numbers that are associated with the name,
2441 but it is not defined which it is.
2442
2443 If you want to get full details of all captured substrings for a given
2444 name, you must use the pcre_get_stringtable_entries() function. The
2445 first argument is the compiled pattern, and the second is the name. The
2446 third and fourth are pointers to variables which are updated by the
2447 function. After it has run, they point to the first and last entries in
2448 the name-to-number table for the given name. The function itself
2449 returns the length of each entry, or PCRE_ERROR_NOSUBSTRING (-7) if
2450 there are none. The format of the table is described above in the sec-
2451 tion entitled Information about a pattern. Given all the relevant
2452 entries for the name, you can extract each of their numbers, and hence
2453 the captured data, if any.
2454
2455
2456 FINDING ALL POSSIBLE MATCHES
2457
2458 The traditional matching function uses a similar algorithm to Perl,
2459 which stops when it finds the first match, starting at a given point in
2460 the subject. If you want to find all possible matches, or the longest
2461 possible match, consider using the alternative matching function (see
2462 below) instead. If you cannot use the alternative function, but still
2463 need to find all possible matches, you can kludge it up by making use
2464 of the callout facility, which is described in the pcrecallout documen-
2465 tation.
2466
2467 What you have to do is to insert a callout right at the end of the pat-
2468 tern. When your callout function is called, extract and save the cur-
2469 rent matched substring. Then return 1, which forces pcre_exec() to
2470 backtrack and try other alternatives. Ultimately, when it runs out of
2471 matches, pcre_exec() will yield PCRE_ERROR_NOMATCH.
2472
2473
2474 MATCHING A PATTERN: THE ALTERNATIVE FUNCTION
2475
2476 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
2477 const char *subject, int length, int startoffset,
2478 int options, int *ovector, int ovecsize,
2479 int *workspace, int wscount);
2480
2481 The function pcre_dfa_exec() is called to match a subject string
2482 against a compiled pattern, using a matching algorithm that scans the
2483 subject string just once, and does not backtrack. This has different
2484 characteristics to the normal algorithm, and is not compatible with
2485 Perl. Some of the features of PCRE patterns are not supported. Never-
2486 theless, there are times when this kind of matching can be useful. For
2487 a discussion of the two matching algorithms, see the pcrematching docu-
2488 mentation.
2489
2490 The arguments for the pcre_dfa_exec() function are the same as for
2491 pcre_exec(), plus two extras. The ovector argument is used in a differ-
2492 ent way, and this is described below. The other common arguments are
2493 used in the same way as for pcre_exec(), so their description is not
2494 repeated here.
2495
2496 The two additional arguments provide workspace for the function. The
2497 workspace vector should contain at least 20 elements. It is used for
2498 keeping track of multiple paths through the pattern tree. More
2499 workspace will be needed for patterns and subjects where there are a
2500 lot of potential matches.
2501
2502 Here is an example of a simple call to pcre_dfa_exec():
2503
2504 int rc;
2505 int ovector[10];
2506 int wspace[20];
2507 rc = pcre_dfa_exec(
2508 re, /* result of pcre_compile() */
2509 NULL, /* we didn't study the pattern */
2510 "some string", /* the subject string */
2511 11, /* the length of the subject string */
2512 0, /* start at offset 0 in the subject */
2513 0, /* default options */
2514 ovector, /* vector of integers for substring information */
2515 10, /* number of elements (NOT size in bytes) */
2516 wspace, /* working space vector */
2517 20); /* number of elements (NOT size in bytes) */
2518
2519 Option bits for pcre_dfa_exec()
2520
2521 The unused bits of the options argument for pcre_dfa_exec() must be
2522 zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NEW-
2523 LINE_xxx, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NO_UTF8_CHECK,
2524 PCRE_PARTIAL_HARD, PCRE_PARTIAL_SOFT, PCRE_DFA_SHORTEST, and
2525 PCRE_DFA_RESTART. All but the last four of these are exactly the same
2526 as for pcre_exec(), so their description is not repeated here.
2527
2528 PCRE_PARTIAL_HARD
2529 PCRE_PARTIAL_SOFT
2530
2531 These have the same general effect as they do for pcre_exec(), but the
2532 details are slightly different. When PCRE_PARTIAL_HARD is set for
2533 pcre_dfa_exec(), it returns PCRE_ERROR_PARTIAL if the end of the sub-
2534 ject is reached and there is still at least one matching possibility
2535 that requires additional characters. This happens even if some complete
2536 matches have also been found. When PCRE_PARTIAL_SOFT is set, the return
2537 code PCRE_ERROR_NOMATCH is converted into PCRE_ERROR_PARTIAL if the end
2538 of the subject is reached, there have been no complete matches, but
2539 there is still at least one matching possibility. The portion of the
2540 string that provided the longest partial match is set as the first
2541 matching string in both cases.
2542
2543 PCRE_DFA_SHORTEST
2544
2545 Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to
2546 stop as soon as it has found one match. Because of the way the alterna-
2547 tive algorithm works, this is necessarily the shortest possible match
2548 at the first possible matching point in the subject string.
2549
2550 PCRE_DFA_RESTART
2551
2552 When pcre_dfa_exec() returns a partial match, it is possible to call it
2553 again, with additional subject characters, and have it continue with
2554 the same match. The PCRE_DFA_RESTART option requests this action; when
2555 it is set, the workspace and wscount options must reference the same
2556 vector as before because data about the match so far is left in them
2557 after a partial match. There is more discussion of this facility in the
2558 pcrepartial documentation.
2559
2560 Successful returns from pcre_dfa_exec()
2561
2562 When pcre_dfa_exec() succeeds, it may have matched more than one sub-
2563 string in the subject. Note, however, that all the matches from one run
2564 of the function start at the same point in the subject. The shorter
2565 matches are all initial substrings of the longer matches. For example,
2566 if the pattern
2567
2568 <.*>
2569
2570 is matched against the string
2571
2572 This is <something> <something else> <something further> no more
2573
2574 the three matched strings are
2575
2576 <something>
2577 <something> <something else>
2578 <something> <something else> <something further>
2579
2580 On success, the yield of the function is a number greater than zero,
2581 which is the number of matched substrings. The substrings themselves
2582 are returned in ovector. Each string uses two elements; the first is
2583 the offset to the start, and the second is the offset to the end. In
2584 fact, all the strings have the same start offset. (Space could have
2585 been saved by giving this only once, but it was decided to retain some
2586 compatibility with the way pcre_exec() returns data, even though the
2587 meaning of the strings is different.)
2588
2589 The strings are returned in reverse order of length; that is, the long-
2590 est matching string is given first. If there were too many matches to
2591 fit into ovector, the yield of the function is zero, and the vector is
2592 filled with the longest matches.
2593
2594 Error returns from pcre_dfa_exec()
2595
2596 The pcre_dfa_exec() function returns a negative number when it fails.
2597 Many of the errors are the same as for pcre_exec(), and these are
2598 described above. There are in addition the following errors that are
2599 specific to pcre_dfa_exec():
2600
2601 PCRE_ERROR_DFA_UITEM (-16)
2602
2603 This return is given if pcre_dfa_exec() encounters an item in the pat-
2604 tern that it does not support, for instance, the use of \C or a back
2605 reference.
2606
2607 PCRE_ERROR_DFA_UCOND (-17)
2608
2609 This return is given if pcre_dfa_exec() encounters a condition item
2610 that uses a back reference for the condition, or a test for recursion
2611 in a specific group. These are not supported.
2612
2613 PCRE_ERROR_DFA_UMLIMIT (-18)
2614
2615 This return is given if pcre_dfa_exec() is called with an extra block
2616 that contains a setting of the match_limit field. This is not supported
2617 (it is meaningless).
2618
2619 PCRE_ERROR_DFA_WSSIZE (-19)
2620
2621 This return is given if pcre_dfa_exec() runs out of space in the
2622 workspace vector.
2623
2624 PCRE_ERROR_DFA_RECURSE (-20)
2625
2626 When a recursive subpattern is processed, the matching function calls
2627 itself recursively, using private vectors for ovector and workspace.
2628 This error is given if the output vector is not large enough. This
2629 should be extremely rare, as a vector of size 1000 is used.
2630
2631
2632 SEE ALSO
2633
2634 pcrebuild(3), pcrecallout(3), pcrecpp(3)(3), pcrematching(3), pcrepar-
2635 tial(3), pcreposix(3), pcreprecompile(3), pcresample(3), pcrestack(3).
2636
2637
2638 AUTHOR
2639
2640 Philip Hazel
2641 University Computing Service
2642 Cambridge CB2 3QH, England.
2643
2644
2645 REVISION
2646
2647 Last updated: 01 September 2009
2648 Copyright (c) 1997-2009 University of Cambridge.
2649 ------------------------------------------------------------------------------
2650
2651
2652 PCRECALLOUT(3) PCRECALLOUT(3)
2653
2654
2655 NAME
2656 PCRE - Perl-compatible regular expressions
2657
2658
2659 PCRE CALLOUTS
2660
2661 int (*pcre_callout)(pcre_callout_block *);
2662
2663 PCRE provides a feature called "callout", which is a means of temporar-
2664 ily passing control to the caller of PCRE in the middle of pattern
2665 matching. The caller of PCRE provides an external function by putting
2666 its entry point in the global variable pcre_callout. By default, this
2667 variable contains NULL, which disables all calling out.
2668
2669 Within a regular expression, (?C) indicates the points at which the
2670 external function is to be called. Different callout points can be
2671 identified by putting a number less than 256 after the letter C. The
2672 default value is zero. For example, this pattern has two callout
2673 points:
2674
2675 (?C1)abc(?C2)def
2676
2677 If the PCRE_AUTO_CALLOUT option bit is set when pcre_compile() is
2678 called, PCRE automatically inserts callouts, all with number 255,
2679 before each item in the pattern. For example, if PCRE_AUTO_CALLOUT is
2680 used with the pattern
2681
2682 A(\d{2}|--)
2683
2684 it is processed as if it were
2685
2686 (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255)
2687
2688 Notice that there is a callout before and after each parenthesis and
2689 alternation bar. Automatic callouts can be used for tracking the
2690 progress of pattern matching. The pcretest command has an option that
2691 sets automatic callouts; when it is used, the output indicates how the
2692 pattern is matched. This is useful information when you are trying to
2693 optimize the performance of a particular pattern.
2694
2695
2696 MISSING CALLOUTS
2697
2698 You should be aware that, because of optimizations in the way PCRE
2699 matches patterns by default, callouts sometimes do not happen. For
2700 example, if the pattern is
2701
2702 ab(?C4)cd
2703
2704 PCRE knows that any matching string must contain the letter "d". If the
2705 subject string is "abyz", the lack of "d" means that matching doesn't
2706 ever start, and the callout is never reached. However, with "abyd",
2707 though the result is still no match, the callout is obeyed.
2708
2709 You can disable these optimizations by passing the PCRE_NO_START_OPTI-
2710 MIZE option to pcre_exec() or pcre_dfa_exec(). This slows down the
2711 matching process, but does ensure that callouts such as the example
2712 above are obeyed.
2713
2714
2715 THE CALLOUT INTERFACE
2716
2717 During matching, when PCRE reaches a callout point, the external func-
2718 tion defined by pcre_callout is called (if it is set). This applies to
2719 both the pcre_exec() and the pcre_dfa_exec() matching functions. The
2720 only argument to the callout function is a pointer to a pcre_callout
2721 block. This structure contains the following fields:
2722
2723 int version;
2724 int callout_number;
2725 int *offset_vector;
2726 const char *subject;
2727 int subject_length;
2728 int start_match;
2729 int current_position;
2730 int capture_top;
2731 int capture_last;
2732 void *callout_data;
2733 int pattern_position;
2734 int next_item_length;
2735
2736 The version field is an integer containing the version number of the
2737 block format. The initial version was 0; the current version is 1. The
2738 version number will change again in future if additional fields are
2739 added, but the intention is never to remove any of the existing fields.
2740
2741 The callout_number field contains the number of the callout, as com-
2742 piled into the pattern (that is, the number after ?C for manual call-
2743 outs, and 255 for automatically generated callouts).
2744
2745 The offset_vector field is a pointer to the vector of offsets that was
2746 passed by the caller to pcre_exec() or pcre_dfa_exec(). When
2747 pcre_exec() is used, the contents can be inspected in order to extract
2748 substrings that have been matched so far, in the same way as for
2749 extracting substrings after a match has completed. For pcre_dfa_exec()
2750 this field is not useful.
2751
2752 The subject and subject_length fields contain copies of the values that
2753 were passed to pcre_exec().
2754
2755 The start_match field normally contains the offset within the subject
2756 at which the current match attempt started. However, if the escape
2757 sequence \K has been encountered, this value is changed to reflect the
2758 modified starting point. If the pattern is not anchored, the callout
2759 function may be called several times from the same point in the pattern
2760 for different starting points in the subject.
2761
2762 The current_position field contains the offset within the subject of
2763 the current match pointer.
2764
2765 When the pcre_exec() function is used, the capture_top field contains
2766 one more than the number of the highest numbered captured substring so
2767 far. If no substrings have been captured, the value of capture_top is
2768 one. This is always the case when pcre_dfa_exec() is used, because it
2769 does not support captured substrings.
2770
2771 The capture_last field contains the number of the most recently cap-
2772 tured substring. If no substrings have been captured, its value is -1.
2773 This is always the case when pcre_dfa_exec() is used.
2774
2775 The callout_data field contains a value that is passed to pcre_exec()
2776 or pcre_dfa_exec() specifically so that it can be passed back in call-
2777 outs. It is passed in the pcre_callout field of the pcre_extra data
2778 structure. If no such data was passed, the value of callout_data in a
2779 pcre_callout block is NULL. There is a description of the pcre_extra
2780 structure in the pcreapi documentation.
2781
2782 The pattern_position field is present from version 1 of the pcre_call-
2783 out structure. It contains the offset to the next item to be matched in
2784 the pattern string.
2785
2786 The next_item_length field is present from version 1 of the pcre_call-
2787 out structure. It contains the length of the next item to be matched in
2788 the pattern string. When the callout immediately precedes an alterna-
2789 tion bar, a closing parenthesis, or the end of the pattern, the length
2790 is zero. When the callout precedes an opening parenthesis, the length
2791 is that of the entire subpattern.
2792
2793 The pattern_position and next_item_length fields are intended to help
2794 in distinguishing between different automatic callouts, which all have
2795 the same callout number. However, they are set for all callouts.
2796
2797
2798 RETURN VALUES
2799
2800 The external callout function returns an integer to PCRE. If the value
2801 is zero, matching proceeds as normal. If the value is greater than
2802 zero, matching fails at the current point, but the testing of other
2803 matching possibilities goes ahead, just as if a lookahead assertion had
2804 failed. If the value is less than zero, the match is abandoned, and
2805 pcre_exec() (or pcre_dfa_exec()) returns the negative value.
2806
2807 Negative values should normally be chosen from the set of
2808 PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan-
2809 dard "no match" failure. The error number PCRE_ERROR_CALLOUT is
2810 reserved for use by callout functions; it will never be used by PCRE
2811 itself.
2812
2813
2814 AUTHOR
2815
2816 Philip Hazel
2817 University Computing Service
2818 Cambridge CB2 3QH, England.
2819
2820
2821 REVISION
2822
2823 Last updated: 15 March 2009
2824 Copyright (c) 1997-2009 University of Cambridge.
2825 ------------------------------------------------------------------------------
2826
2827
2828 PCRECOMPAT(3) PCRECOMPAT(3)
2829
2830
2831 NAME
2832 PCRE - Perl-compatible regular expressions
2833
2834
2835 DIFFERENCES BETWEEN PCRE AND PERL
2836
2837 This document describes the differences in the ways that PCRE and Perl
2838 handle regular expressions. The differences described here are mainly
2839 with respect to Perl 5.8, though PCRE versions 7.0 and later contain
2840 some features that are in Perl 5.10.
2841
2842 1. PCRE has only a subset of Perl's UTF-8 and Unicode support. Details
2843 of what it does have are given in the section on UTF-8 support in the
2844 main pcre page.
2845
2846 2. PCRE does not allow repeat quantifiers on lookahead assertions. Perl
2847 permits them, but they do not mean what you might think. For example,
2848 (?!a){3} does not assert that the next three characters are not "a". It
2849 just asserts that the next character is not "a" three times.
2850
2851 3. Capturing subpatterns that occur inside negative lookahead asser-
2852 tions are counted, but their entries in the offsets vector are never
2853 set. Perl sets its numerical variables from any such patterns that are
2854 matched before the assertion fails to match something (thereby succeed-
2855 ing), but only if the negative lookahead assertion contains just one
2856 branch.
2857
2858 4. Though binary zero characters are supported in the subject string,
2859 they are not allowed in a pattern string because it is passed as a nor-
2860 mal C string, terminated by zero. The escape sequence \0 can be used in
2861 the pattern to represent a binary zero.
2862
2863 5. The following Perl escape sequences are not supported: \l, \u, \L,
2864 \U, and \N. In fact these are implemented by Perl's general string-han-
2865 dling and are not part of its pattern matching engine. If any of these
2866 are encountered by PCRE, an error is generated.
2867
2868 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE
2869 is built with Unicode character property support. The properties that
2870 can be tested with \p and \P are limited to the general category prop-
2871 erties such as Lu and Nd, script names such as Greek or Han, and the
2872 derived properties Any and L&.
2873
2874 7. PCRE does support the \Q...\E escape for quoting substrings. Charac-
2875 ters in between are treated as literals. This is slightly different
2876 from Perl in that $ and @ are also handled as literals inside the
2877 quotes. In Perl, they cause variable interpolation (but of course PCRE
2878 does not have variables). Note the following examples:
2879
2880 Pattern PCRE matches Perl matches
2881
2882 \Qabc$xyz\E abc$xyz abc followed by the
2883 contents of $xyz
2884 \Qabc\$xyz\E abc\$xyz abc\$xyz
2885 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
2886
2887 The \Q...\E sequence is recognized both inside and outside character
2888 classes.
2889
2890 8. Fairly obviously, PCRE does not support the (?{code}) and (??{code})
2891 constructions. However, there is support for recursive patterns. This
2892 is not available in Perl 5.8, but will be in Perl 5.10. Also, the PCRE
2893 "callout" feature allows an external function to be called during pat-
2894 tern matching. See the pcrecallout documentation for details.
