ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1490 - (hide annotations) (download)
Thu Jun 19 07:51:39 2014 UTC (10 months, 1 week ago) by chpe
File size: 135516 byte(s)
Update to Unicode 7.0.0 release
1 ph10 1436 .TH PCREPATTERN 3 "08 January 2014" "PCRE 8.35"
2 nigel 63 .SH NAME
3     PCRE - Perl-compatible regular expressions
5 nigel 63 .rs
6     .sp
7 ph10 208 The syntax and semantics of the regular expressions that are supported by PCRE
8     are described in detail below. There is a quick-reference syntax summary in the
9     .\" HREF
10     \fBpcresyntax\fP
11     .\"
12 ph10 333 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
13     also supports some alternative regular expression syntax (which does not
14     conflict with the Perl syntax) in order to provide some compatibility with
15     regular expressions in Python, .NET, and Oniguruma.
16     .P
17     Perl's regular expressions are described in its own documentation, and
18 ph10 208 regular expressions in general are covered in a number of books, some of which
19     have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
20     published by O'Reilly, covers regular expressions in great detail. This
21     description of PCRE's regular expressions is intended as reference material.
22 nigel 75 .P
23 ph10 1314 This document discusses the patterns that are supported by PCRE when one its
24     main matching functions, \fBpcre_exec()\fP (8-bit) or \fBpcre[16|32]_exec()\fP
25     (16- or 32-bit), is used. PCRE also has alternative matching functions,
26     \fBpcre_dfa_exec()\fP and \fBpcre[16|32_dfa_exec()\fP, which match using a
27     different algorithm that is not Perl-compatible. Some of the features discussed
28     below are not available when DFA matching is used. The advantages and
29     disadvantages of the alternative functions, and how they differ from the normal
30     functions, are discussed in the
31     .\" HREF
32     \fBpcrematching\fP
33     .\"
34     page.
35     .
36     .
38     .rs
39     .sp
40 ph10 1335 A number of options that can be passed to \fBpcre_compile()\fP can also be set
41 ph10 1314 by special items at the start of a pattern. These are not Perl-compatible, but
42     are provided to make these options accessible to pattern writers who are not
43     able to change the program that processes the pattern. Any number of these
44     items may appear, but they must all be together right at the start of the
45     pattern string, and the letters must be in upper case.
46     .
47     .
48     .SS "UTF support"
49     .rs
50     .sp
51 nigel 75 The original operation of PCRE was on strings of one-byte characters. However,
52 chpe 1055 there is now also support for UTF-8 strings in the original library, an
53 ph10 1219 extra library that supports 16-bit and UTF-16 character strings, and a
54     third library that supports 32-bit and UTF-32 character strings. To use these
55 ph10 859 features, PCRE must be built to include appropriate support. When using UTF
56 chpe 1055 strings you must either call the compiling function with the PCRE_UTF8,
57 ph10 1219 PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of
58 chpe 1055 these special sequences:
59 ph10 412 .sp
60     (*UTF8)
61 ph10 903 (*UTF16)
62 chpe 1055 (*UTF32)
63 ph10 1221 (*UTF)
64 ph10 416 .sp
65 ph10 1219 (*UTF) is a generic sequence that can be used with any of the libraries.
66 ph10 859 Starting a pattern with such a sequence is equivalent to setting the relevant
67 ph10 1314 option. How setting a UTF mode affects pattern matching is mentioned in several
68     places below. There is also a summary of features in the
69 nigel 63 .\" HREF
70 ph10 678 \fBpcreunicode\fP
71 nigel 63 .\"
72     page.
73 nigel 75 .P
74 ph10 1335 Some applications that allow their users to supply patterns may wish to
75 ph10 1314 restrict them to non-UTF data for security reasons. If the PCRE_NEVER_UTF
76     option is set at compile time, (*UTF) etc. are not allowed, and their
77     appearance causes an error.
78     .
79     .
80     .SS "Unicode property support"
81     .rs
82 ph10 518 .sp
83 ph10 1395 Another special sequence that may appear at the start of a pattern is (*UCP).
84 ph10 535 This has the same effect as setting the PCRE_UCP option: it causes sequences
85     such as \ed and \ew to use Unicode properties to determine character types,
86     instead of recognizing only characters with codes less than 128 via a lookup
87 ph10 518 table.
88 nigel 93 .
89     .
90 ph10 1395 .SS "Disabling auto-possessification"
91     .rs
92     .sp
93 ph10 1404 If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting
94 ph10 1398 the PCRE_NO_AUTO_POSSESS option at compile time. This stops PCRE from making
95 ph10 1395 quantifiers possessive when what follows cannot match the repeated item. For
96     example, by default a+b is treated as a++b. For more details, see the
97     .\" HREF
98     \fBpcreapi\fP
99     .\"
100     documentation.
101     .
102     .
103 ph10 1314 .SS "Disabling start-up optimizations"
104 ph10 1033 .rs
105     .sp
106 ph10 1314 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
107 ph10 1395 PCRE_NO_START_OPTIMIZE option either at compile or matching time. This disables
108     several optimizations for quickly reaching "no match" results. For more
109     details, see the
110     .\" HREF
111     \fBpcreapi\fP
112     .\"
113     documentation.
114 ph10 1033 .
115     .
116 ph10 556 .\" HTML <a name="newlines"></a>
117 ph10 1314 .SS "Newline conventions"
118 ph10 227 .rs
119     .sp
120     PCRE supports five different conventions for indicating line breaks in
121     strings: a single CR (carriage return) character, a single LF (linefeed)
122     character, the two-character sequence CRLF, any of the three preceding, or any
123     Unicode newline sequence. The
124     .\" HREF
125     \fBpcreapi\fP
126     .\"
127     page has
128     .\" HTML <a href="pcreapi.html#newlines">
129     .\" </a>
130     further discussion
131     .\"
132     about newlines, and shows how to set the newline convention in the
133     \fIoptions\fP arguments for the compiling and matching functions.
134     .P
135     It is also possible to specify a newline convention by starting a pattern
136     string with one of the following five sequences:
137     .sp
138     (*CR) carriage return
139     (*LF) linefeed
140     (*CRLF) carriage return, followed by linefeed
141     (*ANYCRLF) any of the three above
142     (*ANY) all Unicode newline sequences
143     .sp
144 ph10 859 These override the default and the options given to the compiling function. For
145     example, on a Unix system where LF is the default newline sequence, the pattern
146 ph10 227 .sp
147     (*CR)a.b
148     .sp
149     changes the convention to CR. That pattern matches "a\enb" because LF is no
150 ph10 1314 longer a newline. If more than one of these settings is present, the last one
151 ph10 231 is used.
152     .P
153 ph10 1213 The newline convention affects where the circumflex and dollar assertions are
154     true. It also affects the interpretation of the dot metacharacter when
155     PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
156     what the \eR escape sequence matches. By default, this is any Unicode newline
157     sequence, for Perl compatibility. However, this can be changed; see the
158 ph10 514 description of \eR in the section entitled
159 ph10 231 .\" HTML <a href="#newlineseq">
160     .\" </a>
161     "Newline sequences"
162     .\"
163 ph10 247 below. A change of \eR setting can be combined with a change of newline
164 ph10 246 convention.
165 ph10 227 .
166     .
167 ph10 1314 .SS "Setting match and recursion limits"
168     .rs
169     .sp
170 ph10 1335 The caller of \fBpcre_exec()\fP can set a limit on the number of times the
171     internal \fBmatch()\fP function is called and on the maximum depth of
172 ph10 1314 recursive calls. These facilities are provided to catch runaway matches that
173     are provoked by patterns with huge matching trees (a typical example is a
174     pattern with nested unlimited repeats) and to avoid running out of system stack
175     by too much recursion. When one of these limits is reached, \fBpcre_exec()\fP
176 ph10 1335 gives an error return. The limits can also be set by items at the start of the
177 ph10 1314 pattern of the form
178     .sp
179     (*LIMIT_MATCH=d)
180     (*LIMIT_RECURSION=d)
181     .sp
182 ph10 1335 where d is any number of decimal digits. However, the value of the setting must
183 ph10 1391 be less than the value set (or defaulted) by the caller of \fBpcre_exec()\fP
184     for it to have any effect. In other words, the pattern writer can lower the
185     limits set by the programmer, but not raise them. If there is more than one
186     setting of one of these limits, the lower value is used.
187 ph10 1314 .
188     .
190     .rs
191     .sp
192     PCRE can be compiled to run in an environment that uses EBCDIC as its character
193     code rather than ASCII or Unicode (typically a mainframe system). In the
194     sections below, character code values are ASCII or Unicode; in an EBCDIC
195     environment these characters may have different code values, and there are no
196     code points greater than 255.
197     .
198     .
200     .rs
201     .sp
202 nigel 63 A regular expression is a pattern that is matched against a subject string from
203     left to right. Most characters stand for themselves in a pattern, and match the
204     corresponding characters in the subject. As a trivial example, the pattern
205 nigel 75 .sp
206 nigel 63 The quick brown fox
207 nigel 75 .sp
208 nigel 77 matches a portion of a subject string that is identical to itself. When
209     caseless matching is specified (the PCRE_CASELESS option), letters are matched
210 ph10 859 independently of case. In a UTF mode, PCRE always understands the concept of
211 nigel 77 case for characters whose values are less than 128, so caseless matching is
212     always possible. For characters with higher values, the concept of case is
213     supported if PCRE is compiled with Unicode property support, but not otherwise.
214     If you want to use caseless matching for characters 128 and above, you must
215     ensure that PCRE is compiled with Unicode property support as well as with
216 ph10 859 UTF support.
217 nigel 77 .P
218     The power of regular expressions comes from the ability to include alternatives
219     and repetitions in the pattern. These are encoded in the pattern by the use of
220 nigel 75 \fImetacharacters\fP, which do not stand for themselves but instead are
221 nigel 63 interpreted in some special way.
222 nigel 75 .P
223     There are two different sets of metacharacters: those that are recognized
224 nigel 63 anywhere in the pattern except within square brackets, and those that are
225 nigel 93 recognized within square brackets. Outside square brackets, the metacharacters
226     are as follows:
227 nigel 75 .sp
228     \e general escape character with several uses
229 nigel 63 ^ assert start of string (or line, in multiline mode)
230     $ assert end of string (or line, in multiline mode)
231     . match any character except newline (by default)
232     [ start character class definition
233     | start of alternative branch
234     ( start subpattern
235     ) end subpattern
236     ? extends the meaning of (
237     also 0 or 1 quantifier
238     also quantifier minimizer
239     * 0 or more quantifier
240     + 1 or more quantifier
241     also "possessive quantifier"
242     { start min/max quantifier
243 nigel 75 .sp
244 nigel 63 Part of a pattern that is in square brackets is called a "character class". In
245 nigel 75 a character class the only metacharacters are:
246     .sp
247     \e general escape character
248 nigel 63 ^ negate the class, but only if the first character
249     - indicates character range
250 nigel 75 .\" JOIN
251 nigel 63 [ POSIX character class (only if followed by POSIX
252     syntax)
253     ] terminates the character class
254 nigel 75 .sp
255     The following sections describe the use of each of the metacharacters.
256     .
257 nigel 93 .
258 nigel 63 .SH BACKSLASH
259     .rs
260     .sp
261     The backslash character has several uses. Firstly, if it is followed by a
262 ph10 574 character that is not a number or a letter, it takes away any special meaning
263     that character may have. This use of backslash as an escape character applies
264 ph10 579 both inside and outside character classes.
265 nigel 75 .P
266     For example, if you want to match a * character, you write \e* in the pattern.
267 nigel 63 This escaping action applies whether or not the following character would
268 nigel 75 otherwise be interpreted as a metacharacter, so it is always safe to precede a
269     non-alphanumeric with backslash to specify that it stands for itself. In
270     particular, if you want to match a backslash, you write \e\e.
271     .P
272 ph10 859 In a UTF mode, only ASCII numbers and letters have any special meaning after a
273 ph10 579 backslash. All other characters (in particular, those whose codepoints are
274 ph10 574 greater than 127) are treated as literals.
275     .P
276 ph10 1396 If a pattern is compiled with the PCRE_EXTENDED option, most white space in the
277     pattern (other than in a character class), and characters between a # outside a
278     character class and the next newline, inclusive, are ignored. An escaping
279     backslash can be used to include a white space or # character as part of the
280     pattern.
281 nigel 75 .P
282 nigel 63 If you want to remove the special meaning from a sequence of characters, you
283 nigel 75 can do so by putting them between \eQ and \eE. This is different from Perl in
284     that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
285 nigel 63 Perl, $ and @ cause variable interpolation. Note the following examples:
286 nigel 75 .sp
287 nigel 63 Pattern PCRE matches Perl matches
288 nigel 75 .sp
289     .\" JOIN
290     \eQabc$xyz\eE abc$xyz abc followed by the
291 nigel 63 contents of $xyz
292 nigel 75 \eQabc\e$xyz\eE abc\e$xyz abc\e$xyz
293     \eQabc\eE\e$\eQxyz\eE abc$xyz abc$xyz
294     .sp
295     The \eQ...\eE sequence is recognized both inside and outside character classes.
296 ph10 654 An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
297     by \eE later in the pattern, the literal interpretation continues to the end of
298 ph10 607 the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
299     a character class, this causes an error, because the character class is not
300     terminated.
301 nigel 75 .
302     .
303     .\" HTML <a name="digitsafterbackslash"></a>
304     .SS "Non-printing characters"
305     .rs
306     .sp
307 nigel 63 A second use of backslash provides a way of encoding non-printing characters
308     in patterns in a visible manner. There is no restriction on the appearance of
309     non-printing characters, apart from the binary zero that terminates a pattern,
310 ph10 456 but when a pattern is being prepared by text editing, it is often easier to use
311     one of the following escape sequences than the binary character it represents:
312 nigel 75 .sp
313     \ea alarm, that is, the BEL character (hex 07)
314 ph10 574 \ecx "control-x", where x is any ASCII character
315 nigel 75 \ee escape (hex 1B)
316 ph10 968 \ef form feed (hex 0C)
317 ph10 227 \en linefeed (hex 0A)
318 nigel 75 \er carriage return (hex 0D)
319     \et tab (hex 09)
320 ph10 1404 \e0dd character with octal code 0dd
321 ph10 488 \eddd character with octal code ddd, or back reference
322 ph10 1404 \eo{ddd..} character with octal code ddd..
323 nigel 75 \exhh character with hex code hh
324 ph10 745 \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
325 ph10 836 \euhhhh character with hex code hhhh (JavaScript mode only)
326 nigel 75 .sp
327 ph10 1001 The precise effect of \ecx on ASCII characters is as follows: if x is a lower
328     case letter, it is converted to upper case. Then bit 6 of the character (hex
329     40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
330     but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
331     data item (byte or 16-bit value) following \ec has a value greater than 127, a
332     compile-time error occurs. This locks out non-ASCII characters in all modes.
333 nigel 75 .P
334 ph10 1001 The \ec facility was designed for use with ASCII characters, but with the
335     extension to Unicode it is even less useful than it once was. It is, however,
336     recognized when PCRE is compiled in EBCDIC mode, where data items are always
337     bytes. In this mode, all values are valid after \ec. If the next character is a
338     lower case letter, it is converted to upper case. Then the 0xc0 bits of the
339     byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
340 ph10 1221 the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
341 ph10 1001 characters also generate different values.
342     .P
343 nigel 91 After \e0 up to two further octal digits are read. If there are fewer than two
344     digits, just those that are present are used. Thus the sequence \e0\ex\e07
345     specifies two binary zeros followed by a BEL character (code value 7). Make
346     sure you supply two digits after the initial zero if the pattern character that
347     follows is itself an octal digit.
348 nigel 75 .P
349 ph10 1404 The escape \eo must be followed by a sequence of octal digits, enclosed in
350 ph10 1370 braces. An error occurs if this is not the case. This escape is a recent
351     addition to Perl; it provides way of specifying character code points as octal
352     numbers greater than 0777, and it also allows octal numbers and back references
353     to be unambiguously specified.
354     .P
355     For greater clarity and unambiguity, it is best to avoid following \e by a
356     digit greater than zero. Instead, use \eo{} or \ex{} to specify character
357     numbers, and \eg{} to specify back references. The following paragraphs
358     describe the old, ambiguous syntax.
359     .P
360 ph10 1369 The handling of a backslash followed by a digit other than 0 is complicated,
361     and Perl has changed in recent releases, causing PCRE also to change. Outside a
362     character class, PCRE reads the digit and any following digits as a decimal
363     number. If the number is less than 8, or if there have been at least that many
364 nigel 63 previous capturing left parentheses in the expression, the entire sequence is
365 nigel 75 taken as a \fIback reference\fP. A description of how this works is given
366     .\" HTML <a href="#backreferences">
367     .\" </a>
368     later,
369     .\"
370     following the discussion of
371     .\" HTML <a href="#subpattern">
372     .\" </a>
373     parenthesized subpatterns.
374     .\"
375     .P
376 ph10 1369 Inside a character class, or if the decimal number following \e is greater than
377     7 and there have not been that many capturing subpatterns, PCRE handles \e8 and
378     \e9 as the literal characters "8" and "9", and otherwise re-reads up to three
379     octal digits following the backslash, using them to generate a data character.
380 ph10 1370 Any subsequent digits stand for themselves. For example:
381 nigel 75 .sp
382 ph10 1033 \e040 is another way of writing an ASCII space
383 nigel 75 .\" JOIN
384     \e40 is the same, provided there are fewer than 40
385 nigel 63 previous capturing subpatterns
386 nigel 75 \e7 is always a back reference
387     .\" JOIN
388     \e11 might be a back reference, or another way of
389 nigel 63 writing a tab
390 nigel 75 \e011 is always a tab
391     \e0113 is a tab followed by the character "3"
392     .\" JOIN
393     \e113 might be a back reference, otherwise the
394 nigel 63 character with octal code 113
395 nigel 75 .\" JOIN
396     \e377 might be a back reference, otherwise
397 ph10 859 the value 255 (decimal)
398 nigel 75 .\" JOIN
399 ph10 1369 \e81 is either a back reference, or the two
400     characters "8" and "1"
401 nigel 75 .sp
402 ph10 1370 Note that octal values of 100 or greater that are specified using this syntax
403     must not be introduced by a leading zero, because no more than three octal
404     digits are ever read.
405 nigel 75 .P
406 ph10 1370 By default, after \ex that is not followed by {, from zero to two hexadecimal
407     digits are read (letters can be in upper or lower case). Any number of
408     hexadecimal digits may appear between \ex{ and }. If a character other than
409     a hexadecimal digit appears between \ex{ and }, or if there is no terminating
410     }, an error occurs.
