/[pcre]/code/trunk/doc/pcrepattern.3
ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 63 by nigel, Sat Feb 24 21:40:03 2007 UTC revision 1401 by ph10, Tue Nov 12 17:44:07 2013 UTC
# Line 1  Line 1 
1  .TH PCRE 3  .TH PCREPATTERN 3 "12 November 2013" "PCRE 8.34"
2  .SH NAME  .SH NAME
3  PCRE - Perl-compatible regular expressions  PCRE - Perl-compatible regular expressions
4  .SH PCRE REGULAR EXPRESSION DETAILS  .SH "PCRE REGULAR EXPRESSION DETAILS"
5  .rs  .rs
6  .sp  .sp
7  The syntax and semantics of the regular expressions supported by PCRE are  The syntax and semantics of the regular expressions that are supported by PCRE
8  described below. Regular expressions are also described in the Perl  are described in detail below. There is a quick-reference syntax summary in the
9  documentation and in a number of other books, some of which have copious  .\" HREF
10  examples. Jeffrey Friedl's "Mastering Regular Expressions", published by  \fBpcresyntax\fP
11  O'Reilly, covers them in great detail. The description here is intended as  .\"
12  reference documentation.  page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE
13    also supports some alternative regular expression syntax (which does not
14  The basic operation of PCRE is on strings of bytes. However, there is also  conflict with the Perl syntax) in order to provide some compatibility with
15  support for UTF-8 character strings. To use this support you must build PCRE to  regular expressions in Python, .NET, and Oniguruma.
16  include UTF-8 support, and then call \fBpcre_compile()\fR with the PCRE_UTF8  .P
17  option. How this affects the pattern matching is mentioned in several places  Perl's regular expressions are described in its own documentation, and
18  below. There is also a summary of UTF-8 features in the  regular expressions in general are covered in a number of books, some of which
19  .\" HTML <a href="pcre.html#utf8support">  have copious examples. Jeffrey Friedl's "Mastering Regular Expressions",
20  .\" </a>  published by O'Reilly, covers regular expressions in great detail. This
21  section on UTF-8 support  description of PCRE's regular expressions is intended as reference material.
22    .P
23    This document discusses the patterns that are supported by PCRE when one its
24    main matching functions, \fBpcre_exec()\fP (8-bit) or \fBpcre[16|32]_exec()\fP
25    (16- or 32-bit), is used. PCRE also has alternative matching functions,
26    \fBpcre_dfa_exec()\fP and \fBpcre[16|32_dfa_exec()\fP, which match using a
27    different algorithm that is not Perl-compatible. Some of the features discussed
28    below are not available when DFA matching is used. The advantages and
29    disadvantages of the alternative functions, and how they differ from the normal
30    functions, are discussed in the
31    .\" HREF
32    \fBpcrematching\fP
33  .\"  .\"
34  in the main  page.
35    .
36    .
37    .SH "SPECIAL START-OF-PATTERN ITEMS"
38    .rs
39    .sp
40    A number of options that can be passed to \fBpcre_compile()\fP can also be set
41    by special items at the start of a pattern. These are not Perl-compatible, but
42    are provided to make these options accessible to pattern writers who are not
43    able to change the program that processes the pattern. Any number of these
44    items may appear, but they must all be together right at the start of the
45    pattern string, and the letters must be in upper case.
46    .
47    .
48    .SS "UTF support"
49    .rs
50    .sp
51    The original operation of PCRE was on strings of one-byte characters. However,
52    there is now also support for UTF-8 strings in the original library, an
53    extra library that supports 16-bit and UTF-16 character strings, and a
54    third library that supports 32-bit and UTF-32 character strings. To use these
55    features, PCRE must be built to include appropriate support. When using UTF
56    strings you must either call the compiling function with the PCRE_UTF8,
57    PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of
58    these special sequences:
59    .sp
60      (*UTF8)
61      (*UTF16)
62      (*UTF32)
63      (*UTF)
64    .sp
65    (*UTF) is a generic sequence that can be used with any of the libraries.
66    Starting a pattern with such a sequence is equivalent to setting the relevant
67    option. How setting a UTF mode affects pattern matching is mentioned in several
68    places below. There is also a summary of features in the
69  .\" HREF  .\" HREF
70  \fBpcre\fR  \fBpcreunicode\fP
71  .\"  .\"
72  page.  page.
73    .P
74    Some applications that allow their users to supply patterns may wish to
75    restrict them to non-UTF data for security reasons. If the PCRE_NEVER_UTF
76    option is set at compile time, (*UTF) etc. are not allowed, and their
77    appearance causes an error.