2895
2896 9. Subpatterns that are called recursively or as "subroutines" are
2897 always treated as atomic groups in PCRE. This is like Python, but
2898 unlike Perl.
2899
2900 10. There are some differences that are concerned with the settings of
2901 captured strings when part of a pattern is repeated. For example,
2902 matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2
2903 unset, but in PCRE it is set to "b".
2904
2905 11. PCRE does support Perl 5.10's backtracking verbs (*ACCEPT),
2906 (*FAIL), (*F), (*COMMIT), (*PRUNE), (*SKIP), and (*THEN), but only in
2907 the forms without an argument. PCRE does not support (*MARK). If
2908 (*ACCEPT) is within capturing parentheses, PCRE does not set that cap-
2909 ture group; this is different to Perl.
2910
2911 12. PCRE provides some extensions to the Perl regular expression facil-
2912 ities. Perl 5.10 will include new features that are not in earlier
2913 versions, some of which (such as named parentheses) have been in PCRE
2914 for some time. This list is with respect to Perl 5.10:
2915
2916 (a) Although lookbehind assertions must match fixed length strings,
2917 each alternative branch of a lookbehind assertion can match a different
2918 length of string. Perl requires them all to have the same length.
2919
2920 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $
2921 meta-character matches only at the very end of the string.
2922
2923 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe-
2924 cial meaning is faulted. Otherwise, like Perl, the backslash is quietly
2925 ignored. (Perl can be made to issue a warning.)
2926
2927 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti-
2928 fiers is inverted, that is, by default they are not greedy, but if fol-
2929 lowed by a question mark they are.
2930
2931 (e) PCRE_ANCHORED can be used at matching time to force a pattern to be
2932 tried only at the first matching position in the subject string.
2933
2934 (f) The PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, and PCRE_NO_AUTO_CAP-
2935 TURE options for pcre_exec() have no Perl equivalents.
2936
2937 (g) The \R escape sequence can be restricted to match only CR, LF, or
2938 CRLF by the PCRE_BSR_ANYCRLF option.
2939
2940 (h) The callout facility is PCRE-specific.
2941
2942 (i) The partial matching facility is PCRE-specific.
2943
2944 (j) Patterns compiled by PCRE can be saved and re-used at a later time,
2945 even on different hosts that have the other endianness.
2946
2947 (k) The alternative matching function (pcre_dfa_exec()) matches in a
2948 different way and is not Perl-compatible.
2949
2950 (l) PCRE recognizes some special sequences such as (*CR) at the start
2951 of a pattern that set overall options that cannot be changed within the
2952 pattern.
2953
2954
2955 AUTHOR
2956
2957 Philip Hazel
2958 University Computing Service
2959 Cambridge CB2 3QH, England.
2960
2961
2962 REVISION
2963
2964 Last updated: 25 August 2009
2965 Copyright (c) 1997-2009 University of Cambridge.
2966 ------------------------------------------------------------------------------
2967
2968
2969 PCREPATTERN(3) PCREPATTERN(3)
2970
2971
2972 NAME
2973 PCRE - Perl-compatible regular expressions
2974
2975
2976 PCRE REGULAR EXPRESSION DETAILS
2977
2978 The syntax and semantics of the regular expressions that are supported
2979 by PCRE are described in detail below. There is a quick-reference syn-
2980 tax summary in the pcresyntax page. PCRE tries to match Perl syntax and
2981 semantics as closely as it can. PCRE also supports some alternative
2982 regular expression syntax (which does not conflict with the Perl syn-
2983 tax) in order to provide some compatibility with regular expressions in
2984 Python, .NET, and Oniguruma.
2985
2986 Perl's regular expressions are described in its own documentation, and
2987 regular expressions in general are covered in a number of books, some
2988 of which have copious examples. Jeffrey Friedl's "Mastering Regular
2989 Expressions", published by O'Reilly, covers regular expressions in
2990 great detail. This description of PCRE's regular expressions is
2991 intended as reference material.
2992
2993 The original operation of PCRE was on strings of one-byte characters.
2994 However, there is now also support for UTF-8 character strings. To use
2995 this, you must build PCRE to include UTF-8 support, and then call
2996 pcre_compile() with the PCRE_UTF8 option. There is also a special
2997 sequence that can be given at the start of a pattern:
2998
2999 (*UTF8)
3000
3001 Starting a pattern with this sequence is equivalent to setting the
3002 PCRE_UTF8 option. This feature is not Perl-compatible. How setting
3003 UTF-8 mode affects pattern matching is mentioned in several places
3004 below. There is also a summary of UTF-8 features in the section on
3005 UTF-8 support in the main pcre page.
3006
3007 The remainder of this document discusses the patterns that are sup-
3008 ported by PCRE when its main matching function, pcre_exec(), is used.
3009 From release 6.0, PCRE offers a second matching function,
3010 pcre_dfa_exec(), which matches using a different algorithm that is not
3011 Perl-compatible. Some of the features discussed below are not available
3012 when pcre_dfa_exec() is used. The advantages and disadvantages of the
3013 alternative function, and how it differs from the normal function, are
3014 discussed in the pcrematching page.
3015
3016
3017 NEWLINE CONVENTIONS
3018
3019 PCRE supports five different conventions for indicating line breaks in
3020 strings: a single CR (carriage return) character, a single LF (line-
3021 feed) character, the two-character sequence CRLF, any of the three pre-
3022 ceding, or any Unicode newline sequence. The pcreapi page has further
3023 discussion about newlines, and shows how to set the newline convention
3024 in the options arguments for the compiling and matching functions.
3025
3026 It is also possible to specify a newline convention by starting a pat-
3027 tern string with one of the following five sequences:
3028
3029 (*CR) carriage return
3030 (*LF) linefeed
3031 (*CRLF) carriage return, followed by linefeed
3032 (*ANYCRLF) any of the three above
3033 (*ANY) all Unicode newline sequences
3034
3035 These override the default and the options given to pcre_compile(). For
3036 example, on a Unix system where LF is the default newline sequence, the
3037 pattern
3038
3039 (*CR)a.b
3040
3041 changes the convention to CR. That pattern matches "a\nb" because LF is
3042 no longer a newline. Note that these special settings, which are not
3043 Perl-compatible, are recognized only at the very start of a pattern,
3044 and that they must be in upper case. If more than one of them is
3045 present, the last one is used.
3046
3047 The newline convention does not affect what the \R escape sequence
3048 matches. By default, this is any Unicode newline sequence, for Perl
3049 compatibility. However, this can be changed; see the description of \R
3050 in the section entitled "Newline sequences" below. A change of \R set-
3051 ting can be combined with a change of newline convention.
3052
3053
3054 CHARACTERS AND METACHARACTERS
3055
3056 A regular expression is a pattern that is matched against a subject
3057 string from left to right. Most characters stand for themselves in a
3058 pattern, and match the corresponding characters in the subject. As a
3059 trivial example, the pattern
3060
3061 The quick brown fox
3062
3063 matches a portion of a subject string that is identical to itself. When
3064 caseless matching is specified (the PCRE_CASELESS option), letters are
3065 matched independently of case. In UTF-8 mode, PCRE always understands
3066 the concept of case for characters whose values are less than 128, so
3067 caseless matching is always possible. For characters with higher val-
3068 ues, the concept of case is supported if PCRE is compiled with Unicode
3069 property support, but not otherwise. If you want to use caseless
3070 matching for characters 128 and above, you must ensure that PCRE is
3071 compiled with Unicode property support as well as with UTF-8 support.
3072
3073 The power of regular expressions comes from the ability to include
3074 alternatives and repetitions in the pattern. These are encoded in the
3075 pattern by the use of metacharacters, which do not stand for themselves
3076 but instead are interpreted in some special way.
3077
3078 There are two different sets of metacharacters: those that are recog-
3079 nized anywhere in the pattern except within square brackets, and those
3080 that are recognized within square brackets. Outside square brackets,
3081 the metacharacters are as follows:
3082
3083 \ general escape character with several uses
3084 ^ assert start of string (or line, in multiline mode)
3085 $ assert end of string (or line, in multiline mode)
3086 . match any character except newline (by default)
3087 [ start character class definition
3088 | start of alternative branch
3089 ( start subpattern
3090 ) end subpattern
3091 ? extends the meaning of (
3092 also 0 or 1 quantifier
3093 also quantifier minimizer
3094 * 0 or more quantifier
3095 + 1 or more quantifier
3096 also "possessive quantifier"
3097 { start min/max quantifier
3098
3099 Part of a pattern that is in square brackets is called a "character
3100 class". In a character class the only metacharacters are:
3101
3102 \ general escape character
3103 ^ negate the class, but only if the first character
3104 - indicates character range
3105 [ POSIX character class (only if followed by POSIX
3106 syntax)
3107 ] terminates the character class
3108
3109 The following sections describe the use of each of the metacharacters.
3110
3111
3112 BACKSLASH
3113
3114 The backslash character has several uses. Firstly, if it is followed by
3115 a non-alphanumeric character, it takes away any special meaning that
3116 character may have. This use of backslash as an escape character
3117 applies both inside and outside character classes.
3118
3119 For example, if you want to match a * character, you write \* in the
3120 pattern. This escaping action applies whether or not the following
3121 character would otherwise be interpreted as a metacharacter, so it is
3122 always safe to precede a non-alphanumeric with backslash to specify
3123 that it stands for itself. In particular, if you want to match a back-
3124 slash, you write \\.
3125
3126 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in
3127 the pattern (other than in a character class) and characters between a
3128 # outside a character class and the next newline are ignored. An escap-
3129 ing backslash can be used to include a whitespace or # character as
3130 part of the pattern.
3131
3132 If you want to remove the special meaning from a sequence of charac-
3133 ters, you can do so by putting them between \Q and \E. This is differ-
3134 ent from Perl in that $ and @ are handled as literals in \Q...\E
3135 sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
3136 tion. Note the following examples:
3137
3138 Pattern PCRE matches Perl matches
3139
3140 \Qabc$xyz\E abc$xyz abc followed by the
3141 contents of $xyz
3142 \Qabc\$xyz\E abc\$xyz abc\$xyz
3143 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
3144
3145 The \Q...\E sequence is recognized both inside and outside character
3146 classes.
3147
3148 Non-printing characters
3149
3150 A second use of backslash provides a way of encoding non-printing char-
3151 acters in patterns in a visible manner. There is no restriction on the
3152 appearance of non-printing characters, apart from the binary zero that
3153 terminates a pattern, but when a pattern is being prepared by text
3154 editing, it is usually easier to use one of the following escape
3155 sequences than the binary character it represents:
3156
3157 \a alarm, that is, the BEL character (hex 07)
3158 \cx "control-x", where x is any character
3159 \e escape (hex 1B)
3160 \f formfeed (hex 0C)
3161 \n linefeed (hex 0A)
3162 \r carriage return (hex 0D)
3163 \t tab (hex 09)
3164 \ddd character with octal code ddd, or backreference
3165 \xhh character with hex code hh
3166 \x{hhh..} character with hex code hhh..
3167
3168 The precise effect of \cx is as follows: if x is a lower case letter,
3169 it is converted to upper case. Then bit 6 of the character (hex 40) is
3170 inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c;
3171 becomes hex 7B.
3172
3173 After \x, from zero to two hexadecimal digits are read (letters can be
3174 in upper or lower case). Any number of hexadecimal digits may appear
3175 between \x{ and }, but the value of the character code must be less
3176 than 256 in non-UTF-8 mode, and less than 2**31 in UTF-8 mode. That is,
3177 the maximum value in hexadecimal is 7FFFFFFF. Note that this is bigger
3178 than the largest Unicode code point, which is 10FFFF.
3179
3180 If characters other than hexadecimal digits appear between \x{ and },
3181 or if there is no terminating }, this form of escape is not recognized.
3182 Instead, the initial \x will be interpreted as a basic hexadecimal
3183 escape, with no following digits, giving a character whose value is
3184 zero.
3185
3186 Characters whose value is less than 256 can be defined by either of the
3187 two syntaxes for \x. There is no difference in the way they are han-
3188 dled. For example, \xdc is exactly the same as \x{dc}.
3189
3190 After \0 up to two further octal digits are read. If there are fewer
3191 than two digits, just those that are present are used. Thus the
3192 sequence \0\x\07 specifies two binary zeros followed by a BEL character
3193 (code value 7). Make sure you supply two digits after the initial zero
3194 if the pattern character that follows is itself an octal digit.
3195
3196 The handling of a backslash followed by a digit other than 0 is compli-
3197 cated. Outside a character class, PCRE reads it and any following dig-
3198 its as a decimal number. If the number is less than 10, or if there
3199 have been at least that many previous capturing left parentheses in the
3200 expression, the entire sequence is taken as a back reference. A
3201 description of how this works is given later, following the discussion
3202 of parenthesized subpatterns.
3203
3204 Inside a character class, or if the decimal number is greater than 9
3205 and there have not been that many capturing subpatterns, PCRE re-reads
3206 up to three octal digits following the backslash, and uses them to gen-
3207 erate a data character. Any subsequent digits stand for themselves. In
3208 non-UTF-8 mode, the value of a character specified in octal must be
3209 less than \400. In UTF-8 mode, values up to \777 are permitted. For
3210 example:
3211
3212 \040 is another way of writing a space
3213 \40 is the same, provided there are fewer than 40
3214 previous capturing subpatterns
3215 \7 is always a back reference
3216 \11 might be a back reference, or another way of
3217 writing a tab
3218 \011 is always a tab
3219 \0113 is a tab followed by the character "3"
3220 \113 might be a back reference, otherwise the
3221 character with octal code 113
3222 \377 might be a back reference, otherwise
3223 the byte consisting entirely of 1 bits
3224 \81 is either a back reference, or a binary zero
3225 followed by the two characters "8" and "1"
3226
3227 Note that octal values of 100 or greater must not be introduced by a
3228 leading zero, because no more than three octal digits are ever read.
3229
3230 All the sequences that define a single character value can be used both
3231 inside and outside character classes. In addition, inside a character
3232 class, the sequence \b is interpreted as the backspace character (hex
3233 08), and the sequences \R and \X are interpreted as the characters "R"
3234 and "X", respectively. Outside a character class, these sequences have
3235 different meanings (see below).
3236
3237 Absolute and relative back references
3238
3239 The sequence \g followed by an unsigned or a negative number, option-
3240 ally enclosed in braces, is an absolute or relative back reference. A
3241 named back reference can be coded as \g{name}. Back references are dis-
3242 cussed later, following the discussion of parenthesized subpatterns.
3243
3244 Absolute and relative subroutine calls
3245
3246 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
3247 name or a number enclosed either in angle brackets or single quotes, is
3248 an alternative syntax for referencing a subpattern as a "subroutine".
3249 Details are discussed later. Note that \g{...} (Perl syntax) and
3250 \g<...> (Oniguruma syntax) are not synonymous. The former is a back
3251 reference; the latter is a subroutine call.
3252
3253 Generic character types
3254
3255 Another use of backslash is for specifying generic character types. The
3256 following are always recognized:
3257
3258 \d any decimal digit
3259 \D any character that is not a decimal digit
3260 \h any horizontal whitespace character
3261 \H any character that is not a horizontal whitespace character
3262 \s any whitespace character
3263 \S any character that is not a whitespace character
3264 \v any vertical whitespace character
3265 \V any character that is not a vertical whitespace character
3266 \w any "word" character
3267 \W any "non-word" character
3268
3269 Each pair of escape sequences partitions the complete set of characters
3270 into two disjoint sets. Any given character matches one, and only one,
3271 of each pair.
3272
3273 These character type sequences can appear both inside and outside char-
3274 acter classes. They each match one character of the appropriate type.
3275 If the current matching point is at the end of the subject string, all
3276 of them fail, since there is no character to match.
3277
3278 For compatibility with Perl, \s does not match the VT character (code
3279 11). This makes it different from the the POSIX "space" class. The \s
3280 characters are HT (9), LF (10), FF (12), CR (13), and space (32). If
3281 "use locale;" is included in a Perl script, \s may match the VT charac-
3282 ter. In PCRE, it never does.
3283
3284 In UTF-8 mode, characters with values greater than 128 never match \d,
3285 \s, or \w, and always match \D, \S, and \W. This is true even when Uni-
3286 code character property support is available. These sequences retain
3287 their original meanings from before UTF-8 support was available, mainly
3288 for efficiency reasons. Note that this also affects \b, because it is
3289 defined in terms of \w and \W.
3290
3291 The sequences \h, \H, \v, and \V are Perl 5.10 features. In contrast to
3292 the other sequences, these do match certain high-valued codepoints in
3293 UTF-8 mode. The horizontal space characters are:
3294
3295 U+0009 Horizontal tab
3296 U+0020 Space
3297 U+00A0 Non-break space
3298 U+1680 Ogham space mark
3299 U+180E Mongolian vowel separator
3300 U+2000 En quad
3301 U+2001 Em quad
3302 U+2002 En space
3303 U+2003 Em space
3304 U+2004 Three-per-em space
3305 U+2005 Four-per-em space
3306 U+2006 Six-per-em space
3307 U+2007 Figure space
3308 U+2008 Punctuation space
3309 U+2009 Thin space
3310 U+200A Hair space
3311 U+202F Narrow no-break space
3312 U+205F Medium mathematical space
3313 U+3000 Ideographic space
3314
3315 The vertical space characters are:
3316
3317 U+000A Linefeed
3318 U+000B Vertical tab
3319 U+000C Formfeed
3320 U+000D Carriage return
3321 U+0085 Next line
3322 U+2028 Line separator
3323 U+2029 Paragraph separator
3324
3325 A "word" character is an underscore or any character less than 256 that
3326 is a letter or digit. The definition of letters and digits is con-
3327 trolled by PCRE's low-valued character tables, and may vary if locale-
3328 specific matching is taking place (see "Locale support" in the pcreapi
3329 page). For example, in a French locale such as "fr_FR" in Unix-like
3330 systems, or "french" in Windows, some character codes greater than 128
3331 are used for accented letters, and these are matched by \w. The use of
3332 locales with Unicode is discouraged.