411     .P
412     If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
413     as just described only when it is followed by two hexadecimal digits.
414     Otherwise, it matches a literal "x" character. In JavaScript mode, support for
415     code points greater than 256 is provided by \eu, which must be followed by
416     four hexadecimal digits; otherwise it matches a literal "u" character.
417     .P
418     Characters whose value is less than 256 can be defined by either of the two
419     syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
420     way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
421     \eu00dc in JavaScript mode).
422     .
423     .
424     .SS "Constraints on character values"
425     .rs
426     .sp
427     Characters that are specified using octal or hexadecimal numbers are
428     limited to certain values, as follows:
429     .sp
430     8-bit non-UTF mode less than 0x100
431     8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
432     16-bit non-UTF mode less than 0x10000
433     16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
434 chpe 1373 32-bit non-UTF mode less than 0x100000000
435 ph10 1370 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
436     .sp
437     Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
438 ph10 1404 "surrogate" codepoints), and 0xffef.
439 ph10 1370 .
440     .
441     .SS "Escape sequences in character classes"
442     .rs
443     .sp
444 nigel 91 All the sequences that define a single character value can be used both inside
445 ph10 836 and outside character classes. In addition, inside a character class, \eb is
446     interpreted as the backspace character (hex 08).
447     .P
448     \eN is not allowed in a character class. \eB, \eR, and \eX are not special
449     inside a character class. Like other unrecognized escape sequences, they are
450     treated as the literal characters "B", "R", and "X" by default, but cause an
451     error if the PCRE_EXTRA option is set. Outside a character class, these
452     sequences have different meanings.
453 nigel 75 .
454     .
455 ph10 745 .SS "Unsupported escape sequences"
456     .rs
457     .sp
458     In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
459     handler and used to modify the case of following characters. By default, PCRE
460     does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
461     option is set, \eU matches a "U" character, and \eu can be used to define a
462     character by code point, as described in the previous section.
463     .
464     .
465 nigel 93 .SS "Absolute and relative back references"
466     .rs
467     .sp
468 ph10 208 The sequence \eg followed by an unsigned or a negative number, optionally
469     enclosed in braces, is an absolute or relative back reference. A named back
470     reference can be coded as \eg{name}. Back references are discussed
471 nigel 93 .\" HTML <a href="#backreferences">
472     .\" </a>
473     later,
474     .\"
475     following the discussion of
476     .\" HTML <a href="#subpattern">
477     .\" </a>
478     parenthesized subpatterns.
479     .\"
480     .
481     .
482 ph10 333 .SS "Absolute and relative subroutine calls"
483     .rs
484     .sp
485 ph10 345 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
486     a number enclosed either in angle brackets or single quotes, is an alternative
487     syntax for referencing a subpattern as a "subroutine". Details are discussed
488 ph10 333 .\" HTML <a href="#onigurumasubroutines">
489     .\" </a>
490     later.
491     .\"
492 ph10 345 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
493 ph10 461 synonymous. The former is a back reference; the latter is a
494 ph10 454 .\" HTML <a href="#subpatternsassubroutines">
495     .\" </a>
496     subroutine
497     .\"
498     call.
499 ph10 333 .
500     .
501 ph10 518 .\" HTML <a name="genericchartypes"></a>
502 nigel 75 .SS "Generic character types"
503     .rs
504     .sp
505 ph10 514 Another use of backslash is for specifying generic character types:
506 nigel 75 .sp
507 ph10 182 \ed any decimal digit
508 nigel 75 \eD any character that is not a decimal digit
509 ph10 968 \eh any horizontal white space character
510     \eH any character that is not a horizontal white space character
511     \es any white space character
512     \eS any character that is not a white space character
513     \ev any vertical white space character
514     \eV any character that is not a vertical white space character
515 nigel 75 \ew any "word" character
516     \eW any "non-word" character
517     .sp
518 ph10 535 There is also the single sequence \eN, which matches a non-newline character.
519     This is the same as
520 ph10 514 .\" HTML <a href="#fullstopdot">
521     .\" </a>
522 ph10 535 the "." metacharacter
523 ph10 514 .\"
524 ph10 836 when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
525 ph10 745 PCRE does not support this.
526 nigel 75 .P
527 ph10 514 Each pair of lower and upper case escape sequences partitions the complete set
528     of characters into two disjoint sets. Any given character matches one, and only
529 ph10 518 one, of each pair. The sequences can appear both inside and outside character
530 nigel 75 classes. They each match one character of the appropriate type. If the current
531 ph10 518 matching point is at the end of the subject string, all of them fail, because
532 nigel 75 there is no character to match.
533     .P
534 ph10 1364 For compatibility with Perl, \es did not used to match the VT character (code
535     11), which made it different from the the POSIX "space" class. However, Perl
536 ph10 1401 added VT at release 5.18, and PCRE followed suit at release 8.34. The default
537     \es characters are now HT (9), LF (10), VT (11), FF (12), CR (13), and space
538 ph10 1404 (32), which are defined as white space in the "C" locale. This list may vary if
539 ph10 1405 locale-specific matching is taking place. For example, in some locales the
540 ph10 1412 "non-breaking space" character (\exA0) is recognized as white space, and in
541 ph10 1405 others the VT character is not.
542 nigel 75 .P
543 ph10 518 A "word" character is an underscore or any character that is a letter or digit.
544     By default, the definition of letters and digits is controlled by PCRE's
545     low-valued character tables, and may vary if locale-specific matching is taking
546     place (see
547     .\" HTML <a href="pcreapi.html#localesupport">
548     .\" </a>
549     "Locale support"
550     .\"
551     in the
552     .\" HREF
553     \fBpcreapi\fP
554     .\"
555     page). For example, in a French locale such as "fr_FR" in Unix-like systems,
556 ph10 1401 or "french" in Windows, some character codes greater than 127 are used for
557 ph10 518 accented letters, and these are then matched by \ew. The use of locales with
558     Unicode is discouraged.
559 ph10 178 .P
560 ph10 1401 By default, characters whose code points are greater than 127 never match \ed,
561     \es, or \ew, and always match \eD, \eS, and \eW, although this may vary for
562     characters in the range 128-255 when locale-specific matching is happening.
563     These escape sequences retain their original meanings from before Unicode
564     support was available, mainly for efficiency reasons. If PCRE is compiled with
565     Unicode property support, and the PCRE_UCP option is set, the behaviour is
566     changed so that Unicode properties are used to determine character types, as
567     follows:
568 ph10 518 .sp
569 ph10 1376 \ed any character that matches \ep{Nd} (decimal digit)
570     \es any character that matches \ep{Z} or \eh or \ev
571     \ew any character that matches \ep{L} or \ep{N}, plus underscore
572 ph10 518 .sp
573     The upper case escapes match the inverse sets of characters. Note that \ed
574 ph10 535 matches only decimal digits, whereas \ew matches any Unicode digit, as well as
575 ph10 518 any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
576     \eB because they are defined in terms of \ew and \eW. Matching these sequences
577     is noticeably slower when PCRE_UCP is set.
578     .P
579 ph10 579 The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
580 ph10 572 release 5.10. In contrast to the other sequences, which match only ASCII
581 ph10 1401 characters by default, these always match certain high-valued code points,
582 ph10 859 whether or not PCRE_UCP is set. The horizontal space characters are:
583 ph10 178 .sp
584 ph10 1033 U+0009 Horizontal tab (HT)
585 ph10 178 U+0020 Space
586     U+00A0 Non-break space
587     U+1680 Ogham space mark
588     U+180E Mongolian vowel separator
589     U+2000 En quad
590     U+2001 Em quad
591     U+2002 En space
592     U+2003 Em space
593     U+2004 Three-per-em space
594     U+2005 Four-per-em space
595     U+2006 Six-per-em space
596     U+2007 Figure space
597     U+2008 Punctuation space
598     U+2009 Thin space
599     U+200A Hair space
600     U+202F Narrow no-break space
601     U+205F Medium mathematical space
602     U+3000 Ideographic space
603     .sp
604     The vertical space characters are:
605     .sp
606 ph10 1033 U+000A Linefeed (LF)
607     U+000B Vertical tab (VT)
608     U+000C Form feed (FF)
609     U+000D Carriage return (CR)
610     U+0085 Next line (NEL)
611 ph10 178 U+2028 Line separator
612     U+2029 Paragraph separator
613 ph10 859 .sp
614 ph10 903 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
615     relevant.
616 nigel 75 .
617     .
618 ph10 231 .\" HTML <a name="newlineseq"></a>
619 nigel 93 .SS "Newline sequences"
620     .rs
621     .sp
622 ph10 231 Outside a character class, by default, the escape sequence \eR matches any
623 ph10 859 Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
624     following:
625 nigel 93 .sp
626     (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
627     .sp
628     This is an example of an "atomic group", details of which are given
629     .\" HTML <a href="#atomicgroup">
630     .\" </a>
631     below.
632     .\"
633     This particular group matches either the two-character sequence CR followed by
634     LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
635 ph10 968 U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
636 nigel 93 line, U+0085). The two-character sequence is treated as a single unit that
637     cannot be split.
638     .P
639 ph10 859 In other modes, two additional characters whose codepoints are greater than 255
640 nigel 93 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
641     Unicode character property support is not needed for these characters to be
642     recognized.
643     .P
644 ph10 231 It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
645     complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
646 ph10 247 either at compile time or when the pattern is matched. (BSR is an abbrevation
647 ph10 246 for "backslash R".) This can be made the default when PCRE is built; if this is
648     the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
649     It is also possible to specify these settings by starting a pattern string with
650     one of the following sequences:
651 ph10 231 .sp
652     (*BSR_ANYCRLF) CR, LF, or CRLF only
653     (*BSR_UNICODE) any Unicode newline sequence
654     .sp
655 ph10 859 These override the default and the options given to the compiling function, but
656     they can themselves be overridden by options given to a matching function. Note
657     that these special settings, which are not Perl-compatible, are recognized only
658     at the very start of a pattern, and that they must be in upper case. If more
659     than one of them is present, the last one is used. They can be combined with a
660     change of newline convention; for example, a pattern can start with:
661 ph10 246 .sp
662     (*ANY)(*BSR_ANYCRLF)
663     .sp
664 ph10 1219 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
665     (*UCP) special sequences. Inside a character class, \eR is treated as an
666     unrecognized escape sequence, and so matches the letter "R" by default, but
667     causes an error if PCRE_EXTRA is set.
668 nigel 93 .
669     .
670 nigel 75 .\" HTML <a name="uniextseq"></a>
671     .SS Unicode character properties
672     .rs
673     .sp
674     When PCRE is built with Unicode character property support, three additional
675 ph10 184 escape sequences that match characters with specific properties are available.
676 ph10 859 When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
677 ph10 184 characters whose codepoints are less than 256, but they do work in this mode.
678     The extra escape sequences are:
679 nigel 75 .sp
680 nigel 87 \ep{\fIxx\fP} a character with the \fIxx\fP property
681     \eP{\fIxx\fP} a character without the \fIxx\fP property
682 ph10 1011 \eX a Unicode extended grapheme cluster
683 nigel 75 .sp
684 nigel 87 The property names represented by \fIxx\fP above are limited to the Unicode
685 ph10 517 script names, the general category properties, "Any", which matches any
686     character (including newline), and some special PCRE properties (described
687 ph10 535 in the
688 ph10 517 .\" HTML <a href="#extraprops">
689     .\" </a>
690 ph10 535 next section).
691 ph10 517 .\"
692     Other Perl properties such as "InMusicalSymbols" are not currently supported by
693     PCRE. Note that \eP{Any} does not match any characters, so always causes a
694     match failure.
695 nigel 75 .P
696 nigel 87 Sets of Unicode characters are defined as belonging to certain scripts. A
697     character from one of these sets can be matched using a script name. For
698     example:
699 nigel 75 .sp
700 nigel 87 \ep{Greek}
701     \eP{Han}
702     .sp
703     Those that are not part of an identified script are lumped together as
704     "Common". The current list of scripts is:
705     .P
706     Arabic,
707     Armenian,
708 ph10 491 Avestan,
709 nigel 93 Balinese,
710 ph10 491 Bamum,
711 chpe 1490 Bassa_Vah,
712 ph10 942 Batak,
713 nigel 87 Bengali,
714     Bopomofo,
715 ph10 942 Brahmi,
716 nigel 87 Braille,
717     Buginese,
718     Buhid,
719     Canadian_Aboriginal,
720 ph10 491 Carian,
721 chpe 1490 Caucasian_Albanian,
722 ph10 942 Chakma,
723 ph10 491 Cham,
724 nigel 87 Cherokee,
725     Common,
726     Coptic,
727 nigel 93 Cuneiform,
728 nigel 87 Cypriot,
729     Cyrillic,
730     Deseret,
731     Devanagari,
732 chpe 1490 Duployan,
733 ph10 491 Egyptian_Hieroglyphs,
734 chpe 1490 Elbasan,
735 nigel 87 Ethiopic,
736     Georgian,
737     Glagolitic,
738     Gothic,
739 chpe 1490 Grantha,
740 nigel 87 Greek,
741     Gujarati,
742     Gurmukhi,
743     Han,
744     Hangul,
745     Hanunoo,
746     Hebrew,
747     Hiragana,
748 ph10 491 Imperial_Aramaic,
749 nigel 87 Inherited,
750 ph10 491 Inscriptional_Pahlavi,
751     Inscriptional_Parthian,
752     Javanese,
753     Kaithi,
754 nigel 87 Kannada,
755     Katakana,
756 ph10 491 Kayah_Li,
757 nigel 87 Kharoshthi,
758     Khmer,
759 chpe 1490 Khojki,
760     Khudawadi,
761 nigel 87 Lao,
762     Latin,
763 ph10 491 Lepcha,
764 nigel 87 Limbu,
765 chpe 1490 Linear_A,
766 nigel 87 Linear_B,
767 ph10 491 Lisu,
768     Lycian,
769     Lydian,
770 chpe 1490 Mahajani,
771 nigel 87 Malayalam,
772 ph10 942 Mandaic,
773 chpe 1490 Manichaean,
774 ph10 491 Meetei_Mayek,
775 chpe 1490 Mende_Kikakui,
776 ph10 942 Meroitic_Cursive,
777     Meroitic_Hieroglyphs,
778     Miao,
779 chpe 1490 Modi,
780 nigel 87 Mongolian,
781 chpe 1490 Mro,
782 nigel 87 Myanmar,
783 chpe 1490 Nabataean,
784 nigel 87 New_Tai_Lue,
785 nigel 93 Nko,
786 nigel 87 Ogham,
787 chpe 1490 Ol_Chiki,
788 nigel 87 Old_Italic,
789 chpe 1490 Old_North_Arabian,
790     Old_Permic,
791 nigel 87 Old_Persian,
792 ph10 491 Old_South_Arabian,
793     Old_Turkic,
794 nigel 87 Oriya,
795     Osmanya,
796 chpe 1490 Pahawh_Hmong,
797     Palmyrene,
798     Pau_Cin_Hau,
799 nigel 93 Phags_Pa,
800     Phoenician,
801 chpe 1490 Psalter_Pahlavi,
802 ph10 491 Rejang,
803 nigel 87 Runic,
804 ph10 491 Samaritan,
805     Saurashtra,
806 ph10 942 Sharada,
807 nigel 87 Shavian,
808 chpe 1490 Siddham,
809 nigel 87 Sinhala,
810 ph10 942 Sora_Sompeng,
811 ph10 491 Sundanese,
812 nigel 87 Syloti_Nagri,
813     Syriac,
814     Tagalog,
815     Tagbanwa,
816     Tai_Le,
817 ph10 491 Tai_Tham,
818     Tai_Viet,
819 ph10 942 Takri,
820 nigel 87 Tamil,
821     Telugu,
822     Thaana,
823     Thai,
824     Tibetan,
825     Tifinagh,
826 chpe 1490 Tirhuta,
827 nigel 87 Ugaritic,
828 ph10 491 Vai,
829 chpe 1490 Warang_Citi,
830 nigel 87 Yi.
831     .P
832 ph10 517 Each character has exactly one Unicode general category property, specified by
833     a two-letter abbreviation. For compatibility with Perl, negation can be
834     specified by including a circumflex between the opening brace and the property
835     name. For example, \ep{^Lu} is the same as \eP{Lu}.
836 nigel 87 .P
837     If only one letter is specified with \ep or \eP, it includes all the general
838     category properties that start with that letter. In this case, in the absence
839     of negation, the curly brackets in the escape sequence are optional; these two
840     examples have the same effect:
841     .sp
842 nigel 75 \ep{L}
843     \epL
844     .sp
845 nigel 87 The following general category property codes are supported:
846 nigel 75 .sp
847     C Other
848     Cc Control
849     Cf Format
850     Cn Unassigned
851     Co Private use
852     Cs Surrogate
853     .sp
854     L Letter
855     Ll Lower case letter
856     Lm Modifier letter
857     Lo Other letter
858     Lt Title case letter
859     Lu Upper case letter
860     .sp
861     M Mark
862     Mc Spacing mark
863     Me Enclosing mark
864     Mn Non-spacing mark
865     .sp
866     N Number
867     Nd Decimal number
868     Nl Letter number
869     No Other number
870     .sp
871     P Punctuation
872     Pc Connector punctuation
873     Pd Dash punctuation
874     Pe Close punctuation
875     Pf Final punctuation
876     Pi Initial punctuation
877     Po Other punctuation
878     Ps Open punctuation
879     .sp
880     S Symbol
881     Sc Currency symbol
882     Sk Modifier symbol
883     Sm Mathematical symbol
884     So Other symbol
885     .sp
886     Z Separator
887     Zl Line separator
888     Zp Paragraph separator
889     Zs Space separator
890     .sp
891 nigel 87 The special property L& is also supported: it matches a character that has
892     the Lu, Ll, or Lt property, in other words, a letter that is not classified as
893     a modifier or "other".
894 nigel 75 .P
895 ph10 211 The Cs (Surrogate) property applies only to characters in the range U+D800 to
896 ph10 859 U+DFFF. Such characters are not valid in Unicode strings and so
897     cannot be tested by PCRE, unless UTF validity checking has been turned off
898 chpe 1055 (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
899     PCRE_NO_UTF32_CHECK in the
900 ph10 211 .\" HREF
901     \fBpcreapi\fP
902     .\"
903 ph10 451 page). Perl does not support the Cs property.
904 ph10 211 .P
905 ph10 451 The long synonyms for property names that Perl supports (such as \ep{Letter})
906 nigel 91 are not supported by PCRE, nor is it permitted to prefix any of these
907 nigel 87 properties with "Is".
