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
8 -----------------------------------------------------------------------------
11 PCRE(3) Library Functions Manual PCRE(3)
16 PCRE - Perl-compatible regular expressions (original API)
20 This document relates to PCRE releases that use the original API, with
21 library names libpcre, libpcre16, and libpcre32. January 2015 saw the
22 first release of a new API, known as PCRE2, with release numbers start-
23 ing at 10.00 and library names libpcre2-8, libpcre2-16, and
24 libpcre2-32. The old libraries (now called PCRE1) are still being main-
25 tained for bug fixes, but there will be no new development. New
26 projects are advised to use the new PCRE2 libraries.
31 The PCRE library is a set of functions that implement regular expres-
32 sion pattern matching using the same syntax and semantics as Perl, with
33 just a few differences. Some features that appeared in Python and PCRE
34 before they appeared in Perl are also available using the Python syn-
35 tax, there is some support for one or two .NET and Oniguruma syntax
36 items, and there is an option for requesting some minor changes that
37 give better JavaScript compatibility.
39 Starting with release 8.30, it is possible to compile two separate PCRE
40 libraries: the original, which supports 8-bit character strings
41 (including UTF-8 strings), and a second library that supports 16-bit
42 character strings (including UTF-16 strings). The build process allows
43 either one or both to be built. The majority of the work to make this
44 possible was done by Zoltan Herczeg.
46 Starting with release 8.32 it is possible to compile a third separate
47 PCRE library that supports 32-bit character strings (including UTF-32
48 strings). The build process allows any combination of the 8-, 16- and
49 32-bit libraries. The work to make this possible was done by Christian
52 The three libraries contain identical sets of functions, except that
53 the names in the 16-bit library start with pcre16_ instead of pcre_,
54 and the names in the 32-bit library start with pcre32_ instead of
55 pcre_. To avoid over-complication and reduce the documentation mainte-
56 nance load, most of the documentation describes the 8-bit library, with
57 the differences for the 16-bit and 32-bit libraries described sepa-
58 rately in the pcre16 and pcre32 pages. References to functions or
59 structures of the form pcre[16|32]_xxx should be read as meaning
60 "pcre_xxx when using the 8-bit library, pcre16_xxx when using the
61 16-bit library, or pcre32_xxx when using the 32-bit library".
63 The current implementation of PCRE corresponds approximately with Perl
64 5.12, including support for UTF-8/16/32 encoded strings and Unicode
65 general category properties. However, UTF-8/16/32 and Unicode support
66 has to be explicitly enabled; it is not the default. The Unicode tables
67 correspond to Unicode release 6.3.0.
69 In addition to the Perl-compatible matching function, PCRE contains an
70 alternative function that matches the same compiled patterns in a dif-
71 ferent way. In certain circumstances, the alternative function has some
72 advantages. For a discussion of the two matching algorithms, see the
75 PCRE is written in C and released as a C library. A number of people
76 have written wrappers and interfaces of various kinds. In particular,
77 Google Inc. have provided a comprehensive C++ wrapper for the 8-bit
78 library. This is now included as part of the PCRE distribution. The
79 pcrecpp page has details of this interface. Other people's contribu-
80 tions can be found in the Contrib directory at the primary FTP site,
83 ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre
85 Details of exactly which Perl regular expression features are and are
86 not supported by PCRE are given in separate documents. See the pcrepat-
87 tern and pcrecompat pages. There is a syntax summary in the pcresyntax
90 Some features of PCRE can be included, excluded, or changed when the
91 library is built. The pcre_config() function makes it possible for a
92 client to discover which features are available. The features them-
93 selves are described in the pcrebuild page. Documentation about build-
94 ing PCRE for various operating systems can be found in the README and
95 NON-AUTOTOOLS_BUILD files in the source distribution.
97 The libraries contains a number of undocumented internal functions and
98 data tables that are used by more than one of the exported external
99 functions, but which are not intended for use by external callers.
100 Their names all begin with "_pcre_" or "_pcre16_" or "_pcre32_", which
101 hopefully will not provoke any name clashes. In some environments, it
102 is possible to control which external symbols are exported when a
103 shared library is built, and in these cases the undocumented symbols
107 SECURITY CONSIDERATIONS
109 If you are using PCRE in a non-UTF application that permits users to
110 supply arbitrary patterns for compilation, you should be aware of a
111 feature that allows users to turn on UTF support from within a pattern,
112 provided that PCRE was built with UTF support. For example, an 8-bit
113 pattern that begins with "(*UTF8)" or "(*UTF)" turns on UTF-8 mode,
114 which interprets patterns and subjects as strings of UTF-8 characters
115 instead of individual 8-bit characters. This causes both the pattern
116 and any data against which it is matched to be checked for UTF-8 valid-
117 ity. If the data string is very long, such a check might use suffi-
118 ciently many resources as to cause your application to lose perfor-
121 One way of guarding against this possibility is to use the
122 pcre_fullinfo() function to check the compiled pattern's options for
123 UTF. Alternatively, from release 8.33, you can set the PCRE_NEVER_UTF
124 option at compile time. This causes an compile time error if a pattern
125 contains a UTF-setting sequence.
127 If your application is one that supports UTF, be aware that validity
128 checking can take time. If the same data string is to be matched many
129 times, you can use the PCRE_NO_UTF[8|16|32]_CHECK option for the second
130 and subsequent matches to save redundant checks.
132 Another way that performance can be hit is by running a pattern that
133 has a very large search tree against a string that will never match.
134 Nested unlimited repeats in a pattern are a common example. PCRE pro-
135 vides some protection against this: see the PCRE_EXTRA_MATCH_LIMIT fea-
136 ture in the pcreapi page.
141 The user documentation for PCRE comprises a number of different sec-
142 tions. In the "man" format, each of these is a separate "man page". In
143 the HTML format, each is a separate page, linked from the index page.
144 In the plain text format, the descriptions of the pcregrep and pcretest
145 programs are in files called pcregrep.txt and pcretest.txt, respec-
146 tively. The remaining sections, except for the pcredemo section (which
147 is a program listing), are concatenated in pcre.txt, for ease of
148 searching. The sections are as follows:
151 pcre-config show PCRE installation configuration information
152 pcre16 details of the 16-bit library
153 pcre32 details of the 32-bit library
154 pcreapi details of PCRE's native C API
155 pcrebuild building PCRE
156 pcrecallout details of the callout feature
157 pcrecompat discussion of Perl compatibility
158 pcrecpp details of the C++ wrapper for the 8-bit library
159 pcredemo a demonstration C program that uses PCRE
160 pcregrep description of the pcregrep command (8-bit only)
161 pcrejit discussion of the just-in-time optimization support
162 pcrelimits details of size and other limits
163 pcrematching discussion of the two matching algorithms
164 pcrepartial details of the partial matching facility
165 pcrepattern syntax and semantics of supported
167 pcreperform discussion of performance issues
168 pcreposix the POSIX-compatible C API for the 8-bit library
169 pcreprecompile details of saving and re-using precompiled patterns
170 pcresample discussion of the pcredemo program
171 pcrestack discussion of stack usage
172 pcresyntax quick syntax reference
173 pcretest description of the pcretest testing command
174 pcreunicode discussion of Unicode and UTF-8/16/32 support
176 In the "man" and HTML formats, there is also a short page for each C
177 library function, listing its arguments and results.
183 University Computing Service
184 Cambridge CB2 3QH, England.
186 Putting an actual email address here seems to have been a spam magnet,
187 so I've taken it away. If you want to email me, use my two initials,
188 followed by the two digits 10, at the domain cam.ac.uk.
193 Last updated: 10 February 2015
194 Copyright (c) 1997-2015 University of Cambridge.
195 ------------------------------------------------------------------------------
198 PCRE(3) Library Functions Manual PCRE(3)
203 PCRE - Perl-compatible regular expressions
208 PCRE 16-BIT API BASIC FUNCTIONS
210 pcre16 *pcre16_compile(PCRE_SPTR16 pattern, int options,
211 const char **errptr, int *erroffset,
212 const unsigned char *tableptr);
214 pcre16 *pcre16_compile2(PCRE_SPTR16 pattern, int options,
216 const char **errptr, int *erroffset,
217 const unsigned char *tableptr);
219 pcre16_extra *pcre16_study(const pcre16 *code, int options,
220 const char **errptr);
222 void pcre16_free_study(pcre16_extra *extra);
224 int pcre16_exec(const pcre16 *code, const pcre16_extra *extra,
225 PCRE_SPTR16 subject, int length, int startoffset,
226 int options, int *ovector, int ovecsize);
228 int pcre16_dfa_exec(const pcre16 *code, const pcre16_extra *extra,
229 PCRE_SPTR16 subject, int length, int startoffset,
230 int options, int *ovector, int ovecsize,
231 int *workspace, int wscount);
234 PCRE 16-BIT API STRING EXTRACTION FUNCTIONS
236 int pcre16_copy_named_substring(const pcre16 *code,
237 PCRE_SPTR16 subject, int *ovector,
238 int stringcount, PCRE_SPTR16 stringname,
239 PCRE_UCHAR16 *buffer, int buffersize);
241 int pcre16_copy_substring(PCRE_SPTR16 subject, int *ovector,
242 int stringcount, int stringnumber, PCRE_UCHAR16 *buffer,
245 int pcre16_get_named_substring(const pcre16 *code,
246 PCRE_SPTR16 subject, int *ovector,
247 int stringcount, PCRE_SPTR16 stringname,
248 PCRE_SPTR16 *stringptr);
250 int pcre16_get_stringnumber(const pcre16 *code,
253 int pcre16_get_stringtable_entries(const pcre16 *code,
254 PCRE_SPTR16 name, PCRE_UCHAR16 **first, PCRE_UCHAR16 **last);
256 int pcre16_get_substring(PCRE_SPTR16 subject, int *ovector,
257 int stringcount, int stringnumber,
258 PCRE_SPTR16 *stringptr);
260 int pcre16_get_substring_list(PCRE_SPTR16 subject,
261 int *ovector, int stringcount, PCRE_SPTR16 **listptr);
263 void pcre16_free_substring(PCRE_SPTR16 stringptr);
265 void pcre16_free_substring_list(PCRE_SPTR16 *stringptr);
268 PCRE 16-BIT API AUXILIARY FUNCTIONS
270 pcre16_jit_stack *pcre16_jit_stack_alloc(int startsize, int maxsize);
272 void pcre16_jit_stack_free(pcre16_jit_stack *stack);
274 void pcre16_assign_jit_stack(pcre16_extra *extra,
275 pcre16_jit_callback callback, void *data);
277 const unsigned char *pcre16_maketables(void);
279 int pcre16_fullinfo(const pcre16 *code, const pcre16_extra *extra,
280 int what, void *where);
282 int pcre16_refcount(pcre16 *code, int adjust);
284 int pcre16_config(int what, void *where);
286 const char *pcre16_version(void);
288 int pcre16_pattern_to_host_byte_order(pcre16 *code,
289 pcre16_extra *extra, const unsigned char *tables);
292 PCRE 16-BIT API INDIRECTED FUNCTIONS
294 void *(*pcre16_malloc)(size_t);
296 void (*pcre16_free)(void *);
298 void *(*pcre16_stack_malloc)(size_t);
300 void (*pcre16_stack_free)(void *);
302 int (*pcre16_callout)(pcre16_callout_block *);
305 PCRE 16-BIT API 16-BIT-ONLY FUNCTION
307 int pcre16_utf16_to_host_byte_order(PCRE_UCHAR16 *output,
308 PCRE_SPTR16 input, int length, int *byte_order,
312 THE PCRE 16-BIT LIBRARY
314 Starting with release 8.30, it is possible to compile a PCRE library
315 that supports 16-bit character strings, including UTF-16 strings, as
316 well as or instead of the original 8-bit library. The majority of the
317 work to make this possible was done by Zoltan Herczeg. The two
318 libraries contain identical sets of functions, used in exactly the same
319 way. Only the names of the functions and the data types of their argu-
320 ments and results are different. To avoid over-complication and reduce
321 the documentation maintenance load, most of the PCRE documentation
322 describes the 8-bit library, with only occasional references to the
323 16-bit library. This page describes what is different when you use the
326 WARNING: A single application can be linked with both libraries, but
327 you must take care when processing any particular pattern to use func-
328 tions from just one library. For example, if you want to study a pat-
329 tern that was compiled with pcre16_compile(), you must do so with
330 pcre16_study(), not pcre_study(), and you must free the study data with
336 There is only one header file, pcre.h. It contains prototypes for all
337 the functions in all libraries, as well as definitions of flags, struc-
338 tures, error codes, etc.
343 In Unix-like systems, the 16-bit library is called libpcre16, and can
344 normally be accesss by adding -lpcre16 to the command for linking an
345 application that uses PCRE.
350 In the 8-bit library, strings are passed to PCRE library functions as
351 vectors of bytes with the C type "char *". In the 16-bit library,
352 strings are passed as vectors of unsigned 16-bit quantities. The macro
353 PCRE_UCHAR16 specifies an appropriate data type, and PCRE_SPTR16 is
354 defined as "const PCRE_UCHAR16 *". In very many environments, "short
355 int" is a 16-bit data type. When PCRE is built, it defines PCRE_UCHAR16
356 as "unsigned short int", but checks that it really is a 16-bit data
357 type. If it is not, the build fails with an error message telling the
358 maintainer to modify the definition appropriately.
363 The types of the opaque structures that are used for compiled 16-bit
364 patterns and JIT stacks are pcre16 and pcre16_jit_stack respectively.
365 The type of the user-accessible structure that is returned by
366 pcre16_study() is pcre16_extra, and the type of the structure that is
367 used for passing data to a callout function is pcre16_callout_block.
368 These structures contain the same fields, with the same names, as their
369 8-bit counterparts. The only difference is that pointers to character
370 strings are 16-bit instead of 8-bit types.
375 For every function in the 8-bit library there is a corresponding func-
376 tion in the 16-bit library with a name that starts with pcre16_ instead
377 of pcre_. The prototypes are listed above. In addition, there is one
378 extra function, pcre16_utf16_to_host_byte_order(). This is a utility
379 function that converts a UTF-16 character string to host byte order if
380 necessary. The other 16-bit functions expect the strings they are
381 passed to be in host byte order.
383 The input and output arguments of pcre16_utf16_to_host_byte_order() may
384 point to the same address, that is, conversion in place is supported.
385 The output buffer must be at least as long as the input.
387 The length argument specifies the number of 16-bit data units in the
388 input string; a negative value specifies a zero-terminated string.
390 If byte_order is NULL, it is assumed that the string starts off in host
391 byte order. This may be changed by byte-order marks (BOMs) anywhere in
392 the string (commonly as the first character).
394 If byte_order is not NULL, a non-zero value of the integer to which it
395 points means that the input starts off in host byte order, otherwise
396 the opposite order is assumed. Again, BOMs in the string can change
397 this. The final byte order is passed back at the end of processing.
399 If keep_boms is not zero, byte-order mark characters (0xfeff) are
400 copied into the output string. Otherwise they are discarded.
402 The result of the function is the number of 16-bit units placed into
403 the output buffer, including the zero terminator if the string was
407 SUBJECT STRING OFFSETS
409 The lengths and starting offsets of subject strings must be specified
410 in 16-bit data units, and the offsets within subject strings that are
411 returned by the matching functions are in also 16-bit units rather than
417 The name-to-number translation table that is maintained for named sub-
418 patterns uses 16-bit characters. The pcre16_get_stringtable_entries()
419 function returns the length of each entry in the table as the number of
425 There are two new general option names, PCRE_UTF16 and
426 PCRE_NO_UTF16_CHECK, which correspond to PCRE_UTF8 and
427 PCRE_NO_UTF8_CHECK in the 8-bit library. In fact, these new options
428 define the same bits in the options word. There is a discussion about
429 the validity of UTF-16 strings in the pcreunicode page.
431 For the pcre16_config() function there is an option PCRE_CONFIG_UTF16
432 that returns 1 if UTF-16 support is configured, otherwise 0. If this
433 option is given to pcre_config() or pcre32_config(), or if the
434 PCRE_CONFIG_UTF8 or PCRE_CONFIG_UTF32 option is given to pcre16_con-
435 fig(), the result is the PCRE_ERROR_BADOPTION error.
440 In 16-bit mode, when PCRE_UTF16 is not set, character values are
441 treated in the same way as in 8-bit, non UTF-8 mode, except, of course,
442 that they can range from 0 to 0xffff instead of 0 to 0xff. Character
443 types for characters less than 0xff can therefore be influenced by the
444 locale in the same way as before. Characters greater than 0xff have
445 only one case, and no "type" (such as letter or digit).
447 In UTF-16 mode, the character code is Unicode, in the range 0 to
448 0x10ffff, with the exception of values in the range 0xd800 to 0xdfff
449 because those are "surrogate" values that are used in pairs to encode
450 values greater than 0xffff.
452 A UTF-16 string can indicate its endianness by special code knows as a
453 byte-order mark (BOM). The PCRE functions do not handle this, expecting
454 strings to be in host byte order. A utility function called
455 pcre16_utf16_to_host_byte_order() is provided to help with this (see
461 The errors PCRE_ERROR_BADUTF16_OFFSET and PCRE_ERROR_SHORTUTF16 corre-
462 spond to their 8-bit counterparts. The error PCRE_ERROR_BADMODE is
463 given when a compiled pattern is passed to a function that processes
464 patterns in the other mode, for example, if a pattern compiled with
465 pcre_compile() is passed to pcre16_exec().
467 There are new error codes whose names begin with PCRE_UTF16_ERR for
468 invalid UTF-16 strings, corresponding to the PCRE_UTF8_ERR codes for
469 UTF-8 strings that are described in the section entitled "Reason codes
470 for invalid UTF-8 strings" in the main pcreapi page. The UTF-16 errors
473 PCRE_UTF16_ERR1 Missing low surrogate at end of string
474 PCRE_UTF16_ERR2 Invalid low surrogate follows high surrogate
475 PCRE_UTF16_ERR3 Isolated low surrogate
476 PCRE_UTF16_ERR4 Non-character
481 If there is an error while compiling a pattern, the error text that is
482 passed back by pcre16_compile() or pcre16_compile2() is still an 8-bit
483 character string, zero-terminated.
488 The subject and mark fields in the callout block that is passed to a
489 callout function point to 16-bit vectors.
494 The pcretest program continues to operate with 8-bit input and output
495 files, but it can be used for testing the 16-bit library. If it is run
496 with the command line option -16, patterns and subject strings are con-
497 verted from 8-bit to 16-bit before being passed to PCRE, and the 16-bit
498 library functions are used instead of the 8-bit ones. Returned 16-bit
499 strings are converted to 8-bit for output. If both the 8-bit and the
500 32-bit libraries were not compiled, pcretest defaults to 16-bit and the
501 -16 option is ignored.
503 When PCRE is being built, the RunTest script that is called by "make
504 check" uses the pcretest -C option to discover which of the 8-bit,
505 16-bit and 32-bit libraries has been built, and runs the tests appro-
509 NOT SUPPORTED IN 16-BIT MODE
511 Not all the features of the 8-bit library are available with the 16-bit
512 library. The C++ and POSIX wrapper functions support only the 8-bit
513 library, and the pcregrep program is at present 8-bit only.
519 University Computing Service
520 Cambridge CB2 3QH, England.
525 Last updated: 12 May 2013
526 Copyright (c) 1997-2013 University of Cambridge.
527 ------------------------------------------------------------------------------
530 PCRE(3) Library Functions Manual PCRE(3)
535 PCRE - Perl-compatible regular expressions
540 PCRE 32-BIT API BASIC FUNCTIONS
542 pcre32 *pcre32_compile(PCRE_SPTR32 pattern, int options,
543 const char **errptr, int *erroffset,
544 const unsigned char *tableptr);
546 pcre32 *pcre32_compile2(PCRE_SPTR32 pattern, int options,
548 const unsigned char *tableptr);
550 pcre32_extra *pcre32_study(const pcre32 *code, int options,
551 const char **errptr);
553 void pcre32_free_study(pcre32_extra *extra);
555 int pcre32_exec(const pcre32 *code, const pcre32_extra *extra,
556 PCRE_SPTR32 subject, int length, int startoffset,
557 int options, int *ovector, int ovecsize);
559 int pcre32_dfa_exec(const pcre32 *code, const pcre32_extra *extra,
560 PCRE_SPTR32 subject, int length, int startoffset,
561 int options, int *ovector, int ovecsize,
562 int *workspace, int wscount);
565 PCRE 32-BIT API STRING EXTRACTION FUNCTIONS
567 int pcre32_copy_named_substring(const pcre32 *code,
568 PCRE_SPTR32 subject, int *ovector,
569 int stringcount, PCRE_SPTR32 stringname,
570 PCRE_UCHAR32 *buffer, int buffersize);
572 int pcre32_copy_substring(PCRE_SPTR32 subject, int *ovector,
573 int stringcount, int stringnumber, PCRE_UCHAR32 *buffer,
576 int pcre32_get_named_substring(const pcre32 *code,
577 PCRE_SPTR32 subject, int *ovector,
578 int stringcount, PCRE_SPTR32 stringname,
579 PCRE_SPTR32 *stringptr);
581 int pcre32_get_stringnumber(const pcre32 *code,
584 int pcre32_get_stringtable_entries(const pcre32 *code,
585 PCRE_SPTR32 name, PCRE_UCHAR32 **first, PCRE_UCHAR32 **last);
587 int pcre32_get_substring(PCRE_SPTR32 subject, int *ovector,
588 int stringcount, int stringnumber,
589 PCRE_SPTR32 *stringptr);
591 int pcre32_get_substring_list(PCRE_SPTR32 subject,
592 int *ovector, int stringcount, PCRE_SPTR32 **listptr);
594 void pcre32_free_substring(PCRE_SPTR32 stringptr);
596 void pcre32_free_substring_list(PCRE_SPTR32 *stringptr);
599 PCRE 32-BIT API AUXILIARY FUNCTIONS
601 pcre32_jit_stack *pcre32_jit_stack_alloc(int startsize, int maxsize);
603 void pcre32_jit_stack_free(pcre32_jit_stack *stack);
605 void pcre32_assign_jit_stack(pcre32_extra *extra,
606 pcre32_jit_callback callback, void *data);
608 const unsigned char *pcre32_maketables(void);
610 int pcre32_fullinfo(const pcre32 *code, const pcre32_extra *extra,
611 int what, void *where);
613 int pcre32_refcount(pcre32 *code, int adjust);
615 int pcre32_config(int what, void *where);
617 const char *pcre32_version(void);
619 int pcre32_pattern_to_host_byte_order(pcre32 *code,
620 pcre32_extra *extra, const unsigned char *tables);
623 PCRE 32-BIT API INDIRECTED FUNCTIONS
625 void *(*pcre32_malloc)(size_t);
627 void (*pcre32_free)(void *);
629 void *(*pcre32_stack_malloc)(size_t);
631 void (*pcre32_stack_free)(void *);
633 int (*pcre32_callout)(pcre32_callout_block *);
636 PCRE 32-BIT API 32-BIT-ONLY FUNCTION
638 int pcre32_utf32_to_host_byte_order(PCRE_UCHAR32 *output,
639 PCRE_SPTR32 input, int length, int *byte_order,
643 THE PCRE 32-BIT LIBRARY
645 Starting with release 8.32, it is possible to compile a PCRE library
646 that supports 32-bit character strings, including UTF-32 strings, as
647 well as or instead of the original 8-bit library. This work was done by
648 Christian Persch, based on the work done by Zoltan Herczeg for the
649 16-bit library. All three libraries contain identical sets of func-
650 tions, used in exactly the same way. Only the names of the functions
651 and the data types of their arguments and results are different. To
652 avoid over-complication and reduce the documentation maintenance load,
653 most of the PCRE documentation describes the 8-bit library, with only
654 occasional references to the 16-bit and 32-bit libraries. This page
655 describes what is different when you use the 32-bit library.
657 WARNING: A single application can be linked with all or any of the
658 three libraries, but you must take care when processing any particular
659 pattern to use functions from just one library. For example, if you
660 want to study a pattern that was compiled with pcre32_compile(), you
661 must do so with pcre32_study(), not pcre_study(), and you must free the
662 study data with pcre32_free_study().
667 There is only one header file, pcre.h. It contains prototypes for all
668 the functions in all libraries, as well as definitions of flags, struc-
669 tures, error codes, etc.
674 In Unix-like systems, the 32-bit library is called libpcre32, and can
675 normally be accesss by adding -lpcre32 to the command for linking an
676 application that uses PCRE.
681 In the 8-bit library, strings are passed to PCRE library functions as
682 vectors of bytes with the C type "char *". In the 32-bit library,
683 strings are passed as vectors of unsigned 32-bit quantities. The macro
684 PCRE_UCHAR32 specifies an appropriate data type, and PCRE_SPTR32 is
685 defined as "const PCRE_UCHAR32 *". In very many environments, "unsigned
686 int" is a 32-bit data type. When PCRE is built, it defines PCRE_UCHAR32
687 as "unsigned int", but checks that it really is a 32-bit data type. If
688 it is not, the build fails with an error message telling the maintainer
689 to modify the definition appropriately.
694 The types of the opaque structures that are used for compiled 32-bit
695 patterns and JIT stacks are pcre32 and pcre32_jit_stack respectively.
696 The type of the user-accessible structure that is returned by
697 pcre32_study() is pcre32_extra, and the type of the structure that is
698 used for passing data to a callout function is pcre32_callout_block.
699 These structures contain the same fields, with the same names, as their
700 8-bit counterparts. The only difference is that pointers to character
701 strings are 32-bit instead of 8-bit types.
706 For every function in the 8-bit library there is a corresponding func-
707 tion in the 32-bit library with a name that starts with pcre32_ instead
708 of pcre_. The prototypes are listed above. In addition, there is one
709 extra function, pcre32_utf32_to_host_byte_order(). This is a utility
710 function that converts a UTF-32 character string to host byte order if
711 necessary. The other 32-bit functions expect the strings they are
712 passed to be in host byte order.
714 The input and output arguments of pcre32_utf32_to_host_byte_order() may
715 point to the same address, that is, conversion in place is supported.
716 The output buffer must be at least as long as the input.
718 The length argument specifies the number of 32-bit data units in the
719 input string; a negative value specifies a zero-terminated string.
721 If byte_order is NULL, it is assumed that the string starts off in host
722 byte order. This may be changed by byte-order marks (BOMs) anywhere in
723 the string (commonly as the first character).
725 If byte_order is not NULL, a non-zero value of the integer to which it
726 points means that the input starts off in host byte order, otherwise
727 the opposite order is assumed. Again, BOMs in the string can change
728 this. The final byte order is passed back at the end of processing.
730 If keep_boms is not zero, byte-order mark characters (0xfeff) are
731 copied into the output string. Otherwise they are discarded.
733 The result of the function is the number of 32-bit units placed into
734 the output buffer, including the zero terminator if the string was
738 SUBJECT STRING OFFSETS
740 The lengths and starting offsets of subject strings must be specified
741 in 32-bit data units, and the offsets within subject strings that are
742 returned by the matching functions are in also 32-bit units rather than
748 The name-to-number translation table that is maintained for named sub-
749 patterns uses 32-bit characters. The pcre32_get_stringtable_entries()
750 function returns the length of each entry in the table as the number of
756 There are two new general option names, PCRE_UTF32 and
757 PCRE_NO_UTF32_CHECK, which correspond to PCRE_UTF8 and
758 PCRE_NO_UTF8_CHECK in the 8-bit library. In fact, these new options
759 define the same bits in the options word. There is a discussion about
760 the validity of UTF-32 strings in the pcreunicode page.
762 For the pcre32_config() function there is an option PCRE_CONFIG_UTF32
763 that returns 1 if UTF-32 support is configured, otherwise 0. If this
764 option is given to pcre_config() or pcre16_config(), or if the
765 PCRE_CONFIG_UTF8 or PCRE_CONFIG_UTF16 option is given to pcre32_con-
766 fig(), the result is the PCRE_ERROR_BADOPTION error.
771 In 32-bit mode, when PCRE_UTF32 is not set, character values are
772 treated in the same way as in 8-bit, non UTF-8 mode, except, of course,
773 that they can range from 0 to 0x7fffffff instead of 0 to 0xff. Charac-
774 ter types for characters less than 0xff can therefore be influenced by
775 the locale in the same way as before. Characters greater than 0xff
776 have only one case, and no "type" (such as letter or digit).
778 In UTF-32 mode, the character code is Unicode, in the range 0 to
779 0x10ffff, with the exception of values in the range 0xd800 to 0xdfff
780 because those are "surrogate" values that are ill-formed in UTF-32.
782 A UTF-32 string can indicate its endianness by special code knows as a
783 byte-order mark (BOM). The PCRE functions do not handle this, expecting
784 strings to be in host byte order. A utility function called
785 pcre32_utf32_to_host_byte_order() is provided to help with this (see
791 The error PCRE_ERROR_BADUTF32 corresponds to its 8-bit counterpart.
792 The error PCRE_ERROR_BADMODE is given when a compiled pattern is passed
793 to a function that processes patterns in the other mode, for example,
794 if a pattern compiled with pcre_compile() is passed to pcre32_exec().
796 There are new error codes whose names begin with PCRE_UTF32_ERR for
797 invalid UTF-32 strings, corresponding to the PCRE_UTF8_ERR codes for
798 UTF-8 strings that are described in the section entitled "Reason codes
799 for invalid UTF-8 strings" in the main pcreapi page. The UTF-32 errors
802 PCRE_UTF32_ERR1 Surrogate character (range from 0xd800 to 0xdfff)
803 PCRE_UTF32_ERR2 Non-character
804 PCRE_UTF32_ERR3 Character > 0x10ffff
809 If there is an error while compiling a pattern, the error text that is
810 passed back by pcre32_compile() or pcre32_compile2() is still an 8-bit
811 character string, zero-terminated.
816 The subject and mark fields in the callout block that is passed to a
817 callout function point to 32-bit vectors.
822 The pcretest program continues to operate with 8-bit input and output
823 files, but it can be used for testing the 32-bit library. If it is run
824 with the command line option -32, patterns and subject strings are con-
825 verted from 8-bit to 32-bit before being passed to PCRE, and the 32-bit
826 library functions are used instead of the 8-bit ones. Returned 32-bit
827 strings are converted to 8-bit for output. If both the 8-bit and the
828 16-bit libraries were not compiled, pcretest defaults to 32-bit and the
829 -32 option is ignored.
831 When PCRE is being built, the RunTest script that is called by "make
832 check" uses the pcretest -C option to discover which of the 8-bit,
833 16-bit and 32-bit libraries has been built, and runs the tests appro-
837 NOT SUPPORTED IN 32-BIT MODE
839 Not all the features of the 8-bit library are available with the 32-bit
840 library. The C++ and POSIX wrapper functions support only the 8-bit
841 library, and the pcregrep program is at present 8-bit only.
847 University Computing Service
848 Cambridge CB2 3QH, England.
853 Last updated: 12 May 2013
854 Copyright (c) 1997-2013 University of Cambridge.
855 ------------------------------------------------------------------------------
858 PCREBUILD(3) Library Functions Manual PCREBUILD(3)
863 PCRE - Perl-compatible regular expressions
867 PCRE is distributed with a configure script that can be used to build
868 the library in Unix-like environments using the applications known as
869 Autotools. Also in the distribution are files to support building
870 using CMake instead of configure. The text file README contains general
871 information about building with Autotools (some of which is repeated
872 below), and also has some comments about building on various operating
873 systems. There is a lot more information about building PCRE without
874 using Autotools (including information about using CMake and building
875 "by hand") in the text file called NON-AUTOTOOLS-BUILD. You should
876 consult this file as well as the README file if you are building in a
877 non-Unix-like environment.
880 PCRE BUILD-TIME OPTIONS
882 The rest of this document describes the optional features of PCRE that
883 can be selected when the library is compiled. It assumes use of the
884 configure script, where the optional features are selected or dese-
885 lected by providing options to configure before running the make com-
886 mand. However, the same options can be selected in both Unix-like and
887 non-Unix-like environments using the GUI facility of cmake-gui if you
888 are using CMake instead of configure to build PCRE.
890 If you are not using Autotools or CMake, option selection can be done
891 by editing the config.h file, or by passing parameter settings to the
892 compiler, as described in NON-AUTOTOOLS-BUILD.
894 The complete list of options for configure (which includes the standard
895 ones such as the selection of the installation directory) can be
900 The following sections include descriptions of options whose names
901 begin with --enable or --disable. These settings specify changes to the
902 defaults for the configure command. Because of the way that configure
903 works, --enable and --disable always come in pairs, so the complemen-
904 tary option always exists as well, but as it specifies the default, it
908 BUILDING 8-BIT, 16-BIT AND 32-BIT LIBRARIES
910 By default, a library called libpcre is built, containing functions
911 that take string arguments contained in vectors of bytes, either as
912 single-byte characters, or interpreted as UTF-8 strings. You can also
913 build a separate library, called libpcre16, in which strings are con-
914 tained in vectors of 16-bit data units and interpreted either as sin-
915 gle-unit characters or UTF-16 strings, by adding
919 to the configure command. You can also build yet another separate
920 library, called libpcre32, in which strings are contained in vectors of
921 32-bit data units and interpreted either as single-unit characters or
922 UTF-32 strings, by adding
926 to the configure command. If you do not want the 8-bit library, add
930 as well. At least one of the three libraries must be built. Note that
931 the C++ and POSIX wrappers are for the 8-bit library only, and that
932 pcregrep is an 8-bit program. None of these are built if you select
933 only the 16-bit or 32-bit libraries.
936 BUILDING SHARED AND STATIC LIBRARIES
938 The Autotools PCRE building process uses libtool to build both shared
939 and static libraries by default. You can suppress one of these by
945 to the configure command, as required.
950 By default, if the 8-bit library is being built, the configure script
951 will search for a C++ compiler and C++ header files. If it finds them,
952 it automatically builds the C++ wrapper library (which supports only
953 8-bit strings). You can disable this by adding
957 to the configure command.
960 UTF-8, UTF-16 AND UTF-32 SUPPORT
962 To build PCRE with support for UTF Unicode character strings, add
966 to the configure command. This setting applies to all three libraries,
967 adding support for UTF-8 to the 8-bit library, support for UTF-16 to
968 the 16-bit library, and support for UTF-32 to the to the 32-bit
969 library. There are no separate options for enabling UTF-8, UTF-16 and
970 UTF-32 independently because that would allow ridiculous settings such
971 as requesting UTF-16 support while building only the 8-bit library. It
972 is not possible to build one library with UTF support and another with-
973 out in the same configuration. (For backwards compatibility, --enable-
974 utf8 is a synonym of --enable-utf.)
976 Of itself, this setting does not make PCRE treat strings as UTF-8,
977 UTF-16 or UTF-32. As well as compiling PCRE with this option, you also
978 have have to set the PCRE_UTF8, PCRE_UTF16 or PCRE_UTF32 option (as
979 appropriate) when you call one of the pattern compiling functions.
981 If you set --enable-utf when compiling in an EBCDIC environment, PCRE
982 expects its input to be either ASCII or UTF-8 (depending on the run-
983 time option). It is not possible to support both EBCDIC and UTF-8 codes
984 in the same version of the library. Consequently, --enable-utf and
985 --enable-ebcdic are mutually exclusive.
988 UNICODE CHARACTER PROPERTY SUPPORT
990 UTF support allows the libraries to process character codepoints up to
991 0x10ffff in the strings that they handle. On its own, however, it does
992 not provide any facilities for accessing the properties of such charac-
993 ters. If you want to be able to use the pattern escapes \P, \p, and \X,
994 which refer to Unicode character properties, you must add
996 --enable-unicode-properties
998 to the configure command. This implies UTF support, even if you have
999 not explicitly requested it.
1001 Including Unicode property support adds around 30K of tables to the
1002 PCRE library. Only the general category properties such as Lu and Nd
1003 are supported. Details are given in the pcrepattern documentation.
1006 JUST-IN-TIME COMPILER SUPPORT
1008 Just-in-time compiler support is included in the build by specifying
1012 This support is available only for certain hardware architectures. If
1013 this option is set for an unsupported architecture, a compile time
1014 error occurs. See the pcrejit documentation for a discussion of JIT
1015 usage. When JIT support is enabled, pcregrep automatically makes use of
1018 --disable-pcregrep-jit
1020 to the "configure" command.
1023 CODE VALUE OF NEWLINE
1025 By default, PCRE interprets the linefeed (LF) character as indicating
1026 the end of a line. This is the normal newline character on Unix-like
1027 systems. You can compile PCRE to use carriage return (CR) instead, by
1030 --enable-newline-is-cr
1032 to the configure command. There is also a --enable-newline-is-lf
1033 option, which explicitly specifies linefeed as the newline character.
1035 Alternatively, you can specify that line endings are to be indicated by
1036 the two character sequence CRLF. If you want this, add
1038 --enable-newline-is-crlf
1040 to the configure command. There is a fourth option, specified by
1042 --enable-newline-is-anycrlf
1044 which causes PCRE to recognize any of the three sequences CR, LF, or
1045 CRLF as indicating a line ending. Finally, a fifth option, specified by
1047 --enable-newline-is-any
1049 causes PCRE to recognize any Unicode newline sequence.
1051 Whatever line ending convention is selected when PCRE is built can be
1052 overridden when the library functions are called. At build time it is
1053 conventional to use the standard for your operating system.
1058 By default, the sequence \R in a pattern matches any Unicode newline
1059 sequence, whatever has been selected as the line ending sequence. If
1062 --enable-bsr-anycrlf
1064 the default is changed so that \R matches only CR, LF, or CRLF. What-
1065 ever is selected when PCRE is built can be overridden when the library
1066 functions are called.
1071 When the 8-bit library is called through the POSIX interface (see the
1072 pcreposix documentation), additional working storage is required for
1073 holding the pointers to capturing substrings, because PCRE requires
1074 three integers per substring, whereas the POSIX interface provides only
1075 two. If the number of expected substrings is small, the wrapper func-
1076 tion uses space on the stack, because this is faster than using mal-
1077 loc() for each call. The default threshold above which the stack is no
1078 longer used is 10; it can be changed by adding a setting such as
1080 --with-posix-malloc-threshold=20
1082 to the configure command.
1085 HANDLING VERY LARGE PATTERNS
1087 Within a compiled pattern, offset values are used to point from one
1088 part to another (for example, from an opening parenthesis to an alter-
1089 nation metacharacter). By default, in the 8-bit and 16-bit libraries,
1090 two-byte values are used for these offsets, leading to a maximum size
1091 for a compiled pattern of around 64K. This is sufficient to handle all
1092 but the most gigantic patterns. Nevertheless, some people do want to
1093 process truly enormous patterns, so it is possible to compile PCRE to
1094 use three-byte or four-byte offsets by adding a setting such as
1098 to the configure command. The value given must be 2, 3, or 4. For the
1099 16-bit library, a value of 3 is rounded up to 4. In these libraries,
1100 using longer offsets slows down the operation of PCRE because it has to
1101 load additional data when handling them. For the 32-bit library the
1102 value is always 4 and cannot be overridden; the value of --with-link-
1106 AVOIDING EXCESSIVE STACK USAGE
1108 When matching with the pcre_exec() function, PCRE implements backtrack-
1109 ing by making recursive calls to an internal function called match().
1110 In environments where the size of the stack is limited, this can se-
1111 verely limit PCRE's operation. (The Unix environment does not usually
1112 suffer from this problem, but it may sometimes be necessary to increase
1113 the maximum stack size. There is a discussion in the pcrestack docu-
1114 mentation.) An alternative approach to recursion that uses memory from
1115 the heap to remember data, instead of using recursive function calls,
1116 has been implemented to work round the problem of limited stack size.
1117 If you want to build a version of PCRE that works this way, add
1119 --disable-stack-for-recursion
1121 to the configure command. With this configuration, PCRE will use the
1122 pcre_stack_malloc and pcre_stack_free variables to call memory manage-
1123 ment functions. By default these point to malloc() and free(), but you
1124 can replace the pointers so that your own functions are used instead.
1126 Separate functions are provided rather than using pcre_malloc and
1127 pcre_free because the usage is very predictable: the block sizes
1128 requested are always the same, and the blocks are always freed in
1129 reverse order. A calling program might be able to implement optimized
1130 functions that perform better than malloc() and free(). PCRE runs
1131 noticeably more slowly when built in this way. This option affects only
1132 the pcre_exec() function; it is not relevant for pcre_dfa_exec().
1135 LIMITING PCRE RESOURCE USAGE
1137 Internally, PCRE has a function called match(), which it calls repeat-
1138 edly (sometimes recursively) when matching a pattern with the
1139 pcre_exec() function. By controlling the maximum number of times this
1140 function may be called during a single matching operation, a limit can
1141 be placed on the resources used by a single call to pcre_exec(). The
1142 limit can be changed at run time, as described in the pcreapi documen-
1143 tation. The default is 10 million, but this can be changed by adding a
1146 --with-match-limit=500000
1148 to the configure command. This setting has no effect on the
1149 pcre_dfa_exec() matching function.
1151 In some environments it is desirable to limit the depth of recursive
1152 calls of match() more strictly than the total number of calls, in order
1153 to restrict the maximum amount of stack (or heap, if --disable-stack-
1154 for-recursion is specified) that is used. A second limit controls this;
1155 it defaults to the value that is set for --with-match-limit, which
1156 imposes no additional constraints. However, you can set a lower limit
1157 by adding, for example,
1159 --with-match-limit-recursion=10000
1161 to the configure command. This value can also be overridden at run
1165 CREATING CHARACTER TABLES AT BUILD TIME
1167 PCRE uses fixed tables for processing characters whose code values are
1168 less than 256. By default, PCRE is built with a set of tables that are
1169 distributed in the file pcre_chartables.c.dist. These tables are for
1170 ASCII codes only. If you add
1172 --enable-rebuild-chartables
1174 to the configure command, the distributed tables are no longer used.
1175 Instead, a program called dftables is compiled and run. This outputs
1176 the source for new set of tables, created in the default locale of your
1177 C run-time system. (This method of replacing the tables does not work
1178 if you are cross compiling, because dftables is run on the local host.
1179 If you need to create alternative tables when cross compiling, you will
1180 have to do so "by hand".)
1185 PCRE assumes by default that it will run in an environment where the
1186 character code is ASCII (or Unicode, which is a superset of ASCII).
1187 This is the case for most computer operating systems. PCRE can, how-
1188 ever, be compiled to run in an EBCDIC environment by adding
1192 to the configure command. This setting implies --enable-rebuild-charta-
1193 bles. You should only use it if you know that you are in an EBCDIC
1194 environment (for example, an IBM mainframe operating system). The
1195 --enable-ebcdic option is incompatible with --enable-utf.
1197 The EBCDIC character that corresponds to an ASCII LF is assumed to have
1198 the value 0x15 by default. However, in some EBCDIC environments, 0x25
1199 is used. In such an environment you should use
1201 --enable-ebcdic-nl25
1203 as well as, or instead of, --enable-ebcdic. The EBCDIC character for CR
1204 has the same value as in ASCII, namely, 0x0d. Whichever of 0x15 and
1205 0x25 is not chosen as LF is made to correspond to the Unicode NEL char-
1206 acter (which, in Unicode, is 0x85).
1208 The options that select newline behaviour, such as --enable-newline-is-
1209 cr, and equivalent run-time options, refer to these character values in
1210 an EBCDIC environment.
1213 PCREGREP OPTIONS FOR COMPRESSED FILE SUPPORT
1215 By default, pcregrep reads all files as plain text. You can build it so
1216 that it recognizes files whose names end in .gz or .bz2, and reads them
1217 with libz or libbz2, respectively, by adding one or both of
1219 --enable-pcregrep-libz
1220 --enable-pcregrep-libbz2
1222 to the configure command. These options naturally require that the rel-
1223 evant libraries are installed on your system. Configuration will fail
1227 PCREGREP BUFFER SIZE
1229 pcregrep uses an internal buffer to hold a "window" on the file it is
1230 scanning, in order to be able to output "before" and "after" lines when
1231 it finds a match. The size of the buffer is controlled by a parameter
1232 whose default value is 20K. The buffer itself is three times this size,
1233 but because of the way it is used for holding "before" lines, the long-
1234 est line that is guaranteed to be processable is the parameter size.
1235 You can change the default parameter value by adding, for example,
1237 --with-pcregrep-bufsize=50K
1239 to the configure command. The caller of pcregrep can, however, override
1240 this value by specifying a run-time option.
1243 PCRETEST OPTION FOR LIBREADLINE SUPPORT
1247 --enable-pcretest-libreadline
1249 to the configure command, pcretest is linked with the libreadline
1250 library, and when its input is from a terminal, it reads it using the
1251 readline() function. This provides line-editing and history facilities.
1252 Note that libreadline is GPL-licensed, so if you distribute a binary of
1253 pcretest linked in this way, there may be licensing issues.
1255 Setting this option causes the -lreadline option to be added to the
1256 pcretest build. In many operating environments with a sytem-installed
1257 libreadline this is sufficient. However, in some environments (e.g. if
1258 an unmodified distribution version of readline is in use), some extra
1259 configuration may be necessary. The INSTALL file for libreadline says
1262 "Readline uses the termcap functions, but does not link with the
1263 termcap or curses library itself, allowing applications which link
1264 with readline the to choose an appropriate library."
1266 If your environment has not been set up so that an appropriate library
1267 is automatically included, you may need to add something like
1271 immediately before the configure command.
1274 DEBUGGING WITH VALGRIND SUPPORT
1280 option to to the configure command, PCRE will use valgrind annotations
1281 to mark certain memory regions as unaddressable. This allows it to
1282 detect invalid memory accesses, and is mostly useful for debugging PCRE
1286 CODE COVERAGE REPORTING
1288 If your C compiler is gcc, you can build a version of PCRE that can
1289 generate a code coverage report for its test suite. To enable this, you
1290 must install lcov version 1.6 or above. Then specify
1294 to the configure command and build PCRE in the usual way.
1296 Note that using ccache (a caching C compiler) is incompatible with code
1297 coverage reporting. If you have configured ccache to run automatically
1298 on your system, you must set the environment variable
1302 before running make to build PCRE, so that ccache is not used.
1304 When --enable-coverage is used, the following addition targets are
1305 added to the Makefile:
1309 This creates a fresh coverage report for the PCRE test suite. It is
1310 equivalent to running "make coverage-reset", "make coverage-baseline",
1311 "make check", and then "make coverage-report".
1315 This zeroes the coverage counters, but does nothing else.
1317 make coverage-baseline
1319 This captures baseline coverage information.
1321 make coverage-report
1323 This creates the coverage report.
1325 make coverage-clean-report
1327 This removes the generated coverage report without cleaning the cover-
1330 make coverage-clean-data
1332 This removes the captured coverage data without removing the coverage
1333 files created at compile time (*.gcno).
1337 This cleans all coverage data including the generated coverage report.
1338 For more information about code coverage, see the gcov and lcov docu-
1344 pcreapi(3), pcre16, pcre32, pcre_config(3).
1350 University Computing Service
1351 Cambridge CB2 3QH, England.
1356 Last updated: 12 May 2013
1357 Copyright (c) 1997-2013 University of Cambridge.
1358 ------------------------------------------------------------------------------
1361 PCREMATCHING(3) Library Functions Manual PCREMATCHING(3)
1366 PCRE - Perl-compatible regular expressions
1368 PCRE MATCHING ALGORITHMS
1370 This document describes the two different algorithms that are available
1371 in PCRE for matching a compiled regular expression against a given sub-
1372 ject string. The "standard" algorithm is the one provided by the
1373 pcre_exec(), pcre16_exec() and pcre32_exec() functions. These work in
1374 the same as as Perl's matching function, and provide a Perl-compatible
1375 matching operation. The just-in-time (JIT) optimization that is
1376 described in the pcrejit documentation is compatible with these func-
1379 An alternative algorithm is provided by the pcre_dfa_exec(),
1380 pcre16_dfa_exec() and pcre32_dfa_exec() functions; they operate in a
1381 different way, and are not Perl-compatible. This alternative has advan-
1382 tages and disadvantages compared with the standard algorithm, and these
1383 are described below.
1385 When there is only one possible way in which a given subject string can
1386 match a pattern, the two algorithms give the same answer. A difference
1387 arises, however, when there are multiple possibilities. For example, if
1392 is matched against the string
1394 <something> <something else> <something further>
1396 there are three possible answers. The standard algorithm finds only one
1397 of them, whereas the alternative algorithm finds all three.
1400 REGULAR EXPRESSIONS AS TREES
1402 The set of strings that are matched by a regular expression can be rep-
1403 resented as a tree structure. An unlimited repetition in the pattern
1404 makes the tree of infinite size, but it is still a tree. Matching the
1405 pattern to a given subject string (from a given starting point) can be
1406 thought of as a search of the tree. There are two ways to search a
1407 tree: depth-first and breadth-first, and these correspond to the two
1408 matching algorithms provided by PCRE.
1411 THE STANDARD MATCHING ALGORITHM
1413 In the terminology of Jeffrey Friedl's book "Mastering Regular Expres-
1414 sions", the standard algorithm is an "NFA algorithm". It conducts a
1415 depth-first search of the pattern tree. That is, it proceeds along a
1416 single path through the tree, checking that the subject matches what is
1417 required. When there is a mismatch, the algorithm tries any alterna-
1418 tives at the current point, and if they all fail, it backs up to the
1419 previous branch point in the tree, and tries the next alternative
1420 branch at that level. This often involves backing up (moving to the
1421 left) in the subject string as well. The order in which repetition
1422 branches are tried is controlled by the greedy or ungreedy nature of
1425 If a leaf node is reached, a matching string has been found, and at
1426 that point the algorithm stops. Thus, if there is more than one possi-
1427 ble match, this algorithm returns the first one that it finds. Whether
1428 this is the shortest, the longest, or some intermediate length depends
1429 on the way the greedy and ungreedy repetition quantifiers are specified
1432 Because it ends up with a single path through the tree, it is rela-
1433 tively straightforward for this algorithm to keep track of the sub-
1434 strings that are matched by portions of the pattern in parentheses.