3333
3334 Newline sequences
3335
3336 Outside a character class, by default, the escape sequence \R matches
3337 any Unicode newline sequence. This is a Perl 5.10 feature. In non-UTF-8
3338 mode \R is equivalent to the following:
3339
3340 (?>\r\n|\n|\x0b|\f|\r|\x85)
3341
3342 This is an example of an "atomic group", details of which are given
3343 below. This particular group matches either the two-character sequence
3344 CR followed by LF, or one of the single characters LF (linefeed,
3345 U+000A), VT (vertical tab, U+000B), FF (formfeed, U+000C), CR (carriage
3346 return, U+000D), or NEL (next line, U+0085). The two-character sequence
3347 is treated as a single unit that cannot be split.
3348
3349 In UTF-8 mode, two additional characters whose codepoints are greater
3350 than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa-
3351 rator, U+2029). Unicode character property support is not needed for
3352 these characters to be recognized.
3353
3354 It is possible to restrict \R to match only CR, LF, or CRLF (instead of
3355 the complete set of Unicode line endings) by setting the option
3356 PCRE_BSR_ANYCRLF either at compile time or when the pattern is matched.
3357 (BSR is an abbrevation for "backslash R".) This can be made the default
3358 when PCRE is built; if this is the case, the other behaviour can be
3359 requested via the PCRE_BSR_UNICODE option. It is also possible to
3360 specify these settings by starting a pattern string with one of the
3361 following sequences:
3362
3363 (*BSR_ANYCRLF) CR, LF, or CRLF only
3364 (*BSR_UNICODE) any Unicode newline sequence
3365
3366 These override the default and the options given to pcre_compile(), but
3367 they can be overridden by options given to pcre_exec(). Note that these
3368 special settings, which are not Perl-compatible, are recognized only at
3369 the very start of a pattern, and that they must be in upper case. If
3370 more than one of them is present, the last one is used. They can be
3371 combined with a change of newline convention, for example, a pattern
3372 can start with:
3373
3374 (*ANY)(*BSR_ANYCRLF)
3375
3376 Inside a character class, \R matches the letter "R".
3377
3378 Unicode character properties
3379
3380 When PCRE is built with Unicode character property support, three addi-
3381 tional escape sequences that match characters with specific properties
3382 are available. When not in UTF-8 mode, these sequences are of course
3383 limited to testing characters whose codepoints are less than 256, but
3384 they do work in this mode. The extra escape sequences are:
3385
3386 \p{xx} a character with the xx property
3387 \P{xx} a character without the xx property
3388 \X an extended Unicode sequence
3389
3390 The property names represented by xx above are limited to the Unicode
3391 script names, the general category properties, and "Any", which matches
3392 any character (including newline). Other properties such as "InMusical-
3393 Symbols" are not currently supported by PCRE. Note that \P{Any} does
3394 not match any characters, so always causes a match failure.
3395
3396 Sets of Unicode characters are defined as belonging to certain scripts.
3397 A character from one of these sets can be matched using a script name.
3398 For example:
3399
3400 \p{Greek}
3401 \P{Han}
3402
3403 Those that are not part of an identified script are lumped together as
3404 "Common". The current list of scripts is:
3405
3406 Arabic, Armenian, Balinese, Bengali, Bopomofo, Braille, Buginese,
3407 Buhid, Canadian_Aboriginal, Cherokee, Common, Coptic, Cuneiform,
3408 Cypriot, Cyrillic, Deseret, Devanagari, Ethiopic, Georgian, Glagolitic,
3409 Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hira-
3410 gana, Inherited, Kannada, Katakana, Kharoshthi, Khmer, Lao, Latin,
3411 Limbu, Linear_B, Malayalam, Mongolian, Myanmar, New_Tai_Lue, Nko,
3412 Ogham, Old_Italic, Old_Persian, Oriya, Osmanya, Phags_Pa, Phoenician,
3413 Runic, Shavian, Sinhala, Syloti_Nagri, Syriac, Tagalog, Tagbanwa,
3414 Tai_Le, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, Ugaritic, Yi.
3415
3416 Each character has exactly one general category property, specified by
3417 a two-letter abbreviation. For compatibility with Perl, negation can be
3418 specified by including a circumflex between the opening brace and the
3419 property name. For example, \p{^Lu} is the same as \P{Lu}.
3420
3421 If only one letter is specified with \p or \P, it includes all the gen-
3422 eral category properties that start with that letter. In this case, in
3423 the absence of negation, the curly brackets in the escape sequence are
3424 optional; these two examples have the same effect:
3425
3426 \p{L}
3427 \pL
3428
3429 The following general category property codes are supported:
3430
3431 C Other
3432 Cc Control
3433 Cf Format
3434 Cn Unassigned
3435 Co Private use
3436 Cs Surrogate
3437
3438 L Letter
3439 Ll Lower case letter
3440 Lm Modifier letter
3441 Lo Other letter
3442 Lt Title case letter
3443 Lu Upper case letter
3444
3445 M Mark
3446 Mc Spacing mark
3447 Me Enclosing mark
3448 Mn Non-spacing mark
3449
3450 N Number
3451 Nd Decimal number
3452 Nl Letter number
3453 No Other number
3454
3455 P Punctuation
3456 Pc Connector punctuation
3457 Pd Dash punctuation
3458 Pe Close punctuation
3459 Pf Final punctuation
3460 Pi Initial punctuation
3461 Po Other punctuation
3462 Ps Open punctuation
3463
3464 S Symbol
3465 Sc Currency symbol
3466 Sk Modifier symbol
3467 Sm Mathematical symbol
3468 So Other symbol
3469
3470 Z Separator
3471 Zl Line separator
3472 Zp Paragraph separator
3473 Zs Space separator
3474
3475 The special property L& is also supported: it matches a character that
3476 has the Lu, Ll, or Lt property, in other words, a letter that is not
3477 classified as a modifier or "other".
3478
3479 The Cs (Surrogate) property applies only to characters in the range
3480 U+D800 to U+DFFF. Such characters are not valid in UTF-8 strings (see
3481 RFC 3629) and so cannot be tested by PCRE, unless UTF-8 validity check-
3482 ing has been turned off (see the discussion of PCRE_NO_UTF8_CHECK in
3483 the pcreapi page).
3484
3485 The long synonyms for these properties that Perl supports (such as
3486 \p{Letter}) are not supported by PCRE, nor is it permitted to prefix
3487 any of these properties with "Is".
3488
3489 No character that is in the Unicode table has the Cn (unassigned) prop-
3490 erty. Instead, this property is assumed for any code point that is not
3491 in the Unicode table.
3492
3493 Specifying caseless matching does not affect these escape sequences.
3494 For example, \p{Lu} always matches only upper case letters.
3495
3496 The \X escape matches any number of Unicode characters that form an
3497 extended Unicode sequence. \X is equivalent to
3498
3499 (?>\PM\pM*)
3500
3501 That is, it matches a character without the "mark" property, followed
3502 by zero or more characters with the "mark" property, and treats the
3503 sequence as an atomic group (see below). Characters with the "mark"
3504 property are typically accents that affect the preceding character.
3505 None of them have codepoints less than 256, so in non-UTF-8 mode \X
3506 matches any one character.
3507
3508 Matching characters by Unicode property is not fast, because PCRE has
3509 to search a structure that contains data for over fifteen thousand
3510 characters. That is why the traditional escape sequences such as \d and
3511 \w do not use Unicode properties in PCRE.
3512
3513 Resetting the match start
3514
3515 The escape sequence \K, which is a Perl 5.10 feature, causes any previ-
3516 ously matched characters not to be included in the final matched
3517 sequence. For example, the pattern:
3518
3519 foo\Kbar
3520
3521 matches "foobar", but reports that it has matched "bar". This feature
3522 is similar to a lookbehind assertion (described below). However, in
3523 this case, the part of the subject before the real match does not have
3524 to be of fixed length, as lookbehind assertions do. The use of \K does
3525 not interfere with the setting of captured substrings. For example,
3526 when the pattern
3527
3528 (foo)\Kbar
3529
3530 matches "foobar", the first substring is still set to "foo".
3531
3532 Simple assertions
3533
3534 The final use of backslash is for certain simple assertions. An asser-
3535 tion specifies a condition that has to be met at a particular point in
3536 a match, without consuming any characters from the subject string. The
3537 use of subpatterns for more complicated assertions is described below.
3538 The backslashed assertions are:
3539
3540 \b matches at a word boundary
3541 \B matches when not at a word boundary
3542 \A matches at the start of the subject
3543 \Z matches at the end of the subject
3544 also matches before a newline at the end of the subject
3545 \z matches only at the end of the subject
3546 \G matches at the first matching position in the subject
3547
3548 These assertions may not appear in character classes (but note that \b
3549 has a different meaning, namely the backspace character, inside a char-
3550 acter class).
3551
3552 A word boundary is a position in the subject string where the current
3553 character and the previous character do not both match \w or \W (i.e.
3554 one matches \w and the other matches \W), or the start or end of the
3555 string if the first or last character matches \w, respectively.
3556
3557 The \A, \Z, and \z assertions differ from the traditional circumflex
3558 and dollar (described in the next section) in that they only ever match
3559 at the very start and end of the subject string, whatever options are
3560 set. Thus, they are independent of multiline mode. These three asser-
3561 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
3562 affect only the behaviour of the circumflex and dollar metacharacters.
3563 However, if the startoffset argument of pcre_exec() is non-zero, indi-
3564 cating that matching is to start at a point other than the beginning of
3565 the subject, \A can never match. The difference between \Z and \z is
3566 that \Z matches before a newline at the end of the string as well as at
3567 the very end, whereas \z matches only at the end.
3568
3569 The \G assertion is true only when the current matching position is at
3570 the start point of the match, as specified by the startoffset argument
3571 of pcre_exec(). It differs from \A when the value of startoffset is
3572 non-zero. By calling pcre_exec() multiple times with appropriate argu-
3573 ments, you can mimic Perl's /g option, and it is in this kind of imple-
3574 mentation where \G can be useful.
3575
3576 Note, however, that PCRE's interpretation of \G, as the start of the
3577 current match, is subtly different from Perl's, which defines it as the
3578 end of the previous match. In Perl, these can be different when the
3579 previously matched string was empty. Because PCRE does just one match
3580 at a time, it cannot reproduce this behaviour.
3581
3582 If all the alternatives of a pattern begin with \G, the expression is
3583 anchored to the starting match position, and the "anchored" flag is set
3584 in the compiled regular expression.
3585
3586
3587 CIRCUMFLEX AND DOLLAR
3588
3589 Outside a character class, in the default matching mode, the circumflex
3590 character is an assertion that is true only if the current matching
3591 point is at the start of the subject string. If the startoffset argu-
3592 ment of pcre_exec() is non-zero, circumflex can never match if the
3593 PCRE_MULTILINE option is unset. Inside a character class, circumflex
3594 has an entirely different meaning (see below).
3595
3596 Circumflex need not be the first character of the pattern if a number
3597 of alternatives are involved, but it should be the first thing in each
3598 alternative in which it appears if the pattern is ever to match that
3599 branch. If all possible alternatives start with a circumflex, that is,
3600 if the pattern is constrained to match only at the start of the sub-
3601 ject, it is said to be an "anchored" pattern. (There are also other
3602 constructs that can cause a pattern to be anchored.)
3603
3604 A dollar character is an assertion that is true only if the current
3605 matching point is at the end of the subject string, or immediately
3606 before a newline at the end of the string (by default). Dollar need not
3607 be the last character of the pattern if a number of alternatives are
3608 involved, but it should be the last item in any branch in which it
3609 appears. Dollar has no special meaning in a character class.
3610
3611 The meaning of dollar can be changed so that it matches only at the
3612 very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
3613 compile time. This does not affect the \Z assertion.
3614
3615 The meanings of the circumflex and dollar characters are changed if the
3616 PCRE_MULTILINE option is set. When this is the case, a circumflex
3617 matches immediately after internal newlines as well as at the start of
3618 the subject string. It does not match after a newline that ends the
3619 string. A dollar matches before any newlines in the string, as well as
3620 at the very end, when PCRE_MULTILINE is set. When newline is specified
3621 as the two-character sequence CRLF, isolated CR and LF characters do
3622 not indicate newlines.
3623
3624 For example, the pattern /^abc$/ matches the subject string "def\nabc"
3625 (where \n represents a newline) in multiline mode, but not otherwise.
3626 Consequently, patterns that are anchored in single line mode because
3627 all branches start with ^ are not anchored in multiline mode, and a
3628 match for circumflex is possible when the startoffset argument of
3629 pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
3630 PCRE_MULTILINE is set.
3631
3632 Note that the sequences \A, \Z, and \z can be used to match the start
3633 and end of the subject in both modes, and if all branches of a pattern
3634 start with \A it is always anchored, whether or not PCRE_MULTILINE is
3635 set.
3636
3637
3638 FULL STOP (PERIOD, DOT)
3639
3640 Outside a character class, a dot in the pattern matches any one charac-
3641 ter in the subject string except (by default) a character that signi-
3642 fies the end of a line. In UTF-8 mode, the matched character may be
3643 more than one byte long.
3644
3645 When a line ending is defined as a single character, dot never matches
3646 that character; when the two-character sequence CRLF is used, dot does
3647 not match CR if it is immediately followed by LF, but otherwise it
3648 matches all characters (including isolated CRs and LFs). When any Uni-
3649 code line endings are being recognized, dot does not match CR or LF or
3650 any of the other line ending characters.
3651
3652 The behaviour of dot with regard to newlines can be changed. If the
3653 PCRE_DOTALL option is set, a dot matches any one character, without
3654 exception. If the two-character sequence CRLF is present in the subject
3655 string, it takes two dots to match it.
3656
3657 The handling of dot is entirely independent of the handling of circum-
3658 flex and dollar, the only relationship being that they both involve
3659 newlines. Dot has no special meaning in a character class.
3660
3661
3662 MATCHING A SINGLE BYTE
3663
3664 Outside a character class, the escape sequence \C matches any one byte,
3665 both in and out of UTF-8 mode. Unlike a dot, it always matches any
3666 line-ending characters. The feature is provided in Perl in order to
3667 match individual bytes in UTF-8 mode. Because it breaks up UTF-8 char-
3668 acters into individual bytes, what remains in the string may be a mal-
3669 formed UTF-8 string. For this reason, the \C escape sequence is best
3670 avoided.
3671
3672 PCRE does not allow \C to appear in lookbehind assertions (described
3673 below), because in UTF-8 mode this would make it impossible to calcu-
3674 late the length of the lookbehind.
3675
3676
3677 SQUARE BRACKETS AND CHARACTER CLASSES
3678
3679 An opening square bracket introduces a character class, terminated by a
3680 closing square bracket. A closing square bracket on its own is not spe-
3681 cial. If a closing square bracket is required as a member of the class,
3682 it should be the first data character in the class (after an initial
3683 circumflex, if present) or escaped with a backslash.
3684
3685 A character class matches a single character in the subject. In UTF-8
3686 mode, the character may occupy more than one byte. A matched character
3687 must be in the set of characters defined by the class, unless the first
3688 character in the class definition is a circumflex, in which case the
3689 subject character must not be in the set defined by the class. If a
3690 circumflex is actually required as a member of the class, ensure it is
3691 not the first character, or escape it with a backslash.
3692
3693 For example, the character class [aeiou] matches any lower case vowel,
3694 while [^aeiou] matches any character that is not a lower case vowel.
3695 Note that a circumflex is just a convenient notation for specifying the
3696 characters that are in the class by enumerating those that are not. A
3697 class that starts with a circumflex is not an assertion: it still con-
3698 sumes a character from the subject string, and therefore it fails if
3699 the current pointer is at the end of the string.
3700
3701 In UTF-8 mode, characters with values greater than 255 can be included
3702 in a class as a literal string of bytes, or by using the \x{ escaping
3703 mechanism.
3704
3705 When caseless matching is set, any letters in a class represent both
3706 their upper case and lower case versions, so for example, a caseless
3707 [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
3708 match "A", whereas a caseful version would. In UTF-8 mode, PCRE always
3709 understands the concept of case for characters whose values are less
3710 than 128, so caseless matching is always possible. For characters with
3711 higher values, the concept of case is supported if PCRE is compiled
3712 with Unicode property support, but not otherwise. If you want to use
3713 caseless matching for characters 128 and above, you must ensure that
3714 PCRE is compiled with Unicode property support as well as with UTF-8
3715 support.
3716
3717 Characters that might indicate line breaks are never treated in any
3718 special way when matching character classes, whatever line-ending
3719 sequence is in use, and whatever setting of the PCRE_DOTALL and
3720 PCRE_MULTILINE options is used. A class such as [^a] always matches one
3721 of these characters.
3722
3723 The minus (hyphen) character can be used to specify a range of charac-
3724 ters in a character class. For example, [d-m] matches any letter
3725 between d and m, inclusive. If a minus character is required in a
3726 class, it must be escaped with a backslash or appear in a position
3727 where it cannot be interpreted as indicating a range, typically as the
3728 first or last character in the class.
3729
3730 It is not possible to have the literal character "]" as the end charac-
3731 ter of a range. A pattern such as [W-]46] is interpreted as a class of
3732 two characters ("W" and "-") followed by a literal string "46]", so it
3733 would match "W46]" or "-46]". However, if the "]" is escaped with a
3734 backslash it is interpreted as the end of range, so [W-\]46] is inter-
3735 preted as a class containing a range followed by two other characters.
3736 The octal or hexadecimal representation of "]" can also be used to end
3737 a range.
3738
3739 Ranges operate in the collating sequence of character values. They can
3740 also be used for characters specified numerically, for example
3741 [\000-\037]. In UTF-8 mode, ranges can include characters whose values
3742 are greater than 255, for example [\x{100}-\x{2ff}].
3743
3744 If a range that includes letters is used when caseless matching is set,
3745 it matches the letters in either case. For example, [W-c] is equivalent
3746 to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if
3747 character tables for a French locale are in use, [\xc8-\xcb] matches
3748 accented E characters in both cases. In UTF-8 mode, PCRE supports the
3749 concept of case for characters with values greater than 128 only when
3750 it is compiled with Unicode property support.
3751
3752 The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear
3753 in a character class, and add the characters that they match to the
3754 class. For example, [\dABCDEF] matches any hexadecimal digit. A circum-
3755 flex can conveniently be used with the upper case character types to
3756 specify a more restricted set of characters than the matching lower
3757 case type. For example, the class [^\W_] matches any letter or digit,
3758 but not underscore.
