ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 836 - (hide annotations) (download) (as text)
Wed Dec 28 17:16:11 2011 UTC (3 years, 3 months ago) by ph10
File MIME type: text/html
File size: 123388 byte(s)
Merging all the changes from the pcre16 branch into the trunk.

1 nigel 63 <html>
2     <head>
3     <title>pcrepattern specification</title>
4     </head>
5     <body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6 nigel 75 <h1>pcrepattern man page</h1>
7     <p>
8     Return to the <a href="index.html">PCRE index page</a>.
9     </p>
10 ph10 111 <p>
11 nigel 75 This page is part of the PCRE HTML documentation. It was generated automatically
12     from the original man page. If there is any nonsense in it, please consult the
13     man page, in case the conversion went wrong.
14 ph10 111 <br>
15 nigel 63 <ul>
16     <li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a>
17 ph10 227 <li><a name="TOC2" href="#SEC2">NEWLINE CONVENTIONS</a>
18     <li><a name="TOC3" href="#SEC3">CHARACTERS AND METACHARACTERS</a>
19     <li><a name="TOC4" href="#SEC4">BACKSLASH</a>
20     <li><a name="TOC5" href="#SEC5">CIRCUMFLEX AND DOLLAR</a>
21 ph10 518 <li><a name="TOC6" href="#SEC6">FULL STOP (PERIOD, DOT) AND \N</a>
22 ph10 227 <li><a name="TOC7" href="#SEC7">MATCHING A SINGLE BYTE</a>
23     <li><a name="TOC8" href="#SEC8">SQUARE BRACKETS AND CHARACTER CLASSES</a>
24     <li><a name="TOC9" href="#SEC9">POSIX CHARACTER CLASSES</a>
25     <li><a name="TOC10" href="#SEC10">VERTICAL BAR</a>
26     <li><a name="TOC11" href="#SEC11">INTERNAL OPTION SETTING</a>
27     <li><a name="TOC12" href="#SEC12">SUBPATTERNS</a>
28     <li><a name="TOC13" href="#SEC13">DUPLICATE SUBPATTERN NUMBERS</a>
29     <li><a name="TOC14" href="#SEC14">NAMED SUBPATTERNS</a>
30     <li><a name="TOC15" href="#SEC15">REPETITION</a>
32     <li><a name="TOC17" href="#SEC17">BACK REFERENCES</a>
33     <li><a name="TOC18" href="#SEC18">ASSERTIONS</a>
34     <li><a name="TOC19" href="#SEC19">CONDITIONAL SUBPATTERNS</a>
35     <li><a name="TOC20" href="#SEC20">COMMENTS</a>
36     <li><a name="TOC21" href="#SEC21">RECURSIVE PATTERNS</a>
37     <li><a name="TOC22" href="#SEC22">SUBPATTERNS AS SUBROUTINES</a>
38 ph10 345 <li><a name="TOC23" href="#SEC23">ONIGURUMA SUBROUTINE SYNTAX</a>
39     <li><a name="TOC24" href="#SEC24">CALLOUTS</a>
40     <li><a name="TOC25" href="#SEC25">BACKTRACKING CONTROL</a>
41     <li><a name="TOC26" href="#SEC26">SEE ALSO</a>
42     <li><a name="TOC27" href="#SEC27">AUTHOR</a>
43     <li><a name="TOC28" href="#SEC28">REVISION</a>
44 nigel 63 </ul>
45     <br><a name="SEC1" href="#TOC1">PCRE REGULAR EXPRESSION DETAILS</a><br>
46     <P>
47 ph10 208 The syntax and semantics of the regular expressions that are supported by PCRE
48     are described in detail below. There is a quick-reference syntax summary in the
49     <a href="pcresyntax.html"><b>pcresyntax</b></a>
50 ph10 345 page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
51     also supports some alternative regular expression syntax (which does not
52     conflict with the Perl syntax) in order to provide some compatibility with
53     regular expressions in Python, .NET, and Oniguruma.
54     </P>
55     <P>
56     Perl's regular expressions are described in its own documentation, and
57 ph10 208 regular expressions in general are covered in a number of books, some of which
58     have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
59     published by O'Reilly, covers regular expressions in great detail. This
60     description of PCRE's regular expressions is intended as reference material.
61 nigel 63 </P>
62     <P>
63 nigel 75 The original operation of PCRE was on strings of one-byte characters. However,
64 ph10 461 there is now also support for UTF-8 character strings. To use this,
65     PCRE must be built to include UTF-8 support, and you must call
66     <b>pcre_compile()</b> or <b>pcre_compile2()</b> with the PCRE_UTF8 option. There
67     is also a special sequence that can be given at the start of a pattern:
68 ph10 416 <pre>
69     (*UTF8)
70     </pre>
71     Starting a pattern with this sequence is equivalent to setting the PCRE_UTF8
72     option. This feature is not Perl-compatible. How setting UTF-8 mode affects
73     pattern matching is mentioned in several places below. There is also a summary
74     of UTF-8 features in the
75 ph10 678 <a href="pcreunicode.html"><b>pcreunicode</b></a>
76 nigel 63 page.
77     </P>
78     <P>
79 ph10 535 Another special sequence that may appear at the start of a pattern or in
80     combination with (*UTF8) is:
81     <pre>
82     (*UCP)
83     </pre>
84     This has the same effect as setting the PCRE_UCP option: it causes sequences
85     such as \d and \w to use Unicode properties to determine character types,
86     instead of recognizing only characters with codes less than 128 via a lookup
87     table.
88     </P>
89     <P>
90 ph10 579 If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
91     PCRE_NO_START_OPTIMIZE option either at compile or matching time. There are
92     also some more of these special sequences that are concerned with the handling
93     of newlines; they are described below.
94     </P>
95     <P>
96 nigel 77 The remainder of this document discusses the patterns that are supported by
97     PCRE when its main matching function, <b>pcre_exec()</b>, is used.
98     From release 6.0, PCRE offers a second matching function,
99     <b>pcre_dfa_exec()</b>, which matches using a different algorithm that is not
100 ph10 172 Perl-compatible. Some of the features discussed below are not available when
101     <b>pcre_dfa_exec()</b> is used. The advantages and disadvantages of the
102     alternative function, and how it differs from the normal function, are
103     discussed in the
104 nigel 77 <a href="pcrematching.html"><b>pcrematching</b></a>
105     page.
106 ph10 567 <a name="newlines"></a></P>
107 ph10 227 <br><a name="SEC2" href="#TOC1">NEWLINE CONVENTIONS</a><br>
108 nigel 77 <P>
109 ph10 227 PCRE supports five different conventions for indicating line breaks in
110     strings: a single CR (carriage return) character, a single LF (linefeed)
111     character, the two-character sequence CRLF, any of the three preceding, or any
112     Unicode newline sequence. The
113     <a href="pcreapi.html"><b>pcreapi</b></a>
114     page has
115     <a href="pcreapi.html#newlines">further discussion</a>
116     about newlines, and shows how to set the newline convention in the
117     <i>options</i> arguments for the compiling and matching functions.
118     </P>
119     <P>
120     It is also possible to specify a newline convention by starting a pattern
121     string with one of the following five sequences:
122     <pre>
123     (*CR) carriage return
124     (*LF) linefeed
125     (*CRLF) carriage return, followed by linefeed
126     (*ANYCRLF) any of the three above
127     (*ANY) all Unicode newline sequences
128     </pre>
129 ph10 461 These override the default and the options given to <b>pcre_compile()</b> or
130     <b>pcre_compile2()</b>. For example, on a Unix system where LF is the default
131     newline sequence, the pattern
132 ph10 227 <pre>
133     (*CR)a.b
134     </pre>
135     changes the convention to CR. That pattern matches "a\nb" because LF is no
136     longer a newline. Note that these special settings, which are not
137     Perl-compatible, are recognized only at the very start of a pattern, and that
138 ph10 231 they must be in upper case. If more than one of them is present, the last one
139     is used.
140 ph10 227 </P>
141 ph10 231 <P>
142 ph10 518 The newline convention affects the interpretation of the dot metacharacter when
143     PCRE_DOTALL is not set, and also the behaviour of \N. However, it does not
144     affect what the \R escape sequence matches. By default, this is any Unicode
145     newline sequence, for Perl compatibility. However, this can be changed; see the
146     description of \R in the section entitled
147 ph10 231 <a href="#newlineseq">"Newline sequences"</a>
148 ph10 247 below. A change of \R setting can be combined with a change of newline
149     convention.
150 ph10 231 </P>
151 ph10 227 <br><a name="SEC3" href="#TOC1">CHARACTERS AND METACHARACTERS</a><br>
152     <P>
153 nigel 63 A regular expression is a pattern that is matched against a subject string from
154     left to right. Most characters stand for themselves in a pattern, and match the
155     corresponding characters in the subject. As a trivial example, the pattern
156     <pre>
157     The quick brown fox
158 nigel 75 </pre>
159 nigel 77 matches a portion of a subject string that is identical to itself. When
160     caseless matching is specified (the PCRE_CASELESS option), letters are matched
161     independently of case. In UTF-8 mode, PCRE always understands the concept of
162     case for characters whose values are less than 128, so caseless matching is
163     always possible. For characters with higher values, the concept of case is
164     supported if PCRE is compiled with Unicode property support, but not otherwise.
165     If you want to use caseless matching for characters 128 and above, you must
166     ensure that PCRE is compiled with Unicode property support as well as with
167     UTF-8 support.
168     </P>
169     <P>
170     The power of regular expressions comes from the ability to include alternatives
171     and repetitions in the pattern. These are encoded in the pattern by the use of
172 nigel 75 <i>metacharacters</i>, which do not stand for themselves but instead are
173 nigel 63 interpreted in some special way.
174     </P>
175     <P>
176 nigel 75 There are two different sets of metacharacters: those that are recognized
177 nigel 63 anywhere in the pattern except within square brackets, and those that are
178 nigel 93 recognized within square brackets. Outside square brackets, the metacharacters
179     are as follows:
180 nigel 63 <pre>
181     \ general escape character with several uses
182     ^ assert start of string (or line, in multiline mode)
183     $ assert end of string (or line, in multiline mode)
184     . match any character except newline (by default)
185     [ start character class definition
186     | start of alternative branch
187     ( start subpattern
188     ) end subpattern
189     ? extends the meaning of (
190     also 0 or 1 quantifier
191     also quantifier minimizer
192     * 0 or more quantifier
193     + 1 or more quantifier
194     also "possessive quantifier"
195     { start min/max quantifier
196 nigel 75 </pre>
197 nigel 63 Part of a pattern that is in square brackets is called a "character class". In
198 nigel 75 a character class the only metacharacters are:
199 nigel 63 <pre>
200     \ general escape character
201     ^ negate the class, but only if the first character
202     - indicates character range
203 nigel 75 [ POSIX character class (only if followed by POSIX syntax)
204 nigel 63 ] terminates the character class
205 nigel 75 </pre>
206     The following sections describe the use of each of the metacharacters.
207 nigel 63 </P>
208 ph10 227 <br><a name="SEC4" href="#TOC1">BACKSLASH</a><br>
209 nigel 63 <P>
210     The backslash character has several uses. Firstly, if it is followed by a
211 ph10 579 character that is not a number or a letter, it takes away any special meaning
212     that character may have. This use of backslash as an escape character applies
213     both inside and outside character classes.
214 nigel 63 </P>
215     <P>
216     For example, if you want to match a * character, you write \* in the pattern.
217     This escaping action applies whether or not the following character would
218 nigel 75 otherwise be interpreted as a metacharacter, so it is always safe to precede a
219     non-alphanumeric with backslash to specify that it stands for itself. In
220 nigel 63 particular, if you want to match a backslash, you write \\.
221     </P>
222     <P>
223 ph10 579 In UTF-8 mode, only ASCII numbers and letters have any special meaning after a
224     backslash. All other characters (in particular, those whose codepoints are
225     greater than 127) are treated as literals.
226     </P>
227     <P>
228 nigel 63 If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the
229     pattern (other than in a character class) and characters between a # outside
230 nigel 91 a character class and the next newline are ignored. An escaping backslash can
231     be used to include a whitespace or # character as part of the pattern.
232 nigel 63 </P>
233     <P>
234     If you want to remove the special meaning from a sequence of characters, you
235     can do so by putting them between \Q and \E. This is different from Perl in
236     that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in
237     Perl, $ and @ cause variable interpolation. Note the following examples:
238     <pre>
239     Pattern PCRE matches Perl matches
240 nigel 75
241     \Qabc$xyz\E abc$xyz abc followed by the contents of $xyz
242 nigel 63 \Qabc\$xyz\E abc\$xyz abc\$xyz
243     \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
244 nigel 75 </pre>
245 nigel 63 The \Q...\E sequence is recognized both inside and outside character classes.
246 ph10 654 An isolated \E that is not preceded by \Q is ignored. If \Q is not followed
247     by \E later in the pattern, the literal interpretation continues to the end of
248     the pattern (that is, \E is assumed at the end). If the isolated \Q is inside
249     a character class, this causes an error, because the character class is not
250     terminated.
251 nigel 75 <a name="digitsafterbackslash"></a></P>
252     <br><b>
253     Non-printing characters
254     </b><br>
255 nigel 63 <P>
256     A second use of backslash provides a way of encoding non-printing characters
257     in patterns in a visible manner. There is no restriction on the appearance of
258     non-printing characters, apart from the binary zero that terminates a pattern,
259 ph10 461 but when a pattern is being prepared by text editing, it is often easier to use
260     one of the following escape sequences than the binary character it represents:
261 nigel 63 <pre>
262     \a alarm, that is, the BEL character (hex 07)
263 ph10 579 \cx "control-x", where x is any ASCII character
264 nigel 63 \e escape (hex 1B)
265     \f formfeed (hex 0C)
266 ph10 227 \n linefeed (hex 0A)
267 nigel 63 \r carriage return (hex 0D)
268     \t tab (hex 09)
269 ph10 488 \ddd character with octal code ddd, or back reference
270 nigel 63 \xhh character with hex code hh
271 ph10 836 \x{hhh..} character with hex code hhh.. (non-JavaScript mode)
272     \uhhhh character with hex code hhhh (JavaScript mode only)
273 nigel 75 </pre>
274 nigel 63 The precise effect of \cx is as follows: if x is a lower case letter, it
275     is converted to upper case. Then bit 6 of the character (hex 40) is inverted.
276 ph10 579 Thus \cz becomes hex 1A (z is 7A), but \c{ becomes hex 3B ({ is 7B), while
277     \c; becomes hex 7B (; is 3B). If the byte following \c has a value greater
278     than 127, a compile-time error occurs. This locks out non-ASCII characters in
279     both byte mode and UTF-8 mode. (When PCRE is compiled in EBCDIC mode, all byte
280     values are valid. A lower case letter is converted to upper case, and then the
281     0xc0 bits are flipped.)
282 nigel 63 </P>
283     <P>
284 ph10 836 By default, after \x, from zero to two hexadecimal digits are read (letters
285     can be in upper or lower case). Any number of hexadecimal digits may appear
286     between \x{ and }, but the value of the character code must be less than 256
287     in non-UTF-8 mode, and less than 2**31 in UTF-8 mode. That is, the maximum
288     value in hexadecimal is 7FFFFFFF. Note that this is bigger than the largest
289     Unicode code point, which is 10FFFF.
290 nigel 63 </P>
291     <P>
292 ph10 211 If characters other than hexadecimal digits appear between \x{ and }, or if
293     there is no terminating }, this form of escape is not recognized. Instead, the
294     initial \x will be interpreted as a basic hexadecimal escape, with no
295     following digits, giving a character whose value is zero.
296     </P>
297     <P>
298 ph10 836 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x is
299     as just described only when it is followed by two hexadecimal digits.
300     Otherwise, it matches a literal "x" character. In JavaScript mode, support for
301     code points greater than 256 is provided by \u, which must be followed by
302     four hexadecimal digits; otherwise it matches a literal "u" character.
303     </P>
304     <P>
305 nigel 63 Characters whose value is less than 256 can be defined by either of the two
306 ph10 836 syntaxes for \x (or by \u in JavaScript mode). There is no difference in the
307     way they are handled. For example, \xdc is exactly the same as \x{dc} (or
308     \u00dc in JavaScript mode).
309 nigel 63 </P>
310     <P>
311 nigel 91 After \0 up to two further octal digits are read. If there are fewer than two
312     digits, just those that are present are used. Thus the sequence \0\x\07
313     specifies two binary zeros followed by a BEL character (code value 7). Make
314     sure you supply two digits after the initial zero if the pattern character that
315     follows is itself an octal digit.