1435 This provides support for capturing parentheses and back references.
1438 THE ALTERNATIVE MATCHING ALGORITHM
1440 This algorithm conducts a breadth-first search of the tree. Starting
1441 from the first matching point in the subject, it scans the subject
1442 string from left to right, once, character by character, and as it does
1443 this, it remembers all the paths through the tree that represent valid
1444 matches. In Friedl's terminology, this is a kind of "DFA algorithm",
1445 though it is not implemented as a traditional finite state machine (it
1446 keeps multiple states active simultaneously).
1448 Although the general principle of this matching algorithm is that it
1449 scans the subject string only once, without backtracking, there is one
1450 exception: when a lookaround assertion is encountered, the characters
1451 following or preceding the current point have to be independently
1454 The scan continues until either the end of the subject is reached, or
1455 there are no more unterminated paths. At this point, terminated paths
1456 represent the different matching possibilities (if there are none, the
1457 match has failed). Thus, if there is more than one possible match,
1458 this algorithm finds all of them, and in particular, it finds the long-
1459 est. The matches are returned in decreasing order of length. There is
1460 an option to stop the algorithm after the first match (which is neces-
1461 sarily the shortest) is found.
1463 Note that all the matches that are found start at the same point in the
1464 subject. If the pattern
1468 is matched against the string "the caterpillar catchment", the result
1469 will be the three strings "caterpillar", "cater", and "cat" that start
1470 at the fifth character of the subject. The algorithm does not automati-
1471 cally move on to find matches that start at later positions.
1473 PCRE's "auto-possessification" optimization usually applies to charac-
1474 ter repeats at the end of a pattern (as well as internally). For exam-
1475 ple, the pattern "a\d+" is compiled as if it were "a\d++" because there
1476 is no point even considering the possibility of backtracking into the
1477 repeated digits. For DFA matching, this means that only one possible
1478 match is found. If you really do want multiple matches in such cases,
1479 either use an ungreedy repeat ("a\d+?") or set the PCRE_NO_AUTO_POSSESS
1480 option when compiling.
1482 There are a number of features of PCRE regular expressions that are not
1483 supported by the alternative matching algorithm. They are as follows:
1485 1. Because the algorithm finds all possible matches, the greedy or
1486 ungreedy nature of repetition quantifiers is not relevant. Greedy and
1487 ungreedy quantifiers are treated in exactly the same way. However, pos-
1488 sessive quantifiers can make a difference when what follows could also
1489 match what is quantified, for example in a pattern like this:
1493 This pattern matches "aaab!" but not "aaa!", which would be matched by
1494 a non-possessive quantifier. Similarly, if an atomic group is present,
1495 it is matched as if it were a standalone pattern at the current point,
1496 and the longest match is then "locked in" for the rest of the overall
1499 2. When dealing with multiple paths through the tree simultaneously, it
1500 is not straightforward to keep track of captured substrings for the
1501 different matching possibilities, and PCRE's implementation of this
1502 algorithm does not attempt to do this. This means that no captured sub-
1503 strings are available.
1505 3. Because no substrings are captured, back references within the pat-
1506 tern are not supported, and cause errors if encountered.
1508 4. For the same reason, conditional expressions that use a backrefer-
1509 ence as the condition or test for a specific group recursion are not
1512 5. Because many paths through the tree may be active, the \K escape
1513 sequence, which resets the start of the match when encountered (but may
1514 be on some paths and not on others), is not supported. It causes an
1515 error if encountered.
1517 6. Callouts are supported, but the value of the capture_top field is
1518 always 1, and the value of the capture_last field is always -1.
1520 7. The \C escape sequence, which (in the standard algorithm) always
1521 matches a single data unit, even in UTF-8, UTF-16 or UTF-32 modes, is
1522 not supported in these modes, because the alternative algorithm moves
1523 through the subject string one character (not data unit) at a time, for
1524 all active paths through the tree.
1526 8. Except for (*FAIL), the backtracking control verbs such as (*PRUNE)
1527 are not supported. (*FAIL) is supported, and behaves like a failing
1531 ADVANTAGES OF THE ALTERNATIVE ALGORITHM
1533 Using the alternative matching algorithm provides the following advan-
1536 1. All possible matches (at a single point in the subject) are automat-
1537 ically found, and in particular, the longest match is found. To find
1538 more than one match using the standard algorithm, you have to do kludgy
1539 things with callouts.
1541 2. Because the alternative algorithm scans the subject string just
1542 once, and never needs to backtrack (except for lookbehinds), it is pos-
1543 sible to pass very long subject strings to the matching function in
1544 several pieces, checking for partial matching each time. Although it is
1545 possible to do multi-segment matching using the standard algorithm by
1546 retaining partially matched substrings, it is more complicated. The
1547 pcrepartial documentation gives details of partial matching and dis-
1548 cusses multi-segment matching.
1551 DISADVANTAGES OF THE ALTERNATIVE ALGORITHM
1553 The alternative algorithm suffers from a number of disadvantages:
1555 1. It is substantially slower than the standard algorithm. This is
1556 partly because it has to search for all possible matches, but is also
1557 because it is less susceptible to optimization.
1559 2. Capturing parentheses and back references are not supported.
1561 3. Although atomic groups are supported, their use does not provide the
1562 performance advantage that it does for the standard algorithm.
1568 University Computing Service
1569 Cambridge CB2 3QH, England.
1574 Last updated: 12 November 2013
1575 Copyright (c) 1997-2012 University of Cambridge.
1576 ------------------------------------------------------------------------------
1579 PCREAPI(3) Library Functions Manual PCREAPI(3)
1584 PCRE - Perl-compatible regular expressions
1589 PCRE NATIVE API BASIC FUNCTIONS
1591 pcre *pcre_compile(const char *pattern, int options,
1592 const char **errptr, int *erroffset,
1593 const unsigned char *tableptr);
1595 pcre *pcre_compile2(const char *pattern, int options,
1597 const char **errptr, int *erroffset,
1598 const unsigned char *tableptr);
1600 pcre_extra *pcre_study(const pcre *code, int options,
1601 const char **errptr);
1603 void pcre_free_study(pcre_extra *extra);
1605 int pcre_exec(const pcre *code, const pcre_extra *extra,
1606 const char *subject, int length, int startoffset,
1607 int options, int *ovector, int ovecsize);
1609 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
1610 const char *subject, int length, int startoffset,
1611 int options, int *ovector, int ovecsize,
1612 int *workspace, int wscount);
1615 PCRE NATIVE API STRING EXTRACTION FUNCTIONS
1617 int pcre_copy_named_substring(const pcre *code,
1618 const char *subject, int *ovector,
1619 int stringcount, const char *stringname,
1620 char *buffer, int buffersize);
1622 int pcre_copy_substring(const char *subject, int *ovector,
1623 int stringcount, int stringnumber, char *buffer,
1626 int pcre_get_named_substring(const pcre *code,
1627 const char *subject, int *ovector,
1628 int stringcount, const char *stringname,
1629 const char **stringptr);
1631 int pcre_get_stringnumber(const pcre *code,
1634 int pcre_get_stringtable_entries(const pcre *code,
1635 const char *name, char **first, char **last);
1637 int pcre_get_substring(const char *subject, int *ovector,
1638 int stringcount, int stringnumber,
1639 const char **stringptr);
1641 int pcre_get_substring_list(const char *subject,
1642 int *ovector, int stringcount, const char ***listptr);
1644 void pcre_free_substring(const char *stringptr);
1646 void pcre_free_substring_list(const char **stringptr);
1649 PCRE NATIVE API AUXILIARY FUNCTIONS
1651 int pcre_jit_exec(const pcre *code, const pcre_extra *extra,
1652 const char *subject, int length, int startoffset,
1653 int options, int *ovector, int ovecsize,
1654 pcre_jit_stack *jstack);
1656 pcre_jit_stack *pcre_jit_stack_alloc(int startsize, int maxsize);
1658 void pcre_jit_stack_free(pcre_jit_stack *stack);
1660 void pcre_assign_jit_stack(pcre_extra *extra,
1661 pcre_jit_callback callback, void *data);
1663 const unsigned char *pcre_maketables(void);
1665 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
1666 int what, void *where);
1668 int pcre_refcount(pcre *code, int adjust);
1670 int pcre_config(int what, void *where);
1672 const char *pcre_version(void);
1674 int pcre_pattern_to_host_byte_order(pcre *code,
1675 pcre_extra *extra, const unsigned char *tables);
1678 PCRE NATIVE API INDIRECTED FUNCTIONS
1680 void *(*pcre_malloc)(size_t);
1682 void (*pcre_free)(void *);
1684 void *(*pcre_stack_malloc)(size_t);
1686 void (*pcre_stack_free)(void *);
1688 int (*pcre_callout)(pcre_callout_block *);
1690 int (*pcre_stack_guard)(void);
1693 PCRE 8-BIT, 16-BIT, AND 32-BIT LIBRARIES
1695 As well as support for 8-bit character strings, PCRE also supports
1696 16-bit strings (from release 8.30) and 32-bit strings (from release
1697 8.32), by means of two additional libraries. They can be built as well
1698 as, or instead of, the 8-bit library. To avoid too much complication,
1699 this document describes the 8-bit versions of the functions, with only
1700 occasional references to the 16-bit and 32-bit libraries.
1702 The 16-bit and 32-bit functions operate in the same way as their 8-bit
1703 counterparts; they just use different data types for their arguments
1704 and results, and their names start with pcre16_ or pcre32_ instead of
1705 pcre_. For every option that has UTF8 in its name (for example,
1706 PCRE_UTF8), there are corresponding 16-bit and 32-bit names with UTF8
1707 replaced by UTF16 or UTF32, respectively. This facility is in fact just
1708 cosmetic; the 16-bit and 32-bit option names define the same bit val-
1711 References to bytes and UTF-8 in this document should be read as refer-
1712 ences to 16-bit data units and UTF-16 when using the 16-bit library, or
1713 32-bit data units and UTF-32 when using the 32-bit library, unless
1714 specified otherwise. More details of the specific differences for the
1715 16-bit and 32-bit libraries are given in the pcre16 and pcre32 pages.
1720 PCRE has its own native API, which is described in this document. There
1721 are also some wrapper functions (for the 8-bit library only) that cor-
1722 respond to the POSIX regular expression API, but they do not give
1723 access to all the functionality. They are described in the pcreposix
1724 documentation. Both of these APIs define a set of C function calls. A
1725 C++ wrapper (again for the 8-bit library only) is also distributed with
1726 PCRE. It is documented in the pcrecpp page.
1728 The native API C function prototypes are defined in the header file
1729 pcre.h, and on Unix-like systems the (8-bit) library itself is called
1730 libpcre. It can normally be accessed by adding -lpcre to the command
1731 for linking an application that uses PCRE. The header file defines the
1732 macros PCRE_MAJOR and PCRE_MINOR to contain the major and minor release
1733 numbers for the library. Applications can use these to include support
1734 for different releases of PCRE.
1736 In a Windows environment, if you want to statically link an application
1737 program against a non-dll pcre.a file, you must define PCRE_STATIC
1738 before including pcre.h or pcrecpp.h, because otherwise the pcre_mal-
1739 loc() and pcre_free() exported functions will be declared
1740 __declspec(dllimport), with unwanted results.
1742 The functions pcre_compile(), pcre_compile2(), pcre_study(), and
1743 pcre_exec() are used for compiling and matching regular expressions in
1744 a Perl-compatible manner. A sample program that demonstrates the sim-
1745 plest way of using them is provided in the file called pcredemo.c in
1746 the PCRE source distribution. A listing of this program is given in the
1747 pcredemo documentation, and the pcresample documentation describes how
1748 to compile and run it.
1750 Just-in-time compiler support is an optional feature of PCRE that can
1751 be built in appropriate hardware environments. It greatly speeds up the
1752 matching performance of many patterns. Simple programs can easily
1753 request that it be used if available, by setting an option that is
1754 ignored when it is not relevant. More complicated programs might need
1755 to make use of the functions pcre_jit_stack_alloc(),
1756 pcre_jit_stack_free(), and pcre_assign_jit_stack() in order to control
1757 the JIT code's memory usage.
1759 From release 8.32 there is also a direct interface for JIT execution,
1760 which gives improved performance. The JIT-specific functions are dis-
1761 cussed in the pcrejit documentation.
1763 A second matching function, pcre_dfa_exec(), which is not Perl-compati-
1764 ble, is also provided. This uses a different algorithm for the match-
1765 ing. The alternative algorithm finds all possible matches (at a given
1766 point in the subject), and scans the subject just once (unless there
1767 are lookbehind assertions). However, this algorithm does not return
1768 captured substrings. A description of the two matching algorithms and
1769 their advantages and disadvantages is given in the pcrematching docu-
1772 In addition to the main compiling and matching functions, there are
1773 convenience functions for extracting captured substrings from a subject
1774 string that is matched by pcre_exec(). They are:
1776 pcre_copy_substring()
1777 pcre_copy_named_substring()
1778 pcre_get_substring()
1779 pcre_get_named_substring()
1780 pcre_get_substring_list()
1781 pcre_get_stringnumber()
1782 pcre_get_stringtable_entries()
1784 pcre_free_substring() and pcre_free_substring_list() are also provided,
1785 to free the memory used for extracted strings.
1787 The function pcre_maketables() is used to build a set of character
1788 tables in the current locale for passing to pcre_compile(),
1789 pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is
1790 provided for specialist use. Most commonly, no special tables are
1791 passed, in which case internal tables that are generated when PCRE is
1794 The function pcre_fullinfo() is used to find out information about a
1795 compiled pattern. The function pcre_version() returns a pointer to a
1796 string containing the version of PCRE and its date of release.
1798 The function pcre_refcount() maintains a reference count in a data
1799 block containing a compiled pattern. This is provided for the benefit
1800 of object-oriented applications.
1802 The global variables pcre_malloc and pcre_free initially contain the
1803 entry points of the standard malloc() and free() functions, respec-
1804 tively. PCRE calls the memory management functions via these variables,
1805 so a calling program can replace them if it wishes to intercept the
1806 calls. This should be done before calling any PCRE functions.
1808 The global variables pcre_stack_malloc and pcre_stack_free are also
1809 indirections to memory management functions. These special functions
1810 are used only when PCRE is compiled to use the heap for remembering
1811 data, instead of recursive function calls, when running the pcre_exec()
1812 function. See the pcrebuild documentation for details of how to do
1813 this. It is a non-standard way of building PCRE, for use in environ-
1814 ments that have limited stacks. Because of the greater use of memory
1815 management, it runs more slowly. Separate functions are provided so
1816 that special-purpose external code can be used for this case. When
1817 used, these functions always allocate memory blocks of the same size.
1818 There is a discussion about PCRE's stack usage in the pcrestack docu-
1821 The global variable pcre_callout initially contains NULL. It can be set
1822 by the caller to a "callout" function, which PCRE will then call at
1823 specified points during a matching operation. Details are given in the
1824 pcrecallout documentation.
1826 The global variable pcre_stack_guard initially contains NULL. It can be
1827 set by the caller to a function that is called by PCRE whenever it
1828 starts to compile a parenthesized part of a pattern. When parentheses
1829 are nested, PCRE uses recursive function calls, which use up the system
1830 stack. This function is provided so that applications with restricted
1831 stacks can force a compilation error if the stack runs out. The func-
1832 tion should return zero if all is well, or non-zero to force an error.
1837 PCRE supports five different conventions for indicating line breaks in
1838 strings: a single CR (carriage return) character, a single LF (line-
1839 feed) character, the two-character sequence CRLF, any of the three pre-
1840 ceding, or any Unicode newline sequence. The Unicode newline sequences
1841 are the three just mentioned, plus the single characters VT (vertical
1842 tab, U+000B), FF (form feed, U+000C), NEL (next line, U+0085), LS (line
1843 separator, U+2028), and PS (paragraph separator, U+2029).
1845 Each of the first three conventions is used by at least one operating
1846 system as its standard newline sequence. When PCRE is built, a default
1847 can be specified. The default default is LF, which is the Unix stan-
1848 dard. When PCRE is run, the default can be overridden, either when a
1849 pattern is compiled, or when it is matched.
1851 At compile time, the newline convention can be specified by the options
1852 argument of pcre_compile(), or it can be specified by special text at
1853 the start of the pattern itself; this overrides any other settings. See
1854 the pcrepattern page for details of the special character sequences.
1856 In the PCRE documentation the word "newline" is used to mean "the char-
1857 acter or pair of characters that indicate a line break". The choice of
1858 newline convention affects the handling of the dot, circumflex, and
1859 dollar metacharacters, the handling of #-comments in /x mode, and, when
1860 CRLF is a recognized line ending sequence, the match position advance-
1861 ment for a non-anchored pattern. There is more detail about this in the
1862 section on pcre_exec() options below.
1864 The choice of newline convention does not affect the interpretation of
1865 the \n or \r escape sequences, nor does it affect what \R matches,
1866 which is controlled in a similar way, but by separate options.
1871 The PCRE functions can be used in multi-threading applications, with
1872 the proviso that the memory management functions pointed to by
1873 pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the
1874 callout and stack-checking functions pointed to by pcre_callout and
1875 pcre_stack_guard, are shared by all threads.
1877 The compiled form of a regular expression is not altered during match-
1878 ing, so the same compiled pattern can safely be used by several threads
1881 If the just-in-time optimization feature is being used, it needs sepa-
1882 rate memory stack areas for each thread. See the pcrejit documentation
1886 SAVING PRECOMPILED PATTERNS FOR LATER USE
1888 The compiled form of a regular expression can be saved and re-used at a
1889 later time, possibly by a different program, and even on a host other
1890 than the one on which it was compiled. Details are given in the
1891 pcreprecompile documentation, which includes a description of the
1892 pcre_pattern_to_host_byte_order() function. However, compiling a regu-
1893 lar expression with one version of PCRE for use with a different ver-
1894 sion is not guaranteed to work and may cause crashes.
1897 CHECKING BUILD-TIME OPTIONS
1899 int pcre_config(int what, void *where);
1901 The function pcre_config() makes it possible for a PCRE client to dis-
1902 cover which optional features have been compiled into the PCRE library.
1903 The pcrebuild documentation has more details about these optional fea-
1906 The first argument for pcre_config() is an integer, specifying which
1907 information is required; the second argument is a pointer to a variable
1908 into which the information is placed. The returned value is zero on
1909 success, or the negative error code PCRE_ERROR_BADOPTION if the value
1910 in the first argument is not recognized. The following information is
1915 The output is an integer that is set to one if UTF-8 support is avail-
1916 able; otherwise it is set to zero. This value should normally be given
1917 to the 8-bit version of this function, pcre_config(). If it is given to
1918 the 16-bit or 32-bit version of this function, the result is
1919 PCRE_ERROR_BADOPTION.
1923 The output is an integer that is set to one if UTF-16 support is avail-
1924 able; otherwise it is set to zero. This value should normally be given
1925 to the 16-bit version of this function, pcre16_config(). If it is given
1926 to the 8-bit or 32-bit version of this function, the result is
1927 PCRE_ERROR_BADOPTION.
1931 The output is an integer that is set to one if UTF-32 support is avail-
1932 able; otherwise it is set to zero. This value should normally be given
1933 to the 32-bit version of this function, pcre32_config(). If it is given
1934 to the 8-bit or 16-bit version of this function, the result is
1935 PCRE_ERROR_BADOPTION.
1937 PCRE_CONFIG_UNICODE_PROPERTIES
1939 The output is an integer that is set to one if support for Unicode
1940 character properties is available; otherwise it is set to zero.
1944 The output is an integer that is set to one if support for just-in-time
1945 compiling is available; otherwise it is set to zero.
1947 PCRE_CONFIG_JITTARGET
1949 The output is a pointer to a zero-terminated "const char *" string. If
1950 JIT support is available, the string contains the name of the architec-
1951 ture for which the JIT compiler is configured, for example "x86 32bit
1952 (little endian + unaligned)". If JIT support is not available, the
1957 The output is an integer whose value specifies the default character
1958 sequence that is recognized as meaning "newline". The values that are
1959 supported in ASCII/Unicode environments are: 10 for LF, 13 for CR, 3338
1960 for CRLF, -2 for ANYCRLF, and -1 for ANY. In EBCDIC environments, CR,
1961 ANYCRLF, and ANY yield the same values. However, the value for LF is
1962 normally 21, though some EBCDIC environments use 37. The corresponding
1963 values for CRLF are 3349 and 3365. The default should normally corre-
1964 spond to the standard sequence for your operating system.
1968 The output is an integer whose value indicates what character sequences
1969 the \R escape sequence matches by default. A value of 0 means that \R
1970 matches any Unicode line ending sequence; a value of 1 means that \R
1971 matches only CR, LF, or CRLF. The default can be overridden when a pat-
1972 tern is compiled or matched.
1974 PCRE_CONFIG_LINK_SIZE
1976 The output is an integer that contains the number of bytes used for
1977 internal linkage in compiled regular expressions. For the 8-bit
1978 library, the value can be 2, 3, or 4. For the 16-bit library, the value
1979 is either 2 or 4 and is still a number of bytes. For the 32-bit
1980 library, the value is either 2 or 4 and is still a number of bytes. The
1981 default value of 2 is sufficient for all but the most massive patterns,
1982 since it allows the compiled pattern to be up to 64K in size. Larger
1983 values allow larger regular expressions to be compiled, at the expense
1986 PCRE_CONFIG_POSIX_MALLOC_THRESHOLD
1988 The output is an integer that contains the threshold above which the
1989 POSIX interface uses malloc() for output vectors. Further details are
1990 given in the pcreposix documentation.
1992 PCRE_CONFIG_PARENS_LIMIT
1994 The output is a long integer that gives the maximum depth of nesting of
1995 parentheses (of any kind) in a pattern. This limit is imposed to cap
1996 the amount of system stack used when a pattern is compiled. It is spec-
1997 ified when PCRE is built; the default is 250. This limit does not take
1998 into account the stack that may already be used by the calling applica-
1999 tion. For finer control over compilation stack usage, you can set a
2000 pointer to an external checking function in pcre_stack_guard.
2002 PCRE_CONFIG_MATCH_LIMIT
2004 The output is a long integer that gives the default limit for the num-
2005 ber of internal matching function calls in a pcre_exec() execution.
2006 Further details are given with pcre_exec() below.
2008 PCRE_CONFIG_MATCH_LIMIT_RECURSION
2010 The output is a long integer that gives the default limit for the depth
2011 of recursion when calling the internal matching function in a
2012 pcre_exec() execution. Further details are given with pcre_exec()
2015 PCRE_CONFIG_STACKRECURSE
2017 The output is an integer that is set to one if internal recursion when
2018 running pcre_exec() is implemented by recursive function calls that use
2019 the stack to remember their state. This is the usual way that PCRE is
2020 compiled. The output is zero if PCRE was compiled to use blocks of data
2021 on the heap instead of recursive function calls. In this case,
2022 pcre_stack_malloc and pcre_stack_free are called to manage memory
2023 blocks on the heap, thus avoiding the use of the stack.
2028 pcre *pcre_compile(const char *pattern, int options,
2029 const char **errptr, int *erroffset,
2030 const unsigned char *tableptr);
2032 pcre *pcre_compile2(const char *pattern, int options,
2034 const char **errptr, int *erroffset,
2035 const unsigned char *tableptr);
2037 Either of the functions pcre_compile() or pcre_compile2() can be called
2038 to compile a pattern into an internal form. The only difference between
2039 the two interfaces is that pcre_compile2() has an additional argument,
2040 errorcodeptr, via which a numerical error code can be returned. To
2041 avoid too much repetition, we refer just to pcre_compile() below, but
2042 the information applies equally to pcre_compile2().
2044 The pattern is a C string terminated by a binary zero, and is passed in
2045 the pattern argument. A pointer to a single block of memory that is
2046 obtained via pcre_malloc is returned. This contains the compiled code
2047 and related data. The pcre type is defined for the returned block; this
2048 is a typedef for a structure whose contents are not externally defined.
2049 It is up to the caller to free the memory (via pcre_free) when it is no
2052 Although the compiled code of a PCRE regex is relocatable, that is, it
2053 does not depend on memory location, the complete pcre data block is not
2054 fully relocatable, because it may contain a copy of the tableptr argu-
2055 ment, which is an address (see below).
2057 The options argument contains various bit settings that affect the com-
2058 pilation. It should be zero if no options are required. The available
2059 options are described below. Some of them (in particular, those that
2060 are compatible with Perl, but some others as well) can also be set and
2061 unset from within the pattern (see the detailed description in the
2062 pcrepattern documentation). For those options that can be different in
2063 different parts of the pattern, the contents of the options argument
2064 specifies their settings at the start of compilation and execution. The
2065 PCRE_ANCHORED, PCRE_BSR_xxx, PCRE_NEWLINE_xxx, PCRE_NO_UTF8_CHECK, and
2066 PCRE_NO_START_OPTIMIZE options can be set at the time of matching as
2067 well as at compile time.
2069 If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise,
2070 if compilation of a pattern fails, pcre_compile() returns NULL, and
2071 sets the variable pointed to by errptr to point to a textual error mes-
2072 sage. This is a static string that is part of the library. You must not
2073 try to free it. Normally, the offset from the start of the pattern to
2074 the data unit that was being processed when the error was discovered is
2075 placed in the variable pointed to by erroffset, which must not be NULL
2076 (if it is, an immediate error is given). However, for an invalid UTF-8
2077 or UTF-16 string, the offset is that of the first data unit of the
2080 Some errors are not detected until the whole pattern has been scanned;
2081 in these cases, the offset passed back is the length of the pattern.
2082 Note that the offset is in data units, not characters, even in a UTF
2083 mode. It may sometimes point into the middle of a UTF-8 or UTF-16 char-
2086 If pcre_compile2() is used instead of pcre_compile(), and the error-
2087 codeptr argument is not NULL, a non-zero error code number is returned
2088 via this argument in the event of an error. This is in addition to the
2089 textual error message. Error codes and messages are listed below.
2091 If the final argument, tableptr, is NULL, PCRE uses a default set of
2092 character tables that are built when PCRE is compiled, using the
2093 default C locale. Otherwise, tableptr must be an address that is the
2094 result of a call to pcre_maketables(). This value is stored with the
2095 compiled pattern, and used again by pcre_exec() and pcre_dfa_exec()
2096 when the pattern is matched. For more discussion, see the section on
2097 locale support below.
2099 This code fragment shows a typical straightforward call to pcre_com-
2106 "^A.*Z", /* the pattern */
2107 0, /* default options */
2108 &error, /* for error message */
2109 &erroffset, /* for error offset */
2110 NULL); /* use default character tables */
2112 The following names for option bits are defined in the pcre.h header
2117 If this bit is set, the pattern is forced to be "anchored", that is, it
2118 is constrained to match only at the first matching point in the string
2119 that is being searched (the "subject string"). This effect can also be
2120 achieved by appropriate constructs in the pattern itself, which is the
2121 only way to do it in Perl.
2125 If this bit is set, pcre_compile() automatically inserts callout items,
2126 all with number 255, before each pattern item. For discussion of the
2127 callout facility, see the pcrecallout documentation.
2132 These options (which are mutually exclusive) control what the \R escape
2133 sequence matches. The choice is either to match only CR, LF, or CRLF,
2134 or to match any Unicode newline sequence. The default is specified when
2135 PCRE is built. It can be overridden from within the pattern, or by set-
2136 ting an option when a compiled pattern is matched.
2140 If this bit is set, letters in the pattern match both upper and lower
2141 case letters. It is equivalent to Perl's /i option, and it can be
2142 changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE
2143 always understands the concept of case for characters whose values are
2144 less than 128, so caseless matching is always possible. For characters
2145 with higher values, the concept of case is supported if PCRE is com-
2146 piled with Unicode property support, but not otherwise. If you want to
2147 use caseless matching for characters 128 and above, you must ensure
2148 that PCRE is compiled with Unicode property support as well as with
2153 If this bit is set, a dollar metacharacter in the pattern matches only
2154 at the end of the subject string. Without this option, a dollar also
2155 matches immediately before a newline at the end of the string (but not
2156 before any other newlines). The PCRE_DOLLAR_ENDONLY option is ignored
2157 if PCRE_MULTILINE is set. There is no equivalent to this option in
2158 Perl, and no way to set it within a pattern.
2162 If this bit is set, a dot metacharacter in the pattern matches a char-
2163 acter of any value, including one that indicates a newline. However, it
2164 only ever matches one character, even if newlines are coded as CRLF.
2165 Without this option, a dot does not match when the current position is
2166 at a newline. This option is equivalent to Perl's /s option, and it can
2167 be changed within a pattern by a (?s) option setting. A negative class
2168 such as [^a] always matches newline characters, independent of the set-
2169 ting of this option.
2173 If this bit is set, names used to identify capturing subpatterns need
2174 not be unique. This can be helpful for certain types of pattern when it
2175 is known that only one instance of the named subpattern can ever be
2176 matched. There are more details of named subpatterns below; see also
2177 the pcrepattern documentation.
2181 If this bit is set, most white space characters in the pattern are
2182 totally ignored except when escaped or inside a character class. How-
2183 ever, white space is not allowed within sequences such as (?> that
2184 introduce various parenthesized subpatterns, nor within a numerical
2185 quantifier such as {1,3}. However, ignorable white space is permitted
2186 between an item and a following quantifier and between a quantifier and
2187 a following + that indicates possessiveness.
2189 White space did not used to include the VT character (code 11), because
2190 Perl did not treat this character as white space. However, Perl changed
2191 at release 5.18, so PCRE followed at release 8.34, and VT is now
2192 treated as white space.
2194 PCRE_EXTENDED also causes characters between an unescaped # outside a
2195 character class and the next newline, inclusive, to be ignored.
2196 PCRE_EXTENDED is equivalent to Perl's /x option, and it can be changed
2197 within a pattern by a (?x) option setting.
2199 Which characters are interpreted as newlines is controlled by the
2200 options passed to pcre_compile() or by a special sequence at the start
2201 of the pattern, as described in the section entitled "Newline conven-
2202 tions" in the pcrepattern documentation. Note that the end of this type
2203 of comment is a literal newline sequence in the pattern; escape
2204 sequences that happen to represent a newline do not count.
2206 This option makes it possible to include comments inside complicated
2207 patterns. Note, however, that this applies only to data characters.
2208 White space characters may never appear within special character
2209 sequences in a pattern, for example within the sequence (?( that intro-
2210 duces a conditional subpattern.
2214 This option was invented in order to turn on additional functionality
2215 of PCRE that is incompatible with Perl, but it is currently of very
2216 little use. When set, any backslash in a pattern that is followed by a
2217 letter that has no special meaning causes an error, thus reserving
2218 these combinations for future expansion. By default, as in Perl, a
2219 backslash followed by a letter with no special meaning is treated as a
2220 literal. (Perl can, however, be persuaded to give an error for this, by
2221 running it with the -w option.) There are at present no other features
2222 controlled by this option. It can also be set by a (?X) option setting
2227 If this option is set, an unanchored pattern is required to match
2228 before or at the first newline in the subject string, though the
2229 matched text may continue over the newline.
2231 PCRE_JAVASCRIPT_COMPAT
2233 If this option is set, PCRE's behaviour is changed in some ways so that
2234 it is compatible with JavaScript rather than Perl. The changes are as
2237 (1) A lone closing square bracket in a pattern causes a compile-time
2238 error, because this is illegal in JavaScript (by default it is treated
2239 as a data character). Thus, the pattern AB]CD becomes illegal when this
2242 (2) At run time, a back reference to an unset subpattern group matches
2243 an empty string (by default this causes the current matching alterna-
2244 tive to fail). A pattern such as (\1)(a) succeeds when this option is
2245 set (assuming it can find an "a" in the subject), whereas it fails by
2246 default, for Perl compatibility.
2248 (3) \U matches an upper case "U" character; by default \U causes a com-
2249 pile time error (Perl uses \U to upper case subsequent characters).
2251 (4) \u matches a lower case "u" character unless it is followed by four
2252 hexadecimal digits, in which case the hexadecimal number defines the
2253 code point to match. By default, \u causes a compile time error (Perl
2254 uses it to upper case the following character).
2256 (5) \x matches a lower case "x" character unless it is followed by two
2257 hexadecimal digits, in which case the hexadecimal number defines the
2258 code point to match. By default, as in Perl, a hexadecimal number is
2259 always expected after \x, but it may have zero, one, or two digits (so,
2260 for example, \xz matches a binary zero character followed by z).
2264 By default, for the purposes of matching "start of line" and "end of
2265 line", PCRE treats the subject string as consisting of a single line of
2266 characters, even if it actually contains newlines. The "start of line"
2267 metacharacter (^) matches only at the start of the string, and the "end
2268 of line" metacharacter ($) matches only at the end of the string, or
2269 before a terminating newline (except when PCRE_DOLLAR_ENDONLY is set).
2270 Note, however, that unless PCRE_DOTALL is set, the "any character"
2271 metacharacter (.) does not match at a newline. This behaviour (for ^,
2272 $, and dot) is the same as Perl.
2274 When PCRE_MULTILINE it is set, the "start of line" and "end of line"
2275 constructs match immediately following or immediately before internal
2276 newlines in the subject string, respectively, as well as at the very
2277 start and end. This is equivalent to Perl's /m option, and it can be
2278 changed within a pattern by a (?m) option setting. If there are no new-
2279 lines in a subject string, or no occurrences of ^ or $ in a pattern,
2280 setting PCRE_MULTILINE has no effect.
2284 This option locks out interpretation of the pattern as UTF-8 (or UTF-16
2285 or UTF-32 in the 16-bit and 32-bit libraries). In particular, it pre-
2286 vents the creator of the pattern from switching to UTF interpretation
2287 by starting the pattern with (*UTF). This may be useful in applications
2288 that process patterns from external sources. The combination of
2289 PCRE_UTF8 and PCRE_NEVER_UTF also causes an error.
2294 PCRE_NEWLINE_ANYCRLF
2297 These options override the default newline definition that was chosen
2298 when PCRE was built. Setting the first or the second specifies that a
2299 newline is indicated by a single character (CR or LF, respectively).
2300 Setting PCRE_NEWLINE_CRLF specifies that a newline is indicated by the
2301 two-character CRLF sequence. Setting PCRE_NEWLINE_ANYCRLF specifies
2302 that any of the three preceding sequences should be recognized. Setting
2303 PCRE_NEWLINE_ANY specifies that any Unicode newline sequence should be
2306 In an ASCII/Unicode environment, the Unicode newline sequences are the
2307 three just mentioned, plus the single characters VT (vertical tab,
2308 U+000B), FF (form feed, U+000C), NEL (next line, U+0085), LS (line sep-
2309 arator, U+2028), and PS (paragraph separator, U+2029). For the 8-bit
2310 library, the last two are recognized only in UTF-8 mode.
2312 When PCRE is compiled to run in an EBCDIC (mainframe) environment, the
2313 code for CR is 0x0d, the same as ASCII. However, the character code for
2314 LF is normally 0x15, though in some EBCDIC environments 0x25 is used.
2315 Whichever of these is not LF is made to correspond to Unicode's NEL
2316 character. EBCDIC codes are all less than 256. For more details, see
2317 the pcrebuild documentation.
2319 The newline setting in the options word uses three bits that are
2320 treated as a number, giving eight possibilities. Currently only six are
2321 used (default plus the five values above). This means that if you set
2322 more than one newline option, the combination may or may not be sensi-
2323 ble. For example, PCRE_NEWLINE_CR with PCRE_NEWLINE_LF is equivalent to
2324 PCRE_NEWLINE_CRLF, but other combinations may yield unused numbers and
2327 The only time that a line break in a pattern is specially recognized
2328 when compiling is when PCRE_EXTENDED is set. CR and LF are white space
2329 characters, and so are ignored in this mode. Also, an unescaped # out-
2330 side a character class indicates a comment that lasts until after the
2331 next line break sequence. In other circumstances, line break sequences
2332 in patterns are treated as literal data.
2334 The newline option that is set at compile time becomes the default that
2335 is used for pcre_exec() and pcre_dfa_exec(), but it can be overridden.
2337 PCRE_NO_AUTO_CAPTURE
2339 If this option is set, it disables the use of numbered capturing paren-
2340 theses in the pattern. Any opening parenthesis that is not followed by
2341 ? behaves as if it were followed by ?: but named parentheses can still
2342 be used for capturing (and they acquire numbers in the usual way).
2343 There is no equivalent of this option in Perl.
2345 PCRE_NO_AUTO_POSSESS
2347 If this option is set, it disables "auto-possessification". This is an
2348 optimization that, for example, turns a+b into a++b in order to avoid
2349 backtracks into a+ that can never be successful. However, if callouts
2350 are in use, auto-possessification means that some of them are never
2351 taken. You can set this option if you want the matching functions to do
2352 a full unoptimized search and run all the callouts, but it is mainly
2353 provided for testing purposes.
2355 PCRE_NO_START_OPTIMIZE
2357 This is an option that acts at matching time; that is, it is really an
2358 option for pcre_exec() or pcre_dfa_exec(). If it is set at compile
2359 time, it is remembered with the compiled pattern and assumed at match-
2360 ing time. This is necessary if you want to use JIT execution, because
2361 the JIT compiler needs to know whether or not this option is set. For
2362 details see the discussion of PCRE_NO_START_OPTIMIZE below.
2366 This option changes the way PCRE processes \B, \b, \D, \d, \S, \s, \W,
2367 \w, and some of the POSIX character classes. By default, only ASCII
2368 characters are recognized, but if PCRE_UCP is set, Unicode properties
2369 are used instead to classify characters. More details are given in the
2370 section on generic character types in the pcrepattern page. If you set
2371 PCRE_UCP, matching one of the items it affects takes much longer. The
2372 option is available only if PCRE has been compiled with Unicode prop-
2377 This option inverts the "greediness" of the quantifiers so that they
2378 are not greedy by default, but become greedy if followed by "?". It is
2379 not compatible with Perl. It can also be set by a (?U) option setting
2384 This option causes PCRE to regard both the pattern and the subject as
2385 strings of UTF-8 characters instead of single-byte strings. However, it
2386 is available only when PCRE is built to include UTF support. If not,
2387 the use of this option provokes an error. Details of how this option
2388 changes the behaviour of PCRE are given in the pcreunicode page.
2392 When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is
2393 automatically checked. There is a discussion about the validity of
2394 UTF-8 strings in the pcreunicode page. If an invalid UTF-8 sequence is
2395 found, pcre_compile() returns an error. If you already know that your
2396 pattern is valid, and you want to skip this check for performance rea-
2397 sons, you can set the PCRE_NO_UTF8_CHECK option. When it is set, the
2398 effect of passing an invalid UTF-8 string as a pattern is undefined. It
2399 may cause your program to crash or loop. Note that this option can also
2400 be passed to pcre_exec() and pcre_dfa_exec(), to suppress the validity
2401 checking of subject strings only. If the same string is being matched
2402 many times, the option can be safely set for the second and subsequent
2403 matchings to improve performance.
2406 COMPILATION ERROR CODES
2408 The following table lists the error codes than may be returned by
2409 pcre_compile2(), along with the error messages that may be returned by
2410 both compiling functions. Note that error messages are always 8-bit
2411 ASCII strings, even in 16-bit or 32-bit mode. As PCRE has developed,
2412 some error codes have fallen out of use. To avoid confusion, they have
2416 1 \ at end of pattern
2417 2 \c at end of pattern
2418 3 unrecognized character follows \
2419 4 numbers out of order in {} quantifier
2420 5 number too big in {} quantifier
2421 6 missing terminating ] for character class
2422 7 invalid escape sequence in character class
2423 8 range out of order in character class
2425 10 [this code is not in use]
2426 11 internal error: unexpected repeat
2427 12 unrecognized character after (? or (?-
2428 13 POSIX named classes are supported only within a class
2430 15 reference to non-existent subpattern
2431 16 erroffset passed as NULL
2432 17 unknown option bit(s) set
2433 18 missing ) after comment
2434 19 [this code is not in use]
2435 20 regular expression is too large
2436 21 failed to get memory
2437 22 unmatched parentheses
2438 23 internal error: code overflow
2439 24 unrecognized character after (?<
2440 25 lookbehind assertion is not fixed length
2441 26 malformed number or name after (?(
2442 27 conditional group contains more than two branches
2443 28 assertion expected after (?(
2444 29 (?R or (?[+-]digits must be followed by )
2445 30 unknown POSIX class name
2446 31 POSIX collating elements are not supported
2447 32 this version of PCRE is compiled without UTF support
2448 33 [this code is not in use]
2449 34 character value in \x{} or \o{} is too large
2450 35 invalid condition (?(0)
2451 36 \C not allowed in lookbehind assertion
2452 37 PCRE does not support \L, \l, \N{name}, \U, or \u
2453 38 number after (?C is > 255
2454 39 closing ) for (?C expected
2455 40 recursive call could loop indefinitely
2456 41 unrecognized character after (?P
2457 42 syntax error in subpattern name (missing terminator)
2458 43 two named subpatterns have the same name
2459 44 invalid UTF-8 string (specifically UTF-8)
2460 45 support for \P, \p, and \X has not been compiled
2461 46 malformed \P or \p sequence
2462 47 unknown property name after \P or \p
2463 48 subpattern name is too long (maximum 32 characters)
2464 49 too many named subpatterns (maximum 10000)
2465 50 [this code is not in use]
2466 51 octal value is greater than \377 in 8-bit non-UTF-8 mode
2467 52 internal error: overran compiling workspace
2468 53 internal error: previously-checked referenced subpattern
2470 54 DEFINE group contains more than one branch
2471 55 repeating a DEFINE group is not allowed
2472 56 inconsistent NEWLINE options
2473 57 \g is not followed by a braced, angle-bracketed, or quoted
2474 name/number or by a plain number
2475 58 a numbered reference must not be zero
2476 59 an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)
2477 60 (*VERB) not recognized or malformed
2478 61 number is too big
2479 62 subpattern name expected
2480 63 digit expected after (?+
2481 64 ] is an invalid data character in JavaScript compatibility mode
2482 65 different names for subpatterns of the same number are
2484 66 (*MARK) must have an argument
2485 67 this version of PCRE is not compiled with Unicode property
2487 68 \c must be followed by an ASCII character
2488 69 \k is not followed by a braced, angle-bracketed, or quoted name
2489 70 internal error: unknown opcode in find_fixedlength()
2490 71 \N is not supported in a class
2491 72 too many forward references
2492 73 disallowed Unicode code point (>= 0xd800 && <= 0xdfff)
2493 74 invalid UTF-16 string (specifically UTF-16)
2494 75 name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)
2495 76 character value in \u.... sequence is too large
2496 77 invalid UTF-32 string (specifically UTF-32)
2497 78 setting UTF is disabled by the application
2498 79 non-hex character in \x{} (closing brace missing?)
2499 80 non-octal character in \o{} (closing brace missing?)
2500 81 missing opening brace after \o
2501 82 parentheses are too deeply nested
2502 83 invalid range in character class
2503 84 group name must start with a non-digit
2504 85 parentheses are too deeply nested (stack check)
2506 The numbers 32 and 10000 in errors 48 and 49 are defaults; different
2507 values may be used if the limits were changed when PCRE was built.
2512 pcre_extra *pcre_study(const pcre *code, int options,
2513 const char **errptr);
2515 If a compiled pattern is going to be used several times, it is worth
2516 spending more time analyzing it in order to speed up the time taken for
2517 matching. The function pcre_study() takes a pointer to a compiled pat-
2518 tern as its first argument. If studying the pattern produces additional
2519 information that will help speed up matching, pcre_study() returns a
2520 pointer to a pcre_extra block, in which the study_data field points to
2521 the results of the study.
2523 The returned value from pcre_study() can be passed directly to
2524 pcre_exec() or pcre_dfa_exec(). However, a pcre_extra block also con-
2525 tains other fields that can be set by the caller before the block is
2526 passed; these are described below in the section on matching a pattern.
2528 If studying the pattern does not produce any useful information,
2529 pcre_study() returns NULL by default. In that circumstance, if the
2530 calling program wants to pass any of the other fields to pcre_exec() or
2531 pcre_dfa_exec(), it must set up its own pcre_extra block. However, if
2532 pcre_study() is called with the PCRE_STUDY_EXTRA_NEEDED option, it
2533 returns a pcre_extra block even if studying did not find any additional
2534 information. It may still return NULL, however, if an error occurs in
2537 The second argument of pcre_study() contains option bits. There are
2538 three further options in addition to PCRE_STUDY_EXTRA_NEEDED:
2540 PCRE_STUDY_JIT_COMPILE
2541 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
2542 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
2544 If any of these are set, and the just-in-time compiler is available,
2545 the pattern is further compiled into machine code that executes much
2546 faster than the pcre_exec() interpretive matching function. If the
2547 just-in-time compiler is not available, these options are ignored. All
2548 undefined bits in the options argument must be zero.
2550 JIT compilation is a heavyweight optimization. It can take some time
2551 for patterns to be analyzed, and for one-off matches and simple pat-
2552 terns the benefit of faster execution might be offset by a much slower
2553 study time. Not all patterns can be optimized by the JIT compiler. For
2554 those that cannot be handled, matching automatically falls back to the
2555 pcre_exec() interpreter. For more details, see the pcrejit documenta-
2558 The third argument for pcre_study() is a pointer for an error message.
2559 If studying succeeds (even if no data is returned), the variable it
2560 points to is set to NULL. Otherwise it is set to point to a textual
2561 error message. This is a static string that is part of the library. You
2562 must not try to free it. You should test the error pointer for NULL
2563 after calling pcre_study(), to be sure that it has run successfully.
2565 When you are finished with a pattern, you can free the memory used for
2566 the study data by calling pcre_free_study(). This function was added to
2567 the API for release 8.20. For earlier versions, the memory could be
2568 freed with pcre_free(), just like the pattern itself. This will still
2569 work in cases where JIT optimization is not used, but it is advisable
2570 to change to the new function when convenient.
2572 This is a typical way in which pcre_study() is used (except that in a
2573 real application there should be tests for errors):
2578 re = pcre_compile("pattern", 0, &error, &erroroffset, NULL);
2580 re, /* result of pcre_compile() */
2582 &error); /* set to NULL or points to a message */
2583 rc = pcre_exec( /* see below for details of pcre_exec() options */
2584 re, sd, "subject", 7, 0, 0, ovector, 30);
2586 pcre_free_study(sd);
2589 Studying a pattern does two things: first, a lower bound for the length
2590 of subject string that is needed to match the pattern is computed. This
2591 does not mean that there are any strings of that length that match, but
2592 it does guarantee that no shorter strings match. The value is used to
2593 avoid wasting time by trying to match strings that are shorter than the
2594 lower bound. You can find out the value in a calling program via the
2595 pcre_fullinfo() function.
2597 Studying a pattern is also useful for non-anchored patterns that do not
2598 have a single fixed starting character. A bitmap of possible starting
2599 bytes is created. This speeds up finding a position in the subject at
2600 which to start matching. (In 16-bit mode, the bitmap is used for 16-bit
2601 values less than 256. In 32-bit mode, the bitmap is used for 32-bit
2602 values less than 256.)
2604 These two optimizations apply to both pcre_exec() and pcre_dfa_exec(),
2605 and the information is also used by the JIT compiler. The optimiza-
2606 tions can be disabled by setting the PCRE_NO_START_OPTIMIZE option.
2607 You might want to do this if your pattern contains callouts or (*MARK)
2608 and you want to make use of these facilities in cases where matching
2611 PCRE_NO_START_OPTIMIZE can be specified at either compile time or exe-
2612 cution time. However, if PCRE_NO_START_OPTIMIZE is passed to
2613 pcre_exec(), (that is, after any JIT compilation has happened) JIT exe-
2614 cution is disabled. For JIT execution to work with PCRE_NO_START_OPTI-
2615 MIZE, the option must be set at compile time.
2617 There is a longer discussion of PCRE_NO_START_OPTIMIZE below.
2622 PCRE handles caseless matching, and determines whether characters are
2623 letters, digits, or whatever, by reference to a set of tables, indexed
2624 by character code point. When running in UTF-8 mode, or in the 16- or
2625 32-bit libraries, this applies only to characters with code points less
2626 than 256. By default, higher-valued code points never match escapes
2627 such as \w or \d. However, if PCRE is built with Unicode property sup-
2628 port, all characters can be tested with \p and \P, or, alternatively,
2629 the PCRE_UCP option can be set when a pattern is compiled; this causes
2630 \w and friends to use Unicode property support instead of the built-in
2633 The use of locales with Unicode is discouraged. If you are handling
2634 characters with code points greater than 128, you should either use
2635 Unicode support, or use locales, but not try to mix the two.
2637 PCRE contains an internal set of tables that are used when the final
2638 argument of pcre_compile() is NULL. These are sufficient for many
2639 applications. Normally, the internal tables recognize only ASCII char-
2640 acters. However, when PCRE is built, it is possible to cause the inter-
2641 nal tables to be rebuilt in the default "C" locale of the local system,
2642 which may cause them to be different.
2644 The internal tables can always be overridden by tables supplied by the
2645 application that calls PCRE. These may be created in a different locale
2646 from the default. As more and more applications change to using Uni-
2647 code, the need for this locale support is expected to die away.
2649 External tables are built by calling the pcre_maketables() function,
2650 which has no arguments, in the relevant locale. The result can then be
2651 passed to pcre_compile() as often as necessary. For example, to build
2652 and use tables that are appropriate for the French locale (where
2653 accented characters with values greater than 128 are treated as let-
2654 ters), the following code could be used:
2656 setlocale(LC_CTYPE, "fr_FR");
2657 tables = pcre_maketables();
2658 re = pcre_compile(..., tables);
2660 The locale name "fr_FR" is used on Linux and other Unix-like systems;
2661 if you are using Windows, the name for the French locale is "french".