3759
3760 The only metacharacters that are recognized in character classes are
3761 backslash, hyphen (only where it can be interpreted as specifying a
3762 range), circumflex (only at the start), opening square bracket (only
3763 when it can be interpreted as introducing a POSIX class name - see the
3764 next section), and the terminating closing square bracket. However,
3765 escaping other non-alphanumeric characters does no harm.
3766
3767
3768 POSIX CHARACTER CLASSES
3769
3770 Perl supports the POSIX notation for character classes. This uses names
3771 enclosed by [: and :] within the enclosing square brackets. PCRE also
3772 supports this notation. For example,
3773
3774 [01[:alpha:]%]
3775
3776 matches "0", "1", any alphabetic character, or "%". The supported class
3777 names are
3778
3779 alnum letters and digits
3780 alpha letters
3781 ascii character codes 0 - 127
3782 blank space or tab only
3783 cntrl control characters
3784 digit decimal digits (same as \d)
3785 graph printing characters, excluding space
3786 lower lower case letters
3787 print printing characters, including space
3788 punct printing characters, excluding letters and digits
3789 space white space (not quite the same as \s)
3790 upper upper case letters
3791 word "word" characters (same as \w)
3792 xdigit hexadecimal digits
3793
3794 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
3795 and space (32). Notice that this list includes the VT character (code
3796 11). This makes "space" different to \s, which does not include VT (for
3797 Perl compatibility).
3798
3799 The name "word" is a Perl extension, and "blank" is a GNU extension
3800 from Perl 5.8. Another Perl extension is negation, which is indicated
3801 by a ^ character after the colon. For example,
3802
3803 [12[:^digit:]]
3804
3805 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
3806 POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
3807 these are not supported, and an error is given if they are encountered.
3808
3809 In UTF-8 mode, characters with values greater than 128 do not match any
3810 of the POSIX character classes.
3811
3812
3813 VERTICAL BAR
3814
3815 Vertical bar characters are used to separate alternative patterns. For
3816 example, the pattern
3817
3818 gilbert|sullivan
3819
3820 matches either "gilbert" or "sullivan". Any number of alternatives may
3821 appear, and an empty alternative is permitted (matching the empty
3822 string). The matching process tries each alternative in turn, from left
3823 to right, and the first one that succeeds is used. If the alternatives
3824 are within a subpattern (defined below), "succeeds" means matching the
3825 rest of the main pattern as well as the alternative in the subpattern.
3826
3827
3828 INTERNAL OPTION SETTING
3829
3830 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
3831 PCRE_EXTENDED options (which are Perl-compatible) can be changed from
3832 within the pattern by a sequence of Perl option letters enclosed
3833 between "(?" and ")". The option letters are
3834
3835 i for PCRE_CASELESS
3836 m for PCRE_MULTILINE
3837 s for PCRE_DOTALL
3838 x for PCRE_EXTENDED
3839
3840 For example, (?im) sets caseless, multiline matching. It is also possi-
3841 ble to unset these options by preceding the letter with a hyphen, and a
3842 combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
3843 LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED,
3844 is also permitted. If a letter appears both before and after the
3845 hyphen, the option is unset.
3846
3847 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA
3848 can be changed in the same way as the Perl-compatible options by using
3849 the characters J, U and X respectively.
3850
3851 When one of these option changes occurs at top level (that is, not
3852 inside subpattern parentheses), the change applies to the remainder of
3853 the pattern that follows. If the change is placed right at the start of
3854 a pattern, PCRE extracts it into the global options (and it will there-
3855 fore show up in data extracted by the pcre_fullinfo() function).
3856
3857 An option change within a subpattern (see below for a description of
3858 subpatterns) affects only that part of the current pattern that follows
3859 it, so
3860
3861 (a(?i)b)c
3862
3863 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
3864 used). By this means, options can be made to have different settings
3865 in different parts of the pattern. Any changes made in one alternative
3866 do carry on into subsequent branches within the same subpattern. For
3867 example,
3868
3869 (a(?i)b|c)
3870
3871 matches "ab", "aB", "c", and "C", even though when matching "C" the
3872 first branch is abandoned before the option setting. This is because
3873 the effects of option settings happen at compile time. There would be
3874 some very weird behaviour otherwise.
3875
3876 Note: There are other PCRE-specific options that can be set by the
3877 application when the compile or match functions are called. In some
3878 cases the pattern can contain special leading sequences such as (*CRLF)
3879 to override what the application has set or what has been defaulted.
3880 Details are given in the section entitled "Newline sequences" above.
3881 There is also the (*UTF8) leading sequence that can be used to set
3882 UTF-8 mode; this is equivalent to setting the PCRE_UTF8 option.
3883
3884
3885 SUBPATTERNS
3886
3887 Subpatterns are delimited by parentheses (round brackets), which can be
3888 nested. Turning part of a pattern into a subpattern does two things:
3889
3890 1. It localizes a set of alternatives. For example, the pattern
3891
3892 cat(aract|erpillar|)
3893
3894 matches one of the words "cat", "cataract", or "caterpillar". Without
3895 the parentheses, it would match "cataract", "erpillar" or an empty
3896 string.
3897
3898 2. It sets up the subpattern as a capturing subpattern. This means
3899 that, when the whole pattern matches, that portion of the subject
3900 string that matched the subpattern is passed back to the caller via the
3901 ovector argument of pcre_exec(). Opening parentheses are counted from
3902 left to right (starting from 1) to obtain numbers for the capturing
3903 subpatterns.
3904
3905 For example, if the string "the red king" is matched against the pat-
3906 tern
3907
3908 the ((red|white) (king|queen))
3909
3910 the captured substrings are "red king", "red", and "king", and are num-
3911 bered 1, 2, and 3, respectively.
3912
3913 The fact that plain parentheses fulfil two functions is not always
3914 helpful. There are often times when a grouping subpattern is required
3915 without a capturing requirement. If an opening parenthesis is followed
3916 by a question mark and a colon, the subpattern does not do any captur-
3917 ing, and is not counted when computing the number of any subsequent
3918 capturing subpatterns. For example, if the string "the white queen" is
3919 matched against the pattern
3920
3921 the ((?:red|white) (king|queen))
3922
3923 the captured substrings are "white queen" and "queen", and are numbered
3924 1 and 2. The maximum number of capturing subpatterns is 65535.
3925
3926 As a convenient shorthand, if any option settings are required at the
3927 start of a non-capturing subpattern, the option letters may appear
3928 between the "?" and the ":". Thus the two patterns
3929
3930 (?i:saturday|sunday)
3931 (?:(?i)saturday|sunday)
3932
3933 match exactly the same set of strings. Because alternative branches are
3934 tried from left to right, and options are not reset until the end of
3935 the subpattern is reached, an option setting in one branch does affect
3936 subsequent branches, so the above patterns match "SUNDAY" as well as
3937 "Saturday".
3938
3939
3940 DUPLICATE SUBPATTERN NUMBERS
3941
3942 Perl 5.10 introduced a feature whereby each alternative in a subpattern
3943 uses the same numbers for its capturing parentheses. Such a subpattern
3944 starts with (?| and is itself a non-capturing subpattern. For example,
3945 consider this pattern:
3946
3947 (?|(Sat)ur|(Sun))day
3948
3949 Because the two alternatives are inside a (?| group, both sets of cap-
3950 turing parentheses are numbered one. Thus, when the pattern matches,
3951 you can look at captured substring number one, whichever alternative
3952 matched. This construct is useful when you want to capture part, but
3953 not all, of one of a number of alternatives. Inside a (?| group, paren-
3954 theses are numbered as usual, but the number is reset at the start of
3955 each branch. The numbers of any capturing buffers that follow the sub-
3956 pattern start after the highest number used in any branch. The follow-
3957 ing example is taken from the Perl documentation. The numbers under-
3958 neath show in which buffer the captured content will be stored.
3959
3960 # before ---------------branch-reset----------- after
3961 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
3962 # 1 2 2 3 2 3 4
3963
3964 A backreference or a recursive call to a numbered subpattern always
3965 refers to the first one in the pattern with the given number.
3966
3967 An alternative approach to using this "branch reset" feature is to use
3968 duplicate named subpatterns, as described in the next section.
3969
3970
3971 NAMED SUBPATTERNS
3972
3973 Identifying capturing parentheses by number is simple, but it can be
3974 very hard to keep track of the numbers in complicated regular expres-
3975 sions. Furthermore, if an expression is modified, the numbers may
3976 change. To help with this difficulty, PCRE supports the naming of sub-
3977 patterns. This feature was not added to Perl until release 5.10. Python
3978 had the feature earlier, and PCRE introduced it at release 4.0, using
3979 the Python syntax. PCRE now supports both the Perl and the Python syn-
3980 tax.
3981
3982 In PCRE, a subpattern can be named in one of three ways: (?<name>...)
3983 or (?'name'...) as in Perl, or (?P<name>...) as in Python. References
3984 to capturing parentheses from other parts of the pattern, such as back-
3985 references, recursion, and conditions, can be made by name as well as
3986 by number.
3987
3988 Names consist of up to 32 alphanumeric characters and underscores.
3989 Named capturing parentheses are still allocated numbers as well as
3990 names, exactly as if the names were not present. The PCRE API provides
3991 function calls for extracting the name-to-number translation table from
3992 a compiled pattern. There is also a convenience function for extracting
3993 a captured substring by name.
3994
3995 By default, a name must be unique within a pattern, but it is possible
3996 to relax this constraint by setting the PCRE_DUPNAMES option at compile
3997 time. This can be useful for patterns where only one instance of the
3998 named parentheses can match. Suppose you want to match the name of a
3999 weekday, either as a 3-letter abbreviation or as the full name, and in
4000 both cases you want to extract the abbreviation. This pattern (ignoring
4001 the line breaks) does the job:
4002
4003 (?<DN>Mon|Fri|Sun)(?:day)?|
4004 (?<DN>Tue)(?:sday)?|
4005 (?<DN>Wed)(?:nesday)?|
4006 (?<DN>Thu)(?:rsday)?|
4007 (?<DN>Sat)(?:urday)?
4008
4009 There are five capturing substrings, but only one is ever set after a
4010 match. (An alternative way of solving this problem is to use a "branch
4011 reset" subpattern, as described in the previous section.)
4012
4013 The convenience function for extracting the data by name returns the
4014 substring for the first (and in this example, the only) subpattern of
4015 that name that matched. This saves searching to find which numbered
4016 subpattern it was. If you make a reference to a non-unique named sub-
4017 pattern from elsewhere in the pattern, the one that corresponds to the
4018 lowest number is used. For further details of the interfaces for han-
4019 dling named subpatterns, see the pcreapi documentation.
4020
4021 Warning: You cannot use different names to distinguish between two sub-
4022 patterns with the same number (see the previous section) because PCRE
4023 uses only the numbers when matching.
4024
4025
4026 REPETITION
4027
4028 Repetition is specified by quantifiers, which can follow any of the
4029 following items:
4030
4031 a literal data character
4032 the dot metacharacter
4033 the \C escape sequence
4034 the \X escape sequence (in UTF-8 mode with Unicode properties)
4035 the \R escape sequence
4036 an escape such as \d that matches a single character
4037 a character class
4038 a back reference (see next section)
4039 a parenthesized subpattern (unless it is an assertion)
4040
4041 The general repetition quantifier specifies a minimum and maximum num-
4042 ber of permitted matches, by giving the two numbers in curly brackets
4043 (braces), separated by a comma. The numbers must be less than 65536,
4044 and the first must be less than or equal to the second. For example:
4045
4046 z{2,4}
4047
4048 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
4049 special character. If the second number is omitted, but the comma is
4050 present, there is no upper limit; if the second number and the comma
4051 are both omitted, the quantifier specifies an exact number of required
4052 matches. Thus
4053
4054 [aeiou]{3,}
4055
4056 matches at least 3 successive vowels, but may match many more, while
4057
4058 \d{8}
4059
4060 matches exactly 8 digits. An opening curly bracket that appears in a
4061 position where a quantifier is not allowed, or one that does not match
4062 the syntax of a quantifier, is taken as a literal character. For exam-
4063 ple, {,6} is not a quantifier, but a literal string of four characters.
4064
4065 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to
4066 individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char-
4067 acters, each of which is represented by a two-byte sequence. Similarly,
4068 when Unicode property support is available, \X{3} matches three Unicode
4069 extended sequences, each of which may be several bytes long (and they
4070 may be of different lengths).
4071
4072 The quantifier {0} is permitted, causing the expression to behave as if
4073 the previous item and the quantifier were not present. This may be use-
4074 ful for subpatterns that are referenced as subroutines from elsewhere
4075 in the pattern. Items other than subpatterns that have a {0} quantifier
4076 are omitted from the compiled pattern.
4077
4078 For convenience, the three most common quantifiers have single-charac-
4079 ter abbreviations:
4080
4081 * is equivalent to {0,}
4082 + is equivalent to {1,}
4083 ? is equivalent to {0,1}
4084
4085 It is possible to construct infinite loops by following a subpattern
4086 that can match no characters with a quantifier that has no upper limit,
4087 for example:
4088
4089 (a?)*
4090
4091 Earlier versions of Perl and PCRE used to give an error at compile time
4092 for such patterns. However, because there are cases where this can be
4093 useful, such patterns are now accepted, but if any repetition of the
4094 subpattern does in fact match no characters, the loop is forcibly bro-
4095 ken.
4096
4097 By default, the quantifiers are "greedy", that is, they match as much
4098 as possible (up to the maximum number of permitted times), without
4099 causing the rest of the pattern to fail. The classic example of where
4100 this gives problems is in trying to match comments in C programs. These
4101 appear between /* and */ and within the comment, individual * and /
4102 characters may appear. An attempt to match C comments by applying the
4103 pattern
4104
4105 /\*.*\*/
4106
4107 to the string
4108
4109 /* first comment */ not comment /* second comment */
4110
4111 fails, because it matches the entire string owing to the greediness of
4112 the .* item.
4113
4114 However, if a quantifier is followed by a question mark, it ceases to
4115 be greedy, and instead matches the minimum number of times possible, so
4116 the pattern
4117
4118 /\*.*?\*/
4119
4120 does the right thing with the C comments. The meaning of the various
4121 quantifiers is not otherwise changed, just the preferred number of
4122 matches. Do not confuse this use of question mark with its use as a
4123 quantifier in its own right. Because it has two uses, it can sometimes
4124 appear doubled, as in
4125
4126 \d??\d
4127
4128 which matches one digit by preference, but can match two if that is the
4129 only way the rest of the pattern matches.
4130
4131 If the PCRE_UNGREEDY option is set (an option that is not available in
4132 Perl), the quantifiers are not greedy by default, but individual ones
4133 can be made greedy by following them with a question mark. In other
4134 words, it inverts the default behaviour.
4135
4136 When a parenthesized subpattern is quantified with a minimum repeat
4137 count that is greater than 1 or with a limited maximum, more memory is
4138 required for the compiled pattern, in proportion to the size of the
4139 minimum or maximum.
4140
4141 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
4142 alent to Perl's /s) is set, thus allowing the dot to match newlines,
4143 the pattern is implicitly anchored, because whatever follows will be
4144 tried against every character position in the subject string, so there
4145 is no point in retrying the overall match at any position after the
4146 first. PCRE normally treats such a pattern as though it were preceded
4147 by \A.
4148
4149 In cases where it is known that the subject string contains no new-
4150 lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
4151 mization, or alternatively using ^ to indicate anchoring explicitly.
4152
4153 However, there is one situation where the optimization cannot be used.
4154 When .* is inside capturing parentheses that are the subject of a
4155 backreference elsewhere in the pattern, a match at the start may fail
4156 where a later one succeeds. Consider, for example:
4157
4158 (.*)abc\1
4159
4160 If the subject is "xyz123abc123" the match point is the fourth charac-
4161 ter. For this reason, such a pattern is not implicitly anchored.
4162
4163 When a capturing subpattern is repeated, the value captured is the sub-
4164 string that matched the final iteration. For example, after
4165
4166 (tweedle[dume]{3}\s*)+
4167
4168 has matched "tweedledum tweedledee" the value of the captured substring
4169 is "tweedledee". However, if there are nested capturing subpatterns,
4170 the corresponding captured values may have been set in previous itera-
4171 tions. For example, after
4172
4173 /(a|(b))+/
4174
4175 matches "aba" the value of the second captured substring is "b".
4176
4177
4178 ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS
4179
4180 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
4181 repetition, failure of what follows normally causes the repeated item
4182 to be re-evaluated to see if a different number of repeats allows the
4183 rest of the pattern to match. Sometimes it is useful to prevent this,
4184 either to change the nature of the match, or to cause it fail earlier
4185 than it otherwise might, when the author of the pattern knows there is
4186 no point in carrying on.
4187
4188 Consider, for example, the pattern \d+foo when applied to the subject
4189 line
4190
4191 123456bar
4192
4193 After matching all 6 digits and then failing to match "foo", the normal
4194 action of the matcher is to try again with only 5 digits matching the
4195 \d+ item, and then with 4, and so on, before ultimately failing.
4196 "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
4197 the means for specifying that once a subpattern has matched, it is not
4198 to be re-evaluated in this way.
4199
4200 If we use atomic grouping for the previous example, the matcher gives
4201 up immediately on failing to match "foo" the first time. The notation
4202 is a kind of special parenthesis, starting with (?> as in this example:
4203
4204 (?>\d+)foo
4205
4206 This kind of parenthesis "locks up" the part of the pattern it con-
4207 tains once it has matched, and a failure further into the pattern is
4208 prevented from backtracking into it. Backtracking past it to previous
4209 items, however, works as normal.
4210
4211 An alternative description is that a subpattern of this type matches
4212 the string of characters that an identical standalone pattern would
4213 match, if anchored at the current point in the subject string.
4214
4215 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
4216 such as the above example can be thought of as a maximizing repeat that
4217 must swallow everything it can. So, while both \d+ and \d+? are pre-
4218 pared to adjust the number of digits they match in order to make the
4219 rest of the pattern match, (?>\d+) can only match an entire sequence of
4220 digits.