316 nigel 63 </P>
317     <P>
318     The handling of a backslash followed by a digit other than 0 is complicated.
319     Outside a character class, PCRE reads it and any following digits as a decimal
320     number. If the number is less than 10, or if there have been at least that many
321     previous capturing left parentheses in the expression, the entire sequence is
322     taken as a <i>back reference</i>. A description of how this works is given
323 nigel 75 <a href="#backreferences">later,</a>
324     following the discussion of
325     <a href="#subpattern">parenthesized subpatterns.</a>
326 nigel 63 </P>
327     <P>
328     Inside a character class, or if the decimal number is greater than 9 and there
329     have not been that many capturing subpatterns, PCRE re-reads up to three octal
330 nigel 93 digits following the backslash, and uses them to generate a data character. Any
331 nigel 91 subsequent digits stand for themselves. In non-UTF-8 mode, the value of a
332     character specified in octal must be less than \400. In UTF-8 mode, values up
333     to \777 are permitted. For example:
334 nigel 63 <pre>
335     \040 is another way of writing a space
336 nigel 75 \40 is the same, provided there are fewer than 40 previous capturing subpatterns
337 nigel 63 \7 is always a back reference
338 nigel 75 \11 might be a back reference, or another way of writing a tab
339 nigel 63 \011 is always a tab
340     \0113 is a tab followed by the character "3"
341 nigel 75 \113 might be a back reference, otherwise the character with octal code 113
342     \377 might be a back reference, otherwise the byte consisting entirely of 1 bits
343     \81 is either a back reference, or a binary zero followed by the two characters "8" and "1"
344     </pre>
345 nigel 63 Note that octal values of 100 or greater must not be introduced by a leading
346     zero, because no more than three octal digits are ever read.
347     </P>
348     <P>
349 nigel 91 All the sequences that define a single character value can be used both inside
350 ph10 836 and outside character classes. In addition, inside a character class, \b is
351     interpreted as the backspace character (hex 08).
352 nigel 63 </P>
353 ph10 836 <P>
354     \N is not allowed in a character class. \B, \R, and \X are not special
355     inside a character class. Like other unrecognized escape sequences, they are
356     treated as the literal characters "B", "R", and "X" by default, but cause an
357     error if the PCRE_EXTRA option is set. Outside a character class, these
358     sequences have different meanings.
359     </P>
360 nigel 75 <br><b>
361 ph10 836 Unsupported escape sequences
362     </b><br>
363     <P>
364     In Perl, the sequences \l, \L, \u, and \U are recognized by its string
365     handler and used to modify the case of following characters. By default, PCRE
366     does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
367     option is set, \U matches a "U" character, and \u can be used to define a
368     character by code point, as described in the previous section.
369     </P>
370     <br><b>
371 nigel 93 Absolute and relative back references
372     </b><br>
373     <P>
374 ph10 208 The sequence \g followed by an unsigned or a negative number, optionally
375     enclosed in braces, is an absolute or relative back reference. A named back
376     reference can be coded as \g{name}. Back references are discussed
377 nigel 93 <a href="#backreferences">later,</a>
378     following the discussion of
379     <a href="#subpattern">parenthesized subpatterns.</a>
380     </P>
381     <br><b>
382 ph10 345 Absolute and relative subroutine calls
383     </b><br>
384     <P>
385     For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
386     a number enclosed either in angle brackets or single quotes, is an alternative
387     syntax for referencing a subpattern as a "subroutine". Details are discussed
388     <a href="#onigurumasubroutines">later.</a>
389     Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
390 ph10 461 synonymous. The former is a back reference; the latter is a
391 ph10 454 <a href="#subpatternsassubroutines">subroutine</a>
392     call.
393 ph10 518 <a name="genericchartypes"></a></P>
394 ph10 345 <br><b>
395 nigel 75 Generic character types
396     </b><br>
397 nigel 63 <P>
398 ph10 518 Another use of backslash is for specifying generic character types:
399 nigel 63 <pre>
400     \d any decimal digit
401     \D any character that is not a decimal digit
402 ph10 182 \h any horizontal whitespace character
403     \H any character that is not a horizontal whitespace character
404 nigel 63 \s any whitespace character
405     \S any character that is not a whitespace character
406 ph10 182 \v any vertical whitespace character
407     \V any character that is not a vertical whitespace character
408 nigel 63 \w any "word" character
409     \W any "non-word" character
410 nigel 75 </pre>
411 ph10 535 There is also the single sequence \N, which matches a non-newline character.
412     This is the same as
413 ph10 518 <a href="#fullstopdot">the "." metacharacter</a>
414 ph10 836 when PCRE_DOTALL is not set. Perl also uses \N to match characters by name;
415     PCRE does not support this.
416 nigel 63 </P>
417     <P>
418 ph10 518 Each pair of lower and upper case escape sequences partitions the complete set
419     of characters into two disjoint sets. Any given character matches one, and only
420 ph10 535 one, of each pair. The sequences can appear both inside and outside character
421 nigel 75 classes. They each match one character of the appropriate type. If the current
422 ph10 535 matching point is at the end of the subject string, all of them fail, because
423 nigel 75 there is no character to match.
424 nigel 63 </P>
425     <P>
426     For compatibility with Perl, \s does not match the VT character (code 11).
427     This makes it different from the the POSIX "space" class. The \s characters
428 ph10 182 are HT (9), LF (10), FF (12), CR (13), and space (32). If "use locale;" is
429 nigel 91 included in a Perl script, \s may match the VT character. In PCRE, it never
430 ph10 182 does.
431 nigel 63 </P>
432     <P>
433 ph10 535 A "word" character is an underscore or any character that is a letter or digit.
434     By default, the definition of letters and digits is controlled by PCRE's
435     low-valued character tables, and may vary if locale-specific matching is taking
436     place (see
437     <a href="pcreapi.html#localesupport">"Locale support"</a>
438     in the
439     <a href="pcreapi.html"><b>pcreapi</b></a>
440     page). For example, in a French locale such as "fr_FR" in Unix-like systems,
441     or "french" in Windows, some character codes greater than 128 are used for
442     accented letters, and these are then matched by \w. The use of locales with
443     Unicode is discouraged.
444 ph10 182 </P>
445     <P>
446 ph10 535 By default, in UTF-8 mode, characters with values greater than 128 never match
447     \d, \s, or \w, and always match \D, \S, and \W. These sequences retain
448     their original meanings from before UTF-8 support was available, mainly for
449     efficiency reasons. However, if PCRE is compiled with Unicode property support,
450     and the PCRE_UCP option is set, the behaviour is changed so that Unicode
451     properties are used to determine character types, as follows:
452     <pre>
453     \d any character that \p{Nd} matches (decimal digit)
454     \s any character that \p{Z} matches, plus HT, LF, FF, CR
455     \w any character that \p{L} or \p{N} matches, plus underscore
456     </pre>
457     The upper case escapes match the inverse sets of characters. Note that \d
458     matches only decimal digits, whereas \w matches any Unicode digit, as well as
459     any Unicode letter, and underscore. Note also that PCRE_UCP affects \b, and
460     \B because they are defined in terms of \w and \W. Matching these sequences
461     is noticeably slower when PCRE_UCP is set.
462     </P>
463     <P>
464 ph10 579 The sequences \h, \H, \v, and \V are features that were added to Perl at
465 ph10 572 release 5.10. In contrast to the other sequences, which match only ASCII
466     characters by default, these always match certain high-valued codepoints in
467     UTF-8 mode, whether or not PCRE_UCP is set. The horizontal space characters
468     are:
469 ph10 182 <pre>
470     U+0009 Horizontal tab
471     U+0020 Space
472     U+00A0 Non-break space
473     U+1680 Ogham space mark
474     U+180E Mongolian vowel separator
475     U+2000 En quad
476     U+2001 Em quad
477     U+2002 En space
478     U+2003 Em space
479     U+2004 Three-per-em space
480     U+2005 Four-per-em space
481     U+2006 Six-per-em space
482     U+2007 Figure space
483     U+2008 Punctuation space
484     U+2009 Thin space
485     U+200A Hair space
486     U+202F Narrow no-break space
487     U+205F Medium mathematical space
488     U+3000 Ideographic space
489     </pre>
490     The vertical space characters are:
491     <pre>
492     U+000A Linefeed
493     U+000B Vertical tab
494     U+000C Formfeed
495     U+000D Carriage return
496     U+0085 Next line
497     U+2028 Line separator
498     U+2029 Paragraph separator
499 ph10 535 <a name="newlineseq"></a></PRE>
500 ph10 182 </P>
501 nigel 93 <br><b>
502     Newline sequences
503     </b><br>
504     <P>
505 ph10 231 Outside a character class, by default, the escape sequence \R matches any
506 ph10 572 Unicode newline sequence. In non-UTF-8 mode \R is equivalent to the following:
507 nigel 93 <pre>
508     (?&#62;\r\n|\n|\x0b|\f|\r|\x85)
509     </pre>
510     This is an example of an "atomic group", details of which are given
511     <a href="#atomicgroup">below.</a>
512     This particular group matches either the two-character sequence CR followed by
513     LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
514     U+000B), FF (formfeed, U+000C), CR (carriage return, U+000D), or NEL (next
515     line, U+0085). The two-character sequence is treated as a single unit that
516     cannot be split.
517     </P>
518     <P>
519     In UTF-8 mode, two additional characters whose codepoints are greater than 255
520     are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
521     Unicode character property support is not needed for these characters to be
522     recognized.
523     </P>
524     <P>
525 ph10 231 It is possible to restrict \R to match only CR, LF, or CRLF (instead of the
526     complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
527 ph10 247 either at compile time or when the pattern is matched. (BSR is an abbrevation
528     for "backslash R".) This can be made the default when PCRE is built; if this is
529     the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
530     It is also possible to specify these settings by starting a pattern string with
531     one of the following sequences:
532 ph10 231 <pre>
533     (*BSR_ANYCRLF) CR, LF, or CRLF only
534     (*BSR_UNICODE) any Unicode newline sequence
535     </pre>
536 ph10 461 These override the default and the options given to <b>pcre_compile()</b> or
537     <b>pcre_compile2()</b>, but they can be overridden by options given to
538     <b>pcre_exec()</b> or <b>pcre_dfa_exec()</b>. Note that these special settings,
539     which are not Perl-compatible, are recognized only at the very start of a
540     pattern, and that they must be in upper case. If more than one of them is
541     present, the last one is used. They can be combined with a change of newline
542 ph10 535 convention; for example, a pattern can start with:
543 ph10 247 <pre>
544     (*ANY)(*BSR_ANYCRLF)
545     </pre>
546 ph10 535 They can also be combined with the (*UTF8) or (*UCP) special sequences. Inside
547     a character class, \R is treated as an unrecognized escape sequence, and so
548     matches the letter "R" by default, but causes an error if PCRE_EXTRA is set.
549 nigel 75 <a name="uniextseq"></a></P>
550     <br><b>
551     Unicode character properties
552     </b><br>
553     <P>
554     When PCRE is built with Unicode character property support, three additional
555 ph10 185 escape sequences that match characters with specific properties are available.
556     When not in UTF-8 mode, these sequences are of course limited to testing
557     characters whose codepoints are less than 256, but they do work in this mode.
558     The extra escape sequences are:
559 nigel 75 <pre>
560 nigel 87 \p{<i>xx</i>} a character with the <i>xx</i> property
561     \P{<i>xx</i>} a character without the <i>xx</i> property
562     \X an extended Unicode sequence
563 nigel 75 </pre>
564 nigel 87 The property names represented by <i>xx</i> above are limited to the Unicode
565 ph10 518 script names, the general category properties, "Any", which matches any
566     character (including newline), and some special PCRE properties (described
567 ph10 535 in the
568 ph10 518 <a href="#extraprops">next section).</a>
569     Other Perl properties such as "InMusicalSymbols" are not currently supported by
570     PCRE. Note that \P{Any} does not match any characters, so always causes a
571     match failure.
572 nigel 63 </P>
573     <P>
574 nigel 87 Sets of Unicode characters are defined as belonging to certain scripts. A
575     character from one of these sets can be matched using a script name. For
576     example:
577 nigel 75 <pre>
578 nigel 87 \p{Greek}
579     \P{Han}
580     </pre>
581     Those that are not part of an identified script are lumped together as
582     "Common". The current list of scripts is:
583     </P>
584     <P>
585     Arabic,
586     Armenian,
587 ph10 507 Avestan,
588 nigel 93 Balinese,
589 ph10 507 Bamum,
590 nigel 87 Bengali,
591     Bopomofo,
592     Braille,
593     Buginese,
594     Buhid,
595     Canadian_Aboriginal,
596 ph10 507 Carian,
597     Cham,
598 nigel 87 Cherokee,
599     Common,
600     Coptic,
601 nigel 93 Cuneiform,
602 nigel 87 Cypriot,
603     Cyrillic,
604     Deseret,
605     Devanagari,
606 ph10 507 Egyptian_Hieroglyphs,
607 nigel 87 Ethiopic,
608     Georgian,
609     Glagolitic,
610     Gothic,
611     Greek,
612     Gujarati,
613     Gurmukhi,
614     Han,
615     Hangul,
616     Hanunoo,
617     Hebrew,
618     Hiragana,
619 ph10 507 Imperial_Aramaic,
620 nigel 87 Inherited,
621 ph10 507 Inscriptional_Pahlavi,
622     Inscriptional_Parthian,
623     Javanese,
624     Kaithi,
625 nigel 87 Kannada,
626     Katakana,
627 ph10 507 Kayah_Li,
628 nigel 87 Kharoshthi,
629     Khmer,
630     Lao,
631     Latin,
632 ph10 507 Lepcha,
633 nigel 87 Limbu,
634     Linear_B,
635 ph10 507 Lisu,
636     Lycian,
637     Lydian,
638 nigel 87 Malayalam,
639 ph10 507 Meetei_Mayek,
640 nigel 87 Mongolian,
641     Myanmar,
642     New_Tai_Lue,
643 nigel 93 Nko,
644 nigel 87 Ogham,
645     Old_Italic,
646     Old_Persian,
647 ph10 507 Old_South_Arabian,
648     Old_Turkic,
649     Ol_Chiki,
650 nigel 87 Oriya,
651     Osmanya,
652 nigel 93 Phags_Pa,
653     Phoenician,
654 ph10 507 Rejang,
655 nigel 87 Runic,
656 ph10 507 Samaritan,
657     Saurashtra,
658 nigel 87 Shavian,
659     Sinhala,
660 ph10 507 Sundanese,
661 nigel 87 Syloti_Nagri,
662     Syriac,
663     Tagalog,
664     Tagbanwa,
665     Tai_Le,
666 ph10 507 Tai_Tham,
667     Tai_Viet,
668 nigel 87 Tamil,
669     Telugu,
670     Thaana,
671     Thai,
672     Tibetan,
673     Tifinagh,
674     Ugaritic,
675 ph10 507 Vai,
676 nigel 87 Yi.
677     </P>
678     <P>
679 ph10 518 Each character has exactly one Unicode general category property, specified by
680     a two-letter abbreviation. For compatibility with Perl, negation can be
681     specified by including a circumflex between the opening brace and the property
682     name. For example, \p{^Lu} is the same as \P{Lu}.
683 nigel 87 </P>
684     <P>
685     If only one letter is specified with \p or \P, it includes all the general
686     category properties that start with that letter. In this case, in the absence
687     of negation, the curly brackets in the escape sequence are optional; these two
688     examples have the same effect:
689     <pre>
690 nigel 75 \p{L}
691     \pL
692     </pre>
693 nigel 87 The following general category property codes are supported:
694 nigel 75 <pre>
695     C Other
696     Cc Control
697     Cf Format
698     Cn Unassigned
699     Co Private use
700     Cs Surrogate
702     L Letter
703     Ll Lower case letter
704     Lm Modifier letter
705     Lo Other letter
706     Lt Title case letter
707     Lu Upper case letter
709     M Mark
710     Mc Spacing mark
711     Me Enclosing mark
712     Mn Non-spacing mark
714     N Number
715     Nd Decimal number
716     Nl Letter number
717     No Other number
719     P Punctuation
720     Pc Connector punctuation
721     Pd Dash punctuation
722     Pe Close punctuation
723     Pf Final punctuation
724     Pi Initial punctuation
725     Po Other punctuation
726     Ps Open punctuation
728     S Symbol
729     Sc Currency symbol
730     Sk Modifier symbol
731     Sm Mathematical symbol
732     So Other symbol
734     Z Separator
735     Zl Line separator
736     Zp Paragraph separator
737     Zs Space separator
738     </pre>
739 nigel 87 The special property L& is also supported: it matches a character that has
740     the Lu, Ll, or Lt property, in other words, a letter that is not classified as
741     a modifier or "other".
742 nigel 75 </P>
743     <P>
744 ph10 211 The Cs (Surrogate) property applies only to characters in the range U+D800 to
745     U+DFFF. Such characters are not valid in UTF-8 strings (see RFC 3629) and so
746     cannot be tested by PCRE, unless UTF-8 validity checking has been turned off
747     (see the discussion of PCRE_NO_UTF8_CHECK in the
748     <a href="pcreapi.html"><b>pcreapi</b></a>
749 ph10 453 page). Perl does not support the Cs property.