2663 When pcre_maketables() runs, the tables are built in memory that is
2664 obtained via pcre_malloc. It is the caller's responsibility to ensure
2665 that the memory containing the tables remains available for as long as
2668 The pointer that is passed to pcre_compile() is saved with the compiled
2669 pattern, and the same tables are used via this pointer by pcre_study()
2670 and also by pcre_exec() and pcre_dfa_exec(). Thus, for any single pat-
2671 tern, compilation, studying and matching all happen in the same locale,
2672 but different patterns can be processed in different locales.
2674 It is possible to pass a table pointer or NULL (indicating the use of
2675 the internal tables) to pcre_exec() or pcre_dfa_exec() (see the discus-
2676 sion below in the section on matching a pattern). This facility is pro-
2677 vided for use with pre-compiled patterns that have been saved and
2678 reloaded. Character tables are not saved with patterns, so if a non-
2679 standard table was used at compile time, it must be provided again when
2680 the reloaded pattern is matched. Attempting to use this facility to
2681 match a pattern in a different locale from the one in which it was com-
2682 piled is likely to lead to anomalous (usually incorrect) results.
2685 INFORMATION ABOUT A PATTERN
2687 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
2688 int what, void *where);
2690 The pcre_fullinfo() function returns information about a compiled pat-
2691 tern. It replaces the pcre_info() function, which was removed from the
2692 library at version 8.30, after more than 10 years of obsolescence.
2694 The first argument for pcre_fullinfo() is a pointer to the compiled
2695 pattern. The second argument is the result of pcre_study(), or NULL if
2696 the pattern was not studied. The third argument specifies which piece
2697 of information is required, and the fourth argument is a pointer to a
2698 variable to receive the data. The yield of the function is zero for
2699 success, or one of the following negative numbers:
2701 PCRE_ERROR_NULL the argument code was NULL
2702 the argument where was NULL
2703 PCRE_ERROR_BADMAGIC the "magic number" was not found
2704 PCRE_ERROR_BADENDIANNESS the pattern was compiled with different
2706 PCRE_ERROR_BADOPTION the value of what was invalid
2707 PCRE_ERROR_UNSET the requested field is not set
2709 The "magic number" is placed at the start of each compiled pattern as
2710 an simple check against passing an arbitrary memory pointer. The endi-
2711 anness error can occur if a compiled pattern is saved and reloaded on a
2712 different host. Here is a typical call of pcre_fullinfo(), to obtain
2713 the length of the compiled pattern:
2718 re, /* result of pcre_compile() */
2719 sd, /* result of pcre_study(), or NULL */
2720 PCRE_INFO_SIZE, /* what is required */
2721 &length); /* where to put the data */
2723 The possible values for the third argument are defined in pcre.h, and
2726 PCRE_INFO_BACKREFMAX
2728 Return the number of the highest back reference in the pattern. The
2729 fourth argument should point to an int variable. Zero is returned if
2730 there are no back references.
2732 PCRE_INFO_CAPTURECOUNT
2734 Return the number of capturing subpatterns in the pattern. The fourth
2735 argument should point to an int variable.
2737 PCRE_INFO_DEFAULT_TABLES
2739 Return a pointer to the internal default character tables within PCRE.
2740 The fourth argument should point to an unsigned char * variable. This
2741 information call is provided for internal use by the pcre_study() func-
2742 tion. External callers can cause PCRE to use its internal tables by
2743 passing a NULL table pointer.
2745 PCRE_INFO_FIRSTBYTE (deprecated)
2747 Return information about the first data unit of any matched string, for
2748 a non-anchored pattern. The name of this option refers to the 8-bit
2749 library, where data units are bytes. The fourth argument should point
2750 to an int variable. Negative values are used for special cases. How-
2751 ever, this means that when the 32-bit library is in non-UTF-32 mode,
2752 the full 32-bit range of characters cannot be returned. For this rea-
2753 son, this value is deprecated; use PCRE_INFO_FIRSTCHARACTERFLAGS and
2754 PCRE_INFO_FIRSTCHARACTER instead.
2756 If there is a fixed first value, for example, the letter "c" from a
2757 pattern such as (cat|cow|coyote), its value is returned. In the 8-bit
2758 library, the value is always less than 256. In the 16-bit library the
2759 value can be up to 0xffff. In the 32-bit library the value can be up to
2762 If there is no fixed first value, and if either
2764 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
2765 branch starts with "^", or
2767 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
2768 set (if it were set, the pattern would be anchored),
2770 -1 is returned, indicating that the pattern matches only at the start
2771 of a subject string or after any newline within the string. Otherwise
2772 -2 is returned. For anchored patterns, -2 is returned.
2774 PCRE_INFO_FIRSTCHARACTER
2776 Return the value of the first data unit (non-UTF character) of any
2777 matched string in the situation where PCRE_INFO_FIRSTCHARACTERFLAGS
2778 returns 1; otherwise return 0. The fourth argument should point to an
2781 In the 8-bit library, the value is always less than 256. In the 16-bit
2782 library the value can be up to 0xffff. In the 32-bit library in UTF-32
2783 mode the value can be up to 0x10ffff, and up to 0xffffffff when not
2786 PCRE_INFO_FIRSTCHARACTERFLAGS
2788 Return information about the first data unit of any matched string, for
2789 a non-anchored pattern. The fourth argument should point to an int
2792 If there is a fixed first value, for example, the letter "c" from a
2793 pattern such as (cat|cow|coyote), 1 is returned, and the character
2794 value can be retrieved using PCRE_INFO_FIRSTCHARACTER. If there is no
2795 fixed first value, and if either
2797 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
2798 branch starts with "^", or
2800 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
2801 set (if it were set, the pattern would be anchored),
2803 2 is returned, indicating that the pattern matches only at the start of
2804 a subject string or after any newline within the string. Otherwise 0 is
2805 returned. For anchored patterns, 0 is returned.
2807 PCRE_INFO_FIRSTTABLE
2809 If the pattern was studied, and this resulted in the construction of a
2810 256-bit table indicating a fixed set of values for the first data unit
2811 in any matching string, a pointer to the table is returned. Otherwise
2812 NULL is returned. The fourth argument should point to an unsigned char
2817 Return 1 if the pattern contains any explicit matches for CR or LF
2818 characters, otherwise 0. The fourth argument should point to an int
2819 variable. An explicit match is either a literal CR or LF character, or
2824 Return 1 if the (?J) or (?-J) option setting is used in the pattern,
2825 otherwise 0. The fourth argument should point to an int variable. (?J)
2826 and (?-J) set and unset the local PCRE_DUPNAMES option, respectively.
2830 Return 1 if the pattern was studied with one of the JIT options, and
2831 just-in-time compiling was successful. The fourth argument should point
2832 to an int variable. A return value of 0 means that JIT support is not
2833 available in this version of PCRE, or that the pattern was not studied
2834 with a JIT option, or that the JIT compiler could not handle this par-
2835 ticular pattern. See the pcrejit documentation for details of what can
2836 and cannot be handled.
2840 If the pattern was successfully studied with a JIT option, return the
2841 size of the JIT compiled code, otherwise return zero. The fourth argu-
2842 ment should point to a size_t variable.
2844 PCRE_INFO_LASTLITERAL
2846 Return the value of the rightmost literal data unit that must exist in
2847 any matched string, other than at its start, if such a value has been
2848 recorded. The fourth argument should point to an int variable. If there
2849 is no such value, -1 is returned. For anchored patterns, a last literal
2850 value is recorded only if it follows something of variable length. For
2851 example, for the pattern /^a\d+z\d+/ the returned value is "z", but for
2852 /^a\dz\d/ the returned value is -1.
2854 Since for the 32-bit library using the non-UTF-32 mode, this function
2855 is unable to return the full 32-bit range of characters, this value is
2856 deprecated; instead the PCRE_INFO_REQUIREDCHARFLAGS and
2857 PCRE_INFO_REQUIREDCHAR values should be used.
2859 PCRE_INFO_MATCH_EMPTY
2861 Return 1 if the pattern can match an empty string, otherwise 0. The
2862 fourth argument should point to an int variable.
2864 PCRE_INFO_MATCHLIMIT
2866 If the pattern set a match limit by including an item of the form
2867 (*LIMIT_MATCH=nnnn) at the start, the value is returned. The fourth
2868 argument should point to an unsigned 32-bit integer. If no such value
2869 has been set, the call to pcre_fullinfo() returns the error
2872 PCRE_INFO_MAXLOOKBEHIND
2874 Return the number of characters (NB not data units) in the longest
2875 lookbehind assertion in the pattern. This information is useful when
2876 doing multi-segment matching using the partial matching facilities.
2877 Note that the simple assertions \b and \B require a one-character look-
2878 behind. \A also registers a one-character lookbehind, though it does
2879 not actually inspect the previous character. This is to ensure that at
2880 least one character from the old segment is retained when a new segment
2881 is processed. Otherwise, if there are no lookbehinds in the pattern, \A
2882 might match incorrectly at the start of a new segment.
2886 If the pattern was studied and a minimum length for matching subject
2887 strings was computed, its value is returned. Otherwise the returned
2888 value is -1. The value is a number of characters, which in UTF mode may
2889 be different from the number of data units. The fourth argument should
2890 point to an int variable. A non-negative value is a lower bound to the
2891 length of any matching string. There may not be any strings of that
2892 length that do actually match, but every string that does match is at
2896 PCRE_INFO_NAMEENTRYSIZE
2899 PCRE supports the use of named as well as numbered capturing parenthe-
2900 ses. The names are just an additional way of identifying the parenthe-
2901 ses, which still acquire numbers. Several convenience functions such as
2902 pcre_get_named_substring() are provided for extracting captured sub-
2903 strings by name. It is also possible to extract the data directly, by
2904 first converting the name to a number in order to access the correct
2905 pointers in the output vector (described with pcre_exec() below). To do
2906 the conversion, you need to use the name-to-number map, which is
2907 described by these three values.
2909 The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT
2910 gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size
2911 of each entry; both of these return an int value. The entry size
2912 depends on the length of the longest name. PCRE_INFO_NAMETABLE returns
2913 a pointer to the first entry of the table. This is a pointer to char in
2914 the 8-bit library, where the first two bytes of each entry are the num-
2915 ber of the capturing parenthesis, most significant byte first. In the
2916 16-bit library, the pointer points to 16-bit data units, the first of
2917 which contains the parenthesis number. In the 32-bit library, the
2918 pointer points to 32-bit data units, the first of which contains the
2919 parenthesis number. The rest of the entry is the corresponding name,
2922 The names are in alphabetical order. If (?| is used to create multiple
2923 groups with the same number, as described in the section on duplicate
2924 subpattern numbers in the pcrepattern page, the groups may be given the
2925 same name, but there is only one entry in the table. Different names
2926 for groups of the same number are not permitted. Duplicate names for
2927 subpatterns with different numbers are permitted, but only if PCRE_DUP-
2928 NAMES is set. They appear in the table in the order in which they were
2929 found in the pattern. In the absence of (?| this is the order of
2930 increasing number; when (?| is used this is not necessarily the case
2931 because later subpatterns may have lower numbers.
2933 As a simple example of the name/number table, consider the following
2934 pattern after compilation by the 8-bit library (assume PCRE_EXTENDED is
2935 set, so white space - including newlines - is ignored):
2937 (?<date> (?<year>(\d\d)?\d\d) -
2938 (?<month>\d\d) - (?<day>\d\d) )
2940 There are four named subpatterns, so the table has four entries, and
2941 each entry in the table is eight bytes long. The table is as follows,
2942 with non-printing bytes shows in hexadecimal, and undefined bytes shown
2946 00 05 d a y 00 ?? ??
2950 When writing code to extract data from named subpatterns using the
2951 name-to-number map, remember that the length of the entries is likely
2952 to be different for each compiled pattern.
2956 Return 1 if the pattern can be used for partial matching with
2957 pcre_exec(), otherwise 0. The fourth argument should point to an int
2958 variable. From release 8.00, this always returns 1, because the
2959 restrictions that previously applied to partial matching have been
2960 lifted. The pcrepartial documentation gives details of partial match-
2965 Return a copy of the options with which the pattern was compiled. The
2966 fourth argument should point to an unsigned long int variable. These
2967 option bits are those specified in the call to pcre_compile(), modified
2968 by any top-level option settings at the start of the pattern itself. In
2969 other words, they are the options that will be in force when matching
2970 starts. For example, if the pattern /(?im)abc(?-i)d/ is compiled with
2971 the PCRE_EXTENDED option, the result is PCRE_CASELESS, PCRE_MULTILINE,
2974 A pattern is automatically anchored by PCRE if all of its top-level
2975 alternatives begin with one of the following:
2977 ^ unless PCRE_MULTILINE is set
2980 .* if PCRE_DOTALL is set and there are no back
2981 references to the subpattern in which .* appears
2983 For such patterns, the PCRE_ANCHORED bit is set in the options returned
2986 PCRE_INFO_RECURSIONLIMIT
2988 If the pattern set a recursion limit by including an item of the form
2989 (*LIMIT_RECURSION=nnnn) at the start, the value is returned. The fourth
2990 argument should point to an unsigned 32-bit integer. If no such value
2991 has been set, the call to pcre_fullinfo() returns the error
2996 Return the size of the compiled pattern in bytes (for all three
2997 libraries). The fourth argument should point to a size_t variable. This
2998 value does not include the size of the pcre structure that is returned
2999 by pcre_compile(). The value that is passed as the argument to
3000 pcre_malloc() when pcre_compile() is getting memory in which to place
3001 the compiled data is the value returned by this option plus the size of
3002 the pcre structure. Studying a compiled pattern, with or without JIT,
3003 does not alter the value returned by this option.
3007 Return the size in bytes (for all three libraries) of the data block
3008 pointed to by the study_data field in a pcre_extra block. If pcre_extra
3009 is NULL, or there is no study data, zero is returned. The fourth argu-
3010 ment should point to a size_t variable. The study_data field is set by
3011 pcre_study() to record information that will speed up matching (see the
3012 section entitled "Studying a pattern" above). The format of the
3013 study_data block is private, but its length is made available via this
3014 option so that it can be saved and restored (see the pcreprecompile
3015 documentation for details).
3017 PCRE_INFO_REQUIREDCHARFLAGS
3019 Returns 1 if there is a rightmost literal data unit that must exist in
3020 any matched string, other than at its start. The fourth argument should
3021 point to an int variable. If there is no such value, 0 is returned. If
3022 returning 1, the character value itself can be retrieved using
3023 PCRE_INFO_REQUIREDCHAR.
3025 For anchored patterns, a last literal value is recorded only if it fol-
3026 lows something of variable length. For example, for the pattern
3027 /^a\d+z\d+/ the returned value 1 (with "z" returned from
3028 PCRE_INFO_REQUIREDCHAR), but for /^a\dz\d/ the returned value is 0.
3030 PCRE_INFO_REQUIREDCHAR
3032 Return the value of the rightmost literal data unit that must exist in
3033 any matched string, other than at its start, if such a value has been
3034 recorded. The fourth argument should point to an uint32_t variable. If
3035 there is no such value, 0 is returned.
3040 int pcre_refcount(pcre *code, int adjust);
3042 The pcre_refcount() function is used to maintain a reference count in
3043 the data block that contains a compiled pattern. It is provided for the
3044 benefit of applications that operate in an object-oriented manner,
3045 where different parts of the application may be using the same compiled
3046 pattern, but you want to free the block when they are all done.
3048 When a pattern is compiled, the reference count field is initialized to
3049 zero. It is changed only by calling this function, whose action is to
3050 add the adjust value (which may be positive or negative) to it. The
3051 yield of the function is the new value. However, the value of the count
3052 is constrained to lie between 0 and 65535, inclusive. If the new value
3053 is outside these limits, it is forced to the appropriate limit value.
3055 Except when it is zero, the reference count is not correctly preserved
3056 if a pattern is compiled on one host and then transferred to a host
3057 whose byte-order is different. (This seems a highly unlikely scenario.)
3060 MATCHING A PATTERN: THE TRADITIONAL FUNCTION
3062 int pcre_exec(const pcre *code, const pcre_extra *extra,
3063 const char *subject, int length, int startoffset,
3064 int options, int *ovector, int ovecsize);
3066 The function pcre_exec() is called to match a subject string against a
3067 compiled pattern, which is passed in the code argument. If the pattern
3068 was studied, the result of the study should be passed in the extra
3069 argument. You can call pcre_exec() with the same code and extra argu-
3070 ments as many times as you like, in order to match different subject
3071 strings with the same pattern.
3073 This function is the main matching facility of the library, and it
3074 operates in a Perl-like manner. For specialist use there is also an
3075 alternative matching function, which is described below in the section
3076 about the pcre_dfa_exec() function.
3078 In most applications, the pattern will have been compiled (and option-
3079 ally studied) in the same process that calls pcre_exec(). However, it
3080 is possible to save compiled patterns and study data, and then use them
3081 later in different processes, possibly even on different hosts. For a
3082 discussion about this, see the pcreprecompile documentation.
3084 Here is an example of a simple call to pcre_exec():
3089 re, /* result of pcre_compile() */
3090 NULL, /* we didn't study the pattern */
3091 "some string", /* the subject string */
3092 11, /* the length of the subject string */
3093 0, /* start at offset 0 in the subject */
3094 0, /* default options */
3095 ovector, /* vector of integers for substring information */
3096 30); /* number of elements (NOT size in bytes) */
3098 Extra data for pcre_exec()
3100 If the extra argument is not NULL, it must point to a pcre_extra data
3101 block. The pcre_study() function returns such a block (when it doesn't
3102 return NULL), but you can also create one for yourself, and pass addi-
3103 tional information in it. The pcre_extra block contains the following
3104 fields (not necessarily in this order):
3106 unsigned long int flags;
3108 void *executable_jit;
3109 unsigned long int match_limit;
3110 unsigned long int match_limit_recursion;
3112 const unsigned char *tables;
3113 unsigned char **mark;
3115 In the 16-bit version of this structure, the mark field has type
3118 In the 32-bit version of this structure, the mark field has type
3121 The flags field is used to specify which of the other fields are set.
3124 PCRE_EXTRA_CALLOUT_DATA
3125 PCRE_EXTRA_EXECUTABLE_JIT
3127 PCRE_EXTRA_MATCH_LIMIT
3128 PCRE_EXTRA_MATCH_LIMIT_RECURSION
3129 PCRE_EXTRA_STUDY_DATA
3132 Other flag bits should be set to zero. The study_data field and some-
3133 times the executable_jit field are set in the pcre_extra block that is
3134 returned by pcre_study(), together with the appropriate flag bits. You
3135 should not set these yourself, but you may add to the block by setting
3136 other fields and their corresponding flag bits.
3138 The match_limit field provides a means of preventing PCRE from using up
3139 a vast amount of resources when running patterns that are not going to
3140 match, but which have a very large number of possibilities in their
3141 search trees. The classic example is a pattern that uses nested unlim-
3144 Internally, pcre_exec() uses a function called match(), which it calls
3145 repeatedly (sometimes recursively). The limit set by match_limit is
3146 imposed on the number of times this function is called during a match,
3147 which has the effect of limiting the amount of backtracking that can
3148 take place. For patterns that are not anchored, the count restarts from
3149 zero for each position in the subject string.
3151 When pcre_exec() is called with a pattern that was successfully studied
3152 with a JIT option, the way that the matching is executed is entirely
3153 different. However, there is still the possibility of runaway matching
3154 that goes on for a very long time, and so the match_limit value is also
3155 used in this case (but in a different way) to limit how long the match-
3158 The default value for the limit can be set when PCRE is built; the
3159 default default is 10 million, which handles all but the most extreme
3160 cases. You can override the default by suppling pcre_exec() with a
3161 pcre_extra block in which match_limit is set, and
3162 PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is
3163 exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.
3165 A value for the match limit may also be supplied by an item at the
3166 start of a pattern of the form
3170 where d is a decimal number. However, such a setting is ignored unless
3171 d is less than the limit set by the caller of pcre_exec() or, if no
3172 such limit is set, less than the default.
3174 The match_limit_recursion field is similar to match_limit, but instead
3175 of limiting the total number of times that match() is called, it limits
3176 the depth of recursion. The recursion depth is a smaller number than
3177 the total number of calls, because not all calls to match() are recur-
3178 sive. This limit is of use only if it is set smaller than match_limit.
3180 Limiting the recursion depth limits the amount of machine stack that
3181 can be used, or, when PCRE has been compiled to use memory on the heap
3182 instead of the stack, the amount of heap memory that can be used. This
3183 limit is not relevant, and is ignored, when matching is done using JIT
3186 The default value for match_limit_recursion can be set when PCRE is
3187 built; the default default is the same value as the default for
3188 match_limit. You can override the default by suppling pcre_exec() with
3189 a pcre_extra block in which match_limit_recursion is set, and
3190 PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the
3191 limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT.
3193 A value for the recursion limit may also be supplied by an item at the
3194 start of a pattern of the form
3196 (*LIMIT_RECURSION=d)
3198 where d is a decimal number. However, such a setting is ignored unless
3199 d is less than the limit set by the caller of pcre_exec() or, if no
3200 such limit is set, less than the default.
3202 The callout_data field is used in conjunction with the "callout" fea-
3203 ture, and is described in the pcrecallout documentation.
3205 The tables field is provided for use with patterns that have been pre-
3206 compiled using custom character tables, saved to disc or elsewhere, and
3207 then reloaded, because the tables that were used to compile a pattern
3208 are not saved with it. See the pcreprecompile documentation for a dis-
3209 cussion of saving compiled patterns for later use. If NULL is passed
3210 using this mechanism, it forces PCRE's internal tables to be used.
3212 Warning: The tables that pcre_exec() uses must be the same as those
3213 that were used when the pattern was compiled. If this is not the case,
3214 the behaviour of pcre_exec() is undefined. Therefore, when a pattern is
3215 compiled and matched in the same process, this field should never be
3216 set. In this (the most common) case, the correct table pointer is auto-
3217 matically passed with the compiled pattern from pcre_compile() to
3220 If PCRE_EXTRA_MARK is set in the flags field, the mark field must be
3221 set to point to a suitable variable. If the pattern contains any back-
3222 tracking control verbs such as (*MARK:NAME), and the execution ends up
3223 with a name to pass back, a pointer to the name string (zero termi-
3224 nated) is placed in the variable pointed to by the mark field. The
3225 names are within the compiled pattern; if you wish to retain such a
3226 name you must copy it before freeing the memory of a compiled pattern.
3227 If there is no name to pass back, the variable pointed to by the mark
3228 field is set to NULL. For details of the backtracking control verbs,
3229 see the section entitled "Backtracking control" in the pcrepattern doc-
3232 Option bits for pcre_exec()
3234 The unused bits of the options argument for pcre_exec() must be zero.
3235 The only bits that may be set are PCRE_ANCHORED, PCRE_NEWLINE_xxx,
3236 PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART,
3237 PCRE_NO_START_OPTIMIZE, PCRE_NO_UTF8_CHECK, PCRE_PARTIAL_HARD, and
3240 If the pattern was successfully studied with one of the just-in-time
3241 (JIT) compile options, the only supported options for JIT execution are
3242 PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY,
3243 PCRE_NOTEMPTY_ATSTART, PCRE_PARTIAL_HARD, and PCRE_PARTIAL_SOFT. If an
3244 unsupported option is used, JIT execution is disabled and the normal
3245 interpretive code in pcre_exec() is run.
3249 The PCRE_ANCHORED option limits pcre_exec() to matching at the first
3250 matching position. If a pattern was compiled with PCRE_ANCHORED, or
3251 turned out to be anchored by virtue of its contents, it cannot be made
3252 unachored at matching time.
3257 These options (which are mutually exclusive) control what the \R escape
3258 sequence matches. The choice is either to match only CR, LF, or CRLF,
3259 or to match any Unicode newline sequence. These options override the
3260 choice that was made or defaulted when the pattern was compiled.
3265 PCRE_NEWLINE_ANYCRLF
3268 These options override the newline definition that was chosen or
3269 defaulted when the pattern was compiled. For details, see the descrip-
3270 tion of pcre_compile() above. During matching, the newline choice
3271 affects the behaviour of the dot, circumflex, and dollar metacharac-
3272 ters. It may also alter the way the match position is advanced after a
3273 match failure for an unanchored pattern.
3275 When PCRE_NEWLINE_CRLF, PCRE_NEWLINE_ANYCRLF, or PCRE_NEWLINE_ANY is
3276 set, and a match attempt for an unanchored pattern fails when the cur-
3277 rent position is at a CRLF sequence, and the pattern contains no
3278 explicit matches for CR or LF characters, the match position is
3279 advanced by two characters instead of one, in other words, to after the
3282 The above rule is a compromise that makes the most common cases work as
3283 expected. For example, if the pattern is .+A (and the PCRE_DOTALL
3284 option is not set), it does not match the string "\r\nA" because, after
3285 failing at the start, it skips both the CR and the LF before retrying.
3286 However, the pattern [\r\n]A does match that string, because it con-
3287 tains an explicit CR or LF reference, and so advances only by one char-
3288 acter after the first failure.
3290 An explicit match for CR of LF is either a literal appearance of one of
3291 those characters, or one of the \r or \n escape sequences. Implicit
3292 matches such as [^X] do not count, nor does \s (which includes CR and
3293 LF in the characters that it matches).
3295 Notwithstanding the above, anomalous effects may still occur when CRLF
3296 is a valid newline sequence and explicit \r or \n escapes appear in the
3301 This option specifies that first character of the subject string is not
3302 the beginning of a line, so the circumflex metacharacter should not
3303 match before it. Setting this without PCRE_MULTILINE (at compile time)
3304 causes circumflex never to match. This option affects only the behav-
3305 iour of the circumflex metacharacter. It does not affect \A.
3309 This option specifies that the end of the subject string is not the end
3310 of a line, so the dollar metacharacter should not match it nor (except
3311 in multiline mode) a newline immediately before it. Setting this with-
3312 out PCRE_MULTILINE (at compile time) causes dollar never to match. This
3313 option affects only the behaviour of the dollar metacharacter. It does
3314 not affect \Z or \z.
3318 An empty string is not considered to be a valid match if this option is
3319 set. If there are alternatives in the pattern, they are tried. If all
3320 the alternatives match the empty string, the entire match fails. For
3321 example, if the pattern
3325 is applied to a string not beginning with "a" or "b", it matches an
3326 empty string at the start of the subject. With PCRE_NOTEMPTY set, this
3327 match is not valid, so PCRE searches further into the string for occur-
3328 rences of "a" or "b".
3330 PCRE_NOTEMPTY_ATSTART
3332 This is like PCRE_NOTEMPTY, except that an empty string match that is
3333 not at the start of the subject is permitted. If the pattern is
3334 anchored, such a match can occur only if the pattern contains \K.
3336 Perl has no direct equivalent of PCRE_NOTEMPTY or
3337 PCRE_NOTEMPTY_ATSTART, but it does make a special case of a pattern
3338 match of the empty string within its split() function, and when using
3339 the /g modifier. It is possible to emulate Perl's behaviour after
3340 matching a null string by first trying the match again at the same off-
3341 set with PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED, and then if that
3342 fails, by advancing the starting offset (see below) and trying an ordi-
3343 nary match again. There is some code that demonstrates how to do this
3344 in the pcredemo sample program. In the most general case, you have to
3345 check to see if the newline convention recognizes CRLF as a newline,
3346 and if so, and the current character is CR followed by LF, advance the
3347 starting offset by two characters instead of one.
3349 PCRE_NO_START_OPTIMIZE
3351 There are a number of optimizations that pcre_exec() uses at the start
3352 of a match, in order to speed up the process. For example, if it is
3353 known that an unanchored match must start with a specific character, it
3354 searches the subject for that character, and fails immediately if it
3355 cannot find it, without actually running the main matching function.
3356 This means that a special item such as (*COMMIT) at the start of a pat-
3357 tern is not considered until after a suitable starting point for the
3358 match has been found. Also, when callouts or (*MARK) items are in use,
3359 these "start-up" optimizations can cause them to be skipped if the pat-
3360 tern is never actually used. The start-up optimizations are in effect a
3361 pre-scan of the subject that takes place before the pattern is run.
3363 The PCRE_NO_START_OPTIMIZE option disables the start-up optimizations,
3364 possibly causing performance to suffer, but ensuring that in cases
3365 where the result is "no match", the callouts do occur, and that items
3366 such as (*COMMIT) and (*MARK) are considered at every possible starting
3367 position in the subject string. If PCRE_NO_START_OPTIMIZE is set at
3368 compile time, it cannot be unset at matching time. The use of
3369 PCRE_NO_START_OPTIMIZE at matching time (that is, passing it to
3370 pcre_exec()) disables JIT execution; in this situation, matching is
3371 always done using interpretively.
3373 Setting PCRE_NO_START_OPTIMIZE can change the outcome of a matching
3374 operation. Consider the pattern
3378 When this is compiled, PCRE records the fact that a match must start
3379 with the character "A". Suppose the subject string is "DEFABC". The
3380 start-up optimization scans along the subject, finds "A" and runs the
3381 first match attempt from there. The (*COMMIT) item means that the pat-
3382 tern must match the current starting position, which in this case, it
3383 does. However, if the same match is run with PCRE_NO_START_OPTIMIZE
3384 set, the initial scan along the subject string does not happen. The
3385 first match attempt is run starting from "D" and when this fails,
3386 (*COMMIT) prevents any further matches being tried, so the overall
3387 result is "no match". If the pattern is studied, more start-up opti-
3388 mizations may be used. For example, a minimum length for the subject
3389 may be recorded. Consider the pattern
3393 The minimum length for a match is one character. If the subject is
3394 "ABC", there will be attempts to match "ABC", "BC", "C", and then
3395 finally an empty string. If the pattern is studied, the final attempt
3396 does not take place, because PCRE knows that the subject is too short,
3397 and so the (*MARK) is never encountered. In this case, studying the
3398 pattern does not affect the overall match result, which is still "no
3399 match", but it does affect the auxiliary information that is returned.
3403 When PCRE_UTF8 is set at compile time, the validity of the subject as a
3404 UTF-8 string is automatically checked when pcre_exec() is subsequently
3405 called. The entire string is checked before any other processing takes
3406 place. The value of startoffset is also checked to ensure that it
3407 points to the start of a UTF-8 character. There is a discussion about
3408 the validity of UTF-8 strings in the pcreunicode page. If an invalid
3409 sequence of bytes is found, pcre_exec() returns the error
3410 PCRE_ERROR_BADUTF8 or, if PCRE_PARTIAL_HARD is set and the problem is a
3411 truncated character at the end of the subject, PCRE_ERROR_SHORTUTF8. In
3412 both cases, information about the precise nature of the error may also
3413 be returned (see the descriptions of these errors in the section enti-
3414 tled Error return values from pcre_exec() below). If startoffset con-
3415 tains a value that does not point to the start of a UTF-8 character (or
3416 to the end of the subject), PCRE_ERROR_BADUTF8_OFFSET is returned.
3418 If you already know that your subject is valid, and you want to skip
3419 these checks for performance reasons, you can set the
3420 PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to
3421 do this for the second and subsequent calls to pcre_exec() if you are
3422 making repeated calls to find all the matches in a single subject
3423 string. However, you should be sure that the value of startoffset
3424 points to the start of a character (or the end of the subject). When
3425 PCRE_NO_UTF8_CHECK is set, the effect of passing an invalid string as a
3426 subject or an invalid value of startoffset is undefined. Your program
3432 These options turn on the partial matching feature. For backwards com-
3433 patibility, PCRE_PARTIAL is a synonym for PCRE_PARTIAL_SOFT. A partial
3434 match occurs if the end of the subject string is reached successfully,
3435 but there are not enough subject characters to complete the match. If
3436 this happens when PCRE_PARTIAL_SOFT (but not PCRE_PARTIAL_HARD) is set,
3437 matching continues by testing any remaining alternatives. Only if no
3438 complete match can be found is PCRE_ERROR_PARTIAL returned instead of
3439 PCRE_ERROR_NOMATCH. In other words, PCRE_PARTIAL_SOFT says that the
3440 caller is prepared to handle a partial match, but only if no complete
3443 If PCRE_PARTIAL_HARD is set, it overrides PCRE_PARTIAL_SOFT. In this
3444 case, if a partial match is found, pcre_exec() immediately returns
3445 PCRE_ERROR_PARTIAL, without considering any other alternatives. In
3446 other words, when PCRE_PARTIAL_HARD is set, a partial match is consid-
3447 ered to be more important that an alternative complete match.
3449 In both cases, the portion of the string that was inspected when the
3450 partial match was found is set as the first matching string. There is a
3451 more detailed discussion of partial and multi-segment matching, with
3452 examples, in the pcrepartial documentation.
3454 The string to be matched by pcre_exec()
3456 The subject string is passed to pcre_exec() as a pointer in subject, a
3457 length in length, and a starting offset in startoffset. The units for
3458 length and startoffset are bytes for the 8-bit library, 16-bit data
3459 items for the 16-bit library, and 32-bit data items for the 32-bit
3462 If startoffset is negative or greater than the length of the subject,
3463 pcre_exec() returns PCRE_ERROR_BADOFFSET. When the starting offset is
3464 zero, the search for a match starts at the beginning of the subject,
3465 and this is by far the most common case. In UTF-8 or UTF-16 mode, the
3466 offset must point to the start of a character, or the end of the sub-
3467 ject (in UTF-32 mode, one data unit equals one character, so all off-
3468 sets are valid). Unlike the pattern string, the subject may contain
3471 A non-zero starting offset is useful when searching for another match
3472 in the same subject by calling pcre_exec() again after a previous suc-
3473 cess. Setting startoffset differs from just passing over a shortened
3474 string and setting PCRE_NOTBOL in the case of a pattern that begins
3475 with any kind of lookbehind. For example, consider the pattern
3479 which finds occurrences of "iss" in the middle of words. (\B matches
3480 only if the current position in the subject is not a word boundary.)
3481 When applied to the string "Mississipi" the first call to pcre_exec()
3482 finds the first occurrence. If pcre_exec() is called again with just
3483 the remainder of the subject, namely "issipi", it does not match,
3484 because \B is always false at the start of the subject, which is deemed
3485 to be a word boundary. However, if pcre_exec() is passed the entire
3486 string again, but with startoffset set to 4, it finds the second occur-
3487 rence of "iss" because it is able to look behind the starting point to
3488 discover that it is preceded by a letter.
3490 Finding all the matches in a subject is tricky when the pattern can
3491 match an empty string. It is possible to emulate Perl's /g behaviour by
3492 first trying the match again at the same offset, with the
3493 PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED options, and then if that
3494 fails, advancing the starting offset and trying an ordinary match
3495 again. There is some code that demonstrates how to do this in the pcre-
3496 demo sample program. In the most general case, you have to check to see
3497 if the newline convention recognizes CRLF as a newline, and if so, and
3498 the current character is CR followed by LF, advance the starting offset
3499 by two characters instead of one.
3501 If a non-zero starting offset is passed when the pattern is anchored,
3502 one attempt to match at the given offset is made. This can only succeed
3503 if the pattern does not require the match to be at the start of the
3506 How pcre_exec() returns captured substrings
3508 In general, a pattern matches a certain portion of the subject, and in
3509 addition, further substrings from the subject may be picked out by
3510 parts of the pattern. Following the usage in Jeffrey Friedl's book,
3511 this is called "capturing" in what follows, and the phrase "capturing
3512 subpattern" is used for a fragment of a pattern that picks out a sub-
3513 string. PCRE supports several other kinds of parenthesized subpattern
3514 that do not cause substrings to be captured.
3516 Captured substrings are returned to the caller via a vector of integers
3517 whose address is passed in ovector. The number of elements in the vec-
3518 tor is passed in ovecsize, which must be a non-negative number. Note:
3519 this argument is NOT the size of ovector in bytes.
3521 The first two-thirds of the vector is used to pass back captured sub-
3522 strings, each substring using a pair of integers. The remaining third
3523 of the vector is used as workspace by pcre_exec() while matching cap-
3524 turing subpatterns, and is not available for passing back information.
3525 The number passed in ovecsize should always be a multiple of three. If
3526 it is not, it is rounded down.
3528 When a match is successful, information about captured substrings is
3529 returned in pairs of integers, starting at the beginning of ovector,
3530 and continuing up to two-thirds of its length at the most. The first
3531 element of each pair is set to the offset of the first character in a
3532 substring, and the second is set to the offset of the first character
3533 after the end of a substring. These values are always data unit off-
3534 sets, even in UTF mode. They are byte offsets in the 8-bit library,
3535 16-bit data item offsets in the 16-bit library, and 32-bit data item
3536 offsets in the 32-bit library. Note: they are not character counts.
3538 The first pair of integers, ovector[0] and ovector[1], identify the
3539 portion of the subject string matched by the entire pattern. The next
3540 pair is used for the first capturing subpattern, and so on. The value
3541 returned by pcre_exec() is one more than the highest numbered pair that
3542 has been set. For example, if two substrings have been captured, the
3543 returned value is 3. If there are no capturing subpatterns, the return
3544 value from a successful match is 1, indicating that just the first pair
3545 of offsets has been set.
3547 If a capturing subpattern is matched repeatedly, it is the last portion
3548 of the string that it matched that is returned.
3550 If the vector is too small to hold all the captured substring offsets,
3551 it is used as far as possible (up to two-thirds of its length), and the
3552 function returns a value of zero. If neither the actual string matched
3553 nor any captured substrings are of interest, pcre_exec() may be called
3554 with ovector passed as NULL and ovecsize as zero. However, if the pat-
3555 tern contains back references and the ovector is not big enough to
3556 remember the related substrings, PCRE has to get additional memory for
3557 use during matching. Thus it is usually advisable to supply an ovector
3560 There are some cases where zero is returned (indicating vector over-
3561 flow) when in fact the vector is exactly the right size for the final
3562 match. For example, consider the pattern
3566 If a vector of 6 elements (allowing for only 1 captured substring) is
3567 given with subject string "abd", pcre_exec() will try to set the second
3568 captured string, thereby recording a vector overflow, before failing to
3569 match "c" and backing up to try the second alternative. The zero
3570 return, however, does correctly indicate that the maximum number of
3571 slots (namely 2) have been filled. In similar cases where there is tem-
3572 porary overflow, but the final number of used slots is actually less
3573 than the maximum, a non-zero value is returned.
3575 The pcre_fullinfo() function can be used to find out how many capturing
3576 subpatterns there are in a compiled pattern. The smallest size for
3577 ovector that will allow for n captured substrings, in addition to the
3578 offsets of the substring matched by the whole pattern, is (n+1)*3.
3580 It is possible for capturing subpattern number n+1 to match some part
3581 of the subject when subpattern n has not been used at all. For example,
3582 if the string "abc" is matched against the pattern (a|(z))(bc) the
3583 return from the function is 4, and subpatterns 1 and 3 are matched, but
3584 2 is not. When this happens, both values in the offset pairs corre-
3585 sponding to unused subpatterns are set to -1.
3587 Offset values that correspond to unused subpatterns at the end of the
3588 expression are also set to -1. For example, if the string "abc" is
3589 matched against the pattern (abc)(x(yz)?)? subpatterns 2 and 3 are not
3590 matched. The return from the function is 2, because the highest used
3591 capturing subpattern number is 1, and the offsets for for the second
3592 and third capturing subpatterns (assuming the vector is large enough,
3593 of course) are set to -1.
3595 Note: Elements in the first two-thirds of ovector that do not corre-
3596 spond to capturing parentheses in the pattern are never changed. That
3597 is, if a pattern contains n capturing parentheses, no more than ovec-
3598 tor[0] to ovector[2n+1] are set by pcre_exec(). The other elements (in
3599 the first two-thirds) retain whatever values they previously had.
3601 Some convenience functions are provided for extracting the captured
3602 substrings as separate strings. These are described below.
3604 Error return values from pcre_exec()
3606 If pcre_exec() fails, it returns a negative number. The following are
3607 defined in the header file:
3609 PCRE_ERROR_NOMATCH (-1)
3611 The subject string did not match the pattern.
3613 PCRE_ERROR_NULL (-2)
3615 Either code or subject was passed as NULL, or ovector was NULL and
3616 ovecsize was not zero.
3618 PCRE_ERROR_BADOPTION (-3)
3620 An unrecognized bit was set in the options argument.
3622 PCRE_ERROR_BADMAGIC (-4)
3624 PCRE stores a 4-byte "magic number" at the start of the compiled code,
3625 to catch the case when it is passed a junk pointer and to detect when a
3626 pattern that was compiled in an environment of one endianness is run in
3627 an environment with the other endianness. This is the error that PCRE
3628 gives when the magic number is not present.
3630 PCRE_ERROR_UNKNOWN_OPCODE (-5)
3632 While running the pattern match, an unknown item was encountered in the
3633 compiled pattern. This error could be caused by a bug in PCRE or by
3634 overwriting of the compiled pattern.
3636 PCRE_ERROR_NOMEMORY (-6)
3638 If a pattern contains back references, but the ovector that is passed
3639 to pcre_exec() is not big enough to remember the referenced substrings,
3640 PCRE gets a block of memory at the start of matching to use for this
3641 purpose. If the call via pcre_malloc() fails, this error is given. The
3642 memory is automatically freed at the end of matching.
3644 This error is also given if pcre_stack_malloc() fails in pcre_exec().
3645 This can happen only when PCRE has been compiled with --disable-stack-
3648 PCRE_ERROR_NOSUBSTRING (-7)
3650 This error is used by the pcre_copy_substring(), pcre_get_substring(),
3651 and pcre_get_substring_list() functions (see below). It is never
3652 returned by pcre_exec().
3654 PCRE_ERROR_MATCHLIMIT (-8)
3656 The backtracking limit, as specified by the match_limit field in a
3657 pcre_extra structure (or defaulted) was reached. See the description
3660 PCRE_ERROR_CALLOUT (-9)
3662 This error is never generated by pcre_exec() itself. It is provided for
3663 use by callout functions that want to yield a distinctive error code.
3664 See the pcrecallout documentation for details.
3666 PCRE_ERROR_BADUTF8 (-10)
3668 A string that contains an invalid UTF-8 byte sequence was passed as a
3669 subject, and the PCRE_NO_UTF8_CHECK option was not set. If the size of
3670 the output vector (ovecsize) is at least 2, the byte offset to the
3671 start of the the invalid UTF-8 character is placed in the first ele-
3672 ment, and a reason code is placed in the second element. The reason
3673 codes are listed in the following section. For backward compatibility,
3674 if PCRE_PARTIAL_HARD is set and the problem is a truncated UTF-8 char-
3675 acter at the end of the subject (reason codes 1 to 5),
3676 PCRE_ERROR_SHORTUTF8 is returned instead of PCRE_ERROR_BADUTF8.
3678 PCRE_ERROR_BADUTF8_OFFSET (-11)
3680 The UTF-8 byte sequence that was passed as a subject was checked and
3681 found to be valid (the PCRE_NO_UTF8_CHECK option was not set), but the
3682 value of startoffset did not point to the beginning of a UTF-8 charac-
3683 ter or the end of the subject.
3685 PCRE_ERROR_PARTIAL (-12)
3687 The subject string did not match, but it did match partially. See the
3688 pcrepartial documentation for details of partial matching.
3690 PCRE_ERROR_BADPARTIAL (-13)
3692 This code is no longer in use. It was formerly returned when the
3693 PCRE_PARTIAL option was used with a compiled pattern containing items
3694 that were not supported for partial matching. From release 8.00
3695 onwards, there are no restrictions on partial matching.
3697 PCRE_ERROR_INTERNAL (-14)
3699 An unexpected internal error has occurred. This error could be caused
3700 by a bug in PCRE or by overwriting of the compiled pattern.
3702 PCRE_ERROR_BADCOUNT (-15)
3704 This error is given if the value of the ovecsize argument is negative.
3706 PCRE_ERROR_RECURSIONLIMIT (-21)
3708 The internal recursion limit, as specified by the match_limit_recursion
3709 field in a pcre_extra structure (or defaulted) was reached. See the
3712 PCRE_ERROR_BADNEWLINE (-23)
3714 An invalid combination of PCRE_NEWLINE_xxx options was given.
3716 PCRE_ERROR_BADOFFSET (-24)
3718 The value of startoffset was negative or greater than the length of the
3719 subject, that is, the value in length.
3721 PCRE_ERROR_SHORTUTF8 (-25)
3723 This error is returned instead of PCRE_ERROR_BADUTF8 when the subject
3724 string ends with a truncated UTF-8 character and the PCRE_PARTIAL_HARD
3725 option is set. Information about the failure is returned as for
3726 PCRE_ERROR_BADUTF8. It is in fact sufficient to detect this case, but
3727 this special error code for PCRE_PARTIAL_HARD precedes the implementa-
3728 tion of returned information; it is retained for backwards compatibil-
3731 PCRE_ERROR_RECURSELOOP (-26)
3733 This error is returned when pcre_exec() detects a recursion loop within
3734 the pattern. Specifically, it means that either the whole pattern or a
3735 subpattern has been called recursively for the second time at the same
3736 position in the subject string. Some simple patterns that might do this
3737 are detected and faulted at compile time, but more complicated cases,
3738 in particular mutual recursions between two different subpatterns, can-
3739 not be detected until run time.
3741 PCRE_ERROR_JIT_STACKLIMIT (-27)
3743 This error is returned when a pattern that was successfully studied
3744 using a JIT compile option is being matched, but the memory available
3745 for the just-in-time processing stack is not large enough. See the
3746 pcrejit documentation for more details.
3748 PCRE_ERROR_BADMODE (-28)
3750 This error is given if a pattern that was compiled by the 8-bit library
3751 is passed to a 16-bit or 32-bit library function, or vice versa.
3753 PCRE_ERROR_BADENDIANNESS (-29)
3755 This error is given if a pattern that was compiled and saved is
3756 reloaded on a host with different endianness. The utility function
3757 pcre_pattern_to_host_byte_order() can be used to convert such a pattern
3758 so that it runs on the new host.
3760 PCRE_ERROR_JIT_BADOPTION
3762 This error is returned when a pattern that was successfully studied
3763 using a JIT compile option is being matched, but the matching mode
3764 (partial or complete match) does not correspond to any JIT compilation
3765 mode. When the JIT fast path function is used, this error may be also
3766 given for invalid options. See the pcrejit documentation for more
3769 PCRE_ERROR_BADLENGTH (-32)
3771 This error is given if pcre_exec() is called with a negative value for
3772 the length argument.
3774 Error numbers -16 to -20, -22, and 30 are not used by pcre_exec().
3776 Reason codes for invalid UTF-8 strings
3778 This section applies only to the 8-bit library. The corresponding
3779 information for the 16-bit and 32-bit libraries is given in the pcre16
3782 When pcre_exec() returns either PCRE_ERROR_BADUTF8 or PCRE_ERROR_SHORT-
3783 UTF8, and the size of the output vector (ovecsize) is at least 2, the
3784 offset of the start of the invalid UTF-8 character is placed in the
3785 first output vector element (ovector[0]) and a reason code is placed in
3786 the second element (ovector[1]). The reason codes are given names in
3787 the pcre.h header file:
3795 The string ends with a truncated UTF-8 character; the code specifies
3796 how many bytes are missing (1 to 5). Although RFC 3629 restricts UTF-8
3797 characters to be no longer than 4 bytes, the encoding scheme (origi-
3798 nally defined by RFC 2279) allows for up to 6 bytes, and this is
3799 checked first; hence the possibility of 4 or 5 missing bytes.
3807 The two most significant bits of the 2nd, 3rd, 4th, 5th, or 6th byte of
3808 the character do not have the binary value 0b10 (that is, either the
3809 most significant bit is 0, or the next bit is 1).
3814 A character that is valid by the RFC 2279 rules is either 5 or 6 bytes
3815 long; these code points are excluded by RFC 3629.
3819 A 4-byte character has a value greater than 0x10fff; these code points
3820 are excluded by RFC 3629.
3824 A 3-byte character has a value in the range 0xd800 to 0xdfff; this
3825 range of code points are reserved by RFC 3629 for use with UTF-16, and
3826 so are excluded from UTF-8.
3834 A 2-, 3-, 4-, 5-, or 6-byte character is "overlong", that is, it codes
3835 for a value that can be represented by fewer bytes, which is invalid.
3836 For example, the two bytes 0xc0, 0xae give the value 0x2e, whose cor-
3837 rect coding uses just one byte.
3841 The two most significant bits of the first byte of a character have the
3842 binary value 0b10 (that is, the most significant bit is 1 and the sec-
3843 ond is 0). Such a byte can only validly occur as the second or subse-
3844 quent byte of a multi-byte character.
3848 The first byte of a character has the value 0xfe or 0xff. These values
3849 can never occur in a valid UTF-8 string.
3853 This error code was formerly used when the presence of a so-called
3854 "non-character" caused an error. Unicode corrigendum #9 makes it clear
3855 that such characters should not cause a string to be rejected, and so
3856 this code is no longer in use and is never returned.
3859 EXTRACTING CAPTURED SUBSTRINGS BY NUMBER
3861 int pcre_copy_substring(const char *subject, int *ovector,
3862 int stringcount, int stringnumber, char *buffer,
3865 int pcre_get_substring(const char *subject, int *ovector,
3866 int stringcount, int stringnumber,
3867 const char **stringptr);
3869 int pcre_get_substring_list(const char *subject,
3870 int *ovector, int stringcount, const char ***listptr);
3872 Captured substrings can be accessed directly by using the offsets
3873 returned by pcre_exec() in ovector. For convenience, the functions
3874 pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub-
3875 string_list() are provided for extracting captured substrings as new,
3876 separate, zero-terminated strings. These functions identify substrings
3877 by number. The next section describes functions for extracting named
3880 A substring that contains a binary zero is correctly extracted and has
3881 a further zero added on the end, but the result is not, of course, a C
3882 string. However, you can process such a string by referring to the
3883 length that is returned by pcre_copy_substring() and pcre_get_sub-
3884 string(). Unfortunately, the interface to pcre_get_substring_list() is
3885 not adequate for handling strings containing binary zeros, because the
3886 end of the final string is not independently indicated.