4221
4222 Atomic groups in general can of course contain arbitrarily complicated
4223 subpatterns, and can be nested. However, when the subpattern for an
4224 atomic group is just a single repeated item, as in the example above, a
4225 simpler notation, called a "possessive quantifier" can be used. This
4226 consists of an additional + character following a quantifier. Using
4227 this notation, the previous example can be rewritten as
4228
4229 \d++foo
4230
4231 Note that a possessive quantifier can be used with an entire group, for
4232 example:
4233
4234 (abc|xyz){2,3}+
4235
4236 Possessive quantifiers are always greedy; the setting of the
4237 PCRE_UNGREEDY option is ignored. They are a convenient notation for the
4238 simpler forms of atomic group. However, there is no difference in the
4239 meaning of a possessive quantifier and the equivalent atomic group,
4240 though there may be a performance difference; possessive quantifiers
4241 should be slightly faster.
4242
4243 The possessive quantifier syntax is an extension to the Perl 5.8 syn-
4244 tax. Jeffrey Friedl originated the idea (and the name) in the first
4245 edition of his book. Mike McCloskey liked it, so implemented it when he
4246 built Sun's Java package, and PCRE copied it from there. It ultimately
4247 found its way into Perl at release 5.10.
4248
4249 PCRE has an optimization that automatically "possessifies" certain sim-
4250 ple pattern constructs. For example, the sequence A+B is treated as
4251 A++B because there is no point in backtracking into a sequence of A's
4252 when B must follow.
4253
4254 When a pattern contains an unlimited repeat inside a subpattern that
4255 can itself be repeated an unlimited number of times, the use of an
4256 atomic group is the only way to avoid some failing matches taking a
4257 very long time indeed. The pattern
4258
4259 (\D+|<\d+>)*[!?]
4260
4261 matches an unlimited number of substrings that either consist of non-
4262 digits, or digits enclosed in <>, followed by either ! or ?. When it
4263 matches, it runs quickly. However, if it is applied to
4264
4265 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4266
4267 it takes a long time before reporting failure. This is because the
4268 string can be divided between the internal \D+ repeat and the external
4269 * repeat in a large number of ways, and all have to be tried. (The
4270 example uses [!?] rather than a single character at the end, because
4271 both PCRE and Perl have an optimization that allows for fast failure
4272 when a single character is used. They remember the last single charac-
4273 ter that is required for a match, and fail early if it is not present
4274 in the string.) If the pattern is changed so that it uses an atomic
4275 group, like this:
4276
4277 ((?>\D+)|<\d+>)*[!?]
4278
4279 sequences of non-digits cannot be broken, and failure happens quickly.
4280
4281
4282 BACK REFERENCES
4283
4284 Outside a character class, a backslash followed by a digit greater than
4285 0 (and possibly further digits) is a back reference to a capturing sub-
4286 pattern earlier (that is, to its left) in the pattern, provided there
4287 have been that many previous capturing left parentheses.
4288
4289 However, if the decimal number following the backslash is less than 10,
4290 it is always taken as a back reference, and causes an error only if
4291 there are not that many capturing left parentheses in the entire pat-
4292 tern. In other words, the parentheses that are referenced need not be
4293 to the left of the reference for numbers less than 10. A "forward back
4294 reference" of this type can make sense when a repetition is involved
4295 and the subpattern to the right has participated in an earlier itera-
4296 tion.
4297
4298 It is not possible to have a numerical "forward back reference" to a
4299 subpattern whose number is 10 or more using this syntax because a
4300 sequence such as \50 is interpreted as a character defined in octal.
4301 See the subsection entitled "Non-printing characters" above for further
4302 details of the handling of digits following a backslash. There is no
4303 such problem when named parentheses are used. A back reference to any
4304 subpattern is possible using named parentheses (see below).
4305
4306 Another way of avoiding the ambiguity inherent in the use of digits
4307 following a backslash is to use the \g escape sequence, which is a fea-
4308 ture introduced in Perl 5.10. This escape must be followed by an
4309 unsigned number or a negative number, optionally enclosed in braces.
4310 These examples are all identical:
4311
4312 (ring), \1
4313 (ring), \g1
4314 (ring), \g{1}
4315
4316 An unsigned number specifies an absolute reference without the ambigu-
4317 ity that is present in the older syntax. It is also useful when literal
4318 digits follow the reference. A negative number is a relative reference.
4319 Consider this example:
4320
4321 (abc(def)ghi)\g{-1}
4322
4323 The sequence \g{-1} is a reference to the most recently started captur-
4324 ing subpattern before \g, that is, is it equivalent to \2. Similarly,
4325 \g{-2} would be equivalent to \1. The use of relative references can be
4326 helpful in long patterns, and also in patterns that are created by
4327 joining together fragments that contain references within themselves.
4328
4329 A back reference matches whatever actually matched the capturing sub-
4330 pattern in the current subject string, rather than anything matching
4331 the subpattern itself (see "Subpatterns as subroutines" below for a way
4332 of doing that). So the pattern
4333
4334 (sens|respons)e and \1ibility
4335
4336 matches "sense and sensibility" and "response and responsibility", but
4337 not "sense and responsibility". If caseful matching is in force at the
4338 time of the back reference, the case of letters is relevant. For exam-
4339 ple,
4340
4341 ((?i)rah)\s+\1
4342
4343 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
4344 original capturing subpattern is matched caselessly.
4345
4346 There are several different ways of writing back references to named
4347 subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or
4348 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's
4349 unified back reference syntax, in which \g can be used for both numeric
4350 and named references, is also supported. We could rewrite the above
4351 example in any of the following ways:
4352
4353 (?<p1>(?i)rah)\s+\k<p1>
4354 (?'p1'(?i)rah)\s+\k{p1}
4355 (?P<p1>(?i)rah)\s+(?P=p1)
4356 (?<p1>(?i)rah)\s+\g{p1}
4357
4358 A subpattern that is referenced by name may appear in the pattern
4359 before or after the reference.
4360
4361 There may be more than one back reference to the same subpattern. If a
4362 subpattern has not actually been used in a particular match, any back
4363 references to it always fail. For example, the pattern
4364
4365 (a|(bc))\2
4366
4367 always fails if it starts to match "a" rather than "bc". Because there
4368 may be many capturing parentheses in a pattern, all digits following
4369 the backslash are taken as part of a potential back reference number.
4370 If the pattern continues with a digit character, some delimiter must be
4371 used to terminate the back reference. If the PCRE_EXTENDED option is
4372 set, this can be whitespace. Otherwise an empty comment (see "Com-
4373 ments" below) can be used.
4374
4375 A back reference that occurs inside the parentheses to which it refers
4376 fails when the subpattern is first used, so, for example, (a\1) never
4377 matches. However, such references can be useful inside repeated sub-
4378 patterns. For example, the pattern
4379
4380 (a|b\1)+
4381
4382 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
4383 ation of the subpattern, the back reference matches the character
4384 string corresponding to the previous iteration. In order for this to
4385 work, the pattern must be such that the first iteration does not need
4386 to match the back reference. This can be done using alternation, as in
4387 the example above, or by a quantifier with a minimum of zero.
4388
4389
4390 ASSERTIONS
4391
4392 An assertion is a test on the characters following or preceding the
4393 current matching point that does not actually consume any characters.
4394 The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
4395 described above.
4396
4397 More complicated assertions are coded as subpatterns. There are two
4398 kinds: those that look ahead of the current position in the subject
4399 string, and those that look behind it. An assertion subpattern is
4400 matched in the normal way, except that it does not cause the current
4401 matching position to be changed.
4402
4403 Assertion subpatterns are not capturing subpatterns, and may not be
4404 repeated, because it makes no sense to assert the same thing several
4405 times. If any kind of assertion contains capturing subpatterns within
4406 it, these are counted for the purposes of numbering the capturing sub-
4407 patterns in the whole pattern. However, substring capturing is carried
4408 out only for positive assertions, because it does not make sense for
4409 negative assertions.
4410
4411 Lookahead assertions
4412
4413 Lookahead assertions start with (?= for positive assertions and (?! for
4414 negative assertions. For example,
4415
4416 \w+(?=;)
4417
4418 matches a word followed by a semicolon, but does not include the semi-
4419 colon in the match, and
4420
4421 foo(?!bar)
4422
4423 matches any occurrence of "foo" that is not followed by "bar". Note
4424 that the apparently similar pattern
4425
4426 (?!foo)bar
4427
4428 does not find an occurrence of "bar" that is preceded by something
4429 other than "foo"; it finds any occurrence of "bar" whatsoever, because
4430 the assertion (?!foo) is always true when the next three characters are
4431 "bar". A lookbehind assertion is needed to achieve the other effect.
4432
4433 If you want to force a matching failure at some point in a pattern, the
4434 most convenient way to do it is with (?!) because an empty string
4435 always matches, so an assertion that requires there not to be an empty
4436 string must always fail.
4437
4438 Lookbehind assertions
4439
4440 Lookbehind assertions start with (?<= for positive assertions and (?<!
4441 for negative assertions. For example,
4442
4443 (?<!foo)bar
4444
4445 does find an occurrence of "bar" that is not preceded by "foo". The
4446 contents of a lookbehind assertion are restricted such that all the
4447 strings it matches must have a fixed length. However, if there are sev-
4448 eral top-level alternatives, they do not all have to have the same
4449 fixed length. Thus
4450
4451 (?<=bullock|donkey)
4452
4453 is permitted, but
4454
4455 (?<!dogs?|cats?)
4456
4457 causes an error at compile time. Branches that match different length
4458 strings are permitted only at the top level of a lookbehind assertion.
4459 This is an extension compared with Perl (at least for 5.8), which
4460 requires all branches to match the same length of string. An assertion
4461 such as
4462
4463 (?<=ab(c|de))
4464
4465 is not permitted, because its single top-level branch can match two
4466 different lengths, but it is acceptable if rewritten to use two top-
4467 level branches:
4468
4469 (?<=abc|abde)
4470
4471 In some cases, the Perl 5.10 escape sequence \K (see above) can be used
4472 instead of a lookbehind assertion; this is not restricted to a fixed-
4473 length.
4474
4475 The implementation of lookbehind assertions is, for each alternative,
4476 to temporarily move the current position back by the fixed length and
4477 then try to match. If there are insufficient characters before the cur-
4478 rent position, the assertion fails.
4479
4480 PCRE does not allow the \C escape (which matches a single byte in UTF-8
4481 mode) to appear in lookbehind assertions, because it makes it impossi-
4482 ble to calculate the length of the lookbehind. The \X and \R escapes,
4483 which can match different numbers of bytes, are also not permitted.
4484
4485 Possessive quantifiers can be used in conjunction with lookbehind
4486 assertions to specify efficient matching at the end of the subject
4487 string. Consider a simple pattern such as
4488
4489 abcd$
4490
4491 when applied to a long string that does not match. Because matching
4492 proceeds from left to right, PCRE will look for each "a" in the subject
4493 and then see if what follows matches the rest of the pattern. If the
4494 pattern is specified as
4495
4496 ^.*abcd$
4497
4498 the initial .* matches the entire string at first, but when this fails
4499 (because there is no following "a"), it backtracks to match all but the
4500 last character, then all but the last two characters, and so on. Once
4501 again the search for "a" covers the entire string, from right to left,
4502 so we are no better off. However, if the pattern is written as
4503
4504 ^.*+(?<=abcd)
4505
4506 there can be no backtracking for the .*+ item; it can match only the
4507 entire string. The subsequent lookbehind assertion does a single test
4508 on the last four characters. If it fails, the match fails immediately.
4509 For long strings, this approach makes a significant difference to the
4510 processing time.
4511
4512 Using multiple assertions
4513
4514 Several assertions (of any sort) may occur in succession. For example,
4515
4516 (?<=\d{3})(?<!999)foo
4517
4518 matches "foo" preceded by three digits that are not "999". Notice that
4519 each of the assertions is applied independently at the same point in
4520 the subject string. First there is a check that the previous three
4521 characters are all digits, and then there is a check that the same
4522 three characters are not "999". This pattern does not match "foo" pre-
4523 ceded by six characters, the first of which are digits and the last
4524 three of which are not "999". For example, it doesn't match "123abc-
4525 foo". A pattern to do that is
4526
4527 (?<=\d{3}...)(?<!999)foo
4528
4529 This time the first assertion looks at the preceding six characters,
4530 checking that the first three are digits, and then the second assertion
4531 checks that the preceding three characters are not "999".
4532
4533 Assertions can be nested in any combination. For example,
4534
4535 (?<=(?<!foo)bar)baz
4536
4537 matches an occurrence of "baz" that is preceded by "bar" which in turn
4538 is not preceded by "foo", while
4539
4540 (?<=\d{3}(?!999)...)foo
4541
4542 is another pattern that matches "foo" preceded by three digits and any
4543 three characters that are not "999".
4544
4545
4546 CONDITIONAL SUBPATTERNS
4547
4548 It is possible to cause the matching process to obey a subpattern con-
4549 ditionally or to choose between two alternative subpatterns, depending
4550 on the result of an assertion, or whether a previous capturing subpat-
4551 tern matched or not. The two possible forms of conditional subpattern
4552 are
4553
4554 (?(condition)yes-pattern)
4555 (?(condition)yes-pattern|no-pattern)
4556
4557 If the condition is satisfied, the yes-pattern is used; otherwise the
4558 no-pattern (if present) is used. If there are more than two alterna-
4559 tives in the subpattern, a compile-time error occurs.
4560
4561 There are four kinds of condition: references to subpatterns, refer-
4562 ences to recursion, a pseudo-condition called DEFINE, and assertions.
4563
4564 Checking for a used subpattern by number
4565
4566 If the text between the parentheses consists of a sequence of digits,
4567 the condition is true if the capturing subpattern of that number has
4568 previously matched. An alternative notation is to precede the digits
4569 with a plus or minus sign. In this case, the subpattern number is rela-
4570 tive rather than absolute. The most recently opened parentheses can be
4571 referenced by (?(-1), the next most recent by (?(-2), and so on. In
4572 looping constructs it can also make sense to refer to subsequent groups
4573 with constructs such as (?(+2).
4574
4575 Consider the following pattern, which contains non-significant white
4576 space to make it more readable (assume the PCRE_EXTENDED option) and to
4577 divide it into three parts for ease of discussion:
4578
4579 ( \( )? [^()]+ (?(1) \) )
4580
4581 The first part matches an optional opening parenthesis, and if that
4582 character is present, sets it as the first captured substring. The sec-
4583 ond part matches one or more characters that are not parentheses. The
4584 third part is a conditional subpattern that tests whether the first set
4585 of parentheses matched or not. If they did, that is, if subject started
4586 with an opening parenthesis, the condition is true, and so the yes-pat-
4587 tern is executed and a closing parenthesis is required. Otherwise,
4588 since no-pattern is not present, the subpattern matches nothing. In
4589 other words, this pattern matches a sequence of non-parentheses,
4590 optionally enclosed in parentheses.
4591
4592 If you were embedding this pattern in a larger one, you could use a
4593 relative reference:
4594
4595 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
4596
4597 This makes the fragment independent of the parentheses in the larger
4598 pattern.
4599
4600 Checking for a used subpattern by name
4601
4602 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a
4603 used subpattern by name. For compatibility with earlier versions of
4604 PCRE, which had this facility before Perl, the syntax (?(name)...) is
4605 also recognized. However, there is a possible ambiguity with this syn-
4606 tax, because subpattern names may consist entirely of digits. PCRE
4607 looks first for a named subpattern; if it cannot find one and the name
4608 consists entirely of digits, PCRE looks for a subpattern of that num-
4609 ber, which must be greater than zero. Using subpattern names that con-
4610 sist entirely of digits is not recommended.
4611
4612 Rewriting the above example to use a named subpattern gives this:
4613
4614 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
4615
4616
4617 Checking for pattern recursion
4618
4619 If the condition is the string (R), and there is no subpattern with the
4620 name R, the condition is true if a recursive call to the whole pattern
4621 or any subpattern has been made. If digits or a name preceded by amper-
4622 sand follow the letter R, for example:
4623
4624 (?(R3)...) or (?(R&name)...)
4625
4626 the condition is true if the most recent recursion is into the subpat-
4627 tern whose number or name is given. This condition does not check the
4628 entire recursion stack.
4629
4630 At "top level", all these recursion test conditions are false. Recur-
4631 sive patterns are described below.
4632
4633 Defining subpatterns for use by reference only
4634
4635 If the condition is the string (DEFINE), and there is no subpattern
4636 with the name DEFINE, the condition is always false. In this case,
4637 there may be only one alternative in the subpattern. It is always
4638 skipped if control reaches this point in the pattern; the idea of
4639 DEFINE is that it can be used to define "subroutines" that can be ref-
4640 erenced from elsewhere. (The use of "subroutines" is described below.)
4641 For example, a pattern to match an IPv4 address could be written like
4642 this (ignore whitespace and line breaks):
4643
4644 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
4645 \b (?&byte) (\.(?&byte)){3} \b
4646
4647 The first part of the pattern is a DEFINE group inside which a another
4648 group named "byte" is defined. This matches an individual component of
4649 an IPv4 address (a number less than 256). When matching takes place,
4650 this part of the pattern is skipped because DEFINE acts like a false
4651 condition.
4652
4653 The rest of the pattern uses references to the named group to match the
4654 four dot-separated components of an IPv4 address, insisting on a word
4655 boundary at each end.
4656
4657 Assertion conditions
4658
4659 If the condition is not in any of the above formats, it must be an
4660 assertion. This may be a positive or negative lookahead or lookbehind
4661 assertion. Consider this pattern, again containing non-significant
4662 white space, and with the two alternatives on the second line:
4663
4664 (?(?=[^a-z]*[a-z])
4665 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
4666
4667 The condition is a positive lookahead assertion that matches an
4668 optional sequence of non-letters followed by a letter. In other words,
4669 it tests for the presence of at least one letter in the subject. If a
4670 letter is found, the subject is matched against the first alternative;
4671 otherwise it is matched against the second. This pattern matches
4672 strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are
4673 letters and dd are digits.
4674
4675
4676 COMMENTS
4677
4678 The sequence (?# marks the start of a comment that continues up to the
4679 next closing parenthesis. Nested parentheses are not permitted. The
4680 characters that make up a comment play no part in the pattern matching
4681 at all.
4682
4683 If the PCRE_EXTENDED option is set, an unescaped # character outside a
4684 character class introduces a comment that continues to immediately
4685 after the next newline in the pattern.