750 ph10 211 </P>
751     <P>
752 ph10 453 The long synonyms for property names that Perl supports (such as \p{Letter})
753 nigel 91 are not supported by PCRE, nor is it permitted to prefix any of these
754 nigel 87 properties with "Is".
755     </P>
756     <P>
757     No character that is in the Unicode table has the Cn (unassigned) property.
758     Instead, this property is assumed for any code point that is not in the
759     Unicode table.
760     </P>
761     <P>
762 nigel 75 Specifying caseless matching does not affect these escape sequences. For
763     example, \p{Lu} always matches only upper case letters.
764     </P>
765     <P>
766     The \X escape matches any number of Unicode characters that form an extended
767     Unicode sequence. \X is equivalent to
768     <pre>
769     (?&#62;\PM\pM*)
770     </pre>
771     That is, it matches a character without the "mark" property, followed by zero
772     or more characters with the "mark" property, and treats the sequence as an
773     atomic group
774     <a href="#atomicgroup">(see below).</a>
775     Characters with the "mark" property are typically accents that affect the
776 ph10 185 preceding character. None of them have codepoints less than 256, so in
777     non-UTF-8 mode \X matches any one character.
778 nigel 75 </P>
779     <P>
780 ph10 654 Note that recent versions of Perl have changed \X to match what Unicode calls
781     an "extended grapheme cluster", which has a more complicated definition.
782     </P>
783     <P>
784 nigel 75 Matching characters by Unicode property is not fast, because PCRE has to search
785     a structure that contains data for over fifteen thousand characters. That is
786     why the traditional escape sequences such as \d and \w do not use Unicode
787 ph10 535 properties in PCRE by default, though you can make them do so by setting the
788     PCRE_UCP option for <b>pcre_compile()</b> or by starting the pattern with
789     (*UCP).
790 ph10 518 <a name="extraprops"></a></P>
791     <br><b>
792     PCRE's additional properties
793     </b><br>
794     <P>
795 ph10 535 As well as the standard Unicode properties described in the previous
796     section, PCRE supports four more that make it possible to convert traditional
797 ph10 518 escape sequences such as \w and \s and POSIX character classes to use Unicode
798 ph10 535 properties. PCRE uses these non-standard, non-Perl properties internally when
799     PCRE_UCP is set. They are:
800 ph10 518 <pre>
801     Xan Any alphanumeric character
802     Xps Any POSIX space character
803     Xsp Any Perl space character
804     Xwd Any Perl "word" character
805     </pre>
806 ph10 535 Xan matches characters that have either the L (letter) or the N (number)
807     property. Xps matches the characters tab, linefeed, vertical tab, formfeed, or
808 ph10 518 carriage return, and any other character that has the Z (separator) property.
809 ph10 535 Xsp is the same as Xps, except that vertical tab is excluded. Xwd matches the
810 ph10 518 same characters as Xan, plus underscore.
811 ph10 172 <a name="resetmatchstart"></a></P>
812     <br><b>
813     Resetting the match start
814     </b><br>
815     <P>
816 ph10 572 The escape sequence \K causes any previously matched characters not to be
817     included in the final matched sequence. For example, the pattern:
818 ph10 172 <pre>
819     foo\Kbar
820     </pre>
821     matches "foobar", but reports that it has matched "bar". This feature is
822     similar to a lookbehind assertion
823     <a href="#lookbehind">(described below).</a>
824     However, in this case, the part of the subject before the real match does not
825     have to be of fixed length, as lookbehind assertions do. The use of \K does
826     not interfere with the setting of
827     <a href="#subpattern">captured substrings.</a>
828     For example, when the pattern
829     <pre>
830     (foo)\Kbar
831     </pre>
832     matches "foobar", the first substring is still set to "foo".
833 ph10 507 </P>
834     <P>
835     Perl documents that the use of \K within assertions is "not well defined". In
836     PCRE, \K is acted upon when it occurs inside positive assertions, but is
837     ignored in negative assertions.
838 nigel 75 <a name="smallassertions"></a></P>
839     <br><b>
840     Simple assertions
841     </b><br>
842     <P>
843 nigel 93 The final use of backslash is for certain simple assertions. An assertion
844 nigel 63 specifies a condition that has to be met at a particular point in a match,
845     without consuming any characters from the subject string. The use of
846 nigel 75 subpatterns for more complicated assertions is described
847     <a href="#bigassertions">below.</a>
848 nigel 91 The backslashed assertions are:
849 nigel 63 <pre>
850     \b matches at a word boundary
851     \B matches when not at a word boundary
852 nigel 93 \A matches at the start of the subject
853     \Z matches at the end of the subject
854     also matches before a newline at the end of the subject
855     \z matches only at the end of the subject
856     \G matches at the first matching position in the subject
857 nigel 75 </pre>
858 ph10 518 Inside a character class, \b has a different meaning; it matches the backspace
859 ph10 535 character. If any other of these assertions appears in a character class, by
860 ph10 518 default it matches the corresponding literal character (for example, \B
861     matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
862     escape sequence" error is generated instead.
863 nigel 63 </P>
864     <P>
865     A word boundary is a position in the subject string where the current character
866     and the previous character do not both match \w or \W (i.e. one matches
867     \w and the other matches \W), or the start or end of the string if the
868 ph10 535 first or last character matches \w, respectively. In UTF-8 mode, the meanings
869     of \w and \W can be changed by setting the PCRE_UCP option. When this is
870     done, it also affects \b and \B. Neither PCRE nor Perl has a separate "start
871     of word" or "end of word" metasequence. However, whatever follows \b normally
872     determines which it is. For example, the fragment \ba matches "a" at the start
873     of a word.
874 nigel 63 </P>
875     <P>
876     The \A, \Z, and \z assertions differ from the traditional circumflex and
877 nigel 75 dollar (described in the next section) in that they only ever match at the very
878     start and end of the subject string, whatever options are set. Thus, they are
879     independent of multiline mode. These three assertions are not affected by the
880     PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
881     circumflex and dollar metacharacters. However, if the <i>startoffset</i>
882     argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start
883     at a point other than the beginning of the subject, \A can never match. The
884 nigel 91 difference between \Z and \z is that \Z matches before a newline at the end
885     of the string as well as at the very end, whereas \z matches only at the end.
886 nigel 63 </P>
887     <P>
888     The \G assertion is true only when the current matching position is at the
889     start point of the match, as specified by the <i>startoffset</i> argument of
890     <b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is
891     non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate
892     arguments, you can mimic Perl's /g option, and it is in this kind of
893     implementation where \G can be useful.
894     </P>
895     <P>
896     Note, however, that PCRE's interpretation of \G, as the start of the current
897     match, is subtly different from Perl's, which defines it as the end of the
898     previous match. In Perl, these can be different when the previously matched
899     string was empty. Because PCRE does just one match at a time, it cannot
900     reproduce this behaviour.
901     </P>
902     <P>
903     If all the alternatives of a pattern begin with \G, the expression is anchored
904     to the starting match position, and the "anchored" flag is set in the compiled
905     regular expression.
906     </P>
907 ph10 227 <br><a name="SEC5" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br>
908 nigel 63 <P>
909     Outside a character class, in the default matching mode, the circumflex
910 nigel 75 character is an assertion that is true only if the current matching point is
911 nigel 63 at the start of the subject string. If the <i>startoffset</i> argument of
912     <b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE
913     option is unset. Inside a character class, circumflex has an entirely different
914 nigel 75 meaning
915     <a href="#characterclass">(see below).</a>
916 nigel 63 </P>
917     <P>
918     Circumflex need not be the first character of the pattern if a number of
919     alternatives are involved, but it should be the first thing in each alternative
920     in which it appears if the pattern is ever to match that branch. If all
921     possible alternatives start with a circumflex, that is, if the pattern is
922     constrained to match only at the start of the subject, it is said to be an
923     "anchored" pattern. (There are also other constructs that can cause a pattern
924     to be anchored.)
925     </P>
926     <P>
927 nigel 75 A dollar character is an assertion that is true only if the current matching
928 nigel 63 point is at the end of the subject string, or immediately before a newline
929 nigel 91 at the end of the string (by default). Dollar need not be the last character of
930     the pattern if a number of alternatives are involved, but it should be the last
931     item in any branch in which it appears. Dollar has no special meaning in a
932     character class.
933 nigel 63 </P>
934     <P>
935     The meaning of dollar can be changed so that it matches only at the very end of
936     the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
937     does not affect the \Z assertion.
938     </P>
939     <P>
940     The meanings of the circumflex and dollar characters are changed if the
941 nigel 91 PCRE_MULTILINE option is set. When this is the case, a circumflex matches
942     immediately after internal newlines as well as at the start of the subject
943     string. It does not match after a newline that ends the string. A dollar
944     matches before any newlines in the string, as well as at the very end, when
945     PCRE_MULTILINE is set. When newline is specified as the two-character
946     sequence CRLF, isolated CR and LF characters do not indicate newlines.
947 nigel 63 </P>
948     <P>
949 nigel 91 For example, the pattern /^abc$/ matches the subject string "def\nabc" (where
950     \n represents a newline) in multiline mode, but not otherwise. Consequently,
951     patterns that are anchored in single line mode because all branches start with
952     ^ are not anchored in multiline mode, and a match for circumflex is possible
953     when the <i>startoffset</i> argument of <b>pcre_exec()</b> is non-zero. The
954     PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
955     </P>
956     <P>
957 nigel 63 Note that the sequences \A, \Z, and \z can be used to match the start and
958     end of the subject in both modes, and if all branches of a pattern start with
959 nigel 91 \A it is always anchored, whether or not PCRE_MULTILINE is set.
960 ph10 518 <a name="fullstopdot"></a></P>
961     <br><a name="SEC6" href="#TOC1">FULL STOP (PERIOD, DOT) AND \N</a><br>
962 nigel 63 <P>
963     Outside a character class, a dot in the pattern matches any one character in
964 nigel 91 the subject string except (by default) a character that signifies the end of a
965 nigel 93 line. In UTF-8 mode, the matched character may be more than one byte long.
966 nigel 63 </P>
967 nigel 91 <P>
968 nigel 93 When a line ending is defined as a single character, dot never matches that
969     character; when the two-character sequence CRLF is used, dot does not match CR
970     if it is immediately followed by LF, but otherwise it matches all characters
971     (including isolated CRs and LFs). When any Unicode line endings are being
972     recognized, dot does not match CR or LF or any of the other line ending
973     characters.
974     </P>
975     <P>
976 nigel 91 The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
977 nigel 93 option is set, a dot matches any one character, without exception. If the
978     two-character sequence CRLF is present in the subject string, it takes two dots
979     to match it.
980 nigel 91 </P>
981     <P>
982     The handling of dot is entirely independent of the handling of circumflex and
983     dollar, the only relationship being that they both involve newlines. Dot has no
984     special meaning in a character class.
985     </P>
986 ph10 518 <P>
987 ph10 579 The escape sequence \N behaves like a dot, except that it is not affected by
988 ph10 572 the PCRE_DOTALL option. In other words, it matches any character except one
989 ph10 836 that signifies the end of a line. Perl also uses \N to match characters by
990     name; PCRE does not support this.
991 ph10 518 </P>
992 ph10 227 <br><a name="SEC7" href="#TOC1">MATCHING A SINGLE BYTE</a><br>
993 nigel 63 <P>
994     Outside a character class, the escape sequence \C matches any one byte, both
995 ph10 738 in and out of UTF-8 mode. Unlike a dot, it always matches line-ending
996 nigel 93 characters. The feature is provided in Perl in order to match individual bytes
997 ph10 738 in UTF-8 mode, but it is unclear how it can usefully be used. Because \C
998     breaks up characters into individual bytes, matching one byte with \C in UTF-8
999     mode means that the rest of the string may start with a malformed UTF-8
1000     character. This has undefined results, because PCRE assumes that it is dealing
1001     with valid UTF-8 strings (and by default it checks this at the start of
1002     processing unless the PCRE_NO_UTF8_CHECK option is used).
1003 nigel 63 </P>
1004     <P>
1005 nigel 75 PCRE does not allow \C to appear in lookbehind assertions
1006 ph10 836 <a href="#lookbehind">(described below)</a>
1007     in UTF-8 mode, because this would make it impossible to calculate the length of
1008 nigel 75 the lookbehind.
1009 ph10 738 </P>
1010     <P>
1011     In general, the \C escape sequence is best avoided in UTF-8 mode. However, one
1012     way of using it that avoids the problem of malformed UTF-8 characters is to
1013     use a lookahead to check the length of the next character, as in this pattern
1014     (ignore white space and line breaks):
1015     <pre>
1016     (?| (?=[\x00-\x7f])(\C) |
1017     (?=[\x80-\x{7ff}])(\C)(\C) |
1018     (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
1019     (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))
1020     </pre>
1021     A group that starts with (?| resets the capturing parentheses numbers in each
1022     alternative (see
1023     <a href="#dupsubpatternnumber">"Duplicate Subpattern Numbers"</a>
1024     below). The assertions at the start of each branch check the next UTF-8
1025     character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1026     character's individual bytes are then captured by the appropriate number of
1027     groups.
1028 nigel 75 <a name="characterclass"></a></P>
1029 ph10 227 <br><a name="SEC8" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br>
1030 nigel 63 <P>
1031     An opening square bracket introduces a character class, terminated by a closing
1032 ph10 461 square bracket. A closing square bracket on its own is not special by default.
1033     However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1034     bracket causes a compile-time error. If a closing square bracket is required as
1035     a member of the class, it should be the first data character in the class
1036     (after an initial circumflex, if present) or escaped with a backslash.
1037 nigel 63 </P>
1038     <P>
1039     A character class matches a single character in the subject. In UTF-8 mode, the
1040 ph10 461 character may be more than one byte long. A matched character must be in the
1041     set of characters defined by the class, unless the first character in the class
1042 nigel 63 definition is a circumflex, in which case the subject character must not be in
1043     the set defined by the class. If a circumflex is actually required as a member
1044     of the class, ensure it is not the first character, or escape it with a
1045     backslash.
1046     </P>
1047     <P>
1048     For example, the character class [aeiou] matches any lower case vowel, while
1049     [^aeiou] matches any character that is not a lower case vowel. Note that a
1050 nigel 75 circumflex is just a convenient notation for specifying the characters that
1051     are in the class by enumerating those that are not. A class that starts with a
1052 ph10 461 circumflex is not an assertion; it still consumes a character from the subject
1053 nigel 75 string, and therefore it fails if the current pointer is at the end of the
1054     string.
1055 nigel 63 </P>
1056     <P>
1057     In UTF-8 mode, characters with values greater than 255 can be included in a
1058     class as a literal string of bytes, or by using the \x{ escaping mechanism.
1059     </P>
1060     <P>
1061     When caseless matching is set, any letters in a class represent both their
1062     upper case and lower case versions, so for example, a caseless [aeiou] matches
1063     "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1064 nigel 77 caseful version would. In UTF-8 mode, PCRE always understands the concept of
1065     case for characters whose values are less than 128, so caseless matching is
1066     always possible. For characters with higher values, the concept of case is
1067     supported if PCRE is compiled with Unicode property support, but not otherwise.
1068 ph10 461 If you want to use caseless matching in UTF8-mode for characters 128 and above,
1069     you must ensure that PCRE is compiled with Unicode property support as well as
1070     with UTF-8 support.
1071 nigel 63 </P>
1072     <P>
1073 nigel 93 Characters that might indicate line breaks are never treated in any special way
1074     when matching character classes, whatever line-ending sequence is in use, and
1075     whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1076     such as [^a] always matches one of these characters.
1077 nigel 63 </P>
1078     <P>
1079     The minus (hyphen) character can be used to specify a range of characters in a
1080     character class. For example, [d-m] matches any letter between d and m,
1081     inclusive. If a minus character is required in a class, it must be escaped with
1082     a backslash or appear in a position where it cannot be interpreted as
1083     indicating a range, typically as the first or last character in the class.
1084     </P>
1085     <P>
1086     It is not possible to have the literal character "]" as the end character of a
1087     range. A pattern such as [W-]46] is interpreted as a class of two characters
1088     ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1089     "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1090 nigel 75 the end of range, so [W-\]46] is interpreted as a class containing a range
1091     followed by two other characters. The octal or hexadecimal representation of
1092     "]" can also be used to end a range.
1093 nigel 63 </P>
1094     <P>
1095     Ranges operate in the collating sequence of character values. They can also be
1096     used for characters specified numerically, for example [\000-\037]. In UTF-8
1097     mode, ranges can include characters whose values are greater than 255, for
1098     example [\x{100}-\x{2ff}].
1099     </P>
1100     <P>
1101     If a range that includes letters is used when caseless matching is set, it
1102     matches the letters in either case. For example, [W-c] is equivalent to
1103 nigel 75 [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if character
1104 ph10 142 tables for a French locale are in use, [\xc8-\xcb] matches accented E
1105 nigel 75 characters in both cases. In UTF-8 mode, PCRE supports the concept of case for
1106     characters with values greater than 128 only when it is compiled with Unicode
1107     property support.