78    .
79    .
80    .SS "Unicode property support"
81    .rs
82    .sp
83    Another special sequence that may appear at the start of a pattern is (*UCP).
84    This has the same effect as setting the PCRE_UCP option: it causes sequences
85    such as \ed and \ew to use Unicode properties to determine character types,
86    instead of recognizing only characters with codes less than 128 via a lookup
87    table.
88    .
89    .
90    .SS "Disabling auto-possessification"
91    .rs
92    .sp
93    If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting
94    the PCRE_NO_AUTO_POSSESS option at compile time. This stops PCRE from making
95    quantifiers possessive when what follows cannot match the repeated item. For
96    example, by default a+b is treated as a++b. For more details, see the
97    .\" HREF
98    \fBpcreapi\fP
99    .\"
100    documentation.
101    .
102    .
103    .SS "Disabling start-up optimizations"
104    .rs
105    .sp
106    If a pattern starts with (*NO_START_OPT), it has the same effect as setting the
107    PCRE_NO_START_OPTIMIZE option either at compile or matching time. This disables
108    several optimizations for quickly reaching "no match" results. For more
109    details, see the
110    .\" HREF
111    \fBpcreapi\fP
112    .\"
113    documentation.
114    .
115    .
116    .\" HTML <a name="newlines"></a>
117    .SS "Newline conventions"
118    .rs
119    .sp
120    PCRE supports five different conventions for indicating line breaks in
121    strings: a single CR (carriage return) character, a single LF (linefeed)
122    character, the two-character sequence CRLF, any of the three preceding, or any
123    Unicode newline sequence. The
124    .\" HREF
125    \fBpcreapi\fP
126    .\"
127    page has
128    .\" HTML <a href="pcreapi.html#newlines">
129    .\" </a>
130    further discussion
131    .\"
132    about newlines, and shows how to set the newline convention in the
133    \fIoptions\fP arguments for the compiling and matching functions.
134    .P
135    It is also possible to specify a newline convention by starting a pattern
136    string with one of the following five sequences:
137    .sp
138      (*CR)        carriage return
139      (*LF)        linefeed
140      (*CRLF)      carriage return, followed by linefeed
141      (*ANYCRLF)   any of the three above
142      (*ANY)       all Unicode newline sequences
143    .sp
144    These override the default and the options given to the compiling function. For
145    example, on a Unix system where LF is the default newline sequence, the pattern
146    .sp
147      (*CR)a.b
148    .sp
149    changes the convention to CR. That pattern matches "a\enb" because LF is no
150    longer a newline. If more than one of these settings is present, the last one
151    is used.
152    .P
153    The newline convention affects where the circumflex and dollar assertions are
154    true. It also affects the interpretation of the dot metacharacter when
155    PCRE_DOTALL is not set, and the behaviour of \eN. However, it does not affect
156    what the \eR escape sequence matches. By default, this is any Unicode newline
157    sequence, for Perl compatibility. However, this can be changed; see the
158    description of \eR in the section entitled
159    .\" HTML <a href="#newlineseq">
160    .\" </a>
161    "Newline sequences"
162    .\"
163    below. A change of \eR setting can be combined with a change of newline
164    convention.
165    .
166    .
167    .SS "Setting match and recursion limits"
168    .rs
169    .sp
170    The caller of \fBpcre_exec()\fP can set a limit on the number of times the
171    internal \fBmatch()\fP function is called and on the maximum depth of
172    recursive calls. These facilities are provided to catch runaway matches that
173    are provoked by patterns with huge matching trees (a typical example is a
174    pattern with nested unlimited repeats) and to avoid running out of system stack
175    by too much recursion. When one of these limits is reached, \fBpcre_exec()\fP
176    gives an error return. The limits can also be set by items at the start of the
177    pattern of the form
178    .sp
179      (*LIMIT_MATCH=d)
180      (*LIMIT_RECURSION=d)
181    .sp
182    where d is any number of decimal digits. However, the value of the setting must
183    be less than the value set (or defaulted) by the caller of \fBpcre_exec()\fP
184    for it to have any effect. In other words, the pattern writer can lower the
185    limits set by the programmer, but not raise them. If there is more than one
186    setting of one of these limits, the lower value is used.
187    .
188    .
189    .SH "EBCDIC CHARACTER CODES"
190    .rs
191    .sp
192    PCRE can be compiled to run in an environment that uses EBCDIC as its character
193    code rather than ASCII or Unicode (typically a mainframe system). In the
194    sections below, character code values are ASCII or Unicode; in an EBCDIC
195    environment these characters may have different code values, and there are no
196    code points greater than 255.
