ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 77 - (hide annotations) (download)
Sat Feb 24 21:40:45 2007 UTC (8 years, 2 months ago) by nigel
File MIME type: text/plain
File size: 218708 byte(s)
Load pcre-6.0 into code/trunk.

1 nigel 75 -----------------------------------------------------------------------------
2 nigel 63 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. There are
6     separate text files for the pcregrep and pcretest commands.
7     -----------------------------------------------------------------------------
9 nigel 41
11 nigel 73 NAME
12     PCRE - Perl-compatible regular expressions
14 nigel 77
15 nigel 75 INTRODUCTION
16 nigel 41
17 nigel 73 The PCRE library is a set of functions that implement regular expres-
18     sion pattern matching using the same syntax and semantics as Perl, with
19     just a few differences. The current implementation of PCRE (release
20 nigel 77 6.x) corresponds approximately with Perl 5.8, including support for
21 nigel 75 UTF-8 encoded strings and Unicode general category properties. However,
22     this support has to be explicitly enabled; it is not the default.
23 nigel 63
24 nigel 77 In addition to the Perl-compatible matching function, PCRE also con-
25     tains an alternative matching function that matches the same compiled
26     patterns in a different way. In certain circumstances, the alternative
27     function has some advantages. For a discussion of the two matching
28     algorithms, see the pcrematching page.
30 nigel 75 PCRE is written in C and released as a C library. A number of people
31 nigel 77 have written wrappers and interfaces of various kinds. In particular,
32     Google Inc. have provided a comprehensive C++ wrapper. This is now
33     included as part of the PCRE distribution. The pcrecpp page has details
34     of this interface. Other people's contributions can be found in the
35     Contrib directory at the primary FTP site, which is:
36 nigel 63
37 nigel 73 ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre
38 nigel 63
39 nigel 73 Details of exactly which Perl regular expression features are and are
40     not supported by PCRE are given in separate documents. See the pcrepat-
41     tern and pcrecompat pages.
42 nigel 63
43 nigel 73 Some features of PCRE can be included, excluded, or changed when the
44     library is built. The pcre_config() function makes it possible for a
45 nigel 75 client to discover which features are available. The features them-
46     selves are described in the pcrebuild page. Documentation about build-
47     ing PCRE for various operating systems can be found in the README file
48     in the source distribution.
49 nigel 63
50 nigel 77 The library contains a number of undocumented internal functions and
51     data tables that are used by more than one of the exported external
52     functions, but which are not intended for use by external callers.
53     Their names all begin with "_pcre_", which hopefully will not provoke
54     any name clashes.
55 nigel 63
56 nigel 77
59 nigel 75 The user documentation for PCRE comprises a number of different sec-
60     tions. In the "man" format, each of these is a separate "man page". In
61     the HTML format, each is a separate page, linked from the index page.
62     In the plain text format, all the sections are concatenated, for ease
63     of searching. The sections are as follows:
64 nigel 63
65 nigel 73 pcre this document
66 nigel 77 pcreapi details of PCRE's native C API
67 nigel 73 pcrebuild options for building PCRE
68     pcrecallout details of the callout feature
69     pcrecompat discussion of Perl compatibility
70 nigel 77 pcrecpp details of the C++ wrapper
71 nigel 73 pcregrep description of the pcregrep command
72 nigel 77 pcrematching discussion of the two matching algorithms
73 nigel 75 pcrepartial details of the partial matching facility
74 nigel 73 pcrepattern syntax and semantics of supported
75     regular expressions
76     pcreperform discussion of performance issues
77 nigel 77 pcreposix the POSIX-compatible C API
78 nigel 75 pcreprecompile details of saving and re-using precompiled patterns
79 nigel 73 pcresample discussion of the sample program
80 nigel 75 pcretest description of the pcretest testing command
81 nigel 63
82 nigel 73 In addition, in the "man" and HTML formats, there is a short page for
83 nigel 77 each C library function, listing its arguments and results.
84 nigel 63
88 nigel 73 There are some size limitations in PCRE but it is hoped that they will
89     never in practice be relevant.
90 nigel 63
91 nigel 73 The maximum length of a compiled pattern is 65539 (sic) bytes if PCRE
92     is compiled with the default internal linkage size of 2. If you want to
93     process regular expressions that are truly enormous, you can compile
94     PCRE with an internal linkage size of 3 or 4 (see the README file in
95     the source distribution and the pcrebuild documentation for details).
96 nigel 75 In these cases the limit is substantially larger. However, the speed
97 nigel 73 of execution will be slower.
98 nigel 63
99 nigel 73 All values in repeating quantifiers must be less than 65536. The maxi-
100     mum number of capturing subpatterns is 65535.
101 nigel 63
102 nigel 73 There is no limit to the number of non-capturing subpatterns, but the
103     maximum depth of nesting of all kinds of parenthesized subpattern,
104     including capturing subpatterns, assertions, and other types of subpat-
105     tern, is 200.
106 nigel 63
107 nigel 73 The maximum length of a subject string is the largest positive number
108 nigel 77 that an integer variable can hold. However, when using the traditional
109     matching function, PCRE uses recursion to handle subpatterns and indef-
110     inite repetition. This means that the available stack space may limit
111     the size of a subject string that can be processed by certain patterns.
112 nigel 63
115 nigel 63
116 nigel 77 From release 3.3, PCRE has had some support for character strings
117     encoded in the UTF-8 format. For release 4.0 this was greatly extended
118     to cover most common requirements, and in release 5.0 additional sup-
119 nigel 75 port for Unicode general category properties was added.
120 nigel 63
121 nigel 77 In order process UTF-8 strings, you must build PCRE to include UTF-8
122     support in the code, and, in addition, you must call pcre_compile()
123     with the PCRE_UTF8 option flag. When you do this, both the pattern and
124     any subject strings that are matched against it are treated as UTF-8
125 nigel 73 strings instead of just strings of bytes.
126 nigel 63
127 nigel 77 If you compile PCRE with UTF-8 support, but do not use it at run time,
128     the library will be a bit bigger, but the additional run time overhead
129     is limited to testing the PCRE_UTF8 flag in several places, so should
130 nigel 73 not be very large.
131 nigel 63
132 nigel 75 If PCRE is built with Unicode character property support (which implies
133 nigel 77 UTF-8 support), the escape sequences \p{..}, \P{..}, and \X are sup-
134 nigel 75 ported. The available properties that can be tested are limited to the
135 nigel 77 general category properties such as Lu for an upper case letter or Nd
136     for a decimal number. A full list is given in the pcrepattern documen-
137 nigel 75 tation. The PCRE library is increased in size by about 90K when Unicode
138     property support is included.
140 nigel 73 The following comments apply when PCRE is running in UTF-8 mode:
141 nigel 63
142 nigel 77 1. When you set the PCRE_UTF8 flag, the strings passed as patterns and
143     subjects are checked for validity on entry to the relevant functions.
144 nigel 73 If an invalid UTF-8 string is passed, an error return is given. In some
145 nigel 77 situations, you may already know that your strings are valid, and
146 nigel 73 therefore want to skip these checks in order to improve performance. If
147 nigel 77 you set the PCRE_NO_UTF8_CHECK flag at compile time or at run time,
148     PCRE assumes that the pattern or subject it is given (respectively)
149     contains only valid UTF-8 codes. In this case, it does not diagnose an
150     invalid UTF-8 string. If you pass an invalid UTF-8 string to PCRE when
151     PCRE_NO_UTF8_CHECK is set, the results are undefined. Your program may
152 nigel 73 crash.
153 nigel 63
154 nigel 73 2. In a pattern, the escape sequence \x{...}, where the contents of the
155 nigel 77 braces is a string of hexadecimal digits, is interpreted as a UTF-8
156     character whose code number is the given hexadecimal number, for exam-
157     ple: \x{1234}. If a non-hexadecimal digit appears between the braces,
158 nigel 73 the item is not recognized. This escape sequence can be used either as
159     a literal, or within a character class.
160 nigel 63
161 nigel 77 3. The original hexadecimal escape sequence, \xhh, matches a two-byte
162 nigel 73 UTF-8 character if the value is greater than 127.
163 nigel 63
164 nigel 77 4. Repeat quantifiers apply to complete UTF-8 characters, not to indi-
165 nigel 73 vidual bytes, for example: \x{100}{3}.
166 nigel 63
167 nigel 77 5. The dot metacharacter matches one UTF-8 character instead of a sin-
168 nigel 75 gle byte.
169 nigel 63
170 nigel 77 6. The escape sequence \C can be used to match a single byte in UTF-8
171     mode, but its use can lead to some strange effects. This facility is
172     not available in the alternative matching function, pcre_dfa_exec().
173 nigel 63
174 nigel 75 7. The character escapes \b, \B, \d, \D, \s, \S, \w, and \W correctly
175     test characters of any code value, but the characters that PCRE recog-
176     nizes as digits, spaces, or word characters remain the same set as
177     before, all with values less than 256. This remains true even when PCRE
178     includes Unicode property support, because to do otherwise would slow
179     down PCRE in many common cases. If you really want to test for a wider
180     sense of, say, "digit", you must use Unicode property tests such as
181     \p{Nd}.
182 nigel 63
183 nigel 75 8. Similarly, characters that match the POSIX named character classes
184     are all low-valued characters.
185 nigel 63
186 nigel 75 9. Case-insensitive matching applies only to characters whose values
187     are less than 128, unless PCRE is built with Unicode property support.
188     Even when Unicode property support is available, PCRE still uses its
189     own character tables when checking the case of low-valued characters,
190     so as not to degrade performance. The Unicode property information is
191     used only for characters with higher values.
192 nigel 63
194     AUTHOR
196 nigel 77 Philip Hazel
197 nigel 73 University Computing Service,
198     Cambridge CB2 3QG, England.
199 nigel 63
200 nigel 77 Putting an actual email address here seems to have been a spam magnet,
201     so I've taken it away. If you want to email me, use my initial and sur-
202     name, separated by a dot, at the domain ucs.cam.ac.uk.
204     Last updated: 07 March 2005
205     Copyright (c) 1997-2005 University of Cambridge.
206 nigel 63 -----------------------------------------------------------------------------
210 nigel 73 NAME
211     PCRE - Perl-compatible regular expressions
213 nigel 77
216 nigel 73 This document describes the optional features of PCRE that can be
217     selected when the library is compiled. They are all selected, or dese-
218 nigel 75 lected, by providing options to the configure script that is run before
219     the make command. The complete list of options for configure (which
220     includes the standard ones such as the selection of the installation
221     directory) can be obtained by running
222 nigel 63
223 nigel 73 ./configure --help
224 nigel 63
225 nigel 73 The following sections describe certain options whose names begin with
226     --enable or --disable. These settings specify changes to the defaults
227     for the configure command. Because of the way that configure works,
228     --enable and --disable always come in pairs, so the complementary
229     option always exists as well, but as it specifies the default, it is
230     not described.
231 nigel 63
233     UTF-8 SUPPORT
235 nigel 73 To build PCRE with support for UTF-8 character strings, add
236 nigel 63
237 nigel 73 --enable-utf8
238 nigel 63
239 nigel 73 to the configure command. Of itself, this does not make PCRE treat
240     strings as UTF-8. As well as compiling PCRE with this option, you also
241     have have to set the PCRE_UTF8 option when you call the pcre_compile()
242     function.
243 nigel 63
247     UTF-8 support allows PCRE to process character values greater than 255
248     in the strings that it handles. On its own, however, it does not pro-
249     vide any facilities for accessing the properties of such characters. If
250     you want to be able to use the pattern escapes \P, \p, and \X, which
251     refer to Unicode character properties, you must add
253     --enable-unicode-properties
255     to the configure command. This implies UTF-8 support, even if you have
256     not explicitly requested it.
258     Including Unicode property support adds around 90K of tables to the
259     PCRE library, approximately doubling its size. Only the general cate-
260     gory properties such as Lu and Nd are supported. Details are given in
261     the pcrepattern documentation.
266 nigel 73 By default, PCRE treats character 10 (linefeed) as the newline charac-
267     ter. This is the normal newline character on Unix-like systems. You can
268     compile PCRE to use character 13 (carriage return) instead by adding
269 nigel 63
270 nigel 73 --enable-newline-is-cr
271 nigel 63
272 nigel 73 to the configure command. For completeness there is also a --enable-
273     newline-is-lf option, which explicitly specifies linefeed as the new-
274     line character.
275 nigel 63
279 nigel 73 The PCRE building process uses libtool to build both shared and static
280     Unix libraries by default. You can suppress one of these by adding one
281     of
282 nigel 63
283 nigel 73 --disable-shared
284     --disable-static
285 nigel 63
286 nigel 73 to the configure command, as required.
287 nigel 63
291 nigel 75 When PCRE is called through the POSIX interface (see the pcreposix doc-
292     umentation), additional working storage is required for holding the
293     pointers to capturing substrings, because PCRE requires three integers
294 nigel 73 per substring, whereas the POSIX interface provides only two. If the
295     number of expected substrings is small, the wrapper function uses space
296     on the stack, because this is faster than using malloc() for each call.
297     The default threshold above which the stack is no longer used is 10; it
298     can be changed by adding a setting such as
299 nigel 63
300 nigel 73 --with-posix-malloc-threshold=20
301 nigel 63
302 nigel 73 to the configure command.
303 nigel 63
307 nigel 75 Internally, PCRE has a function called match(), which it calls repeat-
308 nigel 77 edly (possibly recursively) when matching a pattern with the
309     pcre_exec() function. By controlling the maximum number of times this
310     function may be called during a single matching operation, a limit can
311     be placed on the resources used by a single call to pcre_exec(). The
312     limit can be changed at run time, as described in the pcreapi documen-
313     tation. The default is 10 million, but this can be changed by adding a
314     setting such as
315 nigel 63
316 nigel 73 --with-match-limit=500000
317 nigel 63
318 nigel 77 to the configure command. This setting has no effect on the
319     pcre_dfa_exec() matching function.
320 nigel 63
324 nigel 73 Within a compiled pattern, offset values are used to point from one
325     part to another (for example, from an opening parenthesis to an alter-
326 nigel 75 nation metacharacter). By default, two-byte values are used for these
327 nigel 73 offsets, leading to a maximum size for a compiled pattern of around
328     64K. This is sufficient to handle all but the most gigantic patterns.
329     Nevertheless, some people do want to process enormous patterns, so it
330     is possible to compile PCRE to use three-byte or four-byte offsets by
331     adding a setting such as
332 nigel 63
333 nigel 73 --with-link-size=3
334 nigel 63
335 nigel 73 to the configure command. The value given must be 2, 3, or 4. Using
336     longer offsets slows down the operation of PCRE because it has to load
337     additional bytes when handling them.
338 nigel 63
339 nigel 73 If you build PCRE with an increased link size, test 2 (and test 5 if
340     you are using UTF-8) will fail. Part of the output of these tests is a
341     representation of the compiled pattern, and this changes with the link
342     size.
343 nigel 63
344 nigel 73
347 nigel 77 When matching with the pcre_exec() function, PCRE implements backtrack-
348     ing by making recursive calls to an internal function called match().
349     In environments where the size of the stack is limited, this can se-
350     verely limit PCRE's operation. (The Unix environment does not usually
351     suffer from this problem.) An alternative approach that uses memory
352     from the heap to remember data, instead of using recursive function
353     calls, has been implemented to work round this problem. If you want to
354     build a version of PCRE that works this way, add
355 nigel 73
356     --disable-stack-for-recursion
358     to the configure command. With this configuration, PCRE will use the
359 nigel 75 pcre_stack_malloc and pcre_stack_free variables to call memory manage-
360     ment functions. Separate functions are provided because the usage is
361     very predictable: the block sizes requested are always the same, and
362 nigel 73 the blocks are always freed in reverse order. A calling program might
363     be able to implement optimized functions that perform better than the
364     standard malloc() and free() functions. PCRE runs noticeably more
365 nigel 77 slowly when built in this way. This option affects only the pcre_exec()
366     function; it is not relevant for the the pcre_dfa_exec() function.
367 nigel 73
371 nigel 77 PCRE assumes by default that it will run in an environment where the
372     character code is ASCII (or Unicode, which is a superset of ASCII).
373     PCRE can, however, be compiled to run in an EBCDIC environment by
374 nigel 75 adding
375 nigel 73
376     --enable-ebcdic
378     to the configure command.
380 nigel 77 Last updated: 28 February 2005
381     Copyright (c) 1997-2005 University of Cambridge.
382 nigel 63 -----------------------------------------------------------------------------
386 nigel 77 NAME
387     PCRE - Perl-compatible regular expressions
388 nigel 73
389 nigel 77
392     This document describes the two different algorithms that are available
393     in PCRE for matching a compiled regular expression against a given sub-
394     ject string. The "standard" algorithm is the one provided by the
395     pcre_exec() function. This works in the same was as Perl's matching
396     function, and provides a Perl-compatible matching operation.
398     An alternative algorithm is provided by the pcre_dfa_exec() function;
399     this operates in a different way, and is not Perl-compatible. It has
400     advantages and disadvantages compared with the standard algorithm, and
401     these are described below.
403     When there is only one possible way in which a given subject string can
404     match a pattern, the two algorithms give the same answer. A difference
405     arises, however, when there are multiple possibilities. For example, if
406     the pattern
408     ^<.*>
410     is matched against the string
412     <something> <something else> <something further>
414     there are three possible answers. The standard algorithm finds only one
415     of them, whereas the DFA algorithm finds all three.
420     The set of strings that are matched by a regular expression can be rep-
421     resented as a tree structure. An unlimited repetition in the pattern
422     makes the tree of infinite size, but it is still a tree. Matching the
423     pattern to a given subject string (from a given starting point) can be
424     thought of as a search of the tree. There are two standard ways to
425     search a tree: depth-first and breadth-first, and these correspond to
426     the two matching algorithms provided by PCRE.
431     In the terminology of Jeffrey Friedl's book Mastering Regular Expres-
432     sions, the standard algorithm is an "NFA algorithm". It conducts a
433     depth-first search of the pattern tree. That is, it proceeds along a
434     single path through the tree, checking that the subject matches what is
435     required. When there is a mismatch, the algorithm tries any alterna-
436     tives at the current point, and if they all fail, it backs up to the
437     previous branch point in the tree, and tries the next alternative
438     branch at that level. This often involves backing up (moving to the
439     left) in the subject string as well. The order in which repetition
440     branches are tried is controlled by the greedy or ungreedy nature of
441     the quantifier.
443     If a leaf node is reached, a matching string has been found, and at
444     that point the algorithm stops. Thus, if there is more than one possi-
445     ble match, this algorithm returns the first one that it finds. Whether
446     this is the shortest, the longest, or some intermediate length depends
447     on the way the greedy and ungreedy repetition quantifiers are specified
448     in the pattern.
450     Because it ends up with a single path through the tree, it is rela-
451     tively straightforward for this algorithm to keep track of the sub-
452     strings that are matched by portions of the pattern in parentheses.
453     This provides support for capturing parentheses and back references.
458     DFA stands for "deterministic finite automaton", but you do not need to
459     understand the origins of that name. This algorithm conducts a breadth-
460     first search of the tree. Starting from the first matching point in the
461     subject, it scans the subject string from left to right, once, charac-
462     ter by character, and as it does this, it remembers all the paths
463     through the tree that represent valid matches.
465     The scan continues until either the end of the subject is reached, or
466     there are no more unterminated paths. At this point, terminated paths
467     represent the different matching possibilities (if there are none, the
468     match has failed). Thus, if there is more than one possible match,
469     this algorithm finds all of them, and in particular, it finds the long-
470     est. In PCRE, there is an option to stop the algorithm after the first
471     match (which is necessarily the shortest) has been found.
473     Note that all the matches that are found start at the same point in the
474     subject. If the pattern
476     cat(er(pillar)?)