908     .P
909     No character that is in the Unicode table has the Cn (unassigned) property.
910     Instead, this property is assumed for any code point that is not in the
911     Unicode table.
912     .P
913 nigel 75 Specifying caseless matching does not affect these escape sequences. For
914 ph10 1335 example, \ep{Lu} always matches only upper case letters. This is different from
915 ph10 1258 the behaviour of current versions of Perl.
916 nigel 75 .P
917 ph10 1011 Matching characters by Unicode property is not fast, because PCRE has to do a
918     multistage table lookup in order to find a character's property. That is why
919     the traditional escape sequences such as \ed and \ew do not use Unicode
920     properties in PCRE by default, though you can make them do so by setting the
921     PCRE_UCP option or by starting the pattern with (*UCP).
922     .
923     .
924     .SS Extended grapheme clusters
925     .rs
926 nigel 75 .sp
927 ph10 1011 The \eX escape matches any number of Unicode characters that form an "extended
928     grapheme cluster", and treats the sequence as an atomic group
929 nigel 75 .\" HTML <a href="#atomicgroup">
930     .\" </a>
931     (see below).
932     .\"
933 ph10 1011 Up to and including release 8.31, PCRE matched an earlier, simpler definition
934     that was equivalent to
935     .sp
936     (?>\ePM\epM*)
937     .sp
938     That is, it matched a character without the "mark" property, followed by zero
939     or more characters with the "mark" property. Characters with the "mark"
940     property are typically non-spacing accents that affect the preceding character.
941 nigel 75 .P
942 ph10 1011 This simple definition was extended in Unicode to include more complicated
943     kinds of composite character by giving each character a grapheme breaking
944     property, and creating rules that use these properties to define the boundaries
945     of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
946     one of these clusters.
947 ph10 628 .P
948 ph10 1221 \eX always matches at least one character. Then it decides whether to add
949 ph10 1011 additional characters according to the following rules for ending a cluster:
950     .P
951     1. End at the end of the subject string.
952     .P
953     2. Do not end between CR and LF; otherwise end after any control character.
954     .P
955 ph10 1221 3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
956 ph10 1011 are of five types: L, V, T, LV, and LVT. An L character may be followed by an
957 ph10 1221 L, V, LV, or LVT character; an LV or V character may be followed by a V or T
958 ph10 1011 character; an LVT or T character may be follwed only by a T character.
959     .P
960     4. Do not end before extending characters or spacing marks. Characters with
961     the "mark" property always have the "extend" grapheme breaking property.
962     .P
963     5. Do not end after prepend characters.
964     .P
965     6. Otherwise, end the cluster.
966 nigel 75 .
967     .
968 ph10 517 .\" HTML <a name="extraprops"></a>
969     .SS PCRE's additional properties
970     .rs
971     .sp
972 ph10 1011 As well as the standard Unicode properties described above, PCRE supports four
973     more that make it possible to convert traditional escape sequences such as \ew
974 ph10 1387 and \es to use Unicode properties. PCRE uses these non-standard, non-Perl
975     properties internally when PCRE_UCP is set. However, they may also be used
976     explicitly. These properties are:
977 ph10 517 .sp
978     Xan Any alphanumeric character
979     Xps Any POSIX space character
980     Xsp Any Perl space character
981     Xwd Any Perl "word" character
982     .sp
983 ph10 535 Xan matches characters that have either the L (letter) or the N (number)
984 ph10 968 property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
985 ph10 517 carriage return, and any other character that has the Z (separator) property.
986 ph10 1387 Xsp is the same as Xps; it used to exclude vertical tab, for Perl
987     compatibility, but Perl changed, and so PCRE followed at release 8.34. Xwd
988     matches the same characters as Xan, plus underscore.
989 ph10 1260 .P
990 ph10 1335 There is another non-standard property, Xuc, which matches any character that
991     can be represented by a Universal Character Name in C++ and other programming
992     languages. These are the characters $, @, ` (grave accent), and all characters
993     with Unicode code points greater than or equal to U+00A0, except for the
994     surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are
995 ph10 1260 excluded. (Universal Character Names are of the form \euHHHH or \eUHHHHHHHH
996 ph10 1335 where H is a hexadecimal digit. Note that the Xuc property does not match these
997 ph10 1260 sequences but the characters that they represent.)
998 ph10 517 .
999     .
1000 ph10 168 .\" HTML <a name="resetmatchstart"></a>
1001     .SS "Resetting the match start"
1002     .rs
1003     .sp
1004 ph10 572 The escape sequence \eK causes any previously matched characters not to be
1005     included in the final matched sequence. For example, the pattern:
1006 ph10 168 .sp
1007     foo\eKbar
1008     .sp
1009 ph10 172 matches "foobar", but reports that it has matched "bar". This feature is
1010 ph10 168 similar to a lookbehind assertion
1011     .\" HTML <a href="#lookbehind">
1012     .\" </a>
1013     (described below).
1014     .\"
1015 ph10 172 However, in this case, the part of the subject before the real match does not
1016     have to be of fixed length, as lookbehind assertions do. The use of \eK does
1017 ph10 168 not interfere with the setting of
1018     .\" HTML <a href="#subpattern">
1019     .\" </a>
1020     captured substrings.
1021 ph10 172 .\"
1022 ph10 168 For example, when the pattern
1023     .sp
1024     (foo)\eKbar
1025     .sp
1026 ph10 172 matches "foobar", the first substring is still set to "foo".
1027 ph10 500 .P
1028 ph10 507 Perl documents that the use of \eK within assertions is "not well defined". In
1029     PCRE, \eK is acted upon when it occurs inside positive assertions, but is
1030 ph10 1418 ignored in negative assertions. Note that when a pattern such as (?=ab\eK)
1031 ph10 1459 matches, the reported start of the match can be greater than the end of the
1032 ph10 1418 match.
1033 ph10 168 .
1034     .
1035 nigel 75 .\" HTML <a name="smallassertions"></a>
1036     .SS "Simple assertions"
1037     .rs
1038     .sp
1039 nigel 93 The final use of backslash is for certain simple assertions. An assertion
1040 nigel 63 specifies a condition that has to be met at a particular point in a match,
1041     without consuming any characters from the subject string. The use of
1042 nigel 75 subpatterns for more complicated assertions is described
1043     .\" HTML <a href="#bigassertions">
1044     .\" </a>
1045     below.
1046     .\"
1047 nigel 91 The backslashed assertions are:
1048 nigel 75 .sp
1049     \eb matches at a word boundary
1050     \eB matches when not at a word boundary
1051 nigel 93 \eA matches at the start of the subject
1052     \eZ matches at the end of the subject
1053     also matches before a newline at the end of the subject
1054     \ez matches only at the end of the subject
1055     \eG matches at the first matching position in the subject
1056 nigel 75 .sp
1057 ph10 513 Inside a character class, \eb has a different meaning; it matches the backspace
1058 ph10 535 character. If any other of these assertions appears in a character class, by
1059 ph10 513 default it matches the corresponding literal character (for example, \eB
1060     matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
1061     escape sequence" error is generated instead.
1062 nigel 75 .P
1063 nigel 63 A word boundary is a position in the subject string where the current character
1064 nigel 75 and the previous character do not both match \ew or \eW (i.e. one matches
1065     \ew and the other matches \eW), or the start or end of the string if the
1066 ph10 859 first or last character matches \ew, respectively. In a UTF mode, the meanings
1067 ph10 518 of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
1068     done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
1069     of word" or "end of word" metasequence. However, whatever follows \eb normally
1070     determines which it is. For example, the fragment \eba matches "a" at the start
1071     of a word.
1072 nigel 75 .P
1073     The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
1074     dollar (described in the next section) in that they only ever match at the very
1075     start and end of the subject string, whatever options are set. Thus, they are
1076     independent of multiline mode. These three assertions are not affected by the
1077     PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
1078     circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
1079     argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
1080     at a point other than the beginning of the subject, \eA can never match. The
1081 nigel 91 difference between \eZ and \ez is that \eZ matches before a newline at the end
1082     of the string as well as at the very end, whereas \ez matches only at the end.
1083 nigel 75 .P
1084     The \eG assertion is true only when the current matching position is at the
1085     start point of the match, as specified by the \fIstartoffset\fP argument of
1086     \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
1087     non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
1088 nigel 63 arguments, you can mimic Perl's /g option, and it is in this kind of
1089 nigel 75 implementation where \eG can be useful.
1090     .P
1091     Note, however, that PCRE's interpretation of \eG, as the start of the current
1092 nigel 63 match, is subtly different from Perl's, which defines it as the end of the
1093     previous match. In Perl, these can be different when the previously matched
1094     string was empty. Because PCRE does just one match at a time, it cannot
1095     reproduce this behaviour.
1096 nigel 75 .P
1097     If all the alternatives of a pattern begin with \eG, the expression is anchored
1098 nigel 63 to the starting match position, and the "anchored" flag is set in the compiled
1099     regular expression.
1100 nigel 75 .
1101     .
1103 nigel 63 .rs
1104     .sp
1105 ph10 1221 The circumflex and dollar metacharacters are zero-width assertions. That is,
1106     they test for a particular condition being true without consuming any
1107 ph10 1213 characters from the subject string.
1108     .P
1109 nigel 63 Outside a character class, in the default matching mode, the circumflex
1110 ph10 1213 character is an assertion that is true only if the current matching point is at
1111     the start of the subject string. If the \fIstartoffset\fP argument of
1112 nigel 75 \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
1113 nigel 63 option is unset. Inside a character class, circumflex has an entirely different
1114 nigel 75 meaning
1115     .\" HTML <a href="#characterclass">
1116     .\" </a>
1117     (see below).
1118     .\"
1119     .P
1120 nigel 63 Circumflex need not be the first character of the pattern if a number of
1121     alternatives are involved, but it should be the first thing in each alternative
1122     in which it appears if the pattern is ever to match that branch. If all
1123     possible alternatives start with a circumflex, that is, if the pattern is
1124     constrained to match only at the start of the subject, it is said to be an
1125     "anchored" pattern. (There are also other constructs that can cause a pattern
1126     to be anchored.)
1127 nigel 75 .P
1128 ph10 1213 The dollar character is an assertion that is true only if the current matching
1129     point is at the end of the subject string, or immediately before a newline at
1130     the end of the string (by default). Note, however, that it does not actually
1131     match the newline. Dollar need not be the last character of the pattern if a
1132     number of alternatives are involved, but it should be the last item in any
1133     branch in which it appears. Dollar has no special meaning in a character class.
1134 nigel 75 .P
1135 nigel 63 The meaning of dollar can be changed so that it matches only at the very end of
1136     the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1137 nigel 75 does not affect the \eZ assertion.
1138     .P
1139 nigel 63 The meanings of the circumflex and dollar characters are changed if the
1140 nigel 91 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1141     immediately after internal newlines as well as at the start of the subject
1142     string. It does not match after a newline that ends the string. A dollar
1143     matches before any newlines in the string, as well as at the very end, when
1144     PCRE_MULTILINE is set. When newline is specified as the two-character
1145     sequence CRLF, isolated CR and LF characters do not indicate newlines.
1146 nigel 75 .P
1147 nigel 91 For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1148     \en represents a newline) in multiline mode, but not otherwise. Consequently,
1149     patterns that are anchored in single line mode because all branches start with
1150     ^ are not anchored in multiline mode, and a match for circumflex is possible
1151     when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1152     PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1153     .P
1154 nigel 75 Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1155 nigel 63 end of the subject in both modes, and if all branches of a pattern start with
1156 nigel 91 \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1157 nigel 75 .
1158     .
1159 ph10 514 .\" HTML <a name="fullstopdot"></a>
1160     .SH "FULL STOP (PERIOD, DOT) AND \eN"
1161 nigel 63 .rs
1162     .sp
1163     Outside a character class, a dot in the pattern matches any one character in
1164 nigel 91 the subject string except (by default) a character that signifies the end of a
1165 ph10 903 line.
1166 nigel 91 .P
1167 nigel 93 When a line ending is defined as a single character, dot never matches that
1168     character; when the two-character sequence CRLF is used, dot does not match CR
1169     if it is immediately followed by LF, but otherwise it matches all characters
1170     (including isolated CRs and LFs). When any Unicode line endings are being
1171     recognized, dot does not match CR or LF or any of the other line ending
1172     characters.
1173     .P
1174 nigel 91 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1175 nigel 93 option is set, a dot matches any one character, without exception. If the
1176     two-character sequence CRLF is present in the subject string, it takes two dots
1177     to match it.
1178 nigel 91 .P
1179     The handling of dot is entirely independent of the handling of circumflex and
1180     dollar, the only relationship being that they both involve newlines. Dot has no
1181     special meaning in a character class.
1182 ph10 514 .P
1183 ph10 579 The escape sequence \eN behaves like a dot, except that it is not affected by
1184 ph10 572 the PCRE_DOTALL option. In other words, it matches any character except one
1185 ph10 745 that signifies the end of a line. Perl also uses \eN to match characters by
1186     name; PCRE does not support this.
1187 nigel 75 .
1188     .
1190 nigel 63 .rs
1191     .sp
1192 ph10 859 Outside a character class, the escape sequence \eC matches any one data unit,
1193     whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1194 chpe 1055 byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1195     a 32-bit unit. Unlike a dot, \eC always
1196 ph10 859 matches line-ending characters. The feature is provided in Perl in order to
1197     match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1198     used. Because \eC breaks up characters into individual data units, matching one
1199     unit with \eC in a UTF mode means that the rest of the string may start with a
1200     malformed UTF character. This has undefined results, because PCRE assumes that
1201     it is dealing with valid UTF strings (and by default it checks this at the
1202 chpe 1055 start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1203     PCRE_NO_UTF32_CHECK option is used).
1204 nigel 75 .P
1205     PCRE does not allow \eC to appear in lookbehind assertions
1206     .\" HTML <a href="#lookbehind">
1207     .\" </a>
1208 ph10 754 (described below)
1209 nigel 75 .\"
1210 ph10 859 in a UTF mode, because this would make it impossible to calculate the length of
1211 nigel 75 the lookbehind.
1212 ph10 737 .P
1213 ph10 859 In general, the \eC escape sequence is best avoided. However, one
1214     way of using it that avoids the problem of malformed UTF characters is to use a
1215     lookahead to check the length of the next character, as in this pattern, which
1216     could be used with a UTF-8 string (ignore white space and line breaks):
1217 ph10 737 .sp
1218     (?| (?=[\ex00-\ex7f])(\eC) |
1219 ph10 738 (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1220     (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1221 ph10 737 (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1222     .sp
1223 ph10 738 A group that starts with (?| resets the capturing parentheses numbers in each
1224     alternative (see
1225 ph10 737 .\" HTML <a href="#dupsubpatternnumber">
1226     .\" </a>
1227     "Duplicate Subpattern Numbers"
1228     .\"
1229 ph10 738 below). The assertions at the start of each branch check the next UTF-8
1230     character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1231 ph10 737 character's individual bytes are then captured by the appropriate number of
1232     groups.
1233 nigel 75 .
1234     .
1235     .\" HTML <a name="characterclass"></a>
1237 nigel 63 .rs
1238     .sp
1239     An opening square bracket introduces a character class, terminated by a closing
1240 ph10 461 square bracket. A closing square bracket on its own is not special by default.
1241     However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1242 ph10 456 bracket causes a compile-time error. If a closing square bracket is required as
1243     a member of the class, it should be the first data character in the class
1244     (after an initial circumflex, if present) or escaped with a backslash.
1245 nigel 75 .P
1246 ph10 859 A character class matches a single character in the subject. In a UTF mode, the
1247     character may be more than one data unit long. A matched character must be in
1248     the set of characters defined by the class, unless the first character in the
1249     class definition is a circumflex, in which case the subject character must not
1250     be in the set defined by the class. If a circumflex is actually required as a
1251     member of the class, ensure it is not the first character, or escape it with a
1252 nigel 63 backslash.
1253 nigel 75 .P
1254 nigel 63 For example, the character class [aeiou] matches any lower case vowel, while
1255     [^aeiou] matches any character that is not a lower case vowel. Note that a
1256 nigel 75 circumflex is just a convenient notation for specifying the characters that
1257     are in the class by enumerating those that are not. A class that starts with a
1258 ph10 456 circumflex is not an assertion; it still consumes a character from the subject
1259 nigel 75 string, and therefore it fails if the current pointer is at the end of the
1260     string.
1261     .P
1262 chpe 1055 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1263     can be included in a class as a literal string of data units, or by using the
1264     \ex{ escaping mechanism.
1265 nigel 75 .P
1266 nigel 63 When caseless matching is set, any letters in a class represent both their
1267     upper case and lower case versions, so for example, a caseless [aeiou] matches
1268     "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1269 ph10 859 caseful version would. In a UTF mode, PCRE always understands the concept of
1270 nigel 77 case for characters whose values are less than 128, so caseless matching is
1271     always possible. For characters with higher values, the concept of case is
1272     supported if PCRE is compiled with Unicode property support, but not otherwise.
1273 ph10 859 If you want to use caseless matching in a UTF mode for characters 128 and
1274     above, you must ensure that PCRE is compiled with Unicode property support as
1275     well as with UTF support.
1276 nigel 75 .P
1277 nigel 93 Characters that might indicate line breaks are never treated in any special way
1278     when matching character classes, whatever line-ending sequence is in use, and
1279     whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1280     such as [^a] always matches one of these characters.
1281 nigel 75 .P
1282 nigel 63 The minus (hyphen) character can be used to specify a range of characters in a
1283     character class. For example, [d-m] matches any letter between d and m,
1284     inclusive. If a minus character is required in a class, it must be escaped with
1285     a backslash or appear in a position where it cannot be interpreted as
1286 ph10 1404 indicating a range, typically as the first or last character in the class, or
1287 ph10 1392 immediately after a range. For example, [b-d-z] matches letters in the range b
1288     to d, a hyphen character, or z.
1289 nigel 75 .P
1290 nigel 63 It is not possible to have the literal character "]" as the end character of a
1291     range. A pattern such as [W-]46] is interpreted as a class of two characters
1292     ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1293     "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1294 nigel 75 the end of range, so [W-\e]46] is interpreted as a class containing a range
1295     followed by two other characters. The octal or hexadecimal representation of
1296     "]" can also be used to end a range.
1297     .P
1298 ph10 1392 An error is generated if a POSIX character class (see below) or an escape
1299     sequence other than one that defines a single character appears at a point
1300     where a range ending character is expected. For example, [z-\exff] is valid,
1301     but [A-\ed] and [A-[:digit:]] are not.