4686
4687
4688 RECURSIVE PATTERNS
4689
4690 Consider the problem of matching a string in parentheses, allowing for
4691 unlimited nested parentheses. Without the use of recursion, the best
4692 that can be done is to use a pattern that matches up to some fixed
4693 depth of nesting. It is not possible to handle an arbitrary nesting
4694 depth.
4695
4696 For some time, Perl has provided a facility that allows regular expres-
4697 sions to recurse (amongst other things). It does this by interpolating
4698 Perl code in the expression at run time, and the code can refer to the
4699 expression itself. A Perl pattern using code interpolation to solve the
4700 parentheses problem can be created like this:
4701
4702 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
4703
4704 The (?p{...}) item interpolates Perl code at run time, and in this case
4705 refers recursively to the pattern in which it appears.
4706
4707 Obviously, PCRE cannot support the interpolation of Perl code. Instead,
4708 it supports special syntax for recursion of the entire pattern, and
4709 also for individual subpattern recursion. After its introduction in
4710 PCRE and Python, this kind of recursion was introduced into Perl at
4711 release 5.10.
4712
4713 A special item that consists of (? followed by a number greater than
4714 zero and a closing parenthesis is a recursive call of the subpattern of
4715 the given number, provided that it occurs inside that subpattern. (If
4716 not, it is a "subroutine" call, which is described in the next sec-
4717 tion.) The special item (?R) or (?0) is a recursive call of the entire
4718 regular expression.
4719
4720 In PCRE (like Python, but unlike Perl), a recursive subpattern call is
4721 always treated as an atomic group. That is, once it has matched some of
4722 the subject string, it is never re-entered, even if it contains untried
4723 alternatives and there is a subsequent matching failure.
4724
4725 This PCRE pattern solves the nested parentheses problem (assume the
4726 PCRE_EXTENDED option is set so that white space is ignored):
4727
4728 \( ( (?>[^()]+) | (?R) )* \)
4729
4730 First it matches an opening parenthesis. Then it matches any number of
4731 substrings which can either be a sequence of non-parentheses, or a
4732 recursive match of the pattern itself (that is, a correctly parenthe-
4733 sized substring). Finally there is a closing parenthesis.
4734
4735 If this were part of a larger pattern, you would not want to recurse
4736 the entire pattern, so instead you could use this:
4737
4738 ( \( ( (?>[^()]+) | (?1) )* \) )
4739
4740 We have put the pattern into parentheses, and caused the recursion to
4741 refer to them instead of the whole pattern.
4742
4743 In a larger pattern, keeping track of parenthesis numbers can be
4744 tricky. This is made easier by the use of relative references. (A Perl
4745 5.10 feature.) Instead of (?1) in the pattern above you can write
4746 (?-2) to refer to the second most recently opened parentheses preceding
4747 the recursion. In other words, a negative number counts capturing
4748 parentheses leftwards from the point at which it is encountered.
4749
4750 It is also possible to refer to subsequently opened parentheses, by
4751 writing references such as (?+2). However, these cannot be recursive
4752 because the reference is not inside the parentheses that are refer-
4753 enced. They are always "subroutine" calls, as described in the next
4754 section.
4755
4756 An alternative approach is to use named parentheses instead. The Perl
4757 syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also
4758 supported. We could rewrite the above example as follows:
4759
4760 (?<pn> \( ( (?>[^()]+) | (?&pn) )* \) )
4761
4762 If there is more than one subpattern with the same name, the earliest
4763 one is used.
4764
4765 This particular example pattern that we have been looking at contains
4766 nested unlimited repeats, and so the use of atomic grouping for match-
4767 ing strings of non-parentheses is important when applying the pattern
4768 to strings that do not match. For example, when this pattern is applied
4769 to
4770
4771 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
4772
4773 it yields "no match" quickly. However, if atomic grouping is not used,
4774 the match runs for a very long time indeed because there are so many
4775 different ways the + and * repeats can carve up the subject, and all
4776 have to be tested before failure can be reported.
4777
4778 At the end of a match, the values set for any capturing subpatterns are
4779 those from the outermost level of the recursion at which the subpattern
4780 value is set. If you want to obtain intermediate values, a callout
4781 function can be used (see below and the pcrecallout documentation). If
4782 the pattern above is matched against
4783
4784 (ab(cd)ef)
4785
4786 the value for the capturing parentheses is "ef", which is the last
4787 value taken on at the top level. If additional parentheses are added,
4788 giving
4789
4790 \( ( ( (?>[^()]+) | (?R) )* ) \)
4791 ^ ^
4792 ^ ^
4793
4794 the string they capture is "ab(cd)ef", the contents of the top level
4795 parentheses. If there are more than 15 capturing parentheses in a pat-
4796 tern, PCRE has to obtain extra memory to store data during a recursion,
4797 which it does by using pcre_malloc, freeing it via pcre_free after-
4798 wards. If no memory can be obtained, the match fails with the
4799 PCRE_ERROR_NOMEMORY error.
4800
4801 Do not confuse the (?R) item with the condition (R), which tests for
4802 recursion. Consider this pattern, which matches text in angle brack-
4803 ets, allowing for arbitrary nesting. Only digits are allowed in nested
4804 brackets (that is, when recursing), whereas any characters are permit-
4805 ted at the outer level.
4806
4807 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
4808
4809 In this pattern, (?(R) is the start of a conditional subpattern, with
4810 two different alternatives for the recursive and non-recursive cases.
4811 The (?R) item is the actual recursive call.
4812
4813
4814 SUBPATTERNS AS SUBROUTINES
4815
4816 If the syntax for a recursive subpattern reference (either by number or
4817 by name) is used outside the parentheses to which it refers, it oper-
4818 ates like a subroutine in a programming language. The "called" subpat-
4819 tern may be defined before or after the reference. A numbered reference
4820 can be absolute or relative, as in these examples:
4821
4822 (...(absolute)...)...(?2)...
4823 (...(relative)...)...(?-1)...
4824 (...(?+1)...(relative)...
4825
4826 An earlier example pointed out that the pattern
4827
4828 (sens|respons)e and \1ibility
4829
4830 matches "sense and sensibility" and "response and responsibility", but
4831 not "sense and responsibility". If instead the pattern
4832
4833 (sens|respons)e and (?1)ibility
4834
4835 is used, it does match "sense and responsibility" as well as the other
4836 two strings. Another example is given in the discussion of DEFINE
4837 above.
4838
4839 Like recursive subpatterns, a "subroutine" call is always treated as an
4840 atomic group. That is, once it has matched some of the subject string,
4841 it is never re-entered, even if it contains untried alternatives and
4842 there is a subsequent matching failure.
4843
4844 When a subpattern is used as a subroutine, processing options such as
4845 case-independence are fixed when the subpattern is defined. They cannot
4846 be changed for different calls. For example, consider this pattern:
4847
4848 (abc)(?i:(?-1))
4849
4850 It matches "abcabc". It does not match "abcABC" because the change of
4851 processing option does not affect the called subpattern.
4852
4853
4854 ONIGURUMA SUBROUTINE SYNTAX
4855
4856 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
4857 name or a number enclosed either in angle brackets or single quotes, is
4858 an alternative syntax for referencing a subpattern as a subroutine,
4859 possibly recursively. Here are two of the examples used above, rewrit-
4860 ten using this syntax:
4861
4862 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
4863 (sens|respons)e and \g'1'ibility
4864
4865 PCRE supports an extension to Oniguruma: if a number is preceded by a
4866 plus or a minus sign it is taken as a relative reference. For example:
4867
4868 (abc)(?i:\g<-1>)
4869
4870 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not
4871 synonymous. The former is a back reference; the latter is a subroutine
4872 call.
4873
4874
4875 CALLOUTS
4876
4877 Perl has a feature whereby using the sequence (?{...}) causes arbitrary
4878 Perl code to be obeyed in the middle of matching a regular expression.
4879 This makes it possible, amongst other things, to extract different sub-
4880 strings that match the same pair of parentheses when there is a repeti-
4881 tion.
4882
4883 PCRE provides a similar feature, but of course it cannot obey arbitrary
4884 Perl code. The feature is called "callout". The caller of PCRE provides
4885 an external function by putting its entry point in the global variable
4886 pcre_callout. By default, this variable contains NULL, which disables
4887 all calling out.
4888
4889 Within a regular expression, (?C) indicates the points at which the
4890 external function is to be called. If you want to identify different
4891 callout points, you can put a number less than 256 after the letter C.
4892 The default value is zero. For example, this pattern has two callout
4893 points:
4894
4895 (?C1)abc(?C2)def
4896
4897 If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are
4898 automatically installed before each item in the pattern. They are all
4899 numbered 255.
4900
4901 During matching, when PCRE reaches a callout point (and pcre_callout is
4902 set), the external function is called. It is provided with the number
4903 of the callout, the position in the pattern, and, optionally, one item
4904 of data originally supplied by the caller of pcre_exec(). The callout
4905 function may cause matching to proceed, to backtrack, or to fail alto-
4906 gether. A complete description of the interface to the callout function
4907 is given in the pcrecallout documentation.
4908
4909
4910 BACKTRACKING CONTROL
4911
4912 Perl 5.10 introduced a number of "Special Backtracking Control Verbs",
4913 which are described in the Perl documentation as "experimental and sub-
4914 ject to change or removal in a future version of Perl". It goes on to
4915 say: "Their usage in production code should be noted to avoid problems
4916 during upgrades." The same remarks apply to the PCRE features described
4917 in this section.
4918
4919 Since these verbs are specifically related to backtracking, most of
4920 them can be used only when the pattern is to be matched using
4921 pcre_exec(), which uses a backtracking algorithm. With the exception of
4922 (*FAIL), which behaves like a failing negative assertion, they cause an
4923 error if encountered by pcre_dfa_exec().
4924
4925 The new verbs make use of what was previously invalid syntax: an open-
4926 ing parenthesis followed by an asterisk. In Perl, they are generally of
4927 the form (*VERB:ARG) but PCRE does not support the use of arguments, so
4928 its general form is just (*VERB). Any number of these verbs may occur
4929 in a pattern. There are two kinds:
4930
4931 Verbs that act immediately
4932
4933 The following verbs act as soon as they are encountered:
4934
4935 (*ACCEPT)
4936
4937 This verb causes the match to end successfully, skipping the remainder
4938 of the pattern. When inside a recursion, only the innermost pattern is
4939 ended immediately. PCRE differs from Perl in what happens if the
4940 (*ACCEPT) is inside capturing parentheses. In Perl, the data so far is
4941 captured: in PCRE no data is captured. For example:
4942
4943 A(A|B(*ACCEPT)|C)D
4944
4945 This matches "AB", "AAD", or "ACD", but when it matches "AB", no data
4946 is captured.
4947
4948 (*FAIL) or (*F)
4949
4950 This verb causes the match to fail, forcing backtracking to occur. It
4951 is equivalent to (?!) but easier to read. The Perl documentation notes
4952 that it is probably useful only when combined with (?{}) or (??{}).
4953 Those are, of course, Perl features that are not present in PCRE. The
4954 nearest equivalent is the callout feature, as for example in this pat-
4955 tern:
4956
4957 a+(?C)(*FAIL)
4958
4959 A match with the string "aaaa" always fails, but the callout is taken
4960 before each backtrack happens (in this example, 10 times).
4961
4962 Verbs that act after backtracking
4963
4964 The following verbs do nothing when they are encountered. Matching con-
4965 tinues with what follows, but if there is no subsequent match, a fail-
4966 ure is forced. The verbs differ in exactly what kind of failure
4967 occurs.
4968
4969 (*COMMIT)
4970
4971 This verb causes the whole match to fail outright if the rest of the
4972 pattern does not match. Even if the pattern is unanchored, no further
4973 attempts to find a match by advancing the start point take place. Once
4974 (*COMMIT) has been passed, pcre_exec() is committed to finding a match
4975 at the current starting point, or not at all. For example:
4976
4977 a+(*COMMIT)b
4978
4979 This matches "xxaab" but not "aacaab". It can be thought of as a kind
4980 of dynamic anchor, or "I've started, so I must finish."
4981
4982 (*PRUNE)
4983
4984 This verb causes the match to fail at the current position if the rest
4985 of the pattern does not match. If the pattern is unanchored, the normal
4986 "bumpalong" advance to the next starting character then happens. Back-
4987 tracking can occur as usual to the left of (*PRUNE), or when matching
4988 to the right of (*PRUNE), but if there is no match to the right, back-
4989 tracking cannot cross (*PRUNE). In simple cases, the use of (*PRUNE)
4990 is just an alternative to an atomic group or possessive quantifier, but
4991 there are some uses of (*PRUNE) that cannot be expressed in any other
4992 way.
4993
4994 (*SKIP)
4995
4996 This verb is like (*PRUNE), except that if the pattern is unanchored,
4997 the "bumpalong" advance is not to the next character, but to the posi-
4998 tion in the subject where (*SKIP) was encountered. (*SKIP) signifies
4999 that whatever text was matched leading up to it cannot be part of a
5000 successful match. Consider:
5001
5002 a+(*SKIP)b
5003
5004 If the subject is "aaaac...", after the first match attempt fails
5005 (starting at the first character in the string), the starting point
5006 skips on to start the next attempt at "c". Note that a possessive quan-
5007 tifer does not have the same effect in this example; although it would
5008 suppress backtracking during the first match attempt, the second
5009 attempt would start at the second character instead of skipping on to
5010 "c".
5011
5012 (*THEN)
5013
5014 This verb causes a skip to the next alternation if the rest of the pat-
5015 tern does not match. That is, it cancels pending backtracking, but only
5016 within the current alternation. Its name comes from the observation
5017 that it can be used for a pattern-based if-then-else block:
5018
5019 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
5020
5021 If the COND1 pattern matches, FOO is tried (and possibly further items
5022 after the end of the group if FOO succeeds); on failure the matcher
5023 skips to the second alternative and tries COND2, without backtracking
5024 into COND1. If (*THEN) is used outside of any alternation, it acts
5025 exactly like (*PRUNE).
5026
5027
5028 SEE ALSO
5029
5030 pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
5031
5032
5033 AUTHOR
5034
5035 Philip Hazel
5036 University Computing Service
5037 Cambridge CB2 3QH, England.
5038
5039
5040 REVISION
5041
5042 Last updated: 11 April 2009
5043 Copyright (c) 1997-2009 University of Cambridge.
5044 ------------------------------------------------------------------------------
5045
5046
5047 PCRESYNTAX(3) PCRESYNTAX(3)
5048
5049
5050 NAME
5051 PCRE - Perl-compatible regular expressions
5052
5053
5054 PCRE REGULAR EXPRESSION SYNTAX SUMMARY
5055
5056 The full syntax and semantics of the regular expressions that are sup-
5057 ported by PCRE are described in the pcrepattern documentation. This
5058 document contains just a quick-reference summary of the syntax.
5059
5060
5061 QUOTING
5062
5063 \x where x is non-alphanumeric is a literal x
5064 \Q...\E treat enclosed characters as literal
5065
5066
5067 CHARACTERS
5068
5069 \a alarm, that is, the BEL character (hex 07)
5070 \cx "control-x", where x is any character
5071 \e escape (hex 1B)
5072 \f formfeed (hex 0C)
5073 \n newline (hex 0A)
5074 \r carriage return (hex 0D)
5075 \t tab (hex 09)
5076 \ddd character with octal code ddd, or backreference
5077 \xhh character with hex code hh
5078 \x{hhh..} character with hex code hhh..
5079
5080
5081 CHARACTER TYPES
5082
5083 . any character except newline;
5084 in dotall mode, any character whatsoever
5085 \C one byte, even in UTF-8 mode (best avoided)
5086 \d a decimal digit
5087 \D a character that is not a decimal digit
5088 \h a horizontal whitespace character
5089 \H a character that is not a horizontal whitespace character
5090 \p{xx} a character with the xx property
5091 \P{xx} a character without the xx property
5092 \R a newline sequence
5093 \s a whitespace character
5094 \S a character that is not a whitespace character
5095 \v a vertical whitespace character
5096 \V a character that is not a vertical whitespace character
5097 \w a "word" character
5098 \W a "non-word" character
5099 \X an extended Unicode sequence
5100
5101 In PCRE, \d, \D, \s, \S, \w, and \W recognize only ASCII characters.
5102
5103
5104 GENERAL CATEGORY PROPERTY CODES FOR \p and \P
5105
5106 C Other
5107 Cc Control
5108 Cf Format
5109 Cn Unassigned
5110 Co Private use
5111 Cs Surrogate
5112
5113 L Letter
5114 Ll Lower case letter
5115 Lm Modifier letter
5116 Lo Other letter
5117 Lt Title case letter
5118 Lu Upper case letter
5119 L& Ll, Lu, or Lt
5120
5121 M Mark
5122 Mc Spacing mark
5123 Me Enclosing mark
5124 Mn Non-spacing mark
5125
5126 N Number
5127 Nd Decimal number
5128 Nl Letter number
5129 No Other number
5130
5131 P Punctuation
5132 Pc Connector punctuation
5133 Pd Dash punctuation
5134 Pe Close punctuation
5135 Pf Final punctuation
5136 Pi Initial punctuation
5137 Po Other punctuation
5138 Ps Open punctuation
5139
5140 S Symbol
5141 Sc Currency symbol
5142 Sk Modifier symbol
5143 Sm Mathematical symbol
5144 So Other symbol
5145
5146 Z Separator
5147 Zl Line separator
5148 Zp Paragraph separator
5149 Zs Space separator
5150
5151
5152 SCRIPT NAMES FOR \p AND \P
5153
5154 Arabic, Armenian, Balinese, Bengali, Bopomofo, Braille, Buginese,
5155 Buhid, Canadian_Aboriginal, Carian, Cham, Cherokee, Common, Coptic, Cu-
5156 neiform, Cypriot, Cyrillic, Deseret, Devanagari, Ethiopic, Georgian,
5157 Glagolitic, Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo,
5158 Hebrew, Hiragana, Inherited, Kannada, Katakana, Kayah_Li, Kharoshthi,
5159 Khmer, Lao, Latin, Lepcha, Limbu, Linear_B, Lycian, Lydian, Malayalam,
5160 Mongolian, Myanmar, New_Tai_Lue, Nko, Ogham, Old_Italic, Old_Persian,
5161 Ol_Chiki, Oriya, Osmanya, Phags_Pa, Phoenician, Rejang, Runic, Saurash-
5162 tra, Shavian, Sinhala, Sudanese, Syloti_Nagri, Syriac, Tagalog, Tag-
5163 banwa, Tai_Le, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh,
5164 Ugaritic, Vai, Yi.
5165
5166
5167 CHARACTER CLASSES
5168
5169 [...] positive character class
5170 [^...] negative character class
5171 [x-y] range (can be used for hex characters)
5172 [[:xxx:]] positive POSIX named set
5173 [[:^xxx:]] negative POSIX named set
5174
5175 alnum alphanumeric
5176 alpha alphabetic
5177 ascii 0-127
5178 blank space or tab
5179 cntrl control character
5180 digit decimal digit
5181 graph printing, excluding space
5182 lower lower case letter
5183 print printing, including space
5184 punct printing, excluding alphanumeric
5185 space whitespace
5186 upper upper case letter
5187 word same as \w
5188 xdigit hexadecimal digit
5189
5190 In PCRE, POSIX character set names recognize only ASCII characters. You
5191 can use \Q...\E inside a character class.