478     is matched against the string "the caterpillar catchment", the result
479     will be the three strings "cat", "cater", and "caterpillar" that start
480     at the fourth character of the subject. The algorithm does not automat-
481     ically move on to find matches that start at later positions.
483     There are a number of features of PCRE regular expressions that are not
484     supported by the DFA matching algorithm. They are as follows:
486     1. Because the algorithm finds all possible matches, the greedy or
487     ungreedy nature of repetition quantifiers is not relevant. Greedy and
488     ungreedy quantifiers are treated in exactly the same way.
490     2. When dealing with multiple paths through the tree simultaneously, it
491     is not straightforward to keep track of captured substrings for the
492     different matching possibilities, and PCRE's implementation of this
493     algorithm does not attempt to do this. This means that no captured sub-
494     strings are available.
496     3. Because no substrings are captured, back references within the pat-
497     tern are not supported, and cause errors if encountered.
499     4. For the same reason, conditional expressions that use a backrefer-
500     ence as the condition are not supported.
502     5. Callouts are supported, but the value of the capture_top field is
503     always 1, and the value of the capture_last field is always -1.
505     6. The \C escape sequence, which (in the standard algorithm) matches a
506     single byte, even in UTF-8 mode, is not supported because the DFA algo-
507     rithm moves through the subject string one character at a time, for all
508     active paths through the tree.
513     Using the DFA matching algorithm provides the following advantages:
515     1. All possible matches (at a single point in the subject) are automat-
516     ically found, and in particular, the longest match is found. To find
517     more than one match using the standard algorithm, you have to do kludgy
518     things with callouts.
520     2. There is much better support for partial matching. The restrictions
521     on the content of the pattern that apply when using the standard algo-
522     rithm for partial matching do not apply to the DFA algorithm. For non-
523     anchored patterns, the starting position of a partial match is avail-
524     able.
526     3. Because the DFA algorithm scans the subject string just once, and
527     never needs to backtrack, it is possible to pass very long subject
528     strings to the matching function in several pieces, checking for par-
529     tial matching each time.
534     The DFA algorithm suffers from a number of disadvantages:
536     1. It is substantially slower than the standard algorithm. This is
537     partly because it has to search for all possible matches, but is also
538     because it is less susceptible to optimization.
540     2. Capturing parentheses and back references are not supported.
542     3. The "atomic group" feature of PCRE regular expressions is supported,
543     but does not provide the advantage that it does for the standard algo-
544     rithm.
546     Last updated: 28 February 2005
547     Copyright (c) 1997-2005 University of Cambridge.
548     -----------------------------------------------------------------------------
552 nigel 73 NAME
553     PCRE - Perl-compatible regular expressions
555 nigel 77
556 nigel 75 PCRE NATIVE API
557 nigel 63
558 nigel 73 #include <pcre.h>
559 nigel 41
560 nigel 73 pcre *pcre_compile(const char *pattern, int options,
561     const char **errptr, int *erroffset,
562     const unsigned char *tableptr);
563 nigel 41
564 nigel 77 pcre *pcre_compile2(const char *pattern, int options,
565     int *errorcodeptr,
566     const char **errptr, int *erroffset,
567     const unsigned char *tableptr);
569 nigel 73 pcre_extra *pcre_study(const pcre *code, int options,
570     const char **errptr);
571 nigel 41
572 nigel 73 int pcre_exec(const pcre *code, const pcre_extra *extra,
573     const char *subject, int length, int startoffset,
574     int options, int *ovector, int ovecsize);
575 nigel 41
576 nigel 77 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
577     const char *subject, int length, int startoffset,
578     int options, int *ovector, int ovecsize,
579     int *workspace, int wscount);
581 nigel 73 int pcre_copy_named_substring(const pcre *code,
582     const char *subject, int *ovector,
583     int stringcount, const char *stringname,
584     char *buffer, int buffersize);
585 nigel 63
586 nigel 73 int pcre_copy_substring(const char *subject, int *ovector,
587     int stringcount, int stringnumber, char *buffer,
588     int buffersize);
589 nigel 41
590 nigel 73 int pcre_get_named_substring(const pcre *code,
591     const char *subject, int *ovector,
592     int stringcount, const char *stringname,
593     const char **stringptr);
594 nigel 63
595 nigel 73 int pcre_get_stringnumber(const pcre *code,
596     const char *name);
597 nigel 63
598 nigel 73 int pcre_get_substring(const char *subject, int *ovector,
599     int stringcount, int stringnumber,
600     const char **stringptr);
601 nigel 41
602 nigel 73 int pcre_get_substring_list(const char *subject,
603     int *ovector, int stringcount, const char ***listptr);
604 nigel 41
605 nigel 73 void pcre_free_substring(const char *stringptr);
606 nigel 49
607 nigel 73 void pcre_free_substring_list(const char **stringptr);
608 nigel 49
609 nigel 73 const unsigned char *pcre_maketables(void);
610 nigel 41
611 nigel 73 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
612     int what, void *where);
613 nigel 43
614 nigel 73 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
615 nigel 63
616 nigel 77 int pcre_refcount(pcre *code, int adjust);
618 nigel 73 int pcre_config(int what, void *where);
619 nigel 41
620 nigel 73 char *pcre_version(void);
621 nigel 63
622 nigel 73 void *(*pcre_malloc)(size_t);
623 nigel 41
624 nigel 73 void (*pcre_free)(void *);
625 nigel 41
626 nigel 73 void *(*pcre_stack_malloc)(size_t);
627 nigel 41
628 nigel 73 void (*pcre_stack_free)(void *);
629 nigel 41
630 nigel 73 int (*pcre_callout)(pcre_callout_block *);
631 nigel 41
632 nigel 73
633 nigel 75 PCRE API OVERVIEW
634 nigel 41
635 nigel 73 PCRE has its own native API, which is described in this document. There
636     is also a set of wrapper functions that correspond to the POSIX regular
637 nigel 77 expression API. These are described in the pcreposix documentation.
638     Both of these APIs define a set of C function calls. A C++ wrapper is
639     distributed with PCRE. It is documented in the pcrecpp page.
640 nigel 43
641 nigel 77 The native API C function prototypes are defined in the header file
642     pcre.h, and on Unix systems the library itself is called libpcre. It
643 nigel 75 can normally be accessed by adding -lpcre to the command for linking an
644     application that uses PCRE. The header file defines the macros
645     PCRE_MAJOR and PCRE_MINOR to contain the major and minor release num-
646     bers for the library. Applications can use these to include support
647     for different releases of PCRE.
648 nigel 41
649 nigel 77 The functions pcre_compile(), pcre_compile2(), pcre_study(), and
650     pcre_exec() are used for compiling and matching regular expressions in
651     a Perl-compatible manner. A sample program that demonstrates the sim-
652     plest way of using them is provided in the file called pcredemo.c in
653     the source distribution. The pcresample documentation describes how to
654     run it.
655 nigel 49
656 nigel 77 A second matching function, pcre_dfa_exec(), which is not Perl-compati-
657     ble, is also provided. This uses a different algorithm for the match-
658     ing. This allows it to find all possible matches (at a given point in
659     the subject), not just one. However, this algorithm does not return
660     captured substrings. A description of the two matching algorithms and
661     their advantages and disadvantages is given in the pcrematching docu-
662     mentation.
663 nigel 63
664 nigel 77 In addition to the main compiling and matching functions, there are
665     convenience functions for extracting captured substrings from a subject
666     string that is matched by pcre_exec(). They are:
668 nigel 73 pcre_copy_substring()
669     pcre_copy_named_substring()
670     pcre_get_substring()
671     pcre_get_named_substring()
672     pcre_get_substring_list()
673 nigel 75 pcre_get_stringnumber()
674 nigel 63
675 nigel 73 pcre_free_substring() and pcre_free_substring_list() are also provided,
676     to free the memory used for extracted strings.
677 nigel 41
678 nigel 77 The function pcre_maketables() is used to build a set of character
679     tables in the current locale for passing to pcre_compile(),
680     pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is
681     provided for specialist use. Most commonly, no special tables are
682     passed, in which case internal tables that are generated when PCRE is
683     built are used.
684 nigel 49
685 nigel 75 The function pcre_fullinfo() is used to find out information about a
686     compiled pattern; pcre_info() is an obsolete version that returns only
687     some of the available information, but is retained for backwards com-
688     patibility. The function pcre_version() returns a pointer to a string
689 nigel 73 containing the version of PCRE and its date of release.
690 nigel 41
691 nigel 77 The function pcre_refcount() maintains a reference count in a data
692     block containing a compiled pattern. This is provided for the benefit
693     of object-oriented applications.
695 nigel 75 The global variables pcre_malloc and pcre_free initially contain the
696     entry points of the standard malloc() and free() functions, respec-
697 nigel 73 tively. PCRE calls the memory management functions via these variables,
698 nigel 75 so a calling program can replace them if it wishes to intercept the
699 nigel 73 calls. This should be done before calling any PCRE functions.
700 nigel 41
701 nigel 75 The global variables pcre_stack_malloc and pcre_stack_free are also
702     indirections to memory management functions. These special functions
703     are used only when PCRE is compiled to use the heap for remembering
704 nigel 77 data, instead of recursive function calls, when running the pcre_exec()
705     function. This is a non-standard way of building PCRE, for use in envi-
706     ronments that have limited stacks. Because of the greater use of memory
707     management, it runs more slowly. Separate functions are provided so
708     that special-purpose external code can be used for this case. When
709     used, these functions are always called in a stack-like manner (last
710     obtained, first freed), and always for memory blocks of the same size.
711 nigel 41
712 nigel 73 The global variable pcre_callout initially contains NULL. It can be set
713 nigel 77 by the caller to a "callout" function, which PCRE will then call at
714     specified points during a matching operation. Details are given in the
715 nigel 73 pcrecallout documentation.
716 nigel 41
717 nigel 73
720 nigel 77 The PCRE functions can be used in multi-threading applications, with
721 nigel 73 the proviso that the memory management functions pointed to by
722     pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the
723     callout function pointed to by pcre_callout, are shared by all threads.
724 nigel 41
725 nigel 77 The compiled form of a regular expression is not altered during match-
726 nigel 73 ing, so the same compiled pattern can safely be used by several threads
727     at once.
728 nigel 41
732     The compiled form of a regular expression can be saved and re-used at a
733 nigel 77 later time, possibly by a different program, and even on a host other
734     than the one on which it was compiled. Details are given in the
735 nigel 75 pcreprecompile documentation.
739 nigel 41
740 nigel 73 int pcre_config(int what, void *where);
741 nigel 63
742 nigel 77 The function pcre_config() makes it possible for a PCRE client to dis-
743 nigel 73 cover which optional features have been compiled into the PCRE library.
744 nigel 77 The pcrebuild documentation has more details about these optional fea-
745 nigel 73 tures.
746 nigel 63
747 nigel 77 The first argument for pcre_config() is an integer, specifying which
748 nigel 73 information is required; the second argument is a pointer to a variable
749 nigel 77 into which the information is placed. The following information is
750 nigel 73 available:
751 nigel 63
752 nigel 73 PCRE_CONFIG_UTF8
753 nigel 63
754 nigel 77 The output is an integer that is set to one if UTF-8 support is avail-
755 nigel 73 able; otherwise it is set to zero.
756 nigel 63
759 nigel 77 The output is an integer that is set to one if support for Unicode
760 nigel 75 character properties is available; otherwise it is set to zero.
763 nigel 63
764 nigel 77 The output is an integer that is set to the value of the code that is
765     used for the newline character. It is either linefeed (10) or carriage
766     return (13), and should normally be the standard character for your
767 nigel 73 operating system.
768 nigel 63
770 nigel 63
771 nigel 77 The output is an integer that contains the number of bytes used for
772 nigel 73 internal linkage in compiled regular expressions. The value is 2, 3, or
773 nigel 77 4. Larger values allow larger regular expressions to be compiled, at
774     the expense of slower matching. The default value of 2 is sufficient
775     for all but the most massive patterns, since it allows the compiled
776 nigel 73 pattern to be up to 64K in size.
777 nigel 63
779 nigel 63
780 nigel 77 The output is an integer that contains the threshold above which the
781     POSIX interface uses malloc() for output vectors. Further details are
782 nigel 73 given in the pcreposix documentation.
783 nigel 63
785 nigel 63
786 nigel 73 The output is an integer that gives the default limit for the number of
787 nigel 77 internal matching function calls in a pcre_exec() execution. Further
788 nigel 73 details are given with pcre_exec() below.
789 nigel 63
791 nigel 63
792 nigel 77 The output is an integer that is set to one if internal recursion when
793     running pcre_exec() is implemented by recursive function calls that use
794     the stack to remember their state. This is the usual way that PCRE is
795     compiled. The output is zero if PCRE was compiled to use blocks of data
796     on the heap instead of recursive function calls. In this case,
797     pcre_stack_malloc and pcre_stack_free are called to manage memory
798     blocks on the heap, thus avoiding the use of the stack.
799 nigel 73
802 nigel 63
803 nigel 73 pcre *pcre_compile(const char *pattern, int options,
804     const char **errptr, int *erroffset,
805     const unsigned char *tableptr);
806 nigel 63
807 nigel 77 pcre *pcre_compile2(const char *pattern, int options,
808     int *errorcodeptr,
809     const char **errptr, int *erroffset,
810     const unsigned char *tableptr);
811 nigel 41
812 nigel 77 Either of the functions pcre_compile() or pcre_compile2() can be called
813     to compile a pattern into an internal form. The only difference between
814     the two interfaces is that pcre_compile2() has an additional argument,
815     errorcodeptr, via which a numerical error code can be returned.
817     The pattern is a C string terminated by a binary zero, and is passed in
818     the pattern argument. A pointer to a single block of memory that is
819     obtained via pcre_malloc is returned. This contains the compiled code
820     and related data. The pcre type is defined for the returned block; this
821     is a typedef for a structure whose contents are not externally defined.
822     It is up to the caller to free the memory when it is no longer
823     required.
825     Although the compiled code of a PCRE regex is relocatable, that is, it
826 nigel 73 does not depend on memory location, the complete pcre data block is not
827 nigel 77 fully relocatable, because it may contain a copy of the tableptr argu-
828 nigel 75 ment, which is an address (see below).
829 nigel 41
830 nigel 73 The options argument contains independent bits that affect the compila-
831 nigel 77 tion. It should be zero if no options are required. The available
832     options are described below. Some of them, in particular, those that
833     are compatible with Perl, can also be set and unset from within the
834     pattern (see the detailed description in the pcrepattern documenta-
835     tion). For these options, the contents of the options argument speci-
836     fies their initial settings at the start of compilation and execution.
837     The PCRE_ANCHORED option can be set at the time of matching as well as
838 nigel 75 at compile time.
839 nigel 41
840 nigel 73 If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise,
841 nigel 77 if compilation of a pattern fails, pcre_compile() returns NULL, and
842 nigel 73 sets the variable pointed to by errptr to point to a textual error mes-
843 nigel 77 sage. The offset from the start of the pattern to the character where
844     the error was discovered is placed in the variable pointed to by
845     erroffset, which must not be NULL. If it is, an immediate error is
846 nigel 73 given.
847 nigel 53
848 nigel 77 If pcre_compile2() is used instead of pcre_compile(), and the error-
849     codeptr argument is not NULL, a non-zero error code number is returned
850     via this argument in the event of an error. This is in addition to the
851     textual error message. Error codes and messages are listed below.
853 nigel 73 If the final argument, tableptr, is NULL, PCRE uses a default set of
854 nigel 75 character tables that are built when PCRE is compiled, using the
855     default C locale. Otherwise, tableptr must be an address that is the
856     result of a call to pcre_maketables(). This value is stored with the
857     compiled pattern, and used again by pcre_exec(), unless another table
858     pointer is passed to it. For more discussion, see the section on locale
859     support below.
860 nigel 53
861 nigel 75 This code fragment shows a typical straightforward call to pcre_com-
862 nigel 73 pile():
863 nigel 41
864 nigel 73 pcre *re;
865     const char *error;
866     int erroffset;
867     re = pcre_compile(
868     "^A.*Z", /* the pattern */
869     0, /* default options */
870     &error, /* for error message */
871     &erroffset, /* for error offset */
872     NULL); /* use default character tables */
873 nigel 41
874 nigel 75 The following names for option bits are defined in the pcre.h header
875     file:
876 nigel 41
877 nigel 73 PCRE_ANCHORED
878 nigel 41
879 nigel 73 If this bit is set, the pattern is forced to be "anchored", that is, it
880     is constrained to match only at the first matching point in the string
881 nigel 75 that is being searched (the "subject string"). This effect can also be
882 nigel 73 achieved by appropriate constructs in the pattern itself, which is the
883     only way to do it in Perl.
884 nigel 41
885 nigel 75 PCRE_AUTO_CALLOUT
887     If this bit is set, pcre_compile() automatically inserts callout items,
888     all with number 255, before each pattern item. For discussion of the
889     callout facility, see the pcrecallout documentation.
891 nigel 73 PCRE_CASELESS
892 nigel 41
893 nigel 73 If this bit is set, letters in the pattern match both upper and lower
894     case letters. It is equivalent to Perl's /i option, and it can be
895 nigel 77 changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE
896     always understands the concept of case for characters whose values are
897     less than 128, so caseless matching is always possible. For characters
898     with higher values, the concept of case is supported if PCRE is com-
899     piled with Unicode property support, but not otherwise. If you want to
900     use caseless matching for characters 128 and above, you must ensure
901     that PCRE is compiled with Unicode property support as well as with
902     UTF-8 support.
903 nigel 41
905 nigel 41
906 nigel 77 If this bit is set, a dollar metacharacter in the pattern matches only
907     at the end of the subject string. Without this option, a dollar also
908     matches immediately before the final character if it is a newline (but
909     not before any other newlines). The PCRE_DOLLAR_ENDONLY option is
910 nigel 73 ignored if PCRE_MULTILINE is set. There is no equivalent to this option
911     in Perl, and no way to set it within a pattern.
912 nigel 41
913 nigel 73 PCRE_DOTALL
914 nigel 41
915 nigel 73 If this bit is set, a dot metacharater in the pattern matches all char-
916 nigel 77 acters, including newlines. Without it, newlines are excluded. This
917     option is equivalent to Perl's /s option, and it can be changed within
918     a pattern by a (?s) option setting. A negative class such as [^a]
919     always matches a newline character, independent of the setting of this
920 nigel 73 option.
921 nigel 63
922 nigel 73 PCRE_EXTENDED
923 nigel 41
924 nigel 77 If this bit is set, whitespace data characters in the pattern are
925     totally ignored except when escaped or inside a character class. White-
926     space does not include the VT character (code 11). In addition, charac-
927     ters between an unescaped # outside a character class and the next new-
928     line character, inclusive, are also ignored. This is equivalent to
929     Perl's /x option, and it can be changed within a pattern by a (?x)
930 nigel 73 option setting.
931 nigel 41
932 nigel 77 This option makes it possible to include comments inside complicated
933     patterns. Note, however, that this applies only to data characters.
934     Whitespace characters may never appear within special character
935     sequences in a pattern, for example within the sequence (?( which
936 nigel 73 introduces a conditional subpattern.
937 nigel 41
938 nigel 73 PCRE_EXTRA
939 nigel 41
940 nigel 77 This option was invented in order to turn on additional functionality
941     of PCRE that is incompatible with Perl, but it is currently of very
942     little use. When set, any backslash in a pattern that is followed by a
943     letter that has no special meaning causes an error, thus reserving
944     these combinations for future expansion. By default, as in Perl, a
945     backslash followed by a letter with no special meaning is treated as a
946     literal. There are at present no other features controlled by this
947 nigel 73 option. It can also be set by a (?X) option setting within a pattern.