1302     .P
1303 nigel 63 Ranges operate in the collating sequence of character values. They can also be
1304 ph10 903 used for characters specified numerically, for example [\e000-\e037]. Ranges
1305 ph10 859 can include any characters that are valid for the current mode.
1306 nigel 75 .P
1307 nigel 63 If a range that includes letters is used when caseless matching is set, it
1308     matches the letters in either case. For example, [W-c] is equivalent to
1309 ph10 859 [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1310 ph10 139 tables for a French locale are in use, [\exc8-\excb] matches accented E
1311 ph10 859 characters in both cases. In UTF modes, PCRE supports the concept of case for
1312 nigel 75 characters with values greater than 128 only when it is compiled with Unicode
1313     property support.
1314     .P
1315 ph10 575 The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1316     \eV, \ew, and \eW may appear in a character class, and add the characters that
1317     they match to the class. For example, [\edABCDEF] matches any hexadecimal
1318 ph10 859 digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1319 ph10 575 and their upper case partners, just as it does when they appear outside a
1320     character class, as described in the section entitled
1321     .\" HTML <a href="#genericchartypes">
1322     .\" </a>
1323     "Generic character types"
1324     .\"
1325     above. The escape sequence \eb has a different meaning inside a character
1326     class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1327     are not special inside a character class. Like any other unrecognized escape
1328     sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1329     default, but cause an error if the PCRE_EXTRA option is set.
1330     .P
1331     A circumflex can conveniently be used with the upper case character types to
1332 ph10 518 specify a more restricted set of characters than the matching lower case type.
1333 ph10 575 For example, the class [^\eW_] matches any letter or digit, but not underscore,
1334 ph10 579 whereas [\ew] includes underscore. A positive character class should be read as
1335     "something OR something OR ..." and a negative class as "NOT something AND NOT
1336 ph10 575 something AND NOT ...".
1337 nigel 75 .P
1338     The only metacharacters that are recognized in character classes are backslash,
1339     hyphen (only where it can be interpreted as specifying a range), circumflex
1340     (only at the start), opening square bracket (only when it can be interpreted as
1341 ph10 1408 introducing a POSIX class name, or for a special compatibility feature - see
1342     the next two sections), and the terminating closing square bracket. However,
1343     escaping other non-alphanumeric characters does no harm.
1344 nigel 75 .
1345     .
1347 nigel 63 .rs
1348     .sp
1349 nigel 75 Perl supports the POSIX notation for character classes. This uses names
1350 nigel 63 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1351     this notation. For example,
1352 nigel 75 .sp
1353 nigel 63 [01[:alpha:]%]
1354 nigel 75 .sp
1355 nigel 63 matches "0", "1", any alphabetic character, or "%". The supported class names
1356 ph10 518 are:
1357 nigel 75 .sp
1358 nigel 63 alnum letters and digits
1359     alpha letters
1360     ascii character codes 0 - 127
1361     blank space or tab only
1362     cntrl control characters
1363 nigel 75 digit decimal digits (same as \ed)
1364 nigel 63 graph printing characters, excluding space
1365     lower lower case letters
1366     print printing characters, including space
1367 ph10 518 punct printing characters, excluding letters and digits and space
1368 ph10 1376 space white space (the same as \es from PCRE 8.34)
1369 nigel 63 upper upper case letters
1370 nigel 75 word "word" characters (same as \ew)
1371 nigel 63 xdigit hexadecimal digits
1372 nigel 75 .sp
1373 ph10 1401 The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
1374 ph10 1405 and space (32). If locale-specific matching is taking place, the list of space
1375     characters may be different; there may be fewer or more of them. "Space" used
1376     to be different to \es, which did not include VT, for Perl compatibility.
1377     However, Perl changed at release 5.18, and PCRE followed at release 8.34.
1378     "Space" and \es now match the same set of characters.
1379 nigel 75 .P
1380 nigel 63 The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1381     5.8. Another Perl extension is negation, which is indicated by a ^ character
1382     after the colon. For example,
1383 nigel 75 .sp
1384 nigel 63 [12[:^digit:]]
1385 nigel 75 .sp
1386 nigel 63 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1387     syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1388     supported, and an error is given if they are encountered.
1389 nigel 75 .P
1390 ph10 1401 By default, characters with values greater than 128 do not match any of the
1391     POSIX character classes. However, if the PCRE_UCP option is passed to
1392     \fBpcre_compile()\fP, some of the classes are changed so that Unicode character
1393     properties are used. This is achieved by replacing certain POSIX classes by
1394     other sequences, as follows:
1395 ph10 518 .sp
1396     [:alnum:] becomes \ep{Xan}
1397     [:alpha:] becomes \ep{L}
1398 ph10 535 [:blank:] becomes \eh
1399 ph10 518 [:digit:] becomes \ep{Nd}
1400     [:lower:] becomes \ep{Ll}
1401 ph10 535 [:space:] becomes \ep{Xps}
1402 ph10 518 [:upper:] becomes \ep{Lu}
1403     [:word:] becomes \ep{Xwd}
1404     .sp
1405 ph10 1404 Negated versions, such as [:^alpha:] use \eP instead of \ep. Three other POSIX
1406 ph10 1387 classes are handled specially in UCP mode:
1407     .TP 10
1408     [:graph:]
1409 ph10 1404 This matches characters that have glyphs that mark the page when printed. In
1410     Unicode property terms, it matches all characters with the L, M, N, P, S, or Cf
1411 ph10 1387 properties, except for:
1412     .sp
1413     U+061C Arabic Letter Mark
1414 ph10 1404 U+180E Mongolian Vowel Separator
1415 ph10 1387 U+2066 - U+2069 Various "isolate"s
1416     .sp
1417     .TP 10
1418     [:print:]
1419 ph10 1404 This matches the same characters as [:graph:] plus space characters that are
1420 ph10 1387 not controls, that is, characters with the Zs property.
1421     .TP 10
1422     [:punct:]
1423     This matches all characters that have the Unicode P (punctuation) property,
1424     plus those characters whose code points are less than 128 that have the S
1425     (Symbol) property.
1426     .P
1427     The other POSIX classes are unchanged, and match only characters with code
1428     points less than 128.
1429 nigel 75 .
1430     .
1432     .rs
1433     .sp
1434 ph10 1412 In the POSIX.2 compliant library that was included in 4.4BSD Unix, the ugly
1435     syntax [[:<:]] and [[:>:]] is used for matching "start of word" and "end of
1436 ph10 1408 word". PCRE treats these items as follows:
1437     .sp
1438     [[:<:]] is converted to \eb(?=\ew)
1439     [[:>:]] is converted to \eb(?<=\ew)
1440     .sp
1441     Only these exact character sequences are recognized. A sequence such as
1442 ph10 1412 [a[:<:]b] provokes error for an unrecognized POSIX class name. This support is
1443     not compatible with Perl. It is provided to help migrations from other
1444     environments, and is best not used in any new patterns. Note that \eb matches
1445     at the start and the end of a word (see
1446 ph10 1408 .\" HTML <a href="#smallassertions">
1447     .\" </a>
1448     "Simple assertions"
1449     .\"
1450 ph10 1412 above), and in a Perl-style pattern the preceding or following character
1451     normally shows which is wanted, without the need for the assertions that are
1452 ph10 1408 used above in order to give exactly the POSIX behaviour.
1453     .
1454     .
1455 nigel 75 .SH "VERTICAL BAR"
1456 nigel 63 .rs
1457     .sp
1458     Vertical bar characters are used to separate alternative patterns. For example,
1459     the pattern
1460 nigel 75 .sp
1461 nigel 63 gilbert|sullivan
1462 nigel 75 .sp
1463 nigel 63 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1464 nigel 91 and an empty alternative is permitted (matching the empty string). The matching
1465     process tries each alternative in turn, from left to right, and the first one
1466     that succeeds is used. If the alternatives are within a subpattern
1467 nigel 75 .\" HTML <a href="#subpattern">
1468     .\" </a>
1469     (defined below),
1470     .\"
1471     "succeeds" means matching the rest of the main pattern as well as the
1472     alternative in the subpattern.
1473     .
1474     .
1476 nigel 63 .rs
1477     .sp
1478     The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1479 ph10 231 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1480     the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1481     The option letters are
1482 nigel 75 .sp
1483 nigel 63 i for PCRE_CASELESS
1484     m for PCRE_MULTILINE
1485     s for PCRE_DOTALL
1486     x for PCRE_EXTENDED
1487 nigel 75 .sp
1488 nigel 63 For example, (?im) sets caseless, multiline matching. It is also possible to
1489     unset these options by preceding the letter with a hyphen, and a combined
1490     setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1491     PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1492     permitted. If a letter appears both before and after the hyphen, the option is
1493     unset.
1494 nigel 75 .P
1495 ph10 231 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1496     changed in the same way as the Perl-compatible options by using the characters
1497     J, U and X respectively.
1498     .P
1499 ph10 412 When one of these option changes occurs at top level (that is, not inside
1500     subpattern parentheses), the change applies to the remainder of the pattern
1501     that follows. If the change is placed right at the start of a pattern, PCRE
1502     extracts it into the global options (and it will therefore show up in data
1503     extracted by the \fBpcre_fullinfo()\fP function).
1504 nigel 75 .P
1505 nigel 93 An option change within a subpattern (see below for a description of
1506 ph10 572 subpatterns) affects only that part of the subpattern that follows it, so
1507 nigel 75 .sp
1508 nigel 63 (a(?i)b)c
1509 nigel 75 .sp
1510 nigel 63 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1511     By this means, options can be made to have different settings in different
1512     parts of the pattern. Any changes made in one alternative do carry on
1513     into subsequent branches within the same subpattern. For example,
1514 nigel 75 .sp
1515 nigel 63 (a(?i)b|c)
1516 nigel 75 .sp
1517 nigel 63 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1518     branch is abandoned before the option setting. This is because the effects of
1519     option settings happen at compile time. There would be some very weird
1520     behaviour otherwise.
1521 ph10 251 .P
1522     \fBNote:\fP There are other PCRE-specific options that can be set by the
1523 ph10 859 application when the compiling or matching functions are called. In some cases
1524     the pattern can contain special leading sequences such as (*CRLF) to override
1525     what the application has set or what has been defaulted. Details are given in
1526     the section entitled
1527 ph10 251 .\" HTML <a href="#newlineseq">
1528     .\" </a>
1529     "Newline sequences"
1530     .\"
1531 ph10 1219 above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1532 chpe 1055 sequences that can be used to set UTF and Unicode property modes; they are
1533     equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1534 ph10 1219 options, respectively. The (*UTF) sequence is a generic version that can be
1535 ph10 1335 used with any of the libraries. However, the application can set the
1536 ph10 1309 PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences.
1537 nigel 75 .
1538     .
1539     .\" HTML <a name="subpattern"></a>
1540 nigel 63 .SH SUBPATTERNS
1541     .rs
1542     .sp
1543     Subpatterns are delimited by parentheses (round brackets), which can be nested.
1544 nigel 75 Turning part of a pattern into a subpattern does two things:
1545     .sp
1546 nigel 63 1. It localizes a set of alternatives. For example, the pattern
1547 nigel 75 .sp
1548 nigel 63 cat(aract|erpillar|)
1549 nigel 75 .sp
1550 ph10 572 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1551     match "cataract", "erpillar" or an empty string.
1552 nigel 75 .sp
1553     2. It sets up the subpattern as a capturing subpattern. This means that, when
1554     the whole pattern matches, that portion of the subject string that matched the
1555 ph10 903 subpattern is passed back to the caller via the \fIovector\fP argument of the
1556     matching function. (This applies only to the traditional matching functions;
1557 ph10 859 the DFA matching functions do not support capturing.)
1558     .P
1559     Opening parentheses are counted from left to right (starting from 1) to obtain
1560     numbers for the capturing subpatterns. For example, if the string "the red
1561     king" is matched against the pattern
1562 nigel 75 .sp
1563 nigel 63 the ((red|white) (king|queen))
1564 nigel 75 .sp
1565 nigel 63 the captured substrings are "red king", "red", and "king", and are numbered 1,
1566     2, and 3, respectively.
1567 nigel 75 .P
1568 nigel 63 The fact that plain parentheses fulfil two functions is not always helpful.
1569     There are often times when a grouping subpattern is required without a
1570     capturing requirement. If an opening parenthesis is followed by a question mark
1571     and a colon, the subpattern does not do any capturing, and is not counted when
1572     computing the number of any subsequent capturing subpatterns. For example, if
1573     the string "the white queen" is matched against the pattern
1574 nigel 75 .sp
1575 nigel 63 the ((?:red|white) (king|queen))
1576 nigel 75 .sp
1577 nigel 63 the captured substrings are "white queen" and "queen", and are numbered 1 and
1578 nigel 93 2. The maximum number of capturing subpatterns is 65535.
1579 nigel 75 .P
1580 nigel 63 As a convenient shorthand, if any option settings are required at the start of
1581     a non-capturing subpattern, the option letters may appear between the "?" and
1582     the ":". Thus the two patterns
1583 nigel 75 .sp
1584 nigel 63 (?i:saturday|sunday)
1585     (?:(?i)saturday|sunday)
1586 nigel 75 .sp
1587 nigel 63 match exactly the same set of strings. Because alternative branches are tried
1588     from left to right, and options are not reset until the end of the subpattern
1589     is reached, an option setting in one branch does affect subsequent branches, so
1590     the above patterns match "SUNDAY" as well as "Saturday".
1591 nigel 75 .
1592     .
1593 ph10 456 .\" HTML <a name="dupsubpatternnumber"></a>
1595     .rs
1596     .sp
1597 ph10 182 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1598     the same numbers for its capturing parentheses. Such a subpattern starts with
1599     (?| and is itself a non-capturing subpattern. For example, consider this
1600 ph10 175 pattern:
1601     .sp
1602     (?|(Sat)ur|(Sun))day
1603 ph10 182 .sp
1604     Because the two alternatives are inside a (?| group, both sets of capturing
1605     parentheses are numbered one. Thus, when the pattern matches, you can look
1606     at captured substring number one, whichever alternative matched. This construct
1607     is useful when you want to capture part, but not all, of one of a number of
1608     alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1609 ph10 175 number is reset at the start of each branch. The numbers of any capturing
1610 ph10 572 parentheses that follow the subpattern start after the highest number used in
1611     any branch. The following example is taken from the Perl documentation. The
1612     numbers underneath show in which buffer the captured content will be stored.
1613 ph10 175 .sp
1614     # before ---------------branch-reset----------- after
1615     / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1616     # 1 2 2 3 2 3 4
1617 ph10 182 .sp
1618 ph10 488 A back reference to a numbered subpattern uses the most recent value that is
1619     set for that number by any subpattern. The following pattern matches "abcabc"
1620     or "defdef":
1621 ph10 456 .sp
1622 ph10 461 /(?|(abc)|(def))\e1/
1623 ph10 456 .sp
1624 ph10 716 In contrast, a subroutine call to a numbered subpattern always refers to the
1625     first one in the pattern with the given number. The following pattern matches
1626     "abcabc" or "defabc":
1627 ph10 456 .sp
1628     /(?|(abc)|(def))(?1)/
1629     .sp
1630 ph10 459 If a
1631     .\" HTML <a href="#conditions">
1632     .\" </a>
1633     condition test
1634     .\"
1635     for a subpattern's having matched refers to a non-unique number, the test is
1636     true if any of the subpatterns of that number have matched.
1637     .P
1638     An alternative approach to using this "branch reset" feature is to use
1639 ph10 175 duplicate named subpatterns, as described in the next section.
1640     .
1641     .
1642 nigel 75 .SH "NAMED SUBPATTERNS"
1643 nigel 63 .rs
1644     .sp
1645     Identifying capturing parentheses by number is simple, but it can be very hard
1646     to keep track of the numbers in complicated regular expressions. Furthermore,
1647 nigel 75 if an expression is modified, the numbers may change. To help with this
1648 nigel 93 difficulty, PCRE supports the naming of subpatterns. This feature was not
1649     added to Perl until release 5.10. Python had the feature earlier, and PCRE
1650     introduced it at release 4.0, using the Python syntax. PCRE now supports both
1651 ph10 459 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1652     have different names, but PCRE does not.
1653 nigel 93 .P
1654     In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1655     (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1656 nigel 91 parentheses from other parts of the pattern, such as
1657     .\" HTML <a href="#backreferences">
1658     .\" </a>
1659 ph10 488 back references,
1660 nigel 91 .\"
1661     .\" HTML <a href="#recursion">
1662     .\" </a>
1663     recursion,
1664     .\"
1665     and
1666     .\" HTML <a href="#conditions">
1667     .\" </a>
1668     conditions,
1669     .\"
1670     can be made by name as well as by number.
1671 nigel 75 .P
1672 ph10 1404 Names consist of up to 32 alphanumeric characters and underscores, but must
1673 ph10 1394 start with a non-digit. Named capturing parentheses are still allocated numbers
1674     as well as names, exactly as if the names were not present. The PCRE API
1675     provides function calls for extracting the name-to-number translation table
1676     from a compiled pattern. There is also a convenience function for extracting a
1677     captured substring by name.
1678 nigel 91 .P
1679     By default, a name must be unique within a pattern, but it is possible to relax
1680 ph10 457 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1681 ph10 461 names are also always permitted for subpatterns with the same number, set up as
1682 ph10 457 described in the previous section.) Duplicate names can be useful for patterns
1683     where only one instance of the named parentheses can match. Suppose you want to
1684     match the name of a weekday, either as a 3-letter abbreviation or as the full
1685     name, and in both cases you want to extract the abbreviation. This pattern
1686     (ignoring the line breaks) does the job:
1687 nigel 91 .sp
1688 nigel 93 (?<DN>Mon|Fri|Sun)(?:day)?|
1689     (?<DN>Tue)(?:sday)?|
1690     (?<DN>Wed)(?:nesday)?|
1691     (?<DN>Thu)(?:rsday)?|
1692     (?<DN>Sat)(?:urday)?
1693 nigel 91 .sp
1694     There are five capturing substrings, but only one is ever set after a match.
1695 ph10 182 (An alternative way of solving this problem is to use a "branch reset"
1696 ph10 175 subpattern, as described in the previous section.)
1697     .P
1698 nigel 91 The convenience function for extracting the data by name returns the substring
1699 nigel 93 for the first (and in this example, the only) subpattern of that name that
1700 ph10 461 matched. This saves searching to find which numbered subpattern it was.