3888 The first three arguments are the same for all three of these func-
3889 tions: subject is the subject string that has just been successfully
3890 matched, ovector is a pointer to the vector of integer offsets that was
3891 passed to pcre_exec(), and stringcount is the number of substrings that
3892 were captured by the match, including the substring that matched the
3893 entire regular expression. This is the value returned by pcre_exec() if
3894 it is greater than zero. If pcre_exec() returned zero, indicating that
3895 it ran out of space in ovector, the value passed as stringcount should
3896 be the number of elements in the vector divided by three.
3898 The functions pcre_copy_substring() and pcre_get_substring() extract a
3899 single substring, whose number is given as stringnumber. A value of
3900 zero extracts the substring that matched the entire pattern, whereas
3901 higher values extract the captured substrings. For pcre_copy_sub-
3902 string(), the string is placed in buffer, whose length is given by
3903 buffersize, while for pcre_get_substring() a new block of memory is
3904 obtained via pcre_malloc, and its address is returned via stringptr.
3905 The yield of the function is the length of the string, not including
3906 the terminating zero, or one of these error codes:
3908 PCRE_ERROR_NOMEMORY (-6)
3910 The buffer was too small for pcre_copy_substring(), or the attempt to
3911 get memory failed for pcre_get_substring().
3913 PCRE_ERROR_NOSUBSTRING (-7)
3915 There is no substring whose number is stringnumber.
3917 The pcre_get_substring_list() function extracts all available sub-
3918 strings and builds a list of pointers to them. All this is done in a
3919 single block of memory that is obtained via pcre_malloc. The address of
3920 the memory block is returned via listptr, which is also the start of
3921 the list of string pointers. The end of the list is marked by a NULL
3922 pointer. The yield of the function is zero if all went well, or the
3925 PCRE_ERROR_NOMEMORY (-6)
3927 if the attempt to get the memory block failed.
3929 When any of these functions encounter a substring that is unset, which
3930 can happen when capturing subpattern number n+1 matches some part of
3931 the subject, but subpattern n has not been used at all, they return an
3932 empty string. This can be distinguished from a genuine zero-length sub-
3933 string by inspecting the appropriate offset in ovector, which is nega-
3934 tive for unset substrings.
3936 The two convenience functions pcre_free_substring() and pcre_free_sub-
3937 string_list() can be used to free the memory returned by a previous
3938 call of pcre_get_substring() or pcre_get_substring_list(), respec-
3939 tively. They do nothing more than call the function pointed to by
3940 pcre_free, which of course could be called directly from a C program.
3941 However, PCRE is used in some situations where it is linked via a spe-
3942 cial interface to another programming language that cannot use
3943 pcre_free directly; it is for these cases that the functions are pro-
3947 EXTRACTING CAPTURED SUBSTRINGS BY NAME
3949 int pcre_get_stringnumber(const pcre *code,
3952 int pcre_copy_named_substring(const pcre *code,
3953 const char *subject, int *ovector,
3954 int stringcount, const char *stringname,
3955 char *buffer, int buffersize);
3957 int pcre_get_named_substring(const pcre *code,
3958 const char *subject, int *ovector,
3959 int stringcount, const char *stringname,
3960 const char **stringptr);
3962 To extract a substring by name, you first have to find associated num-
3963 ber. For example, for this pattern
3967 the number of the subpattern called "xxx" is 2. If the name is known to
3968 be unique (PCRE_DUPNAMES was not set), you can find the number from the
3969 name by calling pcre_get_stringnumber(). The first argument is the com-
3970 piled pattern, and the second is the name. The yield of the function is
3971 the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if there is no
3972 subpattern of that name.
3974 Given the number, you can extract the substring directly, or use one of
3975 the functions described in the previous section. For convenience, there
3976 are also two functions that do the whole job.
3978 Most of the arguments of pcre_copy_named_substring() and
3979 pcre_get_named_substring() are the same as those for the similarly
3980 named functions that extract by number. As these are described in the
3981 previous section, they are not re-described here. There are just two
3984 First, instead of a substring number, a substring name is given. Sec-
3985 ond, there is an extra argument, given at the start, which is a pointer
3986 to the compiled pattern. This is needed in order to gain access to the
3987 name-to-number translation table.
3989 These functions call pcre_get_stringnumber(), and if it succeeds, they
3990 then call pcre_copy_substring() or pcre_get_substring(), as appropri-
3991 ate. NOTE: If PCRE_DUPNAMES is set and there are duplicate names, the
3992 behaviour may not be what you want (see the next section).
3994 Warning: If the pattern uses the (?| feature to set up multiple subpat-
3995 terns with the same number, as described in the section on duplicate
3996 subpattern numbers in the pcrepattern page, you cannot use names to
3997 distinguish the different subpatterns, because names are not included
3998 in the compiled code. The matching process uses only numbers. For this
3999 reason, the use of different names for subpatterns of the same number
4000 causes an error at compile time.
4003 DUPLICATE SUBPATTERN NAMES
4005 int pcre_get_stringtable_entries(const pcre *code,
4006 const char *name, char **first, char **last);
4008 When a pattern is compiled with the PCRE_DUPNAMES option, names for
4009 subpatterns are not required to be unique. (Duplicate names are always
4010 allowed for subpatterns with the same number, created by using the (?|
4011 feature. Indeed, if such subpatterns are named, they are required to
4012 use the same names.)
4014 Normally, patterns with duplicate names are such that in any one match,
4015 only one of the named subpatterns participates. An example is shown in
4016 the pcrepattern documentation.
4018 When duplicates are present, pcre_copy_named_substring() and
4019 pcre_get_named_substring() return the first substring corresponding to
4020 the given name that is set. If none are set, PCRE_ERROR_NOSUBSTRING
4021 (-7) is returned; no data is returned. The pcre_get_stringnumber()
4022 function returns one of the numbers that are associated with the name,
4023 but it is not defined which it is.
4025 If you want to get full details of all captured substrings for a given
4026 name, you must use the pcre_get_stringtable_entries() function. The
4027 first argument is the compiled pattern, and the second is the name. The
4028 third and fourth are pointers to variables which are updated by the
4029 function. After it has run, they point to the first and last entries in
4030 the name-to-number table for the given name. The function itself
4031 returns the length of each entry, or PCRE_ERROR_NOSUBSTRING (-7) if
4032 there are none. The format of the table is described above in the sec-
4033 tion entitled Information about a pattern above. Given all the rele-
4034 vant entries for the name, you can extract each of their numbers, and
4035 hence the captured data, if any.
4038 FINDING ALL POSSIBLE MATCHES
4040 The traditional matching function uses a similar algorithm to Perl,
4041 which stops when it finds the first match, starting at a given point in
4042 the subject. If you want to find all possible matches, or the longest
4043 possible match, consider using the alternative matching function (see
4044 below) instead. If you cannot use the alternative function, but still
4045 need to find all possible matches, you can kludge it up by making use
4046 of the callout facility, which is described in the pcrecallout documen-
4049 What you have to do is to insert a callout right at the end of the pat-
4050 tern. When your callout function is called, extract and save the cur-
4051 rent matched substring. Then return 1, which forces pcre_exec() to
4052 backtrack and try other alternatives. Ultimately, when it runs out of
4053 matches, pcre_exec() will yield PCRE_ERROR_NOMATCH.
4056 OBTAINING AN ESTIMATE OF STACK USAGE
4058 Matching certain patterns using pcre_exec() can use a lot of process
4059 stack, which in certain environments can be rather limited in size.
4060 Some users find it helpful to have an estimate of the amount of stack
4061 that is used by pcre_exec(), to help them set recursion limits, as
4062 described in the pcrestack documentation. The estimate that is output
4063 by pcretest when called with the -m and -C options is obtained by call-
4064 ing pcre_exec with the values NULL, NULL, NULL, -999, and -999 for its
4065 first five arguments.
4067 Normally, if its first argument is NULL, pcre_exec() immediately
4068 returns the negative error code PCRE_ERROR_NULL, but with this special
4069 combination of arguments, it returns instead a negative number whose
4070 absolute value is the approximate stack frame size in bytes. (A nega-
4071 tive number is used so that it is clear that no match has happened.)
4072 The value is approximate because in some cases, recursive calls to
4073 pcre_exec() occur when there are one or two additional variables on the
4076 If PCRE has been compiled to use the heap instead of the stack for
4077 recursion, the value returned is the size of each block that is
4078 obtained from the heap.
4081 MATCHING A PATTERN: THE ALTERNATIVE FUNCTION
4083 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
4084 const char *subject, int length, int startoffset,
4085 int options, int *ovector, int ovecsize,
4086 int *workspace, int wscount);
4088 The function pcre_dfa_exec() is called to match a subject string
4089 against a compiled pattern, using a matching algorithm that scans the
4090 subject string just once, and does not backtrack. This has different
4091 characteristics to the normal algorithm, and is not compatible with
4092 Perl. Some of the features of PCRE patterns are not supported. Never-
4093 theless, there are times when this kind of matching can be useful. For
4094 a discussion of the two matching algorithms, and a list of features
4095 that pcre_dfa_exec() does not support, see the pcrematching documenta-
4098 The arguments for the pcre_dfa_exec() function are the same as for
4099 pcre_exec(), plus two extras. The ovector argument is used in a differ-
4100 ent way, and this is described below. The other common arguments are
4101 used in the same way as for pcre_exec(), so their description is not
4104 The two additional arguments provide workspace for the function. The
4105 workspace vector should contain at least 20 elements. It is used for
4106 keeping track of multiple paths through the pattern tree. More
4107 workspace will be needed for patterns and subjects where there are a
4108 lot of potential matches.
4110 Here is an example of a simple call to pcre_dfa_exec():
4116 re, /* result of pcre_compile() */
4117 NULL, /* we didn't study the pattern */
4118 "some string", /* the subject string */
4119 11, /* the length of the subject string */
4120 0, /* start at offset 0 in the subject */
4121 0, /* default options */
4122 ovector, /* vector of integers for substring information */
4123 10, /* number of elements (NOT size in bytes) */
4124 wspace, /* working space vector */
4125 20); /* number of elements (NOT size in bytes) */
4127 Option bits for pcre_dfa_exec()
4129 The unused bits of the options argument for pcre_dfa_exec() must be
4130 zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NEW-
4131 LINE_xxx, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY,
4132 PCRE_NOTEMPTY_ATSTART, PCRE_NO_UTF8_CHECK, PCRE_BSR_ANYCRLF,
4133 PCRE_BSR_UNICODE, PCRE_NO_START_OPTIMIZE, PCRE_PARTIAL_HARD, PCRE_PAR-
4134 TIAL_SOFT, PCRE_DFA_SHORTEST, and PCRE_DFA_RESTART. All but the last
4135 four of these are exactly the same as for pcre_exec(), so their
4136 description is not repeated here.
4141 These have the same general effect as they do for pcre_exec(), but the
4142 details are slightly different. When PCRE_PARTIAL_HARD is set for
4143 pcre_dfa_exec(), it returns PCRE_ERROR_PARTIAL if the end of the sub-
4144 ject is reached and there is still at least one matching possibility
4145 that requires additional characters. This happens even if some complete
4146 matches have also been found. When PCRE_PARTIAL_SOFT is set, the return
4147 code PCRE_ERROR_NOMATCH is converted into PCRE_ERROR_PARTIAL if the end
4148 of the subject is reached, there have been no complete matches, but
4149 there is still at least one matching possibility. The portion of the
4150 string that was inspected when the longest partial match was found is
4151 set as the first matching string in both cases. There is a more
4152 detailed discussion of partial and multi-segment matching, with exam-
4153 ples, in the pcrepartial documentation.
4157 Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to
4158 stop as soon as it has found one match. Because of the way the alterna-
4159 tive algorithm works, this is necessarily the shortest possible match
4160 at the first possible matching point in the subject string.
4164 When pcre_dfa_exec() returns a partial match, it is possible to call it
4165 again, with additional subject characters, and have it continue with
4166 the same match. The PCRE_DFA_RESTART option requests this action; when
4167 it is set, the workspace and wscount options must reference the same
4168 vector as before because data about the match so far is left in them
4169 after a partial match. There is more discussion of this facility in the
4170 pcrepartial documentation.
4172 Successful returns from pcre_dfa_exec()
4174 When pcre_dfa_exec() succeeds, it may have matched more than one sub-
4175 string in the subject. Note, however, that all the matches from one run
4176 of the function start at the same point in the subject. The shorter
4177 matches are all initial substrings of the longer matches. For example,
4182 is matched against the string
4184 This is <something> <something else> <something further> no more
4186 the three matched strings are
4189 <something> <something else>
4190 <something> <something else> <something further>
4192 On success, the yield of the function is a number greater than zero,
4193 which is the number of matched substrings. The substrings themselves
4194 are returned in ovector. Each string uses two elements; the first is
4195 the offset to the start, and the second is the offset to the end. In
4196 fact, all the strings have the same start offset. (Space could have
4197 been saved by giving this only once, but it was decided to retain some
4198 compatibility with the way pcre_exec() returns data, even though the
4199 meaning of the strings is different.)
4201 The strings are returned in reverse order of length; that is, the long-
4202 est matching string is given first. If there were too many matches to
4203 fit into ovector, the yield of the function is zero, and the vector is
4204 filled with the longest matches. Unlike pcre_exec(), pcre_dfa_exec()
4205 can use the entire ovector for returning matched strings.
4207 NOTE: PCRE's "auto-possessification" optimization usually applies to
4208 character repeats at the end of a pattern (as well as internally). For
4209 example, the pattern "a\d+" is compiled as if it were "a\d++" because
4210 there is no point even considering the possibility of backtracking into
4211 the repeated digits. For DFA matching, this means that only one possi-
4212 ble match is found. If you really do want multiple matches in such
4213 cases, either use an ungreedy repeat ("a\d+?") or set the
4214 PCRE_NO_AUTO_POSSESS option when compiling.
4216 Error returns from pcre_dfa_exec()
4218 The pcre_dfa_exec() function returns a negative number when it fails.
4219 Many of the errors are the same as for pcre_exec(), and these are
4220 described above. There are in addition the following errors that are
4221 specific to pcre_dfa_exec():
4223 PCRE_ERROR_DFA_UITEM (-16)
4225 This return is given if pcre_dfa_exec() encounters an item in the pat-
4226 tern that it does not support, for instance, the use of \C or a back
4229 PCRE_ERROR_DFA_UCOND (-17)
4231 This return is given if pcre_dfa_exec() encounters a condition item
4232 that uses a back reference for the condition, or a test for recursion
4233 in a specific group. These are not supported.
4235 PCRE_ERROR_DFA_UMLIMIT (-18)
4237 This return is given if pcre_dfa_exec() is called with an extra block
4238 that contains a setting of the match_limit or match_limit_recursion
4239 fields. This is not supported (these fields are meaningless for DFA
4242 PCRE_ERROR_DFA_WSSIZE (-19)
4244 This return is given if pcre_dfa_exec() runs out of space in the
4247 PCRE_ERROR_DFA_RECURSE (-20)
4249 When a recursive subpattern is processed, the matching function calls
4250 itself recursively, using private vectors for ovector and workspace.
4251 This error is given if the output vector is not large enough. This
4252 should be extremely rare, as a vector of size 1000 is used.
4254 PCRE_ERROR_DFA_BADRESTART (-30)
4256 When pcre_dfa_exec() is called with the PCRE_DFA_RESTART option, some
4257 plausibility checks are made on the contents of the workspace, which
4258 should contain data about the previous partial match. If any of these
4259 checks fail, this error is given.
4264 pcre16(3), pcre32(3), pcrebuild(3), pcrecallout(3), pcrecpp(3)(3),
4265 pcrematching(3), pcrepartial(3), pcreposix(3), pcreprecompile(3), pcre-
4266 sample(3), pcrestack(3).
4272 University Computing Service
4273 Cambridge CB2 3QH, England.
4278 Last updated: 18 December 2015
4279 Copyright (c) 1997-2015 University of Cambridge.
4280 ------------------------------------------------------------------------------
4283 PCRECALLOUT(3) Library Functions Manual PCRECALLOUT(3)
4288 PCRE - Perl-compatible regular expressions
4294 int (*pcre_callout)(pcre_callout_block *);
4296 int (*pcre16_callout)(pcre16_callout_block *);
4298 int (*pcre32_callout)(pcre32_callout_block *);
4303 PCRE provides a feature called "callout", which is a means of temporar-
4304 ily passing control to the caller of PCRE in the middle of pattern
4305 matching. The caller of PCRE provides an external function by putting
4306 its entry point in the global variable pcre_callout (pcre16_callout for
4307 the 16-bit library, pcre32_callout for the 32-bit library). By default,
4308 this variable contains NULL, which disables all calling out.
4310 Within a regular expression, (?C) indicates the points at which the
4311 external function is to be called. Different callout points can be
4312 identified by putting a number less than 256 after the letter C. The
4313 default value is zero. For example, this pattern has two callout
4318 If the PCRE_AUTO_CALLOUT option bit is set when a pattern is compiled,
4319 PCRE automatically inserts callouts, all with number 255, before each
4320 item in the pattern. For example, if PCRE_AUTO_CALLOUT is used with the
4325 it is processed as if it were
4327 (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255)
4329 Notice that there is a callout before and after each parenthesis and
4330 alternation bar. If the pattern contains a conditional group whose con-
4331 dition is an assertion, an automatic callout is inserted immediately
4332 before the condition. Such a callout may also be inserted explicitly,
4337 This applies only to assertion conditions (because they are themselves
4338 independent groups).
4340 Automatic callouts can be used for tracking the progress of pattern
4341 matching. The pcretest program has a pattern qualifier (/C) that sets
4342 automatic callouts; when it is used, the output indicates how the pat-
4343 tern is being matched. This is useful information when you are trying
4344 to optimize the performance of a particular pattern.
4349 You should be aware that, because of optimizations in the way PCRE com-
4350 piles and matches patterns, callouts sometimes do not happen exactly as
4353 At compile time, PCRE "auto-possessifies" repeated items when it knows
4354 that what follows cannot be part of the repeat. For example, a+[bc] is
4355 compiled as if it were a++[bc]. The pcretest output when this pattern
4356 is anchored and then applied with automatic callouts to the string
4365 This indicates that when matching [bc] fails, there is no backtracking
4366 into a+ and therefore the callouts that would be taken for the back-
4367 tracks do not occur. You can disable the auto-possessify feature by
4368 passing PCRE_NO_AUTO_POSSESS to pcre_compile(), or starting the pattern
4369 with (*NO_AUTO_POSSESS). If this is done in pcretest (using the /O
4370 qualifier), the output changes to this:
4381 This time, when matching [bc] fails, the matcher backtracks into a+ and
4382 tries again, repeatedly, until a+ itself fails.
4384 Other optimizations that provide fast "no match" results also affect
4385 callouts. For example, if the pattern is
4389 PCRE knows that any matching string must contain the letter "d". If the
4390 subject string is "abyz", the lack of "d" means that matching doesn't
4391 ever start, and the callout is never reached. However, with "abyd",
4392 though the result is still no match, the callout is obeyed.
4394 If the pattern is studied, PCRE knows the minimum length of a matching
4395 string, and will immediately give a "no match" return without actually
4396 running a match if the subject is not long enough, or, for unanchored
4397 patterns, if it has been scanned far enough.
4399 You can disable these optimizations by passing the PCRE_NO_START_OPTI-
4400 MIZE option to the matching function, or by starting the pattern with
4401 (*NO_START_OPT). This slows down the matching process, but does ensure
4402 that callouts such as the example above are obeyed.
4405 THE CALLOUT INTERFACE
4407 During matching, when PCRE reaches a callout point, the external func-
4408 tion defined by pcre_callout or pcre[16|32]_callout is called (if it is
4409 set). This applies to both normal and DFA matching. The only argument
4410 to the callout function is a pointer to a pcre_callout or
4411 pcre[16|32]_callout block. These structures contains the following
4417 const char *subject; (8-bit version)
4418 PCRE_SPTR16 subject; (16-bit version)
4419 PCRE_SPTR32 subject; (32-bit version)
4422 int current_position;
4426 int pattern_position;
4427 int next_item_length;
4428 const unsigned char *mark; (8-bit version)
4429 const PCRE_UCHAR16 *mark; (16-bit version)
4430 const PCRE_UCHAR32 *mark; (32-bit version)
4432 The version field is an integer containing the version number of the
4433 block format. The initial version was 0; the current version is 2. The
4434 version number will change again in future if additional fields are
4435 added, but the intention is never to remove any of the existing fields.
4437 The callout_number field contains the number of the callout, as com-
4438 piled into the pattern (that is, the number after ?C for manual call-
4439 outs, and 255 for automatically generated callouts).
4441 The offset_vector field is a pointer to the vector of offsets that was
4442 passed by the caller to the matching function. When pcre_exec() or
4443 pcre[16|32]_exec() is used, the contents can be inspected, in order to
4444 extract substrings that have been matched so far, in the same way as
4445 for extracting substrings after a match has completed. For the DFA
4446 matching functions, this field is not useful.
4448 The subject and subject_length fields contain copies of the values that
4449 were passed to the matching function.
4451 The start_match field normally contains the offset within the subject
4452 at which the current match attempt started. However, if the escape
4453 sequence \K has been encountered, this value is changed to reflect the
4454 modified starting point. If the pattern is not anchored, the callout
4455 function may be called several times from the same point in the pattern
4456 for different starting points in the subject.
4458 The current_position field contains the offset within the subject of
4459 the current match pointer.
4461 When the pcre_exec() or pcre[16|32]_exec() is used, the capture_top
4462 field contains one more than the number of the highest numbered cap-
4463 tured substring so far. If no substrings have been captured, the value
4464 of capture_top is one. This is always the case when the DFA functions
4465 are used, because they do not support captured substrings.
4467 The capture_last field contains the number of the most recently cap-
4468 tured substring. However, when a recursion exits, the value reverts to
4469 what it was outside the recursion, as do the values of all captured
4470 substrings. If no substrings have been captured, the value of cap-
4471 ture_last is -1. This is always the case for the DFA matching func-
4474 The callout_data field contains a value that is passed to a matching
4475 function specifically so that it can be passed back in callouts. It is
4476 passed in the callout_data field of a pcre_extra or pcre[16|32]_extra
4477 data structure. If no such data was passed, the value of callout_data
4478 in a callout block is NULL. There is a description of the pcre_extra
4479 structure in the pcreapi documentation.
4481 The pattern_position field is present from version 1 of the callout
4482 structure. It contains the offset to the next item to be matched in the
4485 The next_item_length field is present from version 1 of the callout
4486 structure. It contains the length of the next item to be matched in the
4487 pattern string. When the callout immediately precedes an alternation
4488 bar, a closing parenthesis, or the end of the pattern, the length is
4489 zero. When the callout precedes an opening parenthesis, the length is
4490 that of the entire subpattern.
4492 The pattern_position and next_item_length fields are intended to help
4493 in distinguishing between different automatic callouts, which all have
4494 the same callout number. However, they are set for all callouts.
4496 The mark field is present from version 2 of the callout structure. In
4497 callouts from pcre_exec() or pcre[16|32]_exec() it contains a pointer
4498 to the zero-terminated name of the most recently passed (*MARK),
4499 (*PRUNE), or (*THEN) item in the match, or NULL if no such items have
4500 been passed. Instances of (*PRUNE) or (*THEN) without a name do not
4501 obliterate a previous (*MARK). In callouts from the DFA matching func-
4502 tions this field always contains NULL.
4507 The external callout function returns an integer to PCRE. If the value
4508 is zero, matching proceeds as normal. If the value is greater than
4509 zero, matching fails at the current point, but the testing of other
4510 matching possibilities goes ahead, just as if a lookahead assertion had
4511 failed. If the value is less than zero, the match is abandoned, the
4512 matching function returns the negative value.
4514 Negative values should normally be chosen from the set of
4515 PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan-
4516 dard "no match" failure. The error number PCRE_ERROR_CALLOUT is
4517 reserved for use by callout functions; it will never be used by PCRE
4524 University Computing Service
4525 Cambridge CB2 3QH, England.
4530 Last updated: 12 November 2013
4531 Copyright (c) 1997-2013 University of Cambridge.
4532 ------------------------------------------------------------------------------
4535 PCRECOMPAT(3) Library Functions Manual PCRECOMPAT(3)
4540 PCRE - Perl-compatible regular expressions
4542 DIFFERENCES BETWEEN PCRE AND PERL
4544 This document describes the differences in the ways that PCRE and Perl
4545 handle regular expressions. The differences described here are with
4546 respect to Perl versions 5.10 and above.
4548 1. PCRE has only a subset of Perl's Unicode support. Details of what it
4549 does have are given in the pcreunicode page.
4551 2. PCRE allows repeat quantifiers only on parenthesized assertions, but
4552 they do not mean what you might think. For example, (?!a){3} does not
4553 assert that the next three characters are not "a". It just asserts that
4554 the next character is not "a" three times (in principle: PCRE optimizes
4555 this to run the assertion just once). Perl allows repeat quantifiers on
4556 other assertions such as \b, but these do not seem to have any use.
4558 3. Capturing subpatterns that occur inside negative lookahead asser-
4559 tions are counted, but their entries in the offsets vector are never
4560 set. Perl sometimes (but not always) sets its numerical variables from
4561 inside negative assertions.
4563 4. Though binary zero characters are supported in the subject string,
4564 they are not allowed in a pattern string because it is passed as a nor-
4565 mal C string, terminated by zero. The escape sequence \0 can be used in
4566 the pattern to represent a binary zero.
4568 5. The following Perl escape sequences are not supported: \l, \u, \L,
4569 \U, and \N when followed by a character name or Unicode value. (\N on
4570 its own, matching a non-newline character, is supported.) In fact these
4571 are implemented by Perl's general string-handling and are not part of
4572 its pattern matching engine. If any of these are encountered by PCRE,
4573 an error is generated by default. However, if the PCRE_JAVASCRIPT_COM-
4574 PAT option is set, \U and \u are interpreted as JavaScript interprets
4577 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE
4578 is built with Unicode character property support. The properties that
4579 can be tested with \p and \P are limited to the general category prop-
4580 erties such as Lu and Nd, script names such as Greek or Han, and the
4581 derived properties Any and L&. PCRE does support the Cs (surrogate)
4582 property, which Perl does not; the Perl documentation says "Because
4583 Perl hides the need for the user to understand the internal representa-
4584 tion of Unicode characters, there is no need to implement the somewhat
4585 messy concept of surrogates."
4587 7. PCRE does support the \Q...\E escape for quoting substrings. Charac-
4588 ters in between are treated as literals. This is slightly different
4589 from Perl in that $ and @ are also handled as literals inside the
4590 quotes. In Perl, they cause variable interpolation (but of course PCRE
4591 does not have variables). Note the following examples:
4593 Pattern PCRE matches Perl matches
4595 \Qabc$xyz\E abc$xyz abc followed by the
4597 \Qabc\$xyz\E abc\$xyz abc\$xyz
4598 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
4600 The \Q...\E sequence is recognized both inside and outside character
4603 8. Fairly obviously, PCRE does not support the (?{code}) and (??{code})
4604 constructions. However, there is support for recursive patterns. This
4605 is not available in Perl 5.8, but it is in Perl 5.10. Also, the PCRE
4606 "callout" feature allows an external function to be called during pat-
4607 tern matching. See the pcrecallout documentation for details.
4609 9. Subpatterns that are called as subroutines (whether or not recur-
4610 sively) are always treated as atomic groups in PCRE. This is like
4611 Python, but unlike Perl. Captured values that are set outside a sub-
4612 routine call can be reference from inside in PCRE, but not in Perl.
4613 There is a discussion that explains these differences in more detail in
4614 the section on recursion differences from Perl in the pcrepattern page.
4616 10. If any of the backtracking control verbs are used in a subpattern
4617 that is called as a subroutine (whether or not recursively), their
4618 effect is confined to that subpattern; it does not extend to the sur-
4619 rounding pattern. This is not always the case in Perl. In particular,
4620 if (*THEN) is present in a group that is called as a subroutine, its
4621 action is limited to that group, even if the group does not contain any
4622 | characters. Note that such subpatterns are processed as anchored at
4623 the point where they are tested.
4625 11. If a pattern contains more than one backtracking control verb, the
4626 first one that is backtracked onto acts. For example, in the pattern
4627 A(*COMMIT)B(*PRUNE)C a failure in B triggers (*COMMIT), but a failure
4628 in C triggers (*PRUNE). Perl's behaviour is more complex; in many cases
4629 it is the same as PCRE, but there are examples where it differs.
4631 12. Most backtracking verbs in assertions have their normal actions.
4632 They are not confined to the assertion.
4634 13. There are some differences that are concerned with the settings of
4635 captured strings when part of a pattern is repeated. For example,
4636 matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2
4637 unset, but in PCRE it is set to "b".
4639 14. PCRE's handling of duplicate subpattern numbers and duplicate sub-
4640 pattern names is not as general as Perl's. This is a consequence of the
4641 fact the PCRE works internally just with numbers, using an external ta-
4642 ble to translate between numbers and names. In particular, a pattern
4643 such as (?|(?<a>A)|(?<b)B), where the two capturing parentheses have
4644 the same number but different names, is not supported, and causes an
4645 error at compile time. If it were allowed, it would not be possible to
4646 distinguish which parentheses matched, because both names map to cap-
4647 turing subpattern number 1. To avoid this confusing situation, an error
4648 is given at compile time.
4650 15. Perl recognizes comments in some places that PCRE does not, for
4651 example, between the ( and ? at the start of a subpattern. If the /x
4652 modifier is set, Perl allows white space between ( and ? (though cur-
4653 rent Perls warn that this is deprecated) but PCRE never does, even if
4654 the PCRE_EXTENDED option is set.
4656 16. Perl, when in warning mode, gives warnings for character classes
4657 such as [A-\d] or [a-[:digit:]]. It then treats the hyphens as liter-
4658 als. PCRE has no warning features, so it gives an error in these cases
4659 because they are almost certainly user mistakes.
4661 17. In PCRE, the upper/lower case character properties Lu and Ll are
4662 not affected when case-independent matching is specified. For example,
4663 \p{Lu} always matches an upper case letter. I think Perl has changed in
4664 this respect; in the release at the time of writing (5.16), \p{Lu} and
4665 \p{Ll} match all letters, regardless of case, when case independence is
4668 18. PCRE provides some extensions to the Perl regular expression facil-
4669 ities. Perl 5.10 includes new features that are not in earlier ver-
4670 sions of Perl, some of which (such as named parentheses) have been in
4671 PCRE for some time. This list is with respect to Perl 5.10:
4673 (a) Although lookbehind assertions in PCRE must match fixed length
4674 strings, each alternative branch of a lookbehind assertion can match a
4675 different length of string. Perl requires them all to have the same
4678 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $
4679 meta-character matches only at the very end of the string.
4681 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe-
4682 cial meaning is faulted. Otherwise, like Perl, the backslash is quietly
4683 ignored. (Perl can be made to issue a warning.)
4685 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti-
4686 fiers is inverted, that is, by default they are not greedy, but if fol-
4687 lowed by a question mark they are.
4689 (e) PCRE_ANCHORED can be used at matching time to force a pattern to be
4690 tried only at the first matching position in the subject string.
4692 (f) The PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART,
4693 and PCRE_NO_AUTO_CAPTURE options for pcre_exec() have no Perl equiva-
4696 (g) The \R escape sequence can be restricted to match only CR, LF, or
4697 CRLF by the PCRE_BSR_ANYCRLF option.
4699 (h) The callout facility is PCRE-specific.
4701 (i) The partial matching facility is PCRE-specific.
4703 (j) Patterns compiled by PCRE can be saved and re-used at a later time,
4704 even on different hosts that have the other endianness. However, this
4705 does not apply to optimized data created by the just-in-time compiler.
4707 (k) The alternative matching functions (pcre_dfa_exec(),
4708 pcre16_dfa_exec() and pcre32_dfa_exec(),) match in a different way and
4709 are not Perl-compatible.
4711 (l) PCRE recognizes some special sequences such as (*CR) at the start
4712 of a pattern that set overall options that cannot be changed within the
4719 University Computing Service
4720 Cambridge CB2 3QH, England.
4725 Last updated: 10 November 2013
4726 Copyright (c) 1997-2013 University of Cambridge.
4727 ------------------------------------------------------------------------------
4730 PCREPATTERN(3) Library Functions Manual PCREPATTERN(3)
4735 PCRE - Perl-compatible regular expressions
4737 PCRE REGULAR EXPRESSION DETAILS
4739 The syntax and semantics of the regular expressions that are supported
4740 by PCRE are described in detail below. There is a quick-reference syn-
4741 tax summary in the pcresyntax page. PCRE tries to match Perl syntax and
4742 semantics as closely as it can. PCRE also supports some alternative
4743 regular expression syntax (which does not conflict with the Perl syn-
4744 tax) in order to provide some compatibility with regular expressions in
4745 Python, .NET, and Oniguruma.
4747 Perl's regular expressions are described in its own documentation, and
4748 regular expressions in general are covered in a number of books, some
4749 of which have copious examples. Jeffrey Friedl's "Mastering Regular
4750 Expressions", published by O'Reilly, covers regular expressions in
4751 great detail. This description of PCRE's regular expressions is
4752 intended as reference material.
4754 This document discusses the patterns that are supported by PCRE when
4755 one its main matching functions, pcre_exec() (8-bit) or
4756 pcre[16|32]_exec() (16- or 32-bit), is used. PCRE also has alternative
4757 matching functions, pcre_dfa_exec() and pcre[16|32_dfa_exec(), which
4758 match using a different algorithm that is not Perl-compatible. Some of
4759 the features discussed below are not available when DFA matching is
4760 used. The advantages and disadvantages of the alternative functions,
4761 and how they differ from the normal functions, are discussed in the
4765 SPECIAL START-OF-PATTERN ITEMS
4767 A number of options that can be passed to pcre_compile() can also be
4768 set by special items at the start of a pattern. These are not Perl-com-
4769 patible, but are provided to make these options accessible to pattern
4770 writers who are not able to change the program that processes the pat-
4771 tern. Any number of these items may appear, but they must all be
4772 together right at the start of the pattern string, and the letters must
4777 The original operation of PCRE was on strings of one-byte characters.
4778 However, there is now also support for UTF-8 strings in the original
4779 library, an extra library that supports 16-bit and UTF-16 character
4780 strings, and a third library that supports 32-bit and UTF-32 character
4781 strings. To use these features, PCRE must be built to include appropri-
4782 ate support. When using UTF strings you must either call the compiling
4783 function with the PCRE_UTF8, PCRE_UTF16, or PCRE_UTF32 option, or the
4784 pattern must start with one of these special sequences:
4791 (*UTF) is a generic sequence that can be used with any of the
4792 libraries. Starting a pattern with such a sequence is equivalent to
4793 setting the relevant option. How setting a UTF mode affects pattern
4794 matching is mentioned in several places below. There is also a summary
4795 of features in the pcreunicode page.
4797 Some applications that allow their users to supply patterns may wish to
4798 restrict them to non-UTF data for security reasons. If the
4799 PCRE_NEVER_UTF option is set at compile time, (*UTF) etc. are not
4800 allowed, and their appearance causes an error.
4802 Unicode property support
4804 Another special sequence that may appear at the start of a pattern is
4805 (*UCP). This has the same effect as setting the PCRE_UCP option: it
4806 causes sequences such as \d and \w to use Unicode properties to deter-
4807 mine character types, instead of recognizing only characters with codes
4808 less than 128 via a lookup table.
4810 Disabling auto-possessification
4812 If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as
4813 setting the PCRE_NO_AUTO_POSSESS option at compile time. This stops
4814 PCRE from making quantifiers possessive when what follows cannot match
4815 the repeated item. For example, by default a+b is treated as a++b. For
4816 more details, see the pcreapi documentation.
4818 Disabling start-up optimizations
4820 If a pattern starts with (*NO_START_OPT), it has the same effect as
4821 setting the PCRE_NO_START_OPTIMIZE option either at compile or matching
4822 time. This disables several optimizations for quickly reaching "no
4823 match" results. For more details, see the pcreapi documentation.
4827 PCRE supports five different conventions for indicating line breaks in
4828 strings: a single CR (carriage return) character, a single LF (line-
4829 feed) character, the two-character sequence CRLF, any of the three pre-
4830 ceding, or any Unicode newline sequence. The pcreapi page has further
4831 discussion about newlines, and shows how to set the newline convention
4832 in the options arguments for the compiling and matching functions.
4834 It is also possible to specify a newline convention by starting a pat-
4835 tern string with one of the following five sequences:
4837 (*CR) carriage return
4839 (*CRLF) carriage return, followed by linefeed
4840 (*ANYCRLF) any of the three above
4841 (*ANY) all Unicode newline sequences
4843 These override the default and the options given to the compiling func-
4844 tion. For example, on a Unix system where LF is the default newline
4845 sequence, the pattern
4849 changes the convention to CR. That pattern matches "a\nb" because LF is
4850 no longer a newline. If more than one of these settings is present, the
4853 The newline convention affects where the circumflex and dollar asser-
4854 tions are true. It also affects the interpretation of the dot metachar-
4855 acter when PCRE_DOTALL is not set, and the behaviour of \N. However, it
4856 does not affect what the \R escape sequence matches. By default, this
4857 is any Unicode newline sequence, for Perl compatibility. However, this
4858 can be changed; see the description of \R in the section entitled "New-
4859 line sequences" below. A change of \R setting can be combined with a
4860 change of newline convention.
4862 Setting match and recursion limits
4864 The caller of pcre_exec() can set a limit on the number of times the
4865 internal match() function is called and on the maximum depth of recur-
4866 sive calls. These facilities are provided to catch runaway matches that
4867 are provoked by patterns with huge matching trees (a typical example is
4868 a pattern with nested unlimited repeats) and to avoid running out of
4869 system stack by too much recursion. When one of these limits is
4870 reached, pcre_exec() gives an error return. The limits can also be set
4871 by items at the start of the pattern of the form
4874 (*LIMIT_RECURSION=d)
4876 where d is any number of decimal digits. However, the value of the set-
4877 ting must be less than the value set (or defaulted) by the caller of
4878 pcre_exec() for it to have any effect. In other words, the pattern
4879 writer can lower the limits set by the programmer, but not raise them.
4880 If there is more than one setting of one of these limits, the lower
4884 EBCDIC CHARACTER CODES
4886 PCRE can be compiled to run in an environment that uses EBCDIC as its
4887 character code rather than ASCII or Unicode (typically a mainframe sys-
4888 tem). In the sections below, character code values are ASCII or Uni-
4889 code; in an EBCDIC environment these characters may have different code
4890 values, and there are no code points greater than 255.
4893 CHARACTERS AND METACHARACTERS
4895 A regular expression is a pattern that is matched against a subject
4896 string from left to right. Most characters stand for themselves in a
4897 pattern, and match the corresponding characters in the subject. As a
4898 trivial example, the pattern
4902 matches a portion of a subject string that is identical to itself. When
4903 caseless matching is specified (the PCRE_CASELESS option), letters are
4904 matched independently of case. In a UTF mode, PCRE always understands
4905 the concept of case for characters whose values are less than 128, so
4906 caseless matching is always possible. For characters with higher val-
4907 ues, the concept of case is supported if PCRE is compiled with Unicode
4908 property support, but not otherwise. If you want to use caseless
4909 matching for characters 128 and above, you must ensure that PCRE is
4910 compiled with Unicode property support as well as with UTF support.
4912 The power of regular expressions comes from the ability to include
4913 alternatives and repetitions in the pattern. These are encoded in the
4914 pattern by the use of metacharacters, which do not stand for themselves
4915 but instead are interpreted in some special way.
4917 There are two different sets of metacharacters: those that are recog-
4918 nized anywhere in the pattern except within square brackets, and those
4919 that are recognized within square brackets. Outside square brackets,
4920 the metacharacters are as follows:
4922 \ general escape character with several uses
4923 ^ assert start of string (or line, in multiline mode)
4924 $ assert end of string (or line, in multiline mode)
4925 . match any character except newline (by default)
4926 [ start character class definition
4927 | start of alternative branch
4930 ? extends the meaning of (
4931 also 0 or 1 quantifier
4932 also quantifier minimizer
4933 * 0 or more quantifier
4934 + 1 or more quantifier
4935 also "possessive quantifier"
4936 { start min/max quantifier
4938 Part of a pattern that is in square brackets is called a "character
4939 class". In a character class the only metacharacters are:
4941 \ general escape character
4942 ^ negate the class, but only if the first character
4943 - indicates character range
4944 [ POSIX character class (only if followed by POSIX
4946 ] terminates the character class
4948 The following sections describe the use of each of the metacharacters.
4953 The backslash character has several uses. Firstly, if it is followed by
4954 a character that is not a number or a letter, it takes away any special
4955 meaning that character may have. This use of backslash as an escape
4956 character applies both inside and outside character classes.
4958 For example, if you want to match a * character, you write \* in the
4959 pattern. This escaping action applies whether or not the following
4960 character would otherwise be interpreted as a metacharacter, so it is
4961 always safe to precede a non-alphanumeric with backslash to specify
4962 that it stands for itself. In particular, if you want to match a back-
4963 slash, you write \\.
4965 In a UTF mode, only ASCII numbers and letters have any special meaning
4966 after a backslash. All other characters (in particular, those whose
4967 codepoints are greater than 127) are treated as literals.
4969 If a pattern is compiled with the PCRE_EXTENDED option, most white
4970 space in the pattern (other than in a character class), and characters
4971 between a # outside a character class and the next newline, inclusive,
4972 are ignored. An escaping backslash can be used to include a white space
4973 or # character as part of the pattern.
4975 If you want to remove the special meaning from a sequence of charac-
4976 ters, you can do so by putting them between \Q and \E. This is differ-
4977 ent from Perl in that $ and @ are handled as literals in \Q...\E
4978 sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
4979 tion. Note the following examples:
4981 Pattern PCRE matches Perl matches
4983 \Qabc$xyz\E abc$xyz abc followed by the
4985 \Qabc\$xyz\E abc\$xyz abc\$xyz
4986 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
4988 The \Q...\E sequence is recognized both inside and outside character
4989 classes. An isolated \E that is not preceded by \Q is ignored. If \Q
4990 is not followed by \E later in the pattern, the literal interpretation
4991 continues to the end of the pattern (that is, \E is assumed at the
4992 end). If the isolated \Q is inside a character class, this causes an
4993 error, because the character class is not terminated.
4995 Non-printing characters
4997 A second use of backslash provides a way of encoding non-printing char-
4998 acters in patterns in a visible manner. There is no restriction on the
4999 appearance of non-printing characters, apart from the binary zero that
5000 terminates a pattern, but when a pattern is being prepared by text
5001 editing, it is often easier to use one of the following escape
5002 sequences than the binary character it represents. In an ASCII or Uni-
5003 code environment, these escapes are as follows:
5005 \a alarm, that is, the BEL character (hex 07)
5006 \cx "control-x", where x is any ASCII character
5008 \f form feed (hex 0C)
5009 \n linefeed (hex 0A)
5010 \r carriage return (hex 0D)
5012 \0dd character with octal code 0dd
5013 \ddd character with octal code ddd, or back reference
5014 \o{ddd..} character with octal code ddd..
5015 \xhh character with hex code hh
5016 \x{hhh..} character with hex code hhh.. (non-JavaScript mode)
5017 \uhhhh character with hex code hhhh (JavaScript mode only)
5019 The precise effect of \cx on ASCII characters is as follows: if x is a
5020 lower case letter, it is converted to upper case. Then bit 6 of the
5021 character (hex 40) is inverted. Thus \cA to \cZ become hex 01 to hex 1A
5022 (A is 41, Z is 5A), but \c{ becomes hex 3B ({ is 7B), and \c; becomes
5023 hex 7B (; is 3B). If the data item (byte or 16-bit value) following \c
5024 has a value greater than 127, a compile-time error occurs. This locks
5025 out non-ASCII characters in all modes.
5027 When PCRE is compiled in EBCDIC mode, \a, \e, \f, \n, \r, and \t gener-
5028 ate the appropriate EBCDIC code values. The \c escape is processed as
5029 specified for Perl in the perlebcdic document. The only characters that
5030 are allowed after \c are A-Z, a-z, or one of @, [, \, ], ^, _, or ?.
5031 Any other character provokes a compile-time error. The sequence \@
5032 encodes character code 0; the letters (in either case) encode charac-
5033 ters 1-26 (hex 01 to hex 1A); [, \, ], ^, and _ encode characters 27-31
5034 (hex 1B to hex 1F), and \? becomes either 255 (hex FF) or 95 (hex 5F).
5036 Thus, apart from \?, these escapes generate the same character code
5037 values as they do in an ASCII environment, though the meanings of the
5038 values mostly differ. For example, \G always generates code value 7,
5039 which is BEL in ASCII but DEL in EBCDIC.
5041 The sequence \? generates DEL (127, hex 7F) in an ASCII environment,
5042 but because 127 is not a control character in EBCDIC, Perl makes it
5043 generate the APC character. Unfortunately, there are several variants
5044 of EBCDIC. In most of them the APC character has the value 255 (hex
5045 FF), but in the one Perl calls POSIX-BC its value is 95 (hex 5F). If
5046 certain other characters have POSIX-BC values, PCRE makes \? generate
5047 95; otherwise it generates 255.
5049 After \0 up to two further octal digits are read. If there are fewer
5050 than two digits, just those that are present are used. Thus the
5051 sequence \0\x\015 specifies two binary zeros followed by a CR character
5052 (code value 13). Make sure you supply two digits after the initial zero
5053 if the pattern character that follows is itself an octal digit.
5055 The escape \o must be followed by a sequence of octal digits, enclosed
5056 in braces. An error occurs if this is not the case. This escape is a
5057 recent addition to Perl; it provides way of specifying character code
5058 points as octal numbers greater than 0777, and it also allows octal
5059 numbers and back references to be unambiguously specified.
5061 For greater clarity and unambiguity, it is best to avoid following \ by
5062 a digit greater than zero. Instead, use \o{} or \x{} to specify charac-
5063 ter numbers, and \g{} to specify back references. The following para-
5064 graphs describe the old, ambiguous syntax.
5066 The handling of a backslash followed by a digit other than 0 is compli-
5067 cated, and Perl has changed in recent releases, causing PCRE also to
5068 change. Outside a character class, PCRE reads the digit and any follow-
5069 ing digits as a decimal number. If the number is less than 8, or if
5070 there have been at least that many previous capturing left parentheses
5071 in the expression, the entire sequence is taken as a back reference. A
5072 description of how this works is given later, following the discussion
5073 of parenthesized subpatterns.
5075 Inside a character class, or if the decimal number following \ is
5076 greater than 7 and there have not been that many capturing subpatterns,
5077 PCRE handles \8 and \9 as the literal characters "8" and "9", and oth-
5078 erwise re-reads up to three octal digits following the backslash, using
5079 them to generate a data character. Any subsequent digits stand for
5080 themselves. For example:
5082 \040 is another way of writing an ASCII space
5083 \40 is the same, provided there are fewer than 40
5084 previous capturing subpatterns
5085 \7 is always a back reference
5086 \11 might be a back reference, or another way of
5088 \011 is always a tab
5089 \0113 is a tab followed by the character "3"
5090 \113 might be a back reference, otherwise the
5091 character with octal code 113
5092 \377 might be a back reference, otherwise
5093 the value 255 (decimal)
5094 \81 is either a back reference, or the two
5095 characters "8" and "1"
5097 Note that octal values of 100 or greater that are specified using this
5098 syntax must not be introduced by a leading zero, because no more than
5099 three octal digits are ever read.
5101 By default, after \x that is not followed by {, from zero to two hexa-
5102 decimal digits are read (letters can be in upper or lower case). Any
5103 number of hexadecimal digits may appear between \x{ and }. If a charac-
5104 ter other than a hexadecimal digit appears between \x{ and }, or if
5105 there is no terminating }, an error occurs.
5107 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x
5108 is as just described only when it is followed by two hexadecimal dig-
5109 its. Otherwise, it matches a literal "x" character. In JavaScript
5110 mode, support for code points greater than 256 is provided by \u, which
5111 must be followed by four hexadecimal digits; otherwise it matches a
5112 literal "u" character.
5114 Characters whose value is less than 256 can be defined by either of the
5115 two syntaxes for \x (or by \u in JavaScript mode). There is no differ-
5116 ence in the way they are handled. For example, \xdc is exactly the same
5117 as \x{dc} (or \u00dc in JavaScript mode).
5119 Constraints on character values
5121 Characters that are specified using octal or hexadecimal numbers are
5122 limited to certain values, as follows:
5124 8-bit non-UTF mode less than 0x100
5125 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
5126 16-bit non-UTF mode less than 0x10000
5127 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
5128 32-bit non-UTF mode less than 0x100000000
5129 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
5131 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-
5132 called "surrogate" codepoints), and 0xffef.
5134 Escape sequences in character classes
5136 All the sequences that define a single character value can be used both
5137 inside and outside character classes. In addition, inside a character
5138 class, \b is interpreted as the backspace character (hex 08).
5140 \N is not allowed in a character class. \B, \R, and \X are not special
5141 inside a character class. Like other unrecognized escape sequences,
5142 they are treated as the literal characters "B", "R", and "X" by
5143 default, but cause an error if the PCRE_EXTRA option is set. Outside a
5144 character class, these sequences have different meanings.