948 nigel 41
949 nigel 77 PCRE_FIRSTLINE
951     If this option is set, an unanchored pattern is required to match
952     before or at the first newline character in the subject string, though
953     the matched text may continue over the newline.
955 nigel 73 PCRE_MULTILINE
956 nigel 41
957 nigel 77 By default, PCRE treats the subject string as consisting of a single
958     line of characters (even if it actually contains newlines). The "start
959     of line" metacharacter (^) matches only at the start of the string,
960     while the "end of line" metacharacter ($) matches only at the end of
961 nigel 75 the string, or before a terminating newline (unless PCRE_DOLLAR_ENDONLY
962     is set). This is the same as Perl.
963 nigel 63
964 nigel 77 When PCRE_MULTILINE it is set, the "start of line" and "end of line"
965     constructs match immediately following or immediately before any new-
966     line in the subject string, respectively, as well as at the very start
967     and end. This is equivalent to Perl's /m option, and it can be changed
968 nigel 73 within a pattern by a (?m) option setting. If there are no "\n" charac-
969 nigel 77 ters in a subject string, or no occurrences of ^ or $ in a pattern,
970 nigel 73 setting PCRE_MULTILINE has no effect.
971 nigel 63
973 nigel 41
974 nigel 73 If this option is set, it disables the use of numbered capturing paren-
975 nigel 77 theses in the pattern. Any opening parenthesis that is not followed by
976     ? behaves as if it were followed by ?: but named parentheses can still
977     be used for capturing (and they acquire numbers in the usual way).
978 nigel 73 There is no equivalent of this option in Perl.
979 nigel 41
980 nigel 73 PCRE_UNGREEDY
981 nigel 41
982 nigel 77 This option inverts the "greediness" of the quantifiers so that they
983     are not greedy by default, but become greedy if followed by "?". It is
984     not compatible with Perl. It can also be set by a (?U) option setting
985 nigel 73 within the pattern.
986 nigel 41
987 nigel 73 PCRE_UTF8
988 nigel 49
989 nigel 77 This option causes PCRE to regard both the pattern and the subject as
990     strings of UTF-8 characters instead of single-byte character strings.
991     However, it is available only when PCRE is built to include UTF-8 sup-
992     port. If not, the use of this option provokes an error. Details of how
993     this option changes the behaviour of PCRE are given in the section on
994 nigel 75 UTF-8 support in the main pcre page.
995 nigel 71
996 nigel 73 PCRE_NO_UTF8_CHECK
997 nigel 71
998 nigel 73 When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is
999 nigel 77 automatically checked. If an invalid UTF-8 sequence of bytes is found,
1000     pcre_compile() returns an error. If you already know that your pattern
1001     is valid, and you want to skip this check for performance reasons, you
1002     can set the PCRE_NO_UTF8_CHECK option. When it is set, the effect of
1003 nigel 73 passing an invalid UTF-8 string as a pattern is undefined. It may cause
1004 nigel 77 your program to crash. Note that this option can also be passed to
1005     pcre_exec() and pcre_dfa_exec(), to suppress the UTF-8 validity check-
1006     ing of subject strings.
1007 nigel 71
1008 nigel 73
1011     The following table lists the error codes than may be returned by
1012     pcre_compile2(), along with the error messages that may be returned by
1013     both compiling functions.
1015     0 no error
1016     1 \ at end of pattern
1017     2 \c at end of pattern
1018     3 unrecognized character follows \
1019     4 numbers out of order in {} quantifier
1020     5 number too big in {} quantifier
1021     6 missing terminating ] for character class
1022     7 invalid escape sequence in character class
1023     8 range out of order in character class
1024     9 nothing to repeat
1025     10 operand of unlimited repeat could match the empty string
1026     11 internal error: unexpected repeat
1027     12 unrecognized character after (?
1028     13 POSIX named classes are supported only within a class
1029     14 missing )
1030     15 reference to non-existent subpattern
1031     16 erroffset passed as NULL
1032     17 unknown option bit(s) set
1033     18 missing ) after comment
1034     19 parentheses nested too deeply
1035     20 regular expression too large
1036     21 failed to get memory
1037     22 unmatched parentheses
1038     23 internal error: code overflow
1039     24 unrecognized character after (?<
1040     25 lookbehind assertion is not fixed length
1041     26 malformed number after (?(
1042     27 conditional group contains more than two branches
1043     28 assertion expected after (?(
1044     29 (?R or (?digits must be followed by )
1045     30 unknown POSIX class name
1046     31 POSIX collating elements are not supported
1047     32 this version of PCRE is not compiled with PCRE_UTF8 support
1048     33 spare error
1049     34 character value in \x{...} sequence is too large
1050     35 invalid condition (?(0)
1051     36 \C not allowed in lookbehind assertion
1052     37 PCRE does not support \L, \l, \N, \U, or \u
1053     38 number after (?C is > 255
1054     39 closing ) for (?C expected
1055     40 recursive call could loop indefinitely
1056     41 unrecognized character after (?P
1057     42 syntax error after (?P
1058     43 two named groups have the same name
1059     44 invalid UTF-8 string
1060     45 support for \P, \p, and \X has not been compiled
1061     46 malformed \P or \p sequence
1062     47 unknown property name after \P or \p
1065 nigel 63 STUDYING A PATTERN
1066 nigel 49
1067 nigel 77 pcre_extra *pcre_study(const pcre *code, int options
1068 nigel 73 const char **errptr);
1069 nigel 63
1070 nigel 77 If a compiled pattern is going to be used several times, it is worth
1071 nigel 75 spending more time analyzing it in order to speed up the time taken for
1072 nigel 77 matching. The function pcre_study() takes a pointer to a compiled pat-
1073 nigel 75 tern as its first argument. If studying the pattern produces additional
1074 nigel 77 information that will help speed up matching, pcre_study() returns a
1075     pointer to a pcre_extra block, in which the study_data field points to
1076 nigel 75 the results of the study.
1077 nigel 41
1078 nigel 75 The returned value from pcre_study() can be passed directly to
1079 nigel 77 pcre_exec(). However, a pcre_extra block also contains other fields
1080     that can be set by the caller before the block is passed; these are
1081 nigel 75 described below in the section on matching a pattern.
1082 nigel 63
1083 nigel 77 If studying the pattern does not produce any additional information
1084 nigel 75 pcre_study() returns NULL. In that circumstance, if the calling program
1085 nigel 77 wants to pass any of the other fields to pcre_exec(), it must set up
1086 nigel 75 its own pcre_extra block.
1087 nigel 41
1088 nigel 77 The second argument of pcre_study() contains option bits. At present,
1089 nigel 75 no options are defined, and this argument should always be zero.
1091 nigel 77 The third argument for pcre_study() is a pointer for an error message.
1092     If studying succeeds (even if no data is returned), the variable it
1093     points to is set to NULL. Otherwise it points to a textual error mes-
1094     sage. You should therefore test the error pointer for NULL after call-
1095 nigel 73 ing pcre_study(), to be sure that it has run successfully.
1096 nigel 41
1097 nigel 73 This is a typical call to pcre_study():
1098 nigel 53
1099 nigel 73 pcre_extra *pe;
1100     pe = pcre_study(
1101     re, /* result of pcre_compile() */
1102     0, /* no options exist */
1103     &error); /* set to NULL or points to a message */
1104 nigel 53
1105 nigel 73 At present, studying a pattern is useful only for non-anchored patterns
1106 nigel 77 that do not have a single fixed starting character. A bitmap of possi-
1107 nigel 75 ble starting bytes is created.
1108 nigel 41
1110 nigel 63 LOCALE SUPPORT
1111 nigel 41
1112 nigel 77 PCRE handles caseless matching, and determines whether characters are
1113     letters digits, or whatever, by reference to a set of tables, indexed
1114     by character value. When running in UTF-8 mode, this applies only to
1115     characters with codes less than 128. Higher-valued codes never match
1116     escapes such as \w or \d, but can be tested with \p if PCRE is built
1117     with Unicode character property support.
1118 nigel 41
1119 nigel 77 An internal set of tables is created in the default C locale when PCRE
1120     is built. This is used when the final argument of pcre_compile() is
1121     NULL, and is sufficient for many applications. An alternative set of
1122     tables can, however, be supplied. These may be created in a different
1123     locale from the default. As more and more applications change to using
1124 nigel 75 Unicode, the need for this locale support is expected to die away.
1125 nigel 41
1126 nigel 77 External tables are built by calling the pcre_maketables() function,
1127     which has no arguments, in the relevant locale. The result can then be
1128     passed to pcre_compile() or pcre_exec() as often as necessary. For
1129     example, to build and use tables that are appropriate for the French
1130     locale (where accented characters with values greater than 128 are
1131 nigel 75 treated as letters), the following code could be used:
1133     setlocale(LC_CTYPE, "fr_FR");
1134 nigel 73 tables = pcre_maketables();
1135     re = pcre_compile(..., tables);
1136 nigel 41
1137 nigel 77 When pcre_maketables() runs, the tables are built in memory that is
1138     obtained via pcre_malloc. It is the caller's responsibility to ensure
1139     that the memory containing the tables remains available for as long as
1140 nigel 75 it is needed.
1141 nigel 41
1142 nigel 75 The pointer that is passed to pcre_compile() is saved with the compiled
1143 nigel 77 pattern, and the same tables are used via this pointer by pcre_study()
1144 nigel 75 and normally also by pcre_exec(). Thus, by default, for any single pat-
1145     tern, compilation, studying and matching all happen in the same locale,
1146     but different patterns can be compiled in different locales.
1147 nigel 41
1148 nigel 77 It is possible to pass a table pointer or NULL (indicating the use of
1149     the internal tables) to pcre_exec(). Although not intended for this
1150     purpose, this facility could be used to match a pattern in a different
1151 nigel 75 locale from the one in which it was compiled. Passing table pointers at
1152     run time is discussed below in the section on matching a pattern.
1156 nigel 41
1157 nigel 73 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
1158     int what, void *where);
1159 nigel 63
1160 nigel 77 The pcre_fullinfo() function returns information about a compiled pat-
1161 nigel 73 tern. It replaces the obsolete pcre_info() function, which is neverthe-
1162     less retained for backwards compability (and is documented below).
1163 nigel 43
1164 nigel 77 The first argument for pcre_fullinfo() is a pointer to the compiled
1165     pattern. The second argument is the result of pcre_study(), or NULL if
1166     the pattern was not studied. The third argument specifies which piece
1167     of information is required, and the fourth argument is a pointer to a
1168     variable to receive the data. The yield of the function is zero for
1169 nigel 73 success, or one of the following negative numbers:
1170 nigel 41
1171 nigel 73 PCRE_ERROR_NULL the argument code was NULL
1172     the argument where was NULL
1173     PCRE_ERROR_BADMAGIC the "magic number" was not found
1174     PCRE_ERROR_BADOPTION the value of what was invalid
1175 nigel 53
1176 nigel 77 The "magic number" is placed at the start of each compiled pattern as
1177     an simple check against passing an arbitrary memory pointer. Here is a
1178     typical call of pcre_fullinfo(), to obtain the length of the compiled
1179 nigel 75 pattern:
1180 nigel 53
1181 nigel 73 int rc;
1182     unsigned long int length;
1183     rc = pcre_fullinfo(
1184     re, /* result of pcre_compile() */
1185     pe, /* result of pcre_study(), or NULL */
1186     PCRE_INFO_SIZE, /* what is required */
1187     &length); /* where to put the data */
1188 nigel 43
1189 nigel 77 The possible values for the third argument are defined in pcre.h, and
1190 nigel 73 are as follows:
1191 nigel 43
1193 nigel 41
1194 nigel 77 Return the number of the highest back reference in the pattern. The
1195     fourth argument should point to an int variable. Zero is returned if
1196 nigel 73 there are no back references.
1197 nigel 43
1199 nigel 43
1200 nigel 77 Return the number of capturing subpatterns in the pattern. The fourth
1201 nigel 73 argument should point to an int variable.
1202 nigel 43
1204 nigel 75
1205 nigel 77 Return a pointer to the internal default character tables within PCRE.
1206     The fourth argument should point to an unsigned char * variable. This
1207 nigel 75 information call is provided for internal use by the pcre_study() func-
1208 nigel 77 tion. External callers can cause PCRE to use its internal tables by
1209 nigel 75 passing a NULL table pointer.
1212 nigel 43
1213 nigel 77 Return information about the first byte of any matched string, for a
1214     non-anchored pattern. (This option used to be called
1215     PCRE_INFO_FIRSTCHAR; the old name is still recognized for backwards
1216 nigel 73 compatibility.)
1217 nigel 41
1218 nigel 77 If there is a fixed first byte, for example, from a pattern such as
1219     (cat|cow|coyote), it is returned in the integer pointed to by where.
1220 nigel 73 Otherwise, if either
1221 nigel 41
1222 nigel 77 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
1223 nigel 73 branch starts with "^", or
1224 nigel 43
1225 nigel 73 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
1226     set (if it were set, the pattern would be anchored),
1227 nigel 41
1228 nigel 77 -1 is returned, indicating that the pattern matches only at the start
1229     of a subject string or after any newline within the string. Otherwise
1230 nigel 73 -2 is returned. For anchored patterns, -2 is returned.
1231 nigel 41
1233 nigel 41
1234 nigel 77 If the pattern was studied, and this resulted in the construction of a
1235 nigel 73 256-bit table indicating a fixed set of bytes for the first byte in any
1236 nigel 77 matching string, a pointer to the table is returned. Otherwise NULL is
1237     returned. The fourth argument should point to an unsigned char * vari-
1238 nigel 73 able.
1239 nigel 43
1241 nigel 43
1242 nigel 77 Return the value of the rightmost literal byte that must exist in any
1243     matched string, other than at its start, if such a byte has been
1244 nigel 73 recorded. The fourth argument should point to an int variable. If there
1245 nigel 77 is no such byte, -1 is returned. For anchored patterns, a last literal
1246     byte is recorded only if it follows something of variable length. For
1247 nigel 73 example, for the pattern /^a\d+z\d+/ the returned value is "z", but for
1248     /^a\dz\d/ the returned value is -1.
1249 nigel 63
1253 nigel 63
1254 nigel 77 PCRE supports the use of named as well as numbered capturing parenthe-
1255     ses. The names are just an additional way of identifying the parenthe-
1256 nigel 75 ses, which still acquire numbers. A convenience function called
1257 nigel 77 pcre_get_named_substring() is provided for extracting an individual
1258     captured substring by name. It is also possible to extract the data
1259     directly, by first converting the name to a number in order to access
1260     the correct pointers in the output vector (described with pcre_exec()
1261     below). To do the conversion, you need to use the name-to-number map,
1262 nigel 75 which is described by these three values.
1263 nigel 63
1264 nigel 73 The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT
1265     gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size
1266 nigel 77 of each entry; both of these return an int value. The entry size
1267     depends on the length of the longest name. PCRE_INFO_NAMETABLE returns
1268     a pointer to the first entry of the table (a pointer to char). The
1269 nigel 73 first two bytes of each entry are the number of the capturing parenthe-
1270 nigel 77 sis, most significant byte first. The rest of the entry is the corre-
1271     sponding name, zero terminated. The names are in alphabetical order.
1272     For example, consider the following pattern (assume PCRE_EXTENDED is
1273 nigel 73 set, so white space - including newlines - is ignored):
1274 nigel 63
1275 nigel 73 (?P<date> (?P<year>(\d\d)?\d\d) -
1276     (?P<month>\d\d) - (?P<day>\d\d) )
1277 nigel 63
1278 nigel 77 There are four named subpatterns, so the table has four entries, and
1279     each entry in the table is eight bytes long. The table is as follows,
1280 nigel 75 with non-printing bytes shows in hexadecimal, and undefined bytes shown
1281     as ??:
1282 nigel 63
1283 nigel 73 00 01 d a t e 00 ??
1284     00 05 d a y 00 ?? ??
1285     00 04 m o n t h 00
1286     00 02 y e a r 00 ??
1287 nigel 63
1288 nigel 77 When writing code to extract data from named subpatterns using the
1289 nigel 75 name-to-number map, remember that the length of each entry is likely to
1290     be different for each compiled pattern.
1291 nigel 63
1292 nigel 73 PCRE_INFO_OPTIONS
1293 nigel 63
1294 nigel 77 Return a copy of the options with which the pattern was compiled. The
1295     fourth argument should point to an unsigned long int variable. These
1296 nigel 73 option bits are those specified in the call to pcre_compile(), modified
1297     by any top-level option settings within the pattern itself.
1298 nigel 63
1299 nigel 77 A pattern is automatically anchored by PCRE if all of its top-level
1300 nigel 73 alternatives begin with one of the following:
1301 nigel 63
1302 nigel 73 ^ unless PCRE_MULTILINE is set
1303     \A always
1304     \G always
1305     .* if PCRE_DOTALL is set and there are no back
1306     references to the subpattern in which .* appears
1307 nigel 63
1308 nigel 73 For such patterns, the PCRE_ANCHORED bit is set in the options returned
1309     by pcre_fullinfo().
1310 nigel 63
1311 nigel 73 PCRE_INFO_SIZE
1312 nigel 63
1313 nigel 77 Return the size of the compiled pattern, that is, the value that was
1314 nigel 73 passed as the argument to pcre_malloc() when PCRE was getting memory in
1315     which to place the compiled data. The fourth argument should point to a
1316     size_t variable.
1317 nigel 63
1319 nigel 63
1320 nigel 75 Return the size of the data block pointed to by the study_data field in
1321 nigel 77 a pcre_extra block. That is, it is the value that was passed to
1322 nigel 73 pcre_malloc() when PCRE was getting memory into which to place the data
1323 nigel 77 created by pcre_study(). The fourth argument should point to a size_t
1324 nigel 73 variable.
1325 nigel 63
1326 nigel 73
1329 nigel 73 int pcre_info(const pcre *code, int *optptr, int *firstcharptr);
1330 nigel 63
1331 nigel 77 The pcre_info() function is now obsolete because its interface is too
1332     restrictive to return all the available data about a compiled pattern.
1333     New programs should use pcre_fullinfo() instead. The yield of
1334     pcre_info() is the number of capturing subpatterns, or one of the fol-
1335 nigel 73 lowing negative numbers:
1336 nigel 43
1337 nigel 73 PCRE_ERROR_NULL the argument code was NULL
1338     PCRE_ERROR_BADMAGIC the "magic number" was not found
1339 nigel 43
1340 nigel 77 If the optptr argument is not NULL, a copy of the options with which
1341     the pattern was compiled is placed in the integer it points to (see
1342 nigel 73 PCRE_INFO_OPTIONS above).
1343 nigel 43
1344 nigel 77 If the pattern is not anchored and the firstcharptr argument is not
1345     NULL, it is used to pass back information about the first character of
1346 nigel 73 any matched string (see PCRE_INFO_FIRSTBYTE above).
1347 nigel 43
1349 nigel 77 REFERENCE COUNTS
1350 nigel 53
1351 nigel 77 int pcre_refcount(pcre *code, int adjust);
1353     The pcre_refcount() function is used to maintain a reference count in
1354     the data block that contains a compiled pattern. It is provided for the
1355     benefit of applications that operate in an object-oriented manner,
1356     where different parts of the application may be using the same compiled
1357     pattern, but you want to free the block when they are all done.
1359     When a pattern is compiled, the reference count field is initialized to
1360     zero. It is changed only by calling this function, whose action is to
1361     add the adjust value (which may be positive or negative) to it. The
1362     yield of the function is the new value. However, the value of the count
1363     is constrained to lie between 0 and 65535, inclusive. If the new value
1364     is outside these limits, it is forced to the appropriate limit value.