1701 ph10 459 .P
1702 ph10 488 If you make a back reference to a non-unique named subpattern from elsewhere in
1703 ph10 1404 the pattern, the subpatterns to which the name refers are checked in the order
1704     in which they appear in the overall pattern. The first one that is set is used
1705     for the reference. For example, this pattern matches both "foofoo" and
1706 ph10 1361 "barbar" but not "foobar" or "barfoo":
1707     .sp
1708 ph10 1404 (?:(?<n>foo)|(?<n>bar))\ek<n>
1709 ph10 1361 .sp
1710     .P
1711     If you make a subroutine call to a non-unique named subpattern, the one that
1712     corresponds to the first occurrence of the name is used. In the absence of
1713     duplicate numbers (see the previous section) this is the one with the lowest
1714     number.
1715     .P
1716     If you use a named reference in a condition
1717 ph10 459 test (see the
1718     .\"
1719     .\" HTML <a href="#conditions">
1720     .\" </a>
1721     section about conditions
1722     .\"
1723 ph10 461 below), either to check whether a subpattern has matched, or to check for
1724 ph10 459 recursion, all subpatterns with the same name are tested. If the condition is
1725     true for any one of them, the overall condition is true. This is the same
1726     behaviour as testing by number. For further details of the interfaces for
1727     handling named subpatterns, see the
1728 nigel 63 .\" HREF
1729 nigel 75 \fBpcreapi\fP
1730 nigel 63 .\"
1731     documentation.
1732 ph10 385 .P
1733     \fBWarning:\fP You cannot use different names to distinguish between two
1734 ph10 457 subpatterns with the same number because PCRE uses only the numbers when
1735     matching. For this reason, an error is given at compile time if different names
1736 ph10 1361 are given to subpatterns with the same number. However, you can always give the
1737     same name to subpatterns with the same number, even when PCRE_DUPNAMES is not
1738     set.
1739 nigel 75 .
1740     .
1741 nigel 63 .SH REPETITION
1742     .rs
1743     .sp
1744     Repetition is specified by quantifiers, which can follow any of the following
1745     items:
1746 nigel 75 .sp
1747 nigel 63 a literal data character
1748 nigel 93 the dot metacharacter
1749 nigel 75 the \eC escape sequence
1750 ph10 859 the \eX escape sequence
1751 nigel 93 the \eR escape sequence
1752 ph10 572 an escape such as \ed or \epL that matches a single character
1753 nigel 63 a character class
1754     a back reference (see next section)
1755 ph10 637 a parenthesized subpattern (including assertions)
1756 ph10 716 a subroutine call to a subpattern (recursive or otherwise)
1757 nigel 75 .sp
1758 nigel 63 The general repetition quantifier specifies a minimum and maximum number of
1759     permitted matches, by giving the two numbers in curly brackets (braces),
1760     separated by a comma. The numbers must be less than 65536, and the first must
1761     be less than or equal to the second. For example:
1762 nigel 75 .sp
1763 nigel 63 z{2,4}
1764 nigel 75 .sp
1765 nigel 63 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1766     character. If the second number is omitted, but the comma is present, there is
1767     no upper limit; if the second number and the comma are both omitted, the
1768     quantifier specifies an exact number of required matches. Thus
1769 nigel 75 .sp
1770 nigel 63 [aeiou]{3,}
1771 nigel 75 .sp
1772 nigel 63 matches at least 3 successive vowels, but may match many more, while
1773 nigel 75 .sp
1774     \ed{8}
1775     .sp
1776 nigel 63 matches exactly 8 digits. An opening curly bracket that appears in a position
1777     where a quantifier is not allowed, or one that does not match the syntax of a
1778     quantifier, is taken as a literal character. For example, {,6} is not a
1779     quantifier, but a literal string of four characters.
1780 nigel 75 .P
1781 ph10 859 In UTF modes, quantifiers apply to characters rather than to individual data
1782     units. Thus, for example, \ex{100}{2} matches two characters, each of
1783     which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1784 ph10 1011 \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1785     several data units long (and they may be of different lengths).
1786 nigel 75 .P
1787 nigel 63 The quantifier {0} is permitted, causing the expression to behave as if the
1788 ph10 345 previous item and the quantifier were not present. This may be useful for
1789     subpatterns that are referenced as
1790 ph10 335 .\" HTML <a href="#subpatternsassubroutines">
1791     .\" </a>
1792     subroutines
1793     .\"
1794 ph10 572 from elsewhere in the pattern (but see also the section entitled
1795     .\" HTML <a href="#subdefine">
1796     .\" </a>
1797     "Defining subpatterns for use by reference only"
1798     .\"
1799     below). Items other than subpatterns that have a {0} quantifier are omitted
1800     from the compiled pattern.
1801 nigel 75 .P
1802 nigel 93 For convenience, the three most common quantifiers have single-character
1803     abbreviations:
1804 nigel 75 .sp
1805 nigel 63 * is equivalent to {0,}
1806     + is equivalent to {1,}
1807     ? is equivalent to {0,1}
1808 nigel 75 .sp
1809 nigel 63 It is possible to construct infinite loops by following a subpattern that can
1810     match no characters with a quantifier that has no upper limit, for example:
1811 nigel 75 .sp
1812 nigel 63 (a?)*
1813 nigel 75 .sp
1814 nigel 63 Earlier versions of Perl and PCRE used to give an error at compile time for
1815     such patterns. However, because there are cases where this can be useful, such
1816     patterns are now accepted, but if any repetition of the subpattern does in fact
1817     match no characters, the loop is forcibly broken.
1818 nigel 75 .P
1819 nigel 63 By default, the quantifiers are "greedy", that is, they match as much as
1820     possible (up to the maximum number of permitted times), without causing the
1821     rest of the pattern to fail. The classic example of where this gives problems
1822 nigel 75 is in trying to match comments in C programs. These appear between /* and */
1823     and within the comment, individual * and / characters may appear. An attempt to
1824     match C comments by applying the pattern
1825     .sp
1826     /\e*.*\e*/
1827     .sp
1828 nigel 63 to the string
1829 nigel 75 .sp
1830     /* first comment */ not comment /* second comment */
1831     .sp
1832 nigel 63 fails, because it matches the entire string owing to the greediness of the .*
1833     item.
1834 nigel 75 .P
1835 nigel 63 However, if a quantifier is followed by a question mark, it ceases to be
1836     greedy, and instead matches the minimum number of times possible, so the
1837     pattern
1838 nigel 75 .sp
1839     /\e*.*?\e*/
1840     .sp
1841 nigel 63 does the right thing with the C comments. The meaning of the various
1842     quantifiers is not otherwise changed, just the preferred number of matches.
1843     Do not confuse this use of question mark with its use as a quantifier in its
1844     own right. Because it has two uses, it can sometimes appear doubled, as in
1845 nigel 75 .sp
1846     \ed??\ed
1847     .sp
1848 nigel 63 which matches one digit by preference, but can match two if that is the only
1849     way the rest of the pattern matches.
1850 nigel 75 .P
1851 nigel 93 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1852 nigel 63 the quantifiers are not greedy by default, but individual ones can be made
1853     greedy by following them with a question mark. In other words, it inverts the
1854     default behaviour.
1855 nigel 75 .P
1856 nigel 63 When a parenthesized subpattern is quantified with a minimum repeat count that
1857 nigel 75 is greater than 1 or with a limited maximum, more memory is required for the
1858 nigel 63 compiled pattern, in proportion to the size of the minimum or maximum.
1859 nigel 75 .P
1860 nigel 63 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1861 nigel 93 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1862 nigel 63 implicitly anchored, because whatever follows will be tried against every
1863     character position in the subject string, so there is no point in retrying the
1864     overall match at any position after the first. PCRE normally treats such a
1865 nigel 75 pattern as though it were preceded by \eA.
1866     .P
1867 nigel 63 In cases where it is known that the subject string contains no newlines, it is
1868     worth setting PCRE_DOTALL in order to obtain this optimization, or
1869     alternatively using ^ to indicate anchoring explicitly.
1870 nigel 75 .P
1871 ph10 994 However, there are some cases where the optimization cannot be used. When .*
1872 ph10 488 is inside capturing parentheses that are the subject of a back reference
1873 nigel 93 elsewhere in the pattern, a match at the start may fail where a later one
1874     succeeds. Consider, for example:
1875 nigel 75 .sp
1876     (.*)abc\e1
1877     .sp
1878 nigel 63 If the subject is "xyz123abc123" the match point is the fourth character. For
1879     this reason, such a pattern is not implicitly anchored.
1880 nigel 75 .P
1881 ph10 994 Another case where implicit anchoring is not applied is when the leading .* is
1882     inside an atomic group. Once again, a match at the start may fail where a later
1883     one succeeds. Consider this pattern:
1884     .sp
1885     (?>.*?a)b
1886     .sp
1887 ph10 1221 It matches "ab" in the subject "aab". The use of the backtracking control verbs
1888 ph10 994 (*PRUNE) and (*SKIP) also disable this optimization.
1889     .P
1890 nigel 63 When a capturing subpattern is repeated, the value captured is the substring
1891     that matched the final iteration. For example, after
1892 nigel 75 .sp
1893     (tweedle[dume]{3}\es*)+
1894     .sp
1895 nigel 63 has matched "tweedledum tweedledee" the value of the captured substring is
1896     "tweedledee". However, if there are nested capturing subpatterns, the
1897     corresponding captured values may have been set in previous iterations. For
1898     example, after
1899 nigel 75 .sp
1900 nigel 63 /(a|(b))+/
1901 nigel 75 .sp
1902 nigel 63 matches "aba" the value of the second captured substring is "b".
1903 nigel 75 .
1904     .
1905     .\" HTML <a name="atomicgroup"></a>
1907 nigel 63 .rs
1908     .sp
1909 nigel 93 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1910     repetition, failure of what follows normally causes the repeated item to be
1911     re-evaluated to see if a different number of repeats allows the rest of the
1912     pattern to match. Sometimes it is useful to prevent this, either to change the
1913     nature of the match, or to cause it fail earlier than it otherwise might, when
1914     the author of the pattern knows there is no point in carrying on.
1915 nigel 75 .P
1916     Consider, for example, the pattern \ed+foo when applied to the subject line
1917     .sp
1918 nigel 63 123456bar
1919 nigel 75 .sp
1920 nigel 63 After matching all 6 digits and then failing to match "foo", the normal
1921 nigel 75 action of the matcher is to try again with only 5 digits matching the \ed+
1922 nigel 63 item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1923     (a term taken from Jeffrey Friedl's book) provides the means for specifying
1924     that once a subpattern has matched, it is not to be re-evaluated in this way.
1925 nigel 75 .P
1926 nigel 93 If we use atomic grouping for the previous example, the matcher gives up
1927 nigel 63 immediately on failing to match "foo" the first time. The notation is a kind of
1928     special parenthesis, starting with (?> as in this example:
1929 nigel 75 .sp
1930     (?>\ed+)foo
1931     .sp
1932 nigel 63 This kind of parenthesis "locks up" the part of the pattern it contains once
1933     it has matched, and a failure further into the pattern is prevented from
1934     backtracking into it. Backtracking past it to previous items, however, works as
1935     normal.
1936 nigel 75 .P
1937 nigel 63 An alternative description is that a subpattern of this type matches the string
1938     of characters that an identical standalone pattern would match, if anchored at
1939     the current point in the subject string.
1940 nigel 75 .P
1941 nigel 63 Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1942     the above example can be thought of as a maximizing repeat that must swallow
1943 nigel 75 everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1944 nigel 63 number of digits they match in order to make the rest of the pattern match,
1945 nigel 75 (?>\ed+) can only match an entire sequence of digits.
1946     .P
1947 nigel 63 Atomic groups in general can of course contain arbitrarily complicated
1948     subpatterns, and can be nested. However, when the subpattern for an atomic
1949     group is just a single repeated item, as in the example above, a simpler
1950     notation, called a "possessive quantifier" can be used. This consists of an
1951     additional + character following a quantifier. Using this notation, the
1952     previous example can be rewritten as
1953 nigel 75 .sp
1954     \ed++foo
1955     .sp
1956 ph10 208 Note that a possessive quantifier can be used with an entire group, for
1957     example:
1958     .sp
1959     (abc|xyz){2,3}+
1960     .sp
1961 nigel 63 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1962     option is ignored. They are a convenient notation for the simpler forms of
1963 nigel 93 atomic group. However, there is no difference in the meaning of a possessive
1964     quantifier and the equivalent atomic group, though there may be a performance
1965     difference; possessive quantifiers should be slightly faster.
1966 nigel 75 .P
1967 nigel 93 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1968     Jeffrey Friedl originated the idea (and the name) in the first edition of his
1969     book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1970     package, and PCRE copied it from there. It ultimately found its way into Perl
1971     at release 5.10.
1972 nigel 75 .P
1973 nigel 93 PCRE has an optimization that automatically "possessifies" certain simple
1974     pattern constructs. For example, the sequence A+B is treated as A++B because
1975     there is no point in backtracking into a sequence of A's when B must follow.
1976     .P
1977 nigel 63 When a pattern contains an unlimited repeat inside a subpattern that can itself
1978     be repeated an unlimited number of times, the use of an atomic group is the
1979     only way to avoid some failing matches taking a very long time indeed. The
1980     pattern
1981 nigel 75 .sp
1982     (\eD+|<\ed+>)*[!?]
1983     .sp
1984 nigel 63 matches an unlimited number of substrings that either consist of non-digits, or
1985     digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1986     quickly. However, if it is applied to
1987 nigel 75 .sp
1988 nigel 63 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1989 nigel 75 .sp
1990 nigel 63 it takes a long time before reporting failure. This is because the string can
1991 nigel 75 be divided between the internal \eD+ repeat and the external * repeat in a
1992     large number of ways, and all have to be tried. (The example uses [!?] rather
1993     than a single character at the end, because both PCRE and Perl have an
1994     optimization that allows for fast failure when a single character is used. They
1995     remember the last single character that is required for a match, and fail early
1996     if it is not present in the string.) If the pattern is changed so that it uses
1997     an atomic group, like this:
1998     .sp
1999     ((?>\eD+)|<\ed+>)*[!?]
2000     .sp
2001 nigel 63 sequences of non-digits cannot be broken, and failure happens quickly.
2002 nigel 75 .
2003     .
2004     .\" HTML <a name="backreferences"></a>
2006 nigel 63 .rs
2007     .sp
2008     Outside a character class, a backslash followed by a digit greater than 0 (and
2009     possibly further digits) is a back reference to a capturing subpattern earlier
2010     (that is, to its left) in the pattern, provided there have been that many
2011     previous capturing left parentheses.
2012 nigel 75 .P
2013 nigel 63 However, if the decimal number following the backslash is less than 10, it is
2014     always taken as a back reference, and causes an error only if there are not
2015     that many capturing left parentheses in the entire pattern. In other words, the
2016     parentheses that are referenced need not be to the left of the reference for
2017 nigel 91 numbers less than 10. A "forward back reference" of this type can make sense
2018     when a repetition is involved and the subpattern to the right has participated
2019     in an earlier iteration.
2020     .P
2021 nigel 93 It is not possible to have a numerical "forward back reference" to a subpattern
2022     whose number is 10 or more using this syntax because a sequence such as \e50 is
2023     interpreted as a character defined in octal. See the subsection entitled
2024 nigel 91 "Non-printing characters"
2025 nigel 75 .\" HTML <a href="#digitsafterbackslash">
2026     .\" </a>
2027     above
2028     .\"
2029 nigel 93 for further details of the handling of digits following a backslash. There is
2030     no such problem when named parentheses are used. A back reference to any
2031     subpattern is possible using named parentheses (see below).
2032 nigel 75 .P
2033 nigel 93 Another way of avoiding the ambiguity inherent in the use of digits following a
2034 ph10 572 backslash is to use the \eg escape sequence. This escape must be followed by an
2035     unsigned number or a negative number, optionally enclosed in braces. These
2036     examples are all identical:
2037 nigel 93 .sp
2038     (ring), \e1
2039     (ring), \eg1
2040     (ring), \eg{1}
2041     .sp
2042 ph10 208 An unsigned number specifies an absolute reference without the ambiguity that
2043     is present in the older syntax. It is also useful when literal digits follow
2044     the reference. A negative number is a relative reference. Consider this
2045     example:
2046 nigel 93 .sp
2047     (abc(def)ghi)\eg{-1}
2048     .sp
2049     The sequence \eg{-1} is a reference to the most recently started capturing
2050 ph10 572 subpattern before \eg, that is, is it equivalent to \e2 in this example.
2051     Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
2052     can be helpful in long patterns, and also in patterns that are created by
2053     joining together fragments that contain references within themselves.
2054 nigel 93 .P
2055 nigel 63 A back reference matches whatever actually matched the capturing subpattern in
2056     the current subject string, rather than anything matching the subpattern
2057     itself (see
2058     .\" HTML <a href="#subpatternsassubroutines">
2059     .\" </a>
2060     "Subpatterns as subroutines"
2061     .\"
2062     below for a way of doing that). So the pattern
2063 nigel 75 .sp
2064     (sens|respons)e and \e1ibility
2065     .sp
2066 nigel 63 matches "sense and sensibility" and "response and responsibility", but not
2067     "sense and responsibility". If caseful matching is in force at the time of the
2068     back reference, the case of letters is relevant. For example,
2069 nigel 75 .sp
2070     ((?i)rah)\es+\e1
2071     .sp
2072 nigel 63 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
2073     capturing subpattern is matched caselessly.
2074 nigel 75 .P
2075 ph10 171 There are several different ways of writing back references to named
2076     subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
2077     \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
2078     back reference syntax, in which \eg can be used for both numeric and named
2079     references, is also supported. We could rewrite the above example in any of
2080 nigel 93 the following ways:
2081 nigel 75 .sp
2082 nigel 93 (?<p1>(?i)rah)\es+\ek<p1>
2083 ph10 171 (?'p1'(?i)rah)\es+\ek{p1}
2084 nigel 91 (?P<p1>(?i)rah)\es+(?P=p1)
2085 ph10 171 (?<p1>(?i)rah)\es+\eg{p1}
2086 nigel 75 .sp
2087 nigel 91 A subpattern that is referenced by name may appear in the pattern before or
2088     after the reference.
2089     .P
2090 nigel 63 There may be more than one back reference to the same subpattern. If a
2091     subpattern has not actually been used in a particular match, any back
2092 ph10 456 references to it always fail by default. For example, the pattern
2093 nigel 75 .sp
2094     (a|(bc))\e2
2095     .sp
2096 ph10 461 always fails if it starts to match "a" rather than "bc". However, if the
2097     PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
2098 ph10 456 unset value matches an empty string.
2099     .P
2100     Because there may be many capturing parentheses in a pattern, all digits
2101     following a backslash are taken as part of a potential back reference number.