5192
5193
5194 QUANTIFIERS
5195
5196 ? 0 or 1, greedy
5197 ?+ 0 or 1, possessive
5198 ?? 0 or 1, lazy
5199 * 0 or more, greedy
5200 *+ 0 or more, possessive
5201 *? 0 or more, lazy
5202 + 1 or more, greedy
5203 ++ 1 or more, possessive
5204 +? 1 or more, lazy
5205 {n} exactly n
5206 {n,m} at least n, no more than m, greedy
5207 {n,m}+ at least n, no more than m, possessive
5208 {n,m}? at least n, no more than m, lazy
5209 {n,} n or more, greedy
5210 {n,}+ n or more, possessive
5211 {n,}? n or more, lazy
5212
5213
5214 ANCHORS AND SIMPLE ASSERTIONS
5215
5216 \b word boundary (only ASCII letters recognized)
5217 \B not a word boundary
5218 ^ start of subject
5219 also after internal newline in multiline mode
5220 \A start of subject
5221 $ end of subject
5222 also before newline at end of subject
5223 also before internal newline in multiline mode
5224 \Z end of subject
5225 also before newline at end of subject
5226 \z end of subject
5227 \G first matching position in subject
5228
5229
5230 MATCH POINT RESET
5231
5232 \K reset start of match
5233
5234
5235 ALTERNATION
5236
5237 expr|expr|expr...
5238
5239
5240 CAPTURING
5241
5242 (...) capturing group
5243 (?<name>...) named capturing group (Perl)
5244 (?'name'...) named capturing group (Perl)
5245 (?P<name>...) named capturing group (Python)
5246 (?:...) non-capturing group
5247 (?|...) non-capturing group; reset group numbers for
5248 capturing groups in each alternative
5249
5250
5251 ATOMIC GROUPS
5252
5253 (?>...) atomic, non-capturing group
5254
5255
5256 COMMENT
5257
5258 (?#....) comment (not nestable)
5259
5260
5261 OPTION SETTING
5262
5263 (?i) caseless
5264 (?J) allow duplicate names
5265 (?m) multiline
5266 (?s) single line (dotall)
5267 (?U) default ungreedy (lazy)
5268 (?x) extended (ignore white space)
5269 (?-...) unset option(s)
5270
5271 The following is recognized only at the start of a pattern or after one
5272 of the newline-setting options with similar syntax:
5273
5274 (*UTF8) set UTF-8 mode
5275
5276
5277 LOOKAHEAD AND LOOKBEHIND ASSERTIONS
5278
5279 (?=...) positive look ahead
5280 (?!...) negative look ahead
5281 (?<=...) positive look behind
5282 (?<!...) negative look behind
5283
5284 Each top-level branch of a look behind must be of a fixed length.
5285
5286
5287 BACKREFERENCES
5288
5289 \n reference by number (can be ambiguous)
5290 \gn reference by number
5291 \g{n} reference by number
5292 \g{-n} relative reference by number
5293 \k<name> reference by name (Perl)
5294 \k'name' reference by name (Perl)
5295 \g{name} reference by name (Perl)
5296 \k{name} reference by name (.NET)
5297 (?P=name) reference by name (Python)
5298
5299
5300 SUBROUTINE REFERENCES (POSSIBLY RECURSIVE)
5301
5302 (?R) recurse whole pattern
5303 (?n) call subpattern by absolute number
5304 (?+n) call subpattern by relative number
5305 (?-n) call subpattern by relative number
5306 (?&name) call subpattern by name (Perl)
5307 (?P>name) call subpattern by name (Python)
5308 \g<name> call subpattern by name (Oniguruma)
5309 \g'name' call subpattern by name (Oniguruma)
5310 \g<n> call subpattern by absolute number (Oniguruma)
5311 \g'n' call subpattern by absolute number (Oniguruma)
5312 \g<+n> call subpattern by relative number (PCRE extension)
5313 \g'+n' call subpattern by relative number (PCRE extension)
5314 \g<-n> call subpattern by relative number (PCRE extension)
5315 \g'-n' call subpattern by relative number (PCRE extension)
5316
5317
5318 CONDITIONAL PATTERNS
5319
5320 (?(condition)yes-pattern)
5321 (?(condition)yes-pattern|no-pattern)
5322
5323 (?(n)... absolute reference condition
5324 (?(+n)... relative reference condition
5325 (?(-n)... relative reference condition
5326 (?(<name>)... named reference condition (Perl)
5327 (?('name')... named reference condition (Perl)
5328 (?(name)... named reference condition (PCRE)
5329 (?(R)... overall recursion condition
5330 (?(Rn)... specific group recursion condition
5331 (?(R&name)... specific recursion condition
5332 (?(DEFINE)... define subpattern for reference
5333 (?(assert)... assertion condition
5334
5335
5336 BACKTRACKING CONTROL
5337
5338 The following act immediately they are reached:
5339
5340 (*ACCEPT) force successful match
5341 (*FAIL) force backtrack; synonym (*F)
5342
5343 The following act only when a subsequent match failure causes a back-
5344 track to reach them. They all force a match failure, but they differ in
5345 what happens afterwards. Those that advance the start-of-match point do
5346 so only if the pattern is not anchored.
5347
5348 (*COMMIT) overall failure, no advance of starting point
5349 (*PRUNE) advance to next starting character
5350 (*SKIP) advance start to current matching position
5351 (*THEN) local failure, backtrack to next alternation
5352
5353
5354 NEWLINE CONVENTIONS
5355
5356 These are recognized only at the very start of the pattern or after a
5357 (*BSR_...) or (*UTF8) option.
5358
5359 (*CR) carriage return only
5360 (*LF) linefeed only
5361 (*CRLF) carriage return followed by linefeed
5362 (*ANYCRLF) all three of the above
5363 (*ANY) any Unicode newline sequence
5364
5365
5366 WHAT \R MATCHES
5367
5368 These are recognized only at the very start of the pattern or after a
5369 (*...) option that sets the newline convention or UTF-8 mode.
5370
5371 (*BSR_ANYCRLF) CR, LF, or CRLF
5372 (*BSR_UNICODE) any Unicode newline sequence
5373
5374
5375 CALLOUTS
5376
5377 (?C) callout
5378 (?Cn) callout with data n
5379
5380
5381 SEE ALSO
5382
5383 pcrepattern(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
5384
5385
5386 AUTHOR
5387
5388 Philip Hazel
5389 University Computing Service
5390 Cambridge CB2 3QH, England.
5391
5392
5393 REVISION
5394
5395 Last updated: 11 April 2009
5396 Copyright (c) 1997-2009 University of Cambridge.
5397 ------------------------------------------------------------------------------
5398
5399
5400 PCREPARTIAL(3) PCREPARTIAL(3)
5401
5402
5403 NAME
5404 PCRE - Perl-compatible regular expressions
5405
5406
5407 PARTIAL MATCHING IN PCRE
5408
5409 In normal use of PCRE, if the subject string that is passed to
5410 pcre_exec() or pcre_dfa_exec() matches as far as it goes, but is too
5411 short to match the entire pattern, PCRE_ERROR_NOMATCH is returned.
5412 There are circumstances where it might be helpful to distinguish this
5413 case from other cases in which there is no match.
5414
5415 Consider, for example, an application where a human is required to type
5416 in data for a field with specific formatting requirements. An example
5417 might be a date in the form ddmmmyy, defined by this pattern:
5418
5419 ^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$
5420
5421 If the application sees the user's keystrokes one by one, and can check
5422 that what has been typed so far is potentially valid, it is able to
5423 raise an error as soon as a mistake is made, by beeping and not
5424 reflecting the character that has been typed, for example. This immedi-
5425 ate feedback is likely to be a better user interface than a check that
5426 is delayed until the entire string has been entered. Partial matching
5427 can also sometimes be useful when the subject string is very long and
5428 is not all available at once.
5429
5430 PCRE supports partial matching by means of the PCRE_PARTIAL_SOFT and
5431 PCRE_PARTIAL_HARD options, which can be set when calling pcre_exec() or
5432 pcre_dfa_exec(). For backwards compatibility, PCRE_PARTIAL is a synonym
5433 for PCRE_PARTIAL_SOFT. The essential difference between the two options
5434 is whether or not a partial match is preferred to an alternative com-
5435 plete match, though the details differ between the two matching func-
5436 tions. If both options are set, PCRE_PARTIAL_HARD takes precedence.
5437
5438 Setting a partial matching option disables one of PCRE's optimizations.
5439 PCRE remembers the last literal byte in a pattern, and abandons match-
5440 ing immediately if such a byte is not present in the subject string.
5441 This optimization cannot be used for a subject string that might match
5442 only partially.
5443
5444
5445 PARTIAL MATCHING USING pcre_exec()
5446
5447 A partial match occurs during a call to pcre_exec() whenever the end of
5448 the subject string is reached successfully, but matching cannot con-
5449 tinue because more characters are needed. However, at least one charac-
5450 ter must have been matched. (In other words, a partial match can never
5451 be an empty string.)
5452
5453 If PCRE_PARTIAL_SOFT is set, the partial match is remembered, but
5454 matching continues as normal, and other alternatives in the pattern are
5455 tried. If no complete match can be found, pcre_exec() returns
5456 PCRE_ERROR_PARTIAL instead of PCRE_ERROR_NOMATCH, and if there are at
5457 least two slots in the offsets vector, they are filled in with the off-
5458 sets of the longest string that partially matched. Consider this pat-
5459 tern:
5460
5461 /123\w+X|dogY/
5462
5463 If this is matched against the subject string "abc123dog", both alter-
5464 natives fail to match, but the end of the subject is reached during
5465 matching, so PCRE_ERROR_PARTIAL is returned instead of
5466 PCRE_ERROR_NOMATCH. The offsets are set to 3 and 9, identifying
5467 "123dog" as the longest partial match that was found. (In this example,
5468 there are two partial matches, because "dog" on its own partially
5469 matches the second alternative.)
5470
5471 If PCRE_PARTIAL_HARD is set for pcre_exec(), it returns PCRE_ERROR_PAR-
5472 TIAL as soon as a partial match is found, without continuing to search
5473 for possible complete matches. The difference between the two options
5474 can be illustrated by a pattern such as:
5475
5476 /dog(sbody)?/
5477
5478 This matches either "dog" or "dogsbody", greedily (that is, it prefers
5479 the longer string if possible). If it is matched against the string
5480 "dog" with PCRE_PARTIAL_SOFT, it yields a complete match for "dog".
5481 However, if PCRE_PARTIAL_HARD is set, the result is PCRE_ERROR_PARTIAL.
5482 On the other hand, if the pattern is made ungreedy the result is dif-
5483 ferent:
5484
5485 /dog(sbody)??/
5486
5487 In this case the result is always a complete match because pcre_exec()
5488 finds that first, and it never continues after finding a match. It
5489 might be easier to follow this explanation by thinking of the two pat-
5490 terns like this:
5491
5492 /dog(sbody)?/ is the same as /dogsbody|dog/
5493 /dog(sbody)??/ is the same as /dog|dogsbody/
5494
5495 The second pattern will never match "dogsbody" when pcre_exec() is
5496 used, because it will always find the shorter match first.
5497
5498
5499 PARTIAL MATCHING USING pcre_dfa_exec()
5500
5501 The pcre_dfa_exec() function moves along the subject string character
5502 by character, without backtracking, searching for all possible matches
5503 simultaneously. If the end of the subject is reached before the end of
5504 the pattern, there is the possibility of a partial match, again pro-
5505 vided that at least one character has matched.
5506
5507 When PCRE_PARTIAL_SOFT is set, PCRE_ERROR_PARTIAL is returned only if
5508 there have been no complete matches. Otherwise, the complete matches
5509 are returned. However, if PCRE_PARTIAL_HARD is set, a partial match
5510 takes precedence over any complete matches. The portion of the string
5511 that provided the longest partial match is set as the first matching
5512 string, provided there are at least two slots in the offsets vector.
5513
5514 Because pcre_dfa_exec() always searches for all possible matches, and
5515 there is no difference between greedy and ungreedy repetition, its be-
5516 haviour is different from pcre_exec when PCRE_PARTIAL_HARD is set. Con-
5517 sider the string "dog" matched against the ungreedy pattern shown
5518 above:
5519
5520 /dog(sbody)??/
5521
5522 Whereas pcre_exec() stops as soon as it finds the complete match for
5523 "dog", pcre_dfa_exec() also finds the partial match for "dogsbody", and
5524 so returns that when PCRE_PARTIAL_HARD is set.
5525
5526
5527 PARTIAL MATCHING AND WORD BOUNDARIES
5528
5529 If a pattern ends with one of sequences \w or \W, which test for word
5530 boundaries, partial matching with PCRE_PARTIAL_SOFT can give counter-
5531 intuitive results. Consider this pattern:
5532
5533 /\bcat\b/
5534
5535 This matches "cat", provided there is a word boundary at either end. If
5536 the subject string is "the cat", the comparison of the final "t" with a
5537 following character cannot take place, so a partial match is found.
5538 However, pcre_exec() carries on with normal matching, which matches \b
5539 at the end of the subject when the last character is a letter, thus
5540 finding a complete match. The result, therefore, is not PCRE_ERROR_PAR-
5541 TIAL. The same thing happens with pcre_dfa_exec(), because it also
5542 finds the complete match.
5543
5544 Using PCRE_PARTIAL_HARD in this case does yield PCRE_ERROR_PARTIAL,
5545 because then the partial match takes precedence.
5546
5547
5548 FORMERLY RESTRICTED PATTERNS
5549
5550 For releases of PCRE prior to 8.00, because of the way certain internal
5551 optimizations were implemented in the pcre_exec() function, the
5552 PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be
5553 used with all patterns. From release 8.00 onwards, the restrictions no
5554 longer apply, and partial matching with pcre_exec() can be requested
5555 for any pattern.
5556
5557 Items that were formerly restricted were repeated single characters and
5558 repeated metasequences. If PCRE_PARTIAL was set for a pattern that did
5559 not conform to the restrictions, pcre_exec() returned the error code
5560 PCRE_ERROR_BADPARTIAL (-13). This error code is no longer in use. The
5561 PCRE_INFO_OKPARTIAL call to pcre_fullinfo() to find out if a compiled
5562 pattern can be used for partial matching now always returns 1.
5563
5564
5565 EXAMPLE OF PARTIAL MATCHING USING PCRETEST
5566
5567 If the escape sequence \P is present in a pcretest data line, the
5568 PCRE_PARTIAL_SOFT option is used for the match. Here is a run of
5569 pcretest that uses the date example quoted above:
5570
5571 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
5572 data> 25jun04\P
5573 0: 25jun04
5574 1: jun
5575 data> 25dec3\P
5576 Partial match: 23dec3
5577 data> 3ju\P
5578 Partial match: 3ju
5579 data> 3juj\P
5580 No match
5581 data> j\P
5582 No match
5583
5584 The first data string is matched completely, so pcretest shows the
5585 matched substrings. The remaining four strings do not match the com-
5586 plete pattern, but the first two are partial matches. Similar output is
5587 obtained when pcre_dfa_exec() is used.
5588
5589 If the escape sequence \P is present more than once in a pcretest data
5590 line, the PCRE_PARTIAL_HARD option is set for the match.
5591
5592
5593 MULTI-SEGMENT MATCHING WITH pcre_dfa_exec()
5594
5595 When a partial match has been found using pcre_dfa_exec(), it is possi-
5596 ble to continue the match by providing additional subject data and
5597 calling pcre_dfa_exec() again with the same compiled regular expres-
5598 sion, this time setting the PCRE_DFA_RESTART option. You must pass the
5599 same working space as before, because this is where details of the pre-
5600 vious partial match are stored. Here is an example using pcretest,
5601 using the \R escape sequence to set the PCRE_DFA_RESTART option (\D
5602 specifies the use of pcre_dfa_exec()):
5603
5604 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
5605 data> 23ja\P\D
5606 Partial match: 23ja
5607 data> n05\R\D
5608 0: n05
5609
5610 The first call has "23ja" as the subject, and requests partial match-
5611 ing; the second call has "n05" as the subject for the continued
5612 (restarted) match. Notice that when the match is complete, only the
5613 last part is shown; PCRE does not retain the previously partially-
5614 matched string. It is up to the calling program to do that if it needs
5615 to.
5616
5617 You can set the PCRE_PARTIAL_SOFT or PCRE_PARTIAL_HARD options with
5618 PCRE_DFA_RESTART to continue partial matching over multiple segments.
5619 This facility can be used to pass very long subject strings to
5620 pcre_dfa_exec().
5621
5622
5623 MULTI-SEGMENT MATCHING WITH pcre_exec()
5624
5625 From release 8.00, pcre_exec() can also be used to do multi-segment
5626 matching. Unlike pcre_dfa_exec(), it is not possible to restart the
5627 previous match with a new segment of data. Instead, new data must be
5628 added to the previous subject string, and the entire match re-run,
5629 starting from the point where the partial match occurred. Earlier data
5630 can be discarded. Consider an unanchored pattern that matches dates:
5631
5632 re> /\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d/
5633 data> The date is 23ja\P
5634 Partial match: 23ja
5635
5636 The this stage, an application could discard the text preceding "23ja",
5637 add on text from the next segment, and call pcre_exec() again. Unlike
5638 pcre_dfa_exec(), the entire matching string must always be available,
5639 and the complete matching process occurs for each call, so more memory
5640 and more processing time is needed.