1366     Except when it is zero, the reference count is not correctly preserved
1367     if a pattern is compiled on one host and then transferred to a host
1368     whose byte-order is different. (This seems a highly unlikely scenario.)
1373 nigel 73 int pcre_exec(const pcre *code, const pcre_extra *extra,
1374     const char *subject, int length, int startoffset,
1375     int options, int *ovector, int ovecsize);
1376 nigel 53
1377 nigel 77 The function pcre_exec() is called to match a subject string against a
1378     compiled pattern, which is passed in the code argument. If the pattern
1379 nigel 75 has been studied, the result of the study should be passed in the extra
1380 nigel 77 argument. This function is the main matching facility of the library,
1381     and it operates in a Perl-like manner. For specialist use there is also
1382     an alternative matching function, which is described below in the sec-
1383     tion about the pcre_dfa_exec() function.
1384 nigel 41
1385 nigel 75 In most applications, the pattern will have been compiled (and option-
1386     ally studied) in the same process that calls pcre_exec(). However, it
1387     is possible to save compiled patterns and study data, and then use them
1388     later in different processes, possibly even on different hosts. For a
1389     discussion about this, see the pcreprecompile documentation.
1391 nigel 73 Here is an example of a simple call to pcre_exec():
1392 nigel 53
1393 nigel 73 int rc;
1394     int ovector[30];
1395     rc = pcre_exec(
1396     re, /* result of pcre_compile() */
1397     NULL, /* we didn't study the pattern */
1398     "some string", /* the subject string */
1399     11, /* the length of the subject string */
1400     0, /* start at offset 0 in the subject */
1401     0, /* default options */
1402 nigel 75 ovector, /* vector of integers for substring information */
1403 nigel 77 30); /* number of elements (NOT size in bytes) */
1404 nigel 53
1405 nigel 75 Extra data for pcre_exec()
1406 nigel 63
1407 nigel 77 If the extra argument is not NULL, it must point to a pcre_extra data
1408     block. The pcre_study() function returns such a block (when it doesn't
1409     return NULL), but you can also create one for yourself, and pass addi-
1410     tional information in it. The fields in a pcre_extra block are as fol-
1411 nigel 75 lows:
1413 nigel 73 unsigned long int flags;
1414     void *study_data;
1415     unsigned long int match_limit;
1416     void *callout_data;
1417 nigel 75 const unsigned char *tables;
1418 nigel 63
1419 nigel 77 The flags field is a bitmap that specifies which of the other fields
1420 nigel 73 are set. The flag bits are:
1421 nigel 63
1425 nigel 75 PCRE_EXTRA_TABLES
1426 nigel 63
1427 nigel 77 Other flag bits should be set to zero. The study_data field is set in
1428     the pcre_extra block that is returned by pcre_study(), together with
1429 nigel 75 the appropriate flag bit. You should not set this yourself, but you may
1430 nigel 77 add to the block by setting the other fields and their corresponding
1431 nigel 75 flag bits.
1432 nigel 63
1433 nigel 73 The match_limit field provides a means of preventing PCRE from using up
1434 nigel 77 a vast amount of resources when running patterns that are not going to
1435     match, but which have a very large number of possibilities in their
1436     search trees. The classic example is the use of nested unlimited
1437 nigel 75 repeats.
1438 nigel 63
1439 nigel 77 Internally, PCRE uses a function called match() which it calls repeat-
1440     edly (sometimes recursively). The limit is imposed on the number of
1441     times this function is called during a match, which has the effect of
1442     limiting the amount of recursion and backtracking that can take place.
1443 nigel 75 For patterns that are not anchored, the count starts from zero for each
1444     position in the subject string.
1446 nigel 77 The default limit for the library can be set when PCRE is built; the
1447     default default is 10 million, which handles all but the most extreme
1448     cases. You can reduce the default by suppling pcre_exec() with a
1449     pcre_extra block in which match_limit is set to a smaller value, and
1450     PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is
1451 nigel 73 exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.
1452 nigel 63
1453 nigel 77 The pcre_callout field is used in conjunction with the "callout" fea-
1454 nigel 73 ture, which is described in the pcrecallout documentation.
1455 nigel 63
1456 nigel 77 The tables field is used to pass a character tables pointer to
1457     pcre_exec(); this overrides the value that is stored with the compiled
1458     pattern. A non-NULL value is stored with the compiled pattern only if
1459     custom tables were supplied to pcre_compile() via its tableptr argu-
1460 nigel 75 ment. If NULL is passed to pcre_exec() using this mechanism, it forces
1461 nigel 77 PCRE's internal tables to be used. This facility is helpful when re-
1462     using patterns that have been saved after compiling with an external
1463     set of tables, because the external tables might be at a different
1464     address when pcre_exec() is called. See the pcreprecompile documenta-
1465 nigel 75 tion for a discussion of saving compiled patterns for later use.
1466 nigel 41
1467 nigel 75 Option bits for pcre_exec()
1468 nigel 71
1469 nigel 77 The unused bits of the options argument for pcre_exec() must be zero.
1470     The only bits that may be set are PCRE_ANCHORED, PCRE_NOTBOL,
1472 nigel 41
1473 nigel 75 PCRE_ANCHORED
1474 nigel 41
1475 nigel 77 The PCRE_ANCHORED option limits pcre_exec() to matching at the first
1476     matching position. If a pattern was compiled with PCRE_ANCHORED, or
1477     turned out to be anchored by virtue of its contents, it cannot be made
1478 nigel 75 unachored at matching time.
1480 nigel 73 PCRE_NOTBOL
1481 nigel 41
1482 nigel 75 This option specifies that first character of the subject string is not
1483 nigel 77 the beginning of a line, so the circumflex metacharacter should not
1484     match before it. Setting this without PCRE_MULTILINE (at compile time)
1485     causes circumflex never to match. This option affects only the behav-
1486     iour of the circumflex metacharacter. It does not affect \A.
1487 nigel 41
1488 nigel 73 PCRE_NOTEOL
1489 nigel 41
1490 nigel 75 This option specifies that the end of the subject string is not the end
1491 nigel 77 of a line, so the dollar metacharacter should not match it nor (except
1492     in multiline mode) a newline immediately before it. Setting this with-
1493 nigel 75 out PCRE_MULTILINE (at compile time) causes dollar never to match. This
1494 nigel 77 option affects only the behaviour of the dollar metacharacter. It does
1495 nigel 75 not affect \Z or \z.
1496 nigel 41
1497 nigel 73 PCRE_NOTEMPTY
1498 nigel 41
1499 nigel 73 An empty string is not considered to be a valid match if this option is
1500 nigel 77 set. If there are alternatives in the pattern, they are tried. If all
1501     the alternatives match the empty string, the entire match fails. For
1502 nigel 73 example, if the pattern
1503 nigel 41
1504 nigel 73 a?b?
1505 nigel 41
1506 nigel 77 is applied to a string not beginning with "a" or "b", it matches the
1507     empty string at the start of the subject. With PCRE_NOTEMPTY set, this
1508 nigel 73 match is not valid, so PCRE searches further into the string for occur-
1509     rences of "a" or "b".
1510 nigel 41
1511 nigel 73 Perl has no direct equivalent of PCRE_NOTEMPTY, but it does make a spe-
1512 nigel 77 cial case of a pattern match of the empty string within its split()
1513     function, and when using the /g modifier. It is possible to emulate
1514 nigel 73 Perl's behaviour after matching a null string by first trying the match
1515 nigel 75 again at the same offset with PCRE_NOTEMPTY and PCRE_ANCHORED, and then
1516 nigel 77 if that fails by advancing the starting offset (see below) and trying
1517 nigel 75 an ordinary match again. There is some code that demonstrates how to do
1518     this in the pcredemo.c sample program.
1519 nigel 41
1520 nigel 75 PCRE_NO_UTF8_CHECK
1522     When PCRE_UTF8 is set at compile time, the validity of the subject as a
1523 nigel 77 UTF-8 string is automatically checked when pcre_exec() is subsequently
1524     called. The value of startoffset is also checked to ensure that it
1525     points to the start of a UTF-8 character. If an invalid UTF-8 sequence
1526 nigel 75 of bytes is found, pcre_exec() returns the error PCRE_ERROR_BADUTF8. If
1527 nigel 77 startoffset contains an invalid value, PCRE_ERROR_BADUTF8_OFFSET is
1528 nigel 75 returned.
1530 nigel 77 If you already know that your subject is valid, and you want to skip
1531     these checks for performance reasons, you can set the
1532     PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to
1533     do this for the second and subsequent calls to pcre_exec() if you are
1534     making repeated calls to find all the matches in a single subject
1535     string. However, you should be sure that the value of startoffset
1536     points to the start of a UTF-8 character. When PCRE_NO_UTF8_CHECK is
1537     set, the effect of passing an invalid UTF-8 string as a subject, or a
1538     value of startoffset that does not point to the start of a UTF-8 char-
1539 nigel 75 acter, is undefined. Your program may crash.
1543 nigel 77 This option turns on the partial matching feature. If the subject
1544     string fails to match the pattern, but at some point during the match-
1545     ing process the end of the subject was reached (that is, the subject
1546     partially matches the pattern and the failure to match occurred only
1547     because there were not enough subject characters), pcre_exec() returns
1549     used, there are restrictions on what may appear in the pattern. These
1550 nigel 75 are discussed in the pcrepartial documentation.
1552     The string to be matched by pcre_exec()
1554 nigel 77 The subject string is passed to pcre_exec() as a pointer in subject, a
1555     length in length, and a starting byte offset in startoffset. In UTF-8
1556     mode, the byte offset must point to the start of a UTF-8 character.
1557     Unlike the pattern string, the subject may contain binary zero bytes.
1558     When the starting offset is zero, the search for a match starts at the
1559 nigel 75 beginning of the subject, and this is by far the most common case.
1560 nigel 63
1561 nigel 77 A non-zero starting offset is useful when searching for another match
1562     in the same subject by calling pcre_exec() again after a previous suc-
1563     cess. Setting startoffset differs from just passing over a shortened
1564     string and setting PCRE_NOTBOL in the case of a pattern that begins
1565 nigel 73 with any kind of lookbehind. For example, consider the pattern
1566 nigel 41
1567 nigel 73 \Biss\B
1568 nigel 41
1569 nigel 77 which finds occurrences of "iss" in the middle of words. (\B matches
1570     only if the current position in the subject is not a word boundary.)
1571     When applied to the string "Mississipi" the first call to pcre_exec()
1572     finds the first occurrence. If pcre_exec() is called again with just
1573     the remainder of the subject, namely "issipi", it does not match,
1574 nigel 73 because \B is always false at the start of the subject, which is deemed
1575 nigel 77 to be a word boundary. However, if pcre_exec() is passed the entire
1576 nigel 75 string again, but with startoffset set to 4, it finds the second occur-
1577 nigel 77 rence of "iss" because it is able to look behind the starting point to
1578 nigel 75 discover that it is preceded by a letter.
1579 nigel 41
1580 nigel 77 If a non-zero starting offset is passed when the pattern is anchored,
1581 nigel 75 one attempt to match at the given offset is made. This can only succeed
1582 nigel 77 if the pattern does not require the match to be at the start of the
1583 nigel 75 subject.
1584 nigel 41
1585 nigel 75 How pcre_exec() returns captured substrings
1587 nigel 77 In general, a pattern matches a certain portion of the subject, and in
1588     addition, further substrings from the subject may be picked out by
1589     parts of the pattern. Following the usage in Jeffrey Friedl's book,
1590     this is called "capturing" in what follows, and the phrase "capturing
1591     subpattern" is used for a fragment of a pattern that picks out a sub-
1592     string. PCRE supports several other kinds of parenthesized subpattern
1593 nigel 73 that do not cause substrings to be captured.
1594 nigel 65
1595 nigel 77 Captured substrings are returned to the caller via a vector of integer
1596     offsets whose address is passed in ovector. The number of elements in
1597     the vector is passed in ovecsize, which must be a non-negative number.
1598 nigel 75 Note: this argument is NOT the size of ovector in bytes.
1599 nigel 41
1600 nigel 77 The first two-thirds of the vector is used to pass back captured sub-
1601     strings, each substring using a pair of integers. The remaining third
1602     of the vector is used as workspace by pcre_exec() while matching cap-
1603     turing subpatterns, and is not available for passing back information.
1604     The length passed in ovecsize should always be a multiple of three. If
1605 nigel 75 it is not, it is rounded down.
1607 nigel 77 When a match is successful, information about captured substrings is
1608     returned in pairs of integers, starting at the beginning of ovector,
1609     and continuing up to two-thirds of its length at the most. The first
1610 nigel 73 element of a pair is set to the offset of the first character in a sub-
1611 nigel 77 string, and the second is set to the offset of the first character
1612     after the end of a substring. The first pair, ovector[0] and ovec-
1613     tor[1], identify the portion of the subject string matched by the
1614     entire pattern. The next pair is used for the first capturing subpat-
1615     tern, and so on. The value returned by pcre_exec() is the number of
1616     pairs that have been set. If there are no capturing subpatterns, the
1617     return value from a successful match is 1, indicating that just the
1618 nigel 73 first pair of offsets has been set.
1619 nigel 41
1620 nigel 77 Some convenience functions are provided for extracting the captured
1621     substrings as separate strings. These are described in the following
1622 nigel 73 section.
1623 nigel 41
1624 nigel 77 It is possible for an capturing subpattern number n+1 to match some
1625     part of the subject when subpattern n has not been used at all. For
1626 nigel 73 example, if the string "abc" is matched against the pattern (a|(z))(bc)
1627 nigel 77 subpatterns 1 and 3 are matched, but 2 is not. When this happens, both
1628 nigel 73 offset values corresponding to the unused subpattern are set to -1.
1629 nigel 41
1630 nigel 73 If a capturing subpattern is matched repeatedly, it is the last portion
1631 nigel 75 of the string that it matched that is returned.
1632 nigel 41
1633 nigel 77 If the vector is too small to hold all the captured substring offsets,
1634 nigel 75 it is used as far as possible (up to two-thirds of its length), and the
1635 nigel 77 function returns a value of zero. In particular, if the substring off-
1636 nigel 75 sets are not of interest, pcre_exec() may be called with ovector passed
1637 nigel 77 as NULL and ovecsize as zero. However, if the pattern contains back
1638     references and the ovector is not big enough to remember the related
1639     substrings, PCRE has to get additional memory for use during matching.
1640 nigel 73 Thus it is usually advisable to supply an ovector.
1641 nigel 41
1642 nigel 77 Note that pcre_info() can be used to find out how many capturing sub-
1643 nigel 73 patterns there are in a compiled pattern. The smallest size for ovector
1644 nigel 77 that will allow for n captured substrings, in addition to the offsets
1645 nigel 73 of the substring matched by the whole pattern, is (n+1)*3.
1646 nigel 41
1647 nigel 75 Return values from pcre_exec()
1649 nigel 77 If pcre_exec() fails, it returns a negative number. The following are
1650 nigel 73 defined in the header file:
1651 nigel 41
1652 nigel 73 PCRE_ERROR_NOMATCH (-1)
1653 nigel 41
1654 nigel 73 The subject string did not match the pattern.
1655 nigel 41
1656 nigel 73 PCRE_ERROR_NULL (-2)
1657 nigel 41
1658 nigel 77 Either code or subject was passed as NULL, or ovector was NULL and
1659 nigel 73 ovecsize was not zero.
1660 nigel 41
1661 nigel 73 PCRE_ERROR_BADOPTION (-3)
1662 nigel 41
1663 nigel 73 An unrecognized bit was set in the options argument.
1664 nigel 41
1665 nigel 73 PCRE_ERROR_BADMAGIC (-4)
1666 nigel 41
1667 nigel 77 PCRE stores a 4-byte "magic number" at the start of the compiled code,
1668 nigel 75 to catch the case when it is passed a junk pointer and to detect when a
1669     pattern that was compiled in an environment of one endianness is run in
1670 nigel 77 an environment with the other endianness. This is the error that PCRE
1671 nigel 75 gives when the magic number is not present.
1672 nigel 41
1673 nigel 73 PCRE_ERROR_UNKNOWN_NODE (-5)
1674 nigel 41
1675 nigel 73 While running the pattern match, an unknown item was encountered in the
1676 nigel 77 compiled pattern. This error could be caused by a bug in PCRE or by
1677 nigel 73 overwriting of the compiled pattern.
1678 nigel 41
1679 nigel 73 PCRE_ERROR_NOMEMORY (-6)
1680 nigel 41
1681 nigel 77 If a pattern contains back references, but the ovector that is passed
1682 nigel 73 to pcre_exec() is not big enough to remember the referenced substrings,
1683 nigel 77 PCRE gets a block of memory at the start of matching to use for this
1684     purpose. If the call via pcre_malloc() fails, this error is given. The
1685 nigel 75 memory is automatically freed at the end of matching.
1686 nigel 41
1687 nigel 73 PCRE_ERROR_NOSUBSTRING (-7)
1688 nigel 53
1689 nigel 77 This error is used by the pcre_copy_substring(), pcre_get_substring(),
1690 nigel 73 and pcre_get_substring_list() functions (see below). It is never
1691     returned by pcre_exec().
1692 nigel 63
1693 nigel 73 PCRE_ERROR_MATCHLIMIT (-8)
1694 nigel 63
1695 nigel 77 The recursion and backtracking limit, as specified by the match_limit
1696     field in a pcre_extra structure (or defaulted) was reached. See the
1697 nigel 73 description above.
1698 nigel 63
1699 nigel 73 PCRE_ERROR_CALLOUT (-9)
1700 nigel 63
1701 nigel 73 This error is never generated by pcre_exec() itself. It is provided for
1702 nigel 77 use by callout functions that want to yield a distinctive error code.
1703 nigel 73 See the pcrecallout documentation for details.
1704 nigel 71
1705 nigel 73 PCRE_ERROR_BADUTF8 (-10)
1706 nigel 71
1707 nigel 77 A string that contains an invalid UTF-8 byte sequence was passed as a
1708 nigel 73 subject.
1712     The UTF-8 byte sequence that was passed as a subject was valid, but the
1713 nigel 77 value of startoffset did not point to the beginning of a UTF-8 charac-
1714 nigel 73 ter.
1716 nigel 77 PCRE_ERROR_PARTIAL (-12)
1717 nigel 73
1718 nigel 77 The subject string did not match, but it did match partially. See the
1719 nigel 75 pcrepartial documentation for details of partial matching.
1721 nigel 77 PCRE_ERROR_BADPARTIAL (-13)
1722 nigel 75
1723 nigel 77 The PCRE_PARTIAL option was used with a compiled pattern containing
1724     items that are not supported for partial matching. See the pcrepartial
1725 nigel 75 documentation for details of partial matching.
1727 nigel 77 PCRE_ERROR_INTERNAL (-14)
1728 nigel 75
1729 nigel 77 An unexpected internal error has occurred. This error could be caused
1730 nigel 75 by a bug in PCRE or by overwriting of the compiled pattern.
1732 nigel 77 PCRE_ERROR_BADCOUNT (-15)
1733 nigel 75
1734 nigel 77 This error is given if the value of the ovecsize argument is negative.
1735 nigel 75
1739 nigel 73 int pcre_copy_substring(const char *subject, int *ovector,
1740     int stringcount, int stringnumber, char *buffer,
1741     int buffersize);
1742 nigel 63
1743 nigel 73 int pcre_get_substring(const char *subject, int *ovector,
1744     int stringcount, int stringnumber,
1745     const char **stringptr);
1746 nigel 63
1747 nigel 73 int pcre_get_substring_list(const char *subject,
1748     int *ovector, int stringcount, const char ***listptr);
1749 nigel 63
1750 nigel 77 Captured substrings can be accessed directly by using the offsets
1751     returned by pcre_exec() in ovector. For convenience, the functions
1752 nigel 73 pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub-
1753 nigel 77 string_list() are provided for extracting captured substrings as new,
1754     separate, zero-terminated strings. These functions identify substrings
1755     by number. The next section describes functions for extracting named
1756     substrings. A substring that contains a binary zero is correctly
1757     extracted and has a further zero added on the end, but the result is
1758 nigel 73 not, of course, a C string.