5146 Unsupported escape sequences
5148 In Perl, the sequences \l, \L, \u, and \U are recognized by its string
5149 handler and used to modify the case of following characters. By
5150 default, PCRE does not support these escape sequences. However, if the
5151 PCRE_JAVASCRIPT_COMPAT option is set, \U matches a "U" character, and
5152 \u can be used to define a character by code point, as described in the
5155 Absolute and relative back references
5157 The sequence \g followed by an unsigned or a negative number, option-
5158 ally enclosed in braces, is an absolute or relative back reference. A
5159 named back reference can be coded as \g{name}. Back references are dis-
5160 cussed later, following the discussion of parenthesized subpatterns.
5162 Absolute and relative subroutine calls
5164 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
5165 name or a number enclosed either in angle brackets or single quotes, is
5166 an alternative syntax for referencing a subpattern as a "subroutine".
5167 Details are discussed later. Note that \g{...} (Perl syntax) and
5168 \g<...> (Oniguruma syntax) are not synonymous. The former is a back
5169 reference; the latter is a subroutine call.
5171 Generic character types
5173 Another use of backslash is for specifying generic character types:
5175 \d any decimal digit
5176 \D any character that is not a decimal digit
5177 \h any horizontal white space character
5178 \H any character that is not a horizontal white space character
5179 \s any white space character
5180 \S any character that is not a white space character
5181 \v any vertical white space character
5182 \V any character that is not a vertical white space character
5183 \w any "word" character
5184 \W any "non-word" character
5186 There is also the single sequence \N, which matches a non-newline char-
5187 acter. This is the same as the "." metacharacter when PCRE_DOTALL is
5188 not set. Perl also uses \N to match characters by name; PCRE does not
5191 Each pair of lower and upper case escape sequences partitions the com-
5192 plete set of characters into two disjoint sets. Any given character
5193 matches one, and only one, of each pair. The sequences can appear both
5194 inside and outside character classes. They each match one character of
5195 the appropriate type. If the current matching point is at the end of
5196 the subject string, all of them fail, because there is no character to
5199 For compatibility with Perl, \s did not used to match the VT character
5200 (code 11), which made it different from the the POSIX "space" class.
5201 However, Perl added VT at release 5.18, and PCRE followed suit at
5202 release 8.34. The default \s characters are now HT (9), LF (10), VT
5203 (11), FF (12), CR (13), and space (32), which are defined as white
5204 space in the "C" locale. This list may vary if locale-specific matching
5205 is taking place. For example, in some locales the "non-breaking space"
5206 character (\xA0) is recognized as white space, and in others the VT
5209 A "word" character is an underscore or any character that is a letter
5210 or digit. By default, the definition of letters and digits is con-
5211 trolled by PCRE's low-valued character tables, and may vary if locale-
5212 specific matching is taking place (see "Locale support" in the pcreapi
5213 page). For example, in a French locale such as "fr_FR" in Unix-like
5214 systems, or "french" in Windows, some character codes greater than 127
5215 are used for accented letters, and these are then matched by \w. The
5216 use of locales with Unicode is discouraged.
5218 By default, characters whose code points are greater than 127 never
5219 match \d, \s, or \w, and always match \D, \S, and \W, although this may
5220 vary for characters in the range 128-255 when locale-specific matching
5221 is happening. These escape sequences retain their original meanings
5222 from before Unicode support was available, mainly for efficiency rea-
5223 sons. If PCRE is compiled with Unicode property support, and the
5224 PCRE_UCP option is set, the behaviour is changed so that Unicode prop-
5225 erties are used to determine character types, as follows:
5227 \d any character that matches \p{Nd} (decimal digit)
5228 \s any character that matches \p{Z} or \h or \v
5229 \w any character that matches \p{L} or \p{N}, plus underscore
5231 The upper case escapes match the inverse sets of characters. Note that
5232 \d matches only decimal digits, whereas \w matches any Unicode digit,
5233 as well as any Unicode letter, and underscore. Note also that PCRE_UCP
5234 affects \b, and \B because they are defined in terms of \w and \W.
5235 Matching these sequences is noticeably slower when PCRE_UCP is set.
5237 The sequences \h, \H, \v, and \V are features that were added to Perl
5238 at release 5.10. In contrast to the other sequences, which match only
5239 ASCII characters by default, these always match certain high-valued
5240 code points, whether or not PCRE_UCP is set. The horizontal space char-
5243 U+0009 Horizontal tab (HT)
5245 U+00A0 Non-break space
5246 U+1680 Ogham space mark
5247 U+180E Mongolian vowel separator
5252 U+2004 Three-per-em space
5253 U+2005 Four-per-em space
5254 U+2006 Six-per-em space
5256 U+2008 Punctuation space
5259 U+202F Narrow no-break space
5260 U+205F Medium mathematical space
5261 U+3000 Ideographic space
5263 The vertical space characters are:
5265 U+000A Linefeed (LF)
5266 U+000B Vertical tab (VT)
5267 U+000C Form feed (FF)
5268 U+000D Carriage return (CR)
5269 U+0085 Next line (NEL)
5270 U+2028 Line separator
5271 U+2029 Paragraph separator
5273 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than
5278 Outside a character class, by default, the escape sequence \R matches
5279 any Unicode newline sequence. In 8-bit non-UTF-8 mode \R is equivalent
5282 (?>\r\n|\n|\x0b|\f|\r|\x85)
5284 This is an example of an "atomic group", details of which are given
5285 below. This particular group matches either the two-character sequence
5286 CR followed by LF, or one of the single characters LF (linefeed,
5287 U+000A), VT (vertical tab, U+000B), FF (form feed, U+000C), CR (car-
5288 riage return, U+000D), or NEL (next line, U+0085). The two-character
5289 sequence is treated as a single unit that cannot be split.
5291 In other modes, two additional characters whose codepoints are greater
5292 than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa-
5293 rator, U+2029). Unicode character property support is not needed for
5294 these characters to be recognized.
5296 It is possible to restrict \R to match only CR, LF, or CRLF (instead of
5297 the complete set of Unicode line endings) by setting the option
5298 PCRE_BSR_ANYCRLF either at compile time or when the pattern is matched.
5299 (BSR is an abbrevation for "backslash R".) This can be made the default
5300 when PCRE is built; if this is the case, the other behaviour can be
5301 requested via the PCRE_BSR_UNICODE option. It is also possible to
5302 specify these settings by starting a pattern string with one of the
5303 following sequences:
5305 (*BSR_ANYCRLF) CR, LF, or CRLF only
5306 (*BSR_UNICODE) any Unicode newline sequence
5308 These override the default and the options given to the compiling func-
5309 tion, but they can themselves be overridden by options given to a
5310 matching function. Note that these special settings, which are not
5311 Perl-compatible, are recognized only at the very start of a pattern,
5312 and that they must be in upper case. If more than one of them is
5313 present, the last one is used. They can be combined with a change of
5314 newline convention; for example, a pattern can start with:
5316 (*ANY)(*BSR_ANYCRLF)
5318 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF)
5319 or (*UCP) special sequences. Inside a character class, \R is treated as
5320 an unrecognized escape sequence, and so matches the letter "R" by
5321 default, but causes an error if PCRE_EXTRA is set.
5323 Unicode character properties
5325 When PCRE is built with Unicode character property support, three addi-
5326 tional escape sequences that match characters with specific properties
5327 are available. When in 8-bit non-UTF-8 mode, these sequences are of
5328 course limited to testing characters whose codepoints are less than
5329 256, but they do work in this mode. The extra escape sequences are:
5331 \p{xx} a character with the xx property
5332 \P{xx} a character without the xx property
5333 \X a Unicode extended grapheme cluster
5335 The property names represented by xx above are limited to the Unicode
5336 script names, the general category properties, "Any", which matches any
5337 character (including newline), and some special PCRE properties
5338 (described in the next section). Other Perl properties such as "InMu-
5339 sicalSymbols" are not currently supported by PCRE. Note that \P{Any}
5340 does not match any characters, so always causes a match failure.
5342 Sets of Unicode characters are defined as belonging to certain scripts.
5343 A character from one of these sets can be matched using a script name.
5349 Those that are not part of an identified script are lumped together as
5350 "Common". The current list of scripts is:
5352 Arabic, Armenian, Avestan, Balinese, Bamum, Bassa_Vah, Batak, Bengali,
5353 Bopomofo, Brahmi, Braille, Buginese, Buhid, Canadian_Aboriginal, Car-
5354 ian, Caucasian_Albanian, Chakma, Cham, Cherokee, Common, Coptic, Cunei-
5355 form, Cypriot, Cyrillic, Deseret, Devanagari, Duployan, Egyptian_Hiero-
5356 glyphs, Elbasan, Ethiopic, Georgian, Glagolitic, Gothic, Grantha,
5357 Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana,
5358 Imperial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscrip-
5359 tional_Parthian, Javanese, Kaithi, Kannada, Katakana, Kayah_Li,
5360 Kharoshthi, Khmer, Khojki, Khudawadi, Lao, Latin, Lepcha, Limbu, Lin-
5361 ear_A, Linear_B, Lisu, Lycian, Lydian, Mahajani, Malayalam, Mandaic,
5362 Manichaean, Meetei_Mayek, Mende_Kikakui, Meroitic_Cursive,
5363 Meroitic_Hieroglyphs, Miao, Modi, Mongolian, Mro, Myanmar, Nabataean,
5364 New_Tai_Lue, Nko, Ogham, Ol_Chiki, Old_Italic, Old_North_Arabian,
5365 Old_Permic, Old_Persian, Old_South_Arabian, Old_Turkic, Oriya, Osmanya,
5366 Pahawh_Hmong, Palmyrene, Pau_Cin_Hau, Phags_Pa, Phoenician,
5367 Psalter_Pahlavi, Rejang, Runic, Samaritan, Saurashtra, Sharada, Sha-
5368 vian, Siddham, Sinhala, Sora_Sompeng, Sundanese, Syloti_Nagri, Syriac,
5369 Tagalog, Tagbanwa, Tai_Le, Tai_Tham, Tai_Viet, Takri, Tamil, Telugu,
5370 Thaana, Thai, Tibetan, Tifinagh, Tirhuta, Ugaritic, Vai, Warang_Citi,
5373 Each character has exactly one Unicode general category property, spec-
5374 ified by a two-letter abbreviation. For compatibility with Perl, nega-
5375 tion can be specified by including a circumflex between the opening
5376 brace and the property name. For example, \p{^Lu} is the same as
5379 If only one letter is specified with \p or \P, it includes all the gen-
5380 eral category properties that start with that letter. In this case, in
5381 the absence of negation, the curly brackets in the escape sequence are
5382 optional; these two examples have the same effect:
5387 The following general category property codes are supported:
5397 Ll Lower case letter
5400 Lt Title case letter
5401 Lu Upper case letter
5414 Pc Connector punctuation
5416 Pe Close punctuation
5417 Pf Final punctuation
5418 Pi Initial punctuation
5419 Po Other punctuation
5425 Sm Mathematical symbol
5430 Zp Paragraph separator
5433 The special property L& is also supported: it matches a character that
5434 has the Lu, Ll, or Lt property, in other words, a letter that is not
5435 classified as a modifier or "other".
5437 The Cs (Surrogate) property applies only to characters in the range
5438 U+D800 to U+DFFF. Such characters are not valid in Unicode strings and
5439 so cannot be tested by PCRE, unless UTF validity checking has been
5440 turned off (see the discussion of PCRE_NO_UTF8_CHECK,
5441 PCRE_NO_UTF16_CHECK and PCRE_NO_UTF32_CHECK in the pcreapi page). Perl
5442 does not support the Cs property.
5444 The long synonyms for property names that Perl supports (such as
5445 \p{Letter}) are not supported by PCRE, nor is it permitted to prefix
5446 any of these properties with "Is".
5448 No character that is in the Unicode table has the Cn (unassigned) prop-
5449 erty. Instead, this property is assumed for any code point that is not
5450 in the Unicode table.
5452 Specifying caseless matching does not affect these escape sequences.
5453 For example, \p{Lu} always matches only upper case letters. This is
5454 different from the behaviour of current versions of Perl.
5456 Matching characters by Unicode property is not fast, because PCRE has
5457 to do a multistage table lookup in order to find a character's prop-
5458 erty. That is why the traditional escape sequences such as \d and \w do
5459 not use Unicode properties in PCRE by default, though you can make them
5460 do so by setting the PCRE_UCP option or by starting the pattern with
5463 Extended grapheme clusters
5465 The \X escape matches any number of Unicode characters that form an
5466 "extended grapheme cluster", and treats the sequence as an atomic group
5467 (see below). Up to and including release 8.31, PCRE matched an ear-
5468 lier, simpler definition that was equivalent to
5472 That is, it matched a character without the "mark" property, followed
5473 by zero or more characters with the "mark" property. Characters with
5474 the "mark" property are typically non-spacing accents that affect the
5475 preceding character.
5477 This simple definition was extended in Unicode to include more compli-
5478 cated kinds of composite character by giving each character a grapheme
5479 breaking property, and creating rules that use these properties to
5480 define the boundaries of extended grapheme clusters. In releases of
5481 PCRE later than 8.31, \X matches one of these clusters.
5483 \X always matches at least one character. Then it decides whether to
5484 add additional characters according to the following rules for ending a
5487 1. End at the end of the subject string.
5489 2. Do not end between CR and LF; otherwise end after any control char-
5492 3. Do not break Hangul (a Korean script) syllable sequences. Hangul
5493 characters are of five types: L, V, T, LV, and LVT. An L character may
5494 be followed by an L, V, LV, or LVT character; an LV or V character may
5495 be followed by a V or T character; an LVT or T character may be follwed
5496 only by a T character.
5498 4. Do not end before extending characters or spacing marks. Characters
5499 with the "mark" property always have the "extend" grapheme breaking
5502 5. Do not end after prepend characters.
5504 6. Otherwise, end the cluster.
5506 PCRE's additional properties
5508 As well as the standard Unicode properties described above, PCRE sup-
5509 ports four more that make it possible to convert traditional escape
5510 sequences such as \w and \s to use Unicode properties. PCRE uses these
5511 non-standard, non-Perl properties internally when PCRE_UCP is set. How-
5512 ever, they may also be used explicitly. These properties are:
5514 Xan Any alphanumeric character
5515 Xps Any POSIX space character
5516 Xsp Any Perl space character
5517 Xwd Any Perl "word" character
5519 Xan matches characters that have either the L (letter) or the N (num-
5520 ber) property. Xps matches the characters tab, linefeed, vertical tab,
5521 form feed, or carriage return, and any other character that has the Z
5522 (separator) property. Xsp is the same as Xps; it used to exclude ver-
5523 tical tab, for Perl compatibility, but Perl changed, and so PCRE fol-
5524 lowed at release 8.34. Xwd matches the same characters as Xan, plus
5527 There is another non-standard property, Xuc, which matches any charac-
5528 ter that can be represented by a Universal Character Name in C++ and
5529 other programming languages. These are the characters $, @, ` (grave
5530 accent), and all characters with Unicode code points greater than or
5531 equal to U+00A0, except for the surrogates U+D800 to U+DFFF. Note that
5532 most base (ASCII) characters are excluded. (Universal Character Names
5533 are of the form \uHHHH or \UHHHHHHHH where H is a hexadecimal digit.
5534 Note that the Xuc property does not match these sequences but the char-
5535 acters that they represent.)
5537 Resetting the match start
5539 The escape sequence \K causes any previously matched characters not to
5540 be included in the final matched sequence. For example, the pattern:
5544 matches "foobar", but reports that it has matched "bar". This feature
5545 is similar to a lookbehind assertion (described below). However, in
5546 this case, the part of the subject before the real match does not have
5547 to be of fixed length, as lookbehind assertions do. The use of \K does
5548 not interfere with the setting of captured substrings. For example,
5553 matches "foobar", the first substring is still set to "foo".
5555 Perl documents that the use of \K within assertions is "not well
5556 defined". In PCRE, \K is acted upon when it occurs inside positive
5557 assertions, but is ignored in negative assertions. Note that when a
5558 pattern such as (?=ab\K) matches, the reported start of the match can
5559 be greater than the end of the match.
5563 The final use of backslash is for certain simple assertions. An asser-
5564 tion specifies a condition that has to be met at a particular point in
5565 a match, without consuming any characters from the subject string. The
5566 use of subpatterns for more complicated assertions is described below.
5567 The backslashed assertions are:
5569 \b matches at a word boundary
5570 \B matches when not at a word boundary
5571 \A matches at the start of the subject
5572 \Z matches at the end of the subject
5573 also matches before a newline at the end of the subject
5574 \z matches only at the end of the subject
5575 \G matches at the first matching position in the subject
5577 Inside a character class, \b has a different meaning; it matches the
5578 backspace character. If any other of these assertions appears in a
5579 character class, by default it matches the corresponding literal char-
5580 acter (for example, \B matches the letter B). However, if the
5581 PCRE_EXTRA option is set, an "invalid escape sequence" error is gener-
5584 A word boundary is a position in the subject string where the current
5585 character and the previous character do not both match \w or \W (i.e.
5586 one matches \w and the other matches \W), or the start or end of the
5587 string if the first or last character matches \w, respectively. In a
5588 UTF mode, the meanings of \w and \W can be changed by setting the
5589 PCRE_UCP option. When this is done, it also affects \b and \B. Neither
5590 PCRE nor Perl has a separate "start of word" or "end of word" metase-
5591 quence. However, whatever follows \b normally determines which it is.
5592 For example, the fragment \ba matches "a" at the start of a word.
5594 The \A, \Z, and \z assertions differ from the traditional circumflex
5595 and dollar (described in the next section) in that they only ever match
5596 at the very start and end of the subject string, whatever options are
5597 set. Thus, they are independent of multiline mode. These three asser-
5598 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
5599 affect only the behaviour of the circumflex and dollar metacharacters.
5600 However, if the startoffset argument of pcre_exec() is non-zero, indi-
5601 cating that matching is to start at a point other than the beginning of
5602 the subject, \A can never match. The difference between \Z and \z is
5603 that \Z matches before a newline at the end of the string as well as at
5604 the very end, whereas \z matches only at the end.
5606 The \G assertion is true only when the current matching position is at
5607 the start point of the match, as specified by the startoffset argument
5608 of pcre_exec(). It differs from \A when the value of startoffset is
5609 non-zero. By calling pcre_exec() multiple times with appropriate argu-
5610 ments, you can mimic Perl's /g option, and it is in this kind of imple-
5611 mentation where \G can be useful.
5613 Note, however, that PCRE's interpretation of \G, as the start of the
5614 current match, is subtly different from Perl's, which defines it as the
5615 end of the previous match. In Perl, these can be different when the
5616 previously matched string was empty. Because PCRE does just one match
5617 at a time, it cannot reproduce this behaviour.
5619 If all the alternatives of a pattern begin with \G, the expression is
5620 anchored to the starting match position, and the "anchored" flag is set
5621 in the compiled regular expression.
5624 CIRCUMFLEX AND DOLLAR
5626 The circumflex and dollar metacharacters are zero-width assertions.
5627 That is, they test for a particular condition being true without con-
5628 suming any characters from the subject string.
5630 Outside a character class, in the default matching mode, the circumflex
5631 character is an assertion that is true only if the current matching
5632 point is at the start of the subject string. If the startoffset argu-
5633 ment of pcre_exec() is non-zero, circumflex can never match if the
5634 PCRE_MULTILINE option is unset. Inside a character class, circumflex
5635 has an entirely different meaning (see below).
5637 Circumflex need not be the first character of the pattern if a number
5638 of alternatives are involved, but it should be the first thing in each
5639 alternative in which it appears if the pattern is ever to match that
5640 branch. If all possible alternatives start with a circumflex, that is,
5641 if the pattern is constrained to match only at the start of the sub-
5642 ject, it is said to be an "anchored" pattern. (There are also other
5643 constructs that can cause a pattern to be anchored.)
5645 The dollar character is an assertion that is true only if the current
5646 matching point is at the end of the subject string, or immediately
5647 before a newline at the end of the string (by default). Note, however,
5648 that it does not actually match the newline. Dollar need not be the
5649 last character of the pattern if a number of alternatives are involved,
5650 but it should be the last item in any branch in which it appears. Dol-
5651 lar has no special meaning in a character class.
5653 The meaning of dollar can be changed so that it matches only at the
5654 very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
5655 compile time. This does not affect the \Z assertion.
5657 The meanings of the circumflex and dollar characters are changed if the
5658 PCRE_MULTILINE option is set. When this is the case, a circumflex
5659 matches immediately after internal newlines as well as at the start of
5660 the subject string. It does not match after a newline that ends the
5661 string. A dollar matches before any newlines in the string, as well as
5662 at the very end, when PCRE_MULTILINE is set. When newline is specified
5663 as the two-character sequence CRLF, isolated CR and LF characters do
5664 not indicate newlines.
5666 For example, the pattern /^abc$/ matches the subject string "def\nabc"
5667 (where \n represents a newline) in multiline mode, but not otherwise.
5668 Consequently, patterns that are anchored in single line mode because
5669 all branches start with ^ are not anchored in multiline mode, and a
5670 match for circumflex is possible when the startoffset argument of
5671 pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
5672 PCRE_MULTILINE is set.
5674 Note that the sequences \A, \Z, and \z can be used to match the start
5675 and end of the subject in both modes, and if all branches of a pattern
5676 start with \A it is always anchored, whether or not PCRE_MULTILINE is
5680 FULL STOP (PERIOD, DOT) AND \N
5682 Outside a character class, a dot in the pattern matches any one charac-
5683 ter in the subject string except (by default) a character that signi-
5684 fies the end of a line.
5686 When a line ending is defined as a single character, dot never matches
5687 that character; when the two-character sequence CRLF is used, dot does
5688 not match CR if it is immediately followed by LF, but otherwise it
5689 matches all characters (including isolated CRs and LFs). When any Uni-
5690 code line endings are being recognized, dot does not match CR or LF or
5691 any of the other line ending characters.
5693 The behaviour of dot with regard to newlines can be changed. If the
5694 PCRE_DOTALL option is set, a dot matches any one character, without
5695 exception. If the two-character sequence CRLF is present in the subject
5696 string, it takes two dots to match it.
5698 The handling of dot is entirely independent of the handling of circum-
5699 flex and dollar, the only relationship being that they both involve
5700 newlines. Dot has no special meaning in a character class.
5702 The escape sequence \N behaves like a dot, except that it is not
5703 affected by the PCRE_DOTALL option. In other words, it matches any
5704 character except one that signifies the end of a line. Perl also uses
5705 \N to match characters by name; PCRE does not support this.
5708 MATCHING A SINGLE DATA UNIT
5710 Outside a character class, the escape sequence \C matches any one data
5711 unit, whether or not a UTF mode is set. In the 8-bit library, one data
5712 unit is one byte; in the 16-bit library it is a 16-bit unit; in the
5713 32-bit library it is a 32-bit unit. Unlike a dot, \C always matches
5714 line-ending characters. The feature is provided in Perl in order to
5715 match individual bytes in UTF-8 mode, but it is unclear how it can use-
5716 fully be used. Because \C breaks up characters into individual data
5717 units, matching one unit with \C in a UTF mode means that the rest of
5718 the string may start with a malformed UTF character. This has undefined
5719 results, because PCRE assumes that it is dealing with valid UTF strings
5720 (and by default it checks this at the start of processing unless the
5721 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or PCRE_NO_UTF32_CHECK option
5724 PCRE does not allow \C to appear in lookbehind assertions (described
5725 below) in a UTF mode, because this would make it impossible to calcu-
5726 late the length of the lookbehind.
5728 In general, the \C escape sequence is best avoided. However, one way of
5729 using it that avoids the problem of malformed UTF characters is to use
5730 a lookahead to check the length of the next character, as in this pat-
5731 tern, which could be used with a UTF-8 string (ignore white space and
5734 (?| (?=[\x00-\x7f])(\C) |
5735 (?=[\x80-\x{7ff}])(\C)(\C) |
5736 (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
5737 (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))
5739 A group that starts with (?| resets the capturing parentheses numbers
5740 in each alternative (see "Duplicate Subpattern Numbers" below). The
5741 assertions at the start of each branch check the next UTF-8 character
5742 for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
5743 character's individual bytes are then captured by the appropriate num-
5747 SQUARE BRACKETS AND CHARACTER CLASSES
5749 An opening square bracket introduces a character class, terminated by a
5750 closing square bracket. A closing square bracket on its own is not spe-
5751 cial by default. However, if the PCRE_JAVASCRIPT_COMPAT option is set,
5752 a lone closing square bracket causes a compile-time error. If a closing
5753 square bracket is required as a member of the class, it should be the
5754 first data character in the class (after an initial circumflex, if
5755 present) or escaped with a backslash.
5757 A character class matches a single character in the subject. In a UTF
5758 mode, the character may be more than one data unit long. A matched
5759 character must be in the set of characters defined by the class, unless
5760 the first character in the class definition is a circumflex, in which
5761 case the subject character must not be in the set defined by the class.
5762 If a circumflex is actually required as a member of the class, ensure
5763 it is not the first character, or escape it with a backslash.
5765 For example, the character class [aeiou] matches any lower case vowel,
5766 while [^aeiou] matches any character that is not a lower case vowel.
5767 Note that a circumflex is just a convenient notation for specifying the
5768 characters that are in the class by enumerating those that are not. A
5769 class that starts with a circumflex is not an assertion; it still con-
5770 sumes a character from the subject string, and therefore it fails if
5771 the current pointer is at the end of the string.
5773 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255
5774 (0xffff) can be included in a class as a literal string of data units,
5775 or by using the \x{ escaping mechanism.
5777 When caseless matching is set, any letters in a class represent both
5778 their upper case and lower case versions, so for example, a caseless
5779 [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
5780 match "A", whereas a caseful version would. In a UTF mode, PCRE always
5781 understands the concept of case for characters whose values are less
5782 than 128, so caseless matching is always possible. For characters with
5783 higher values, the concept of case is supported if PCRE is compiled
5784 with Unicode property support, but not otherwise. If you want to use
5785 caseless matching in a UTF mode for characters 128 and above, you must
5786 ensure that PCRE is compiled with Unicode property support as well as
5789 Characters that might indicate line breaks are never treated in any
5790 special way when matching character classes, whatever line-ending
5791 sequence is in use, and whatever setting of the PCRE_DOTALL and
5792 PCRE_MULTILINE options is used. A class such as [^a] always matches one
5793 of these characters.
5795 The minus (hyphen) character can be used to specify a range of charac-
5796 ters in a character class. For example, [d-m] matches any letter
5797 between d and m, inclusive. If a minus character is required in a
5798 class, it must be escaped with a backslash or appear in a position
5799 where it cannot be interpreted as indicating a range, typically as the
5800 first or last character in the class, or immediately after a range. For
5801 example, [b-d-z] matches letters in the range b to d, a hyphen charac-
5804 It is not possible to have the literal character "]" as the end charac-
5805 ter of a range. A pattern such as [W-]46] is interpreted as a class of
5806 two characters ("W" and "-") followed by a literal string "46]", so it
5807 would match "W46]" or "-46]". However, if the "]" is escaped with a
5808 backslash it is interpreted as the end of range, so [W-\]46] is inter-
5809 preted as a class containing a range followed by two other characters.
5810 The octal or hexadecimal representation of "]" can also be used to end
5813 An error is generated if a POSIX character class (see below) or an
5814 escape sequence other than one that defines a single character appears
5815 at a point where a range ending character is expected. For example,
5816 [z-\xff] is valid, but [A-\d] and [A-[:digit:]] are not.
5818 Ranges operate in the collating sequence of character values. They can
5819 also be used for characters specified numerically, for example
5820 [\000-\037]. Ranges can include any characters that are valid for the
5823 If a range that includes letters is used when caseless matching is set,
5824 it matches the letters in either case. For example, [W-c] is equivalent
5825 to [][\\^_`wxyzabc], matched caselessly, and in a non-UTF mode, if
5826 character tables for a French locale are in use, [\xc8-\xcb] matches
5827 accented E characters in both cases. In UTF modes, PCRE supports the
5828 concept of case for characters with values greater than 128 only when
5829 it is compiled with Unicode property support.
5831 The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, \V,
5832 \w, and \W may appear in a character class, and add the characters that
5833 they match to the class. For example, [\dABCDEF] matches any hexadeci-
5834 mal digit. In UTF modes, the PCRE_UCP option affects the meanings of
5835 \d, \s, \w and their upper case partners, just as it does when they
5836 appear outside a character class, as described in the section entitled
5837 "Generic character types" above. The escape sequence \b has a different
5838 meaning inside a character class; it matches the backspace character.
5839 The sequences \B, \N, \R, and \X are not special inside a character
5840 class. Like any other unrecognized escape sequences, they are treated
5841 as the literal characters "B", "N", "R", and "X" by default, but cause
5842 an error if the PCRE_EXTRA option is set.
5844 A circumflex can conveniently be used with the upper case character
5845 types to specify a more restricted set of characters than the matching
5846 lower case type. For example, the class [^\W_] matches any letter or
5847 digit, but not underscore, whereas [\w] includes underscore. A positive
5848 character class should be read as "something OR something OR ..." and a
5849 negative class as "NOT something AND NOT something AND NOT ...".
5851 The only metacharacters that are recognized in character classes are
5852 backslash, hyphen (only where it can be interpreted as specifying a
5853 range), circumflex (only at the start), opening square bracket (only
5854 when it can be interpreted as introducing a POSIX class name, or for a
5855 special compatibility feature - see the next two sections), and the
5856 terminating closing square bracket. However, escaping other non-
5857 alphanumeric characters does no harm.
5860 POSIX CHARACTER CLASSES
5862 Perl supports the POSIX notation for character classes. This uses names
5863 enclosed by [: and :] within the enclosing square brackets. PCRE also
5864 supports this notation. For example,
5868 matches "0", "1", any alphabetic character, or "%". The supported class
5871 alnum letters and digits
5873 ascii character codes 0 - 127
5874 blank space or tab only
5875 cntrl control characters
5876 digit decimal digits (same as \d)
5877 graph printing characters, excluding space
5878 lower lower case letters
5879 print printing characters, including space
5880 punct printing characters, excluding letters and digits and space
5881 space white space (the same as \s from PCRE 8.34)
5882 upper upper case letters
5883 word "word" characters (same as \w)
5884 xdigit hexadecimal digits
5886 The default "space" characters are HT (9), LF (10), VT (11), FF (12),
5887 CR (13), and space (32). If locale-specific matching is taking place,
5888 the list of space characters may be different; there may be fewer or
5889 more of them. "Space" used to be different to \s, which did not include
5890 VT, for Perl compatibility. However, Perl changed at release 5.18, and
5891 PCRE followed at release 8.34. "Space" and \s now match the same set
5894 The name "word" is a Perl extension, and "blank" is a GNU extension
5895 from Perl 5.8. Another Perl extension is negation, which is indicated
5896 by a ^ character after the colon. For example,
5900 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
5901 POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
5902 these are not supported, and an error is given if they are encountered.
5904 By default, characters with values greater than 128 do not match any of
5905 the POSIX character classes. However, if the PCRE_UCP option is passed
5906 to pcre_compile(), some of the classes are changed so that Unicode
5907 character properties are used. This is achieved by replacing certain
5908 POSIX classes by other sequences, as follows:
5910 [:alnum:] becomes \p{Xan}
5911 [:alpha:] becomes \p{L}
5912 [:blank:] becomes \h
5913 [:digit:] becomes \p{Nd}
5914 [:lower:] becomes \p{Ll}
5915 [:space:] becomes \p{Xps}
5916 [:upper:] becomes \p{Lu}
5917 [:word:] becomes \p{Xwd}
5919 Negated versions, such as [:^alpha:] use \P instead of \p. Three other
5920 POSIX classes are handled specially in UCP mode:
5922 [:graph:] This matches characters that have glyphs that mark the page
5923 when printed. In Unicode property terms, it matches all char-
5924 acters with the L, M, N, P, S, or Cf properties, except for:
5926 U+061C Arabic Letter Mark
5927 U+180E Mongolian Vowel Separator
5928 U+2066 - U+2069 Various "isolate"s
5931 [:print:] This matches the same characters as [:graph:] plus space
5932 characters that are not controls, that is, characters with
5935 [:punct:] This matches all characters that have the Unicode P (punctua-
5936 tion) property, plus those characters whose code points are
5937 less than 128 that have the S (Symbol) property.
5939 The other POSIX classes are unchanged, and match only characters with
5940 code points less than 128.
5943 COMPATIBILITY FEATURE FOR WORD BOUNDARIES
5945 In the POSIX.2 compliant library that was included in 4.4BSD Unix, the
5946 ugly syntax [[:<:]] and [[:>:]] is used for matching "start of word"
5947 and "end of word". PCRE treats these items as follows:
5949 [[:<:]] is converted to \b(?=\w)
5950 [[:>:]] is converted to \b(?<=\w)
5952 Only these exact character sequences are recognized. A sequence such as
5953 [a[:<:]b] provokes error for an unrecognized POSIX class name. This
5954 support is not compatible with Perl. It is provided to help migrations
5955 from other environments, and is best not used in any new patterns. Note
5956 that \b matches at the start and the end of a word (see "Simple asser-
5957 tions" above), and in a Perl-style pattern the preceding or following
5958 character normally shows which is wanted, without the need for the
5959 assertions that are used above in order to give exactly the POSIX be-
5965 Vertical bar characters are used to separate alternative patterns. For
5966 example, the pattern
5970 matches either "gilbert" or "sullivan". Any number of alternatives may
5971 appear, and an empty alternative is permitted (matching the empty
5972 string). The matching process tries each alternative in turn, from left
5973 to right, and the first one that succeeds is used. If the alternatives
5974 are within a subpattern (defined below), "succeeds" means matching the
5975 rest of the main pattern as well as the alternative in the subpattern.
5978 INTERNAL OPTION SETTING
5980 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
5981 PCRE_EXTENDED options (which are Perl-compatible) can be changed from
5982 within the pattern by a sequence of Perl option letters enclosed
5983 between "(?" and ")". The option letters are
5986 m for PCRE_MULTILINE
5990 For example, (?im) sets caseless, multiline matching. It is also possi-
5991 ble to unset these options by preceding the letter with a hyphen, and a
5992 combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
5993 LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED,
5994 is also permitted. If a letter appears both before and after the
5995 hyphen, the option is unset.
5997 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA
5998 can be changed in the same way as the Perl-compatible options by using
5999 the characters J, U and X respectively.
6001 When one of these option changes occurs at top level (that is, not
6002 inside subpattern parentheses), the change applies to the remainder of
6003 the pattern that follows. If the change is placed right at the start of
6004 a pattern, PCRE extracts it into the global options (and it will there-
6005 fore show up in data extracted by the pcre_fullinfo() function).
6007 An option change within a subpattern (see below for a description of
6008 subpatterns) affects only that part of the subpattern that follows it,
6013 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
6014 used). By this means, options can be made to have different settings
6015 in different parts of the pattern. Any changes made in one alternative
6016 do carry on into subsequent branches within the same subpattern. For
6021 matches "ab", "aB", "c", and "C", even though when matching "C" the
6022 first branch is abandoned before the option setting. This is because
6023 the effects of option settings happen at compile time. There would be
6024 some very weird behaviour otherwise.
6026 Note: There are other PCRE-specific options that can be set by the
6027 application when the compiling or matching functions are called. In
6028 some cases the pattern can contain special leading sequences such as
6029 (*CRLF) to override what the application has set or what has been
6030 defaulted. Details are given in the section entitled "Newline
6031 sequences" above. There are also the (*UTF8), (*UTF16),(*UTF32), and
6032 (*UCP) leading sequences that can be used to set UTF and Unicode prop-
6033 erty modes; they are equivalent to setting the PCRE_UTF8, PCRE_UTF16,
6034 PCRE_UTF32 and the PCRE_UCP options, respectively. The (*UTF) sequence
6035 is a generic version that can be used with any of the libraries. How-
6036 ever, the application can set the PCRE_NEVER_UTF option, which locks
6037 out the use of the (*UTF) sequences.
6042 Subpatterns are delimited by parentheses (round brackets), which can be
6043 nested. Turning part of a pattern into a subpattern does two things:
6045 1. It localizes a set of alternatives. For example, the pattern
6047 cat(aract|erpillar|)
6049 matches "cataract", "caterpillar", or "cat". Without the parentheses,
6050 it would match "cataract", "erpillar" or an empty string.
6052 2. It sets up the subpattern as a capturing subpattern. This means
6053 that, when the whole pattern matches, that portion of the subject
6054 string that matched the subpattern is passed back to the caller via the
6055 ovector argument of the matching function. (This applies only to the
6056 traditional matching functions; the DFA matching functions do not sup-
6059 Opening parentheses are counted from left to right (starting from 1) to
6060 obtain numbers for the capturing subpatterns. For example, if the
6061 string "the red king" is matched against the pattern
6063 the ((red|white) (king|queen))
6065 the captured substrings are "red king", "red", and "king", and are num-
6066 bered 1, 2, and 3, respectively.
6068 The fact that plain parentheses fulfil two functions is not always
6069 helpful. There are often times when a grouping subpattern is required
6070 without a capturing requirement. If an opening parenthesis is followed
6071 by a question mark and a colon, the subpattern does not do any captur-
6072 ing, and is not counted when computing the number of any subsequent
6073 capturing subpatterns. For example, if the string "the white queen" is
6074 matched against the pattern
6076 the ((?:red|white) (king|queen))
6078 the captured substrings are "white queen" and "queen", and are numbered
6079 1 and 2. The maximum number of capturing subpatterns is 65535.
6081 As a convenient shorthand, if any option settings are required at the
6082 start of a non-capturing subpattern, the option letters may appear
6083 between the "?" and the ":". Thus the two patterns
6085 (?i:saturday|sunday)
6086 (?:(?i)saturday|sunday)
6088 match exactly the same set of strings. Because alternative branches are
6089 tried from left to right, and options are not reset until the end of
6090 the subpattern is reached, an option setting in one branch does affect
6091 subsequent branches, so the above patterns match "SUNDAY" as well as
6095 DUPLICATE SUBPATTERN NUMBERS
6097 Perl 5.10 introduced a feature whereby each alternative in a subpattern
6098 uses the same numbers for its capturing parentheses. Such a subpattern
6099 starts with (?| and is itself a non-capturing subpattern. For example,
6100 consider this pattern:
6102 (?|(Sat)ur|(Sun))day
6104 Because the two alternatives are inside a (?| group, both sets of cap-
6105 turing parentheses are numbered one. Thus, when the pattern matches,
6106 you can look at captured substring number one, whichever alternative
6107 matched. This construct is useful when you want to capture part, but
6108 not all, of one of a number of alternatives. Inside a (?| group, paren-
6109 theses are numbered as usual, but the number is reset at the start of
6110 each branch. The numbers of any capturing parentheses that follow the
6111 subpattern start after the highest number used in any branch. The fol-
6112 lowing example is taken from the Perl documentation. The numbers under-
6113 neath show in which buffer the captured content will be stored.
6115 # before ---------------branch-reset----------- after
6116 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
6119 A back reference to a numbered subpattern uses the most recent value
6120 that is set for that number by any subpattern. The following pattern
6121 matches "abcabc" or "defdef":
6125 In contrast, a subroutine call to a numbered subpattern always refers
6126 to the first one in the pattern with the given number. The following
6127 pattern matches "abcabc" or "defabc":
6129 /(?|(abc)|(def))(?1)/
6131 If a condition test for a subpattern's having matched refers to a non-
6132 unique number, the test is true if any of the subpatterns of that num-
6135 An alternative approach to using this "branch reset" feature is to use
6136 duplicate named subpatterns, as described in the next section.
6141 Identifying capturing parentheses by number is simple, but it can be
6142 very hard to keep track of the numbers in complicated regular expres-
6143 sions. Furthermore, if an expression is modified, the numbers may
6144 change. To help with this difficulty, PCRE supports the naming of sub-
6145 patterns. This feature was not added to Perl until release 5.10. Python
6146 had the feature earlier, and PCRE introduced it at release 4.0, using
6147 the Python syntax. PCRE now supports both the Perl and the Python syn-
6148 tax. Perl allows identically numbered subpatterns to have different
6149 names, but PCRE does not.
6151 In PCRE, a subpattern can be named in one of three ways: (?<name>...)
6152 or (?'name'...) as in Perl, or (?P<name>...) as in Python. References
6153 to capturing parentheses from other parts of the pattern, such as back
6154 references, recursion, and conditions, can be made by name as well as
6157 Names consist of up to 32 alphanumeric characters and underscores, but
6158 must start with a non-digit. Named capturing parentheses are still
6159 allocated numbers as well as names, exactly as if the names were not
6160 present. The PCRE API provides function calls for extracting the name-
6161 to-number translation table from a compiled pattern. There is also a
6162 convenience function for extracting a captured substring by name.
6164 By default, a name must be unique within a pattern, but it is possible
6165 to relax this constraint by setting the PCRE_DUPNAMES option at compile
6166 time. (Duplicate names are also always permitted for subpatterns with
6167 the same number, set up as described in the previous section.) Dupli-
6168 cate names can be useful for patterns where only one instance of the
6169 named parentheses can match. Suppose you want to match the name of a
6170 weekday, either as a 3-letter abbreviation or as the full name, and in
6171 both cases you want to extract the abbreviation. This pattern (ignoring
6172 the line breaks) does the job:
6174 (?<DN>Mon|Fri|Sun)(?:day)?|
6175 (?<DN>Tue)(?:sday)?|
6176 (?<DN>Wed)(?:nesday)?|
6177 (?<DN>Thu)(?:rsday)?|
6178 (?<DN>Sat)(?:urday)?
6180 There are five capturing substrings, but only one is ever set after a
6181 match. (An alternative way of solving this problem is to use a "branch
6182 reset" subpattern, as described in the previous section.)
6184 The convenience function for extracting the data by name returns the
6185 substring for the first (and in this example, the only) subpattern of
6186 that name that matched. This saves searching to find which numbered
6189 If you make a back reference to a non-unique named subpattern from
6190 elsewhere in the pattern, the subpatterns to which the name refers are
6191 checked in the order in which they appear in the overall pattern. The
6192 first one that is set is used for the reference. For example, this pat-
6193 tern matches both "foofoo" and "barbar" but not "foobar" or "barfoo":
6195 (?:(?<n>foo)|(?<n>bar))\k<n>
6198 If you make a subroutine call to a non-unique named subpattern, the one
6199 that corresponds to the first occurrence of the name is used. In the
6200 absence of duplicate numbers (see the previous section) this is the one
6201 with the lowest number.
6203 If you use a named reference in a condition test (see the section about
6204 conditions below), either to check whether a subpattern has matched, or
6205 to check for recursion, all subpatterns with the same name are tested.
6206 If the condition is true for any one of them, the overall condition is
6207 true. This is the same behaviour as testing by number. For further
6208 details of the interfaces for handling named subpatterns, see the
6209 pcreapi documentation.
6211 Warning: You cannot use different names to distinguish between two sub-
6212 patterns with the same number because PCRE uses only the numbers when
6213 matching. For this reason, an error is given at compile time if differ-
6214 ent names are given to subpatterns with the same number. However, you
6215 can always give the same name to subpatterns with the same number, even
6216 when PCRE_DUPNAMES is not set.
6221 Repetition is specified by quantifiers, which can follow any of the
6224 a literal data character
6225 the dot metacharacter
6226 the \C escape sequence
6227 the \X escape sequence
6228 the \R escape sequence
6229 an escape such as \d or \pL that matches a single character
6231 a back reference (see next section)
6232 a parenthesized subpattern (including assertions)
6233 a subroutine call to a subpattern (recursive or otherwise)
6235 The general repetition quantifier specifies a minimum and maximum num-
6236 ber of permitted matches, by giving the two numbers in curly brackets
6237 (braces), separated by a comma. The numbers must be less than 65536,
6238 and the first must be less than or equal to the second. For example:
6242 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
6243 special character. If the second number is omitted, but the comma is
6244 present, there is no upper limit; if the second number and the comma
6245 are both omitted, the quantifier specifies an exact number of required
6250 matches at least 3 successive vowels, but may match many more, while
6254 matches exactly 8 digits. An opening curly bracket that appears in a
6255 position where a quantifier is not allowed, or one that does not match
6256 the syntax of a quantifier, is taken as a literal character. For exam-
6257 ple, {,6} is not a quantifier, but a literal string of four characters.
6259 In UTF modes, quantifiers apply to characters rather than to individual
6260 data units. Thus, for example, \x{100}{2} matches two characters, each
6261 of which is represented by a two-byte sequence in a UTF-8 string. Simi-
6262 larly, \X{3} matches three Unicode extended grapheme clusters, each of
6263 which may be several data units long (and they may be of different
6266 The quantifier {0} is permitted, causing the expression to behave as if
6267 the previous item and the quantifier were not present. This may be use-
6268 ful for subpatterns that are referenced as subroutines from elsewhere
6269 in the pattern (but see also the section entitled "Defining subpatterns
6270 for use by reference only" below). Items other than subpatterns that
6271 have a {0} quantifier are omitted from the compiled pattern.
6273 For convenience, the three most common quantifiers have single-charac-
6276 * is equivalent to {0,}
6277 + is equivalent to {1,}
6278 ? is equivalent to {0,1}
6280 It is possible to construct infinite loops by following a subpattern
6281 that can match no characters with a quantifier that has no upper limit,
6286 Earlier versions of Perl and PCRE used to give an error at compile time
6287 for such patterns. However, because there are cases where this can be
6288 useful, such patterns are now accepted, but if any repetition of the
6289 subpattern does in fact match no characters, the loop is forcibly bro-
6292 By default, the quantifiers are "greedy", that is, they match as much
6293 as possible (up to the maximum number of permitted times), without
6294 causing the rest of the pattern to fail. The classic example of where
6295 this gives problems is in trying to match comments in C programs. These
6296 appear between /* and */ and within the comment, individual * and /
6297 characters may appear. An attempt to match C comments by applying the
6304 /* first comment */ not comment /* second comment */
6306 fails, because it matches the entire string owing to the greediness of
6309 However, if a quantifier is followed by a question mark, it ceases to
6310 be greedy, and instead matches the minimum number of times possible, so
6315 does the right thing with the C comments. The meaning of the various
6316 quantifiers is not otherwise changed, just the preferred number of
6317 matches. Do not confuse this use of question mark with its use as a
6318 quantifier in its own right. Because it has two uses, it can sometimes
6319 appear doubled, as in
6323 which matches one digit by preference, but can match two if that is the
6324 only way the rest of the pattern matches.
6326 If the PCRE_UNGREEDY option is set (an option that is not available in
6327 Perl), the quantifiers are not greedy by default, but individual ones
6328 can be made greedy by following them with a question mark. In other
6329 words, it inverts the default behaviour.
6331 When a parenthesized subpattern is quantified with a minimum repeat
6332 count that is greater than 1 or with a limited maximum, more memory is
6333 required for the compiled pattern, in proportion to the size of the
6336 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
6337 alent to Perl's /s) is set, thus allowing the dot to match newlines,
6338 the pattern is implicitly anchored, because whatever follows will be
6339 tried against every character position in the subject string, so there
6340 is no point in retrying the overall match at any position after the
6341 first. PCRE normally treats such a pattern as though it were preceded
6344 In cases where it is known that the subject string contains no new-
6345 lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
6346 mization, or alternatively using ^ to indicate anchoring explicitly.
6348 However, there are some cases where the optimization cannot be used.
6349 When .* is inside capturing parentheses that are the subject of a back
6350 reference elsewhere in the pattern, a match at the start may fail where
6351 a later one succeeds. Consider, for example:
6355 If the subject is "xyz123abc123" the match point is the fourth charac-
6356 ter. For this reason, such a pattern is not implicitly anchored.
6358 Another case where implicit anchoring is not applied is when the lead-
6359 ing .* is inside an atomic group. Once again, a match at the start may
6360 fail where a later one succeeds. Consider this pattern:
6364 It matches "ab" in the subject "aab". The use of the backtracking con-
6365 trol verbs (*PRUNE) and (*SKIP) also disable this optimization.
6367 When a capturing subpattern is repeated, the value captured is the sub-
6368 string that matched the final iteration. For example, after
6370 (tweedle[dume]{3}\s*)+
6372 has matched "tweedledum tweedledee" the value of the captured substring
6373 is "tweedledee". However, if there are nested capturing subpatterns,
6374 the corresponding captured values may have been set in previous itera-
6375 tions. For example, after
6379 matches "aba" the value of the second captured substring is "b".
6382 ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS
6384 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
6385 repetition, failure of what follows normally causes the repeated item
6386 to be re-evaluated to see if a different number of repeats allows the
6387 rest of the pattern to match. Sometimes it is useful to prevent this,
6388 either to change the nature of the match, or to cause it fail earlier
6389 than it otherwise might, when the author of the pattern knows there is
6390 no point in carrying on.
6392 Consider, for example, the pattern \d+foo when applied to the subject
6397 After matching all 6 digits and then failing to match "foo", the normal
6398 action of the matcher is to try again with only 5 digits matching the
6399 \d+ item, and then with 4, and so on, before ultimately failing.
6400 "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
6401 the means for specifying that once a subpattern has matched, it is not
6402 to be re-evaluated in this way.
6404 If we use atomic grouping for the previous example, the matcher gives
6405 up immediately on failing to match "foo" the first time. The notation
6406 is a kind of special parenthesis, starting with (?> as in this example:
6410 This kind of parenthesis "locks up" the part of the pattern it con-
6411 tains once it has matched, and a failure further into the pattern is
6412 prevented from backtracking into it. Backtracking past it to previous
6413 items, however, works as normal.
6415 An alternative description is that a subpattern of this type matches
6416 the string of characters that an identical standalone pattern would
6417 match, if anchored at the current point in the subject string.