5641
5642
5643 ISSUES WITH MULTI-SEGMENT MATCHING
5644
5645 Certain types of pattern may give problems with multi-segment matching,
5646 whichever matching function is used.
5647
5648 1. If the pattern contains tests for the beginning or end of a line,
5649 you need to pass the PCRE_NOTBOL or PCRE_NOTEOL options, as appropri-
5650 ate, when the subject string for any call does not contain the begin-
5651 ning or end of a line.
5652
5653 2. If the pattern contains backward assertions (including \b or \B),
5654 you need to arrange for some overlap in the subject strings to allow
5655 for them to be correctly tested at the start of each substring. For
5656 example, using pcre_dfa_exec(), you could pass the subject in chunks
5657 that are 500 bytes long, but in a buffer of 700 bytes, with the start-
5658 ing offset set to 200 and the previous 200 bytes at the start of the
5659 buffer.
5660
5661 3. Matching a subject string that is split into multiple segments may
5662 not always produce exactly the same result as matching over one single
5663 long string, especially when PCRE_PARTIAL_SOFT is used. The section
5664 "Partial Matching and Word Boundaries" above describes an issue that
5665 arises if the pattern ends with \b or \B. Another kind of difference
5666 may occur when there are multiple matching possibilities, because a
5667 partial match result is given only when there are no completed matches.
5668 This means that as soon as the shortest match has been found, continua-
5669 tion to a new subject segment is no longer possible. Consider again
5670 this pcretest example:
5671
5672 re> /dog(sbody)?/
5673 data> dogsb\P
5674 0: dog
5675 data> do\P\D
5676 Partial match: do
5677 data> gsb\R\P\D
5678 0: g
5679 data> dogsbody\D
5680 0: dogsbody
5681 1: dog
5682
5683 The first data line passes the string "dogsb" to pcre_exec(), setting
5684 the PCRE_PARTIAL_SOFT option. Although the string is a partial match
5685 for "dogsbody", the result is not PCRE_ERROR_PARTIAL, because the
5686 shorter string "dog" is a complete match. Similarly, when the subject
5687 is presented to pcre_dfa_exec() in several parts ("do" and "gsb" being
5688 the first two) the match stops when "dog" has been found, and it is not
5689 possible to continue. On the other hand, if "dogsbody" is presented as
5690 a single string, pcre_dfa_exec() finds both matches.
5691
5692 Because of these problems, it is probably best to use PCRE_PARTIAL_HARD
5693 when matching multi-segment data. The example above then behaves dif-
5694 ferently:
5695
5696 re> /dog(sbody)?/
5697 data> dogsb\P\P
5698 Partial match: dogsb
5699 data> do\P\D
5700 Partial match: do
5701 data> gsb\R\P\P\D
5702 Partial match: gsb
5703
5704
5705 4. Patterns that contain alternatives at the top level which do not all
5706 start with the same pattern item may not work as expected when
5707 pcre_dfa_exec() is used. For example, consider this pattern:
5708
5709 1234|3789
5710
5711 If the first part of the subject is "ABC123", a partial match of the
5712 first alternative is found at offset 3. There is no partial match for
5713 the second alternative, because such a match does not start at the same
5714 point in the subject string. Attempting to continue with the string
5715 "7890" does not yield a match because only those alternatives that
5716 match at one point in the subject are remembered. The problem arises
5717 because the start of the second alternative matches within the first
5718 alternative. There is no problem with anchored patterns or patterns
5719 such as:
5720
5721 1234|ABCD
5722
5723 where no string can be a partial match for both alternatives. This is
5724 not a problem if pcre_exec() is used, because the entire match has to
5725 be rerun each time:
5726
5727 re> /1234|3789/
5728 data> ABC123\P
5729 Partial match: 123
5730 data> 1237890
5731 0: 3789
5732
5733
5734 AUTHOR
5735
5736 Philip Hazel
5737 University Computing Service
5738 Cambridge CB2 3QH, England.
5739
5740
5741 REVISION
5742
5743 Last updated: 31 August 2009
5744 Copyright (c) 1997-2009 University of Cambridge.
5745 ------------------------------------------------------------------------------
5746
5747
5748 PCREPRECOMPILE(3) PCREPRECOMPILE(3)
5749
5750
5751 NAME
5752 PCRE - Perl-compatible regular expressions
5753
5754
5755 SAVING AND RE-USING PRECOMPILED PCRE PATTERNS
5756
5757 If you are running an application that uses a large number of regular
5758 expression patterns, it may be useful to store them in a precompiled
5759 form instead of having to compile them every time the application is
5760 run. If you are not using any private character tables (see the
5761 pcre_maketables() documentation), this is relatively straightforward.
5762 If you are using private tables, it is a little bit more complicated.
5763
5764 If you save compiled patterns to a file, you can copy them to a differ-
5765 ent host and run them there. This works even if the new host has the
5766 opposite endianness to the one on which the patterns were compiled.
5767 There may be a small performance penalty, but it should be insignifi-
5768 cant. However, compiling regular expressions with one version of PCRE
5769 for use with a different version is not guaranteed to work and may
5770 cause crashes.
5771
5772
5773 SAVING A COMPILED PATTERN
5774 The value returned by pcre_compile() points to a single block of memory
5775 that holds the compiled pattern and associated data. You can find the
5776 length of this block in bytes by calling pcre_fullinfo() with an argu-
5777 ment of PCRE_INFO_SIZE. You can then save the data in any appropriate
5778 manner. Here is sample code that compiles a pattern and writes it to a
5779 file. It assumes that the variable fd refers to a file that is open for
5780 output:
5781
5782 int erroroffset, rc, size;
5783 char *error;
5784 pcre *re;
5785
5786 re = pcre_compile("my pattern", 0, &error, &erroroffset, NULL);
5787 if (re == NULL) { ... handle errors ... }
5788 rc = pcre_fullinfo(re, NULL, PCRE_INFO_SIZE, &size);
5789 if (rc < 0) { ... handle errors ... }
5790 rc = fwrite(re, 1, size, fd);
5791 if (rc != size) { ... handle errors ... }
5792
5793 In this example, the bytes that comprise the compiled pattern are
5794 copied exactly. Note that this is binary data that may contain any of
5795 the 256 possible byte values. On systems that make a distinction
5796 between binary and non-binary data, be sure that the file is opened for
5797 binary output.
5798
5799 If you want to write more than one pattern to a file, you will have to
5800 devise a way of separating them. For binary data, preceding each pat-
5801 tern with its length is probably the most straightforward approach.
5802 Another possibility is to write out the data in hexadecimal instead of
5803 binary, one pattern to a line.
5804
5805 Saving compiled patterns in a file is only one possible way of storing
5806 them for later use. They could equally well be saved in a database, or
5807 in the memory of some daemon process that passes them via sockets to
5808 the processes that want them.
5809
5810 If the pattern has been studied, it is also possible to save the study
5811 data in a similar way to the compiled pattern itself. When studying
5812 generates additional information, pcre_study() returns a pointer to a
5813 pcre_extra data block. Its format is defined in the section on matching
5814 a pattern in the pcreapi documentation. The study_data field points to
5815 the binary study data, and this is what you must save (not the
5816 pcre_extra block itself). The length of the study data can be obtained
5817 by calling pcre_fullinfo() with an argument of PCRE_INFO_STUDYSIZE.
5818 Remember to check that pcre_study() did return a non-NULL value before
5819 trying to save the study data.
5820
5821
5822 RE-USING A PRECOMPILED PATTERN
5823
5824 Re-using a precompiled pattern is straightforward. Having reloaded it
5825 into main memory, you pass its pointer to pcre_exec() or
5826 pcre_dfa_exec() in the usual way. This should work even on another
5827 host, and even if that host has the opposite endianness to the one
5828 where the pattern was compiled.
5829
5830 However, if you passed a pointer to custom character tables when the
5831 pattern was compiled (the tableptr argument of pcre_compile()), you
5832 must now pass a similar pointer to pcre_exec() or pcre_dfa_exec(),
5833 because the value saved with the compiled pattern will obviously be
5834 nonsense. A field in a pcre_extra() block is used to pass this data, as
5835 described in the section on matching a pattern in the pcreapi documen-
5836 tation.
5837
5838 If you did not provide custom character tables when the pattern was
5839 compiled, the pointer in the compiled pattern is NULL, which causes
5840 pcre_exec() to use PCRE's internal tables. Thus, you do not need to
5841 take any special action at run time in this case.
5842
5843 If you saved study data with the compiled pattern, you need to create
5844 your own pcre_extra data block and set the study_data field to point to
5845 the reloaded study data. You must also set the PCRE_EXTRA_STUDY_DATA
5846 bit in the flags field to indicate that study data is present. Then
5847 pass the pcre_extra block to pcre_exec() or pcre_dfa_exec() in the
5848 usual way.
5849
5850
5851 COMPATIBILITY WITH DIFFERENT PCRE RELEASES
5852
5853 In general, it is safest to recompile all saved patterns when you
5854 update to a new PCRE release, though not all updates actually require
5855 this. Recompiling is definitely needed for release 7.2.
5856
5857
5858 AUTHOR
5859
5860 Philip Hazel
5861 University Computing Service
5862 Cambridge CB2 3QH, England.
5863
5864
5865 REVISION
5866
5867 Last updated: 13 June 2007
5868 Copyright (c) 1997-2007 University of Cambridge.
5869 ------------------------------------------------------------------------------
5870
5871
5872 PCREPERFORM(3) PCREPERFORM(3)
5873
5874
5875 NAME
5876 PCRE - Perl-compatible regular expressions
5877
5878
5879 PCRE PERFORMANCE
5880
5881 Two aspects of performance are discussed below: memory usage and pro-
5882 cessing time. The way you express your pattern as a regular expression
5883 can affect both of them.
5884
5885
5886 MEMORY USAGE
5887
5888 Patterns are compiled by PCRE into a reasonably efficient byte code, so
5889 that most simple patterns do not use much memory. However, there is one
5890 case where memory usage can be unexpectedly large. When a parenthesized
5891 subpattern has a quantifier with a minimum greater than 1 and/or a lim-
5892 ited maximum, the whole subpattern is repeated in the compiled code.
5893 For example, the pattern
5894
5895 (abc|def){2,4}
5896
5897 is compiled as if it were
5898
5899 (abc|def)(abc|def)((abc|def)(abc|def)?)?
5900
5901 (Technical aside: It is done this way so that backtrack points within
5902 each of the repetitions can be independently maintained.)
5903
5904 For regular expressions whose quantifiers use only small numbers, this
5905 is not usually a problem. However, if the numbers are large, and par-
5906 ticularly if such repetitions are nested, the memory usage can become
5907 an embarrassment. For example, the very simple pattern
5908
5909 ((ab){1,1000}c){1,3}
5910
5911 uses 51K bytes when compiled. When PCRE is compiled with its default
5912 internal pointer size of two bytes, the size limit on a compiled pat-
5913 tern is 64K, and this is reached with the above pattern if the outer
5914 repetition is increased from 3 to 4. PCRE can be compiled to use larger
5915 internal pointers and thus handle larger compiled patterns, but it is
5916 better to try to rewrite your pattern to use less memory if you can.
5917
5918 One way of reducing the memory usage for such patterns is to make use
5919 of PCRE's "subroutine" facility. Re-writing the above pattern as
5920
5921 ((ab)(?2){0,999}c)(?1){0,2}
5922
5923 reduces the memory requirements to 18K, and indeed it remains under 20K
5924 even with the outer repetition increased to 100. However, this pattern
5925 is not exactly equivalent, because the "subroutine" calls are treated
5926 as atomic groups into which there can be no backtracking if there is a
5927 subsequent matching failure. Therefore, PCRE cannot do this kind of
5928 rewriting automatically. Furthermore, there is a noticeable loss of
5929 speed when executing the modified pattern. Nevertheless, if the atomic
5930 grouping is not a problem and the loss of speed is acceptable, this
5931 kind of rewriting will allow you to process patterns that PCRE cannot
5932 otherwise handle.
5933
5934
5935 PROCESSING TIME
5936
5937 Certain items in regular expression patterns are processed more effi-
5938 ciently than others. It is more efficient to use a character class like
5939 [aeiou] than a set of single-character alternatives such as
5940 (a|e|i|o|u). In general, the simplest construction that provides the
5941 required behaviour is usually the most efficient. Jeffrey Friedl's book
5942 contains a lot of useful general discussion about optimizing regular
5943 expressions for efficient performance. This document contains a few
5944 observations about PCRE.
5945
5946 Using Unicode character properties (the \p, \P, and \X escapes) is
5947 slow, because PCRE has to scan a structure that contains data for over
5948 fifteen thousand characters whenever it needs a character's property.
5949 If you can find an alternative pattern that does not use character
5950 properties, it will probably be faster.
5951
5952 When a pattern begins with .* not in parentheses, or in parentheses
5953 that are not the subject of a backreference, and the PCRE_DOTALL option
5954 is set, the pattern is implicitly anchored by PCRE, since it can match
5955 only at the start of a subject string. However, if PCRE_DOTALL is not
5956 set, PCRE cannot make this optimization, because the . metacharacter
5957 does not then match a newline, and if the subject string contains new-
5958 lines, the pattern may match from the character immediately following
5959 one of them instead of from the very start. For example, the pattern
5960
5961 .*second
5962
5963 matches the subject "first\nand second" (where \n stands for a newline
5964 character), with the match starting at the seventh character. In order
5965 to do this, PCRE has to retry the match starting after every newline in
5966 the subject.
5967
5968 If you are using such a pattern with subject strings that do not con-
5969 tain newlines, the best performance is obtained by setting PCRE_DOTALL,
5970 or starting the pattern with ^.* or ^.*? to indicate explicit anchor-
5971 ing. That saves PCRE from having to scan along the subject looking for
5972 a newline to restart at.
5973
5974 Beware of patterns that contain nested indefinite repeats. These can
5975 take a long time to run when applied to a string that does not match.
5976 Consider the pattern fragment
5977
5978 ^(a+)*
5979
5980 This can match "aaaa" in 16 different ways, and this number increases
5981 very rapidly as the string gets longer. (The * repeat can match 0, 1,
5982 2, 3, or 4 times, and for each of those cases other than 0 or 4, the +
5983 repeats can match different numbers of times.) When the remainder of
5984 the pattern is such that the entire match is going to fail, PCRE has in
5985 principle to try every possible variation, and this can take an
5986 extremely long time, even for relatively short strings.
5987
5988 An optimization catches some of the more simple cases such as
5989
5990 (a+)*b
5991
5992 where a literal character follows. Before embarking on the standard
5993 matching procedure, PCRE checks that there is a "b" later in the sub-
5994 ject string, and if there is not, it fails the match immediately. How-
5995 ever, when there is no following literal this optimization cannot be
5996 used. You can see the difference by comparing the behaviour of
5997
5998 (a+)*\d
5999
6000 with the pattern above. The former gives a failure almost instantly
6001 when applied to a whole line of "a" characters, whereas the latter
6002 takes an appreciable time with strings longer than about 20 characters.
6003
6004 In many cases, the solution to this kind of performance issue is to use
6005 an atomic group or a possessive quantifier.
6006
6007
6008 AUTHOR
6009
6010 Philip Hazel
6011 University Computing Service
6012 Cambridge CB2 3QH, England.
6013
6014
6015 REVISION
6016
6017 Last updated: 06 March 2007
6018 Copyright (c) 1997-2007 University of Cambridge.
6019 ------------------------------------------------------------------------------
6020
6021
6022 PCREPOSIX(3) PCREPOSIX(3)
6023
6024
6025 NAME
6026 PCRE - Perl-compatible regular expressions.
6027
6028
6029 SYNOPSIS OF POSIX API
6030
6031 #include <pcreposix.h>
6032
6033 int regcomp(regex_t *preg, const char *pattern,
6034 int cflags);
6035
6036 int regexec(regex_t *preg, const char *string,
6037 size_t nmatch, regmatch_t pmatch[], int eflags);
6038
6039 size_t regerror(int errcode, const regex_t *preg,
6040 char *errbuf, size_t errbuf_size);
6041
6042 void regfree(regex_t *preg);
6043
6044
6045 DESCRIPTION
6046
6047 This set of functions provides a POSIX-style API to the PCRE regular
6048 expression package. See the pcreapi documentation for a description of
6049 PCRE's native API, which contains much additional functionality.
6050
6051 The functions described here are just wrapper functions that ultimately
6052 call the PCRE native API. Their prototypes are defined in the
6053 pcreposix.h header file, and on Unix systems the library itself is
6054 called pcreposix.a, so can be accessed by adding -lpcreposix to the
6055 command for linking an application that uses them. Because the POSIX
6056 functions call the native ones, it is also necessary to add -lpcre.
6057
6058 I have implemented only those POSIX option bits that can be reasonably
6059 mapped to PCRE native options. In addition, the option REG_EXTENDED is
6060 defined with the value zero. This has no effect, but since programs
6061 that are written to the POSIX interface often use it, this makes it
6062 easier to slot in PCRE as a replacement library. Other POSIX options
6063 are not even defined.
6064
6065 When PCRE is called via these functions, it is only the API that is
6066 POSIX-like in style. The syntax and semantics of the regular expres-
6067 sions themselves are still those of Perl, subject to the setting of
6068 various PCRE options, as described below. "POSIX-like in style" means
6069 that the API approximates to the POSIX definition; it is not fully
6070 POSIX-compatible, and in multi-byte encoding domains it is probably
6071 even less compatible.
6072
6073 The header for these functions is supplied as pcreposix.h to avoid any
6074 potential clash with other POSIX libraries. It can, of course, be
6075 renamed or aliased as regex.h, which is the "correct" name. It provides
6076 two structure types, regex_t for compiled internal forms, and reg-
6077 match_t for returning captured substrings. It also defines some con-
6078 stants whose names start with "REG_"; these are used for setting
6079 options and identifying error codes.
6080
6081
6082 COMPILING A PATTERN
6083
6084 The function regcomp() is called to compile a pattern into an internal
6085 form. The pattern is a C string terminated by a binary zero, and is
6086 passed in the argument pattern. The preg argument is a pointer to a
6087 regex_t structure that is used as a base for storing information about
6088 the compiled regular expression.
6089