1759 nigel 41
1760 nigel 77 The first three arguments are the same for all three of these func-
1761     tions: subject is the subject string that has just been successfully
1762 nigel 73 matched, ovector is a pointer to the vector of integer offsets that was
1763     passed to pcre_exec(), and stringcount is the number of substrings that
1764 nigel 77 were captured by the match, including the substring that matched the
1765 nigel 75 entire regular expression. This is the value returned by pcre_exec() if
1766 nigel 77 it is greater than zero. If pcre_exec() returned zero, indicating that
1767     it ran out of space in ovector, the value passed as stringcount should
1768 nigel 75 be the number of elements in the vector divided by three.
1769 nigel 41
1770 nigel 77 The functions pcre_copy_substring() and pcre_get_substring() extract a
1771     single substring, whose number is given as stringnumber. A value of
1772     zero extracts the substring that matched the entire pattern, whereas
1773     higher values extract the captured substrings. For pcre_copy_sub-
1774     string(), the string is placed in buffer, whose length is given by
1775     buffersize, while for pcre_get_substring() a new block of memory is
1776     obtained via pcre_malloc, and its address is returned via stringptr.
1777     The yield of the function is the length of the string, not including
1778 nigel 73 the terminating zero, or one of
1779 nigel 41
1780 nigel 73 PCRE_ERROR_NOMEMORY (-6)
1781 nigel 41
1782 nigel 77 The buffer was too small for pcre_copy_substring(), or the attempt to
1783 nigel 73 get memory failed for pcre_get_substring().
1784 nigel 41
1785 nigel 73 PCRE_ERROR_NOSUBSTRING (-7)
1786 nigel 41
1787 nigel 73 There is no substring whose number is stringnumber.
1788 nigel 41
1789 nigel 77 The pcre_get_substring_list() function extracts all available sub-
1790     strings and builds a list of pointers to them. All this is done in a
1791 nigel 75 single block of memory that is obtained via pcre_malloc. The address of
1792 nigel 77 the memory block is returned via listptr, which is also the start of
1793     the list of string pointers. The end of the list is marked by a NULL
1794 nigel 73 pointer. The yield of the function is zero if all went well, or
1795 nigel 41
1796 nigel 73 PCRE_ERROR_NOMEMORY (-6)
1797 nigel 41
1798 nigel 73 if the attempt to get the memory block failed.
1799 nigel 41
1800 nigel 77 When any of these functions encounter a substring that is unset, which
1801     can happen when capturing subpattern number n+1 matches some part of
1802     the subject, but subpattern n has not been used at all, they return an
1803 nigel 73 empty string. This can be distinguished from a genuine zero-length sub-
1804 nigel 77 string by inspecting the appropriate offset in ovector, which is nega-
1805 nigel 73 tive for unset substrings.
1806 nigel 41
1807 nigel 77 The two convenience functions pcre_free_substring() and pcre_free_sub-
1808     string_list() can be used to free the memory returned by a previous
1809 nigel 75 call of pcre_get_substring() or pcre_get_substring_list(), respec-
1810 nigel 77 tively. They do nothing more than call the function pointed to by
1811     pcre_free, which of course could be called directly from a C program.
1812     However, PCRE is used in some situations where it is linked via a spe-
1813 nigel 73 cial interface to another programming language which cannot use
1814 nigel 77 pcre_free directly; it is for these cases that the functions are pro-
1815     vided.
1816 nigel 41
1817 nigel 73
1819 nigel 41
1820 nigel 75 int pcre_get_stringnumber(const pcre *code,
1821     const char *name);
1823 nigel 73 int pcre_copy_named_substring(const pcre *code,
1824     const char *subject, int *ovector,
1825     int stringcount, const char *stringname,
1826     char *buffer, int buffersize);
1827 nigel 41
1828 nigel 73 int pcre_get_named_substring(const pcre *code,
1829     const char *subject, int *ovector,
1830     int stringcount, const char *stringname,
1831     const char **stringptr);
1832 nigel 41
1833 nigel 77 To extract a substring by name, you first have to find associated num-
1834 nigel 75 ber. For example, for this pattern
1835 nigel 41
1836 nigel 75 (a+)b(?<xxx>\d+)...
1837 nigel 63
1838 nigel 75 the number of the subpattern called "xxx" is 2. You can find the number
1839     from the name by calling pcre_get_stringnumber(). The first argument is
1840 nigel 77 the compiled pattern, and the second is the name. The yield of the
1841     function is the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if
1842 nigel 75 there is no subpattern of that name.
1843 nigel 63
1844 nigel 75 Given the number, you can extract the substring directly, or use one of
1845     the functions described in the previous section. For convenience, there
1846     are also two functions that do the whole job.
1848 nigel 77 Most of the arguments of pcre_copy_named_substring() and
1849     pcre_get_named_substring() are the same as those for the similarly
1850     named functions that extract by number. As these are described in the
1851     previous section, they are not re-described here. There are just two
1852 nigel 75 differences:
1853 nigel 63
1854 nigel 77 First, instead of a substring number, a substring name is given. Sec-
1855 nigel 73 ond, there is an extra argument, given at the start, which is a pointer
1856 nigel 77 to the compiled pattern. This is needed in order to gain access to the
1857 nigel 73 name-to-number translation table.
1858 nigel 63
1859 nigel 77 These functions call pcre_get_stringnumber(), and if it succeeds, they
1860     then call pcre_copy_substring() or pcre_get_substring(), as appropri-
1861 nigel 73 ate.
1862 nigel 63
1863 nigel 77
1866     The traditional matching function uses a similar algorithm to Perl,
1867     which stops when it finds the first match, starting at a given point in
1868     the subject. If you want to find all possible matches, or the longest
1869     possible match, consider using the alternative matching function (see
1870     below) instead. If you cannot use the alternative function, but still
1871     need to find all possible matches, you can kludge it up by making use
1872     of the callout facility, which is described in the pcrecallout documen-
1873     tation.
1875     What you have to do is to insert a callout right at the end of the pat-
1876     tern. When your callout function is called, extract and save the cur-
1877     rent matched substring. Then return 1, which forces pcre_exec() to
1878     backtrack and try other alternatives. Ultimately, when it runs out of
1879     matches, pcre_exec() will yield PCRE_ERROR_NOMATCH.
1884     int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
1885     const char *subject, int length, int startoffset,
1886     int options, int *ovector, int ovecsize,
1887     int *workspace, int wscount);
1889     The function pcre_dfa_exec() is called to match a subject string
1890     against a compiled pattern, using a "DFA" matching algorithm. This has
1891     different characteristics to the normal algorithm, and is not compati-
1892     ble with Perl. Some of the features of PCRE patterns are not supported.
1893     Nevertheless, there are times when this kind of matching can be useful.
1894     For a discussion of the two matching algorithms, see the pcrematching
1895     documentation.
1897     The arguments for the pcre_dfa_exec() function are the same as for
1898     pcre_exec(), plus two extras. The ovector argument is used in a differ-
1899     ent way, and this is described below. The other common arguments are
1900     used in the same way as for pcre_exec(), so their description is not
1901     repeated here.
1903     The two additional arguments provide workspace for the function. The
1904     workspace vector should contain at least 20 elements. It is used for
1905     keeping track of multiple paths through the pattern tree. More
1906     workspace will be needed for patterns and subjects where there are a
1907     lot of possible matches.
1909     Here is an example of a simple call to pcre_exec():
1911     int rc;
1912     int ovector[10];
1913     int wspace[20];
1914     rc = pcre_exec(
1915     re, /* result of pcre_compile() */
1916     NULL, /* we didn't study the pattern */
1917     "some string", /* the subject string */
1918     11, /* the length of the subject string */
1919     0, /* start at offset 0 in the subject */
1920     0, /* default options */
1921     ovector, /* vector of integers for substring information */
1922     10, /* number of elements (NOT size in bytes) */
1923     wspace, /* working space vector */
1924     20); /* number of elements (NOT size in bytes) */
1926     Option bits for pcre_dfa_exec()
1928     The unused bits of the options argument for pcre_dfa_exec() must be
1929     zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NOTBOL,
1931     PCRE_DFA_SHORTEST, and PCRE_DFA_RESTART. All but the last three of
1932     these are the same as for pcre_exec(), so their description is not
1933     repeated here.
1937     This has the same general effect as it does for pcre_exec(), but the
1938     details are slightly different. When PCRE_PARTIAL is set for
1939     pcre_dfa_exec(), the return code PCRE_ERROR_NOMATCH is converted into
1940     PCRE_ERROR_PARTIAL if the end of the subject is reached, there have
1941     been no complete matches, but there is still at least one matching pos-
1942     sibility. The portion of the string that provided the partial match is
1943     set as the first matching string.
1947     Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to
1948     stop as soon as it has found one match. Because of the way the DFA
1949     algorithm works, this is necessarily the shortest possible match at the
1950     first possible matching point in the subject string.
1954     When pcre_dfa_exec() is called with the PCRE_PARTIAL option, and
1955     returns a partial match, it is possible to call it again, with addi-
1956     tional subject characters, and have it continue with the same match.
1957     The PCRE_DFA_RESTART option requests this action; when it is set, the
1958     workspace and wscount options must reference the same vector as before
1959     because data about the match so far is left in them after a partial
1960     match. There is more discussion of this facility in the pcrepartial
1961     documentation.
1963     Successful returns from pcre_dfa_exec()
1965     When pcre_dfa_exec() succeeds, it may have matched more than one sub-
1966     string in the subject. Note, however, that all the matches from one run
1967     of the function start at the same point in the subject. The shorter
1968     matches are all initial substrings of the longer matches. For example,
1969     if the pattern
1971     <.*>
1973     is matched against the string
1975     This is <something> <something else> <something further> no more
1977     the three matched strings are
1979     <something>
1980     <something> <something else>
1981     <something> <something else> <something further>
1983     On success, the yield of the function is a number greater than zero,
1984     which is the number of matched substrings. The substrings themselves
1985     are returned in ovector. Each string uses two elements; the first is
1986     the offset to the start, and the second is the offset to the end. All
1987     the strings have the same start offset. (Space could have been saved by
1988     giving this only once, but it was decided to retain some compatibility
1989     with the way pcre_exec() returns data, even though the meaning of the
1990     strings is different.)
1992     The strings are returned in reverse order of length; that is, the long-
1993     est matching string is given first. If there were too many matches to
1994     fit into ovector, the yield of the function is zero, and the vector is
1995     filled with the longest matches.
1997     Error returns from pcre_dfa_exec()
1999     The pcre_dfa_exec() function returns a negative number when it fails.
2000     Many of the errors are the same as for pcre_exec(), and these are
2001     described above. There are in addition the following errors that are
2002     specific to pcre_dfa_exec():
2004     PCRE_ERROR_DFA_UITEM (-16)
2006     This return is given if pcre_dfa_exec() encounters an item in the pat-
2007     tern that it does not support, for instance, the use of \C or a back
2008     reference.
2010     PCRE_ERROR_DFA_UCOND (-17)
2012     This return is given if pcre_dfa_exec() encounters a condition item in
2013     a pattern that uses a back reference for the condition. This is not
2014     supported.
2018     This return is given if pcre_dfa_exec() is called with an extra block
2019     that contains a setting of the match_limit field. This is not supported
2020     (it is meaningless).
2022     PCRE_ERROR_DFA_WSSIZE (-19)
2024     This return is given if pcre_dfa_exec() runs out of space in the
2025     workspace vector.
2029     When a recursive subpattern is processed, the matching function calls
2030     itself recursively, using private vectors for ovector and workspace.
2031     This error is given if the output vector is not large enough. This
2032     should be extremely rare, as a vector of size 1000 is used.
2034     Last updated: 16 May 2005
2035     Copyright (c) 1997-2005 University of Cambridge.
2036 nigel 63 -----------------------------------------------------------------------------
2040 nigel 73 NAME
2041     PCRE - Perl-compatible regular expressions
2043 nigel 77
2044 nigel 63 PCRE CALLOUTS
2046 nigel 73 int (*pcre_callout)(pcre_callout_block *);
2047 nigel 63
2048 nigel 73 PCRE provides a feature called "callout", which is a means of temporar-
2049     ily passing control to the caller of PCRE in the middle of pattern
2050     matching. The caller of PCRE provides an external function by putting
2051     its entry point in the global variable pcre_callout. By default, this
2052     variable contains NULL, which disables all calling out.
2053 nigel 63
2054 nigel 73 Within a regular expression, (?C) indicates the points at which the
2055     external function is to be called. Different callout points can be
2056     identified by putting a number less than 256 after the letter C. The
2057     default value is zero. For example, this pattern has two callout
2058     points:
2059 nigel 63
2060 nigel 75 (?C1)eabc(?C2)def
2061 nigel 63
2062 nigel 75 If the PCRE_AUTO_CALLOUT option bit is set when pcre_compile() is
2063     called, PCRE automatically inserts callouts, all with number 255,
2064     before each item in the pattern. For example, if PCRE_AUTO_CALLOUT is
2065     used with the pattern
2066 nigel 63
2067 nigel 75 A(\d{2}|--)
2069     it is processed as if it were
2071     (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255)
2073     Notice that there is a callout before and after each parenthesis and
2074     alternation bar. Automatic callouts can be used for tracking the
2075     progress of pattern matching. The pcretest command has an option that
2076     sets automatic callouts; when it is used, the output indicates how the
2077     pattern is matched. This is useful information when you are trying to
2078     optimize the performance of a particular pattern.
2083     You should be aware that, because of optimizations in the way PCRE
2084     matches patterns, callouts sometimes do not happen. For example, if the
2085     pattern is
2087     ab(?C4)cd
2089     PCRE knows that any matching string must contain the letter "d". If the
2090     subject string is "abyz", the lack of "d" means that matching doesn't
2091     ever start, and the callout is never reached. However, with "abyd",
2092     though the result is still no match, the callout is obeyed.
2097     During matching, when PCRE reaches a callout point, the external func-
2098 nigel 77 tion defined by pcre_callout is called (if it is set). This applies to
2099     both the pcre_exec() and the pcre_dfa_exec() matching functions. The
2100     only argument to the callout function is a pointer to a pcre_callout
2101     block. This structure contains the following fields:
2102 nigel 75
2103 nigel 73 int version;
2104     int callout_number;
2105     int *offset_vector;
2106     const char *subject;
2107     int subject_length;
2108     int start_match;
2109     int current_position;
2110     int capture_top;
2111     int capture_last;
2112     void *callout_data;
2113 nigel 75 int pattern_position;
2114     int next_item_length;
2115 nigel 63
2116 nigel 77 The version field is an integer containing the version number of the
2117     block format. The initial version was 0; the current version is 1. The
2118     version number will change again in future if additional fields are
2119 nigel 75 added, but the intention is never to remove any of the existing fields.
2120 nigel 63
2121 nigel 77 The callout_number field contains the number of the callout, as com-
2122     piled into the pattern (that is, the number after ?C for manual call-
2123 nigel 75 outs, and 255 for automatically generated callouts).
2124 nigel 63
2125 nigel 77 The offset_vector field is a pointer to the vector of offsets that was
2126     passed by the caller to pcre_exec() or pcre_dfa_exec(). When
2127     pcre_exec() is used, the contents can be inspected in order to extract
2128     substrings that have been matched so far, in the same way as for
2129     extracting substrings after a match has completed. For pcre_dfa_exec()
2130     this field is not useful.
2131 nigel 63
2132 nigel 75 The subject and subject_length fields contain copies of the values that
2133 nigel 73 were passed to pcre_exec().
2134 nigel 63
2135 nigel 77 The start_match field contains the offset within the subject at which
2136     the current match attempt started. If the pattern is not anchored, the
2137 nigel 75 callout function may be called several times from the same point in the
2138     pattern for different starting points in the subject.
2139 nigel 63
2140 nigel 77 The current_position field contains the offset within the subject of
2141 nigel 73 the current match pointer.
2142 nigel 63
2143 nigel 77 When the pcre_exec() function is used, the capture_top field contains
2144     one more than the number of the highest numbered captured substring so
2145     far. If no substrings have been captured, the value of capture_top is
2146     one. This is always the case when pcre_dfa_exec() is used, because it
2147     does not support captured substrings.
2148 nigel 63
2149 nigel 77 The capture_last field contains the number of the most recently cap-
2150     tured substring. If no substrings have been captured, its value is -1.
2151     This is always the case when pcre_dfa_exec() is used.
2152 nigel 63
2153 nigel 77 The callout_data field contains a value that is passed to pcre_exec()
2154     or pcre_dfa_exec() specifically so that it can be passed back in call-
2155     outs. It is passed in the pcre_callout field of the pcre_extra data
2156     structure. If no such data was passed, the value of callout_data in a
2157     pcre_callout block is NULL. There is a description of the pcre_extra
2158 nigel 73 structure in the pcreapi documentation.
2159 nigel 63
2160 nigel 77 The pattern_position field is present from version 1 of the pcre_call-
2161 nigel 75 out structure. It contains the offset to the next item to be matched in
2162     the pattern string.
2163 nigel 63
2164 nigel 77 The next_item_length field is present from version 1 of the pcre_call-
2165 nigel 75 out structure. It contains the length of the next item to be matched in
2166 nigel 77 the pattern string. When the callout immediately precedes an alterna-
2167     tion bar, a closing parenthesis, or the end of the pattern, the length
2168     is zero. When the callout precedes an opening parenthesis, the length
2169 nigel 75 is that of the entire subpattern.
2170 nigel 73
2171 nigel 77 The pattern_position and next_item_length fields are intended to help
2172     in distinguishing between different automatic callouts, which all have
2173 nigel 75 the same callout number. However, they are set for all callouts.
2176 nigel 63 RETURN VALUES
2178 nigel 77 The external callout function returns an integer to PCRE. If the value
2179     is zero, matching proceeds as normal. If the value is greater than
2180     zero, matching fails at the current point, but the testing of other
2181     matching possibilities goes ahead, just as if a lookahead assertion had
2182     failed. If the value is less than zero, the match is abandoned, and
2183     pcre_exec() (or pcre_dfa_exec()) returns the negative value.
2184 nigel 63
2185 nigel 77 Negative values should normally be chosen from the set of
2186 nigel 73 PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan-
2187 nigel 77 dard "no match" failure. The error number PCRE_ERROR_CALLOUT is
2188     reserved for use by callout functions; it will never be used by PCRE
2189 nigel 73 itself.
2190 nigel 63
2191 nigel 77 Last updated: 28 February 2005
2192     Copyright (c) 1997-2005 University of Cambridge.
2193 nigel 63 -----------------------------------------------------------------------------
2197 nigel 73 NAME
2198     PCRE - Perl-compatible regular expressions
2200 nigel 77
2202 nigel 41
2203 nigel 73 This document describes the differences in the ways that PCRE and Perl
2204     handle regular expressions. The differences described here are with
2205     respect to Perl 5.8.
2206 nigel 41
2207 nigel 73 1. PCRE does not have full UTF-8 support. Details of what it does have
2208     are given in the section on UTF-8 support in the main pcre page.
2209 nigel 41
2210 nigel 73 2. PCRE does not allow repeat quantifiers on lookahead assertions. Perl
2211     permits them, but they do not mean what you might think. For example,
2212     (?!a){3} does not assert that the next three characters are not "a". It
2213     just asserts that the next character is not "a" three times.