6419 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
6420 such as the above example can be thought of as a maximizing repeat that
6421 must swallow everything it can. So, while both \d+ and \d+? are pre-
6422 pared to adjust the number of digits they match in order to make the
6423 rest of the pattern match, (?>\d+) can only match an entire sequence of
6426 Atomic groups in general can of course contain arbitrarily complicated
6427 subpatterns, and can be nested. However, when the subpattern for an
6428 atomic group is just a single repeated item, as in the example above, a
6429 simpler notation, called a "possessive quantifier" can be used. This
6430 consists of an additional + character following a quantifier. Using
6431 this notation, the previous example can be rewritten as
6435 Note that a possessive quantifier can be used with an entire group, for
6440 Possessive quantifiers are always greedy; the setting of the
6441 PCRE_UNGREEDY option is ignored. They are a convenient notation for the
6442 simpler forms of atomic group. However, there is no difference in the
6443 meaning of a possessive quantifier and the equivalent atomic group,
6444 though there may be a performance difference; possessive quantifiers
6445 should be slightly faster.
6447 The possessive quantifier syntax is an extension to the Perl 5.8 syn-
6448 tax. Jeffrey Friedl originated the idea (and the name) in the first
6449 edition of his book. Mike McCloskey liked it, so implemented it when he
6450 built Sun's Java package, and PCRE copied it from there. It ultimately
6451 found its way into Perl at release 5.10.
6453 PCRE has an optimization that automatically "possessifies" certain sim-
6454 ple pattern constructs. For example, the sequence A+B is treated as
6455 A++B because there is no point in backtracking into a sequence of A's
6458 When a pattern contains an unlimited repeat inside a subpattern that
6459 can itself be repeated an unlimited number of times, the use of an
6460 atomic group is the only way to avoid some failing matches taking a
6461 very long time indeed. The pattern
6465 matches an unlimited number of substrings that either consist of non-
6466 digits, or digits enclosed in <>, followed by either ! or ?. When it
6467 matches, it runs quickly. However, if it is applied to
6469 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
6471 it takes a long time before reporting failure. This is because the
6472 string can be divided between the internal \D+ repeat and the external
6473 * repeat in a large number of ways, and all have to be tried. (The
6474 example uses [!?] rather than a single character at the end, because
6475 both PCRE and Perl have an optimization that allows for fast failure
6476 when a single character is used. They remember the last single charac-
6477 ter that is required for a match, and fail early if it is not present
6478 in the string.) If the pattern is changed so that it uses an atomic
6481 ((?>\D+)|<\d+>)*[!?]
6483 sequences of non-digits cannot be broken, and failure happens quickly.
6488 Outside a character class, a backslash followed by a digit greater than
6489 0 (and possibly further digits) is a back reference to a capturing sub-
6490 pattern earlier (that is, to its left) in the pattern, provided there
6491 have been that many previous capturing left parentheses.
6493 However, if the decimal number following the backslash is less than 10,
6494 it is always taken as a back reference, and causes an error only if
6495 there are not that many capturing left parentheses in the entire pat-
6496 tern. In other words, the parentheses that are referenced need not be
6497 to the left of the reference for numbers less than 10. A "forward back
6498 reference" of this type can make sense when a repetition is involved
6499 and the subpattern to the right has participated in an earlier itera-
6502 It is not possible to have a numerical "forward back reference" to a
6503 subpattern whose number is 10 or more using this syntax because a
6504 sequence such as \50 is interpreted as a character defined in octal.
6505 See the subsection entitled "Non-printing characters" above for further
6506 details of the handling of digits following a backslash. There is no
6507 such problem when named parentheses are used. A back reference to any
6508 subpattern is possible using named parentheses (see below).
6510 Another way of avoiding the ambiguity inherent in the use of digits
6511 following a backslash is to use the \g escape sequence. This escape
6512 must be followed by an unsigned number or a negative number, optionally
6513 enclosed in braces. These examples are all identical:
6519 An unsigned number specifies an absolute reference without the ambigu-
6520 ity that is present in the older syntax. It is also useful when literal
6521 digits follow the reference. A negative number is a relative reference.
6522 Consider this example:
6526 The sequence \g{-1} is a reference to the most recently started captur-
6527 ing subpattern before \g, that is, is it equivalent to \2 in this exam-
6528 ple. Similarly, \g{-2} would be equivalent to \1. The use of relative
6529 references can be helpful in long patterns, and also in patterns that
6530 are created by joining together fragments that contain references
6533 A back reference matches whatever actually matched the capturing sub-
6534 pattern in the current subject string, rather than anything matching
6535 the subpattern itself (see "Subpatterns as subroutines" below for a way
6536 of doing that). So the pattern
6538 (sens|respons)e and \1ibility
6540 matches "sense and sensibility" and "response and responsibility", but
6541 not "sense and responsibility". If caseful matching is in force at the
6542 time of the back reference, the case of letters is relevant. For exam-
6547 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
6548 original capturing subpattern is matched caselessly.
6550 There are several different ways of writing back references to named
6551 subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or
6552 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's
6553 unified back reference syntax, in which \g can be used for both numeric
6554 and named references, is also supported. We could rewrite the above
6555 example in any of the following ways:
6557 (?<p1>(?i)rah)\s+\k<p1>
6558 (?'p1'(?i)rah)\s+\k{p1}
6559 (?P<p1>(?i)rah)\s+(?P=p1)
6560 (?<p1>(?i)rah)\s+\g{p1}
6562 A subpattern that is referenced by name may appear in the pattern
6563 before or after the reference.
6565 There may be more than one back reference to the same subpattern. If a
6566 subpattern has not actually been used in a particular match, any back
6567 references to it always fail by default. For example, the pattern
6571 always fails if it starts to match "a" rather than "bc". However, if
6572 the PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back refer-
6573 ence to an unset value matches an empty string.
6575 Because there may be many capturing parentheses in a pattern, all dig-
6576 its following a backslash are taken as part of a potential back refer-
6577 ence number. If the pattern continues with a digit character, some
6578 delimiter must be used to terminate the back reference. If the
6579 PCRE_EXTENDED option is set, this can be white space. Otherwise, the
6580 \g{ syntax or an empty comment (see "Comments" below) can be used.
6582 Recursive back references
6584 A back reference that occurs inside the parentheses to which it refers
6585 fails when the subpattern is first used, so, for example, (a\1) never
6586 matches. However, such references can be useful inside repeated sub-
6587 patterns. For example, the pattern
6591 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
6592 ation of the subpattern, the back reference matches the character
6593 string corresponding to the previous iteration. In order for this to
6594 work, the pattern must be such that the first iteration does not need
6595 to match the back reference. This can be done using alternation, as in
6596 the example above, or by a quantifier with a minimum of zero.
6598 Back references of this type cause the group that they reference to be
6599 treated as an atomic group. Once the whole group has been matched, a
6600 subsequent matching failure cannot cause backtracking into the middle
6606 An assertion is a test on the characters following or preceding the
6607 current matching point that does not actually consume any characters.
6608 The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
6611 More complicated assertions are coded as subpatterns. There are two
6612 kinds: those that look ahead of the current position in the subject
6613 string, and those that look behind it. An assertion subpattern is
6614 matched in the normal way, except that it does not cause the current
6615 matching position to be changed.
6617 Assertion subpatterns are not capturing subpatterns. If such an asser-
6618 tion contains capturing subpatterns within it, these are counted for
6619 the purposes of numbering the capturing subpatterns in the whole pat-
6620 tern. However, substring capturing is carried out only for positive
6621 assertions. (Perl sometimes, but not always, does do capturing in nega-
6624 For compatibility with Perl, assertion subpatterns may be repeated;
6625 though it makes no sense to assert the same thing several times, the
6626 side effect of capturing parentheses may occasionally be useful. In
6627 practice, there only three cases:
6629 (1) If the quantifier is {0}, the assertion is never obeyed during
6630 matching. However, it may contain internal capturing parenthesized
6631 groups that are called from elsewhere via the subroutine mechanism.
6633 (2) If quantifier is {0,n} where n is greater than zero, it is treated
6634 as if it were {0,1}. At run time, the rest of the pattern match is
6635 tried with and without the assertion, the order depending on the greed-
6636 iness of the quantifier.
6638 (3) If the minimum repetition is greater than zero, the quantifier is
6639 ignored. The assertion is obeyed just once when encountered during
6642 Lookahead assertions
6644 Lookahead assertions start with (?= for positive assertions and (?! for
6645 negative assertions. For example,
6649 matches a word followed by a semicolon, but does not include the semi-
6650 colon in the match, and
6654 matches any occurrence of "foo" that is not followed by "bar". Note
6655 that the apparently similar pattern
6659 does not find an occurrence of "bar" that is preceded by something
6660 other than "foo"; it finds any occurrence of "bar" whatsoever, because
6661 the assertion (?!foo) is always true when the next three characters are
6662 "bar". A lookbehind assertion is needed to achieve the other effect.
6664 If you want to force a matching failure at some point in a pattern, the
6665 most convenient way to do it is with (?!) because an empty string
6666 always matches, so an assertion that requires there not to be an empty
6667 string must always fail. The backtracking control verb (*FAIL) or (*F)
6668 is a synonym for (?!).
6670 Lookbehind assertions
6672 Lookbehind assertions start with (?<= for positive assertions and (?<!
6673 for negative assertions. For example,
6677 does find an occurrence of "bar" that is not preceded by "foo". The
6678 contents of a lookbehind assertion are restricted such that all the
6679 strings it matches must have a fixed length. However, if there are sev-
6680 eral top-level alternatives, they do not all have to have the same
6689 causes an error at compile time. Branches that match different length
6690 strings are permitted only at the top level of a lookbehind assertion.
6691 This is an extension compared with Perl, which requires all branches to
6692 match the same length of string. An assertion such as
6696 is not permitted, because its single top-level branch can match two
6697 different lengths, but it is acceptable to PCRE if rewritten to use two
6702 In some cases, the escape sequence \K (see above) can be used instead
6703 of a lookbehind assertion to get round the fixed-length restriction.
6705 The implementation of lookbehind assertions is, for each alternative,
6706 to temporarily move the current position back by the fixed length and
6707 then try to match. If there are insufficient characters before the cur-
6708 rent position, the assertion fails.
6710 In a UTF mode, PCRE does not allow the \C escape (which matches a sin-
6711 gle data unit even in a UTF mode) to appear in lookbehind assertions,
6712 because it makes it impossible to calculate the length of the lookbe-
6713 hind. The \X and \R escapes, which can match different numbers of data
6714 units, are also not permitted.
6716 "Subroutine" calls (see below) such as (?2) or (?&X) are permitted in
6717 lookbehinds, as long as the subpattern matches a fixed-length string.
6718 Recursion, however, is not supported.
6720 Possessive quantifiers can be used in conjunction with lookbehind
6721 assertions to specify efficient matching of fixed-length strings at the
6722 end of subject strings. Consider a simple pattern such as
6726 when applied to a long string that does not match. Because matching
6727 proceeds from left to right, PCRE will look for each "a" in the subject
6728 and then see if what follows matches the rest of the pattern. If the
6729 pattern is specified as
6733 the initial .* matches the entire string at first, but when this fails
6734 (because there is no following "a"), it backtracks to match all but the
6735 last character, then all but the last two characters, and so on. Once
6736 again the search for "a" covers the entire string, from right to left,
6737 so we are no better off. However, if the pattern is written as
6741 there can be no backtracking for the .*+ item; it can match only the
6742 entire string. The subsequent lookbehind assertion does a single test
6743 on the last four characters. If it fails, the match fails immediately.
6744 For long strings, this approach makes a significant difference to the
6747 Using multiple assertions
6749 Several assertions (of any sort) may occur in succession. For example,
6751 (?<=\d{3})(?<!999)foo
6753 matches "foo" preceded by three digits that are not "999". Notice that
6754 each of the assertions is applied independently at the same point in
6755 the subject string. First there is a check that the previous three
6756 characters are all digits, and then there is a check that the same
6757 three characters are not "999". This pattern does not match "foo" pre-
6758 ceded by six characters, the first of which are digits and the last
6759 three of which are not "999". For example, it doesn't match "123abc-
6760 foo". A pattern to do that is
6762 (?<=\d{3}...)(?<!999)foo
6764 This time the first assertion looks at the preceding six characters,
6765 checking that the first three are digits, and then the second assertion
6766 checks that the preceding three characters are not "999".
6768 Assertions can be nested in any combination. For example,
6772 matches an occurrence of "baz" that is preceded by "bar" which in turn
6773 is not preceded by "foo", while
6775 (?<=\d{3}(?!999)...)foo
6777 is another pattern that matches "foo" preceded by three digits and any
6778 three characters that are not "999".
6781 CONDITIONAL SUBPATTERNS
6783 It is possible to cause the matching process to obey a subpattern con-
6784 ditionally or to choose between two alternative subpatterns, depending
6785 on the result of an assertion, or whether a specific capturing subpat-
6786 tern has already been matched. The two possible forms of conditional
6789 (?(condition)yes-pattern)
6790 (?(condition)yes-pattern|no-pattern)
6792 If the condition is satisfied, the yes-pattern is used; otherwise the
6793 no-pattern (if present) is used. If there are more than two alterna-
6794 tives in the subpattern, a compile-time error occurs. Each of the two
6795 alternatives may itself contain nested subpatterns of any form, includ-
6796 ing conditional subpatterns; the restriction to two alternatives
6797 applies only at the level of the condition. This pattern fragment is an
6798 example where the alternatives are complex:
6800 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
6803 There are four kinds of condition: references to subpatterns, refer-
6804 ences to recursion, a pseudo-condition called DEFINE, and assertions.
6806 Checking for a used subpattern by number
6808 If the text between the parentheses consists of a sequence of digits,
6809 the condition is true if a capturing subpattern of that number has pre-
6810 viously matched. If there is more than one capturing subpattern with
6811 the same number (see the earlier section about duplicate subpattern
6812 numbers), the condition is true if any of them have matched. An alter-
6813 native notation is to precede the digits with a plus or minus sign. In
6814 this case, the subpattern number is relative rather than absolute. The
6815 most recently opened parentheses can be referenced by (?(-1), the next
6816 most recent by (?(-2), and so on. Inside loops it can also make sense
6817 to refer to subsequent groups. The next parentheses to be opened can be
6818 referenced as (?(+1), and so on. (The value zero in any of these forms
6819 is not used; it provokes a compile-time error.)
6821 Consider the following pattern, which contains non-significant white
6822 space to make it more readable (assume the PCRE_EXTENDED option) and to
6823 divide it into three parts for ease of discussion:
6825 ( \( )? [^()]+ (?(1) \) )
6827 The first part matches an optional opening parenthesis, and if that
6828 character is present, sets it as the first captured substring. The sec-
6829 ond part matches one or more characters that are not parentheses. The
6830 third part is a conditional subpattern that tests whether or not the
6831 first set of parentheses matched. If they did, that is, if subject
6832 started with an opening parenthesis, the condition is true, and so the
6833 yes-pattern is executed and a closing parenthesis is required. Other-
6834 wise, since no-pattern is not present, the subpattern matches nothing.
6835 In other words, this pattern matches a sequence of non-parentheses,
6836 optionally enclosed in parentheses.
6838 If you were embedding this pattern in a larger one, you could use a
6841 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
6843 This makes the fragment independent of the parentheses in the larger
6846 Checking for a used subpattern by name
6848 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a
6849 used subpattern by name. For compatibility with earlier versions of
6850 PCRE, which had this facility before Perl, the syntax (?(name)...) is
6853 Rewriting the above example to use a named subpattern gives this:
6855 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
6857 If the name used in a condition of this kind is a duplicate, the test
6858 is applied to all subpatterns of the same name, and is true if any one
6859 of them has matched.
6861 Checking for pattern recursion
6863 If the condition is the string (R), and there is no subpattern with the
6864 name R, the condition is true if a recursive call to the whole pattern
6865 or any subpattern has been made. If digits or a name preceded by amper-
6866 sand follow the letter R, for example:
6868 (?(R3)...) or (?(R&name)...)
6870 the condition is true if the most recent recursion is into a subpattern
6871 whose number or name is given. This condition does not check the entire
6872 recursion stack. If the name used in a condition of this kind is a
6873 duplicate, the test is applied to all subpatterns of the same name, and
6874 is true if any one of them is the most recent recursion.
6876 At "top level", all these recursion test conditions are false. The
6877 syntax for recursive patterns is described below.
6879 Defining subpatterns for use by reference only
6881 If the condition is the string (DEFINE), and there is no subpattern
6882 with the name DEFINE, the condition is always false. In this case,
6883 there may be only one alternative in the subpattern. It is always
6884 skipped if control reaches this point in the pattern; the idea of
6885 DEFINE is that it can be used to define subroutines that can be refer-
6886 enced from elsewhere. (The use of subroutines is described below.) For
6887 example, a pattern to match an IPv4 address such as "192.168.23.245"
6888 could be written like this (ignore white space and line breaks):
6890 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
6891 \b (?&byte) (\.(?&byte)){3} \b
6893 The first part of the pattern is a DEFINE group inside which a another
6894 group named "byte" is defined. This matches an individual component of
6895 an IPv4 address (a number less than 256). When matching takes place,
6896 this part of the pattern is skipped because DEFINE acts like a false
6897 condition. The rest of the pattern uses references to the named group
6898 to match the four dot-separated components of an IPv4 address, insist-
6899 ing on a word boundary at each end.
6901 Assertion conditions
6903 If the condition is not in any of the above formats, it must be an
6904 assertion. This may be a positive or negative lookahead or lookbehind
6905 assertion. Consider this pattern, again containing non-significant
6906 white space, and with the two alternatives on the second line:
6909 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
6911 The condition is a positive lookahead assertion that matches an
6912 optional sequence of non-letters followed by a letter. In other words,
6913 it tests for the presence of at least one letter in the subject. If a
6914 letter is found, the subject is matched against the first alternative;
6915 otherwise it is matched against the second. This pattern matches
6916 strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are
6917 letters and dd are digits.
6922 There are two ways of including comments in patterns that are processed
6923 by PCRE. In both cases, the start of the comment must not be in a char-
6924 acter class, nor in the middle of any other sequence of related charac-
6925 ters such as (?: or a subpattern name or number. The characters that
6926 make up a comment play no part in the pattern matching.
6928 The sequence (?# marks the start of a comment that continues up to the
6929 next closing parenthesis. Nested parentheses are not permitted. If the
6930 PCRE_EXTENDED option is set, an unescaped # character also introduces a
6931 comment, which in this case continues to immediately after the next
6932 newline character or character sequence in the pattern. Which charac-
6933 ters are interpreted as newlines is controlled by the options passed to
6934 a compiling function or by a special sequence at the start of the pat-
6935 tern, as described in the section entitled "Newline conventions" above.
6936 Note that the end of this type of comment is a literal newline sequence
6937 in the pattern; escape sequences that happen to represent a newline do
6938 not count. For example, consider this pattern when PCRE_EXTENDED is
6939 set, and the default newline convention is in force:
6941 abc #comment \n still comment
6943 On encountering the # character, pcre_compile() skips along, looking
6944 for a newline in the pattern. The sequence \n is still literal at this
6945 stage, so it does not terminate the comment. Only an actual character
6946 with the code value 0x0a (the default newline) does so.
6951 Consider the problem of matching a string in parentheses, allowing for
6952 unlimited nested parentheses. Without the use of recursion, the best
6953 that can be done is to use a pattern that matches up to some fixed
6954 depth of nesting. It is not possible to handle an arbitrary nesting
6957 For some time, Perl has provided a facility that allows regular expres-
6958 sions to recurse (amongst other things). It does this by interpolating
6959 Perl code in the expression at run time, and the code can refer to the
6960 expression itself. A Perl pattern using code interpolation to solve the
6961 parentheses problem can be created like this:
6963 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
6965 The (?p{...}) item interpolates Perl code at run time, and in this case
6966 refers recursively to the pattern in which it appears.
6968 Obviously, PCRE cannot support the interpolation of Perl code. Instead,
6969 it supports special syntax for recursion of the entire pattern, and
6970 also for individual subpattern recursion. After its introduction in
6971 PCRE and Python, this kind of recursion was subsequently introduced
6972 into Perl at release 5.10.
6974 A special item that consists of (? followed by a number greater than
6975 zero and a closing parenthesis is a recursive subroutine call of the
6976 subpattern of the given number, provided that it occurs inside that
6977 subpattern. (If not, it is a non-recursive subroutine call, which is
6978 described in the next section.) The special item (?R) or (?0) is a
6979 recursive call of the entire regular expression.
6981 This PCRE pattern solves the nested parentheses problem (assume the
6982 PCRE_EXTENDED option is set so that white space is ignored):
6984 \( ( [^()]++ | (?R) )* \)
6986 First it matches an opening parenthesis. Then it matches any number of
6987 substrings which can either be a sequence of non-parentheses, or a
6988 recursive match of the pattern itself (that is, a correctly parenthe-
6989 sized substring). Finally there is a closing parenthesis. Note the use
6990 of a possessive quantifier to avoid backtracking into sequences of non-
6993 If this were part of a larger pattern, you would not want to recurse
6994 the entire pattern, so instead you could use this:
6996 ( \( ( [^()]++ | (?1) )* \) )
6998 We have put the pattern into parentheses, and caused the recursion to
6999 refer to them instead of the whole pattern.
7001 In a larger pattern, keeping track of parenthesis numbers can be
7002 tricky. This is made easier by the use of relative references. Instead
7003 of (?1) in the pattern above you can write (?-2) to refer to the second
7004 most recently opened parentheses preceding the recursion. In other
7005 words, a negative number counts capturing parentheses leftwards from
7006 the point at which it is encountered.
7008 It is also possible to refer to subsequently opened parentheses, by
7009 writing references such as (?+2). However, these cannot be recursive
7010 because the reference is not inside the parentheses that are refer-
7011 enced. They are always non-recursive subroutine calls, as described in
7014 An alternative approach is to use named parentheses instead. The Perl
7015 syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also
7016 supported. We could rewrite the above example as follows:
7018 (?<pn> \( ( [^()]++ | (?&pn) )* \) )
7020 If there is more than one subpattern with the same name, the earliest
7023 This particular example pattern that we have been looking at contains
7024 nested unlimited repeats, and so the use of a possessive quantifier for
7025 matching strings of non-parentheses is important when applying the pat-
7026 tern to strings that do not match. For example, when this pattern is
7029 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
7031 it yields "no match" quickly. However, if a possessive quantifier is
7032 not used, the match runs for a very long time indeed because there are
7033 so many different ways the + and * repeats can carve up the subject,
7034 and all have to be tested before failure can be reported.
7036 At the end of a match, the values of capturing parentheses are those
7037 from the outermost level. If you want to obtain intermediate values, a
7038 callout function can be used (see below and the pcrecallout documenta-
7039 tion). If the pattern above is matched against
7043 the value for the inner capturing parentheses (numbered 2) is "ef",
7044 which is the last value taken on at the top level. If a capturing sub-
7045 pattern is not matched at the top level, its final captured value is
7046 unset, even if it was (temporarily) set at a deeper level during the
7049 If there are more than 15 capturing parentheses in a pattern, PCRE has
7050 to obtain extra memory to store data during a recursion, which it does
7051 by using pcre_malloc, freeing it via pcre_free afterwards. If no memory
7052 can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
7054 Do not confuse the (?R) item with the condition (R), which tests for
7055 recursion. Consider this pattern, which matches text in angle brack-
7056 ets, allowing for arbitrary nesting. Only digits are allowed in nested
7057 brackets (that is, when recursing), whereas any characters are permit-
7058 ted at the outer level.
7060 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
7062 In this pattern, (?(R) is the start of a conditional subpattern, with
7063 two different alternatives for the recursive and non-recursive cases.
7064 The (?R) item is the actual recursive call.
7066 Differences in recursion processing between PCRE and Perl
7068 Recursion processing in PCRE differs from Perl in two important ways.
7069 In PCRE (like Python, but unlike Perl), a recursive subpattern call is
7070 always treated as an atomic group. That is, once it has matched some of
7071 the subject string, it is never re-entered, even if it contains untried
7072 alternatives and there is a subsequent matching failure. This can be
7073 illustrated by the following pattern, which purports to match a palin-
7074 dromic string that contains an odd number of characters (for example,
7075 "a", "aba", "abcba", "abcdcba"):
7079 The idea is that it either matches a single character, or two identical
7080 characters surrounding a sub-palindrome. In Perl, this pattern works;
7081 in PCRE it does not if the pattern is longer than three characters.
7082 Consider the subject string "abcba":
7084 At the top level, the first character is matched, but as it is not at
7085 the end of the string, the first alternative fails; the second alterna-
7086 tive is taken and the recursion kicks in. The recursive call to subpat-
7087 tern 1 successfully matches the next character ("b"). (Note that the
7088 beginning and end of line tests are not part of the recursion).
7090 Back at the top level, the next character ("c") is compared with what
7091 subpattern 2 matched, which was "a". This fails. Because the recursion
7092 is treated as an atomic group, there are now no backtracking points,
7093 and so the entire match fails. (Perl is able, at this point, to re-
7094 enter the recursion and try the second alternative.) However, if the
7095 pattern is written with the alternatives in the other order, things are
7100 This time, the recursing alternative is tried first, and continues to
7101 recurse until it runs out of characters, at which point the recursion
7102 fails. But this time we do have another alternative to try at the
7103 higher level. That is the big difference: in the previous case the
7104 remaining alternative is at a deeper recursion level, which PCRE cannot
7107 To change the pattern so that it matches all palindromic strings, not
7108 just those with an odd number of characters, it is tempting to change
7109 the pattern to this:
7113 Again, this works in Perl, but not in PCRE, and for the same reason.
7114 When a deeper recursion has matched a single character, it cannot be
7115 entered again in order to match an empty string. The solution is to
7116 separate the two cases, and write out the odd and even cases as alter-
7117 natives at the higher level:
7119 ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
7121 If you want to match typical palindromic phrases, the pattern has to
7122 ignore all non-word characters, which can be done like this:
7124 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
7126 If run with the PCRE_CASELESS option, this pattern matches phrases such
7127 as "A man, a plan, a canal: Panama!" and it works well in both PCRE and
7128 Perl. Note the use of the possessive quantifier *+ to avoid backtrack-
7129 ing into sequences of non-word characters. Without this, PCRE takes a
7130 great deal longer (ten times or more) to match typical phrases, and
7131 Perl takes so long that you think it has gone into a loop.
7133 WARNING: The palindrome-matching patterns above work only if the sub-
7134 ject string does not start with a palindrome that is shorter than the
7135 entire string. For example, although "abcba" is correctly matched, if
7136 the subject is "ababa", PCRE finds the palindrome "aba" at the start,
7137 then fails at top level because the end of the string does not follow.
7138 Once again, it cannot jump back into the recursion to try other alter-
7139 natives, so the entire match fails.
7141 The second way in which PCRE and Perl differ in their recursion pro-
7142 cessing is in the handling of captured values. In Perl, when a subpat-
7143 tern is called recursively or as a subpattern (see the next section),
7144 it has no access to any values that were captured outside the recur-
7145 sion, whereas in PCRE these values can be referenced. Consider this
7150 In PCRE, this pattern matches "bab". The first capturing parentheses
7151 match "b", then in the second group, when the back reference \1 fails
7152 to match "b", the second alternative matches "a" and then recurses. In
7153 the recursion, \1 does now match "b" and so the whole match succeeds.
7154 In Perl, the pattern fails to match because inside the recursive call
7155 \1 cannot access the externally set value.
7158 SUBPATTERNS AS SUBROUTINES
7160 If the syntax for a recursive subpattern call (either by number or by
7161 name) is used outside the parentheses to which it refers, it operates
7162 like a subroutine in a programming language. The called subpattern may
7163 be defined before or after the reference. A numbered reference can be
7164 absolute or relative, as in these examples:
7166 (...(absolute)...)...(?2)...
7167 (...(relative)...)...(?-1)...
7168 (...(?+1)...(relative)...
7170 An earlier example pointed out that the pattern
7172 (sens|respons)e and \1ibility
7174 matches "sense and sensibility" and "response and responsibility", but
7175 not "sense and responsibility". If instead the pattern
7177 (sens|respons)e and (?1)ibility
7179 is used, it does match "sense and responsibility" as well as the other
7180 two strings. Another example is given in the discussion of DEFINE
7183 All subroutine calls, whether recursive or not, are always treated as
7184 atomic groups. That is, once a subroutine has matched some of the sub-
7185 ject string, it is never re-entered, even if it contains untried alter-
7186 natives and there is a subsequent matching failure. Any capturing
7187 parentheses that are set during the subroutine call revert to their
7188 previous values afterwards.
7190 Processing options such as case-independence are fixed when a subpat-
7191 tern is defined, so if it is used as a subroutine, such options cannot
7192 be changed for different calls. For example, consider this pattern:
7196 It matches "abcabc". It does not match "abcABC" because the change of
7197 processing option does not affect the called subpattern.
7200 ONIGURUMA SUBROUTINE SYNTAX
7202 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
7203 name or a number enclosed either in angle brackets or single quotes, is
7204 an alternative syntax for referencing a subpattern as a subroutine,
7205 possibly recursively. Here are two of the examples used above, rewrit-
7206 ten using this syntax:
7208 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
7209 (sens|respons)e and \g'1'ibility
7211 PCRE supports an extension to Oniguruma: if a number is preceded by a
7212 plus or a minus sign it is taken as a relative reference. For example:
7216 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not
7217 synonymous. The former is a back reference; the latter is a subroutine
7223 Perl has a feature whereby using the sequence (?{...}) causes arbitrary
7224 Perl code to be obeyed in the middle of matching a regular expression.
7225 This makes it possible, amongst other things, to extract different sub-
7226 strings that match the same pair of parentheses when there is a repeti-
7229 PCRE provides a similar feature, but of course it cannot obey arbitrary
7230 Perl code. The feature is called "callout". The caller of PCRE provides
7231 an external function by putting its entry point in the global variable
7232 pcre_callout (8-bit library) or pcre[16|32]_callout (16-bit or 32-bit
7233 library). By default, this variable contains NULL, which disables all
7236 Within a regular expression, (?C) indicates the points at which the
7237 external function is to be called. If you want to identify different
7238 callout points, you can put a number less than 256 after the letter C.
7239 The default value is zero. For example, this pattern has two callout
7244 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, call-
7245 outs are automatically installed before each item in the pattern. They
7246 are all numbered 255. If there is a conditional group in the pattern
7247 whose condition is an assertion, an additional callout is inserted just
7248 before the condition. An explicit callout may also be set at this posi-
7249 tion, as in this example:
7251 (?(?C9)(?=a)abc|def)
7253 Note that this applies only to assertion conditions, not to other types
7256 During matching, when PCRE reaches a callout point, the external func-
7257 tion is called. It is provided with the number of the callout, the
7258 position in the pattern, and, optionally, one item of data originally
7259 supplied by the caller of the matching function. The callout function
7260 may cause matching to proceed, to backtrack, or to fail altogether.
7262 By default, PCRE implements a number of optimizations at compile time
7263 and matching time, and one side-effect is that sometimes callouts are
7264 skipped. If you need all possible callouts to happen, you need to set
7265 options that disable the relevant optimizations. More details, and a
7266 complete description of the interface to the callout function, are
7267 given in the pcrecallout documentation.
7270 BACKTRACKING CONTROL
7272 Perl 5.10 introduced a number of "Special Backtracking Control Verbs",
7273 which are still described in the Perl documentation as "experimental
7274 and subject to change or removal in a future version of Perl". It goes
7275 on to say: "Their usage in production code should be noted to avoid
7276 problems during upgrades." The same remarks apply to the PCRE features
7277 described in this section.
7279 The new verbs make use of what was previously invalid syntax: an open-
7280 ing parenthesis followed by an asterisk. They are generally of the form
7281 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
7282 differently depending on whether or not a name is present. A name is
7283 any sequence of characters that does not include a closing parenthesis.
7284 The maximum length of name is 255 in the 8-bit library and 65535 in the
7285 16-bit and 32-bit libraries. If the name is empty, that is, if the
7286 closing parenthesis immediately follows the colon, the effect is as if
7287 the colon were not there. Any number of these verbs may occur in a
7290 Since these verbs are specifically related to backtracking, most of
7291 them can be used only when the pattern is to be matched using one of
7292 the traditional matching functions, because these use a backtracking
7293 algorithm. With the exception of (*FAIL), which behaves like a failing
7294 negative assertion, the backtracking control verbs cause an error if
7295 encountered by a DFA matching function.
7297 The behaviour of these verbs in repeated groups, assertions, and in
7298 subpatterns called as subroutines (whether or not recursively) is docu-
7301 Optimizations that affect backtracking verbs
7303 PCRE contains some optimizations that are used to speed up matching by
7304 running some checks at the start of each match attempt. For example, it
7305 may know the minimum length of matching subject, or that a particular
7306 character must be present. When one of these optimizations bypasses the
7307 running of a match, any included backtracking verbs will not, of
7308 course, be processed. You can suppress the start-of-match optimizations
7309 by setting the PCRE_NO_START_OPTIMIZE option when calling pcre_com-
7310 pile() or pcre_exec(), or by starting the pattern with (*NO_START_OPT).
7311 There is more discussion of this option in the section entitled "Option
7312 bits for pcre_exec()" in the pcreapi documentation.
7314 Experiments with Perl suggest that it too has similar optimizations,
7315 sometimes leading to anomalous results.
7317 Verbs that act immediately
7319 The following verbs act as soon as they are encountered. They may not
7320 be followed by a name.
7324 This verb causes the match to end successfully, skipping the remainder
7325 of the pattern. However, when it is inside a subpattern that is called
7326 as a subroutine, only that subpattern is ended successfully. Matching
7327 then continues at the outer level. If (*ACCEPT) in triggered in a posi-
7328 tive assertion, the assertion succeeds; in a negative assertion, the
7331 If (*ACCEPT) is inside capturing parentheses, the data so far is cap-
7334 A((?:A|B(*ACCEPT)|C)D)
7336 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is cap-
7337 tured by the outer parentheses.
7341 This verb causes a matching failure, forcing backtracking to occur. It
7342 is equivalent to (?!) but easier to read. The Perl documentation notes
7343 that it is probably useful only when combined with (?{}) or (??{}).
7344 Those are, of course, Perl features that are not present in PCRE. The
7345 nearest equivalent is the callout feature, as for example in this pat-
7350 A match with the string "aaaa" always fails, but the callout is taken
7351 before each backtrack happens (in this example, 10 times).
7353 Recording which path was taken
7355 There is one verb whose main purpose is to track how a match was
7356 arrived at, though it also has a secondary use in conjunction with
7357 advancing the match starting point (see (*SKIP) below).
7359 (*MARK:NAME) or (*:NAME)
7361 A name is always required with this verb. There may be as many
7362 instances of (*MARK) as you like in a pattern, and their names do not
7365 When a match succeeds, the name of the last-encountered (*MARK:NAME),
7366 (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to
7367 the caller as described in the section entitled "Extra data for
7368 pcre_exec()" in the pcreapi documentation. Here is an example of
7369 pcretest output, where the /K modifier requests the retrieval and out-
7370 putting of (*MARK) data:
7372 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
7380 The (*MARK) name is tagged with "MK:" in this output, and in this exam-
7381 ple it indicates which of the two alternatives matched. This is a more
7382 efficient way of obtaining this information than putting each alterna-
7383 tive in its own capturing parentheses.
7385 If a verb with a name is encountered in a positive assertion that is
7386 true, the name is recorded and passed back if it is the last-encoun-
7387 tered. This does not happen for negative assertions or failing positive
7390 After a partial match or a failed match, the last encountered name in
7391 the entire match process is returned. For example:
7393 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
7397 Note that in this unanchored example the mark is retained from the
7398 match attempt that started at the letter "X" in the subject. Subsequent
7399 match attempts starting at "P" and then with an empty string do not get
7400 as far as the (*MARK) item, but nevertheless do not reset it.
7402 If you are interested in (*MARK) values after failed matches, you
7403 should probably set the PCRE_NO_START_OPTIMIZE option (see above) to
7404 ensure that the match is always attempted.
7406 Verbs that act after backtracking
7408 The following verbs do nothing when they are encountered. Matching con-
7409 tinues with what follows, but if there is no subsequent match, causing
7410 a backtrack to the verb, a failure is forced. That is, backtracking
7411 cannot pass to the left of the verb. However, when one of these verbs
7412 appears inside an atomic group or an assertion that is true, its effect
7413 is confined to that group, because once the group has been matched,
7414 there is never any backtracking into it. In this situation, backtrack-
7415 ing can "jump back" to the left of the entire atomic group or asser-
7416 tion. (Remember also, as stated above, that this localization also
7417 applies in subroutine calls.)
7419 These verbs differ in exactly what kind of failure occurs when back-
7420 tracking reaches them. The behaviour described below is what happens
7421 when the verb is not in a subroutine or an assertion. Subsequent sec-
7422 tions cover these special cases.
7426 This verb, which may not be followed by a name, causes the whole match
7427 to fail outright if there is a later matching failure that causes back-
7428 tracking to reach it. Even if the pattern is unanchored, no further
7429 attempts to find a match by advancing the starting point take place. If
7430 (*COMMIT) is the only backtracking verb that is encountered, once it
7431 has been passed pcre_exec() is committed to finding a match at the cur-
7432 rent starting point, or not at all. For example:
7436 This matches "xxaab" but not "aacaab". It can be thought of as a kind
7437 of dynamic anchor, or "I've started, so I must finish." The name of the
7438 most recently passed (*MARK) in the path is passed back when (*COMMIT)
7439 forces a match failure.
7441 If there is more than one backtracking verb in a pattern, a different
7442 one that follows (*COMMIT) may be triggered first, so merely passing
7443 (*COMMIT) during a match does not always guarantee that a match must be
7444 at this starting point.
7446 Note that (*COMMIT) at the start of a pattern is not the same as an
7447 anchor, unless PCRE's start-of-match optimizations are turned off, as
7448 shown in this output from pcretest:
7456 For this pattern, PCRE knows that any match must start with "a", so the
7457 optimization skips along the subject to "a" before applying the pattern
7458 to the first set of data. The match attempt then succeeds. In the sec-
7459 ond set of data, the escape sequence \Y is interpreted by the pcretest
7460 program. It causes the PCRE_NO_START_OPTIMIZE option to be set when
7461 pcre_exec() is called. This disables the optimization that skips along
7462 to the first character. The pattern is now applied starting at "x", and
7463 so the (*COMMIT) causes the match to fail without trying any other
7466 (*PRUNE) or (*PRUNE:NAME)
7468 This verb causes the match to fail at the current starting position in
7469 the subject if there is a later matching failure that causes backtrack-
7470 ing to reach it. If the pattern is unanchored, the normal "bumpalong"
7471 advance to the next starting character then happens. Backtracking can
7472 occur as usual to the left of (*PRUNE), before it is reached, or when
7473 matching to the right of (*PRUNE), but if there is no match to the
7474 right, backtracking cannot cross (*PRUNE). In simple cases, the use of
7475 (*PRUNE) is just an alternative to an atomic group or possessive quan-
7476 tifier, but there are some uses of (*PRUNE) that cannot be expressed in
7477 any other way. In an anchored pattern (*PRUNE) has the same effect as
7480 The behaviour of (*PRUNE:NAME) is the not the same as
7481 (*MARK:NAME)(*PRUNE). It is like (*MARK:NAME) in that the name is
7482 remembered for passing back to the caller. However, (*SKIP:NAME)
7483 searches only for names set with (*MARK).
7487 This verb, when given without a name, is like (*PRUNE), except that if
7488 the pattern is unanchored, the "bumpalong" advance is not to the next
7489 character, but to the position in the subject where (*SKIP) was encoun-
7490 tered. (*SKIP) signifies that whatever text was matched leading up to
7491 it cannot be part of a successful match. Consider:
7495 If the subject is "aaaac...", after the first match attempt fails
7496 (starting at the first character in the string), the starting point
7497 skips on to start the next attempt at "c". Note that a possessive quan-
7498 tifer does not have the same effect as this example; although it would
7499 suppress backtracking during the first match attempt, the second
7500 attempt would start at the second character instead of skipping on to
7505 When (*SKIP) has an associated name, its behaviour is modified. When it
7506 is triggered, the previous path through the pattern is searched for the
7507 most recent (*MARK) that has the same name. If one is found, the
7508 "bumpalong" advance is to the subject position that corresponds to that
7509 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with
7510 a matching name is found, the (*SKIP) is ignored.
7512 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It
7513 ignores names that are set by (*PRUNE:NAME) or (*THEN:NAME).
7515 (*THEN) or (*THEN:NAME)
7517 This verb causes a skip to the next innermost alternative when back-
7518 tracking reaches it. That is, it cancels any further backtracking
7519 within the current alternative. Its name comes from the observation
7520 that it can be used for a pattern-based if-then-else block:
7522 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
7524 If the COND1 pattern matches, FOO is tried (and possibly further items
7525 after the end of the group if FOO succeeds); on failure, the matcher
7526 skips to the second alternative and tries COND2, without backtracking
7527 into COND1. If that succeeds and BAR fails, COND3 is tried. If subse-
7528 quently BAZ fails, there are no more alternatives, so there is a back-
7529 track to whatever came before the entire group. If (*THEN) is not
7530 inside an alternation, it acts like (*PRUNE).
7532 The behaviour of (*THEN:NAME) is the not the same as
7533 (*MARK:NAME)(*THEN). It is like (*MARK:NAME) in that the name is
7534 remembered for passing back to the caller. However, (*SKIP:NAME)
7535 searches only for names set with (*MARK).
7537 A subpattern that does not contain a | character is just a part of the
7538 enclosing alternative; it is not a nested alternation with only one
7539 alternative. The effect of (*THEN) extends beyond such a subpattern to
7540 the enclosing alternative. Consider this pattern, where A, B, etc. are
7541 complex pattern fragments that do not contain any | characters at this
7546 If A and B are matched, but there is a failure in C, matching does not
7547 backtrack into A; instead it moves to the next alternative, that is, D.
7548 However, if the subpattern containing (*THEN) is given an alternative,
7549 it behaves differently:
7551 A (B(*THEN)C | (*FAIL)) | D
7553 The effect of (*THEN) is now confined to the inner subpattern. After a
7554 failure in C, matching moves to (*FAIL), which causes the whole subpat-
7555 tern to fail because there are no more alternatives to try. In this
7556 case, matching does now backtrack into A.
7558 Note that a conditional subpattern is not considered as having two
7559 alternatives, because only one is ever used. In other words, the |
7560 character in a conditional subpattern has a different meaning. Ignoring
7561 white space, consider:
7563 ^.*? (?(?=a) a | b(*THEN)c )
7565 If the subject is "ba", this pattern does not match. Because .*? is
7566 ungreedy, it initially matches zero characters. The condition (?=a)
7567 then fails, the character "b" is matched, but "c" is not. At this
7568 point, matching does not backtrack to .*? as might perhaps be expected
7569 from the presence of the | character. The conditional subpattern is
7570 part of the single alternative that comprises the whole pattern, and so
7571 the match fails. (If there was a backtrack into .*?, allowing it to
7572 match "b", the match would succeed.)
7574 The verbs just described provide four different "strengths" of control
7575 when subsequent matching fails. (*THEN) is the weakest, carrying on the
7576 match at the next alternative. (*PRUNE) comes next, failing the match
7577 at the current starting position, but allowing an advance to the next
7578 character (for an unanchored pattern). (*SKIP) is similar, except that
7579 the advance may be more than one character. (*COMMIT) is the strongest,
7580 causing the entire match to fail.
7582 More than one backtracking verb
7584 If more than one backtracking verb is present in a pattern, the one
7585 that is backtracked onto first acts. For example, consider this pat-
7586 tern, where A, B, etc. are complex pattern fragments:
7588 (A(*COMMIT)B(*THEN)C|ABD)
7590 If A matches but B fails, the backtrack to (*COMMIT) causes the entire
7591 match to fail. However, if A and B match, but C fails, the backtrack to
7592 (*THEN) causes the next alternative (ABD) to be tried. This behaviour
7593 is consistent, but is not always the same as Perl's. It means that if
7594 two or more backtracking verbs appear in succession, all the the last
7595 of them has no effect. Consider this example:
7597 ...(*COMMIT)(*PRUNE)...
7599 If there is a matching failure to the right, backtracking onto (*PRUNE)
7600 causes it to be triggered, and its action is taken. There can never be
7601 a backtrack onto (*COMMIT).
7603 Backtracking verbs in repeated groups
7605 PCRE differs from Perl in its handling of backtracking verbs in
7606 repeated groups. For example, consider:
7610 If the subject is "abac", Perl matches, but PCRE fails because the
7611 (*COMMIT) in the second repeat of the group acts.
7613 Backtracking verbs in assertions
7615 (*FAIL) in an assertion has its normal effect: it forces an immediate
7618 (*ACCEPT) in a positive assertion causes the assertion to succeed with-
7619 out any further processing. In a negative assertion, (*ACCEPT) causes
7620 the assertion to fail without any further processing.
7622 The other backtracking verbs are not treated specially if they appear
7623 in a positive assertion. In particular, (*THEN) skips to the next
7624 alternative in the innermost enclosing group that has alternations,
7625 whether or not this is within the assertion.
7627 Negative assertions are, however, different, in order to ensure that
7628 changing a positive assertion into a negative assertion changes its
7629 result. Backtracking into (*COMMIT), (*SKIP), or (*PRUNE) causes a neg-
7630 ative assertion to be true, without considering any further alternative
7631 branches in the assertion. Backtracking into (*THEN) causes it to skip
7632 to the next enclosing alternative within the assertion (the normal be-
7633 haviour), but if the assertion does not have such an alternative,
7634 (*THEN) behaves like (*PRUNE).
7636 Backtracking verbs in subroutines
7638 These behaviours occur whether or not the subpattern is called recur-
7639 sively. Perl's treatment of subroutines is different in some cases.
7641 (*FAIL) in a subpattern called as a subroutine has its normal effect:
7642 it forces an immediate backtrack.
7644 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine
7645 match to succeed without any further processing. Matching then contin-
7646 ues after the subroutine call.
7648 (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine
7649 cause the subroutine match to fail.
7651 (*THEN) skips to the next alternative in the innermost enclosing group
7652 within the subpattern that has alternatives. If there is no such group
7653 within the subpattern, (*THEN) causes the subroutine match to fail.
7658 pcreapi(3), pcrecallout(3), pcrematching(3), pcresyntax(3), pcre(3),
7659 pcre16(3), pcre32(3).
7665 University Computing Service
7666 Cambridge CB2 3QH, England.
7671 Last updated: 14 June 2015
7672 Copyright (c) 1997-2015 University of Cambridge.
7673 ------------------------------------------------------------------------------
7676 PCRESYNTAX(3) Library Functions Manual PCRESYNTAX(3)
7681 PCRE - Perl-compatible regular expressions
7683 PCRE REGULAR EXPRESSION SYNTAX SUMMARY
7685 The full syntax and semantics of the regular expressions that are sup-
7686 ported by PCRE are described in the pcrepattern documentation. This
7687 document contains a quick-reference summary of the syntax.
7692 \x where x is non-alphanumeric is a literal x
7693 \Q...\E treat enclosed characters as literal
7698 \a alarm, that is, the BEL character (hex 07)
7699 \cx "control-x", where x is any ASCII character
7701 \f form feed (hex 0C)
7703 \r carriage return (hex 0D)
7705 \0dd character with octal code 0dd
7706 \ddd character with octal code ddd, or backreference
7707 \o{ddd..} character with octal code ddd..
7708 \xhh character with hex code hh
7709 \x{hhh..} character with hex code hhh..
7711 Note that \0dd is always an octal code, and that \8 and \9 are the lit-
7712 eral characters "8" and "9".
7717 . any character except newline;
7718 in dotall mode, any character whatsoever
7719 \C one data unit, even in UTF mode (best avoided)
7721 \D a character that is not a decimal digit
7722 \h a horizontal white space character
7723 \H a character that is not a horizontal white space character
7724 \N a character that is not a newline
7725 \p{xx} a character with the xx property
7726 \P{xx} a character without the xx property
7727 \R a newline sequence
7728 \s a white space character
7729 \S a character that is not a white space character
7730 \v a vertical white space character
7731 \V a character that is not a vertical white space character
7732 \w a "word" character
7733 \W a "non-word" character
7734 \X a Unicode extended grapheme cluster
7736 By default, \d, \s, and \w match only ASCII characters, even in UTF-8
7737 mode or in the 16- bit and 32-bit libraries. However, if locale-spe-
7738 cific matching is happening, \s and \w may also match characters with
7739 code points in the range 128-255. If the PCRE_UCP option is set, the
7740 behaviour of these escape sequences is changed to use Unicode proper-
7741 ties and they match many more characters.
7744 GENERAL CATEGORY PROPERTIES FOR \p and \P
7754 Ll Lower case letter
7757 Lt Title case letter
7758 Lu Upper case letter
7772 Pc Connector punctuation
7774 Pe Close punctuation
7775 Pf Final punctuation
7776 Pi Initial punctuation
7777 Po Other punctuation
7783 Sm Mathematical symbol
7788 Zp Paragraph separator
7792 PCRE SPECIAL CATEGORY PROPERTIES FOR \p and \P
7794 Xan Alphanumeric: union of properties L and N
7795 Xps POSIX space: property Z or tab, NL, VT, FF, CR
7796 Xsp Perl space: property Z or tab, NL, VT, FF, CR
7797 Xuc Univerally-named character: one that can be
7798 represented by a Universal Character Name
7799 Xwd Perl word: property Xan or underscore
7801 Perl and POSIX space are now the same. Perl added VT to its space char-
7802 acter set at release 5.18 and PCRE changed at release 8.34.