2102     If the pattern continues with a digit character, some delimiter must be used to
2103     terminate the back reference. If the PCRE_EXTENDED option is set, this can be
2104 ph10 968 white space. Otherwise, the \eg{ syntax or an empty comment (see
2105 nigel 75 .\" HTML <a href="#comments">
2106     .\" </a>
2107     "Comments"
2108     .\"
2109     below) can be used.
2110 ph10 488 .
2111     .SS "Recursive back references"
2112     .rs
2113     .sp
2114 nigel 63 A back reference that occurs inside the parentheses to which it refers fails
2115 nigel 75 when the subpattern is first used, so, for example, (a\e1) never matches.
2116 nigel 63 However, such references can be useful inside repeated subpatterns. For
2117     example, the pattern
2118 nigel 75 .sp
2119     (a|b\e1)+
2120     .sp
2121 nigel 63 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
2122     the subpattern, the back reference matches the character string corresponding
2123     to the previous iteration. In order for this to work, the pattern must be such
2124     that the first iteration does not need to match the back reference. This can be
2125     done using alternation, as in the example above, or by a quantifier with a
2126     minimum of zero.
2127 ph10 488 .P
2128     Back references of this type cause the group that they reference to be treated
2129     as an
2130     .\" HTML <a href="#atomicgroup">
2131     .\" </a>
2132     atomic group.
2133     .\"
2134     Once the whole group has been matched, a subsequent matching failure cannot
2135     cause backtracking into the middle of the group.
2136 nigel 75 .
2137     .
2138     .\" HTML <a name="bigassertions"></a>
2139 nigel 63 .SH ASSERTIONS
2140     .rs
2141     .sp
2142     An assertion is a test on the characters following or preceding the current
2143     matching point that does not actually consume any characters. The simple
2144 nigel 75 assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
2145     .\" HTML <a href="#smallassertions">
2146     .\" </a>
2147     above.
2148     .\"
2149     .P
2150 nigel 63 More complicated assertions are coded as subpatterns. There are two kinds:
2151     those that look ahead of the current position in the subject string, and those
2152 nigel 75 that look behind it. An assertion subpattern is matched in the normal way,
2153     except that it does not cause the current matching position to be changed.
2154     .P
2155 ph10 637 Assertion subpatterns are not capturing subpatterns. If such an assertion
2156     contains capturing subpatterns within it, these are counted for the purposes of
2157     numbering the capturing subpatterns in the whole pattern. However, substring
2158 ph10 1335 capturing is carried out only for positive assertions. (Perl sometimes, but not
2159 ph10 1292 always, does do capturing in negative assertions.)
2160 ph10 637 .P
2161 ph10 643 For compatibility with Perl, assertion subpatterns may be repeated; though
2162 ph10 654 it makes no sense to assert the same thing several times, the side effect of
2163 ph10 643 capturing parentheses may occasionally be useful. In practice, there only three
2164     cases:
2165 ph10 637 .sp
2166 ph10 654 (1) If the quantifier is {0}, the assertion is never obeyed during matching.
2167     However, it may contain internal capturing parenthesized groups that are called
2168 ph10 637 from elsewhere via the
2169     .\" HTML <a href="#subpatternsassubroutines">
2170     .\" </a>
2171     subroutine mechanism.
2172     .\"
2173     .sp
2174 ph10 654 (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
2175     were {0,1}. At run time, the rest of the pattern match is tried with and
2176 ph10 637 without the assertion, the order depending on the greediness of the quantifier.
2177     .sp
2178 ph10 654 (3) If the minimum repetition is greater than zero, the quantifier is ignored.
2179 ph10 637 The assertion is obeyed just once when encountered during matching.
2180 nigel 75 .
2181     .
2182     .SS "Lookahead assertions"
2183     .rs
2184     .sp
2185 nigel 91 Lookahead assertions start with (?= for positive assertions and (?! for
2186     negative assertions. For example,
2187 nigel 75 .sp
2188     \ew+(?=;)
2189     .sp
2190 nigel 63 matches a word followed by a semicolon, but does not include the semicolon in
2191     the match, and
2192 nigel 75 .sp
2193 nigel 63 foo(?!bar)
2194 nigel 75 .sp
2195 nigel 63 matches any occurrence of "foo" that is not followed by "bar". Note that the
2196     apparently similar pattern
2197 nigel 75 .sp
2198 nigel 63 (?!foo)bar
2199 nigel 75 .sp
2200 nigel 63 does not find an occurrence of "bar" that is preceded by something other than
2201     "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2202     (?!foo) is always true when the next three characters are "bar". A
2203 nigel 75 lookbehind assertion is needed to achieve the other effect.
2204     .P
2205 nigel 63 If you want to force a matching failure at some point in a pattern, the most
2206     convenient way to do it is with (?!) because an empty string always matches, so
2207     an assertion that requires there not to be an empty string must always fail.
2208 ph10 572 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2209 nigel 75 .
2210     .
2211     .\" HTML <a name="lookbehind"></a>
2212     .SS "Lookbehind assertions"
2213     .rs
2214     .sp
2215 nigel 63 Lookbehind assertions start with (?<= for positive assertions and (?<! for
2216     negative assertions. For example,
2217 nigel 75 .sp
2218 nigel 63 (?<!foo)bar
2219 nigel 75 .sp
2220 nigel 63 does find an occurrence of "bar" that is not preceded by "foo". The contents of
2221     a lookbehind assertion are restricted such that all the strings it matches must
2222 nigel 91 have a fixed length. However, if there are several top-level alternatives, they
2223     do not all have to have the same fixed length. Thus
2224 nigel 75 .sp
2225 nigel 63 (?<=bullock|donkey)
2226 nigel 75 .sp
2227 nigel 63 is permitted, but
2228 nigel 75 .sp
2229 nigel 63 (?<!dogs?|cats?)
2230 nigel 75 .sp
2231 nigel 63 causes an error at compile time. Branches that match different length strings
2232     are permitted only at the top level of a lookbehind assertion. This is an
2233 ph10 572 extension compared with Perl, which requires all branches to match the same
2234     length of string. An assertion such as
2235 nigel 75 .sp
2236 nigel 63 (?<=ab(c|de))
2237 nigel 75 .sp
2238 nigel 63 is not permitted, because its single top-level branch can match two different
2239 ph10 454 lengths, but it is acceptable to PCRE if rewritten to use two top-level
2240     branches:
2241 nigel 75 .sp
2242 nigel 63 (?<=abc|abde)
2243 nigel 75 .sp
2244 ph10 572 In some cases, the escape sequence \eK
2245 ph10 168 .\" HTML <a href="#resetmatchstart">
2246     .\" </a>
2247     (see above)
2248     .\"
2249 ph10 461 can be used instead of a lookbehind assertion to get round the fixed-length
2250 ph10 454 restriction.
2251 ph10 168 .P
2252 nigel 63 The implementation of lookbehind assertions is, for each alternative, to
2253 nigel 93 temporarily move the current position back by the fixed length and then try to
2254 nigel 63 match. If there are insufficient characters before the current position, the
2255 nigel 93 assertion fails.
2256 nigel 75 .P
2257 ph10 859 In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2258     unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2259     it impossible to calculate the length of the lookbehind. The \eX and \eR
2260     escapes, which can match different numbers of data units, are also not
2261     permitted.
2262 nigel 75 .P
2263 ph10 454 .\" HTML <a href="#subpatternsassubroutines">
2264     .\" </a>
2265     "Subroutine"
2266     .\"
2267     calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2268 ph10 461 as the subpattern matches a fixed-length string.
2269 ph10 454 .\" HTML <a href="#recursion">
2270     .\" </a>
2271     Recursion,
2272     .\"
2273     however, is not supported.
2274     .P
2275 nigel 93 Possessive quantifiers can be used in conjunction with lookbehind assertions to
2276 ph10 456 specify efficient matching of fixed-length strings at the end of subject
2277     strings. Consider a simple pattern such as
2278 nigel 75 .sp
2279 nigel 63 abcd$
2280 nigel 75 .sp
2281 nigel 63 when applied to a long string that does not match. Because matching proceeds
2282     from left to right, PCRE will look for each "a" in the subject and then see if
2283     what follows matches the rest of the pattern. If the pattern is specified as
2284 nigel 75 .sp
2285 nigel 63 ^.*abcd$
2286 nigel 75 .sp
2287 nigel 63 the initial .* matches the entire string at first, but when this fails (because
2288     there is no following "a"), it backtracks to match all but the last character,
2289     then all but the last two characters, and so on. Once again the search for "a"
2290     covers the entire string, from right to left, so we are no better off. However,
2291     if the pattern is written as
2292 nigel 75 .sp
2293 nigel 63 ^.*+(?<=abcd)
2294 nigel 75 .sp
2295 nigel 93 there can be no backtracking for the .*+ item; it can match only the entire
2296 nigel 63 string. The subsequent lookbehind assertion does a single test on the last four
2297     characters. If it fails, the match fails immediately. For long strings, this
2298     approach makes a significant difference to the processing time.
2299 nigel 75 .
2300     .
2301     .SS "Using multiple assertions"
2302     .rs
2303     .sp
2304 nigel 63 Several assertions (of any sort) may occur in succession. For example,
2305 nigel 75 .sp
2306     (?<=\ed{3})(?<!999)foo
2307     .sp
2308 nigel 63 matches "foo" preceded by three digits that are not "999". Notice that each of
2309     the assertions is applied independently at the same point in the subject
2310     string. First there is a check that the previous three characters are all
2311     digits, and then there is a check that the same three characters are not "999".
2312 nigel 75 This pattern does \fInot\fP match "foo" preceded by six characters, the first
2313 nigel 63 of which are digits and the last three of which are not "999". For example, it
2314     doesn't match "123abcfoo". A pattern to do that is
2315 nigel 75 .sp
2316     (?<=\ed{3}...)(?<!999)foo
2317     .sp
2318 nigel 63 This time the first assertion looks at the preceding six characters, checking
2319     that the first three are digits, and then the second assertion checks that the
2320     preceding three characters are not "999".
2321 nigel 75 .P
2322 nigel 63 Assertions can be nested in any combination. For example,
2323 nigel 75 .sp
2324 nigel 63 (?<=(?<!foo)bar)baz
2325 nigel 75 .sp
2326 nigel 63 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2327     preceded by "foo", while
2328 nigel 75 .sp
2329     (?<=\ed{3}(?!999)...)foo
2330     .sp
2331     is another pattern that matches "foo" preceded by three digits and any three
2332 nigel 63 characters that are not "999".
2333 nigel 75 .
2334     .
2335 nigel 91 .\" HTML <a name="conditions"></a>
2337 nigel 63 .rs
2338     .sp
2339     It is possible to cause the matching process to obey a subpattern
2340     conditionally or to choose between two alternative subpatterns, depending on
2341 ph10 461 the result of an assertion, or whether a specific capturing subpattern has
2342 ph10 456 already been matched. The two possible forms of conditional subpattern are:
2343 nigel 75 .sp
2344 nigel 63 (?(condition)yes-pattern)
2345     (?(condition)yes-pattern|no-pattern)
2346 nigel 75 .sp
2347 nigel 63 If the condition is satisfied, the yes-pattern is used; otherwise the
2348     no-pattern (if present) is used. If there are more than two alternatives in the
2349 ph10 557 subpattern, a compile-time error occurs. Each of the two alternatives may
2350 ph10 579 itself contain nested subpatterns of any form, including conditional
2351 ph10 557 subpatterns; the restriction to two alternatives applies only at the level of
2352 ph10 579 the condition. This pattern fragment is an example where the alternatives are
2353 ph10 557 complex:
2354     .sp
2355     (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2356     .sp
2357 nigel 75 .P
2358 nigel 93 There are four kinds of condition: references to subpatterns, references to
2359     recursion, a pseudo-condition called DEFINE, and assertions.
2360     .
2361     .SS "Checking for a used subpattern by number"
2362     .rs
2363     .sp
2364     If the text between the parentheses consists of a sequence of digits, the
2365 ph10 456 condition is true if a capturing subpattern of that number has previously
2366 ph10 461 matched. If there is more than one capturing subpattern with the same number
2367     (see the earlier
2368 ph10 456 .\"
2369     .\" HTML <a href="#recursion">
2370     .\" </a>
2371     section about duplicate subpattern numbers),
2372     .\"
2373 ph10 572 the condition is true if any of them have matched. An alternative notation is
2374 ph10 456 to precede the digits with a plus or minus sign. In this case, the subpattern
2375     number is relative rather than absolute. The most recently opened parentheses
2376 ph10 572 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2377 ph10 579 loops it can also make sense to refer to subsequent groups. The next
2378 ph10 572 parentheses to be opened can be referenced as (?(+1), and so on. (The value
2379     zero in any of these forms is not used; it provokes a compile-time error.)
2380 nigel 91 .P
2381     Consider the following pattern, which contains non-significant white space to
2382     make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2383     three parts for ease of discussion:
2384 nigel 75 .sp
2385     ( \e( )? [^()]+ (?(1) \e) )
2386     .sp
2387 nigel 63 The first part matches an optional opening parenthesis, and if that
2388     character is present, sets it as the first captured substring. The second part
2389     matches one or more characters that are not parentheses. The third part is a
2390 ph10 572 conditional subpattern that tests whether or not the first set of parentheses
2391     matched. If they did, that is, if subject started with an opening parenthesis,
2392 nigel 63 the condition is true, and so the yes-pattern is executed and a closing
2393     parenthesis is required. Otherwise, since no-pattern is not present, the
2394     subpattern matches nothing. In other words, this pattern matches a sequence of
2395 nigel 93 non-parentheses, optionally enclosed in parentheses.
2396 ph10 167 .P
2397 ph10 172 If you were embedding this pattern in a larger one, you could use a relative
2398 ph10 167 reference:
2399     .sp
2400     ...other stuff... ( \e( )? [^()]+ (?(-1) \e) ) ...
2401     .sp
2402     This makes the fragment independent of the parentheses in the larger pattern.
2403 nigel 93 .
2404     .SS "Checking for a used subpattern by name"
2405     .rs
2406 nigel 91 .sp
2407 nigel 93 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2408     subpattern by name. For compatibility with earlier versions of PCRE, which had
2409 ph10 1404 this facility before Perl, the syntax (?(name)...) is also recognized.
2410 nigel 93 .P
2411     Rewriting the above example to use a named subpattern gives this:
2412 nigel 91 .sp
2413 nigel 93 (?<OPEN> \e( )? [^()]+ (?(<OPEN>) \e) )
2414     .sp
2415 ph10 461 If the name used in a condition of this kind is a duplicate, the test is
2416     applied to all subpatterns of the same name, and is true if any one of them has
2417 ph10 459 matched.
2418 nigel 93 .
2419     .SS "Checking for pattern recursion"
2420     .rs
2421     .sp
2422 nigel 91 If the condition is the string (R), and there is no subpattern with the name R,
2423 nigel 93 the condition is true if a recursive call to the whole pattern or any
2424     subpattern has been made. If digits or a name preceded by ampersand follow the
2425     letter R, for example:
2426     .sp
2427     (?(R3)...) or (?(R&name)...)
2428     .sp
2429 ph10 456 the condition is true if the most recent recursion is into a subpattern whose
2430 nigel 93 number or name is given. This condition does not check the entire recursion
2431 ph10 461 stack. If the name used in a condition of this kind is a duplicate, the test is
2432     applied to all subpatterns of the same name, and is true if any one of them is
2433     the most recent recursion.
2434 nigel 75 .P
2435 ph10 461 At "top level", all these recursion test conditions are false.
2436 ph10 454 .\" HTML <a href="#recursion">
2437     .\" </a>
2438 ph10 459 The syntax for recursive patterns
2439 ph10 454 .\"
2440 ph10 459 is described below.
2441 nigel 93 .
2442 ph10 572 .\" HTML <a name="subdefine"></a>
2443 nigel 93 .SS "Defining subpatterns for use by reference only"
2444     .rs
2445     .sp
2446     If the condition is the string (DEFINE), and there is no subpattern with the
2447     name DEFINE, the condition is always false. In this case, there may be only one
2448     alternative in the subpattern. It is always skipped if control reaches this
2449     point in the pattern; the idea of DEFINE is that it can be used to define
2450 ph10 716 subroutines that can be referenced from elsewhere. (The use of
2451 ph10 454 .\" HTML <a href="#subpatternsassubroutines">
2452     .\" </a>
2453 ph10 716 subroutines
2454 ph10 454 .\"
2455 ph10 572 is described below.) For example, a pattern to match an IPv4 address such as
2456 ph10 968 "" could be written like this (ignore white space and line
2457 ph10 572 breaks):
2458 nigel 93 .sp
2459     (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2460     \eb (?&byte) (\e.(?&byte)){3} \eb
2461     .sp
2462     The first part of the pattern is a DEFINE group inside which a another group
2463     named "byte" is defined. This matches an individual component of an IPv4
2464     address (a number less than 256). When matching takes place, this part of the
2465 ph10 456 pattern is skipped because DEFINE acts like a false condition. The rest of the
2466     pattern uses references to the named group to match the four dot-separated
2467     components of an IPv4 address, insisting on a word boundary at each end.
2468 nigel 93 .
2469     .SS "Assertion conditions"
2470     .rs
2471     .sp
2472     If the condition is not in any of the above formats, it must be an assertion.
2473 nigel 63 This may be a positive or negative lookahead or lookbehind assertion. Consider
2474     this pattern, again containing non-significant white space, and with the two
2475     alternatives on the second line:
2476 nigel 75 .sp
2477 nigel 63 (?(?=[^a-z]*[a-z])
2478 nigel 75 \ed{2}-[a-z]{3}-\ed{2} | \ed{2}-\ed{2}-\ed{2} )
2479     .sp
2480 nigel 63 The condition is a positive lookahead assertion that matches an optional
2481     sequence of non-letters followed by a letter. In other words, it tests for the
2482     presence of at least one letter in the subject. If a letter is found, the
2483     subject is matched against the first alternative; otherwise it is matched
2484     against the second. This pattern matches strings in one of the two forms
2485     dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2486 nigel 75 .
2487     .
2488     .\" HTML <a name="comments"></a>
2489 nigel 63 .SH COMMENTS
2490     .rs
2491     .sp
2492 ph10 579 There are two ways of including comments in patterns that are processed by
2493 ph10 562 PCRE. In both cases, the start of the comment must not be in a character class,
2494     nor in the middle of any other sequence of related characters such as (?: or a
2495     subpattern name or number. The characters that make up a comment play no part
2496     in the pattern matching.