197    .
198    .
199    .SH "CHARACTERS AND METACHARACTERS"
200    .rs
201    .sp
202  A regular expression is a pattern that is matched against a subject string from  A regular expression is a pattern that is matched against a subject string from
203  left to right. Most characters stand for themselves in a pattern, and match the  left to right. Most characters stand for themselves in a pattern, and match the
204  corresponding characters in the subject. As a trivial example, the pattern  corresponding characters in the subject. As a trivial example, the pattern
205    .sp
206    The quick brown fox    The quick brown fox
207    .sp
208  matches a portion of a subject string that is identical to itself. The power of  matches a portion of a subject string that is identical to itself. When
209  regular expressions comes from the ability to include alternatives and  caseless matching is specified (the PCRE_CASELESS option), letters are matched
210  repetitions in the pattern. These are encoded in the pattern by the use of  independently of case. In a UTF mode, PCRE always understands the concept of
211  \fImeta-characters\fR, which do not stand for themselves but instead are  case for characters whose values are less than 128, so caseless matching is
212    always possible. For characters with higher values, the concept of case is
213    supported if PCRE is compiled with Unicode property support, but not otherwise.
214    If you want to use caseless matching for characters 128 and above, you must
215    ensure that PCRE is compiled with Unicode property support as well as with
216    UTF support.
217    .P
218    The power of regular expressions comes from the ability to include alternatives
219    and repetitions in the pattern. These are encoded in the pattern by the use of
220    \fImetacharacters\fP, which do not stand for themselves but instead are
221  interpreted in some special way.  interpreted in some special way.
222    .P
223  There are two different sets of meta-characters: those that are recognized  There are two different sets of metacharacters: those that are recognized
224  anywhere in the pattern except within square brackets, and those that are  anywhere in the pattern except within square brackets, and those that are
225  recognized in square brackets. Outside square brackets, the meta-characters are  recognized within square brackets. Outside square brackets, the metacharacters
226  as follows:  are as follows:
227    .sp
228    \\      general escape character with several uses    \e      general escape character with several uses
229    ^      assert start of string (or line, in multiline mode)    ^      assert start of string (or line, in multiline mode)
230    $      assert end of string (or line, in multiline mode)    $      assert end of string (or line, in multiline mode)
231    .      match any character except newline (by default)    .      match any character except newline (by default)
# Line 58  as follows: Line 240  as follows:
240    +      1 or more quantifier    +      1 or more quantifier
241           also "possessive quantifier"           also "possessive quantifier"
242    {      start min/max quantifier    {      start min/max quantifier
243    .sp
244  Part of a pattern that is in square brackets is called a "character class". In  Part of a pattern that is in square brackets is called a "character class". In
245  a character class the only meta-characters are:  a character class the only metacharacters are:
246    .sp
247    \\      general escape character    \e      general escape character
248    ^      negate the class, but only if the first character    ^      negate the class, but only if the first character
249    -      indicates character range    -      indicates character range
250    .\" JOIN
251    [      POSIX character class (only if followed by POSIX    [      POSIX character class (only if followed by POSIX
252             syntax)             syntax)
253    ]      terminates the character class    ]      terminates the character class
254    .sp
255  The following sections describe the use of each of the meta-characters.  The following sections describe the use of each of the metacharacters.
256    .
257    .
258  .SH BACKSLASH  .SH BACKSLASH
259  .rs  .rs
260  .sp  .sp
261  The backslash character has several uses. Firstly, if it is followed by a  The backslash character has several uses. Firstly, if it is followed by a
262  non-alphameric character, it takes away any special meaning that character may  character that is not a number or a letter, it takes away any special meaning
263  have. This use of backslash as an escape character applies both inside and  that character may have. This use of backslash as an escape character applies
264  outside character classes.  both inside and outside character classes.
265    .P
266  For example, if you want to match a * character, you write \\* in the pattern.  For example, if you want to match a * character, you write \e* in the pattern.
267  This escaping action applies whether or not the following character would  This escaping action applies whether or not the following character would
268  otherwise be interpreted as a meta-character, so it is always safe to precede a  otherwise be interpreted as a metacharacter, so it is always safe to precede a
269  non-alphameric with backslash to specify that it stands for itself. In  non-alphanumeric with backslash to specify that it stands for itself. In
270  particular, if you want to match a backslash, you write \\\\.  particular, if you want to match a backslash, you write \e\e.