7805 SCRIPT NAMES FOR \p AND \P
7807 Arabic, Armenian, Avestan, Balinese, Bamum, Bassa_Vah, Batak, Bengali,
7808 Bopomofo, Brahmi, Braille, Buginese, Buhid, Canadian_Aboriginal, Car-
7809 ian, Caucasian_Albanian, Chakma, Cham, Cherokee, Common, Coptic, Cunei-
7810 form, Cypriot, Cyrillic, Deseret, Devanagari, Duployan, Egyptian_Hiero-
7811 glyphs, Elbasan, Ethiopic, Georgian, Glagolitic, Gothic, Grantha,
7812 Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana,
7813 Imperial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscrip-
7814 tional_Parthian, Javanese, Kaithi, Kannada, Katakana, Kayah_Li,
7815 Kharoshthi, Khmer, Khojki, Khudawadi, Lao, Latin, Lepcha, Limbu, Lin-
7816 ear_A, Linear_B, Lisu, Lycian, Lydian, Mahajani, Malayalam, Mandaic,
7817 Manichaean, Meetei_Mayek, Mende_Kikakui, Meroitic_Cursive,
7818 Meroitic_Hieroglyphs, Miao, Modi, Mongolian, Mro, Myanmar, Nabataean,
7819 New_Tai_Lue, Nko, Ogham, Ol_Chiki, Old_Italic, Old_North_Arabian,
7820 Old_Permic, Old_Persian, Old_South_Arabian, Old_Turkic, Oriya, Osmanya,
7821 Pahawh_Hmong, Palmyrene, Pau_Cin_Hau, Phags_Pa, Phoenician,
7822 Psalter_Pahlavi, Rejang, Runic, Samaritan, Saurashtra, Sharada, Sha-
7823 vian, Siddham, Sinhala, Sora_Sompeng, Sundanese, Syloti_Nagri, Syriac,
7824 Tagalog, Tagbanwa, Tai_Le, Tai_Tham, Tai_Viet, Takri, Tamil, Telugu,
7825 Thaana, Thai, Tibetan, Tifinagh, Tirhuta, Ugaritic, Vai, Warang_Citi,
7831 [...] positive character class
7832 [^...] negative character class
7833 [x-y] range (can be used for hex characters)
7834 [[:xxx:]] positive POSIX named set
7835 [[:^xxx:]] negative POSIX named set
7841 cntrl control character
7843 graph printing, excluding space
7844 lower lower case letter
7845 print printing, including space
7846 punct printing, excluding alphanumeric
7848 upper upper case letter
7850 xdigit hexadecimal digit
7852 In PCRE, POSIX character set names recognize only ASCII characters by
7853 default, but some of them use Unicode properties if PCRE_UCP is set.
7854 You can use \Q...\E inside a character class.
7860 ?+ 0 or 1, possessive
7863 *+ 0 or more, possessive
7866 ++ 1 or more, possessive
7869 {n,m} at least n, no more than m, greedy
7870 {n,m}+ at least n, no more than m, possessive
7871 {n,m}? at least n, no more than m, lazy
7872 {n,} n or more, greedy
7873 {n,}+ n or more, possessive
7874 {n,}? n or more, lazy
7877 ANCHORS AND SIMPLE ASSERTIONS
7880 \B not a word boundary
7882 also after internal newline in multiline mode
7885 also before newline at end of subject
7886 also before internal newline in multiline mode
7888 also before newline at end of subject
7890 \G first matching position in subject
7895 \K reset start of match
7897 \K is honoured in positive assertions, but ignored in negative ones.
7907 (...) capturing group
7908 (?<name>...) named capturing group (Perl)
7909 (?'name'...) named capturing group (Perl)
7910 (?P<name>...) named capturing group (Python)
7911 (?:...) non-capturing group
7912 (?|...) non-capturing group; reset group numbers for
7913 capturing groups in each alternative
7918 (?>...) atomic, non-capturing group
7923 (?#....) comment (not nestable)
7929 (?J) allow duplicate names
7931 (?s) single line (dotall)
7932 (?U) default ungreedy (lazy)
7933 (?x) extended (ignore white space)
7934 (?-...) unset option(s)
7936 The following are recognized only at the very start of a pattern or
7937 after one of the newline or \R options with similar syntax. More than
7938 one of them may appear.
7940 (*LIMIT_MATCH=d) set the match limit to d (decimal number)
7941 (*LIMIT_RECURSION=d) set the recursion limit to d (decimal number)
7942 (*NO_AUTO_POSSESS) no auto-possessification (PCRE_NO_AUTO_POSSESS)
7943 (*NO_START_OPT) no start-match optimization (PCRE_NO_START_OPTIMIZE)
7944 (*UTF8) set UTF-8 mode: 8-bit library (PCRE_UTF8)
7945 (*UTF16) set UTF-16 mode: 16-bit library (PCRE_UTF16)
7946 (*UTF32) set UTF-32 mode: 32-bit library (PCRE_UTF32)
7947 (*UTF) set appropriate UTF mode for the library in use
7948 (*UCP) set PCRE_UCP (use Unicode properties for \d etc)
7950 Note that LIMIT_MATCH and LIMIT_RECURSION can only reduce the value of
7951 the limits set by the caller of pcre_exec(), not increase them.
7956 These are recognized only at the very start of the pattern or after
7957 option settings with a similar syntax.
7959 (*CR) carriage return only
7961 (*CRLF) carriage return followed by linefeed
7962 (*ANYCRLF) all three of the above
7963 (*ANY) any Unicode newline sequence
7968 These are recognized only at the very start of the pattern or after
7969 option setting with a similar syntax.
7971 (*BSR_ANYCRLF) CR, LF, or CRLF
7972 (*BSR_UNICODE) any Unicode newline sequence
7975 LOOKAHEAD AND LOOKBEHIND ASSERTIONS
7977 (?=...) positive look ahead
7978 (?!...) negative look ahead
7979 (?<=...) positive look behind
7980 (?<!...) negative look behind
7982 Each top-level branch of a look behind must be of a fixed length.
7987 \n reference by number (can be ambiguous)
7988 \gn reference by number
7989 \g{n} reference by number
7990 \g{-n} relative reference by number
7991 \k<name> reference by name (Perl)
7992 \k'name' reference by name (Perl)
7993 \g{name} reference by name (Perl)
7994 \k{name} reference by name (.NET)
7995 (?P=name) reference by name (Python)
7998 SUBROUTINE REFERENCES (POSSIBLY RECURSIVE)
8000 (?R) recurse whole pattern
8001 (?n) call subpattern by absolute number
8002 (?+n) call subpattern by relative number
8003 (?-n) call subpattern by relative number
8004 (?&name) call subpattern by name (Perl)
8005 (?P>name) call subpattern by name (Python)
8006 \g<name> call subpattern by name (Oniguruma)
8007 \g'name' call subpattern by name (Oniguruma)
8008 \g<n> call subpattern by absolute number (Oniguruma)
8009 \g'n' call subpattern by absolute number (Oniguruma)
8010 \g<+n> call subpattern by relative number (PCRE extension)
8011 \g'+n' call subpattern by relative number (PCRE extension)
8012 \g<-n> call subpattern by relative number (PCRE extension)
8013 \g'-n' call subpattern by relative number (PCRE extension)
8016 CONDITIONAL PATTERNS
8018 (?(condition)yes-pattern)
8019 (?(condition)yes-pattern|no-pattern)
8021 (?(n)... absolute reference condition
8022 (?(+n)... relative reference condition
8023 (?(-n)... relative reference condition
8024 (?(<name>)... named reference condition (Perl)
8025 (?('name')... named reference condition (Perl)
8026 (?(name)... named reference condition (PCRE)
8027 (?(R)... overall recursion condition
8028 (?(Rn)... specific group recursion condition
8029 (?(R&name)... specific recursion condition
8030 (?(DEFINE)... define subpattern for reference
8031 (?(assert)... assertion condition
8034 BACKTRACKING CONTROL
8036 The following act immediately they are reached:
8038 (*ACCEPT) force successful match
8039 (*FAIL) force backtrack; synonym (*F)
8040 (*MARK:NAME) set name to be passed back; synonym (*:NAME)
8042 The following act only when a subsequent match failure causes a back-
8043 track to reach them. They all force a match failure, but they differ in
8044 what happens afterwards. Those that advance the start-of-match point do
8045 so only if the pattern is not anchored.
8047 (*COMMIT) overall failure, no advance of starting point
8048 (*PRUNE) advance to next starting character
8049 (*PRUNE:NAME) equivalent to (*MARK:NAME)(*PRUNE)
8050 (*SKIP) advance to current matching position
8051 (*SKIP:NAME) advance to position corresponding to an earlier
8052 (*MARK:NAME); if not found, the (*SKIP) is ignored
8053 (*THEN) local failure, backtrack to next alternation
8054 (*THEN:NAME) equivalent to (*MARK:NAME)(*THEN)
8060 (?Cn) callout with data n
8065 pcrepattern(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
8071 University Computing Service
8072 Cambridge CB2 3QH, England.
8077 Last updated: 08 January 2014
8078 Copyright (c) 1997-2014 University of Cambridge.
8079 ------------------------------------------------------------------------------
8082 PCREUNICODE(3) Library Functions Manual PCREUNICODE(3)
8087 PCRE - Perl-compatible regular expressions
8089 UTF-8, UTF-16, UTF-32, AND UNICODE PROPERTY SUPPORT
8091 As well as UTF-8 support, PCRE also supports UTF-16 (from release 8.30)
8092 and UTF-32 (from release 8.32), by means of two additional libraries.
8093 They can be built as well as, or instead of, the 8-bit library.
8098 In order process UTF-8 strings, you must build PCRE's 8-bit library
8099 with UTF support, and, in addition, you must call pcre_compile() with
8100 the PCRE_UTF8 option flag, or the pattern must start with the sequence
8101 (*UTF8) or (*UTF). When either of these is the case, both the pattern
8102 and any subject strings that are matched against it are treated as
8103 UTF-8 strings instead of strings of individual 1-byte characters.
8106 UTF-16 AND UTF-32 SUPPORT
8108 In order process UTF-16 or UTF-32 strings, you must build PCRE's 16-bit
8109 or 32-bit library with UTF support, and, in addition, you must call
8110 pcre16_compile() or pcre32_compile() with the PCRE_UTF16 or PCRE_UTF32
8111 option flag, as appropriate. Alternatively, the pattern must start with
8112 the sequence (*UTF16), (*UTF32), as appropriate, or (*UTF), which can
8113 be used with either library. When UTF mode is set, both the pattern and
8114 any subject strings that are matched against it are treated as UTF-16
8115 or UTF-32 strings instead of strings of individual 16-bit or 32-bit
8119 UTF SUPPORT OVERHEAD
8121 If you compile PCRE with UTF support, but do not use it at run time,
8122 the library will be a bit bigger, but the additional run time overhead
8123 is limited to testing the PCRE_UTF[8|16|32] flag occasionally, so
8124 should not be very big.
8127 UNICODE PROPERTY SUPPORT
8129 If PCRE is built with Unicode character property support (which implies
8130 UTF support), the escape sequences \p{..}, \P{..}, and \X can be used.
8131 The available properties that can be tested are limited to the general
8132 category properties such as Lu for an upper case letter or Nd for a
8133 decimal number, the Unicode script names such as Arabic or Han, and the
8134 derived properties Any and L&. Full lists is given in the pcrepattern
8135 and pcresyntax documentation. Only the short names for properties are
8136 supported. For example, \p{L} matches a letter. Its Perl synonym,
8137 \p{Letter}, is not supported. Furthermore, in Perl, many properties
8138 may optionally be prefixed by "Is", for compatibility with Perl 5.6.
8139 PCRE does not support this.
8141 Validity of UTF-8 strings
8143 When you set the PCRE_UTF8 flag, the byte strings passed as patterns
8144 and subjects are (by default) checked for validity on entry to the rel-
8145 evant functions. The entire string is checked before any other process-
8146 ing takes place. From release 7.3 of PCRE, the check is according the
8147 rules of RFC 3629, which are themselves derived from the Unicode speci-
8148 fication. Earlier releases of PCRE followed the rules of RFC 2279,
8149 which allows the full range of 31-bit values (0 to 0x7FFFFFFF). The
8150 current check allows only values in the range U+0 to U+10FFFF, exclud-
8151 ing the surrogate area. (From release 8.33 the so-called "non-charac-
8152 ter" code points are no longer excluded because Unicode corrigendum #9
8153 makes it clear that they should not be.)
8155 Characters in the "Surrogate Area" of Unicode are reserved for use by
8156 UTF-16, where they are used in pairs to encode codepoints with values
8157 greater than 0xFFFF. The code points that are encoded by UTF-16 pairs
8158 are available independently in the UTF-8 and UTF-32 encodings. (In
8159 other words, the whole surrogate thing is a fudge for UTF-16 which
8160 unfortunately messes up UTF-8 and UTF-32.)
8162 If an invalid UTF-8 string is passed to PCRE, an error return is given.
8163 At compile time, the only additional information is the offset to the
8164 first byte of the failing character. The run-time functions pcre_exec()
8165 and pcre_dfa_exec() also pass back this information, as well as a more
8166 detailed reason code if the caller has provided memory in which to do
8169 In some situations, you may already know that your strings are valid,
8170 and therefore want to skip these checks in order to improve perfor-
8171 mance, for example in the case of a long subject string that is being
8172 scanned repeatedly. If you set the PCRE_NO_UTF8_CHECK flag at compile
8173 time or at run time, PCRE assumes that the pattern or subject it is
8174 given (respectively) contains only valid UTF-8 codes. In this case, it
8175 does not diagnose an invalid UTF-8 string.
8177 Note that passing PCRE_NO_UTF8_CHECK to pcre_compile() just disables
8178 the check for the pattern; it does not also apply to subject strings.
8179 If you want to disable the check for a subject string you must pass
8180 this option to pcre_exec() or pcre_dfa_exec().
8182 If you pass an invalid UTF-8 string when PCRE_NO_UTF8_CHECK is set, the
8183 result is undefined and your program may crash.
8185 Validity of UTF-16 strings
8187 When you set the PCRE_UTF16 flag, the strings of 16-bit data units that
8188 are passed as patterns and subjects are (by default) checked for valid-
8189 ity on entry to the relevant functions. Values other than those in the
8190 surrogate range U+D800 to U+DFFF are independent code points. Values in
8191 the surrogate range must be used in pairs in the correct manner.
8193 If an invalid UTF-16 string is passed to PCRE, an error return is
8194 given. At compile time, the only additional information is the offset
8195 to the first data unit of the failing character. The run-time functions
8196 pcre16_exec() and pcre16_dfa_exec() also pass back this information, as
8197 well as a more detailed reason code if the caller has provided memory
8198 in which to do this.
8200 In some situations, you may already know that your strings are valid,
8201 and therefore want to skip these checks in order to improve perfor-
8202 mance. If you set the PCRE_NO_UTF16_CHECK flag at compile time or at
8203 run time, PCRE assumes that the pattern or subject it is given (respec-
8204 tively) contains only valid UTF-16 sequences. In this case, it does not
8205 diagnose an invalid UTF-16 string. However, if an invalid string is
8206 passed, the result is undefined.
8208 Validity of UTF-32 strings
8210 When you set the PCRE_UTF32 flag, the strings of 32-bit data units that
8211 are passed as patterns and subjects are (by default) checked for valid-
8212 ity on entry to the relevant functions. This check allows only values
8213 in the range U+0 to U+10FFFF, excluding the surrogate area U+D800 to
8216 If an invalid UTF-32 string is passed to PCRE, an error return is
8217 given. At compile time, the only additional information is the offset
8218 to the first data unit of the failing character. The run-time functions
8219 pcre32_exec() and pcre32_dfa_exec() also pass back this information, as
8220 well as a more detailed reason code if the caller has provided memory
8221 in which to do this.
8223 In some situations, you may already know that your strings are valid,
8224 and therefore want to skip these checks in order to improve perfor-
8225 mance. If you set the PCRE_NO_UTF32_CHECK flag at compile time or at
8226 run time, PCRE assumes that the pattern or subject it is given (respec-
8227 tively) contains only valid UTF-32 sequences. In this case, it does not
8228 diagnose an invalid UTF-32 string. However, if an invalid string is
8229 passed, the result is undefined.
8231 General comments about UTF modes
8233 1. Codepoints less than 256 can be specified in patterns by either
8234 braced or unbraced hexadecimal escape sequences (for example, \x{b3} or
8235 \xb3). Larger values have to use braced sequences.
8237 2. Octal numbers up to \777 are recognized, and in UTF-8 mode they
8238 match two-byte characters for values greater than \177.
8240 3. Repeat quantifiers apply to complete UTF characters, not to individ-
8241 ual data units, for example: \x{100}{3}.
8243 4. The dot metacharacter matches one UTF character instead of a single
8246 5. The escape sequence \C can be used to match a single byte in UTF-8
8247 mode, or a single 16-bit data unit in UTF-16 mode, or a single 32-bit
8248 data unit in UTF-32 mode, but its use can lead to some strange effects
8249 because it breaks up multi-unit characters (see the description of \C
8250 in the pcrepattern documentation). The use of \C is not supported in
8251 the alternative matching function pcre[16|32]_dfa_exec(), nor is it
8252 supported in UTF mode by the JIT optimization of pcre[16|32]_exec(). If
8253 JIT optimization is requested for a UTF pattern that contains \C, it
8254 will not succeed, and so the matching will be carried out by the normal
8255 interpretive function.
8257 6. The character escapes \b, \B, \d, \D, \s, \S, \w, and \W correctly
8258 test characters of any code value, but, by default, the characters that
8259 PCRE recognizes as digits, spaces, or word characters remain the same
8260 set as in non-UTF mode, all with values less than 256. This remains
8261 true even when PCRE is built to include Unicode property support,
8262 because to do otherwise would slow down PCRE in many common cases. Note
8263 in particular that this applies to \b and \B, because they are defined
8264 in terms of \w and \W. If you really want to test for a wider sense of,
8265 say, "digit", you can use explicit Unicode property tests such as
8266 \p{Nd}. Alternatively, if you set the PCRE_UCP option, the way that the
8267 character escapes work is changed so that Unicode properties are used
8268 to determine which characters match. There are more details in the sec-
8269 tion on generic character types in the pcrepattern documentation.
8271 7. Similarly, characters that match the POSIX named character classes
8272 are all low-valued characters, unless the PCRE_UCP option is set.
8274 8. However, the horizontal and vertical white space matching escapes
8275 (\h, \H, \v, and \V) do match all the appropriate Unicode characters,
8276 whether or not PCRE_UCP is set.
8278 9. Case-insensitive matching applies only to characters whose values
8279 are less than 128, unless PCRE is built with Unicode property support.
8280 A few Unicode characters such as Greek sigma have more than two code-
8281 points that are case-equivalent. Up to and including PCRE release 8.31,
8282 only one-to-one case mappings were supported, but later releases (with
8283 Unicode property support) do treat as case-equivalent all versions of
8284 characters such as Greek sigma.
8290 University Computing Service
8291 Cambridge CB2 3QH, England.
8296 Last updated: 27 February 2013
8297 Copyright (c) 1997-2013 University of Cambridge.
8298 ------------------------------------------------------------------------------
8301 PCREJIT(3) Library Functions Manual PCREJIT(3)
8306 PCRE - Perl-compatible regular expressions
8308 PCRE JUST-IN-TIME COMPILER SUPPORT
8310 Just-in-time compiling is a heavyweight optimization that can greatly
8311 speed up pattern matching. However, it comes at the cost of extra pro-
8312 cessing before the match is performed. Therefore, it is of most benefit
8313 when the same pattern is going to be matched many times. This does not
8314 necessarily mean many calls of a matching function; if the pattern is
8315 not anchored, matching attempts may take place many times at various
8316 positions in the subject, even for a single call. Therefore, if the
8317 subject string is very long, it may still pay to use JIT for one-off
8320 JIT support applies only to the traditional Perl-compatible matching
8321 function. It does not apply when the DFA matching function is being
8322 used. The code for this support was written by Zoltan Herczeg.
8325 8-BIT, 16-BIT AND 32-BIT SUPPORT
8327 JIT support is available for all of the 8-bit, 16-bit and 32-bit PCRE
8328 libraries. To keep this documentation simple, only the 8-bit interface
8329 is described in what follows. If you are using the 16-bit library, sub-
8330 stitute the 16-bit functions and 16-bit structures (for example,
8331 pcre16_jit_stack instead of pcre_jit_stack). If you are using the
8332 32-bit library, substitute the 32-bit functions and 32-bit structures
8333 (for example, pcre32_jit_stack instead of pcre_jit_stack).
8336 AVAILABILITY OF JIT SUPPORT
8338 JIT support is an optional feature of PCRE. The "configure" option
8339 --enable-jit (or equivalent CMake option) must be set when PCRE is
8340 built if you want to use JIT. The support is limited to the following
8343 ARM v5, v7, and Thumb2
8344 Intel x86 32-bit and 64-bit
8346 Power PC 32-bit and 64-bit
8347 SPARC 32-bit (experimental)
8349 If --enable-jit is set on an unsupported platform, compilation fails.
8351 A program that is linked with PCRE 8.20 or later can tell if JIT sup-
8352 port is available by calling pcre_config() with the PCRE_CONFIG_JIT
8353 option. The result is 1 when JIT is available, and 0 otherwise. How-
8354 ever, a simple program does not need to check this in order to use JIT.
8355 The normal API is implemented in a way that falls back to the interpre-
8356 tive code if JIT is not available. For programs that need the best pos-
8357 sible performance, there is also a "fast path" API that is JIT-spe-
8360 If your program may sometimes be linked with versions of PCRE that are
8361 older than 8.20, but you want to use JIT when it is available, you can
8362 test the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT
8363 macro such as PCRE_CONFIG_JIT, for compile-time control of your code.
8368 You have to do two things to make use of the JIT support in the sim-
8371 (1) Call pcre_study() with the PCRE_STUDY_JIT_COMPILE option for
8372 each compiled pattern, and pass the resulting pcre_extra block to
8375 (2) Use pcre_free_study() to free the pcre_extra block when it is
8376 no longer needed, instead of just freeing it yourself. This
8378 any JIT data is also freed.
8380 For a program that may be linked with pre-8.20 versions of PCRE, you
8383 #ifndef PCRE_STUDY_JIT_COMPILE
8384 #define PCRE_STUDY_JIT_COMPILE 0
8387 so that no option is passed to pcre_study(), and then use something
8388 like this to free the study data:
8390 #ifdef PCRE_CONFIG_JIT
8391 pcre_free_study(study_ptr);
8393 pcre_free(study_ptr);
8396 PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for
8397 complete matches. If you want to run partial matches using the
8398 PCRE_PARTIAL_HARD or PCRE_PARTIAL_SOFT options of pcre_exec(), you
8399 should set one or both of the following options in addition to, or
8400 instead of, PCRE_STUDY_JIT_COMPILE when you call pcre_study():
8402 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
8403 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
8405 The JIT compiler generates different optimized code for each of the
8406 three modes (normal, soft partial, hard partial). When pcre_exec() is
8407 called, the appropriate code is run if it is available. Otherwise, the
8408 pattern is matched using interpretive code.
8410 In some circumstances you may need to call additional functions. These
8411 are described in the section entitled "Controlling the JIT stack"
8414 If JIT support is not available, PCRE_STUDY_JIT_COMPILE etc. are
8415 ignored, and no JIT data is created. Otherwise, the compiled pattern is
8416 passed to the JIT compiler, which turns it into machine code that exe-
8417 cutes much faster than the normal interpretive code. When pcre_exec()
8418 is passed a pcre_extra block containing a pointer to JIT code of the
8419 appropriate mode (normal or hard/soft partial), it obeys that code
8420 instead of running the interpreter. The result is identical, but the
8421 compiled JIT code runs much faster.
8423 There are some pcre_exec() options that are not supported for JIT exe-
8424 cution. There are also some pattern items that JIT cannot handle.
8425 Details are given below. In both cases, execution automatically falls
8426 back to the interpretive code. If you want to know whether JIT was
8427 actually used for a particular match, you should arrange for a JIT
8428 callback function to be set up as described in the section entitled
8429 "Controlling the JIT stack" below, even if you do not need to supply a
8430 non-default JIT stack. Such a callback function is called whenever JIT
8431 code is about to be obeyed. If the execution options are not right for
8432 JIT execution, the callback function is not obeyed.
8434 If the JIT compiler finds an unsupported item, no JIT data is gener-
8435 ated. You can find out if JIT execution is available after studying a
8436 pattern by calling pcre_fullinfo() with the PCRE_INFO_JIT option. A
8437 result of 1 means that JIT compilation was successful. A result of 0
8438 means that JIT support is not available, or the pattern was not studied
8439 with PCRE_STUDY_JIT_COMPILE etc., or the JIT compiler was not able to
8442 Once a pattern has been studied, with or without JIT, it can be used as
8443 many times as you like for matching different subject strings.
8446 UNSUPPORTED OPTIONS AND PATTERN ITEMS
8448 The only pcre_exec() options that are supported for JIT execution are
8449 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK, PCRE_NO_UTF32_CHECK, PCRE_NOT-
8450 BOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, PCRE_PAR-
8451 TIAL_HARD, and PCRE_PARTIAL_SOFT.
8453 The only unsupported pattern items are \C (match a single data unit)
8454 when running in a UTF mode, and a callout immediately before an asser-
8455 tion condition in a conditional group.
8458 RETURN VALUES FROM JIT EXECUTION
8460 When a pattern is matched using JIT execution, the return values are
8461 the same as those given by the interpretive pcre_exec() code, with the
8462 addition of one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means
8463 that the memory used for the JIT stack was insufficient. See "Control-
8464 ling the JIT stack" below for a discussion of JIT stack usage. For com-
8465 patibility with the interpretive pcre_exec() code, no more than two-
8466 thirds of the ovector argument is used for passing back captured sub-
8469 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if
8470 searching a very large pattern tree goes on for too long, as it is in
8471 the same circumstance when JIT is not used, but the details of exactly
8472 what is counted are not the same. The PCRE_ERROR_RECURSIONLIMIT error
8473 code is never returned by JIT execution.
8476 SAVING AND RESTORING COMPILED PATTERNS
8478 The code that is generated by the JIT compiler is architecture-spe-
8479 cific, and is also position dependent. For those reasons it cannot be
8480 saved (in a file or database) and restored later like the bytecode and
8481 other data of a compiled pattern. Saving and restoring compiled pat-
8482 terns is not something many people do. More detail about this facility
8483 is given in the pcreprecompile documentation. It should be possible to
8484 run pcre_study() on a saved and restored pattern, and thereby recreate
8485 the JIT data, but because JIT compilation uses significant resources,
8486 it is probably not worth doing this; you might as well recompile the
8490 CONTROLLING THE JIT STACK
8492 When the compiled JIT code runs, it needs a block of memory to use as a
8493 stack. By default, it uses 32K on the machine stack. However, some
8494 large or complicated patterns need more than this. The error
8495 PCRE_ERROR_JIT_STACKLIMIT is given when there is not enough stack.
8496 Three functions are provided for managing blocks of memory for use as
8497 JIT stacks. There is further discussion about the use of JIT stacks in
8498 the section entitled "JIT stack FAQ" below.
8500 The pcre_jit_stack_alloc() function creates a JIT stack. Its arguments
8501 are a starting size and a maximum size, and it returns a pointer to an
8502 opaque structure of type pcre_jit_stack, or NULL if there is an error.
8503 The pcre_jit_stack_free() function can be used to free a stack that is
8504 no longer needed. (For the technically minded: the address space is
8505 allocated by mmap or VirtualAlloc.)
8507 JIT uses far less memory for recursion than the interpretive code, and
8508 a maximum stack size of 512K to 1M should be more than enough for any
8511 The pcre_assign_jit_stack() function specifies which stack JIT code
8512 should use. Its arguments are as follows:
8515 pcre_jit_callback callback
8518 The extra argument must be the result of studying a pattern with
8519 PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the
8522 (1) If callback is NULL and data is NULL, an internal 32K block
8523 on the machine stack is used.
8525 (2) If callback is NULL and data is not NULL, data must be
8526 a valid JIT stack, the result of calling pcre_jit_stack_alloc().
8528 (3) If callback is not NULL, it must point to a function that is
8529 called with data as an argument at the start of matching, in
8530 order to set up a JIT stack. If the return from the callback
8531 function is NULL, the internal 32K stack is used; otherwise the
8532 return value must be a valid JIT stack, the result of calling
8533 pcre_jit_stack_alloc().
8535 A callback function is obeyed whenever JIT code is about to be run; it
8536 is not obeyed when pcre_exec() is called with options that are incom-
8537 patible for JIT execution. A callback function can therefore be used to
8538 determine whether a match operation was executed by JIT or by the
8541 You may safely use the same JIT stack for more than one pattern (either
8542 by assigning directly or by callback), as long as the patterns are all
8543 matched sequentially in the same thread. In a multithread application,
8544 if you do not specify a JIT stack, or if you assign or pass back NULL
8545 from a callback, that is thread-safe, because each thread has its own
8546 machine stack. However, if you assign or pass back a non-NULL JIT
8547 stack, this must be a different stack for each thread so that the
8548 application is thread-safe.
8550 Strictly speaking, even more is allowed. You can assign the same non-
8551 NULL stack to any number of patterns as long as they are not used for
8552 matching by multiple threads at the same time. For example, you can
8553 assign the same stack to all compiled patterns, and use a global mutex
8554 in the callback to wait until the stack is available for use. However,
8555 this is an inefficient solution, and not recommended.
8557 This is a suggestion for how a multithreaded program that needs to set
8558 up non-default JIT stacks might operate:
8560 During thread initalization
8561 thread_local_var = pcre_jit_stack_alloc(...)
8564 pcre_jit_stack_free(thread_local_var)
8566 Use a one-line callback function
8567 return thread_local_var
8569 All the functions described in this section do nothing if JIT is not
8570 available, and pcre_assign_jit_stack() does nothing unless the extra
8571 argument is non-NULL and points to a pcre_extra block that is the
8572 result of a successful study with PCRE_STUDY_JIT_COMPILE etc.
8577 (1) Why do we need JIT stacks?
8579 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack
8580 where the local data of the current node is pushed before checking its
8581 child nodes. Allocating real machine stack on some platforms is diffi-
8582 cult. For example, the stack chain needs to be updated every time if we
8583 extend the stack on PowerPC. Although it is possible, its updating
8584 time overhead decreases performance. So we do the recursion in memory.
8586 (2) Why don't we simply allocate blocks of memory with malloc()?
8588 Modern operating systems have a nice feature: they can reserve an
8589 address space instead of allocating memory. We can safely allocate mem-
8590 ory pages inside this address space, so the stack could grow without
8591 moving memory data (this is important because of pointers). Thus we can
8592 allocate 1M address space, and use only a single memory page (usually
8593 4K) if that is enough. However, we can still grow up to 1M anytime if
8596 (3) Who "owns" a JIT stack?
8598 The owner of the stack is the user program, not the JIT studied pattern
8599 or anything else. The user program must ensure that if a stack is used
8600 by pcre_exec(), (that is, it is assigned to the pattern currently run-
8601 ning), that stack must not be used by any other threads (to avoid over-
8602 writing the same memory area). The best practice for multithreaded pro-
8603 grams is to allocate a stack for each thread, and return this stack
8604 through the JIT callback function.
8606 (4) When should a JIT stack be freed?
8608 You can free a JIT stack at any time, as long as it will not be used by
8609 pcre_exec() again. When you assign the stack to a pattern, only a
8610 pointer is set. There is no reference counting or any other magic. You
8611 can free the patterns and stacks in any order, anytime. Just do not
8612 call pcre_exec() with a pattern pointing to an already freed stack, as
8613 that will cause SEGFAULT. (Also, do not free a stack currently used by
8614 pcre_exec() in another thread). You can also replace the stack for a
8615 pattern at any time. You can even free the previous stack before
8616 assigning a replacement.
8618 (5) Should I allocate/free a stack every time before/after calling
8621 No, because this is too costly in terms of resources. However, you
8622 could implement some clever idea which release the stack if it is not
8623 used in let's say two minutes. The JIT callback can help to achieve
8624 this without keeping a list of the currently JIT studied patterns.
8626 (6) OK, the stack is for long term memory allocation. But what happens
8627 if a pattern causes stack overflow with a stack of 1M? Is that 1M kept
8628 until the stack is freed?
8630 Especially on embedded sytems, it might be a good idea to release mem-
8631 ory sometimes without freeing the stack. There is no API for this at
8632 the moment. Probably a function call which returns with the currently
8633 allocated memory for any stack and another which allows releasing mem-
8634 ory (shrinking the stack) would be a good idea if someone needs this.
8636 (7) This is too much of a headache. Isn't there any better solution for
8639 No, thanks to Windows. If POSIX threads were used everywhere, we could
8640 throw out this complicated API.
8645 This is a single-threaded example that specifies a JIT stack without
8652 pcre_jit_stack *jit_stack;
8654 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
8655 /* Check for errors */
8656 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
8657 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
8658 /* Check for error (NULL) */
8659 pcre_assign_jit_stack(extra, NULL, jit_stack);
8660 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
8663 pcre_free_study(extra);
8664 pcre_jit_stack_free(jit_stack);
8669 Because the API described above falls back to interpreted execution
8670 when JIT is not available, it is convenient for programs that are writ-
8671 ten for general use in many environments. However, calling JIT via
8672 pcre_exec() does have a performance impact. Programs that are written
8673 for use where JIT is known to be available, and which need the best
8674 possible performance, can instead use a "fast path" API to call JIT
8675 execution directly instead of calling pcre_exec() (obviously only for
8676 patterns that have been successfully studied by JIT).
8678 The fast path function is called pcre_jit_exec(), and it takes exactly
8679 the same arguments as pcre_exec(), plus one additional argument that
8680 must point to a JIT stack. The JIT stack arrangements described above
8681 do not apply. The return values are the same as for pcre_exec().
8683 When you call pcre_exec(), as well as testing for invalid options, a
8684 number of other sanity checks are performed on the arguments. For exam-
8685 ple, if the subject pointer is NULL, or its length is negative, an
8686 immediate error is given. Also, unless PCRE_NO_UTF[8|16|32] is set, a
8687 UTF subject string is tested for validity. In the interests of speed,
8688 these checks do not happen on the JIT fast path, and if invalid data is
8689 passed, the result is undefined.
8691 Bypassing the sanity checks and the pcre_exec() wrapping can give
8692 speedups of more than 10%.
8702 Philip Hazel (FAQ by Zoltan Herczeg)
8703 University Computing Service
8704 Cambridge CB2 3QH, England.
8709 Last updated: 17 March 2013
8710 Copyright (c) 1997-2013 University of Cambridge.
8711 ------------------------------------------------------------------------------
8714 PCREPARTIAL(3) Library Functions Manual PCREPARTIAL(3)
8719 PCRE - Perl-compatible regular expressions
8721 PARTIAL MATCHING IN PCRE
8723 In normal use of PCRE, if the subject string that is passed to a match-
8724 ing function matches as far as it goes, but is too short to match the
8725 entire pattern, PCRE_ERROR_NOMATCH is returned. There are circumstances
8726 where it might be helpful to distinguish this case from other cases in
8727 which there is no match.
8729 Consider, for example, an application where a human is required to type
8730 in data for a field with specific formatting requirements. An example
8731 might be a date in the form ddmmmyy, defined by this pattern:
8733 ^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$
8735 If the application sees the user's keystrokes one by one, and can check
8736 that what has been typed so far is potentially valid, it is able to
8737 raise an error as soon as a mistake is made, by beeping and not
8738 reflecting the character that has been typed, for example. This immedi-
8739 ate feedback is likely to be a better user interface than a check that
8740 is delayed until the entire string has been entered. Partial matching
8741 can also be useful when the subject string is very long and is not all
8744 PCRE supports partial matching by means of the PCRE_PARTIAL_SOFT and
8745 PCRE_PARTIAL_HARD options, which can be set when calling any of the
8746 matching functions. For backwards compatibility, PCRE_PARTIAL is a syn-
8747 onym for PCRE_PARTIAL_SOFT. The essential difference between the two
8748 options is whether or not a partial match is preferred to an alterna-
8749 tive complete match, though the details differ between the two types of
8750 matching function. If both options are set, PCRE_PARTIAL_HARD takes
8753 If you want to use partial matching with just-in-time optimized code,
8754 you must call pcre_study(), pcre16_study() or pcre32_study() with one
8755 or both of these options:
8757 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
8758 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
8760 PCRE_STUDY_JIT_COMPILE should also be set if you are going to run non-
8761 partial matches on the same pattern. If the appropriate JIT study mode
8762 has not been set for a match, the interpretive matching code is used.
8764 Setting a partial matching option disables two of PCRE's standard opti-
8765 mizations. PCRE remembers the last literal data unit in a pattern, and
8766 abandons matching immediately if it is not present in the subject
8767 string. This optimization cannot be used for a subject string that
8768 might match only partially. If the pattern was studied, PCRE knows the
8769 minimum length of a matching string, and does not bother to run the
8770 matching function on shorter strings. This optimization is also dis-
8771 abled for partial matching.
8774 PARTIAL MATCHING USING pcre_exec() OR pcre[16|32]_exec()
8776 A partial match occurs during a call to pcre_exec() or
8777 pcre[16|32]_exec() when the end of the subject string is reached suc-
8778 cessfully, but matching cannot continue because more characters are
8779 needed. However, at least one character in the subject must have been
8780 inspected. This character need not form part of the final matched
8781 string; lookbehind assertions and the \K escape sequence provide ways
8782 of inspecting characters before the start of a matched substring. The
8783 requirement for inspecting at least one character exists because an
8784 empty string can always be matched; without such a restriction there
8785 would always be a partial match of an empty string at the end of the
8788 If there are at least two slots in the offsets vector when a partial
8789 match is returned, the first slot is set to the offset of the earliest
8790 character that was inspected. For convenience, the second offset points
8791 to the end of the subject so that a substring can easily be identified.
8792 If there are at least three slots in the offsets vector, the third slot
8793 is set to the offset of the character where matching started.
8795 For the majority of patterns, the contents of the first and third slots
8796 will be the same. However, for patterns that contain lookbehind asser-
8797 tions, or begin with \b or \B, characters before the one where matching
8798 started may have been inspected while carrying out the match. For exam-
8799 ple, consider this pattern:
8803 This pattern matches "123", but only if it is preceded by "abc". If the
8804 subject string is "xyzabc12", the first two offsets after a partial
8805 match are for the substring "abc12", because all these characters were
8806 inspected. However, the third offset is set to 6, because that is the
8807 offset where matching began.
8809 What happens when a partial match is identified depends on which of the
8810 two partial matching options are set.
8812 PCRE_PARTIAL_SOFT WITH pcre_exec() OR pcre[16|32]_exec()
8814 If PCRE_PARTIAL_SOFT is set when pcre_exec() or pcre[16|32]_exec()
8815 identifies a partial match, the partial match is remembered, but match-
8816 ing continues as normal, and other alternatives in the pattern are
8817 tried. If no complete match can be found, PCRE_ERROR_PARTIAL is
8818 returned instead of PCRE_ERROR_NOMATCH.
8820 This option is "soft" because it prefers a complete match over a par-
8821 tial match. All the various matching items in a pattern behave as if
8822 the subject string is potentially complete. For example, \z, \Z, and $
8823 match at the end of the subject, as normal, and for \b and \B the end
8824 of the subject is treated as a non-alphanumeric.
8826 If there is more than one partial match, the first one that was found
8827 provides the data that is returned. Consider this pattern:
8831 If this is matched against the subject string "abc123dog", both alter-
8832 natives fail to match, but the end of the subject is reached during
8833 matching, so PCRE_ERROR_PARTIAL is returned. The offsets are set to 3
8834 and 9, identifying "123dog" as the first partial match that was found.
8835 (In this example, there are two partial matches, because "dog" on its
8836 own partially matches the second alternative.)
8838 PCRE_PARTIAL_HARD WITH pcre_exec() OR pcre[16|32]_exec()
8840 If PCRE_PARTIAL_HARD is set for pcre_exec() or pcre[16|32]_exec(),
8841 PCRE_ERROR_PARTIAL is returned as soon as a partial match is found,
8842 without continuing to search for possible complete matches. This option
8843 is "hard" because it prefers an earlier partial match over a later com-
8844 plete match. For this reason, the assumption is made that the end of
8845 the supplied subject string may not be the true end of the available
8846 data, and so, if \z, \Z, \b, \B, or $ are encountered at the end of the
8847 subject, the result is PCRE_ERROR_PARTIAL, provided that at least one
8848 character in the subject has been inspected.
8850 Setting PCRE_PARTIAL_HARD also affects the way UTF-8 and UTF-16 subject
8851 strings are checked for validity. Normally, an invalid sequence causes
8852 the error PCRE_ERROR_BADUTF8 or PCRE_ERROR_BADUTF16. However, in the
8853 special case of a truncated character at the end of the subject,
8854 PCRE_ERROR_SHORTUTF8 or PCRE_ERROR_SHORTUTF16 is returned when
8855 PCRE_PARTIAL_HARD is set.
8857 Comparing hard and soft partial matching
8859 The difference between the two partial matching options can be illus-
8860 trated by a pattern such as:
8864 This matches either "dog" or "dogsbody", greedily (that is, it prefers
8865 the longer string if possible). If it is matched against the string
8866 "dog" with PCRE_PARTIAL_SOFT, it yields a complete match for "dog".
8867 However, if PCRE_PARTIAL_HARD is set, the result is PCRE_ERROR_PARTIAL.
8868 On the other hand, if the pattern is made ungreedy the result is dif-
8873 In this case the result is always a complete match because that is
8874 found first, and matching never continues after finding a complete
8875 match. It might be easier to follow this explanation by thinking of the
8876 two patterns like this:
8878 /dog(sbody)?/ is the same as /dogsbody|dog/
8879 /dog(sbody)??/ is the same as /dog|dogsbody/
8881 The second pattern will never match "dogsbody", because it will always
8882 find the shorter match first.
8885 PARTIAL MATCHING USING pcre_dfa_exec() OR pcre[16|32]_dfa_exec()
8887 The DFA functions move along the subject string character by character,
8888 without backtracking, searching for all possible matches simultane-
8889 ously. If the end of the subject is reached before the end of the pat-
8890 tern, there is the possibility of a partial match, again provided that
8891 at least one character has been inspected.
8893 When PCRE_PARTIAL_SOFT is set, PCRE_ERROR_PARTIAL is returned only if
8894 there have been no complete matches. Otherwise, the complete matches
8895 are returned. However, if PCRE_PARTIAL_HARD is set, a partial match
8896 takes precedence over any complete matches. The portion of the string
8897 that was inspected when the longest partial match was found is set as
8898 the first matching string, provided there are at least two slots in the
8901 Because the DFA functions always search for all possible matches, and
8902 there is no difference between greedy and ungreedy repetition, their
8903 behaviour is different from the standard functions when PCRE_PAR-
8904 TIAL_HARD is set. Consider the string "dog" matched against the
8905 ungreedy pattern shown above:
8909 Whereas the standard functions stop as soon as they find the complete
8910 match for "dog", the DFA functions also find the partial match for
8911 "dogsbody", and so return that when PCRE_PARTIAL_HARD is set.
8914 PARTIAL MATCHING AND WORD BOUNDARIES
8916 If a pattern ends with one of sequences \b or \B, which test for word
8917 boundaries, partial matching with PCRE_PARTIAL_SOFT can give counter-
8918 intuitive results. Consider this pattern:
8922 This matches "cat", provided there is a word boundary at either end. If
8923 the subject string is "the cat", the comparison of the final "t" with a
8924 following character cannot take place, so a partial match is found.
8925 However, normal matching carries on, and \b matches at the end of the
8926 subject when the last character is a letter, so a complete match is
8927 found. The result, therefore, is not PCRE_ERROR_PARTIAL. Using
8928 PCRE_PARTIAL_HARD in this case does yield PCRE_ERROR_PARTIAL, because
8929 then the partial match takes precedence.
8932 FORMERLY RESTRICTED PATTERNS
8934 For releases of PCRE prior to 8.00, because of the way certain internal
8935 optimizations were implemented in the pcre_exec() function, the
8936 PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be
8937 used with all patterns. From release 8.00 onwards, the restrictions no
8938 longer apply, and partial matching with can be requested for any pat-
8941 Items that were formerly restricted were repeated single characters and
8942 repeated metasequences. If PCRE_PARTIAL was set for a pattern that did
8943 not conform to the restrictions, pcre_exec() returned the error code
8944 PCRE_ERROR_BADPARTIAL (-13). This error code is no longer in use. The
8945 PCRE_INFO_OKPARTIAL call to pcre_fullinfo() to find out if a compiled
8946 pattern can be used for partial matching now always returns 1.
8949 EXAMPLE OF PARTIAL MATCHING USING PCRETEST
8951 If the escape sequence \P is present in a pcretest data line, the
8952 PCRE_PARTIAL_SOFT option is used for the match. Here is a run of
8953 pcretest that uses the date example quoted above:
8955 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
8960 Partial match: 23dec3
8968 The first data string is matched completely, so pcretest shows the
8969 matched substrings. The remaining four strings do not match the com-
8970 plete pattern, but the first two are partial matches. Similar output is
8971 obtained if DFA matching is used.
8973 If the escape sequence \P is present more than once in a pcretest data
8974 line, the PCRE_PARTIAL_HARD option is set for the match.
8977 MULTI-SEGMENT MATCHING WITH pcre_dfa_exec() OR pcre[16|32]_dfa_exec()
8979 When a partial match has been found using a DFA matching function, it
8980 is possible to continue the match by providing additional subject data
8981 and calling the function again with the same compiled regular expres-
8982 sion, this time setting the PCRE_DFA_RESTART option. You must pass the
8983 same working space as before, because this is where details of the pre-
8984 vious partial match are stored. Here is an example using pcretest,
8985 using the \R escape sequence to set the PCRE_DFA_RESTART option (\D
8986 specifies the use of the DFA matching function):
8988 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
8994 The first call has "23ja" as the subject, and requests partial match-
8995 ing; the second call has "n05" as the subject for the continued
8996 (restarted) match. Notice that when the match is complete, only the
8997 last part is shown; PCRE does not retain the previously partially-
8998 matched string. It is up to the calling program to do that if it needs
9001 That means that, for an unanchored pattern, if a continued match fails,
9002 it is not possible to try again at a new starting point. All this
9003 facility is capable of doing is continuing with the previous match
9004 attempt. In the previous example, if the second set of data is "ug23"
9005 the result is no match, even though there would be a match for "aug23"
9006 if the entire string were given at once. Depending on the application,
9007 this may or may not be what you want. The only way to allow for start-
9008 ing again at the next character is to retain the matched part of the
9009 subject and try a new complete match.
9011 You can set the PCRE_PARTIAL_SOFT or PCRE_PARTIAL_HARD options with
9012 PCRE_DFA_RESTART to continue partial matching over multiple segments.
9013 This facility can be used to pass very long subject strings to the DFA
9017 MULTI-SEGMENT MATCHING WITH pcre_exec() OR pcre[16|32]_exec()
9019 From release 8.00, the standard matching functions can also be used to
9020 do multi-segment matching. Unlike the DFA functions, it is not possible
9021 to restart the previous match with a new segment of data. Instead, new
9022 data must be added to the previous subject string, and the entire match
9023 re-run, starting from the point where the partial match occurred. Ear-
9024 lier data can be discarded.
9026 It is best to use PCRE_PARTIAL_HARD in this situation, because it does
9027 not treat the end of a segment as the end of the subject when matching
9028 \z, \Z, \b, \B, and $. Consider an unanchored pattern that matches
9031 re> /\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d/
9032 data> The date is 23ja\P\P
9035 At this stage, an application could discard the text preceding "23ja",
9036 add on text from the next segment, and call the matching function
9037 again. Unlike the DFA matching functions, the entire matching string
9038 must always be available, and the complete matching process occurs for
9039 each call, so more memory and more processing time is needed.
9041 Note: If the pattern contains lookbehind assertions, or \K, or starts
9042 with \b or \B, the string that is returned for a partial match includes
9043 characters that precede the start of what would be returned for a com-
9044 plete match, because it contains all the characters that were inspected
9045 during the partial match.
9048 ISSUES WITH MULTI-SEGMENT MATCHING
9050 Certain types of pattern may give problems with multi-segment matching,
9051 whichever matching function is used.
9053 1. If the pattern contains a test for the beginning of a line, you need
9054 to pass the PCRE_NOTBOL option when the subject string for any call
9055 does start at the beginning of a line. There is also a PCRE_NOTEOL
9056 option, but in practice when doing multi-segment matching you should be
9057 using PCRE_PARTIAL_HARD, which includes the effect of PCRE_NOTEOL.
9059 2. Lookbehind assertions that have already been obeyed are catered for
9060 in the offsets that are returned for a partial match. However a lookbe-
9061 hind assertion later in the pattern could require even earlier charac-
9062 ters to be inspected. You can handle this case by using the
9063 PCRE_INFO_MAXLOOKBEHIND option of the pcre_fullinfo() or
9064 pcre[16|32]_fullinfo() functions to obtain the length of the longest
9065 lookbehind in the pattern. This length is given in characters, not
9066 bytes. If you always retain at least that many characters before the
9067 partially matched string, all should be well. (Of course, near the
9068 start of the subject, fewer characters may be present; in that case all
9069 characters should be retained.)
9071 From release 8.33, there is a more accurate way of deciding which char-
9072 acters to retain. Instead of subtracting the length of the longest
9073 lookbehind from the earliest inspected character (offsets[0]), the
9074 match start position (offsets[2]) should be used, and the next match
9075 attempt started at the offsets[2] character by setting the startoffset
9076 argument of pcre_exec() or pcre_dfa_exec().
9078 For example, if the pattern "(?<=123)abc" is partially matched against
9079 the string "xx123a", the three offset values returned are 2, 6, and 5.