2497     .P
2498 nigel 75 The sequence (?# marks the start of a comment that continues up to the next
2499 ph10 562 closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2500     option is set, an unescaped # character also introduces a comment, which in
2501     this case continues to immediately after the next newline character or
2502     character sequence in the pattern. Which characters are interpreted as newlines
2503 ph10 859 is controlled by the options passed to a compiling function or by a special
2504 ph10 562 sequence at the start of the pattern, as described in the section entitled
2505 ph10 572 .\" HTML <a href="#newlines">
2506 ph10 556 .\" </a>
2507     "Newline conventions"
2508     .\"
2509 ph10 572 above. Note that the end of this type of comment is a literal newline sequence
2510     in the pattern; escape sequences that happen to represent a newline do not
2511     count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2512     default newline convention is in force:
2513 ph10 556 .sp
2514     abc #comment \en still comment
2515     .sp
2516 ph10 579 On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2517 ph10 556 a newline in the pattern. The sequence \en is still literal at this stage, so
2518     it does not terminate the comment. Only an actual character with the code value
2519 ph10 562 0x0a (the default newline) does so.
2520 nigel 75 .
2521     .
2522 nigel 91 .\" HTML <a name="recursion"></a>
2523 nigel 75 .SH "RECURSIVE PATTERNS"
2524 nigel 63 .rs
2525     .sp
2526     Consider the problem of matching a string in parentheses, allowing for
2527     unlimited nested parentheses. Without the use of recursion, the best that can
2528     be done is to use a pattern that matches up to some fixed depth of nesting. It
2529 nigel 93 is not possible to handle an arbitrary nesting depth.
2530     .P
2531     For some time, Perl has provided a facility that allows regular expressions to
2532     recurse (amongst other things). It does this by interpolating Perl code in the
2533     expression at run time, and the code can refer to the expression itself. A Perl
2534     pattern using code interpolation to solve the parentheses problem can be
2535     created like this:
2536 nigel 75 .sp
2537     $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2538     .sp
2539 nigel 63 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2540 nigel 93 recursively to the pattern in which it appears.
2541 nigel 75 .P
2542 nigel 93 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2543     supports special syntax for recursion of the entire pattern, and also for
2544     individual subpattern recursion. After its introduction in PCRE and Python,
2545 ph10 453 this kind of recursion was subsequently introduced into Perl at release 5.10.
2546 nigel 75 .P
2547 nigel 93 A special item that consists of (? followed by a number greater than zero and a
2548 ph10 716 closing parenthesis is a recursive subroutine call of the subpattern of the
2549     given number, provided that it occurs inside that subpattern. (If not, it is a
2550 ph10 454 .\" HTML <a href="#subpatternsassubroutines">
2551     .\" </a>
2552 ph10 716 non-recursive subroutine
2553 ph10 454 .\"
2554 nigel 93 call, which is described in the next section.) The special item (?R) or (?0) is
2555     a recursive call of the entire regular expression.
2556 nigel 87 .P
2557     This PCRE pattern solves the nested parentheses problem (assume the
2558     PCRE_EXTENDED option is set so that white space is ignored):
2559 nigel 75 .sp
2560 ph10 456 \e( ( [^()]++ | (?R) )* \e)
2561 nigel 75 .sp
2562 nigel 63 First it matches an opening parenthesis. Then it matches any number of
2563     substrings which can either be a sequence of non-parentheses, or a recursive
2564 nigel 87 match of the pattern itself (that is, a correctly parenthesized substring).
2565 ph10 461 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2566 ph10 456 to avoid backtracking into sequences of non-parentheses.
2567 nigel 75 .P
2568 nigel 63 If this were part of a larger pattern, you would not want to recurse the entire
2569     pattern, so instead you could use this:
2570 nigel 75 .sp
2571 ph10 456 ( \e( ( [^()]++ | (?1) )* \e) )
2572 nigel 75 .sp
2573 nigel 63 We have put the pattern into parentheses, and caused the recursion to refer to
2574 ph10 172 them instead of the whole pattern.
2575 ph10 166 .P
2576     In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2577 ph10 572 is made easier by the use of relative references. Instead of (?1) in the
2578     pattern above you can write (?-2) to refer to the second most recently opened
2579     parentheses preceding the recursion. In other words, a negative number counts
2580     capturing parentheses leftwards from the point at which it is encountered.
2581 ph10 166 .P
2582     It is also possible to refer to subsequently opened parentheses, by writing
2583     references such as (?+2). However, these cannot be recursive because the
2584     reference is not inside the parentheses that are referenced. They are always
2585 ph10 454 .\" HTML <a href="#subpatternsassubroutines">
2586     .\" </a>
2587 ph10 716 non-recursive subroutine
2588 ph10 454 .\"
2589     calls, as described in the next section.
2590 ph10 166 .P
2591     An alternative approach is to use named parentheses instead. The Perl syntax
2592     for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2593     could rewrite the above example as follows:
2594 nigel 75 .sp
2595 ph10 456 (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2596 nigel 75 .sp
2597 nigel 93 If there is more than one subpattern with the same name, the earliest one is
2598 ph10 172 used.
2599 ph10 166 .P
2600     This particular example pattern that we have been looking at contains nested
2601 ph10 456 unlimited repeats, and so the use of a possessive quantifier for matching
2602     strings of non-parentheses is important when applying the pattern to strings
2603     that do not match. For example, when this pattern is applied to
2604 nigel 75 .sp
2605 nigel 63 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2606 nigel 75 .sp
2607 ph10 456 it yields "no match" quickly. However, if a possessive quantifier is not used,
2608 nigel 63 the match runs for a very long time indeed because there are so many different
2609     ways the + and * repeats can carve up the subject, and all have to be tested
2610     before failure can be reported.
2611 nigel 75 .P
2612 ph10 464 At the end of a match, the values of capturing parentheses are those from
2613     the outermost level. If you want to obtain intermediate values, a callout
2614     function can be used (see below and the
2615 nigel 63 .\" HREF
2616 nigel 75 \fBpcrecallout\fP
2617 nigel 63 .\"
2618     documentation). If the pattern above is matched against
2619 nigel 75 .sp
2620 nigel 63 (ab(cd)ef)
2621 nigel 75 .sp
2622 ph10 464 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2623     the last value taken on at the top level. If a capturing subpattern is not
2624 ph10 724 matched at the top level, its final captured value is unset, even if it was
2625     (temporarily) set at a deeper level during the matching process.
2626 nigel 75 .P
2627 ph10 464 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2628     obtain extra memory to store data during a recursion, which it does by using
2629     \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2630     be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2631     .P
2632 nigel 63 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2633     Consider this pattern, which matches text in angle brackets, allowing for
2634     arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2635     recursing), whereas any characters are permitted at the outer level.
2636 nigel 75 .sp
2637     < (?: (?(R) \ed++ | [^<>]*+) | (?R)) * >
2638     .sp
2639 nigel 63 In this pattern, (?(R) is the start of a conditional subpattern, with two
2640     different alternatives for the recursive and non-recursive cases. The (?R) item
2641     is the actual recursive call.
2642 nigel 75 .
2643     .
2644 ph10 453 .\" HTML <a name="recursiondifference"></a>
2645 ph10 724 .SS "Differences in recursion processing between PCRE and Perl"
2646 ph10 453 .rs
2647     .sp
2648 ph10 724 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2649     (like Python, but unlike Perl), a recursive subpattern call is always treated
2650     as an atomic group. That is, once it has matched some of the subject string, it
2651     is never re-entered, even if it contains untried alternatives and there is a
2652     subsequent matching failure. This can be illustrated by the following pattern,
2653     which purports to match a palindromic string that contains an odd number of
2654     characters (for example, "a", "aba", "abcba", "abcdcba"):
2655 ph10 453 .sp
2656     ^(.|(.)(?1)\e2)$
2657     .sp
2658 ph10 461 The idea is that it either matches a single character, or two identical
2659     characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2660 ph10 453 it does not if the pattern is longer than three characters. Consider the
2661     subject string "abcba":
2662     .P
2663 ph10 461 At the top level, the first character is matched, but as it is not at the end
2664 ph10 453 of the string, the first alternative fails; the second alternative is taken
2665     and the recursion kicks in. The recursive call to subpattern 1 successfully
2666     matches the next character ("b"). (Note that the beginning and end of line
2667     tests are not part of the recursion).
2668     .P
2669     Back at the top level, the next character ("c") is compared with what
2670 ph10 461 subpattern 2 matched, which was "a". This fails. Because the recursion is
2671 ph10 453 treated as an atomic group, there are now no backtracking points, and so the
2672     entire match fails. (Perl is able, at this point, to re-enter the recursion and
2673     try the second alternative.) However, if the pattern is written with the
2674     alternatives in the other order, things are different:
2675     .sp
2676     ^((.)(?1)\e2|.)$
2677     .sp
2678 ph10 461 This time, the recursing alternative is tried first, and continues to recurse
2679     until it runs out of characters, at which point the recursion fails. But this
2680     time we do have another alternative to try at the higher level. That is the big
2681 ph10 453 difference: in the previous case the remaining alternative is at a deeper
2682     recursion level, which PCRE cannot use.
2683     .P
2684 ph10 572 To change the pattern so that it matches all palindromic strings, not just
2685     those with an odd number of characters, it is tempting to change the pattern to
2686     this:
2687 ph10 453 .sp
2688     ^((.)(?1)\e2|.?)$
2689     .sp
2690 ph10 461 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2691     deeper recursion has matched a single character, it cannot be entered again in
2692     order to match an empty string. The solution is to separate the two cases, and
2693 ph10 453 write out the odd and even cases as alternatives at the higher level:
2694     .sp
2695     ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2696 ph10 461 .sp
2697     If you want to match typical palindromic phrases, the pattern has to ignore all
2698 ph10 453 non-word characters, which can be done like this:
2699     .sp
2700 ph10 461 ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2701 ph10 453 .sp
2702 ph10 461 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2703     man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2704     the use of the possessive quantifier *+ to avoid backtracking into sequences of
2705 ph10 453 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2706     more) to match typical phrases, and Perl takes so long that you think it has
2707     gone into a loop.
2708 ph10 456 .P
2709     \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2710     string does not start with a palindrome that is shorter than the entire string.
2711     For example, although "abcba" is correctly matched, if the subject is "ababa",
2712     PCRE finds the palindrome "aba" at the start, then fails at top level because
2713     the end of the string does not follow. Once again, it cannot jump back into the
2714     recursion to try other alternatives, so the entire match fails.
2715 ph10 724 .P
2716 ph10 733 The second way in which PCRE and Perl differ in their recursion processing is
2717     in the handling of captured values. In Perl, when a subpattern is called
2718     recursively or as a subpattern (see the next section), it has no access to any
2719     values that were captured outside the recursion, whereas in PCRE these values
2720 ph10 724 can be referenced. Consider this pattern:
2721     .sp
2722     ^(.)(\e1|a(?2))
2723     .sp
2724 ph10 733 In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2725 ph10 724 then in the second group, when the back reference \e1 fails to match "b", the
2726     second alternative matches "a" and then recurses. In the recursion, \e1 does
2727     now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2728     match because inside the recursive call \e1 cannot access the externally set
2729     value.
2730 ph10 453 .
2731     .
2732 nigel 63 .\" HTML <a name="subpatternsassubroutines"></a>
2734 nigel 63 .rs
2735     .sp
2736 ph10 716 If the syntax for a recursive subpattern call (either by number or by
2737 nigel 63 name) is used outside the parentheses to which it refers, it operates like a
2738 ph10 716 subroutine in a programming language. The called subpattern may be defined
2739 ph10 166 before or after the reference. A numbered reference can be absolute or
2740     relative, as in these examples:
2741 nigel 75 .sp
2742 ph10 166 (...(absolute)...)...(?2)...
2743     (...(relative)...)...(?-1)...
2744 ph10 172 (...(?+1)...(relative)...
2745 ph10 166 .sp
2746     An earlier example pointed out that the pattern
2747     .sp
2748 nigel 75 (sens|respons)e and \e1ibility
2749     .sp
2750 nigel 63 matches "sense and sensibility" and "response and responsibility", but not
2751     "sense and responsibility". If instead the pattern
2752 nigel 75 .sp
2753 nigel 63 (sens|respons)e and (?1)ibility
2754 nigel 75 .sp
2755 nigel 63 is used, it does match "sense and responsibility" as well as the other two
2756 nigel 93 strings. Another example is given in the discussion of DEFINE above.
2757 nigel 87 .P
2758 ph10 716 All subroutine calls, whether recursive or not, are always treated as atomic
2759     groups. That is, once a subroutine has matched some of the subject string, it
2760     is never re-entered, even if it contains untried alternatives and there is a
2761     subsequent matching failure. Any capturing parentheses that are set during the
2762     subroutine call revert to their previous values afterwards.
2763 nigel 93 .P
2764 ph10 716 Processing options such as case-independence are fixed when a subpattern is
2765     defined, so if it is used as a subroutine, such options cannot be changed for
2766     different calls. For example, consider this pattern:
2767 nigel 93 .sp
2768 ph10 166 (abc)(?i:(?-1))
2769 nigel 93 .sp
2770     It matches "abcabc". It does not match "abcABC" because the change of
2771     processing option does not affect the called subpattern.
2772 nigel 75 .
2773     .
2774 ph10 333 .\" HTML <a name="onigurumasubroutines"></a>
2776     .rs
2777     .sp
2778 ph10 345 For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2779     a number enclosed either in angle brackets or single quotes, is an alternative
2780     syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2781 ph10 333 are two of the examples used above, rewritten using this syntax:
2782     .sp
2783     (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2784     (sens|respons)e and \eg'1'ibility
2785     .sp
2786 ph10 345 PCRE supports an extension to Oniguruma: if a number is preceded by a
2787 ph10 333 plus or a minus sign it is taken as a relative reference. For example:
2788     .sp
2789     (abc)(?i:\eg<-1>)
2790     .sp
2791 ph10 345 Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2792 ph10 333 synonymous. The former is a back reference; the latter is a subroutine call.
2793     .
2794     .
2795 nigel 63 .SH CALLOUTS
2796     .rs
2797     .sp
2798     Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2799     code to be obeyed in the middle of matching a regular expression. This makes it
2800     possible, amongst other things, to extract different substrings that match the
2801     same pair of parentheses when there is a repetition.
2802 nigel 75 .P
2803 nigel 63 PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2804     code. The feature is called "callout". The caller of PCRE provides an external
2805 ph10 903 function by putting its entry point in the global variable \fIpcre_callout\fP
2806 chpe 1055 (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2807     By default, this variable contains NULL, which disables all calling out.
2808 nigel 75 .P
2809 nigel 63 Within a regular expression, (?C) indicates the points at which the external
2810     function is to be called. If you want to identify different callout points, you
2811     can put a number less than 256 after the letter C. The default value is zero.
2812     For example, this pattern has two callout points:
2813 nigel 75 .sp
2814 ph10 155 (?C1)abc(?C2)def
2815 nigel 75 .sp
2816 ph10 859 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2817 nigel 75 automatically installed before each item in the pattern. They are all numbered
2818 ph10 1266 255. If there is a conditional group in the pattern whose condition is an
2819     assertion, an additional callout is inserted just before the condition. An
2820     explicit callout may also be set at this position, as in this example:
2821     .sp
2822     (?(?C9)(?=a)abc|def)
2823     .sp
2824 ph10 1335 Note that this applies only to assertion conditions, not to other types of
2825     condition.
2826 nigel 75 .P
2827 ph10 859 During matching, when PCRE reaches a callout point, the external function is
2828     called. It is provided with the number of the callout, the position in the
2829     pattern, and, optionally, one item of data originally supplied by the caller of
2830     the matching function. The callout function may cause matching to proceed, to
2831 ph10 1395 backtrack, or to fail altogether.
2832     .P
2833     By default, PCRE implements a number of optimizations at compile time and
2834     matching time, and one side-effect is that sometimes callouts are skipped. If
2835     you need all possible callouts to happen, you need to set options that disable
2836     the relevant optimizations. More details, and a complete description of the
2837     interface to the callout function, are given in the
2838 nigel 63 .\" HREF
2839 nigel 75 \fBpcrecallout\fP
2840 nigel 63 .\"
2841     documentation.
2842 nigel 93 .
2843     .
2844 ph10 510 .\" HTML <a name="backtrackcontrol"></a>
2846 ph10 210 .rs
2847     .sp
2848 ph10 211 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2849 ph10 1297 are still described in the Perl documentation as "experimental and subject to
2850     change or removal in a future version of Perl". It goes on to say: "Their usage
2851     in production code should be noted to avoid problems during upgrades." The same
2852 ph10 210 remarks apply to the PCRE features described in this section.
2853     .P
2854 ph10 211 The new verbs make use of what was previously invalid syntax: an opening
2855 ph10 510 parenthesis followed by an asterisk. They are generally of the form
2856 ph10 1335 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
2857 ph10 1302 differently depending on whether or not a name is present. A name is any
2858     sequence of characters that does not include a closing parenthesis. The maximum
2859     length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
2860     libraries. If the name is empty, that is, if the closing parenthesis
2861     immediately follows the colon, the effect is as if the colon were not there.
2862     Any number of these verbs may occur in a pattern.
2863 ph10 1297 .P
2864     Since these verbs are specifically related to backtracking, most of them can be
2865     used only when the pattern is to be matched using one of the traditional
2866     matching functions, because these use a backtracking algorithm. With the
2867 ph10 1335 exception of (*FAIL), which behaves like a failing negative assertion, the
2868 ph10 1297 backtracking control verbs cause an error if encountered by a DFA matching
2869     function.
2870     .P
2871 ph10 1335 The behaviour of these verbs in
2872 ph10 1298 .\" HTML <a href="#btrepeat">
2873     .\" </a>
2874 ph10 1335 repeated groups,
2875 ph10 1298 .\"
2876     .\" HTML <a href="#btassert">
2877     .\" </a>
2878 ph10 1335 assertions,
2879 ph10 1298 .\"
2880 ph10 1335 and in
2881 ph10 1298 .\" HTML <a href="#btsub">
2882     .\" </a>
2883     subpatterns called as subroutines
2884     .\"
2885     (whether or not recursively) is documented below.
2886 ph10 930 .
2887     .
2888     .\" HTML <a name="nooptimize"></a>
2889     .SS "Optimizations that affect backtracking verbs"
2890     .rs
2891     .sp
2892 ph10 512 PCRE contains some optimizations that are used to speed up matching by running
2893     some checks at the start of each match attempt. For example, it may know the
2894     minimum length of matching subject, or that a particular character must be
2895 ph10 1297 present. When one of these optimizations bypasses the running of a match, any
2896 ph10 512 included backtracking verbs will not, of course, be processed. You can suppress
2897     the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2898 ph10 577 when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2899 ph10 930 pattern with (*NO_START_OPT). There is more discussion of this option in the
2900     section entitled
2901     .\" HTML <a href="pcreapi.html#execoptions">
2902     .\" </a>
2903     "Option bits for \fBpcre_exec()\fP"
2904     .\"
2905     in the
2906     .\" HREF
2907     \fBpcreapi\fP
2908     .\"
2909 ph10 975 documentation.