1108 nigel 63 </P>
1109     <P>
1110 ph10 579 The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v,
1111     \V, \w, and \W may appear in a character class, and add the characters that
1112     they match to the class. For example, [\dABCDEF] matches any hexadecimal
1113     digit. In UTF-8 mode, the PCRE_UCP option affects the meanings of \d, \s, \w
1114     and their upper case partners, just as it does when they appear outside a
1115     character class, as described in the section entitled
1116     <a href="#genericchartypes">"Generic character types"</a>
1117     above. The escape sequence \b has a different meaning inside a character
1118     class; it matches the backspace character. The sequences \B, \N, \R, and \X
1119     are not special inside a character class. Like any other unrecognized escape
1120     sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1121     default, but cause an error if the PCRE_EXTRA option is set.
1122     </P>
1123     <P>
1124     A circumflex can conveniently be used with the upper case character types to
1125 ph10 535 specify a more restricted set of characters than the matching lower case type.
1126 ph10 579 For example, the class [^\W_] matches any letter or digit, but not underscore,
1127     whereas [\w] includes underscore. A positive character class should be read as
1128     "something OR something OR ..." and a negative class as "NOT something AND NOT
1129     something AND NOT ...".
1130 nigel 63 </P>
1131     <P>
1132 nigel 75 The only metacharacters that are recognized in character classes are backslash,
1133     hyphen (only where it can be interpreted as specifying a range), circumflex
1134     (only at the start), opening square bracket (only when it can be interpreted as
1135     introducing a POSIX class name - see the next section), and the terminating
1136     closing square bracket. However, escaping other non-alphanumeric characters
1137     does no harm.
1138 nigel 63 </P>
1139 ph10 227 <br><a name="SEC9" href="#TOC1">POSIX CHARACTER CLASSES</a><br>
1140 nigel 63 <P>
1141 nigel 75 Perl supports the POSIX notation for character classes. This uses names
1142 nigel 63 enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1143     this notation. For example,
1144     <pre>
1145     [01[:alpha:]%]
1146 nigel 75 </pre>
1147 nigel 63 matches "0", "1", any alphabetic character, or "%". The supported class names
1148 ph10 535 are:
1149 nigel 63 <pre>
1150     alnum letters and digits
1151     alpha letters
1152     ascii character codes 0 - 127
1153     blank space or tab only
1154     cntrl control characters
1155     digit decimal digits (same as \d)
1156     graph printing characters, excluding space
1157     lower lower case letters
1158     print printing characters, including space
1159 ph10 535 punct printing characters, excluding letters and digits and space
1160 nigel 63 space white space (not quite the same as \s)
1161     upper upper case letters
1162     word "word" characters (same as \w)
1163     xdigit hexadecimal digits
1164 nigel 75 </pre>
1165 nigel 63 The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and
1166     space (32). Notice that this list includes the VT character (code 11). This
1167     makes "space" different to \s, which does not include VT (for Perl
1168     compatibility).
1169     </P>
1170     <P>
1171     The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1172     5.8. Another Perl extension is negation, which is indicated by a ^ character
1173     after the colon. For example,
1174     <pre>
1175     [12[:^digit:]]
1176 nigel 75 </pre>
1177 nigel 63 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1178     syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1179     supported, and an error is given if they are encountered.
1180     </P>
1181     <P>
1182 ph10 535 By default, in UTF-8 mode, characters with values greater than 128 do not match
1183     any of the POSIX character classes. However, if the PCRE_UCP option is passed
1184     to <b>pcre_compile()</b>, some of the classes are changed so that Unicode
1185     character properties are used. This is achieved by replacing the POSIX classes
1186     by other sequences, as follows:
1187     <pre>
1188     [:alnum:] becomes \p{Xan}
1189     [:alpha:] becomes \p{L}
1190     [:blank:] becomes \h
1191     [:digit:] becomes \p{Nd}
1192     [:lower:] becomes \p{Ll}
1193     [:space:] becomes \p{Xps}
1194     [:upper:] becomes \p{Lu}
1195     [:word:] becomes \p{Xwd}
1196     </pre>
1197     Negated versions, such as [:^alpha:] use \P instead of \p. The other POSIX
1198     classes are unchanged, and match only characters with code points less than
1199     128.
1200 nigel 63 </P>
1201 ph10 227 <br><a name="SEC10" href="#TOC1">VERTICAL BAR</a><br>
1202 nigel 63 <P>
1203     Vertical bar characters are used to separate alternative patterns. For example,
1204     the pattern
1205     <pre>
1206     gilbert|sullivan
1207 nigel 75 </pre>
1208 nigel 63 matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1209 nigel 91 and an empty alternative is permitted (matching the empty string). The matching
1210     process tries each alternative in turn, from left to right, and the first one
1211     that succeeds is used. If the alternatives are within a subpattern
1212 nigel 75 <a href="#subpattern">(defined below),</a>
1213     "succeeds" means matching the rest of the main pattern as well as the
1214     alternative in the subpattern.
1215 nigel 63 </P>
1216 ph10 227 <br><a name="SEC11" href="#TOC1">INTERNAL OPTION SETTING</a><br>
1217 nigel 63 <P>
1218     The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1219 ph10 243 PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1220     the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1221     The option letters are
1222 nigel 63 <pre>
1223     i for PCRE_CASELESS
1224     m for PCRE_MULTILINE
1225     s for PCRE_DOTALL
1226     x for PCRE_EXTENDED
1227 nigel 75 </pre>
1228 nigel 63 For example, (?im) sets caseless, multiline matching. It is also possible to
1229     unset these options by preceding the letter with a hyphen, and a combined
1230     setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1231     PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1232     permitted. If a letter appears both before and after the hyphen, the option is
1233     unset.
1234     </P>
1235     <P>
1236 ph10 243 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1237     changed in the same way as the Perl-compatible options by using the characters
1238     J, U and X respectively.
1239     </P>
1240     <P>
1241 ph10 416 When one of these option changes occurs at top level (that is, not inside
1242     subpattern parentheses), the change applies to the remainder of the pattern
1243     that follows. If the change is placed right at the start of a pattern, PCRE
1244     extracts it into the global options (and it will therefore show up in data
1245     extracted by the <b>pcre_fullinfo()</b> function).
1246 nigel 63 </P>
1247     <P>
1248 nigel 93 An option change within a subpattern (see below for a description of
1249 ph10 572 subpatterns) affects only that part of the subpattern that follows it, so
1250 nigel 63 <pre>
1251     (a(?i)b)c
1252 nigel 75 </pre>
1253 nigel 63 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1254     By this means, options can be made to have different settings in different
1255     parts of the pattern. Any changes made in one alternative do carry on
1256     into subsequent branches within the same subpattern. For example,
1257     <pre>
1258     (a(?i)b|c)
1259 nigel 75 </pre>
1260 nigel 63 matches "ab", "aB", "c", and "C", even though when matching "C" the first
1261     branch is abandoned before the option setting. This is because the effects of
1262     option settings happen at compile time. There would be some very weird
1263     behaviour otherwise.
1264 ph10 254 </P>
1265     <P>
1266     <b>Note:</b> There are other PCRE-specific options that can be set by the
1267     application when the compile or match functions are called. In some cases the
1268 ph10 416 pattern can contain special leading sequences such as (*CRLF) to override what
1269     the application has set or what has been defaulted. Details are given in the
1270     section entitled
1271 ph10 254 <a href="#newlineseq">"Newline sequences"</a>
1272 ph10 535 above. There are also the (*UTF8) and (*UCP) leading sequences that can be used
1273     to set UTF-8 and Unicode property modes; they are equivalent to setting the
1274     PCRE_UTF8 and the PCRE_UCP options, respectively.
1275 nigel 75 <a name="subpattern"></a></P>
1276 ph10 227 <br><a name="SEC12" href="#TOC1">SUBPATTERNS</a><br>
1277 nigel 63 <P>
1278     Subpatterns are delimited by parentheses (round brackets), which can be nested.
1279 nigel 75 Turning part of a pattern into a subpattern does two things:
1280     <br>
1281     <br>
1282 nigel 63 1. It localizes a set of alternatives. For example, the pattern
1283     <pre>
1284     cat(aract|erpillar|)
1285 nigel 75 </pre>
1286 ph10 572 matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1287     match "cataract", "erpillar" or an empty string.
1288 nigel 75 <br>
1289     <br>
1290     2. It sets up the subpattern as a capturing subpattern. This means that, when
1291     the whole pattern matches, that portion of the subject string that matched the
1292     subpattern is passed back to the caller via the <i>ovector</i> argument of
1293 nigel 63 <b>pcre_exec()</b>. Opening parentheses are counted from left to right (starting
1294 ph10 572 from 1) to obtain numbers for the capturing subpatterns. For example, if the
1295     string "the red king" is matched against the pattern
1296 nigel 63 <pre>
1297     the ((red|white) (king|queen))
1298 nigel 75 </pre>
1299 nigel 63 the captured substrings are "red king", "red", and "king", and are numbered 1,
1300     2, and 3, respectively.
1301     </P>
1302     <P>
1303     The fact that plain parentheses fulfil two functions is not always helpful.
1304     There are often times when a grouping subpattern is required without a
1305     capturing requirement. If an opening parenthesis is followed by a question mark
1306     and a colon, the subpattern does not do any capturing, and is not counted when
1307     computing the number of any subsequent capturing subpatterns. For example, if
1308     the string "the white queen" is matched against the pattern
1309     <pre>
1310     the ((?:red|white) (king|queen))
1311 nigel 75 </pre>
1312 nigel 63 the captured substrings are "white queen" and "queen", and are numbered 1 and
1313 nigel 93 2. The maximum number of capturing subpatterns is 65535.
1314 nigel 63 </P>
1315     <P>
1316     As a convenient shorthand, if any option settings are required at the start of
1317     a non-capturing subpattern, the option letters may appear between the "?" and
1318     the ":". Thus the two patterns
1319     <pre>
1320     (?i:saturday|sunday)
1321     (?:(?i)saturday|sunday)
1322 nigel 75 </pre>
1323 nigel 63 match exactly the same set of strings. Because alternative branches are tried
1324     from left to right, and options are not reset until the end of the subpattern
1325     is reached, an option setting in one branch does affect subsequent branches, so
1326     the above patterns match "SUNDAY" as well as "Saturday".
1327 ph10 461 <a name="dupsubpatternnumber"></a></P>
1328 ph10 227 <br><a name="SEC13" href="#TOC1">DUPLICATE SUBPATTERN NUMBERS</a><br>
1329 nigel 63 <P>
1330 ph10 182 Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1331     the same numbers for its capturing parentheses. Such a subpattern starts with
1332     (?| and is itself a non-capturing subpattern. For example, consider this
1333     pattern:
1334     <pre>
1335     (?|(Sat)ur|(Sun))day
1336     </pre>
1337     Because the two alternatives are inside a (?| group, both sets of capturing
1338     parentheses are numbered one. Thus, when the pattern matches, you can look
1339     at captured substring number one, whichever alternative matched. This construct
1340     is useful when you want to capture part, but not all, of one of a number of
1341     alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1342     number is reset at the start of each branch. The numbers of any capturing
1343 ph10 572 parentheses that follow the subpattern start after the highest number used in
1344     any branch. The following example is taken from the Perl documentation. The
1345     numbers underneath show in which buffer the captured content will be stored.
1346 ph10 182 <pre>
1347     # before ---------------branch-reset----------- after
1348     / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1349     # 1 2 2 3 2 3 4
1350     </pre>
1351 ph10 488 A back reference to a numbered subpattern uses the most recent value that is
1352     set for that number by any subpattern. The following pattern matches "abcabc"
1353     or "defdef":
1354 ph10 461 <pre>
1355     /(?|(abc)|(def))\1/
1356     </pre>
1357 ph10 733 In contrast, a subroutine call to a numbered subpattern always refers to the
1358     first one in the pattern with the given number. The following pattern matches
1359     "abcabc" or "defabc":
1360 ph10 461 <pre>
1361     /(?|(abc)|(def))(?1)/
1362     </pre>
1363     If a
1364     <a href="#conditions">condition test</a>
1365     for a subpattern's having matched refers to a non-unique number, the test is
1366     true if any of the subpatterns of that number have matched.
1367 ph10 182 </P>
1368     <P>
1369     An alternative approach to using this "branch reset" feature is to use
1370     duplicate named subpatterns, as described in the next section.
1371     </P>
1372 ph10 227 <br><a name="SEC14" href="#TOC1">NAMED SUBPATTERNS</a><br>
1373 ph10 182 <P>
1374 nigel 63 Identifying capturing parentheses by number is simple, but it can be very hard
1375     to keep track of the numbers in complicated regular expressions. Furthermore,
1376 nigel 75 if an expression is modified, the numbers may change. To help with this
1377 nigel 93 difficulty, PCRE supports the naming of subpatterns. This feature was not
1378     added to Perl until release 5.10. Python had the feature earlier, and PCRE
1379     introduced it at release 4.0, using the Python syntax. PCRE now supports both
1380 ph10 461 the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1381     have different names, but PCRE does not.
1382 nigel 93 </P>
1383     <P>
1384     In PCRE, a subpattern can be named in one of three ways: (?&#60;name&#62;...) or
1385     (?'name'...) as in Perl, or (?P&#60;name&#62;...) as in Python. References to capturing
1386 nigel 91 parentheses from other parts of the pattern, such as
1387 ph10 488 <a href="#backreferences">back references,</a>
1388 nigel 91 <a href="#recursion">recursion,</a>
1389     and
1390     <a href="#conditions">conditions,</a>
1391     can be made by name as well as by number.
1392 nigel 63 </P>
1393     <P>
1394 nigel 91 Names consist of up to 32 alphanumeric characters and underscores. Named
1395 nigel 93 capturing parentheses are still allocated numbers as well as names, exactly as
1396     if the names were not present. The PCRE API provides function calls for
1397     extracting the name-to-number translation table from a compiled pattern. There
1398     is also a convenience function for extracting a captured substring by name.
1399 nigel 91 </P>
1400     <P>
1401     By default, a name must be unique within a pattern, but it is possible to relax
1402 ph10 461 this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1403     names are also always permitted for subpatterns with the same number, set up as
1404     described in the previous section.) Duplicate names can be useful for patterns
1405     where only one instance of the named parentheses can match. Suppose you want to
1406     match the name of a weekday, either as a 3-letter abbreviation or as the full
1407     name, and in both cases you want to extract the abbreviation. This pattern
1408     (ignoring the line breaks) does the job:
1409 nigel 91 <pre>
1410 nigel 93 (?&#60;DN&#62;Mon|Fri|Sun)(?:day)?|
1411     (?&#60;DN&#62;Tue)(?:sday)?|
1412     (?&#60;DN&#62;Wed)(?:nesday)?|
1413     (?&#60;DN&#62;Thu)(?:rsday)?|
1414     (?&#60;DN&#62;Sat)(?:urday)?
1415 nigel 91 </pre>
1416     There are five capturing substrings, but only one is ever set after a match.
1417 ph10 182 (An alternative way of solving this problem is to use a "branch reset"
1418     subpattern, as described in the previous section.)
1419     </P>
1420     <P>
1421 nigel 91 The convenience function for extracting the data by name returns the substring
1422 nigel 93 for the first (and in this example, the only) subpattern of that name that
1423 ph10 461 matched. This saves searching to find which numbered subpattern it was.
1424     </P>
1425     <P>
1426 ph10 488 If you make a back reference to a non-unique named subpattern from elsewhere in
1427 ph10 461 the pattern, the one that corresponds to the first occurrence of the name is
1428     used. In the absence of duplicate numbers (see the previous section) this is
1429     the one with the lowest number. If you use a named reference in a condition
1430     test (see the
1431     <a href="#conditions">section about conditions</a>
1432     below), either to check whether a subpattern has matched, or to check for
1433     recursion, all subpatterns with the same name are tested. If the condition is
1434     true for any one of them, the overall condition is true. This is the same
1435     behaviour as testing by number. For further details of the interfaces for
1436     handling named subpatterns, see the
1437 nigel 63 <a href="pcreapi.html"><b>pcreapi</b></a>
1438     documentation.
1439     </P>
1440 ph10 392 <P>
1441     <b>Warning:</b> You cannot use different names to distinguish between two
1442 ph10 461 subpatterns with the same number because PCRE uses only the numbers when
1443     matching. For this reason, an error is given at compile time if different names
1444     are given to subpatterns with the same number. However, you can give the same
1445     name to subpatterns with the same number, even when PCRE_DUPNAMES is not set.