2214 nigel 41
2215 nigel 73 3. Capturing subpatterns that occur inside negative lookahead asser-
2216     tions are counted, but their entries in the offsets vector are never
2217     set. Perl sets its numerical variables from any such patterns that are
2218     matched before the assertion fails to match something (thereby succeed-
2219     ing), but only if the negative lookahead assertion contains just one
2220     branch.
2221 nigel 41
2222 nigel 73 4. Though binary zero characters are supported in the subject string,
2223     they are not allowed in a pattern string because it is passed as a nor-
2224 nigel 75 mal C string, terminated by zero. The escape sequence \0 can be used in
2225     the pattern to represent a binary zero.
2226 nigel 41
2227 nigel 73 5. The following Perl escape sequences are not supported: \l, \u, \L,
2228 nigel 75 \U, and \N. In fact these are implemented by Perl's general string-han-
2229     dling and are not part of its pattern matching engine. If any of these
2230     are encountered by PCRE, an error is generated.
2231 nigel 41
2232 nigel 75 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE
2233     is built with Unicode character property support. The properties that
2234     can be tested with \p and \P are limited to the general category prop-
2235     erties such as Lu and Nd.
2237     7. PCRE does support the \Q...\E escape for quoting substrings. Charac-
2238     ters in between are treated as literals. This is slightly different
2239     from Perl in that $ and @ are also handled as literals inside the
2240     quotes. In Perl, they cause variable interpolation (but of course PCRE
2241 nigel 73 does not have variables). Note the following examples:
2242 nigel 49
2243 nigel 73 Pattern PCRE matches Perl matches
2244 nigel 41
2245 nigel 73 \Qabc$xyz\E abc$xyz abc followed by the
2246     contents of $xyz
2247     \Qabc\$xyz\E abc\$xyz abc\$xyz
2248     \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
2249 nigel 41
2250 nigel 75 The \Q...\E sequence is recognized both inside and outside character
2251 nigel 73 classes.
2252 nigel 41
2253 nigel 75 8. Fairly obviously, PCRE does not support the (?{code}) and (?p{code})
2254     constructions. However, there is support for recursive patterns using
2255     the non-Perl items (?R), (?number), and (?P>name). Also, the PCRE
2256     "callout" feature allows an external function to be called during pat-
2257     tern matching. See the pcrecallout documentation for details.
2258 nigel 63
2259 nigel 75 9. There are some differences that are concerned with the settings of
2260     captured strings when part of a pattern is repeated. For example,
2261     matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2
2262 nigel 73 unset, but in PCRE it is set to "b".
2263 nigel 41
2264 nigel 75 10. PCRE provides some extensions to the Perl regular expression facil-
2265     ities:
2266 nigel 41
2267 nigel 75 (a) Although lookbehind assertions must match fixed length strings,
2268 nigel 73 each alternative branch of a lookbehind assertion can match a different
2269     length of string. Perl requires them all to have the same length.
2270 nigel 41
2271 nigel 75 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $
2272 nigel 73 meta-character matches only at the very end of the string.
2273 nigel 41
2274 nigel 73 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe-
2275     cial meaning is faulted.
2276 nigel 41
2277 nigel 75 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti-
2278 nigel 73 fiers is inverted, that is, by default they are not greedy, but if fol-
2279     lowed by a question mark they are.
2280 nigel 41
2281 nigel 75 (e) PCRE_ANCHORED can be used at matching time to force a pattern to be
2282     tried only at the first matching position in the subject string.
2283 nigel 41
2285 nigel 73 TURE options for pcre_exec() have no Perl equivalents.
2286 nigel 41
2287 nigel 75 (g) The (?R), (?number), and (?P>name) constructs allows for recursive
2288     pattern matching (Perl can do this using the (?p{code}) construct,
2289 nigel 73 which PCRE cannot support.)
2290 nigel 41
2291 nigel 75 (h) PCRE supports named capturing substrings, using the Python syntax.
2292 nigel 43
2293 nigel 75 (i) PCRE supports the possessive quantifier "++" syntax, taken from
2294 nigel 73 Sun's Java package.
2295 nigel 63
2296 nigel 73 (j) The (R) condition, for testing recursion, is a PCRE extension.
2297 nigel 63
2298 nigel 73 (k) The callout facility is PCRE-specific.
2300 nigel 75 (l) The partial matching facility is PCRE-specific.
2302     (m) Patterns compiled by PCRE can be saved and re-used at a later time,
2303     even on different hosts that have the other endianness.
2305 nigel 77 (n) The alternative matching function (pcre_dfa_exec()) matches in a
2306     different way and is not Perl-compatible.
2308     Last updated: 28 February 2005
2309     Copyright (c) 1997-2005 University of Cambridge.
2310 nigel 63 -----------------------------------------------------------------------------
2314 nigel 73 NAME
2315     PCRE - Perl-compatible regular expressions
2317 nigel 77
2320 nigel 73 The syntax and semantics of the regular expressions supported by PCRE
2321     are described below. Regular expressions are also described in the Perl
2322 nigel 75 documentation and in a number of books, some of which have copious
2323     examples. Jeffrey Friedl's "Mastering Regular Expressions", published
2324     by O'Reilly, covers regular expressions in great detail. This descrip-
2325     tion of PCRE's regular expressions is intended as reference material.
2326 nigel 49
2327 nigel 75 The original operation of PCRE was on strings of one-byte characters.
2328     However, there is now also support for UTF-8 character strings. To use
2329     this, you must build PCRE to include UTF-8 support, and then call
2330     pcre_compile() with the PCRE_UTF8 option. How this affects pattern
2331     matching is mentioned in several places below. There is also a summary
2332     of UTF-8 features in the section on UTF-8 support in the main pcre
2333     page.
2334 nigel 41
2335 nigel 77 The remainder of this document discusses the patterns that are sup-
2336     ported by PCRE when its main matching function, pcre_exec(), is used.
2337     From release 6.0, PCRE offers a second matching function,
2338     pcre_dfa_exec(), which matches using a different algorithm that is not
2339     Perl-compatible. The advantages and disadvantages of the alternative
2340     function, and how it differs from the normal function, are discussed in
2341     the pcrematching page.
2343 nigel 75 A regular expression is a pattern that is matched against a subject
2344     string from left to right. Most characters stand for themselves in a
2345     pattern, and match the corresponding characters in the subject. As a
2346 nigel 73 trivial example, the pattern
2347 nigel 41
2348 nigel 73 The quick brown fox
2349 nigel 41
2350 nigel 77 matches a portion of a subject string that is identical to itself. When
2351     caseless matching is specified (the PCRE_CASELESS option), letters are
2352     matched independently of case. In UTF-8 mode, PCRE always understands
2353     the concept of case for characters whose values are less than 128, so
2354     caseless matching is always possible. For characters with higher val-
2355     ues, the concept of case is supported if PCRE is compiled with Unicode
2356     property support, but not otherwise. If you want to use caseless
2357     matching for characters 128 and above, you must ensure that PCRE is
2358     compiled with Unicode property support as well as with UTF-8 support.
2359 nigel 41
2360 nigel 77 The power of regular expressions comes from the ability to include
2361     alternatives and repetitions in the pattern. These are encoded in the
2362     pattern by the use of metacharacters, which do not stand for themselves
2363     but instead are interpreted in some special way.
2365     There are two different sets of metacharacters: those that are recog-
2366     nized anywhere in the pattern except within square brackets, and those
2367     that are recognized in square brackets. Outside square brackets, the
2368 nigel 75 metacharacters are as follows:
2369 nigel 41
2370 nigel 73 \ general escape character with several uses
2371     ^ assert start of string (or line, in multiline mode)
2372     $ assert end of string (or line, in multiline mode)
2373     . match any character except newline (by default)
2374     [ start character class definition
2375     | start of alternative branch
2376     ( start subpattern
2377     ) end subpattern
2378     ? extends the meaning of (
2379     also 0 or 1 quantifier
2380     also quantifier minimizer
2381     * 0 or more quantifier
2382     + 1 or more quantifier
2383     also "possessive quantifier"
2384     { start min/max quantifier
2385 nigel 41
2386 nigel 77 Part of a pattern that is in square brackets is called a "character
2387 nigel 75 class". In a character class the only metacharacters are:
2388 nigel 41
2389 nigel 73 \ general escape character
2390     ^ negate the class, but only if the first character
2391     - indicates character range
2392     [ POSIX character class (only if followed by POSIX
2393     syntax)
2394     ] terminates the character class
2395 nigel 41
2396 nigel 77 The following sections describe the use of each of the metacharacters.
2397 nigel 41
2399 nigel 63 BACKSLASH
2400 nigel 41
2401 nigel 73 The backslash character has several uses. Firstly, if it is followed by
2402 nigel 77 a non-alphanumeric character, it takes away any special meaning that
2403     character may have. This use of backslash as an escape character
2404 nigel 73 applies both inside and outside character classes.
2405 nigel 41
2406 nigel 77 For example, if you want to match a * character, you write \* in the
2407     pattern. This escaping action applies whether or not the following
2408     character would otherwise be interpreted as a metacharacter, so it is
2409     always safe to precede a non-alphanumeric with backslash to specify
2410     that it stands for itself. In particular, if you want to match a back-
2411 nigel 75 slash, you write \\.
2412 nigel 41
2413 nigel 77 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in
2414     the pattern (other than in a character class) and characters between a
2415 nigel 73 # outside a character class and the next newline character are ignored.
2416 nigel 77 An escaping backslash can be used to include a whitespace or # charac-
2417 nigel 73 ter as part of the pattern.
2418 nigel 41
2419 nigel 77 If you want to remove the special meaning from a sequence of charac-
2420     ters, you can do so by putting them between \Q and \E. This is differ-
2421     ent from Perl in that $ and @ are handled as literals in \Q...\E
2422     sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
2423 nigel 73 tion. Note the following examples:
2424 nigel 63
2425 nigel 73 Pattern PCRE matches Perl matches
2426 nigel 63
2427 nigel 73 \Qabc$xyz\E abc$xyz abc followed by the
2428     contents of $xyz
2429     \Qabc\$xyz\E abc\$xyz abc\$xyz
2430     \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
2431 nigel 63
2432 nigel 77 The \Q...\E sequence is recognized both inside and outside character
2433 nigel 73 classes.
2434 nigel 63
2435 nigel 75 Non-printing characters
2437 nigel 73 A second use of backslash provides a way of encoding non-printing char-
2438 nigel 77 acters in patterns in a visible manner. There is no restriction on the
2439     appearance of non-printing characters, apart from the binary zero that
2440     terminates a pattern, but when a pattern is being prepared by text
2441     editing, it is usually easier to use one of the following escape
2442 nigel 73 sequences than the binary character it represents:
2443 nigel 63
2444 nigel 73 \a alarm, that is, the BEL character (hex 07)
2445     \cx "control-x", where x is any character
2446     \e escape (hex 1B)
2447     \f formfeed (hex 0C)
2448     \n newline (hex 0A)
2449     \r carriage return (hex 0D)
2450     \t tab (hex 09)
2451     \ddd character with octal code ddd, or backreference
2452     \xhh character with hex code hh
2453     \x{hhh..} character with hex code hhh... (UTF-8 mode only)
2454 nigel 41
2455 nigel 77 The precise effect of \cx is as follows: if x is a lower case letter,
2456     it is converted to upper case. Then bit 6 of the character (hex 40) is
2457     inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c;
2458 nigel 73 becomes hex 7B.
2459 nigel 41
2460 nigel 77 After \x, from zero to two hexadecimal digits are read (letters can be
2461     in upper or lower case). In UTF-8 mode, any number of hexadecimal dig-
2462     its may appear between \x{ and }, but the value of the character code
2463     must be less than 2**31 (that is, the maximum hexadecimal value is
2464     7FFFFFFF). If characters other than hexadecimal digits appear between
2465     \x{ and }, or if there is no terminating }, this form of escape is not
2466     recognized. Instead, the initial \x will be interpreted as a basic
2467     hexadecimal escape, with no following digits, giving a character whose
2468 nigel 75 value is zero.
2469 nigel 41
2470 nigel 73 Characters whose value is less than 256 can be defined by either of the
2471 nigel 77 two syntaxes for \x when PCRE is in UTF-8 mode. There is no difference
2472     in the way they are handled. For example, \xdc is exactly the same as
2473 nigel 73 \x{dc}.
2474 nigel 41
2475 nigel 77 After \0 up to two further octal digits are read. In both cases, if
2476     there are fewer than two digits, just those that are present are used.
2477     Thus the sequence \0\x\07 specifies two binary zeros followed by a BEL
2478     character (code value 7). Make sure you supply two digits after the
2479     initial zero if the pattern character that follows is itself an octal
2480 nigel 75 digit.
2481 nigel 63
2482 nigel 73 The handling of a backslash followed by a digit other than 0 is compli-
2483     cated. Outside a character class, PCRE reads it and any following dig-
2484 nigel 77 its as a decimal number. If the number is less than 10, or if there
2485 nigel 73 have been at least that many previous capturing left parentheses in the
2486 nigel 77 expression, the entire sequence is taken as a back reference. A
2487     description of how this works is given later, following the discussion
2488 nigel 73 of parenthesized subpatterns.
2489 nigel 41
2490 nigel 77 Inside a character class, or if the decimal number is greater than 9
2491     and there have not been that many capturing subpatterns, PCRE re-reads
2492     up to three octal digits following the backslash, and generates a sin-
2493 nigel 73 gle byte from the least significant 8 bits of the value. Any subsequent
2494     digits stand for themselves. For example:
2495 nigel 41
2496 nigel 73 \040 is another way of writing a space
2497     \40 is the same, provided there are fewer than 40
2498     previous capturing subpatterns
2499     \7 is always a back reference
2500     \11 might be a back reference, or another way of
2501     writing a tab
2502     \011 is always a tab
2503     \0113 is a tab followed by the character "3"
2504     \113 might be a back reference, otherwise the
2505     character with octal code 113
2506     \377 might be a back reference, otherwise
2507     the byte consisting entirely of 1 bits
2508     \81 is either a back reference, or a binary zero
2509     followed by the two characters "8" and "1"
2510 nigel 41
2511 nigel 77 Note that octal values of 100 or greater must not be introduced by a
2512 nigel 73 leading zero, because no more than three octal digits are ever read.
2513 nigel 41
2514 nigel 77 All the sequences that define a single byte value or a single UTF-8
2515 nigel 73 character (in UTF-8 mode) can be used both inside and outside character
2516 nigel 77 classes. In addition, inside a character class, the sequence \b is
2517 nigel 75 interpreted as the backspace character (hex 08), and the sequence \X is
2518 nigel 77 interpreted as the character "X". Outside a character class, these
2519 nigel 75 sequences have different meanings (see below).
2520 nigel 43
2521 nigel 75 Generic character types
2522 nigel 41
2523 nigel 77 The third use of backslash is for specifying generic character types.
2524 nigel 75 The following are always recognized:
2526 nigel 73 \d any decimal digit
2527     \D any character that is not a decimal digit
2528     \s any whitespace character
2529     \S any character that is not a whitespace character
2530     \w any "word" character
2531     \W any "non-word" character
2532 nigel 41
2533 nigel 73 Each pair of escape sequences partitions the complete set of characters
2534 nigel 77 into two disjoint sets. Any given character matches one, and only one,
2535 nigel 73 of each pair.
2536 nigel 41
2537 nigel 75 These character type sequences can appear both inside and outside char-
2538 nigel 77 acter classes. They each match one character of the appropriate type.
2539     If the current matching point is at the end of the subject string, all
2540 nigel 75 of them fail, since there is no character to match.
2541 nigel 41
2542 nigel 77 For compatibility with Perl, \s does not match the VT character (code
2543     11). This makes it different from the the POSIX "space" class. The \s
2544 nigel 73 characters are HT (9), LF (10), FF (12), CR (13), and space (32).
2545 nigel 63
2546 nigel 75 A "word" character is an underscore or any character less than 256 that
2547 nigel 77 is a letter or digit. The definition of letters and digits is con-
2548     trolled by PCRE's low-valued character tables, and may vary if locale-
2549     specific matching is taking place (see "Locale support" in the pcreapi
2550     page). For example, in the "fr_FR" (French) locale, some character
2551     codes greater than 128 are used for accented letters, and these are
2552 nigel 75 matched by \w.
2553 nigel 63
2554 nigel 77 In UTF-8 mode, characters with values greater than 128 never match \d,
2555 nigel 75 \s, or \w, and always match \D, \S, and \W. This is true even when Uni-
2556     code character property support is available.
2557 nigel 41
2558 nigel 75 Unicode character properties
2560     When PCRE is built with Unicode character property support, three addi-
2561 nigel 77 tional escape sequences to match generic character types are available
2562 nigel 75 when UTF-8 mode is selected. They are:
2564     \p{xx} a character with the xx property
2565     \P{xx} a character without the xx property
2566     \X an extended Unicode sequence
2568 nigel 77 The property names represented by xx above are limited to the Unicode
2569     general category properties. Each character has exactly one such prop-
2570     erty, specified by a two-letter abbreviation. For compatibility with
2571     Perl, negation can be specified by including a circumflex between the
2572     opening brace and the property name. For example, \p{^Lu} is the same
2573 nigel 75 as \P{Lu}.
2575 nigel 77 If only one letter is specified with \p or \P, it includes all the
2576 nigel 75 properties that start with that letter. In this case, in the absence of
2577     negation, the curly brackets in the escape sequence are optional; these
2578     two examples have the same effect:
2580     \p{L}
2581     \pL
2583     The following property codes are supported:
2585     C Other
2586     Cc Control
2587     Cf Format
2588     Cn Unassigned
2589     Co Private use
2590     Cs Surrogate
2592     L Letter
2593     Ll Lower case letter
2594     Lm Modifier letter
2595     Lo Other letter
2596     Lt Title case letter
2597     Lu Upper case letter
2599     M Mark
2600     Mc Spacing mark
2601     Me Enclosing mark
2602     Mn Non-spacing mark
2604     N Number
2605     Nd Decimal number
2606     Nl Letter number
2607     No Other number
2609     P Punctuation
2610     Pc Connector punctuation
2611     Pd Dash punctuation
2612     Pe Close punctuation
2613     Pf Final punctuation
2614     Pi Initial punctuation
2615     Po Other punctuation
2616     Ps Open punctuation
2618     S Symbol
2619     Sc Currency symbol
2620     Sk Modifier symbol
2621     Sm Mathematical symbol
2622     So Other symbol
2624     Z Separator
2625     Zl Line separator
2626     Zp Paragraph separator
2627     Zs Space separator
2629 nigel 77 Extended properties such as "Greek" or "InMusicalSymbols" are not sup-
2630 nigel 75 ported by PCRE.
2632 nigel 77 Specifying caseless matching does not affect these escape sequences.
2633 nigel 75 For example, \p{Lu} always matches only upper case letters.
2635 nigel 77 The \X escape matches any number of Unicode characters that form an
2636 nigel 75 extended Unicode sequence. \X is equivalent to
2638     (?>\PM\pM*)
2640 nigel 77 That is, it matches a character without the "mark" property, followed
2641     by zero or more characters with the "mark" property, and treats the
2642     sequence as an atomic group (see below). Characters with the "mark"
2643 nigel 75 property are typically accents that affect the preceding character.
2645 nigel 77 Matching characters by Unicode property is not fast, because PCRE has
2646     to search a structure that contains data for over fifteen thousand
2647 nigel 75 characters. That is why the traditional escape sequences such as \d and
2648     \w do not use Unicode properties in PCRE.
2650     Simple assertions
2652 nigel 73 The fourth use of backslash is for certain simple assertions. An asser-
2653 nigel 77 tion specifies a condition that has to be met at a particular point in
2654     a match, without consuming any characters from the subject string. The
2655     use of subpatterns for more complicated assertions is described below.