2910 ph10 836 .P
2911     Experiments with Perl suggest that it too has similar optimizations, sometimes
2912     leading to anomalous results.
2913 ph10 210 .
2914 ph10 510 .
2915 ph10 210 .SS "Verbs that act immediately"
2916     .rs
2917     .sp
2918 ph10 512 The following verbs act as soon as they are encountered. They may not be
2919 ph10 510 followed by a name.
2920 ph10 210 .sp
2921     (*ACCEPT)
2922     .sp
2923     This verb causes the match to end successfully, skipping the remainder of the
2924 ph10 716 pattern. However, when it is inside a subpattern that is called as a
2925     subroutine, only that subpattern is ended successfully. Matching then continues
2926 ph10 1335 at the outer level. If (*ACCEPT) in triggered in a positive assertion, the
2927 ph10 1297 assertion succeeds; in a negative assertion, the assertion fails.
2928     .P
2929     If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For
2930     example:
2931 ph10 210 .sp
2932 ph10 447 A((?:A|B(*ACCEPT)|C)D)
2933 ph10 210 .sp
2934 ph10 461 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2935 ph10 447 the outer parentheses.
2936 ph10 210 .sp
2937     (*FAIL) or (*F)
2938     .sp
2939 ph10 716 This verb causes a matching failure, forcing backtracking to occur. It is
2940 ph10 210 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2941     probably useful only when combined with (?{}) or (??{}). Those are, of course,
2942     Perl features that are not present in PCRE. The nearest equivalent is the
2943     callout feature, as for example in this pattern:
2944     .sp
2945     a+(?C)(*FAIL)
2946     .sp
2947 ph10 211 A match with the string "aaaa" always fails, but the callout is taken before
2948     each backtrack happens (in this example, 10 times).
2949 ph10 210 .
2950 ph10 510 .
2951     .SS "Recording which path was taken"
2952     .rs
2953     .sp
2954 ph10 512 There is one verb whose main purpose is to track how a match was arrived at,
2955     though it also has a secondary use in conjunction with advancing the match
2956 ph10 510 starting point (see (*SKIP) below).
2957     .sp
2958     (*MARK:NAME) or (*:NAME)
2959     .sp
2960     A name is always required with this verb. There may be as many instances of
2961     (*MARK) as you like in a pattern, and their names do not have to be unique.
2962     .P
2963 ph10 1287 When a match succeeds, the name of the last-encountered (*MARK:NAME),
2964     (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
2965     caller as described in the section entitled
2966 ph10 510 .\" HTML <a href="pcreapi.html#extradata">
2967     .\" </a>
2968 ph10 859 "Extra data for \fBpcre_exec()\fP"
2969 ph10 510 .\"
2970 ph10 512 in the
2971 ph10 510 .\" HREF
2972     \fBpcreapi\fP
2973     .\"
2974 ph10 836 documentation. Here is an example of \fBpcretest\fP output, where the /K
2975     modifier requests the retrieval and outputting of (*MARK) data:
2976 ph10 510 .sp
2977 ph10 836 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2978     data> XY
2979 ph10 510 0: XY
2980     MK: A
2981     XZ
2982     0: XZ
2983     MK: B
2984     .sp
2985 ph10 512 The (*MARK) name is tagged with "MK:" in this output, and in this example it
2986     indicates which of the two alternatives matched. This is a more efficient way
2987 ph10 510 of obtaining this information than putting each alternative in its own
2988     capturing parentheses.
2989     .P
2990 ph10 1302 If a verb with a name is encountered in a positive assertion that is true, the
2991     name is recorded and passed back if it is the last-encountered. This does not
2992     happen for negative assertions or failing positive assertions.
2993 ph10 630 .P
2994 ph10 1287 After a partial match or a failed match, the last encountered name in the
2995     entire match process is returned. For example:
2996 ph10 510 .sp
2997 ph10 836 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2998     data> XP
2999 ph10 510 No match, mark = B
3000     .sp
3001 ph10 836 Note that in this unanchored example the mark is retained from the match
3002 ph10 930 attempt that started at the letter "X" in the subject. Subsequent match
3003     attempts starting at "P" and then with an empty string do not get as far as the
3004     (*MARK) item, but nevertheless do not reset it.
3005     .P
3006     If you are interested in (*MARK) values after failed matches, you should
3007 ph10 975 probably set the PCRE_NO_START_OPTIMIZE option
3008 ph10 930 .\" HTML <a href="#nooptimize">
3009     .\" </a>
3010 ph10 975 (see above)
3011 ph10 930 .\"
3012     to ensure that the match is always attempted.
3013 ph10 510 .
3014     .
3015 ph10 210 .SS "Verbs that act after backtracking"
3016     .rs
3017     .sp
3018 ph10 211 The following verbs do nothing when they are encountered. Matching continues
3019 ph10 510 with what follows, but if there is no subsequent match, causing a backtrack to
3020     the verb, a failure is forced. That is, backtracking cannot pass to the left of
3021 ph10 1335 the verb. However, when one of these verbs appears inside an atomic group or an
3022 ph10 1302 assertion that is true, its effect is confined to that group, because once the
3023     group has been matched, there is never any backtracking into it. In this
3024     situation, backtracking can "jump back" to the left of the entire atomic group
3025     or assertion. (Remember also, as stated above, that this localization also
3026     applies in subroutine calls.)
3027 ph10 510 .P
3028     These verbs differ in exactly what kind of failure occurs when backtracking
3029 ph10 1302 reaches them. The behaviour described below is what happens when the verb is
3030 ph10 1335 not in a subroutine or an assertion. Subsequent sections cover these special
3031 ph10 1302 cases.
3032 ph10 210 .sp
3033     (*COMMIT)
3034     .sp
3035 ph10 510 This verb, which may not be followed by a name, causes the whole match to fail
3036 ph10 1335 outright if there is a later matching failure that causes backtracking to reach
3037 ph10 1297 it. Even if the pattern is unanchored, no further attempts to find a match by
3038     advancing the starting point take place. If (*COMMIT) is the only backtracking
3039     verb that is encountered, once it has been passed \fBpcre_exec()\fP is
3040     committed to finding a match at the current starting point, or not at all. For
3041     example:
3042 ph10 210 .sp
3043     a+(*COMMIT)b
3044     .sp
3045 ph10 211 This matches "xxaab" but not "aacaab". It can be thought of as a kind of
3046 ph10 512 dynamic anchor, or "I've started, so I must finish." The name of the most
3047     recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
3048 ph10 510 match failure.
3049     .P
3050 ph10 1335 If there is more than one backtracking verb in a pattern, a different one that
3051     follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a
3052 ph10 1297 match does not always guarantee that a match must be at this starting point.
3053     .P
3054 ph10 512 Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
3055     unless PCRE's start-of-match optimizations are turned off, as shown in this
3056 ph10 1436 output from \fBpcretest\fP:
3057 ph10 210 .sp
3058 ph10 836 re> /(*COMMIT)abc/
3059     data> xyzabc
3060 ph10 510 0: abc
3061 ph10 1436 data> xyzabc\eY
3062 ph10 510 No match
3063 ph10 210 .sp
3064 ph10 1436 For this pattern, PCRE knows that any match must start with "a", so the
3065     optimization skips along the subject to "a" before applying the pattern to the
3066     first set of data. The match attempt then succeeds. In the second set of data,
3067     the escape sequence \eY is interpreted by the \fBpcretest\fP program. It causes
3068     the PCRE_NO_START_OPTIMIZE option to be set when \fBpcre_exec()\fP is called.
3069     This disables the optimization that skips along to the first character. The
3070     pattern is now applied starting at "x", and so the (*COMMIT) causes the match
3071     to fail without trying any other starting points.
3072 ph10 210 .sp
3073 ph10 510 (*PRUNE) or (*PRUNE:NAME)
3074     .sp
3075 ph10 512 This verb causes the match to fail at the current starting position in the
3076 ph10 1297 subject if there is a later matching failure that causes backtracking to reach
3077     it. If the pattern is unanchored, the normal "bumpalong" advance to the next
3078     starting character then happens. Backtracking can occur as usual to the left of
3079     (*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but
3080     if there is no match to the right, backtracking cannot cross (*PRUNE). In
3081     simple cases, the use of (*PRUNE) is just an alternative to an atomic group or
3082     possessive quantifier, but there are some uses of (*PRUNE) that cannot be
3083     expressed in any other way. In an anchored pattern (*PRUNE) has the same effect
3084     as (*COMMIT).
3085 ph10 1287 .P
3086 ph10 1335 The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE).
3087 ph10 1287 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3088     caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3089 ph10 510 .sp
3090 ph10 210 (*SKIP)
3091     .sp
3092 ph10 510 This verb, when given without a name, is like (*PRUNE), except that if the
3093     pattern is unanchored, the "bumpalong" advance is not to the next character,
3094     but to the position in the subject where (*SKIP) was encountered. (*SKIP)
3095     signifies that whatever text was matched leading up to it cannot be part of a
3096     successful match. Consider:
3097 ph10 210 .sp
3098     a+(*SKIP)b
3099     .sp
3100 ph10 211 If the subject is "aaaac...", after the first match attempt fails (starting at
3101 ph10 210 the first character in the string), the starting point skips on to start the
3102 ph10 211 next attempt at "c". Note that a possessive quantifer does not have the same
3103 ph10 456 effect as this example; although it would suppress backtracking during the
3104 ph10 210 first match attempt, the second attempt would start at the second character
3105     instead of skipping on to "c".
3106     .sp
3107 ph10 510 (*SKIP:NAME)
3108 ph10 211 .sp
3109 ph10 1335 When (*SKIP) has an associated name, its behaviour is modified. When it is
3110 ph10 1297 triggered, the previous path through the pattern is searched for the most
3111     recent (*MARK) that has the same name. If one is found, the "bumpalong" advance
3112     is to the subject position that corresponds to that (*MARK) instead of to where
3113     (*SKIP) was encountered. If no (*MARK) with a matching name is found, the
3114     (*SKIP) is ignored.
3115 ph10 1287 .P
3116 ph10 1335 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores
3117 ph10 1287 names that are set by (*PRUNE:NAME) or (*THEN:NAME).
3118 ph10 510 .sp
3119     (*THEN) or (*THEN:NAME)
3120     .sp
3121 ph10 1335 This verb causes a skip to the next innermost alternative when backtracking
3122 ph10 1297 reaches it. That is, it cancels any further backtracking within the current
3123     alternative. Its name comes from the observation that it can be used for a
3124     pattern-based if-then-else block:
3125 ph10 210 .sp
3126     ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
3127     .sp
3128 ph10 211 If the COND1 pattern matches, FOO is tried (and possibly further items after
3129 ph10 716 the end of the group if FOO succeeds); on failure, the matcher skips to the
3130 ph10 1302 second alternative and tries COND2, without backtracking into COND1. If that
3131     succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no
3132     more alternatives, so there is a backtrack to whatever came before the entire
3133     group. If (*THEN) is not inside an alternation, it acts like (*PRUNE).
3134 ph10 551 .P
3135 ph10 1335 The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
3136 ph10 1287 It is like (*MARK:NAME) in that the name is remembered for passing back to the
3137     caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3138     .P
3139 ph10 1297 A subpattern that does not contain a | character is just a part of the
3140     enclosing alternative; it is not a nested alternation with only one
3141 ph10 716 alternative. The effect of (*THEN) extends beyond such a subpattern to the
3142     enclosing alternative. Consider this pattern, where A, B, etc. are complex
3143     pattern fragments that do not contain any | characters at this level:
3144     .sp
3145     A (B(*THEN)C) | D
3146     .sp
3147 ph10 733 If A and B are matched, but there is a failure in C, matching does not
3148 ph10 716 backtrack into A; instead it moves to the next alternative, that is, D.
3149     However, if the subpattern containing (*THEN) is given an alternative, it
3150     behaves differently:
3151     .sp
3152     A (B(*THEN)C | (*FAIL)) | D
3153     .sp
3154     The effect of (*THEN) is now confined to the inner subpattern. After a failure
3155 ph10 733 in C, matching moves to (*FAIL), which causes the whole subpattern to fail
3156     because there are no more alternatives to try. In this case, matching does now
3157 ph10 716 backtrack into A.
3158 ph10 551 .P
3159 ph10 1297 Note that a conditional subpattern is not considered as having two
3160 ph10 733 alternatives, because only one is ever used. In other words, the | character in
3161 ph10 716 a conditional subpattern has a different meaning. Ignoring white space,
3162     consider:
3163 ph10 551 .sp
3164 ph10 716 ^.*? (?(?=a) a | b(*THEN)c )
3165     .sp
3166 ph10 733 If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
3167     it initially matches zero characters. The condition (?=a) then fails, the
3168 ph10 716 character "b" is matched, but "c" is not. At this point, matching does not
3169     backtrack to .*? as might perhaps be expected from the presence of the |
3170     character. The conditional subpattern is part of the single alternative that
3171 ph10 733 comprises the whole pattern, and so the match fails. (If there was a backtrack
3172 ph10 716 into .*?, allowing it to match "b", the match would succeed.)
3173     .P
3174     The verbs just described provide four different "strengths" of control when
3175     subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
3176     next alternative. (*PRUNE) comes next, failing the match at the current
3177     starting position, but allowing an advance to the next character (for an
3178     unanchored pattern). (*SKIP) is similar, except that the advance may be more
3179     than one character. (*COMMIT) is the strongest, causing the entire match to
3180     fail.
3181 ph10 1297 .
3182     .
3183     .SS "More than one backtracking verb"
3184     .rs
3185 ph10 716 .sp
3186 ph10 1297 If more than one backtracking verb is present in a pattern, the one that is
3187     backtracked onto first acts. For example, consider this pattern, where A, B,
3188     etc. are complex pattern fragments:
3189     .sp
3190 ph10 1293 (A(*COMMIT)B(*THEN)C|ABD)
3191 ph10 551 .sp
3192 ph10 1335 If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to
3193     fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes
3194 ph10 1293 the next alternative (ABD) to be tried. This behaviour is consistent, but is
3195 ph10 1297 not always the same as Perl's. It means that if two or more backtracking verbs
3196     appear in succession, all the the last of them has no effect. Consider this
3197     example:
3198     .sp
3199     ...(*COMMIT)(*PRUNE)...
3200     .sp
3201 ph10 1341 If there is a matching failure to the right, backtracking onto (*PRUNE) causes
3202 ph10 1335 it to be triggered, and its action is taken. There can never be a backtrack
3203     onto (*COMMIT).
3204 ph10 210 .
3205 ph10 551 .
3206 ph10 1298 .\" HTML <a name="btrepeat"></a>
3207 ph10 1297 .SS "Backtracking verbs in repeated groups"
3208     .rs
3209     .sp
3210 ph10 1335 PCRE differs from Perl in its handling of backtracking verbs in repeated
3211 ph10 1297 groups. For example, consider:
3212     .sp
3213     /(a(*COMMIT)b)+ac/
3214     .sp
3215 ph10 1335 If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in
3216     the second repeat of the group acts.
3217 ph10 1297 .
3218     .
3219 ph10 1298 .\" HTML <a name="btassert"></a>
3220     .SS "Backtracking verbs in assertions"
3221     .rs
3222     .sp
3223     (*FAIL) in an assertion has its normal effect: it forces an immediate backtrack.
3224     .P
3225 ph10 1335 (*ACCEPT) in a positive assertion causes the assertion to succeed without any
3226     further processing. In a negative assertion, (*ACCEPT) causes the assertion to
3227 ph10 1298 fail without any further processing.
3228     .P
3229 ph10 1302 The other backtracking verbs are not treated specially if they appear in a
3230     positive assertion. In particular, (*THEN) skips to the next alternative in the
3231 ph10 1298 innermost enclosing group that has alternations, whether or not this is within
3232     the assertion.
3233 ph10 1302 .P
3234     Negative assertions are, however, different, in order to ensure that changing a
3235     positive assertion into a negative assertion changes its result. Backtracking
3236 ph10 1335 into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true,
3237     without considering any further alternative branches in the assertion.
3238 ph10 1302 Backtracking into (*THEN) causes it to skip to the next enclosing alternative
3239 ph10 1335 within the assertion (the normal behaviour), but if the assertion does not have
3240 ph10 1302 such an alternative, (*THEN) behaves like (*PRUNE).
3241 ph10 1298 .
3242     .
3243     .\" HTML <a name="btsub"></a>
3244     .SS "Backtracking verbs in subroutines"
3245     .rs
3246     .sp
3247 ph10 1335 These behaviours occur whether or not the subpattern is called recursively.
3248 ph10 1298 Perl's treatment of subroutines is different in some cases.
3249     .P
3250     (*FAIL) in a subpattern called as a subroutine has its normal effect: it forces
3251     an immediate backtrack.
3252     .P
3253 ph10 1335 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to
3254     succeed without any further processing. Matching then continues after the
3255 ph10 1298 subroutine call.
3256     .P
3257     (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause
3258     the subroutine match to fail.
3259     .P
3260     (*THEN) skips to the next alternative in the innermost enclosing group within
3261 ph10 1335 the subpattern that has alternatives. If there is no such group within the
3262 ph10 1298 subpattern, (*THEN) causes the subroutine match to fail.
3263     .
3264     .
3265 nigel 93 .SH "SEE ALSO"
3266     .rs
3267     .sp
3268 ph10 461 \fBpcreapi\fP(3), \fBpcrecallout\fP(3), \fBpcrematching\fP(3),
3269 chpe 1055 \fBpcresyntax\fP(3), \fBpcre\fP(3), \fBpcre16(3)\fP, \fBpcre32(3)\fP.
3270 ph10 99 .
3271     .
3272     .SH AUTHOR
3273     .rs
3274     .sp
3275     .nf
3276     Philip Hazel
3277     University Computing Service
3278     Cambridge CB2 3QH, England.
3279     .fi
3280     .
3281     .
3282     .SH REVISION
3283     .rs
3284     .sp
3285     .nf
3286 ph10 1436 Last updated: 08 January 2014
3287     Copyright (c) 1997-2014 University of Cambridge.
3288 ph10 99 .fi


Name Value
svn:eol-style native
svn:keywords "Author Date Id Revision Url"

ViewVC Help
Powered by ViewVC 1.1.12