1446 ph10 392 </P>
1447 ph10 227 <br><a name="SEC15" href="#TOC1">REPETITION</a><br>
1448 nigel 63 <P>
1449     Repetition is specified by quantifiers, which can follow any of the following
1450     items:
1451     <pre>
1452     a literal data character
1453 nigel 93 the dot metacharacter
1454 nigel 63 the \C escape sequence
1455 nigel 75 the \X escape sequence (in UTF-8 mode with Unicode properties)
1456 nigel 93 the \R escape sequence
1457 ph10 572 an escape such as \d or \pL that matches a single character
1458 nigel 63 a character class
1459     a back reference (see next section)
1460 ph10 654 a parenthesized subpattern (including assertions)
1461 ph10 733 a subroutine call to a subpattern (recursive or otherwise)
1462 nigel 75 </pre>
1463 nigel 63 The general repetition quantifier specifies a minimum and maximum number of
1464     permitted matches, by giving the two numbers in curly brackets (braces),
1465     separated by a comma. The numbers must be less than 65536, and the first must
1466     be less than or equal to the second. For example:
1467     <pre>
1468     z{2,4}
1469 nigel 75 </pre>
1470 nigel 63 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1471     character. If the second number is omitted, but the comma is present, there is
1472     no upper limit; if the second number and the comma are both omitted, the
1473     quantifier specifies an exact number of required matches. Thus
1474     <pre>
1475     [aeiou]{3,}
1476 nigel 75 </pre>
1477 nigel 63 matches at least 3 successive vowels, but may match many more, while
1478     <pre>
1479     \d{8}
1480 nigel 75 </pre>
1481 nigel 63 matches exactly 8 digits. An opening curly bracket that appears in a position
1482     where a quantifier is not allowed, or one that does not match the syntax of a
1483     quantifier, is taken as a literal character. For example, {,6} is not a
1484     quantifier, but a literal string of four characters.
1485     </P>
1486     <P>
1487     In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual
1488     bytes. Thus, for example, \x{100}{2} matches two UTF-8 characters, each of
1489 nigel 75 which is represented by a two-byte sequence. Similarly, when Unicode property
1490     support is available, \X{3} matches three Unicode extended sequences, each of
1491     which may be several bytes long (and they may be of different lengths).
1492 nigel 63 </P>
1493     <P>
1494     The quantifier {0} is permitted, causing the expression to behave as if the
1495 ph10 345 previous item and the quantifier were not present. This may be useful for
1496     subpatterns that are referenced as
1497     <a href="#subpatternsassubroutines">subroutines</a>
1498 ph10 572 from elsewhere in the pattern (but see also the section entitled
1499     <a href="#subdefine">"Defining subpatterns for use by reference only"</a>
1500     below). Items other than subpatterns that have a {0} quantifier are omitted
1501     from the compiled pattern.
1502 nigel 63 </P>
1503     <P>
1504 nigel 93 For convenience, the three most common quantifiers have single-character
1505     abbreviations:
1506 nigel 63 <pre>
1507     * is equivalent to {0,}
1508     + is equivalent to {1,}
1509     ? is equivalent to {0,1}
1510 nigel 75 </pre>
1511 nigel 63 It is possible to construct infinite loops by following a subpattern that can
1512     match no characters with a quantifier that has no upper limit, for example:
1513     <pre>
1514     (a?)*
1515 nigel 75 </pre>
1516 nigel 63 Earlier versions of Perl and PCRE used to give an error at compile time for
1517     such patterns. However, because there are cases where this can be useful, such
1518     patterns are now accepted, but if any repetition of the subpattern does in fact
1519     match no characters, the loop is forcibly broken.
1520     </P>
1521     <P>
1522     By default, the quantifiers are "greedy", that is, they match as much as
1523     possible (up to the maximum number of permitted times), without causing the
1524     rest of the pattern to fail. The classic example of where this gives problems
1525 nigel 75 is in trying to match comments in C programs. These appear between /* and */
1526     and within the comment, individual * and / characters may appear. An attempt to
1527     match C comments by applying the pattern
1528 nigel 63 <pre>
1529     /\*.*\*/
1530 nigel 75 </pre>
1531 nigel 63 to the string
1532     <pre>
1533 nigel 75 /* first comment */ not comment /* second comment */
1534     </pre>
1535 nigel 63 fails, because it matches the entire string owing to the greediness of the .*
1536     item.
1537     </P>
1538     <P>
1539     However, if a quantifier is followed by a question mark, it ceases to be
1540     greedy, and instead matches the minimum number of times possible, so the
1541     pattern
1542     <pre>
1543     /\*.*?\*/
1544 nigel 75 </pre>
1545 nigel 63 does the right thing with the C comments. The meaning of the various
1546     quantifiers is not otherwise changed, just the preferred number of matches.
1547     Do not confuse this use of question mark with its use as a quantifier in its
1548     own right. Because it has two uses, it can sometimes appear doubled, as in
1549     <pre>
1550     \d??\d
1551 nigel 75 </pre>
1552 nigel 63 which matches one digit by preference, but can match two if that is the only
1553     way the rest of the pattern matches.
1554     </P>
1555     <P>
1556 nigel 93 If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1557 nigel 63 the quantifiers are not greedy by default, but individual ones can be made
1558     greedy by following them with a question mark. In other words, it inverts the
1559     default behaviour.
1560     </P>
1561     <P>
1562     When a parenthesized subpattern is quantified with a minimum repeat count that
1563 nigel 75 is greater than 1 or with a limited maximum, more memory is required for the
1564 nigel 63 compiled pattern, in proportion to the size of the minimum or maximum.
1565     </P>
1566     <P>
1567     If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1568 nigel 93 to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1569 nigel 63 implicitly anchored, because whatever follows will be tried against every
1570     character position in the subject string, so there is no point in retrying the
1571     overall match at any position after the first. PCRE normally treats such a
1572     pattern as though it were preceded by \A.
1573     </P>
1574     <P>
1575     In cases where it is known that the subject string contains no newlines, it is
1576     worth setting PCRE_DOTALL in order to obtain this optimization, or
1577     alternatively using ^ to indicate anchoring explicitly.
1578     </P>
1579     <P>
1580     However, there is one situation where the optimization cannot be used. When .*
1581 ph10 488 is inside capturing parentheses that are the subject of a back reference
1582 nigel 93 elsewhere in the pattern, a match at the start may fail where a later one
1583     succeeds. Consider, for example:
1584 nigel 63 <pre>
1585     (.*)abc\1
1586 nigel 75 </pre>
1587 nigel 63 If the subject is "xyz123abc123" the match point is the fourth character. For
1588     this reason, such a pattern is not implicitly anchored.
1589     </P>
1590     <P>
1591     When a capturing subpattern is repeated, the value captured is the substring
1592     that matched the final iteration. For example, after
1593     <pre>
1594     (tweedle[dume]{3}\s*)+
1595 nigel 75 </pre>
1596 nigel 63 has matched "tweedledum tweedledee" the value of the captured substring is
1597     "tweedledee". However, if there are nested capturing subpatterns, the
1598     corresponding captured values may have been set in previous iterations. For
1599     example, after
1600     <pre>
1601     /(a|(b))+/
1602 nigel 75 </pre>
1603 nigel 63 matches "aba" the value of the second captured substring is "b".
1604 nigel 75 <a name="atomicgroup"></a></P>
1605 ph10 227 <br><a name="SEC16" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br>
1606 nigel 63 <P>
1607 nigel 93 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1608     repetition, failure of what follows normally causes the repeated item to be
1609     re-evaluated to see if a different number of repeats allows the rest of the
1610     pattern to match. Sometimes it is useful to prevent this, either to change the
1611     nature of the match, or to cause it fail earlier than it otherwise might, when
1612     the author of the pattern knows there is no point in carrying on.
1613 nigel 63 </P>
1614     <P>
1615     Consider, for example, the pattern \d+foo when applied to the subject line
1616     <pre>
1617     123456bar
1618 nigel 75 </pre>
1619 nigel 63 After matching all 6 digits and then failing to match "foo", the normal
1620     action of the matcher is to try again with only 5 digits matching the \d+
1621     item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1622     (a term taken from Jeffrey Friedl's book) provides the means for specifying
1623     that once a subpattern has matched, it is not to be re-evaluated in this way.
1624     </P>
1625     <P>
1626 nigel 93 If we use atomic grouping for the previous example, the matcher gives up
1627 nigel 63 immediately on failing to match "foo" the first time. The notation is a kind of
1628     special parenthesis, starting with (?&#62; as in this example:
1629     <pre>
1630 nigel 73 (?&#62;\d+)foo
1631 nigel 75 </pre>
1632 nigel 63 This kind of parenthesis "locks up" the part of the pattern it contains once
1633     it has matched, and a failure further into the pattern is prevented from
1634     backtracking into it. Backtracking past it to previous items, however, works as
1635     normal.
1636     </P>
1637     <P>
1638     An alternative description is that a subpattern of this type matches the string
1639     of characters that an identical standalone pattern would match, if anchored at
1640     the current point in the subject string.
1641     </P>
1642     <P>
1643     Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1644     the above example can be thought of as a maximizing repeat that must swallow
1645     everything it can. So, while both \d+ and \d+? are prepared to adjust the
1646     number of digits they match in order to make the rest of the pattern match,
1647     (?&#62;\d+) can only match an entire sequence of digits.
1648     </P>
1649     <P>
1650     Atomic groups in general can of course contain arbitrarily complicated
1651     subpatterns, and can be nested. However, when the subpattern for an atomic
1652     group is just a single repeated item, as in the example above, a simpler
1653     notation, called a "possessive quantifier" can be used. This consists of an
1654     additional + character following a quantifier. Using this notation, the
1655     previous example can be rewritten as
1656     <pre>
1657 nigel 75 \d++foo
1658     </pre>
1659 ph10 208 Note that a possessive quantifier can be used with an entire group, for
1660     example:
1661     <pre>
1662     (abc|xyz){2,3}+
1663     </pre>
1664 nigel 63 Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1665     option is ignored. They are a convenient notation for the simpler forms of
1666 nigel 93 atomic group. However, there is no difference in the meaning of a possessive
1667     quantifier and the equivalent atomic group, though there may be a performance
1668     difference; possessive quantifiers should be slightly faster.
1669 nigel 63 </P>
1670     <P>
1671 nigel 93 The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1672     Jeffrey Friedl originated the idea (and the name) in the first edition of his
1673     book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1674     package, and PCRE copied it from there. It ultimately found its way into Perl
1675     at release 5.10.
1676 nigel 63 </P>
1677     <P>
1678 nigel 93 PCRE has an optimization that automatically "possessifies" certain simple
1679     pattern constructs. For example, the sequence A+B is treated as A++B because
1680     there is no point in backtracking into a sequence of A's when B must follow.
1681     </P>
1682     <P>
1683 nigel 63 When a pattern contains an unlimited repeat inside a subpattern that can itself
1684     be repeated an unlimited number of times, the use of an atomic group is the
1685     only way to avoid some failing matches taking a very long time indeed. The
1686     pattern
1687     <pre>
1688     (\D+|&#60;\d+&#62;)*[!?]
1689 nigel 75 </pre>
1690 nigel 63 matches an unlimited number of substrings that either consist of non-digits, or
1691     digits enclosed in &#60;&#62;, followed by either ! or ?. When it matches, it runs
1692     quickly. However, if it is applied to
1693     <pre>
1694     aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1695 nigel 75 </pre>
1696 nigel 63 it takes a long time before reporting failure. This is because the string can
1697 nigel 75 be divided between the internal \D+ repeat and the external * repeat in a
1698     large number of ways, and all have to be tried. (The example uses [!?] rather
1699     than a single character at the end, because both PCRE and Perl have an
1700     optimization that allows for fast failure when a single character is used. They
1701     remember the last single character that is required for a match, and fail early
1702     if it is not present in the string.) If the pattern is changed so that it uses
1703     an atomic group, like this:
1704 nigel 63 <pre>
1705     ((?&#62;\D+)|&#60;\d+&#62;)*[!?]
1706 nigel 75 </pre>
1707 nigel 63 sequences of non-digits cannot be broken, and failure happens quickly.
1708 nigel 75 <a name="backreferences"></a></P>
1709 ph10 227 <br><a name="SEC17" href="#TOC1">BACK REFERENCES</a><br>
1710 nigel 63 <P>
1711     Outside a character class, a backslash followed by a digit greater than 0 (and
1712     possibly further digits) is a back reference to a capturing subpattern earlier
1713     (that is, to its left) in the pattern, provided there have been that many
1714     previous capturing left parentheses.
1715     </P>
1716     <P>
1717     However, if the decimal number following the backslash is less than 10, it is
1718     always taken as a back reference, and causes an error only if there are not
1719     that many capturing left parentheses in the entire pattern. In other words, the
1720     parentheses that are referenced need not be to the left of the reference for
1721 nigel 91 numbers less than 10. A "forward back reference" of this type can make sense
1722     when a repetition is involved and the subpattern to the right has participated
1723     in an earlier iteration.
1724     </P>
1725     <P>
1726 nigel 93 It is not possible to have a numerical "forward back reference" to a subpattern
1727     whose number is 10 or more using this syntax because a sequence such as \50 is
1728     interpreted as a character defined in octal. See the subsection entitled
1729 nigel 91 "Non-printing characters"
1730 nigel 75 <a href="#digitsafterbackslash">above</a>
1731 nigel 93 for further details of the handling of digits following a backslash. There is
1732     no such problem when named parentheses are used. A back reference to any
1733     subpattern is possible using named parentheses (see below).
1734 nigel 63 </P>
1735     <P>
1736 nigel 93 Another way of avoiding the ambiguity inherent in the use of digits following a
1737 ph10 572 backslash is to use the \g escape sequence. This escape must be followed by an
1738     unsigned number or a negative number, optionally enclosed in braces. These
1739     examples are all identical:
1740 nigel 93 <pre>
1741     (ring), \1
1742     (ring), \g1
1743     (ring), \g{1}
1744     </pre>
1745 ph10 208 An unsigned number specifies an absolute reference without the ambiguity that
1746     is present in the older syntax. It is also useful when literal digits follow
1747     the reference. A negative number is a relative reference. Consider this
1748     example:
1749 nigel 93 <pre>
1750     (abc(def)ghi)\g{-1}
1751     </pre>
1752     The sequence \g{-1} is a reference to the most recently started capturing
1753 ph10 579 subpattern before \g, that is, is it equivalent to \2 in this example.
1754     Similarly, \g{-2} would be equivalent to \1. The use of relative references
1755     can be helpful in long patterns, and also in patterns that are created by
1756     joining together fragments that contain references within themselves.
1757 nigel 93 </P>
1758     <P>
1759 nigel 63 A back reference matches whatever actually matched the capturing subpattern in
1760     the current subject string, rather than anything matching the subpattern
1761     itself (see
1762     <a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a>
1763     below for a way of doing that). So the pattern
1764     <pre>
1765     (sens|respons)e and \1ibility
1766 nigel 75 </pre>
1767 nigel 63 matches "sense and sensibility" and "response and responsibility", but not
1768     "sense and responsibility". If caseful matching is in force at the time of the
1769     back reference, the case of letters is relevant. For example,
1770     <pre>
1771     ((?i)rah)\s+\1
1772 nigel 75 </pre>
1773 nigel 63 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
1774     capturing subpattern is matched caselessly.
1775     </P>
1776     <P>
1777 ph10 172 There are several different ways of writing back references to named
1778     subpatterns. The .NET syntax \k{name} and the Perl syntax \k&#60;name&#62; or
1779     \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
1780     back reference syntax, in which \g can be used for both numeric and named
1781     references, is also supported. We could rewrite the above example in any of
1782 nigel 93 the following ways:
1783 nigel 63 <pre>
1784 nigel 93 (?&#60;p1&#62;(?i)rah)\s+\k&#60;p1&#62;
1785 ph10 172 (?'p1'(?i)rah)\s+\k{p1}
1786 nigel 91 (?P&#60;p1&#62;(?i)rah)\s+(?P=p1)
1787 ph10 172 (?&#60;p1&#62;(?i)rah)\s+\g{p1}
1788 nigel 75 </pre>
1789 nigel 91 A subpattern that is referenced by name may appear in the pattern before or
1790     after the reference.
1791     </P>
1792     <P>
1793 nigel 63 There may be more than one back reference to the same subpattern. If a
1794     subpattern has not actually been used in a particular match, any back
1795 ph10 461 references to it always fail by default. For example, the pattern
1796 nigel 63 <pre>
1797     (a|(bc))\2
1798 nigel 75 </pre>
1799 ph10 461 always fails if it starts to match "a" rather than "bc". However, if the
1800     PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
1801     unset value matches an empty string.
1802     </P>
1803     <P>
1804     Because there may be many capturing parentheses in a pattern, all digits
1805     following a backslash are taken as part of a potential back reference number.
1806     If the pattern continues with a digit character, some delimiter must be used to
1807     terminate the back reference. If the PCRE_EXTENDED option is set, this can be
1808     whitespace. Otherwise, the \g{ syntax or an empty comment (see
1809 nigel 75 <a href="#comments">"Comments"</a>
1810     below) can be used.