2656 nigel 75 The backslashed assertions are:
2657 nigel 41
2658 nigel 73 \b matches at a word boundary
2659     \B matches when not at a word boundary
2660     \A matches at start of subject
2661     \Z matches at end of subject or before newline at end
2662     \z matches at end of subject
2663     \G matches at first matching position in subject
2664 nigel 41
2665 nigel 77 These assertions may not appear in character classes (but note that \b
2666 nigel 73 has a different meaning, namely the backspace character, inside a char-
2667     acter class).
2668 nigel 41
2669 nigel 77 A word boundary is a position in the subject string where the current
2670     character and the previous character do not both match \w or \W (i.e.
2671     one matches \w and the other matches \W), or the start or end of the
2672 nigel 73 string if the first or last character matches \w, respectively.
2673 nigel 43
2674 nigel 77 The \A, \Z, and \z assertions differ from the traditional circumflex
2675 nigel 75 and dollar (described in the next section) in that they only ever match
2676 nigel 77 at the very start and end of the subject string, whatever options are
2677     set. Thus, they are independent of multiline mode. These three asser-
2678 nigel 75 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
2679 nigel 77 affect only the behaviour of the circumflex and dollar metacharacters.
2680     However, if the startoffset argument of pcre_exec() is non-zero, indi-
2681 nigel 75 cating that matching is to start at a point other than the beginning of
2682 nigel 77 the subject, \A can never match. The difference between \Z and \z is
2683     that \Z matches before a newline that is the last character of the
2684     string as well as at the end of the string, whereas \z matches only at
2685 nigel 75 the end.
2686 nigel 63
2687 nigel 77 The \G assertion is true only when the current matching position is at
2688     the start point of the match, as specified by the startoffset argument
2689     of pcre_exec(). It differs from \A when the value of startoffset is
2690     non-zero. By calling pcre_exec() multiple times with appropriate argu-
2691 nigel 73 ments, you can mimic Perl's /g option, and it is in this kind of imple-
2692     mentation where \G can be useful.
2693 nigel 41
2694 nigel 77 Note, however, that PCRE's interpretation of \G, as the start of the
2695 nigel 73 current match, is subtly different from Perl's, which defines it as the
2696 nigel 77 end of the previous match. In Perl, these can be different when the
2697     previously matched string was empty. Because PCRE does just one match
2698 nigel 73 at a time, it cannot reproduce this behaviour.
2699 nigel 41
2700 nigel 77 If all the alternatives of a pattern begin with \G, the expression is
2701 nigel 73 anchored to the starting match position, and the "anchored" flag is set
2702     in the compiled regular expression.
2703 nigel 63
2706 nigel 63
2707 nigel 73 Outside a character class, in the default matching mode, the circumflex
2708 nigel 77 character is an assertion that is true only if the current matching
2709     point is at the start of the subject string. If the startoffset argu-
2710     ment of pcre_exec() is non-zero, circumflex can never match if the
2711     PCRE_MULTILINE option is unset. Inside a character class, circumflex
2712 nigel 73 has an entirely different meaning (see below).
2713 nigel 41
2714 nigel 77 Circumflex need not be the first character of the pattern if a number
2715     of alternatives are involved, but it should be the first thing in each
2716     alternative in which it appears if the pattern is ever to match that
2717     branch. If all possible alternatives start with a circumflex, that is,
2718     if the pattern is constrained to match only at the start of the sub-
2719     ject, it is said to be an "anchored" pattern. (There are also other
2720 nigel 73 constructs that can cause a pattern to be anchored.)
2721 nigel 41
2722 nigel 77 A dollar character is an assertion that is true only if the current
2723     matching point is at the end of the subject string, or immediately
2724 nigel 73 before a newline character that is the last character in the string (by
2725 nigel 77 default). Dollar need not be the last character of the pattern if a
2726     number of alternatives are involved, but it should be the last item in
2727     any branch in which it appears. Dollar has no special meaning in a
2728 nigel 73 character class.
2729 nigel 41
2730 nigel 77 The meaning of dollar can be changed so that it matches only at the
2731     very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
2732 nigel 73 compile time. This does not affect the \Z assertion.
2733 nigel 41
2734 nigel 73 The meanings of the circumflex and dollar characters are changed if the
2735     PCRE_MULTILINE option is set. When this is the case, they match immedi-
2736 nigel 77 ately after and immediately before an internal newline character,
2737     respectively, in addition to matching at the start and end of the sub-
2738     ject string. For example, the pattern /^abc$/ matches the subject
2739     string "def\nabc" (where \n represents a newline character) in multi-
2740 nigel 75 line mode, but not otherwise. Consequently, patterns that are anchored
2741 nigel 77 in single line mode because all branches start with ^ are not anchored
2742     in multiline mode, and a match for circumflex is possible when the
2743     startoffset argument of pcre_exec() is non-zero. The PCRE_DOL-
2744 nigel 75 LAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
2745 nigel 41
2746 nigel 77 Note that the sequences \A, \Z, and \z can be used to match the start
2747     and end of the subject in both modes, and if all branches of a pattern
2748     start with \A it is always anchored, whether PCRE_MULTILINE is set or
2749 nigel 73 not.
2750 nigel 41
2752 nigel 63 FULL STOP (PERIOD, DOT)
2753 nigel 41
2754 nigel 73 Outside a character class, a dot in the pattern matches any one charac-
2755 nigel 77 ter in the subject, including a non-printing character, but not (by
2756     default) newline. In UTF-8 mode, a dot matches any UTF-8 character,
2757 nigel 75 which might be more than one byte long, except (by default) newline. If
2758 nigel 77 the PCRE_DOTALL option is set, dots match newlines as well. The han-
2759     dling of dot is entirely independent of the handling of circumflex and
2760     dollar, the only relationship being that they both involve newline
2761 nigel 75 characters. Dot has no special meaning in a character class.
2762 nigel 41
2766 nigel 73 Outside a character class, the escape sequence \C matches any one byte,
2767 nigel 77 both in and out of UTF-8 mode. Unlike a dot, it can match a newline.
2768     The feature is provided in Perl in order to match individual bytes in
2769     UTF-8 mode. Because it breaks up UTF-8 characters into individual
2770     bytes, what remains in the string may be a malformed UTF-8 string. For
2771 nigel 75 this reason, the \C escape sequence is best avoided.
2772 nigel 63
2773 nigel 77 PCRE does not allow \C to appear in lookbehind assertions (described
2774     below), because in UTF-8 mode this would make it impossible to calcu-
2775 nigel 75 late the length of the lookbehind.
2776 nigel 63
2779 nigel 63
2780 nigel 73 An opening square bracket introduces a character class, terminated by a
2781     closing square bracket. A closing square bracket on its own is not spe-
2782     cial. If a closing square bracket is required as a member of the class,
2783 nigel 77 it should be the first data character in the class (after an initial
2784 nigel 73 circumflex, if present) or escaped with a backslash.
2785 nigel 41
2786 nigel 77 A character class matches a single character in the subject. In UTF-8
2787     mode, the character may occupy more than one byte. A matched character
2788 nigel 73 must be in the set of characters defined by the class, unless the first
2789 nigel 77 character in the class definition is a circumflex, in which case the
2790     subject character must not be in the set defined by the class. If a
2791     circumflex is actually required as a member of the class, ensure it is
2792 nigel 73 not the first character, or escape it with a backslash.
2793 nigel 41
2794 nigel 77 For example, the character class [aeiou] matches any lower case vowel,
2795     while [^aeiou] matches any character that is not a lower case vowel.
2796 nigel 73 Note that a circumflex is just a convenient notation for specifying the
2797 nigel 77 characters that are in the class by enumerating those that are not. A
2798     class that starts with a circumflex is not an assertion: it still con-
2799     sumes a character from the subject string, and therefore it fails if
2800 nigel 75 the current pointer is at the end of the string.
2801 nigel 41
2802 nigel 77 In UTF-8 mode, characters with values greater than 255 can be included
2803     in a class as a literal string of bytes, or by using the \x{ escaping
2804 nigel 73 mechanism.
2805 nigel 63
2806 nigel 77 When caseless matching is set, any letters in a class represent both
2807     their upper case and lower case versions, so for example, a caseless
2808     [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
2809     match "A", whereas a caseful version would. In UTF-8 mode, PCRE always
2810     understands the concept of case for characters whose values are less
2811     than 128, so caseless matching is always possible. For characters with
2812     higher values, the concept of case is supported if PCRE is compiled
2813     with Unicode property support, but not otherwise. If you want to use
2814     caseless matching for characters 128 and above, you must ensure that
2815     PCRE is compiled with Unicode property support as well as with UTF-8
2816     support.
2817 nigel 41
2818 nigel 75 The newline character is never treated in any special way in character
2819     classes, whatever the setting of the PCRE_DOTALL or PCRE_MULTILINE
2820 nigel 73 options is. A class such as [^a] will always match a newline.
2821 nigel 41
2822 nigel 75 The minus (hyphen) character can be used to specify a range of charac-
2823     ters in a character class. For example, [d-m] matches any letter
2824     between d and m, inclusive. If a minus character is required in a
2825     class, it must be escaped with a backslash or appear in a position
2826     where it cannot be interpreted as indicating a range, typically as the
2827 nigel 73 first or last character in the class.
2828 nigel 41
2829 nigel 73 It is not possible to have the literal character "]" as the end charac-
2830 nigel 75 ter of a range. A pattern such as [W-]46] is interpreted as a class of
2831     two characters ("W" and "-") followed by a literal string "46]", so it
2832     would match "W46]" or "-46]". However, if the "]" is escaped with a
2833     backslash it is interpreted as the end of range, so [W-\]46] is inter-
2834     preted as a class containing a range followed by two other characters.
2835     The octal or hexadecimal representation of "]" can also be used to end
2836     a range.
2837 nigel 41
2838 nigel 75 Ranges operate in the collating sequence of character values. They can
2839     also be used for characters specified numerically, for example
2840     [\000-\037]. In UTF-8 mode, ranges can include characters whose values
2841 nigel 73 are greater than 255, for example [\x{100}-\x{2ff}].
2842 nigel 63
2843 nigel 73 If a range that includes letters is used when caseless matching is set,
2844     it matches the letters in either case. For example, [W-c] is equivalent
2845 nigel 75 to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if
2846     character tables for the "fr_FR" locale are in use, [\xc8-\xcb] matches
2847     accented E characters in both cases. In UTF-8 mode, PCRE supports the
2848     concept of case for characters with values greater than 128 only when
2849     it is compiled with Unicode property support.
2850 nigel 41
2851 nigel 75 The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear
2852     in a character class, and add the characters that they match to the
2853     class. For example, [\dABCDEF] matches any hexadecimal digit. A circum-
2854     flex can conveniently be used with the upper case character types to
2855     specify a more restricted set of characters than the matching lower
2856     case type. For example, the class [^\W_] matches any letter or digit,
2857     but not underscore.
2858 nigel 41
2859 nigel 75 The only metacharacters that are recognized in character classes are
2860     backslash, hyphen (only where it can be interpreted as specifying a
2861     range), circumflex (only at the start), opening square bracket (only
2862     when it can be interpreted as introducing a POSIX class name - see the
2863     next section), and the terminating closing square bracket. However,
2864     escaping other non-alphanumeric characters does no harm.
2865 nigel 41
2866 nigel 73
2869 nigel 75 Perl supports the POSIX notation for character classes. This uses names
2870     enclosed by [: and :] within the enclosing square brackets. PCRE also
2871     supports this notation. For example,
2872 nigel 63
2873 nigel 73 [01[:alpha:]%]
2874 nigel 43
2875 nigel 73 matches "0", "1", any alphabetic character, or "%". The supported class
2876     names are
2877 nigel 43
2878 nigel 73 alnum letters and digits
2879     alpha letters
2880     ascii character codes 0 - 127
2881     blank space or tab only
2882     cntrl control characters
2883     digit decimal digits (same as \d)
2884     graph printing characters, excluding space
2885     lower lower case letters
2886     print printing characters, including space
2887     punct printing characters, excluding letters and digits
2888     space white space (not quite the same as \s)
2889     upper upper case letters
2890     word "word" characters (same as \w)
2891     xdigit hexadecimal digits
2892 nigel 43
2893 nigel 73 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
2894     and space (32). Notice that this list includes the VT character (code
2895     11). This makes "space" different to \s, which does not include VT (for
2896     Perl compatibility).
2897 nigel 43
2898 nigel 73 The name "word" is a Perl extension, and "blank" is a GNU extension
2899     from Perl 5.8. Another Perl extension is negation, which is indicated
2900     by a ^ character after the colon. For example,
2901 nigel 63
2902 nigel 73 [12[:^digit:]]
2903 nigel 43
2904 nigel 73 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
2905     POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
2906     these are not supported, and an error is given if they are encountered.
2907 nigel 43
2908 nigel 75 In UTF-8 mode, characters with values greater than 128 do not match any
2909 nigel 73 of the POSIX character classes.
2910 nigel 43
2912 nigel 41 VERTICAL BAR
2913 nigel 63
2914 nigel 73 Vertical bar characters are used to separate alternative patterns. For
2915     example, the pattern
2916 nigel 41
2917 nigel 73 gilbert|sullivan
2918 nigel 41
2919 nigel 73 matches either "gilbert" or "sullivan". Any number of alternatives may
2920     appear, and an empty alternative is permitted (matching the empty
2921     string). The matching process tries each alternative in turn, from
2922     left to right, and the first one that succeeds is used. If the alterna-
2923     tives are within a subpattern (defined below), "succeeds" means match-
2924     ing the rest of the main pattern as well as the alternative in the sub-
2925     pattern.
2926 nigel 41
2930 nigel 73 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
2931     PCRE_EXTENDED options can be changed from within the pattern by a
2932     sequence of Perl option letters enclosed between "(?" and ")". The
2933     option letters are
2934 nigel 63
2935 nigel 73 i for PCRE_CASELESS
2936     m for PCRE_MULTILINE
2937     s for PCRE_DOTALL
2938     x for PCRE_EXTENDED
2939 nigel 41
2940 nigel 73 For example, (?im) sets caseless, multiline matching. It is also possi-
2941     ble to unset these options by preceding the letter with a hyphen, and a
2942     combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
2943     LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED,
2944     is also permitted. If a letter appears both before and after the
2945     hyphen, the option is unset.
2946 nigel 41
2947 nigel 73 When an option change occurs at top level (that is, not inside subpat-
2948     tern parentheses), the change applies to the remainder of the pattern
2949     that follows. If the change is placed right at the start of a pattern,
2950     PCRE extracts it into the global options (and it will therefore show up
2951     in data extracted by the pcre_fullinfo() function).
2952 nigel 41
2953 nigel 73 An option change within a subpattern affects only that part of the cur-
2954     rent pattern that follows it, so
2955 nigel 41
2956 nigel 73 (a(?i)b)c
2957 nigel 41
2958 nigel 73 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
2959     used). By this means, options can be made to have different settings
2960     in different parts of the pattern. Any changes made in one alternative
2961     do carry on into subsequent branches within the same subpattern. For
2962     example,
2963 nigel 41
2964 nigel 73 (a(?i)b|c)
2965 nigel 41
2966 nigel 73 matches "ab", "aB", "c", and "C", even though when matching "C" the
2967     first branch is abandoned before the option setting. This is because
2968     the effects of option settings happen at compile time. There would be
2969     some very weird behaviour otherwise.
2970 nigel 41
2971 nigel 73 The PCRE-specific options PCRE_UNGREEDY and PCRE_EXTRA can be changed
2972     in the same way as the Perl-compatible options by using the characters
2973     U and X respectively. The (?X) flag setting is special in that it must
2974     always occur earlier in the pattern than any of the additional features
2975 nigel 75 it turns on, even when it is at top level. It is best to put it at the
2976     start.
2977 nigel 41
2979 nigel 63 SUBPATTERNS
2980 nigel 41
2981 nigel 73 Subpatterns are delimited by parentheses (round brackets), which can be
2982 nigel 75 nested. Turning part of a pattern into a subpattern does two things:
2983 nigel 41
2984 nigel 73 1. It localizes a set of alternatives. For example, the pattern
2985 nigel 41
2986 nigel 73 cat(aract|erpillar|)
2987 nigel 41
2988 nigel 73 matches one of the words "cat", "cataract", or "caterpillar". Without
2989     the parentheses, it would match "cataract", "erpillar" or the empty
2990     string.
2991 nigel 41
2992 nigel 75 2. It sets up the subpattern as a capturing subpattern. This means
2993     that, when the whole pattern matches, that portion of the subject
2994 nigel 73 string that matched the subpattern is passed back to the caller via the
2995     ovector argument of pcre_exec(). Opening parentheses are counted from
2996 nigel 75 left to right (starting from 1) to obtain numbers for the capturing
2997 nigel 73 subpatterns.
2998 nigel 41
2999 nigel 73 For example, if the string "the red king" is matched against the pat-
3000     tern
3001 nigel 41
3002 nigel 73 the ((red|white) (king|queen))
3003 nigel 41
3004 nigel 73 the captured substrings are "red king", "red", and "king", and are num-
3005     bered 1, 2, and 3, respectively.
3006 nigel 41
3007 nigel 73 The fact that plain parentheses fulfil two functions is not always
3008     helpful. There are often times when a grouping subpattern is required
3009     without a capturing requirement. If an opening parenthesis is followed
3010     by a question mark and a colon, the subpattern does not do any captur-
3011     ing, and is not counted when computing the number of any subsequent
3012     capturing subpatterns. For example, if the string "the white queen" is
3013     matched against the pattern
3014 nigel 41
3015 nigel 73 the ((?:red|white) (king|queen))
3016 nigel 41
3017 nigel 73 the captured substrings are "white queen" and "queen", and are numbered
3018     1 and 2. The maximum number of capturing subpatterns is 65535, and the
3019     maximum depth of nesting of all subpatterns, both capturing and non-
3020     capturing, is 200.
3021 nigel 41
3022 nigel 73 As a convenient shorthand, if any option settings are required at the
3023     start of a non-capturing subpattern, the option letters may appear
3024     between the "?" and the ":". Thus the two patterns
3025 nigel 41
3026 nigel 73 (?i:saturday|sunday)
3027     (?:(?i)saturday|sunday)
3028 nigel 41
3029 nigel 73 match exactly the same set of strings. Because alternative branches are
3030     tried from left to right, and options are not reset until the end of
3031     the subpattern is reached, an option setting in one branch does affect
3032     subsequent branches, so the above patterns match "SUNDAY" as well as
3033     "Saturday".
3034 nigel 41
3037 nigel 41
3038 nigel 73 Identifying capturing parentheses by number is simple, but it can be
3039     very hard to keep track of the numbers in complicated regular expres-
3040     sions. Furthermore, if an expression is modified, the numbers may
3041 nigel 75 change. To help with this difficulty, PCRE supports the naming of sub-
3042 nigel 73 patterns, something that Perl does not provide. The Python syntax
3043     (?P<name>...) is used. Names consist of alphanumeric characters and
3044     underscores, and must be unique within a pattern.
3045 nigel 63
3046 nigel 73 Named capturing parentheses are still allocated numbers as well as
3047     names. The PCRE API provides function calls for extracting the name-to-
3048 nigel 75 number translation table from a compiled pattern. There is also a con-
3049     venience function for extracting a captured substring by name. For fur-
3050     ther details see the pcreapi documentation.