271    .P
272  If a pattern is compiled with the PCRE_EXTENDED option, whitespace in the  In a UTF mode, only ASCII numbers and letters have any special meaning after a
273  pattern (other than in a character class) and characters between a # outside  backslash. All other characters (in particular, those whose codepoints are
274  a character class and the next newline character are ignored. An escaping  greater than 127) are treated as literals.
275  backslash can be used to include a whitespace or # character as part of the  .P
276    If a pattern is compiled with the PCRE_EXTENDED option, most white space in the
277    pattern (other than in a character class), and characters between a # outside a
278    character class and the next newline, inclusive, are ignored. An escaping
279    backslash can be used to include a white space or # character as part of the
280  pattern.  pattern.
281    .P
282  If you want to remove the special meaning from a sequence of characters, you  If you want to remove the special meaning from a sequence of characters, you
283  can do so by putting them between \\Q and \\E. This is different from Perl in  can do so by putting them between \eQ and \eE. This is different from Perl in
284  that $ and @ are handled as literals in \\Q...\\E sequences in PCRE, whereas in  that $ and @ are handled as literals in \eQ...\eE sequences in PCRE, whereas in
285  Perl, $ and @ cause variable interpolation. Note the following examples:  Perl, $ and @ cause variable interpolation. Note the following examples:
286    .sp
287    Pattern            PCRE matches   Perl matches    Pattern            PCRE matches   Perl matches
288    .sp
289    \\Qabc$xyz\\E        abc$xyz        abc followed by the  .\" JOIN
290      \eQabc$xyz\eE        abc$xyz        abc followed by the
291                                        contents of $xyz                                        contents of $xyz
292    \\Qabc\\$xyz\\E       abc\\$xyz       abc\\$xyz    \eQabc\e$xyz\eE       abc\e$xyz       abc\e$xyz
293    \\Qabc\\E\\$\\Qxyz\\E   abc$xyz        abc$xyz    \eQabc\eE\e$\eQxyz\eE   abc$xyz        abc$xyz
294    .sp
295  The \\Q...\\E sequence is recognized both inside and outside character classes.  The \eQ...\eE sequence is recognized both inside and outside character classes.
296    An isolated \eE that is not preceded by \eQ is ignored. If \eQ is not followed
297    by \eE later in the pattern, the literal interpretation continues to the end of
298    the pattern (that is, \eE is assumed at the end). If the isolated \eQ is inside
299    a character class, this causes an error, because the character class is not
300    terminated.
301    .
302    .
303    .\" HTML <a name="digitsafterbackslash"></a>
304    .SS "Non-printing characters"
305    .rs
306    .sp
307  A second use of backslash provides a way of encoding non-printing characters  A second use of backslash provides a way of encoding non-printing characters
308  in patterns in a visible manner. There is no restriction on the appearance of  in patterns in a visible manner. There is no restriction on the appearance of
309  non-printing characters, apart from the binary zero that terminates a pattern,  non-printing characters, apart from the binary zero that terminates a pattern,
310  but when a pattern is being prepared by text editing, it is usually easier to  but when a pattern is being prepared by text editing, it is often easier to use
311  use one of the following escape sequences than the binary character it  one of the following escape sequences than the binary character it represents:
312  represents:  .sp
313      \ea        alarm, that is, the BEL character (hex 07)
314    \\a        alarm, that is, the BEL character (hex 07)    \ecx       "control-x", where x is any ASCII character
315    \\cx       "control-x", where x is any character    \ee        escape (hex 1B)
316    \\e        escape (hex 1B)    \ef        form feed (hex 0C)
317    \\f        formfeed (hex 0C)    \en        linefeed (hex 0A)
318    \\n        newline (hex 0A)    \er        carriage return (hex 0D)
319    \\r        carriage return (hex 0D)    \et        tab (hex 09)
320    \\t        tab (hex 09)    \e0dd      character with octal code 0dd
321    \\ddd      character with octal code ddd, or backreference    \eddd      character with octal code ddd, or back reference
322    \\xhh      character with hex code hh    \eo{ddd..} character with octal code ddd..