1811 nigel 63 </P>
1812 ph10 488 <br><b>
1813     Recursive back references
1814     </b><br>
1815 nigel 63 <P>
1816     A back reference that occurs inside the parentheses to which it refers fails
1817     when the subpattern is first used, so, for example, (a\1) never matches.
1818     However, such references can be useful inside repeated subpatterns. For
1819     example, the pattern
1820     <pre>
1821     (a|b\1)+
1822 nigel 75 </pre>
1823 nigel 63 matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
1824     the subpattern, the back reference matches the character string corresponding
1825     to the previous iteration. In order for this to work, the pattern must be such
1826     that the first iteration does not need to match the back reference. This can be
1827     done using alternation, as in the example above, or by a quantifier with a
1828     minimum of zero.
1829 ph10 488 </P>
1830     <P>
1831     Back references of this type cause the group that they reference to be treated
1832     as an
1833     <a href="#atomicgroup">atomic group.</a>
1834     Once the whole group has been matched, a subsequent matching failure cannot
1835     cause backtracking into the middle of the group.
1836 nigel 75 <a name="bigassertions"></a></P>
1837 ph10 227 <br><a name="SEC18" href="#TOC1">ASSERTIONS</a><br>
1838 nigel 63 <P>
1839     An assertion is a test on the characters following or preceding the current
1840     matching point that does not actually consume any characters. The simple
1841 nigel 75 assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described
1842     <a href="#smallassertions">above.</a>
1843     </P>
1844     <P>
1845 nigel 63 More complicated assertions are coded as subpatterns. There are two kinds:
1846     those that look ahead of the current position in the subject string, and those
1847 nigel 75 that look behind it. An assertion subpattern is matched in the normal way,
1848     except that it does not cause the current matching position to be changed.
1849 nigel 63 </P>
1850     <P>
1851 ph10 654 Assertion subpatterns are not capturing subpatterns. If such an assertion
1852     contains capturing subpatterns within it, these are counted for the purposes of
1853     numbering the capturing subpatterns in the whole pattern. However, substring
1854     capturing is carried out only for positive assertions, because it does not make
1855     sense for negative assertions.
1856 nigel 63 </P>
1857 ph10 654 <P>
1858     For compatibility with Perl, assertion subpatterns may be repeated; though
1859     it makes no sense to assert the same thing several times, the side effect of
1860     capturing parentheses may occasionally be useful. In practice, there only three
1861     cases:
1862     <br>
1863     <br>
1864     (1) If the quantifier is {0}, the assertion is never obeyed during matching.
1865     However, it may contain internal capturing parenthesized groups that are called
1866     from elsewhere via the
1867     <a href="#subpatternsassubroutines">subroutine mechanism.</a>
1868     <br>
1869     <br>
1870     (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
1871     were {0,1}. At run time, the rest of the pattern match is tried with and
1872     without the assertion, the order depending on the greediness of the quantifier.
1873     <br>
1874     <br>
1875     (3) If the minimum repetition is greater than zero, the quantifier is ignored.
1876     The assertion is obeyed just once when encountered during matching.
1877     </P>
1878 nigel 75 <br><b>
1879     Lookahead assertions
1880     </b><br>
1881 nigel 63 <P>
1882 nigel 91 Lookahead assertions start with (?= for positive assertions and (?! for
1883     negative assertions. For example,
1884 nigel 63 <pre>
1885     \w+(?=;)
1886 nigel 75 </pre>
1887 nigel 63 matches a word followed by a semicolon, but does not include the semicolon in
1888     the match, and
1889     <pre>
1890     foo(?!bar)
1891 nigel 75 </pre>
1892 nigel 63 matches any occurrence of "foo" that is not followed by "bar". Note that the
1893     apparently similar pattern
1894     <pre>
1895     (?!foo)bar
1896 nigel 75 </pre>
1897 nigel 63 does not find an occurrence of "bar" that is preceded by something other than
1898     "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
1899     (?!foo) is always true when the next three characters are "bar". A
1900 nigel 75 lookbehind assertion is needed to achieve the other effect.
1901 nigel 63 </P>
1902     <P>
1903     If you want to force a matching failure at some point in a pattern, the most
1904     convenient way to do it is with (?!) because an empty string always matches, so
1905     an assertion that requires there not to be an empty string must always fail.
1906 ph10 579 The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
1907 nigel 75 <a name="lookbehind"></a></P>
1908     <br><b>
1909     Lookbehind assertions
1910     </b><br>
1911 nigel 63 <P>
1912     Lookbehind assertions start with (?&#60;= for positive assertions and (?&#60;! for
1913     negative assertions. For example,
1914     <pre>
1915     (?&#60;!foo)bar
1916 nigel 75 </pre>
1917 nigel 63 does find an occurrence of "bar" that is not preceded by "foo". The contents of
1918     a lookbehind assertion are restricted such that all the strings it matches must
1919 nigel 91 have a fixed length. However, if there are several top-level alternatives, they
1920     do not all have to have the same fixed length. Thus
1921 nigel 63 <pre>
1922     (?&#60;=bullock|donkey)
1923 nigel 75 </pre>
1924 nigel 63 is permitted, but
1925     <pre>
1926     (?&#60;!dogs?|cats?)
1927 nigel 75 </pre>
1928 nigel 63 causes an error at compile time. Branches that match different length strings
1929     are permitted only at the top level of a lookbehind assertion. This is an
1930 ph10 572 extension compared with Perl, which requires all branches to match the same
1931     length of string. An assertion such as
1932 nigel 63 <pre>
1933     (?&#60;=ab(c|de))
1934 nigel 75 </pre>
1935 nigel 63 is not permitted, because its single top-level branch can match two different
1936 ph10 454 lengths, but it is acceptable to PCRE if rewritten to use two top-level
1937     branches:
1938 nigel 63 <pre>
1939     (?&#60;=abc|abde)
1940 nigel 75 </pre>
1941 ph10 572 In some cases, the escape sequence \K
1942 ph10 172 <a href="#resetmatchstart">(see above)</a>
1943 ph10 461 can be used instead of a lookbehind assertion to get round the fixed-length
1944 ph10 454 restriction.
1945 ph10 172 </P>
1946     <P>
1947 nigel 63 The implementation of lookbehind assertions is, for each alternative, to
1948 nigel 93 temporarily move the current position back by the fixed length and then try to
1949 nigel 63 match. If there are insufficient characters before the current position, the
1950 nigel 93 assertion fails.
1951 nigel 63 </P>
1952     <P>
1953 ph10 836 In UTF-8 mode, PCRE does not allow the \C escape (which matches a single byte,
1954     even in UTF-8 mode) to appear in lookbehind assertions, because it makes it
1955     impossible to calculate the length of the lookbehind. The \X and \R escapes,
1956     which can match different numbers of bytes, are also not permitted.
1957 nigel 63 </P>
1958     <P>
1959 ph10 454 <a href="#subpatternsassubroutines">"Subroutine"</a>
1960     calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
1961 ph10 461 as the subpattern matches a fixed-length string.
1962 ph10 454 <a href="#recursion">Recursion,</a>
1963     however, is not supported.
1964     </P>
1965     <P>
1966 nigel 93 Possessive quantifiers can be used in conjunction with lookbehind assertions to
1967 ph10 461 specify efficient matching of fixed-length strings at the end of subject
1968     strings. Consider a simple pattern such as
1969 nigel 63 <pre>
1970     abcd$
1971 nigel 75 </pre>
1972 nigel 63 when applied to a long string that does not match. Because matching proceeds
1973     from left to right, PCRE will look for each "a" in the subject and then see if
1974     what follows matches the rest of the pattern. If the pattern is specified as
1975     <pre>
1976     ^.*abcd$
1977 nigel 75 </pre>
1978 nigel 63 the initial .* matches the entire string at first, but when this fails (because
1979     there is no following "a"), it backtracks to match all but the last character,
1980     then all but the last two characters, and so on. Once again the search for "a"
1981     covers the entire string, from right to left, so we are no better off. However,
1982     if the pattern is written as
1983     <pre>
1984     ^.*+(?&#60;=abcd)
1985 nigel 75 </pre>
1986 nigel 93 there can be no backtracking for the .*+ item; it can match only the entire
1987 nigel 63 string. The subsequent lookbehind assertion does a single test on the last four
1988     characters. If it fails, the match fails immediately. For long strings, this
1989     approach makes a significant difference to the processing time.
1990     </P>
1991 nigel 75 <br><b>
1992     Using multiple assertions
1993     </b><br>
1994 nigel 63 <P>
1995     Several assertions (of any sort) may occur in succession. For example,
1996     <pre>
1997     (?&#60;=\d{3})(?&#60;!999)foo
1998 nigel 75 </pre>
1999 nigel 63 matches "foo" preceded by three digits that are not "999". Notice that each of
2000     the assertions is applied independently at the same point in the subject
2001     string. First there is a check that the previous three characters are all
2002     digits, and then there is a check that the same three characters are not "999".
2003     This pattern does <i>not</i> match "foo" preceded by six characters, the first
2004     of which are digits and the last three of which are not "999". For example, it
2005     doesn't match "123abcfoo". A pattern to do that is
2006     <pre>
2007     (?&#60;=\d{3}...)(?&#60;!999)foo
2008 nigel 75 </pre>
2009 nigel 63 This time the first assertion looks at the preceding six characters, checking
2010     that the first three are digits, and then the second assertion checks that the
2011     preceding three characters are not "999".
2012     </P>
2013     <P>
2014     Assertions can be nested in any combination. For example,
2015     <pre>
2016     (?&#60;=(?&#60;!foo)bar)baz
2017 nigel 75 </pre>
2018 nigel 63 matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2019     preceded by "foo", while
2020     <pre>
2021     (?&#60;=\d{3}(?!999)...)foo
2022 nigel 75 </pre>
2023     is another pattern that matches "foo" preceded by three digits and any three
2024 nigel 63 characters that are not "999".
2025 nigel 91 <a name="conditions"></a></P>
2026 ph10 227 <br><a name="SEC19" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br>
2027 nigel 63 <P>
2028     It is possible to cause the matching process to obey a subpattern
2029     conditionally or to choose between two alternative subpatterns, depending on
2030 ph10 461 the result of an assertion, or whether a specific capturing subpattern has
2031     already been matched. The two possible forms of conditional subpattern are:
2032 nigel 63 <pre>
2033     (?(condition)yes-pattern)
2034     (?(condition)yes-pattern|no-pattern)
2035 nigel 75 </pre>
2036 nigel 63 If the condition is satisfied, the yes-pattern is used; otherwise the
2037     no-pattern (if present) is used. If there are more than two alternatives in the
2038 ph10 567 subpattern, a compile-time error occurs. Each of the two alternatives may
2039 ph10 579 itself contain nested subpatterns of any form, including conditional
2040 ph10 567 subpatterns; the restriction to two alternatives applies only at the level of
2041 ph10 579 the condition. This pattern fragment is an example where the alternatives are
2042 ph10 567 complex:
2043     <pre>
2044     (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2046     </PRE>
2047 nigel 63 </P>
2048     <P>
2049 nigel 93 There are four kinds of condition: references to subpatterns, references to
2050     recursion, a pseudo-condition called DEFINE, and assertions.
2051 nigel 91 </P>
2052 nigel 93 <br><b>
2053     Checking for a used subpattern by number
2054     </b><br>
2055 nigel 91 <P>
2056 nigel 93 If the text between the parentheses consists of a sequence of digits, the
2057 ph10 461 condition is true if a capturing subpattern of that number has previously
2058     matched. If there is more than one capturing subpattern with the same number
2059     (see the earlier
2060     <a href="#recursion">section about duplicate subpattern numbers),</a>
2061 ph10 579 the condition is true if any of them have matched. An alternative notation is
2062 ph10 461 to precede the digits with a plus or minus sign. In this case, the subpattern
2063     number is relative rather than absolute. The most recently opened parentheses
2064 ph10 579 can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2065     loops it can also make sense to refer to subsequent groups. The next
2066     parentheses to be opened can be referenced as (?(+1), and so on. (The value
2067     zero in any of these forms is not used; it provokes a compile-time error.)
2068 nigel 93 </P>
2069     <P>
2070 nigel 91 Consider the following pattern, which contains non-significant white space to
2071     make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2072     three parts for ease of discussion:
2073 nigel 63 <pre>
2074     ( \( )? [^()]+ (?(1) \) )
2075 nigel 75 </pre>
2076 nigel 63 The first part matches an optional opening parenthesis, and if that
2077     character is present, sets it as the first captured substring. The second part
2078     matches one or more characters that are not parentheses. The third part is a
2079 ph10 579 conditional subpattern that tests whether or not the first set of parentheses
2080     matched. If they did, that is, if subject started with an opening parenthesis,
2081 nigel 63 the condition is true, and so the yes-pattern is executed and a closing
2082     parenthesis is required. Otherwise, since no-pattern is not present, the
2083     subpattern matches nothing. In other words, this pattern matches a sequence of
2084 nigel 93 non-parentheses, optionally enclosed in parentheses.
2085     </P>
2086 ph10 172 <P>
2087     If you were embedding this pattern in a larger one, you could use a relative
2088     reference:
2089     <pre>
2090     ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
2091     </pre>
2092     This makes the fragment independent of the parentheses in the larger pattern.
2093     </P>
2094 nigel 93 <br><b>
2095     Checking for a used subpattern by name
2096     </b><br>
2097     <P>
2098     Perl uses the syntax (?(&#60;name&#62;)...) or (?('name')...) to test for a used
2099     subpattern by name. For compatibility with earlier versions of PCRE, which had
2100     this facility before Perl, the syntax (?(name)...) is also recognized. However,
2101     there is a possible ambiguity with this syntax, because subpattern names may
2102     consist entirely of digits. PCRE looks first for a named subpattern; if it
2103     cannot find one and the name consists entirely of digits, PCRE looks for a
2104     subpattern of that number, which must be greater than zero. Using subpattern
2105     names that consist entirely of digits is not recommended.
2106     </P>
2107     <P>
2108     Rewriting the above example to use a named subpattern gives this:
2109 nigel 91 <pre>
2110 nigel 93 (?&#60;OPEN&#62; \( )? [^()]+ (?(&#60;OPEN&#62;) \) )
2111 ph10 461 </pre>
2112     If the name used in a condition of this kind is a duplicate, the test is
2113     applied to all subpatterns of the same name, and is true if any one of them has
2114     matched.
2115 nigel 93 </P>
2116     <br><b>
2117     Checking for pattern recursion
2118     </b><br>
2119     <P>
2120     If the condition is the string (R), and there is no subpattern with the name R,
2121     the condition is true if a recursive call to the whole pattern or any
2122     subpattern has been made. If digits or a name preceded by ampersand follow the
2123     letter R, for example:
2124     <pre>
2125     (?(R3)...) or (?(R&name)...)
2126 nigel 91 </pre>
2127 ph10 461 the condition is true if the most recent recursion is into a subpattern whose
2128 nigel 93 number or name is given. This condition does not check the entire recursion
2129 ph10 461 stack. If the name used in a condition of this kind is a duplicate, the test is
2130     applied to all subpatterns of the same name, and is true if any one of them is
2131     the most recent recursion.
2132 nigel 63 </P>
2133     <P>
2134 ph10 461 At "top level", all these recursion test conditions are false.
2135     <a href="#recursion">The syntax for recursive patterns</a>
2136     is described below.
2137 ph10 572 <a name="subdefine"></a></P>
2138 nigel 93 <br><b>
2139     Defining subpatterns for use by reference only
2140     </b><br>
2141     <P>
2142     If the condition is the string (DEFINE), and there is no subpattern with the
2143     name DEFINE, the condition is always false. In this case, there may be only one
2144     alternative in the subpattern. It is always skipped if control reaches this
2145     point in the pattern; the idea of DEFINE is that it can be used to define
2146 ph10 733 subroutines that can be referenced from elsewhere. (The use of
2147     <a href="#subpatternsassubroutines">subroutines</a>
2148 ph10 579 is described below.) For example, a pattern to match an IPv4 address such as
2149     "" could be written like this (ignore whitespace and line
2150     breaks):
2151 nigel 93 <pre>
2152     (?(DEFINE) (?&#60;byte&#62; 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
2153     \b (?&byte) (\.(?&byte)){3} \b
2154     </pre>
2155     The first part of the pattern is a DEFINE group inside which a another group
2156     named "byte" is defined. This matches an individual component of an IPv4
2157     address (a number less than 256). When matching takes place, this part of the
2158 ph10 461 pattern is skipped because DEFINE acts like a false condition. The rest of the
2159     pattern uses references to the named group to match the four dot-separated
2160     components of an IPv4 address, insisting on a word boundary at each end.