3051 nigel 63
3053 nigel 41 REPETITION
3054 nigel 63
3055 nigel 75 Repetition is specified by quantifiers, which can follow any of the
3056 nigel 73 following items:
3057 nigel 41
3058 nigel 73 a literal data character
3059     the . metacharacter
3060     the \C escape sequence
3061 nigel 75 the \X escape sequence (in UTF-8 mode with Unicode properties)
3062     an escape such as \d that matches a single character
3063 nigel 73 a character class
3064     a back reference (see next section)
3065     a parenthesized subpattern (unless it is an assertion)
3066 nigel 41
3067 nigel 75 The general repetition quantifier specifies a minimum and maximum num-
3068     ber of permitted matches, by giving the two numbers in curly brackets
3069     (braces), separated by a comma. The numbers must be less than 65536,
3070 nigel 73 and the first must be less than or equal to the second. For example:
3071 nigel 41
3072 nigel 73 z{2,4}
3073 nigel 41
3074 nigel 75 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
3075     special character. If the second number is omitted, but the comma is
3076     present, there is no upper limit; if the second number and the comma
3077     are both omitted, the quantifier specifies an exact number of required
3078 nigel 73 matches. Thus
3079 nigel 41
3080 nigel 73 [aeiou]{3,}
3081 nigel 41
3082 nigel 73 matches at least 3 successive vowels, but may match many more, while
3083 nigel 41
3084 nigel 73 \d{8}
3085 nigel 41
3086 nigel 75 matches exactly 8 digits. An opening curly bracket that appears in a
3087     position where a quantifier is not allowed, or one that does not match
3088     the syntax of a quantifier, is taken as a literal character. For exam-
3089 nigel 73 ple, {,6} is not a quantifier, but a literal string of four characters.
3090 nigel 63
3091 nigel 75 In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to
3092 nigel 73 individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char-
3093 nigel 75 acters, each of which is represented by a two-byte sequence. Similarly,
3094     when Unicode property support is available, \X{3} matches three Unicode
3095     extended sequences, each of which may be several bytes long (and they
3096     may be of different lengths).
3097 nigel 63
3098 nigel 73 The quantifier {0} is permitted, causing the expression to behave as if
3099     the previous item and the quantifier were not present.
3100 nigel 41
3101 nigel 73 For convenience (and historical compatibility) the three most common
3102     quantifiers have single-character abbreviations:
3103 nigel 41
3104 nigel 73 * is equivalent to {0,}
3105     + is equivalent to {1,}
3106     ? is equivalent to {0,1}
3107 nigel 41
3108 nigel 73 It is possible to construct infinite loops by following a subpattern
3109     that can match no characters with a quantifier that has no upper limit,
3110     for example:
3111 nigel 41
3112 nigel 73 (a?)*
3113 nigel 41
3114 nigel 73 Earlier versions of Perl and PCRE used to give an error at compile time
3115     for such patterns. However, because there are cases where this can be
3116     useful, such patterns are now accepted, but if any repetition of the
3117     subpattern does in fact match no characters, the loop is forcibly bro-
3118     ken.
3119 nigel 41
3120 nigel 73 By default, the quantifiers are "greedy", that is, they match as much
3121     as possible (up to the maximum number of permitted times), without
3122     causing the rest of the pattern to fail. The classic example of where
3123     this gives problems is in trying to match comments in C programs. These
3124 nigel 75 appear between /* and */ and within the comment, individual * and /
3125     characters may appear. An attempt to match C comments by applying the
3126     pattern
3127 nigel 41
3128 nigel 73 /\*.*\*/
3129 nigel 41
3130 nigel 73 to the string
3131 nigel 41
3132 nigel 75 /* first comment */ not comment /* second comment */
3133 nigel 41
3134 nigel 73 fails, because it matches the entire string owing to the greediness of
3135     the .* item.
3136 nigel 41
3137 nigel 73 However, if a quantifier is followed by a question mark, it ceases to
3138     be greedy, and instead matches the minimum number of times possible, so
3139     the pattern
3140 nigel 41
3141 nigel 73 /\*.*?\*/
3142 nigel 41
3143 nigel 73 does the right thing with the C comments. The meaning of the various
3144     quantifiers is not otherwise changed, just the preferred number of
3145     matches. Do not confuse this use of question mark with its use as a
3146     quantifier in its own right. Because it has two uses, it can sometimes
3147     appear doubled, as in
3148 nigel 41
3149 nigel 73 \d??\d
3150 nigel 41
3151 nigel 73 which matches one digit by preference, but can match two if that is the
3152     only way the rest of the pattern matches.
3153 nigel 41
3154 nigel 73 If the PCRE_UNGREEDY option is set (an option which is not available in
3155     Perl), the quantifiers are not greedy by default, but individual ones
3156     can be made greedy by following them with a question mark. In other
3157     words, it inverts the default behaviour.
3158 nigel 41
3159 nigel 73 When a parenthesized subpattern is quantified with a minimum repeat
3160 nigel 75 count that is greater than 1 or with a limited maximum, more memory is
3161 nigel 73 required for the compiled pattern, in proportion to the size of the
3162     minimum or maximum.
3163 nigel 41
3164 nigel 73 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
3165     alent to Perl's /s) is set, thus allowing the . to match newlines, the
3166     pattern is implicitly anchored, because whatever follows will be tried
3167     against every character position in the subject string, so there is no
3168     point in retrying the overall match at any position after the first.
3169     PCRE normally treats such a pattern as though it were preceded by \A.
3170 nigel 63
3171 nigel 73 In cases where it is known that the subject string contains no new-
3172     lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
3173     mization, or alternatively using ^ to indicate anchoring explicitly.
3174 nigel 63
3175 nigel 73 However, there is one situation where the optimization cannot be used.
3176     When .* is inside capturing parentheses that are the subject of a
3177     backreference elsewhere in the pattern, a match at the start may fail,
3178     and a later one succeed. Consider, for example:
3179 nigel 63
3180 nigel 73 (.*)abc\1
3181 nigel 63
3182 nigel 73 If the subject is "xyz123abc123" the match point is the fourth charac-
3183     ter. For this reason, such a pattern is not implicitly anchored.
3184 nigel 41
3185 nigel 73 When a capturing subpattern is repeated, the value captured is the sub-
3186     string that matched the final iteration. For example, after
3187 nigel 41
3188 nigel 73 (tweedle[dume]{3}\s*)+
3189 nigel 41
3190 nigel 73 has matched "tweedledum tweedledee" the value of the captured substring
3191     is "tweedledee". However, if there are nested capturing subpatterns,
3192     the corresponding captured values may have been set in previous itera-
3193     tions. For example, after
3194 nigel 41
3195 nigel 73 /(a|(b))+/
3196 nigel 41
3197 nigel 73 matches "aba" the value of the second captured substring is "b".
3198 nigel 41
3199 nigel 73
3201 nigel 41
3202 nigel 73 With both maximizing and minimizing repetition, failure of what follows
3203     normally causes the repeated item to be re-evaluated to see if a dif-
3204     ferent number of repeats allows the rest of the pattern to match. Some-
3205     times it is useful to prevent this, either to change the nature of the
3206     match, or to cause it fail earlier than it otherwise might, when the
3207     author of the pattern knows there is no point in carrying on.
3208 nigel 53
3209 nigel 73 Consider, for example, the pattern \d+foo when applied to the subject
3210     line
3211 nigel 53
3212 nigel 73 123456bar
3213 nigel 53
3214 nigel 73 After matching all 6 digits and then failing to match "foo", the normal
3215     action of the matcher is to try again with only 5 digits matching the
3216     \d+ item, and then with 4, and so on, before ultimately failing.
3217     "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
3218     the means for specifying that once a subpattern has matched, it is not
3219     to be re-evaluated in this way.
3220 nigel 53
3221 nigel 73 If we use atomic grouping for the previous example, the matcher would
3222     give up immediately on failing to match "foo" the first time. The nota-
3223     tion is a kind of special parenthesis, starting with (?> as in this
3224     example:
3225 nigel 53
3226 nigel 73 (?>\d+)foo
3227 nigel 53
3228 nigel 73 This kind of parenthesis "locks up" the part of the pattern it con-
3229     tains once it has matched, and a failure further into the pattern is
3230     prevented from backtracking into it. Backtracking past it to previous
3231     items, however, works as normal.
3232 nigel 53
3233 nigel 73 An alternative description is that a subpattern of this type matches
3234     the string of characters that an identical standalone pattern would
3235     match, if anchored at the current point in the subject string.
3236 nigel 63
3237 nigel 73 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
3238     such as the above example can be thought of as a maximizing repeat that
3239     must swallow everything it can. So, while both \d+ and \d+? are pre-
3240     pared to adjust the number of digits they match in order to make the
3241     rest of the pattern match, (?>\d+) can only match an entire sequence of
3242     digits.
3243 nigel 63
3244 nigel 73 Atomic groups in general can of course contain arbitrarily complicated
3245     subpatterns, and can be nested. However, when the subpattern for an
3246     atomic group is just a single repeated item, as in the example above, a
3247     simpler notation, called a "possessive quantifier" can be used. This
3248     consists of an additional + character following a quantifier. Using
3249     this notation, the previous example can be rewritten as
3250 nigel 63
3251 nigel 75 \d++foo
3252 nigel 63
3253 nigel 73 Possessive quantifiers are always greedy; the setting of the
3254     PCRE_UNGREEDY option is ignored. They are a convenient notation for the
3255     simpler forms of atomic group. However, there is no difference in the
3256     meaning or processing of a possessive quantifier and the equivalent
3257     atomic group.
3258 nigel 63
3259 nigel 73 The possessive quantifier syntax is an extension to the Perl syntax. It
3260     originates in Sun's Java package.
3261 nigel 63
3262 nigel 73 When a pattern contains an unlimited repeat inside a subpattern that
3263     can itself be repeated an unlimited number of times, the use of an
3264     atomic group is the only way to avoid some failing matches taking a
3265     very long time indeed. The pattern
3266 nigel 63
3267 nigel 73 (\D+|<\d+>)*[!?]
3268 nigel 63
3269 nigel 73 matches an unlimited number of substrings that either consist of non-
3270     digits, or digits enclosed in <>, followed by either ! or ?. When it
3271     matches, it runs quickly. However, if it is applied to
3272 nigel 63
3273 nigel 73 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
3274 nigel 63
3275 nigel 73 it takes a long time before reporting failure. This is because the
3276 nigel 75 string can be divided between the internal \D+ repeat and the external
3277     * repeat in a large number of ways, and all have to be tried. (The
3278     example uses [!?] rather than a single character at the end, because
3279     both PCRE and Perl have an optimization that allows for fast failure
3280     when a single character is used. They remember the last single charac-
3281     ter that is required for a match, and fail early if it is not present
3282     in the string.) If the pattern is changed so that it uses an atomic
3283     group, like this:
3284 nigel 63
3285 nigel 73 ((?>\D+)|<\d+>)*[!?]
3286 nigel 63
3287 nigel 75 sequences of non-digits cannot be broken, and failure happens quickly.
3288 nigel 63
3292 nigel 73 Outside a character class, a backslash followed by a digit greater than
3293     0 (and possibly further digits) is a back reference to a capturing sub-
3294 nigel 75 pattern earlier (that is, to its left) in the pattern, provided there
3295 nigel 73 have been that many previous capturing left parentheses.
3296 nigel 41
3297 nigel 73 However, if the decimal number following the backslash is less than 10,
3298 nigel 75 it is always taken as a back reference, and causes an error only if
3299     there are not that many capturing left parentheses in the entire pat-
3300     tern. In other words, the parentheses that are referenced need not be
3301     to the left of the reference for numbers less than 10. See the subsec-
3302     tion entitled "Non-printing characters" above for further details of
3303     the handling of digits following a backslash.
3304 nigel 41
3305 nigel 75 A back reference matches whatever actually matched the capturing sub-
3306     pattern in the current subject string, rather than anything matching
3307 nigel 73 the subpattern itself (see "Subpatterns as subroutines" below for a way
3308     of doing that). So the pattern
3309 nigel 41
3310 nigel 73 (sens|respons)e and \1ibility
3311 nigel 41
3312 nigel 75 matches "sense and sensibility" and "response and responsibility", but
3313     not "sense and responsibility". If caseful matching is in force at the
3314     time of the back reference, the case of letters is relevant. For exam-
3315 nigel 73 ple,
3316 nigel 41
3317 nigel 73 ((?i)rah)\s+\1
3318 nigel 41
3319 nigel 75 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
3320 nigel 73 original capturing subpattern is matched caselessly.
3321 nigel 41
3322 nigel 75 Back references to named subpatterns use the Python syntax (?P=name).
3323 nigel 73 We could rewrite the above example as follows:
3324 nigel 63
3325 nigel 73 (?<p1>(?i)rah)\s+(?P=p1)
3326 nigel 63
3327 nigel 75 There may be more than one back reference to the same subpattern. If a
3328     subpattern has not actually been used in a particular match, any back
3329 nigel 73 references to it always fail. For example, the pattern
3330 nigel 41
3331 nigel 73 (a|(bc))\2
3332 nigel 41
3333 nigel 75 always fails if it starts to match "a" rather than "bc". Because there
3334     may be many capturing parentheses in a pattern, all digits following
3335     the backslash are taken as part of a potential back reference number.
3336 nigel 73 If the pattern continues with a digit character, some delimiter must be
3337 nigel 75 used to terminate the back reference. If the PCRE_EXTENDED option is
3338     set, this can be whitespace. Otherwise an empty comment (see "Com-
3339     ments" below) can be used.
3340 nigel 41
3341 nigel 73 A back reference that occurs inside the parentheses to which it refers
3342     fails when the subpattern is first used, so, for example, (a\1) never
3343     matches. However, such references can be useful inside repeated sub-
3344     patterns. For example, the pattern
3345 nigel 41
3346 nigel 73 (a|b\1)+
3347 nigel 41
3348 nigel 73 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
3349     ation of the subpattern, the back reference matches the character
3350     string corresponding to the previous iteration. In order for this to
3351     work, the pattern must be such that the first iteration does not need
3352     to match the back reference. This can be done using alternation, as in
3353     the example above, or by a quantifier with a minimum of zero.
3354 nigel 41
3356 nigel 63 ASSERTIONS
3357 nigel 41
3358 nigel 73 An assertion is a test on the characters following or preceding the
3359     current matching point that does not actually consume any characters.
3360     The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
3361 nigel 75 described above.
3362 nigel 43
3363 nigel 75 More complicated assertions are coded as subpatterns. There are two
3364     kinds: those that look ahead of the current position in the subject
3365     string, and those that look behind it. An assertion subpattern is
3366     matched in the normal way, except that it does not cause the current
3367     matching position to be changed.
3368 nigel 41
3369 nigel 75 Assertion subpatterns are not capturing subpatterns, and may not be
3370     repeated, because it makes no sense to assert the same thing several
3371     times. If any kind of assertion contains capturing subpatterns within
3372     it, these are counted for the purposes of numbering the capturing sub-
3373     patterns in the whole pattern. However, substring capturing is carried
3374     out only for positive assertions, because it does not make sense for
3375     negative assertions.
3377     Lookahead assertions
3379     Lookahead assertions start with (?= for positive assertions and (?! for
3380     negative assertions. For example,
3382 nigel 73 \w+(?=;)
3383 nigel 41
3384 nigel 73 matches a word followed by a semicolon, but does not include the semi-
3385     colon in the match, and
3386 nigel 41
3387 nigel 73 foo(?!bar)
3388 nigel 41
3389 nigel 73 matches any occurrence of "foo" that is not followed by "bar". Note
3390     that the apparently similar pattern
3391 nigel 41
3392 nigel 73 (?!foo)bar
3393 nigel 41
3394 nigel 73 does not find an occurrence of "bar" that is preceded by something
3395     other than "foo"; it finds any occurrence of "bar" whatsoever, because
3396     the assertion (?!foo) is always true when the next three characters are
3397 nigel 75 "bar". A lookbehind assertion is needed to achieve the other effect.
3398 nigel 41
3399 nigel 73 If you want to force a matching failure at some point in a pattern, the
3400     most convenient way to do it is with (?!) because an empty string
3401     always matches, so an assertion that requires there not to be an empty
3402     string must always fail.
3403 nigel 63
3404 nigel 75 Lookbehind assertions
3406 nigel 73 Lookbehind assertions start with (?<= for positive assertions and (?<!
3407     for negative assertions. For example,
3408 nigel 41
3409 nigel 73 (?<!foo)bar
3410 nigel 41
3411 nigel 73 does find an occurrence of "bar" that is not preceded by "foo". The
3412     contents of a lookbehind assertion are restricted such that all the
3413     strings it matches must have a fixed length. However, if there are sev-
3414     eral alternatives, they do not all have to have the same fixed length.
3415     Thus
3416 nigel 41
3417 nigel 73 (?<=bullock|donkey)
3418 nigel 41
3419 nigel 73 is permitted, but
3420 nigel 41
3421 nigel 73 (?<!dogs?|cats?)
3422 nigel 41
3423 nigel 73 causes an error at compile time. Branches that match different length
3424     strings are permitted only at the top level of a lookbehind assertion.
3425     This is an extension compared with Perl (at least for 5.8), which
3426     requires all branches to match the same length of string. An assertion
3427     such as
3428 nigel 41
3429 nigel 73 (?<=ab(c|de))
3430 nigel 41
3431 nigel 73 is not permitted, because its single top-level branch can match two
3432     different lengths, but it is acceptable if rewritten to use two top-
3433     level branches:
3434 nigel 41
3435 nigel 73 (?<=abc|abde)
3436 nigel 41
3437 nigel 73 The implementation of lookbehind assertions is, for each alternative,
3438     to temporarily move the current position back by the fixed width and
3439     then try to match. If there are insufficient characters before the cur-
3440     rent position, the match is deemed to fail.
3441 nigel 41
3442 nigel 73 PCRE does not allow the \C escape (which matches a single byte in UTF-8
3443     mode) to appear in lookbehind assertions, because it makes it impossi-
3444 nigel 75 ble to calculate the length of the lookbehind. The \X escape, which can
3445     match different numbers of bytes, is also not permitted.
3446 nigel 63
3447 nigel 75 Atomic groups can be used in conjunction with lookbehind assertions to
3448 nigel 73 specify efficient matching at the end of the subject string. Consider a
3449     simple pattern such as
3450 nigel 63
3451 nigel 73 abcd$
3452 nigel 63
3453 nigel 75 when applied to a long string that does not match. Because matching
3454 nigel 73 proceeds from left to right, PCRE will look for each "a" in the subject
3455 nigel 75 and then see if what follows matches the rest of the pattern. If the
3456 nigel 73 pattern is specified as
3457 nigel 63
3458 nigel 73 ^.*abcd$
3459 nigel 63
3460 nigel 75 the initial .* matches the entire string at first, but when this fails
3461 nigel 73 (because there is no following "a"), it backtracks to match all but the
3462 nigel 75 last character, then all but the last two characters, and so on. Once
3463     again the search for "a" covers the entire string, from right to left,
3464 nigel 73 so we are no better off. However, if the pattern is written as
3465 nigel 63
3466 nigel 73 ^(?>.*)(?<=abcd)
3467 nigel 63
3468 nigel 75 or, equivalently, using the possessive quantifier syntax,
3469 nigel 63
3470 nigel 73 ^.*+(?<=abcd)
3471 nigel 63
3472 nigel 75 there can be no backtracking for the .* item; it can match only the
3473     entire string. The subsequent lookbehind assertion does a single test
3474     on the last four characters. If it fails, the match fails immediately.
3475     For long strings, this approach makes a significant difference to the
3476 nigel 73 processing time.
3477 nigel 63
3478 nigel 75 Using multiple assertions
3480 nigel 73 Several assertions (of any sort) may occur in succession. For example,
3481 nigel 41
3482 nigel 73 (?<=\d{3})(?<!999)foo
3483 nigel 41
3484 nigel 75 matches "foo" preceded by three digits that are not "999". Notice that
3485     each of the assertions is applied independently at the same point in
3486     the subject string. First there is a check that the previous three
3487     characters are all digits, and then there is a check that