9080 This indicates that the matching process that gave a partial match
9081 started at offset 5, but the characters "123a" were all inspected. The
9082 maximum lookbehind for that pattern is 3, so taking that away from 5
9083 shows that we need only keep "123a", and the next match attempt can be
9084 started at offset 3 (that is, at "a") when further characters have been
9085 added. When the match start is not the earliest inspected character,
9086 pcretest shows it explicitly:
9090 Partial match at offset 5: 123a
9092 3. Because a partial match must always contain at least one character,
9093 what might be considered a partial match of an empty string actually
9094 gives a "no match" result. For example:
9100 If the next segment begins "cx", a match should be found, but this will
9101 only happen if characters from the previous segment are retained. For
9102 this reason, a "no match" result should be interpreted as "partial
9103 match of an empty string" when the pattern contains lookbehinds.
9105 4. Matching a subject string that is split into multiple segments may
9106 not always produce exactly the same result as matching over one single
9107 long string, especially when PCRE_PARTIAL_SOFT is used. The section
9108 "Partial Matching and Word Boundaries" above describes an issue that
9109 arises if the pattern ends with \b or \B. Another kind of difference
9110 may occur when there are multiple matching possibilities, because (for
9111 PCRE_PARTIAL_SOFT) a partial match result is given only when there are
9112 no completed matches. This means that as soon as the shortest match has
9113 been found, continuation to a new subject segment is no longer possi-
9114 ble. Consider again this pcretest example:
9127 The first data line passes the string "dogsb" to a standard matching
9128 function, setting the PCRE_PARTIAL_SOFT option. Although the string is
9129 a partial match for "dogsbody", the result is not PCRE_ERROR_PARTIAL,
9130 because the shorter string "dog" is a complete match. Similarly, when
9131 the subject is presented to a DFA matching function in several parts
9132 ("do" and "gsb" being the first two) the match stops when "dog" has
9133 been found, and it is not possible to continue. On the other hand, if
9134 "dogsbody" is presented as a single string, a DFA matching function
9137 Because of these problems, it is best to use PCRE_PARTIAL_HARD when
9138 matching multi-segment data. The example above then behaves differ-
9143 Partial match: dogsb
9149 5. Patterns that contain alternatives at the top level which do not all
9150 start with the same pattern item may not work as expected when
9151 PCRE_DFA_RESTART is used. For example, consider this pattern:
9155 If the first part of the subject is "ABC123", a partial match of the
9156 first alternative is found at offset 3. There is no partial match for
9157 the second alternative, because such a match does not start at the same
9158 point in the subject string. Attempting to continue with the string
9159 "7890" does not yield a match because only those alternatives that
9160 match at one point in the subject are remembered. The problem arises
9161 because the start of the second alternative matches within the first
9162 alternative. There is no problem with anchored patterns or patterns
9167 where no string can be a partial match for both alternatives. This is
9168 not a problem if a standard matching function is used, because the
9169 entire match has to be rerun each time:
9177 Of course, instead of using PCRE_DFA_RESTART, the same technique of re-
9178 running the entire match can also be used with the DFA matching func-
9179 tions. Another possibility is to work with two buffers. If a partial
9180 match at offset n in the first buffer is followed by "no match" when
9181 PCRE_DFA_RESTART is used on the second buffer, you can then try a new
9182 match starting at offset n+1 in the first buffer.
9188 University Computing Service
9189 Cambridge CB2 3QH, England.
9194 Last updated: 02 July 2013
9195 Copyright (c) 1997-2013 University of Cambridge.
9196 ------------------------------------------------------------------------------
9199 PCREPRECOMPILE(3) Library Functions Manual PCREPRECOMPILE(3)
9204 PCRE - Perl-compatible regular expressions
9206 SAVING AND RE-USING PRECOMPILED PCRE PATTERNS
9208 If you are running an application that uses a large number of regular
9209 expression patterns, it may be useful to store them in a precompiled
9210 form instead of having to compile them every time the application is
9211 run. If you are not using any private character tables (see the
9212 pcre_maketables() documentation), this is relatively straightforward.
9213 If you are using private tables, it is a little bit more complicated.
9214 However, if you are using the just-in-time optimization feature, it is
9215 not possible to save and reload the JIT data.
9217 If you save compiled patterns to a file, you can copy them to a differ-
9218 ent host and run them there. If the two hosts have different endianness
9219 (byte order), you should run the pcre[16|32]_pat-
9220 tern_to_host_byte_order() function on the new host before trying to
9221 match the pattern. The matching functions return PCRE_ERROR_BADENDIAN-
9222 NESS if they detect a pattern with the wrong endianness.
9224 Compiling regular expressions with one version of PCRE for use with a
9225 different version is not guaranteed to work and may cause crashes, and
9226 saving and restoring a compiled pattern loses any JIT optimization
9230 SAVING A COMPILED PATTERN
9232 The value returned by pcre[16|32]_compile() points to a single block of
9233 memory that holds the compiled pattern and associated data. You can
9234 find the length of this block in bytes by calling
9235 pcre[16|32]_fullinfo() with an argument of PCRE_INFO_SIZE. You can then
9236 save the data in any appropriate manner. Here is sample code for the
9237 8-bit library that compiles a pattern and writes it to a file. It
9238 assumes that the variable fd refers to a file that is open for output:
9240 int erroroffset, rc, size;
9244 re = pcre_compile("my pattern", 0, &error, &erroroffset, NULL);
9245 if (re == NULL) { ... handle errors ... }
9246 rc = pcre_fullinfo(re, NULL, PCRE_INFO_SIZE, &size);
9247 if (rc < 0) { ... handle errors ... }
9248 rc = fwrite(re, 1, size, fd);
9249 if (rc != size) { ... handle errors ... }
9251 In this example, the bytes that comprise the compiled pattern are
9252 copied exactly. Note that this is binary data that may contain any of
9253 the 256 possible byte values. On systems that make a distinction
9254 between binary and non-binary data, be sure that the file is opened for
9257 If you want to write more than one pattern to a file, you will have to
9258 devise a way of separating them. For binary data, preceding each pat-
9259 tern with its length is probably the most straightforward approach.
9260 Another possibility is to write out the data in hexadecimal instead of
9261 binary, one pattern to a line.
9263 Saving compiled patterns in a file is only one possible way of storing
9264 them for later use. They could equally well be saved in a database, or
9265 in the memory of some daemon process that passes them via sockets to
9266 the processes that want them.
9268 If the pattern has been studied, it is also possible to save the normal
9269 study data in a similar way to the compiled pattern itself. However, if
9270 the PCRE_STUDY_JIT_COMPILE was used, the just-in-time data that is cre-
9271 ated cannot be saved because it is too dependent on the current envi-
9272 ronment. When studying generates additional information,
9273 pcre[16|32]_study() returns a pointer to a pcre[16|32]_extra data
9274 block. Its format is defined in the section on matching a pattern in
9275 the pcreapi documentation. The study_data field points to the binary
9276 study data, and this is what you must save (not the pcre[16|32]_extra
9277 block itself). The length of the study data can be obtained by calling
9278 pcre[16|32]_fullinfo() with an argument of PCRE_INFO_STUDYSIZE. Remem-
9279 ber to check that pcre[16|32]_study() did return a non-NULL value
9280 before trying to save the study data.
9283 RE-USING A PRECOMPILED PATTERN
9285 Re-using a precompiled pattern is straightforward. Having reloaded it
9286 into main memory, called pcre[16|32]_pattern_to_host_byte_order() if
9287 necessary, you pass its pointer to pcre[16|32]_exec() or
9288 pcre[16|32]_dfa_exec() in the usual way.
9290 However, if you passed a pointer to custom character tables when the
9291 pattern was compiled (the tableptr argument of pcre[16|32]_compile()),
9292 you must now pass a similar pointer to pcre[16|32]_exec() or
9293 pcre[16|32]_dfa_exec(), because the value saved with the compiled pat-
9294 tern will obviously be nonsense. A field in a pcre[16|32]_extra() block
9295 is used to pass this data, as described in the section on matching a
9296 pattern in the pcreapi documentation.
9298 Warning: The tables that pcre_exec() and pcre_dfa_exec() use must be
9299 the same as those that were used when the pattern was compiled. If this
9300 is not the case, the behaviour is undefined.
9302 If you did not provide custom character tables when the pattern was
9303 compiled, the pointer in the compiled pattern is NULL, which causes the
9304 matching functions to use PCRE's internal tables. Thus, you do not need
9305 to take any special action at run time in this case.
9307 If you saved study data with the compiled pattern, you need to create
9308 your own pcre[16|32]_extra data block and set the study_data field to
9309 point to the reloaded study data. You must also set the
9310 PCRE_EXTRA_STUDY_DATA bit in the flags field to indicate that study
9311 data is present. Then pass the pcre[16|32]_extra block to the matching
9312 function in the usual way. If the pattern was studied for just-in-time
9313 optimization, that data cannot be saved, and so is lost by a
9317 COMPATIBILITY WITH DIFFERENT PCRE RELEASES
9319 In general, it is safest to recompile all saved patterns when you
9320 update to a new PCRE release, though not all updates actually require
9327 University Computing Service
9328 Cambridge CB2 3QH, England.
9333 Last updated: 12 November 2013
9334 Copyright (c) 1997-2013 University of Cambridge.
9335 ------------------------------------------------------------------------------
9338 PCREPERFORM(3) Library Functions Manual PCREPERFORM(3)
9343 PCRE - Perl-compatible regular expressions
9347 Two aspects of performance are discussed below: memory usage and pro-
9348 cessing time. The way you express your pattern as a regular expression
9349 can affect both of them.
9352 COMPILED PATTERN MEMORY USAGE
9354 Patterns are compiled by PCRE into a reasonably efficient interpretive
9355 code, so that most simple patterns do not use much memory. However,
9356 there is one case where the memory usage of a compiled pattern can be
9357 unexpectedly large. If a parenthesized subpattern has a quantifier with
9358 a minimum greater than 1 and/or a limited maximum, the whole subpattern
9359 is repeated in the compiled code. For example, the pattern
9363 is compiled as if it were
9365 (abc|def)(abc|def)((abc|def)(abc|def)?)?
9367 (Technical aside: It is done this way so that backtrack points within
9368 each of the repetitions can be independently maintained.)
9370 For regular expressions whose quantifiers use only small numbers, this
9371 is not usually a problem. However, if the numbers are large, and par-
9372 ticularly if such repetitions are nested, the memory usage can become
9373 an embarrassment. For example, the very simple pattern
9375 ((ab){1,1000}c){1,3}
9377 uses 51K bytes when compiled using the 8-bit library. When PCRE is com-
9378 piled with its default internal pointer size of two bytes, the size
9379 limit on a compiled pattern is 64K data units, and this is reached with
9380 the above pattern if the outer repetition is increased from 3 to 4.
9381 PCRE can be compiled to use larger internal pointers and thus handle
9382 larger compiled patterns, but it is better to try to rewrite your pat-
9383 tern to use less memory if you can.
9385 One way of reducing the memory usage for such patterns is to make use
9386 of PCRE's "subroutine" facility. Re-writing the above pattern as
9388 ((ab)(?2){0,999}c)(?1){0,2}
9390 reduces the memory requirements to 18K, and indeed it remains under 20K
9391 even with the outer repetition increased to 100. However, this pattern
9392 is not exactly equivalent, because the "subroutine" calls are treated
9393 as atomic groups into which there can be no backtracking if there is a
9394 subsequent matching failure. Therefore, PCRE cannot do this kind of
9395 rewriting automatically. Furthermore, there is a noticeable loss of
9396 speed when executing the modified pattern. Nevertheless, if the atomic
9397 grouping is not a problem and the loss of speed is acceptable, this
9398 kind of rewriting will allow you to process patterns that PCRE cannot
9402 STACK USAGE AT RUN TIME
9404 When pcre_exec() or pcre[16|32]_exec() is used for matching, certain
9405 kinds of pattern can cause it to use large amounts of the process
9406 stack. In some environments the default process stack is quite small,
9407 and if it runs out the result is often SIGSEGV. This issue is probably
9408 the most frequently raised problem with PCRE. Rewriting your pattern
9409 can often help. The pcrestack documentation discusses this issue in
9415 Certain items in regular expression patterns are processed more effi-
9416 ciently than others. It is more efficient to use a character class like
9417 [aeiou] than a set of single-character alternatives such as
9418 (a|e|i|o|u). In general, the simplest construction that provides the
9419 required behaviour is usually the most efficient. Jeffrey Friedl's book
9420 contains a lot of useful general discussion about optimizing regular
9421 expressions for efficient performance. This document contains a few
9422 observations about PCRE.
9424 Using Unicode character properties (the \p, \P, and \X escapes) is
9425 slow, because PCRE has to use a multi-stage table lookup whenever it
9426 needs a character's property. If you can find an alternative pattern
9427 that does not use character properties, it will probably be faster.
9429 By default, the escape sequences \b, \d, \s, and \w, and the POSIX
9430 character classes such as [:alpha:] do not use Unicode properties,
9431 partly for backwards compatibility, and partly for performance reasons.
9432 However, you can set PCRE_UCP if you want Unicode character properties
9433 to be used. This can double the matching time for items such as \d,
9434 when matched with a traditional matching function; the performance loss
9435 is less with a DFA matching function, and in both cases there is not
9436 much difference for \b.
9438 When a pattern begins with .* not in parentheses, or in parentheses
9439 that are not the subject of a backreference, and the PCRE_DOTALL option
9440 is set, the pattern is implicitly anchored by PCRE, since it can match
9441 only at the start of a subject string. However, if PCRE_DOTALL is not
9442 set, PCRE cannot make this optimization, because the . metacharacter
9443 does not then match a newline, and if the subject string contains new-
9444 lines, the pattern may match from the character immediately following
9445 one of them instead of from the very start. For example, the pattern
9449 matches the subject "first\nand second" (where \n stands for a newline
9450 character), with the match starting at the seventh character. In order
9451 to do this, PCRE has to retry the match starting after every newline in
9454 If you are using such a pattern with subject strings that do not con-
9455 tain newlines, the best performance is obtained by setting PCRE_DOTALL,
9456 or starting the pattern with ^.* or ^.*? to indicate explicit anchor-
9457 ing. That saves PCRE from having to scan along the subject looking for
9458 a newline to restart at.
9460 Beware of patterns that contain nested indefinite repeats. These can
9461 take a long time to run when applied to a string that does not match.
9462 Consider the pattern fragment
9466 This can match "aaaa" in 16 different ways, and this number increases
9467 very rapidly as the string gets longer. (The * repeat can match 0, 1,
9468 2, 3, or 4 times, and for each of those cases other than 0 or 4, the +
9469 repeats can match different numbers of times.) When the remainder of
9470 the pattern is such that the entire match is going to fail, PCRE has in
9471 principle to try every possible variation, and this can take an
9472 extremely long time, even for relatively short strings.
9474 An optimization catches some of the more simple cases such as
9478 where a literal character follows. Before embarking on the standard
9479 matching procedure, PCRE checks that there is a "b" later in the sub-
9480 ject string, and if there is not, it fails the match immediately. How-
9481 ever, when there is no following literal this optimization cannot be
9482 used. You can see the difference by comparing the behaviour of
9486 with the pattern above. The former gives a failure almost instantly
9487 when applied to a whole line of "a" characters, whereas the latter
9488 takes an appreciable time with strings longer than about 20 characters.
9490 In many cases, the solution to this kind of performance issue is to use
9491 an atomic group or a possessive quantifier.
9497 University Computing Service
9498 Cambridge CB2 3QH, England.
9503 Last updated: 25 August 2012
9504 Copyright (c) 1997-2012 University of Cambridge.
9505 ------------------------------------------------------------------------------
9508 PCREPOSIX(3) Library Functions Manual PCREPOSIX(3)
9513 PCRE - Perl-compatible regular expressions.
9517 #include <pcreposix.h>
9519 int regcomp(regex_t *preg, const char *pattern,
9522 int regexec(regex_t *preg, const char *string,
9523 size_t nmatch, regmatch_t pmatch[], int eflags);
9524 size_t regerror(int errcode, const regex_t *preg,
9525 char *errbuf, size_t errbuf_size);
9527 void regfree(regex_t *preg);
9532 This set of functions provides a POSIX-style API for the PCRE regular
9533 expression 8-bit library. See the pcreapi documentation for a descrip-
9534 tion of PCRE's native API, which contains much additional functional-
9535 ity. There is no POSIX-style wrapper for PCRE's 16-bit and 32-bit
9538 The functions described here are just wrapper functions that ultimately
9539 call the PCRE native API. Their prototypes are defined in the
9540 pcreposix.h header file, and on Unix systems the library itself is
9541 called pcreposix.a, so can be accessed by adding -lpcreposix to the
9542 command for linking an application that uses them. Because the POSIX
9543 functions call the native ones, it is also necessary to add -lpcre.
9545 I have implemented only those POSIX option bits that can be reasonably
9546 mapped to PCRE native options. In addition, the option REG_EXTENDED is
9547 defined with the value zero. This has no effect, but since programs
9548 that are written to the POSIX interface often use it, this makes it
9549 easier to slot in PCRE as a replacement library. Other POSIX options
9550 are not even defined.
9552 There are also some other options that are not defined by POSIX. These
9553 have been added at the request of users who want to make use of certain
9554 PCRE-specific features via the POSIX calling interface.
9556 When PCRE is called via these functions, it is only the API that is
9557 POSIX-like in style. The syntax and semantics of the regular expres-
9558 sions themselves are still those of Perl, subject to the setting of
9559 various PCRE options, as described below. "POSIX-like in style" means
9560 that the API approximates to the POSIX definition; it is not fully
9561 POSIX-compatible, and in multi-byte encoding domains it is probably
9562 even less compatible.
9564 The header for these functions is supplied as pcreposix.h to avoid any
9565 potential clash with other POSIX libraries. It can, of course, be
9566 renamed or aliased as regex.h, which is the "correct" name. It provides
9567 two structure types, regex_t for compiled internal forms, and reg-
9568 match_t for returning captured substrings. It also defines some con-
9569 stants whose names start with "REG_"; these are used for setting
9570 options and identifying error codes.
9575 The function regcomp() is called to compile a pattern into an internal
9576 form. The pattern is a C string terminated by a binary zero, and is
9577 passed in the argument pattern. The preg argument is a pointer to a
9578 regex_t structure that is used as a base for storing information about
9579 the compiled regular expression.
9581 The argument cflags is either zero, or contains one or more of the bits
9582 defined by the following macros:
9586 The PCRE_DOTALL option is set when the regular expression is passed for
9587 compilation to the native function. Note that REG_DOTALL is not part of
9592 The PCRE_CASELESS option is set when the regular expression is passed
9593 for compilation to the native function.
9597 The PCRE_MULTILINE option is set when the regular expression is passed
9598 for compilation to the native function. Note that this does not mimic
9599 the defined POSIX behaviour for REG_NEWLINE (see the following sec-
9604 The PCRE_NO_AUTO_CAPTURE option is set when the regular expression is
9605 passed for compilation to the native function. In addition, when a pat-
9606 tern that is compiled with this flag is passed to regexec() for match-
9607 ing, the nmatch and pmatch arguments are ignored, and no captured
9608 strings are returned.
9612 The PCRE_UCP option is set when the regular expression is passed for
9613 compilation to the native function. This causes PCRE to use Unicode
9614 properties when matchine \d, \w, etc., instead of just recognizing
9615 ASCII values. Note that REG_UTF8 is not part of the POSIX standard.
9619 The PCRE_UNGREEDY option is set when the regular expression is passed
9620 for compilation to the native function. Note that REG_UNGREEDY is not
9621 part of the POSIX standard.
9625 The PCRE_UTF8 option is set when the regular expression is passed for
9626 compilation to the native function. This causes the pattern itself and
9627 all data strings used for matching it to be treated as UTF-8 strings.
9628 Note that REG_UTF8 is not part of the POSIX standard.
9630 In the absence of these flags, no options are passed to the native
9631 function. This means the the regex is compiled with PCRE default
9632 semantics. In particular, the way it handles newline characters in the
9633 subject string is the Perl way, not the POSIX way. Note that setting
9634 PCRE_MULTILINE has only some of the effects specified for REG_NEWLINE.
9635 It does not affect the way newlines are matched by . (they are not) or
9636 by a negative class such as [^a] (they are).
9638 The yield of regcomp() is zero on success, and non-zero otherwise. The
9639 preg structure is filled in on success, and one member of the structure
9640 is public: re_nsub contains the number of capturing subpatterns in the
9641 regular expression. Various error codes are defined in the header file.
9643 NOTE: If the yield of regcomp() is non-zero, you must not attempt to
9644 use the contents of the preg structure. If, for example, you pass it to
9645 regexec(), the result is undefined and your program is likely to crash.
9648 MATCHING NEWLINE CHARACTERS
9650 This area is not simple, because POSIX and Perl take different views of
9651 things. It is not possible to get PCRE to obey POSIX semantics, but
9652 then PCRE was never intended to be a POSIX engine. The following table
9653 lists the different possibilities for matching newline characters in
9658 . matches newline no PCRE_DOTALL
9659 newline matches [^a] yes not changeable
9660 $ matches \n at end yes PCRE_DOLLARENDONLY
9661 $ matches \n in middle no PCRE_MULTILINE
9662 ^ matches \n in middle no PCRE_MULTILINE
9664 This is the equivalent table for POSIX:
9668 . matches newline yes REG_NEWLINE
9669 newline matches [^a] yes REG_NEWLINE
9670 $ matches \n at end no REG_NEWLINE
9671 $ matches \n in middle no REG_NEWLINE
9672 ^ matches \n in middle no REG_NEWLINE
9674 PCRE's behaviour is the same as Perl's, except that there is no equiva-
9675 lent for PCRE_DOLLAR_ENDONLY in Perl. In both PCRE and Perl, there is
9676 no way to stop newline from matching [^a].
9678 The default POSIX newline handling can be obtained by setting
9679 PCRE_DOTALL and PCRE_DOLLAR_ENDONLY, but there is no way to make PCRE
9680 behave exactly as for the REG_NEWLINE action.
9685 The function regexec() is called to match a compiled pattern preg
9686 against a given string, which is by default terminated by a zero byte
9687 (but see REG_STARTEND below), subject to the options in eflags. These
9692 The PCRE_NOTBOL option is set when calling the underlying PCRE matching
9697 The PCRE_NOTEMPTY option is set when calling the underlying PCRE match-
9698 ing function. Note that REG_NOTEMPTY is not part of the POSIX standard.
9699 However, setting this option can give more POSIX-like behaviour in some
9704 The PCRE_NOTEOL option is set when calling the underlying PCRE matching
9709 The string is considered to start at string + pmatch[0].rm_so and to
9710 have a terminating NUL located at string + pmatch[0].rm_eo (there need
9711 not actually be a NUL at that location), regardless of the value of
9712 nmatch. This is a BSD extension, compatible with but not specified by
9713 IEEE Standard 1003.2 (POSIX.2), and should be used with caution in
9714 software intended to be portable to other systems. Note that a non-zero
9715 rm_so does not imply REG_NOTBOL; REG_STARTEND affects only the location
9716 of the string, not how it is matched.
9718 If the pattern was compiled with the REG_NOSUB flag, no data about any
9719 matched strings is returned. The nmatch and pmatch arguments of
9720 regexec() are ignored.
9722 If the value of nmatch is zero, or if the value pmatch is NULL, no data
9723 about any matched strings is returned.
9725 Otherwise,the portion of the string that was matched, and also any cap-
9726 tured substrings, are returned via the pmatch argument, which points to
9727 an array of nmatch structures of type regmatch_t, containing the mem-
9728 bers rm_so and rm_eo. These contain the offset to the first character
9729 of each substring and the offset to the first character after the end
9730 of each substring, respectively. The 0th element of the vector relates
9731 to the entire portion of string that was matched; subsequent elements
9732 relate to the capturing subpatterns of the regular expression. Unused
9733 entries in the array have both structure members set to -1.
9735 A successful match yields a zero return; various error codes are
9736 defined in the header file, of which REG_NOMATCH is the "expected"
9742 The regerror() function maps a non-zero errorcode from either regcomp()
9743 or regexec() to a printable message. If preg is not NULL, the error
9744 should have arisen from the use of that structure. A message terminated
9745 by a binary zero is placed in errbuf. The length of the message,
9746 including the zero, is limited to errbuf_size. The yield of the func-
9747 tion is the size of buffer needed to hold the whole message.
9752 Compiling a regular expression causes memory to be allocated and asso-
9753 ciated with the preg structure. The function regfree() frees all such
9754 memory, after which preg may no longer be used as a compiled expres-
9761 University Computing Service
9762 Cambridge CB2 3QH, England.
9767 Last updated: 09 January 2012
9768 Copyright (c) 1997-2012 University of Cambridge.
9769 ------------------------------------------------------------------------------
9772 PCRECPP(3) Library Functions Manual PCRECPP(3)
9777 PCRE - Perl-compatible regular expressions.
9779 SYNOPSIS OF C++ WRAPPER
9781 #include <pcrecpp.h>
9786 The C++ wrapper for PCRE was provided by Google Inc. Some additional
9787 functionality was added by Giuseppe Maxia. This brief man page was con-
9788 structed from the notes in the pcrecpp.h file, which should be con-
9789 sulted for further details. Note that the C++ wrapper supports only the
9790 original 8-bit PCRE library. There is no 16-bit or 32-bit support at
9796 The "FullMatch" operation checks that supplied text matches a supplied
9797 pattern exactly. If pointer arguments are supplied, it copies matched
9798 sub-strings that match sub-patterns into them.
9800 Example: successful match
9801 pcrecpp::RE re("h.*o");
9802 re.FullMatch("hello");
9804 Example: unsuccessful match (requires full match):
9805 pcrecpp::RE re("e");
9806 !re.FullMatch("hello");
9808 Example: creating a temporary RE object:
9809 pcrecpp::RE("h.*o").FullMatch("hello");
9811 You can pass in a "const char*" or a "string" for "text". The examples
9812 below tend to use a const char*. You can, as in the different examples
9813 above, store the RE object explicitly in a variable or use a temporary
9814 RE object. The examples below use one mode or the other arbitrarily.
9815 Either could correctly be used for any of these examples.
9817 You must supply extra pointer arguments to extract matched subpieces.
9819 Example: extracts "ruby" into "s" and 1234 into "i"
9822 pcrecpp::RE re("(\\w+):(\\d+)");
9823 re.FullMatch("ruby:1234", &s, &i);
9825 Example: does not try to extract any extra sub-patterns
9826 re.FullMatch("ruby:1234", &s);
9828 Example: does not try to extract into NULL
9829 re.FullMatch("ruby:1234", NULL, &i);
9831 Example: integer overflow causes failure
9832 !re.FullMatch("ruby:1234567891234", NULL, &i);
9834 Example: fails because there aren't enough sub-patterns:
9835 !pcrecpp::RE("\\w+:\\d+").FullMatch("ruby:1234", &s);
9837 Example: fails because string cannot be stored in integer
9838 !pcrecpp::RE("(.*)").FullMatch("ruby", &i);
9840 The provided pointer arguments can be pointers to any scalar numeric
9843 string (matched piece is copied to string)
9844 StringPiece (StringPiece is mutated to point to matched piece)
9845 T (where "bool T::ParseFrom(const char*, int)" exists)
9846 NULL (the corresponding matched sub-pattern is not copied)
9848 The function returns true iff all of the following conditions are sat-
9851 a. "text" matches "pattern" exactly;
9853 b. The number of matched sub-patterns is >= number of supplied
9856 c. The "i"th argument has a suitable type for holding the
9857 string captured as the "i"th sub-pattern. If you pass in
9858 void * NULL for the "i"th argument, or a non-void * NULL
9859 of the correct type, or pass fewer arguments than the
9860 number of sub-patterns, "i"th captured sub-pattern is
9863 CAVEAT: An optional sub-pattern that does not exist in the matched
9864 string is assigned the empty string. Therefore, the following will
9865 return false (because the empty string is not a valid number):
9868 pcrecpp::RE::FullMatch("abc", "[a-z]+(\\d+)?", &number);
9870 The matching interface supports at most 16 arguments per call. If you
9871 need more, consider using the more general interface
9872 pcrecpp::RE::DoMatch. See pcrecpp.h for the signature for DoMatch.
9874 NOTE: Do not use no_arg, which is used internally to mark the end of a
9875 list of optional arguments, as a placeholder for missing arguments, as
9876 this can lead to segfaults.
9879 QUOTING METACHARACTERS
9881 You can use the "QuoteMeta" operation to insert backslashes before all
9882 potentially meaningful characters in a string. The returned string,
9883 used as a regular expression, will exactly match the original string.
9886 string quoted = RE::QuoteMeta(unquoted);
9888 Note that it's legal to escape a character even if it has no special
9889 meaning in a regular expression -- so this function does that. (This
9890 also makes it identical to the perl function of the same name; see
9891 "perldoc -f quotemeta".) For example, "1.5-2.0?" becomes
9897 You can use the "PartialMatch" operation when you want the pattern to
9898 match any substring of the text.
9900 Example: simple search for a string:
9901 pcrecpp::RE("ell").PartialMatch("hello");
9903 Example: find first number in a string:
9905 pcrecpp::RE re("(\\d+)");
9906 re.PartialMatch("x*100 + 20", &number);
9907 assert(number == 100);
9910 UTF-8 AND THE MATCHING INTERFACE
9912 By default, pattern and text are plain text, one byte per character.
9913 The UTF8 flag, passed to the constructor, causes both pattern and
9914 string to be treated as UTF-8 text, still a byte stream but potentially
9915 multiple bytes per character. In practice, the text is likelier to be
9916 UTF-8 than the pattern, but the match returned may depend on the UTF8
9917 flag, so always use it when matching UTF8 text. For example, "." will
9918 match one byte normally but with UTF8 set may match up to three bytes
9919 of a multi-byte character.
9922 pcrecpp::RE_Options options;
9924 pcrecpp::RE re(utf8_pattern, options);
9925 re.FullMatch(utf8_string);
9927 Example: using the convenience function UTF8():
9928 pcrecpp::RE re(utf8_pattern, pcrecpp::UTF8());
9929 re.FullMatch(utf8_string);
9931 NOTE: The UTF8 flag is ignored if pcre was not configured with the
9935 PASSING MODIFIERS TO THE REGULAR EXPRESSION ENGINE
9937 PCRE defines some modifiers to change the behavior of the regular
9938 expression engine. The C++ wrapper defines an auxiliary class,
9939 RE_Options, as a vehicle to pass such modifiers to a RE class. Cur-
9940 rently, the following modifiers are supported:
9942 modifier description Perl corresponding
9944 PCRE_CASELESS case insensitive match /i
9945 PCRE_MULTILINE multiple lines match /m
9946 PCRE_DOTALL dot matches newlines /s
9947 PCRE_DOLLAR_ENDONLY $ matches only at end N/A
9948 PCRE_EXTRA strict escape parsing N/A
9949 PCRE_EXTENDED ignore white spaces /x
9950 PCRE_UTF8 handles UTF8 chars built-in
9951 PCRE_UNGREEDY reverses * and *? N/A
9952 PCRE_NO_AUTO_CAPTURE disables capturing parens N/A (*)
9954 (*) Both Perl and PCRE allow non capturing parentheses by means of the
9955 "?:" modifier within the pattern itself. e.g. (?:ab|cd) does not cap-
9956 ture, while (ab|cd) does.
9958 For a full account on how each modifier works, please check the PCRE
9961 For each modifier, there are two member functions whose name is made
9962 out of the modifier in lowercase, without the "PCRE_" prefix. For
9963 instance, PCRE_CASELESS is handled by
9967 which returns true if the modifier is set, and
9969 RE_Options & set_caseless(bool)
9971 which sets or unsets the modifier. Moreover, PCRE_EXTRA_MATCH_LIMIT can
9972 be accessed through the set_match_limit() and match_limit() member
9973 functions. Setting match_limit to a non-zero value will limit the exe-
9974 cution of pcre to keep it from doing bad things like blowing the stack
9975 or taking an eternity to return a result. A value of 5000 is good
9976 enough to stop stack blowup in a 2MB thread stack. Setting match_limit
9977 to zero disables match limiting. Alternatively, you can call
9978 match_limit_recursion() which uses PCRE_EXTRA_MATCH_LIMIT_RECURSION to
9979 limit how much PCRE recurses. match_limit() limits the number of
9980 matches PCRE does; match_limit_recursion() limits the depth of internal
9981 recursion, and therefore the amount of stack that is used.
9983 Normally, to pass one or more modifiers to a RE class, you declare a
9984 RE_Options object, set the appropriate options, and pass this object to
9985 a RE constructor. Example:
9988 opt.set_caseless(true);
9989 if (RE("HELLO", opt).PartialMatch("hello world")) ...
9991 RE_options has two constructors. The default constructor takes no argu-
9992 ments and creates a set of flags that are off by default. The optional
9993 parameter option_flags is to facilitate transfer of legacy code from C
9994 programs. This lets you do
9997 RE_Options(PCRE_CASELESS|PCRE_MULTILINE)).PartialMatch(str);
9999 However, new code is better off doing
10002 RE_Options().set_caseless(true).set_multiline(true))
10003 .PartialMatch(str);
10005 If you are going to pass one of the most used modifiers, there are some
10006 convenience functions that return a RE_Options class with the appropri-
10007 ate modifier already set: CASELESS(), UTF8(), MULTILINE(), DOTALL(),
10010 If you need to set several options at once, and you don't want to go
10011 through the pains of declaring a RE_Options object and setting several
10012 options, there is a parallel method that give you such ability on the
10013 fly. You can concatenate several set_xxxxx() member functions, since
10014 each of them returns a reference to its class object. For example, to
10015 pass PCRE_CASELESS, PCRE_EXTENDED, and PCRE_MULTILINE to a RE with one
10016 statement, you may write:
10018 RE(" ^ xyz \\s+ .* blah$",
10020 .set_caseless(true)
10021 .set_extended(true)
10022 .set_multiline(true)).PartialMatch(sometext);
10025 SCANNING TEXT INCREMENTALLY
10027 The "Consume" operation may be useful if you want to repeatedly match
10028 regular expressions at the front of a string and skip over them as they
10029 match. This requires use of the "StringPiece" type, which represents a
10030 sub-range of a real string. Like RE, StringPiece is defined in the
10033 Example: read lines of the form "var = value" from a string.
10034 string contents = ...; // Fill string somehow
10035 pcrecpp::StringPiece input(contents); // Wrap in a StringPiece
10039 pcrecpp::RE re("(\\w+) = (\\d+)\n");
10040 while (re.Consume(&input, &var, &value)) {
10044 Each successful call to "Consume" will set "var/value", and also
10045 advance "input" so it points past the matched text.
10047 The "FindAndConsume" operation is similar to "Consume" but does not
10048 anchor your match at the beginning of the string. For example, you
10049 could extract all words from a string by repeatedly calling
10051 pcrecpp::RE("(\\w+)").FindAndConsume(&input, &word)
10054 PARSING HEX/OCTAL/C-RADIX NUMBERS
10056 By default, if you pass a pointer to a numeric value, the corresponding
10057 text is interpreted as a base-10 number. You can instead wrap the
10058 pointer with a call to one of the operators Hex(), Octal(), or CRadix()
10059 to interpret the text in another base. The CRadix operator interprets
10060 C-style "0" (base-8) and "0x" (base-16) prefixes, but defaults to
10065 pcrecpp::RE re("(.*) (.*) (.*) (.*)");
10066 re.FullMatch("100 40 0100 0x40",
10067 pcrecpp::Octal(&a), pcrecpp::Hex(&b),
10068 pcrecpp::CRadix(&c), pcrecpp::CRadix(&d));
10070 will leave 64 in a, b, c, and d.
10073 REPLACING PARTS OF STRINGS
10075 You can replace the first match of "pattern" in "str" with "rewrite".
10076 Within "rewrite", backslash-escaped digits (\1 to \9) can be used to
10077 insert text matching corresponding parenthesized group from the pat-
10078 tern. \0 in "rewrite" refers to the entire matching text. For example:
10080 string s = "yabba dabba doo";
10081 pcrecpp::RE("b+").Replace("d", &s);
10083 will leave "s" containing "yada dabba doo". The result is true if the
10084 pattern matches and a replacement occurs, false otherwise.
10086 GlobalReplace is like Replace except that it replaces all occurrences
10087 of the pattern in the string with the rewrite. Replacements are not
10088 subject to re-matching. For example:
10090 string s = "yabba dabba doo";
10091 pcrecpp::RE("b+").GlobalReplace("d", &s);
10093 will leave "s" containing "yada dada doo". It returns the number of
10096 Extract is like Replace, except that if the pattern matches, "rewrite"
10097 is copied into "out" (an additional argument) with substitutions. The
10098 non-matching portions of "text" are ignored. Returns true iff a match
10099 occurred and the extraction happened successfully; if no match occurs,
10100 the string is left unaffected.
10105 The C++ wrapper was contributed by Google Inc.
10106 Copyright (c) 2007 Google Inc.
10111 Last updated: 08 January 2012
10112 ------------------------------------------------------------------------------
10115 PCRESAMPLE(3) Library Functions Manual PCRESAMPLE(3)
10120 PCRE - Perl-compatible regular expressions
10122 PCRE SAMPLE PROGRAM
10124 A simple, complete demonstration program, to get you started with using
10125 PCRE, is supplied in the file pcredemo.c in the PCRE distribution. A
10126 listing of this program is given in the pcredemo documentation. If you
10127 do not have a copy of the PCRE distribution, you can save this listing
10128 to re-create pcredemo.c.
10130 The demonstration program, which uses the original PCRE 8-bit library,
10131 compiles the regular expression that is its first argument, and matches
10132 it against the subject string in its second argument. No PCRE options
10133 are set, and default character tables are used. If matching succeeds,
10134 the program outputs the portion of the subject that matched, together
10135 with the contents of any captured substrings.
10137 If the -g option is given on the command line, the program then goes on
10138 to check for further matches of the same regular expression in the same
10139 subject string. The logic is a little bit tricky because of the possi-
10140 bility of matching an empty string. Comments in the code explain what
10143 If PCRE is installed in the standard include and library directories
10144 for your operating system, you should be able to compile the demonstra-
10145 tion program using this command:
10147 gcc -o pcredemo pcredemo.c -lpcre
10149 If PCRE is installed elsewhere, you may need to add additional options
10150 to the command line. For example, on a Unix-like system that has PCRE
10151 installed in /usr/local, you can compile the demonstration program
10152 using a command like this:
10154 gcc -o pcredemo -I/usr/local/include pcredemo.c \
10155 -L/usr/local/lib -lpcre
10157 In a Windows environment, if you want to statically link the program
10158 against a non-dll pcre.a file, you must uncomment the line that defines
10159 PCRE_STATIC before including pcre.h, because otherwise the pcre_mal-
10160 loc() and pcre_free() exported functions will be declared
10161 __declspec(dllimport), with unwanted results.
10163 Once you have compiled and linked the demonstration program, you can
10164 run simple tests like this:
10166 ./pcredemo 'cat|dog' 'the cat sat on the mat'
10167 ./pcredemo -g 'cat|dog' 'the dog sat on the cat'
10169 Note that there is a much more comprehensive test program, called
10170 pcretest, which supports many more facilities for testing regular
10171 expressions and both PCRE libraries. The pcredemo program is provided
10172 as a simple coding example.
10174 If you try to run pcredemo when PCRE is not installed in the standard
10175 library directory, you may get an error like this on some operating
10176 systems (e.g. Solaris):
10178 ld.so.1: a.out: fatal: libpcre.so.0: open failed: No such file or
10181 This is caused by the way shared library support works on those sys-
10182 tems. You need to add
10186 (for example) to the compile command to get round this problem.
10192 University Computing Service
10193 Cambridge CB2 3QH, England.
10198 Last updated: 10 January 2012
10199 Copyright (c) 1997-2012 University of Cambridge.
10200 ------------------------------------------------------------------------------
10201 PCRELIMITS(3) Library Functions Manual PCRELIMITS(3)
10206 PCRE - Perl-compatible regular expressions
10208 SIZE AND OTHER LIMITATIONS
10210 There are some size limitations in PCRE but it is hoped that they will
10211 never in practice be relevant.
10213 The maximum length of a compiled pattern is approximately 64K data
10214 units (bytes for the 8-bit library, 16-bit units for the 16-bit
10215 library, and 32-bit units for the 32-bit library) if PCRE is compiled
10216 with the default internal linkage size, which is 2 bytes for the 8-bit
10217 and 16-bit libraries, and 4 bytes for the 32-bit library. If you want
10218 to process regular expressions that are truly enormous, you can compile
10219 PCRE with an internal linkage size of 3 or 4 (when building the 16-bit
10220 or 32-bit library, 3 is rounded up to 4). See the README file in the
10221 source distribution and the pcrebuild documentation for details. In
10222 these cases the limit is substantially larger. However, the speed of
10223 execution is slower.
10225 All values in repeating quantifiers must be less than 65536.
10227 There is no limit to the number of parenthesized subpatterns, but there
10228 can be no more than 65535 capturing subpatterns. There is, however, a
10229 limit to the depth of nesting of parenthesized subpatterns of all
10230 kinds. This is imposed in order to limit the amount of system stack
10231 used at compile time. The limit can be specified when PCRE is built;
10232 the default is 250.
10234 There is a limit to the number of forward references to subsequent sub-
10235 patterns of around 200,000. Repeated forward references with fixed
10236 upper limits, for example, (?2){0,100} when subpattern number 2 is to
10237 the right, are included in the count. There is no limit to the number
10238 of backward references.
10240 The maximum length of name for a named subpattern is 32 characters, and
10241 the maximum number of named subpatterns is 10000.
10243 The maximum length of a name in a (*MARK), (*PRUNE), (*SKIP), or
10244 (*THEN) verb is 255 for the 8-bit library and 65535 for the 16-bit and
10247 The maximum length of a subject string is the largest positive number
10248 that an integer variable can hold. However, when using the traditional
10249 matching function, PCRE uses recursion to handle subpatterns and indef-
10250 inite repetition. This means that the available stack space may limit
10251 the size of a subject string that can be processed by certain patterns.
10252 For a discussion of stack issues, see the pcrestack documentation.
10258 University Computing Service
10259 Cambridge CB2 3QH, England.
10264 Last updated: 05 November 2013
10265 Copyright (c) 1997-2013 University of Cambridge.
10266 ------------------------------------------------------------------------------
10269 PCRESTACK(3) Library Functions Manual PCRESTACK(3)
10274 PCRE - Perl-compatible regular expressions
10276 PCRE DISCUSSION OF STACK USAGE
10278 When you call pcre[16|32]_exec(), it makes use of an internal function
10279 called match(). This calls itself recursively at branch points in the
10280 pattern, in order to remember the state of the match so that it can
10281 back up and try a different alternative if the first one fails. As
10282 matching proceeds deeper and deeper into the tree of possibilities, the
10283 recursion depth increases. The match() function is also called in other
10284 circumstances, for example, whenever a parenthesized sub-pattern is
10285 entered, and in certain cases of repetition.
10287 Not all calls of match() increase the recursion depth; for an item such
10288 as a* it may be called several times at the same level, after matching
10289 different numbers of a's. Furthermore, in a number of cases where the
10290 result of the recursive call would immediately be passed back as the
10291 result of the current call (a "tail recursion"), the function is just
10294 The above comments apply when pcre[16|32]_exec() is run in its normal
10295 interpretive manner. If the pattern was studied with the
10296 PCRE_STUDY_JIT_COMPILE option, and just-in-time compiling was success-
10297 ful, and the options passed to pcre[16|32]_exec() were not incompati-
10298 ble, the matching process uses the JIT-compiled code instead of the
10299 match() function. In this case, the memory requirements are handled
10300 entirely differently. See the pcrejit documentation for details.
10302 The pcre[16|32]_dfa_exec() function operates in an entirely different
10303 way, and uses recursion only when there is a regular expression recur-
10304 sion or subroutine call in the pattern. This includes the processing of
10305 assertion and "once-only" subpatterns, which are handled like subrou-
10306 tine calls. Normally, these are never very deep, and the limit on the
10307 complexity of pcre[16|32]_dfa_exec() is controlled by the amount of
10308 workspace it is given. However, it is possible to write patterns with
10309 runaway infinite recursions; such patterns will cause
10310 pcre[16|32]_dfa_exec() to run out of stack. At present, there is no
10311 protection against this.
10313 The comments that follow do NOT apply to pcre[16|32]_dfa_exec(); they
10314 are relevant only for pcre[16|32]_exec() without the JIT optimization.
10316 Reducing pcre[16|32]_exec()'s stack usage
10318 Each time that match() is actually called recursively, it uses memory
10319 from the process stack. For certain kinds of pattern and data, very
10320 large amounts of stack may be needed, despite the recognition of "tail
10321 recursion". You can often reduce the amount of recursion, and there-
10322 fore the amount of stack used, by modifying the pattern that is being
10323 matched. Consider, for example, this pattern:
10327 It matches from wherever it starts until it encounters "<inet" or the
10328 end of the data, and is the kind of pattern that might be used when
10329 processing an XML file. Each iteration of the outer parentheses matches
10330 either one character that is not "<" or a "<" that is not followed by
10331 "inet". However, each time a parenthesis is processed, a recursion
10332 occurs, so this formulation uses a stack frame for each matched charac-
10333 ter. For a long string, a lot of stack is required. Consider now this
10334 rewritten pattern, which matches exactly the same strings:
10336 ([^<]++|<(?!inet))+
10338 This uses very much less stack, because runs of characters that do not
10339 contain "<" are "swallowed" in one item inside the parentheses. Recur-
10340 sion happens only when a "<" character that is not followed by "inet"
10341 is encountered (and we assume this is relatively rare). A possessive
10342 quantifier is used to stop any backtracking into the runs of non-"<"
10343 characters, but that is not related to stack usage.
10345 This example shows that one way of avoiding stack problems when match-
10346 ing long subject strings is to write repeated parenthesized subpatterns
10347 to match more than one character whenever possible.
10349 Compiling PCRE to use heap instead of stack for pcre[16|32]_exec()
10351 In environments where stack memory is constrained, you might want to
10352 compile PCRE to use heap memory instead of stack for remembering back-
10353 up points when pcre[16|32]_exec() is running. This makes it run a lot
10354 more slowly, however. Details of how to do this are given in the pcre-
10355 build documentation. When built in this way, instead of using the
10356 stack, PCRE obtains and frees memory by calling the functions that are
10357 pointed to by the pcre[16|32]_stack_malloc and pcre[16|32]_stack_free
10358 variables. By default, these point to malloc() and free(), but you can
10359 replace the pointers to cause PCRE to use your own functions. Since the
10360 block sizes are always the same, and are always freed in reverse order,
10361 it may be possible to implement customized memory handlers that are
10362 more efficient than the standard functions.
10364 Limiting pcre[16|32]_exec()'s stack usage
10366 You can set limits on the number of times that match() is called, both
10367 in total and recursively. If a limit is exceeded, pcre[16|32]_exec()
10368 returns an error code. Setting suitable limits should prevent it from
10369 running out of stack. The default values of the limits are very large,
10370 and unlikely ever to operate. They can be changed when PCRE is built,
10371 and they can also be set when pcre[16|32]_exec() is called. For details
10372 of these interfaces, see the pcrebuild documentation and the section on
10373 extra data for pcre[16|32]_exec() in the pcreapi documentation.
10375 As a very rough rule of thumb, you should reckon on about 500 bytes per
10376 recursion. Thus, if you want to limit your stack usage to 8Mb, you
10377 should set the limit at 16000 recursions. A 64Mb stack, on the other
10378 hand, can support around 128000 recursions.
10380 In Unix-like environments, the pcretest test program has a command line
10381 option (-S) that can be used to increase the size of its stack. As long
10382 as the stack is large enough, another option (-M) can be used to find
10383 the smallest limits that allow a particular pattern to match a given
10384 subject string. This is done by calling pcre[16|32]_exec() repeatedly
10385 with different limits.
10387 Obtaining an estimate of stack usage
10389 The actual amount of stack used per recursion can vary quite a lot,
10390 depending on the compiler that was used to build PCRE and the optimiza-
10391 tion or debugging options that were set for it. The rule of thumb value
10392 of 500 bytes mentioned above may be larger or smaller than what is
10393 actually needed. A better approximation can be obtained by running this
10398 The -C option causes pcretest to output information about the options
10399 with which PCRE was compiled. When -m is also given (before -C), infor-
10400 mation about stack use is given in a line like this:
10402 Match recursion uses stack: approximate frame size = 640 bytes
10404 The value is approximate because some recursions need a bit more (up to
10405 perhaps 16 more bytes).
10407 If the above command is given when PCRE is compiled to use the heap
10408 instead of the stack for recursion, the value that is output is the
10409 size of each block that is obtained from the heap.
10411 Changing stack size in Unix-like systems
10413 In Unix-like environments, there is not often a problem with the stack
10414 unless very long strings are involved, though the default limit on
10415 stack size varies from system to system. Values from 8Mb to 64Mb are
10416 common. You can find your default limit by running the command:
10420 Unfortunately, the effect of running out of stack is often SIGSEGV,
10421 though sometimes a more explicit error message is given. You can nor-
10422 mally increase the limit on stack size by code such as this:
10424 struct rlimit rlim;
10425 getrlimit(RLIMIT_STACK, &rlim);
10426 rlim.rlim_cur = 100*1024*1024;
10427 setrlimit(RLIMIT_STACK, &rlim);
10429 This reads the current limits (soft and hard) using getrlimit(), then
10430 attempts to increase the soft limit to 100Mb using setrlimit(). You
10431 must do this before calling pcre[16|32]_exec().
10433 Changing stack size in Mac OS X
10435 Using setrlimit(), as described above, should also work on Mac OS X. It
10436 is also possible to set a stack size when linking a program. There is a
10437 discussion about stack sizes in Mac OS X at this web site:
10438 http://developer.apple.com/qa/qa2005/qa1419.html.
10444 University Computing Service
10445 Cambridge CB2 3QH, England.
10450 Last updated: 24 June 2012
10451 Copyright (c) 1997-2012 University of Cambridge.
10452 ------------------------------------------------------------------------------
~ian / pcre3.git / blob