2161 nigel 93 </P>
2162     <br><b>
2163     Assertion conditions
2164     </b><br>
2165     <P>
2166     If the condition is not in any of the above formats, it must be an assertion.
2167 nigel 63 This may be a positive or negative lookahead or lookbehind assertion. Consider
2168     this pattern, again containing non-significant white space, and with the two
2169     alternatives on the second line:
2170     <pre>
2171     (?(?=[^a-z]*[a-z])
2172     \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
2173 nigel 75 </pre>
2174 nigel 63 The condition is a positive lookahead assertion that matches an optional
2175     sequence of non-letters followed by a letter. In other words, it tests for the
2176     presence of at least one letter in the subject. If a letter is found, the
2177     subject is matched against the first alternative; otherwise it is matched
2178     against the second. This pattern matches strings in one of the two forms
2179     dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2180 nigel 75 <a name="comments"></a></P>
2181 ph10 227 <br><a name="SEC20" href="#TOC1">COMMENTS</a><br>
2182 nigel 63 <P>
2183 ph10 579 There are two ways of including comments in patterns that are processed by
2184 ph10 567 PCRE. In both cases, the start of the comment must not be in a character class,
2185     nor in the middle of any other sequence of related characters such as (?: or a
2186     subpattern name or number. The characters that make up a comment play no part
2187     in the pattern matching.
2188 nigel 63 </P>
2189     <P>
2190 ph10 567 The sequence (?# marks the start of a comment that continues up to the next
2191     closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2192     option is set, an unescaped # character also introduces a comment, which in
2193     this case continues to immediately after the next newline character or
2194     character sequence in the pattern. Which characters are interpreted as newlines
2195     is controlled by the options passed to <b>pcre_compile()</b> or by a special
2196     sequence at the start of the pattern, as described in the section entitled
2197 ph10 572 <a href="#newlines">"Newline conventions"</a>
2198     above. Note that the end of this type of comment is a literal newline sequence
2199     in the pattern; escape sequences that happen to represent a newline do not
2200     count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2201     default newline convention is in force:
2202 ph10 567 <pre>
2203     abc #comment \n still comment
2204     </pre>
2205 ph10 579 On encountering the # character, <b>pcre_compile()</b> skips along, looking for
2206 ph10 567 a newline in the pattern. The sequence \n is still literal at this stage, so
2207     it does not terminate the comment. Only an actual character with the code value
2208     0x0a (the default newline) does so.
2209 nigel 91 <a name="recursion"></a></P>
2210 ph10 227 <br><a name="SEC21" href="#TOC1">RECURSIVE PATTERNS</a><br>
2211 nigel 63 <P>
2212     Consider the problem of matching a string in parentheses, allowing for
2213     unlimited nested parentheses. Without the use of recursion, the best that can
2214     be done is to use a pattern that matches up to some fixed depth of nesting. It
2215 nigel 93 is not possible to handle an arbitrary nesting depth.
2216     </P>
2217     <P>
2218     For some time, Perl has provided a facility that allows regular expressions to
2219     recurse (amongst other things). It does this by interpolating Perl code in the
2220     expression at run time, and the code can refer to the expression itself. A Perl
2221     pattern using code interpolation to solve the parentheses problem can be
2222     created like this:
2223 nigel 63 <pre>
2224     $re = qr{\( (?: (?&#62;[^()]+) | (?p{$re}) )* \)}x;
2225 nigel 75 </pre>
2226 nigel 63 The (?p{...}) item interpolates Perl code at run time, and in this case refers
2227 nigel 93 recursively to the pattern in which it appears.
2228 nigel 63 </P>
2229     <P>
2230 nigel 93 Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2231     supports special syntax for recursion of the entire pattern, and also for
2232     individual subpattern recursion. After its introduction in PCRE and Python,
2233 ph10 453 this kind of recursion was subsequently introduced into Perl at release 5.10.
2234 nigel 63 </P>
2235     <P>
2236 nigel 93 A special item that consists of (? followed by a number greater than zero and a
2237 ph10 733 closing parenthesis is a recursive subroutine call of the subpattern of the
2238     given number, provided that it occurs inside that subpattern. (If not, it is a
2239     <a href="#subpatternsassubroutines">non-recursive subroutine</a>
2240 nigel 93 call, which is described in the next section.) The special item (?R) or (?0) is
2241     a recursive call of the entire regular expression.
2242 nigel 87 </P>
2243     <P>
2244     This PCRE pattern solves the nested parentheses problem (assume the
2245     PCRE_EXTENDED option is set so that white space is ignored):
2246 nigel 63 <pre>
2247 ph10 461 \( ( [^()]++ | (?R) )* \)
2248 nigel 75 </pre>
2249 nigel 63 First it matches an opening parenthesis. Then it matches any number of
2250     substrings which can either be a sequence of non-parentheses, or a recursive
2251 nigel 87 match of the pattern itself (that is, a correctly parenthesized substring).
2252 ph10 461 Finally there is a closing parenthesis. Note the use of a possessive quantifier
2253     to avoid backtracking into sequences of non-parentheses.
2254 nigel 63 </P>
2255     <P>
2256     If this were part of a larger pattern, you would not want to recurse the entire
2257     pattern, so instead you could use this:
2258     <pre>
2259 ph10 461 ( \( ( [^()]++ | (?1) )* \) )
2260 nigel 75 </pre>
2261 nigel 63 We have put the pattern into parentheses, and caused the recursion to refer to
2262 ph10 172 them instead of the whole pattern.
2263     </P>
2264     <P>
2265     In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2266 ph10 572 is made easier by the use of relative references. Instead of (?1) in the
2267     pattern above you can write (?-2) to refer to the second most recently opened
2268     parentheses preceding the recursion. In other words, a negative number counts
2269     capturing parentheses leftwards from the point at which it is encountered.
2270 ph10 172 </P>
2271     <P>
2272     It is also possible to refer to subsequently opened parentheses, by writing
2273     references such as (?+2). However, these cannot be recursive because the
2274     reference is not inside the parentheses that are referenced. They are always
2275 ph10 733 <a href="#subpatternsassubroutines">non-recursive subroutine</a>
2276 ph10 454 calls, as described in the next section.
2277 ph10 172 </P>
2278     <P>
2279     An alternative approach is to use named parentheses instead. The Perl syntax
2280     for this is (?&name); PCRE's earlier syntax (?P&#62;name) is also supported. We
2281     could rewrite the above example as follows:
2282 nigel 63 <pre>
2283 ph10 461 (?&#60;pn&#62; \( ( [^()]++ | (?&pn) )* \) )
2284 nigel 75 </pre>
2285 nigel 93 If there is more than one subpattern with the same name, the earliest one is
2286 ph10 172 used.
2287     </P>
2288     <P>
2289     This particular example pattern that we have been looking at contains nested
2290 ph10 461 unlimited repeats, and so the use of a possessive quantifier for matching
2291     strings of non-parentheses is important when applying the pattern to strings
2292     that do not match. For example, when this pattern is applied to
2293 nigel 63 <pre>
2294     (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2295 nigel 75 </pre>
2296 ph10 461 it yields "no match" quickly. However, if a possessive quantifier is not used,
2297 nigel 63 the match runs for a very long time indeed because there are so many different
2298     ways the + and * repeats can carve up the subject, and all have to be tested
2299     before failure can be reported.
2300     </P>
2301     <P>
2302 ph10 469 At the end of a match, the values of capturing parentheses are those from
2303     the outermost level. If you want to obtain intermediate values, a callout
2304     function can be used (see below and the
2305 nigel 63 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2306     documentation). If the pattern above is matched against
2307     <pre>
2308     (ab(cd)ef)
2309 nigel 75 </pre>
2310 ph10 469 the value for the inner capturing parentheses (numbered 2) is "ef", which is
2311     the last value taken on at the top level. If a capturing subpattern is not
2312 ph10 733 matched at the top level, its final captured value is unset, even if it was
2313     (temporarily) set at a deeper level during the matching process.
2314 nigel 63 </P>
2315     <P>
2316 ph10 469 If there are more than 15 capturing parentheses in a pattern, PCRE has to
2317     obtain extra memory to store data during a recursion, which it does by using
2318     <b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no memory can
2319     be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2320     </P>
2321     <P>
2322 nigel 63 Do not confuse the (?R) item with the condition (R), which tests for recursion.
2323     Consider this pattern, which matches text in angle brackets, allowing for
2324     arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2325     recursing), whereas any characters are permitted at the outer level.
2326     <pre>
2327     &#60; (?: (?(R) \d++ | [^&#60;&#62;]*+) | (?R)) * &#62;
2328 nigel 75 </pre>
2329 nigel 63 In this pattern, (?(R) is the start of a conditional subpattern, with two
2330     different alternatives for the recursive and non-recursive cases. The (?R) item
2331     is the actual recursive call.
2332 ph10 453 <a name="recursiondifference"></a></P>
2333     <br><b>
2334 ph10 733 Differences in recursion processing between PCRE and Perl
2335 ph10 453 </b><br>
2336     <P>
2337 ph10 733 Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2338     (like Python, but unlike Perl), a recursive subpattern call is always treated
2339     as an atomic group. That is, once it has matched some of the subject string, it
2340     is never re-entered, even if it contains untried alternatives and there is a
2341     subsequent matching failure. This can be illustrated by the following pattern,
2342     which purports to match a palindromic string that contains an odd number of
2343     characters (for example, "a", "aba", "abcba", "abcdcba"):
2344 ph10 453 <pre>
2345     ^(.|(.)(?1)\2)$
2346     </pre>
2347 ph10 461 The idea is that it either matches a single character, or two identical
2348     characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2349 ph10 453 it does not if the pattern is longer than three characters. Consider the
2350     subject string "abcba":
2351     </P>
2352     <P>
2353 ph10 461 At the top level, the first character is matched, but as it is not at the end
2354 ph10 453 of the string, the first alternative fails; the second alternative is taken
2355     and the recursion kicks in. The recursive call to subpattern 1 successfully
2356     matches the next character ("b"). (Note that the beginning and end of line
2357     tests are not part of the recursion).
2358     </P>
2359     <P>
2360     Back at the top level, the next character ("c") is compared with what
2361 ph10 461 subpattern 2 matched, which was "a". This fails. Because the recursion is
2362 ph10 453 treated as an atomic group, there are now no backtracking points, and so the
2363     entire match fails. (Perl is able, at this point, to re-enter the recursion and
2364     try the second alternative.) However, if the pattern is written with the
2365     alternatives in the other order, things are different:
2366     <pre>
2367     ^((.)(?1)\2|.)$
2368     </pre>
2369 ph10 461 This time, the recursing alternative is tried first, and continues to recurse
2370     until it runs out of characters, at which point the recursion fails. But this
2371     time we do have another alternative to try at the higher level. That is the big
2372 ph10 453 difference: in the previous case the remaining alternative is at a deeper
2373     recursion level, which PCRE cannot use.
2374     </P>
2375     <P>
2376 ph10 579 To change the pattern so that it matches all palindromic strings, not just
2377     those with an odd number of characters, it is tempting to change the pattern to
2378     this:
2379 ph10 453 <pre>
2380     ^((.)(?1)\2|.?)$
2381     </pre>
2382 ph10 461 Again, this works in Perl, but not in PCRE, and for the same reason. When a
2383     deeper recursion has matched a single character, it cannot be entered again in
2384     order to match an empty string. The solution is to separate the two cases, and
2385 ph10 453 write out the odd and even cases as alternatives at the higher level:
2386     <pre>
2387     ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
2388     </pre>
2389 ph10 461 If you want to match typical palindromic phrases, the pattern has to ignore all
2390 ph10 453 non-word characters, which can be done like this:
2391     <pre>
2392     ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
2393     </pre>
2394 ph10 461 If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2395     man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2396     the use of the possessive quantifier *+ to avoid backtracking into sequences of
2397 ph10 453 non-word characters. Without this, PCRE takes a great deal longer (ten times or
2398     more) to match typical phrases, and Perl takes so long that you think it has
2399     gone into a loop.
2400 ph10 461 </P>
2401     <P>
2402     <b>WARNING</b>: The palindrome-matching patterns above work only if the subject
2403     string does not start with a palindrome that is shorter than the entire string.
2404     For example, although "abcba" is correctly matched, if the subject is "ababa",
2405     PCRE finds the palindrome "aba" at the start, then fails at top level because
2406     the end of the string does not follow. Once again, it cannot jump back into the
2407     recursion to try other alternatives, so the entire match fails.
2408 ph10 733 </P>
2409     <P>
2410     The second way in which PCRE and Perl differ in their recursion processing is
2411     in the handling of captured values. In Perl, when a subpattern is called
2412     recursively or as a subpattern (see the next section), it has no access to any
2413     values that were captured outside the recursion, whereas in PCRE these values
2414     can be referenced. Consider this pattern:
2415     <pre>
2416     ^(.)(\1|a(?2))
2417     </pre>
2418     In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2419     then in the second group, when the back reference \1 fails to match "b", the
2420     second alternative matches "a" and then recurses. In the recursion, \1 does
2421     now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2422     match because inside the recursive call \1 cannot access the externally set
2423     value.
2424 nigel 75 <a name="subpatternsassubroutines"></a></P>
2425 ph10 227 <br><a name="SEC22" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br>
2426 nigel 63 <P>
2427 ph10 733 If the syntax for a recursive subpattern call (either by number or by
2428 nigel 63 name) is used outside the parentheses to which it refers, it operates like a
2429 ph10 733 subroutine in a programming language. The called subpattern may be defined
2430 ph10 172 before or after the reference. A numbered reference can be absolute or
2431     relative, as in these examples:
2432 nigel 63 <pre>
2433 ph10 172 (...(absolute)...)...(?2)...
2434     (...(relative)...)...(?-1)...
2435     (...(?+1)...(relative)...
2436     </pre>
2437     An earlier example pointed out that the pattern
2438     <pre>
2439 nigel 63 (sens|respons)e and \1ibility
2440 nigel 75 </pre>
2441 nigel 63 matches "sense and sensibility" and "response and responsibility", but not
2442     "sense and responsibility". If instead the pattern
2443     <pre>
2444     (sens|respons)e and (?1)ibility
2445 nigel 75 </pre>
2446 nigel 63 is used, it does match "sense and responsibility" as well as the other two
2447 nigel 93 strings. Another example is given in the discussion of DEFINE above.
2448 nigel 63 </P>
2449 nigel 87 <P>
2450 ph10 733 All subroutine calls, whether recursive or not, are always treated as atomic
2451     groups. That is, once a subroutine has matched some of the subject string, it
2452     is never re-entered, even if it contains untried alternatives and there is a
2453     subsequent matching failure. Any capturing parentheses that are set during the
2454     subroutine call revert to their previous values afterwards.
2455 nigel 87 </P>
2456 nigel 63 <P>
2457 ph10 733 Processing options such as case-independence are fixed when a subpattern is
2458     defined, so if it is used as a subroutine, such options cannot be changed for
2459     different calls. For example, consider this pattern:
2460 nigel 93 <pre>
2461 ph10 172 (abc)(?i:(?-1))
2462 nigel 93 </pre>
2463     It matches "abcabc". It does not match "abcABC" because the change of
2464     processing option does not affect the called subpattern.
2465 ph10 345 <a name="onigurumasubroutines"></a></P>
2466     <br><a name="SEC23" href="#TOC1">ONIGURUMA SUBROUTINE SYNTAX</a><br>
2467     <P>
2468     For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or
2469     a number enclosed either in angle brackets or single quotes, is an alternative
2470     syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2471     are two of the examples used above, rewritten using this syntax:
2472     <pre>
2473     (?&#60;pn&#62; \( ( (?&#62;[^()]+) | \g&#60;pn&#62; )* \) )
2474     (sens|respons)e and \g'1'ibility
2475     </pre>
2476     PCRE supports an extension to Oniguruma: if a number is preceded by a
2477     plus or a minus sign it is taken as a relative reference. For example:
2478     <pre>
2479     (abc)(?i:\g&#60;-1&#62;)
2480     </pre>
2481     Note that \g{...} (Perl syntax) and \g&#60;...&#62; (Oniguruma syntax) are <i>not</i>
2482     synonymous. The former is a back reference; the latter is a subroutine call.
2483 nigel 93 </P>
2484 ph10 345 <br><a name="SEC24" href="#TOC1">CALLOUTS</a><br>
2485 nigel 93 <P>
2486 nigel 63 Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl
2487     code to be obeyed in the middle of matching a regular expression. This makes it
2488     possible, amongst other things, to extract different substrings that match the
2489     same pair of parentheses when there is a repetition.
2490     </P>
2491     <P>
2492     PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2493     code. The feature is called "callout". The caller of PCRE provides an external
2494     function by putting its entry point in the global variable <i>pcre_callout</i>.
2495     By default, this variable contains NULL, which disables all calling out.
2496     </P>
2497     <P>
2498     Within a regular expression, (?C) indicates the points at which the external
2499     function is to be called. If you want to identify different callout points, you
2500     can put a number less than 256 after the letter C. The default value is zero.
2501     For example, this pattern has two callout points:
2502     <pre>
2503 ph10 155 (?C1)abc(?C2)def
2504 nigel 75 </pre>
2505     If the PCRE_AUTO_CALLOUT flag is passed to <b>pcre_compile()</b>, callouts are
2506     automatically installed before each item in the pattern. They are all numbered
2507     255.
2508 nigel 63 </P>
2509     <P>
2510     During matching, when PCRE reaches a callout point (and <i>pcre_callout</i> is
2511     set), the external function is called. It is provided with the number of the
2512 nigel 75 callout, the position in the pattern, and, optionally, one item of data
2513     originally supplied by the caller of <b>pcre_exec()</b>. The callout function
2514     may cause matching to proceed, to backtrack, or to fail altogether. A complete
2515     description of the interface to the callout function is given in the
2516 nigel 63 <a href="pcrecallout.html"><b>pcrecallout</b></a>
2517     documentation.
2518 ph10 512 <a name="backtrackcontrol"></a></P>
2519 ph10 345 <br><a name="SEC25" href="#TOC1">BACKTRACKING CONTROL</a><br>
2520 nigel 63 <P>
2521 ph10 211 Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2522     are described in the Perl documentation as "experimental and subject to change
2523     or removal in a future version of Perl". It goes on to say: "Their usage in
2524     production code should be noted to avoid problems during upgrades." The same
2525     remarks apply to the PCRE features described in this section.
2526     </P>
2527     <P>
2528 ph10 345 Since these verbs are specifically related to backtracking, most of them can be
2529     used only when the pattern is to be matched using <b>pcre_exec()</b>, which uses
2530     a backtracking algorithm. With the exception of (*FAIL), which behaves like a
2531     failing negative assertion, they cause an error if encountered by
2532 ph10 211 <b>pcre_dfa_exec()</b>.
2533     </P>
2534     <P>
2535 ph10 733 If any of these verbs are used in an assertion or in a subpattern that is
2536     called as a subroutine (whether or not recursively), their effect is confined
2537     to that subpattern; it does not extend to the surrounding pattern, with one
2538 ph10 836 exception: the name from a *(MARK), (*PRUNE), or (*THEN) that is encountered in
2539     a successful positive assertion <i>is</i> passed back when a match succeeds
2540     (compare capturing parentheses in assertions). Note that such subpatterns are
2541     processed as anchored at the point where they are tested. Note also that Perl's
2542     treatment of subroutines is different in some cases.
2543 ph10 453 </P>
2544     <P>
2545 ph10 211 The new verbs make use of what was previously invalid syntax: an opening
2546 ph10 512 parenthesis followed by an asterisk. They are generally of the form
2547     (*VERB) or (*VERB:NAME). Some may take either form, with differing behaviour,
2548 ph10 733 depending on whether or not an argument is present. A name is any sequence of
2549     characters that does not include a closing parenthesis. If the name is empty,
2550     that is, if the closing parenthesis immediately follows the colon, the effect
2551     is as if the colon were not there. Any number of these verbs may occur in a
2552     pattern.
2553 ph10 211 </P>
2554 ph10 512 <P>
2555     PCRE contains some optimizations that are used to speed up matching by running
2556     some checks at the start of each match attempt. For example, it may know the
2557     minimum length of matching subject, or that a particular character must be
2558     present. When one of these optimizations suppresses the running of a match, any
2559     included backtracking verbs will not, of course, be processed. You can suppress
2560     the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2561 ph10 579 when calling <b>pcre_compile()</b> or <b>pcre_exec()</b>, or by starting the
2562     pattern with (*NO_START_OPT).
2563 ph10 512 </P>
2564 ph10 836 <P>
2565     Experiments with Perl suggest that it too has similar optimizations, sometimes
2566     leading to anomalous results.
2567     </P>
2568 ph10 211 <br><b>
2569     Verbs that act immediately
2570     </b><br>
2571     <P>
2572 ph10 512 The following verbs act as soon as they are encountered. They may not be
2573     followed by a name.
2574 ph10 211 <pre>
2575     (*ACCEPT)
2576     </pre>
2577     This verb causes the match to end successfully, skipping the remainder of the
2578 ph10 733 pattern. However, when it is inside a subpattern that is called as a
2579     subroutine, only that subpattern is ended successfully. Matching then continues
2580     at the outer level. If (*ACCEPT) is inside capturing parentheses, the data so
2581     far is captured. For example:
2582 ph10 211 <pre>
2583 ph10 453 A((?:A|B(*ACCEPT)|C)D)
2584 ph10 211 </pre>
2585 ph10 461 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2586 ph10 453 the outer parentheses.
2587 ph10 211 <pre>
2588     (*FAIL) or (*F)
2589     </pre>
2590 ph10 733 This verb causes a matching failure, forcing backtracking to occur. It is
2591 ph10 211 equivalent to (?!) but easier to read. The Perl documentation notes that it is
2592     probably useful only when combined with (?{}) or (??{}). Those are, of course,
2593     Perl features that are not present in PCRE. The nearest equivalent is the
2594     callout feature, as for example in this pattern:
2595     <pre>
2596     a+(?C)(*FAIL)
2597     </pre>
2598     A match with the string "aaaa" always fails, but the callout is taken before
2599     each backtrack happens (in this example, 10 times).
2600     </P>
2601     <br><b>
2602 ph10 512 Recording which path was taken
2603     </b><br>
2604     <P>
2605     There is one verb whose main purpose is to track how a match was arrived at,
2606     though it also has a secondary use in conjunction with advancing the match
2607     starting point (see (*SKIP) below).
2608     <pre>
2609     (*MARK:NAME) or (*:NAME)
2610     </pre>
2611     A name is always required with this verb. There may be as many instances of
2612     (*MARK) as you like in a pattern, and their names do not have to be unique.
2613     </P>
2614     <P>
2615 ph10 836 When a match succeeds, the name of the last-encountered (*MARK) on the matching
2616     path is passed back to the caller via the <i>pcre_extra</i> data structure, as
2617     described in the
2618 ph10 512 <a href="pcreapi.html#extradata">section on <i>pcre_extra</i></a>
2619     in the
2620     <a href="pcreapi.html"><b>pcreapi</b></a>
2621 ph10 836 documentation. Here is an example of <b>pcretest</b> output, where the /K
2622     modifier requests the retrieval and outputting of (*MARK) data:
2623 ph10 512 <pre>
2624 ph10 836 re&#62; /X(*MARK:A)Y|X(*MARK:B)Z/K
2625     data&#62; XY
2626 ph10 512 0: XY
2627     MK: A
2628     XZ
2629     0: XZ
2630     MK: B
2631     </pre>
2632     The (*MARK) name is tagged with "MK:" in this output, and in this example it
2633     indicates which of the two alternatives matched. This is a more efficient way
2634     of obtaining this information than putting each alternative in its own
2635     capturing parentheses.
2636     </P>
2637     <P>
2638 ph10 654 If (*MARK) is encountered in a positive assertion, its name is recorded and
2639     passed back if it is the last-encountered. This does not happen for negative
2640 ph10 733 assertions.
2641 ph10 654 </P>
2642     <P>
2643 ph10 836 After a partial match or a failed match, the name of the last encountered
2644     (*MARK) in the entire match process is returned. For example:
2645 ph10 512 <pre>
2646 ph10 836 re&#62; /X(*MARK:A)Y|X(*MARK:B)Z/K
2647     data&#62; XP
2648 ph10 512 No match, mark = B
2649     </pre>
2650 ph10 836 Note that in this unanchored example the mark is retained from the match
2651     attempt that started at the letter "X". Subsequent match attempts starting at
2652     "P" and then with an empty string do not get as far as the (*MARK) item, but
2653     nevertheless do not reset it.
2654 ph10 512 </P>
2655     <br><b>
2656 ph10 211 Verbs that act after backtracking
2657     </b><br>
2658     <P>
2659     The following verbs do nothing when they are encountered. Matching continues
2660 ph10 512 with what follows, but if there is no subsequent match, causing a backtrack to
2661     the verb, a failure is forced. That is, backtracking cannot pass to the left of
2662     the verb. However, when one of these verbs appears inside an atomic group, its
2663     effect is confined to that group, because once the group has been matched,
2664     there is never any backtracking into it. In this situation, backtracking can
2665     "jump back" to the left of the entire atomic group. (Remember also, as stated
2666     above, that this localization also applies in subroutine calls and assertions.)
2667     </P>
2668     <P>
2669     These verbs differ in exactly what kind of failure occurs when backtracking
2670     reaches them.
2671 ph10 211 <pre>
2672     (*COMMIT)
2673     </pre>
2674 ph10 512 This verb, which may not be followed by a name, causes the whole match to fail
2675     outright if the rest of the pattern does not match. Even if the pattern is
2676     unanchored, no further attempts to find a match by advancing the starting point
2677     take place. Once (*COMMIT) has been passed, <b>pcre_exec()</b> is committed to
2678     finding a match at the current starting point, or not at all. For example:
2679 ph10 211 <pre>
2680     a+(*COMMIT)b
2681     </pre>
2682     This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2683 ph10 512 dynamic anchor, or "I've started, so I must finish." The name of the most
2684     recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2685     match failure.
2686     </P>
2687     <P>
2688     Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
2689     unless PCRE's start-of-match optimizations are turned off, as shown in this
2690     <b>pcretest</b> example:
2691 ph10 211 <pre>
2692 ph10 836 re&#62; /(*COMMIT)abc/
2693     data&#62; xyzabc
2694 ph10 512 0: abc
2695     xyzabc\Y
2696     No match
2697 ph10 211 </pre>
2698 ph10 512 PCRE knows that any match must start with "a", so the optimization skips along
2699     the subject to "a" before running the first match attempt, which succeeds. When
2700     the optimization is disabled by the \Y escape in the second subject, the match
2701     starts at "x" and so the (*COMMIT) causes it to fail without trying any other
2702     starting points.
2703 ph10 211 <pre>
2704 ph10 512 (*PRUNE) or (*PRUNE:NAME)
2705     </pre>
2706     This verb causes the match to fail at the current starting position in the
2707     subject if the rest of the pattern does not match. If the pattern is
2708     unanchored, the normal "bumpalong" advance to the next starting character then
2709     happens. Backtracking can occur as usual to the left of (*PRUNE), before it is
2710     reached, or when matching to the right of (*PRUNE), but if there is no match to
2711     the right, backtracking cannot cross (*PRUNE). In simple cases, the use of
2712     (*PRUNE) is just an alternative to an atomic group or possessive quantifier,
2713     but there are some uses of (*PRUNE) that cannot be expressed in any other way.
2714 ph10 836 The behaviour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE). In an
2715     anchored pattern (*PRUNE) has the same effect as (*COMMIT).
2716 ph10 512 <pre>
2717 ph10 211 (*SKIP)
2718     </pre>
2719 ph10 512 This verb, when given without a name, is like (*PRUNE), except that if the
2720     pattern is unanchored, the "bumpalong" advance is not to the next character,
2721     but to the position in the subject where (*SKIP) was encountered. (*SKIP)
2722     signifies that whatever text was matched leading up to it cannot be part of a
2723     successful match. Consider:
2724 ph10 211 <pre>
2725     a+(*SKIP)b
2726     </pre>
2727     If the subject is "aaaac...", after the first match attempt fails (starting at
2728     the first character in the string), the starting point skips on to start the
2729     next attempt at "c". Note that a possessive quantifer does not have the same
2730 ph10 461 effect as this example; although it would suppress backtracking during the
2731 ph10 211 first match attempt, the second attempt would start at the second character
2732     instead of skipping on to "c".
2733     <pre>
2734 ph10 512 (*SKIP:NAME)
2735 ph10 211 </pre>
2736 ph10 512 When (*SKIP) has an associated name, its behaviour is modified. If the
2737     following pattern fails to match, the previous path through the pattern is
2738     searched for the most recent (*MARK) that has the same name. If one is found,
2739     the "bumpalong" advance is to the subject position that corresponds to that
2740     (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
2741 ph10 836 matching name is found, the (*SKIP) is ignored.
2742 ph10 512 <pre>
2743     (*THEN) or (*THEN:NAME)
2744     </pre>
2745 ph10 733 This verb causes a skip to the next innermost alternative if the rest of the
2746     pattern does not match. That is, it cancels pending backtracking, but only
2747     within the current alternative. Its name comes from the observation that it can
2748     be used for a pattern-based if-then-else block:
2749 ph10 211 <pre>
2750     ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
2751     </pre>
2752     If the COND1 pattern matches, FOO is tried (and possibly further items after
2753 ph10 733 the end of the group if FOO succeeds); on failure, the matcher skips to the
2754 ph10 512 second alternative and tries COND2, without backtracking into COND1. The
2755 ph10 836 behaviour of (*THEN:NAME) is exactly the same as (*MARK:NAME)(*THEN).
2756     If (*THEN) is not inside an alternation, it acts like (*PRUNE).
2757 ph10 211 </P>
2758 ph10 567 <P>
2759 ph10 733 Note that a subpattern that does not contain a | character is just a part of
2760     the enclosing alternative; it is not a nested alternation with only one
2761     alternative. The effect of (*THEN) extends beyond such a subpattern to the
2762     enclosing alternative. Consider this pattern, where A, B, etc. are complex
2763     pattern fragments that do not contain any | characters at this level:
2764     <pre>
2765     A (B(*THEN)C) | D
2766     </pre>
2767     If A and B are matched, but there is a failure in C, matching does not
2768     backtrack into A; instead it moves to the next alternative, that is, D.
2769     However, if the subpattern containing (*THEN) is given an alternative, it
2770     behaves differently:
2771     <pre>
2772     A (B(*THEN)C | (*FAIL)) | D
2773     </pre>
2774     The effect of (*THEN) is now confined to the inner subpattern. After a failure
2775     in C, matching moves to (*FAIL), which causes the whole subpattern to fail
2776     because there are no more alternatives to try. In this case, matching does now
2777     backtrack into A.
2778 ph10 567 </P>
2779     <P>
2780 ph10 733 Note also that a conditional subpattern is not considered as having two
2781     alternatives, because only one is ever used. In other words, the | character in
2782     a conditional subpattern has a different meaning. Ignoring white space,
2783     consider:
2784 ph10 567 <pre>
2785 ph10 733 ^.*? (?(?=a) a | b(*THEN)c )
2786     </pre>
2787     If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
2788     it initially matches zero characters. The condition (?=a) then fails, the
2789     character "b" is matched, but "c" is not. At this point, matching does not
2790     backtrack to .*? as might perhaps be expected from the presence of the |
2791     character. The conditional subpattern is part of the single alternative that
2792     comprises the whole pattern, and so the match fails. (If there was a backtrack
2793     into .*?, allowing it to match "b", the match would succeed.)
2794     </P>
2795     <P>
2796     The verbs just described provide four different "strengths" of control when
2797     subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
2798     next alternative. (*PRUNE) comes next, failing the match at the current
2799     starting position, but allowing an advance to the next character (for an
2800     unanchored pattern). (*SKIP) is similar, except that the advance may be more
2801     than one character. (*COMMIT) is the strongest, causing the entire match to
2802     fail.
2803     </P>
2804     <P>
2805     If more than one such verb is present in a pattern, the "strongest" one wins.
2806     For example, consider this pattern, where A, B, etc. are complex pattern
2807     fragments:
2808     <pre>
2809 ph10 567 (A(*COMMIT)B(*THEN)C|D)
2810     </pre>
2811     Once A has matched, PCRE is committed to this match, at the current starting
2812     position. If subsequently B matches, but C does not, the normal (*THEN) action
2813 ph10 733 of trying the next alternative (that is, D) does not happen because (*COMMIT)
2814 ph10 567 overrides.
2815     </P>
2816 ph10 345 <br><a name="SEC26" href="#TOC1">SEE ALSO</a><br>
2817 ph10 211 <P>
2818 ph10 461 <b>pcreapi</b>(3), <b>pcrecallout</b>(3), <b>pcrematching</b>(3),
2819     <b>pcresyntax</b>(3), <b>pcre</b>(3).
2820 nigel 93 </P>
2821 ph10 345 <br><a name="SEC27" href="#TOC1">AUTHOR</a><br>
2822 nigel 93 <P>
2823 ph10 99 Philip Hazel
2824 nigel 63 <br>
2825 ph10 99 University Computing Service
2826     <br>
2827     Cambridge CB2 3QH, England.
2828     <br>
2829     </P>
2830 ph10 345 <br><a name="SEC28" href="#TOC1">REVISION</a><br>
2831 ph10 99 <P>
2832 ph10 836 Last updated: 29 November 2011
2833 ph10 99 <br>
2834 ph10 654 Copyright &copy; 1997-2011 University of Cambridge.
2835 ph10 99 <br>
2836 nigel 75 <p>
2837     Return to the <a href="index.html">PCRE index page</a>.
2838     </p>


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

ViewVC Help
Powered by ViewVC 1.1.12