323    \\x{hhh..} character with hex code hhh... (UTF-8 mode only)    \exhh      character with hex code hh
324      \ex{hhh..} character with hex code hhh.. (non-JavaScript mode)
325  The precise effect of \\cx is as follows: if x is a lower case letter, it    \euhhhh    character with hex code hhhh (JavaScript mode only)
326  is converted to upper case. Then bit 6 of the character (hex 40) is inverted.  .sp
327  Thus \\cz becomes hex 1A, but \\c{ becomes hex 3B, while \\c; becomes hex  The precise effect of \ecx on ASCII characters is as follows: if x is a lower
328  7B.  case letter, it is converted to upper case. Then bit 6 of the character (hex
329    40) is inverted. Thus \ecA to \ecZ become hex 01 to hex 1A (A is 41, Z is 5A),
330  After \\x, from zero to two hexadecimal digits are read (letters can be in  but \ec{ becomes hex 3B ({ is 7B), and \ec; becomes hex 7B (; is 3B). If the
331  upper or lower case). In UTF-8 mode, any number of hexadecimal digits may  data item (byte or 16-bit value) following \ec has a value greater than 127, a
332  appear between \\x{ and }, but the value of the character code must be less  compile-time error occurs. This locks out non-ASCII characters in all modes.
333  than 2**31 (that is, the maximum hexadecimal value is 7FFFFFFF). If characters  .P
334  other than hexadecimal digits appear between \\x{ and }, or if there is no  The \ec facility was designed for use with ASCII characters, but with the
335  terminating }, this form of escape is not recognized. Instead, the initial  extension to Unicode it is even less useful than it once was. It is, however,
336  \\x will be interpreted as a basic hexadecimal escape, with no following  recognized when PCRE is compiled in EBCDIC mode, where data items are always
337  digits, giving a byte whose value is zero.  bytes. In this mode, all values are valid after \ec. If the next character is a
338    lower case letter, it is converted to upper case. Then the 0xc0 bits of the
339  Characters whose value is less than 256 can be defined by either of the two  byte are inverted. Thus \ecA becomes hex 01, as in ASCII (A is C1), but because
340  syntaxes for \\x when PCRE is in UTF-8 mode. There is no difference in the  the EBCDIC letters are disjoint, \ecZ becomes hex 29 (Z is E9), and other
341  way they are handled. For example, \\xdc is exactly the same as \\x{dc}.  characters also generate different values.
342    .P
343  After \\0 up to two further octal digits are read. In both cases, if there  After \e0 up to two further octal digits are read. If there are fewer than two
344  are fewer than two digits, just those that are present are used. Thus the  digits, just those that are present are used. Thus the sequence \e0\ex\e07
345  sequence \\0\\x\\07 specifies two binary zeros followed by a BEL character  specifies two binary zeros followed by a BEL character (code value 7). Make
346  (code value 7). Make sure you supply two digits after the initial zero if the  sure you supply two digits after the initial zero if the pattern character that
347  character that follows is itself an octal digit.  follows is itself an octal digit.
348    .P
349  The handling of a backslash followed by a digit other than 0 is complicated.  The escape \eo must be followed by a sequence of octal digits, enclosed in
350  Outside a character class, PCRE reads it and any following digits as a decimal  braces. An error occurs if this is not the case. This escape is a recent
351  number. If the number is less than 10, or if there have been at least that many  addition to Perl; it provides way of specifying character code points as octal
352    numbers greater than 0777, and it also allows octal numbers and back references
353    to be unambiguously specified.
354    .P
355    For greater clarity and unambiguity, it is best to avoid following \e by a
356    digit greater than zero. Instead, use \eo{} or \ex{} to specify character
357    numbers, and \eg{} to specify back references. The following paragraphs
358    describe the old, ambiguous syntax.
359    .P
360    The handling of a backslash followed by a digit other than 0 is complicated,
361    and Perl has changed in recent releases, causing PCRE also to change. Outside a
362    character class, PCRE reads the digit and any following digits as a decimal
363    number. If the number is less than 8, or if there have been at least that many
364  previous capturing left parentheses in the expression, the entire sequence is  previous capturing left parentheses in the expression, the entire sequence is
365  taken as a \fIback reference\fR. A description of how this works is given  taken as a \fIback reference\fP. A description of how this works is given
366  later, following the discussion of parenthesized subpatterns.  .\" HTML <a href="#backreferences">
367    .\" </a>
368  Inside a character class, or if the decimal number is greater than 9 and there  later,
369  have not been that many capturing subpatterns, PCRE re-reads up to three octal  .\"
370  digits following the backslash, and generates a single byte from the least  following the discussion of
371  significant 8 bits of the value. Any subsequent digits stand for themselves.  .\" HTML <a href="#subpattern">
372  For example:  .\" </a>
373    parenthesized subpatterns.
374    \\040   is another way of writing a space  .\"
375    \\40    is the same, provided there are fewer than 40  .P
376    Inside a character class, or if the decimal number following \e is greater than
377    7 and there have not been that many capturing subpatterns, PCRE handles \e8 and
378    \e9 as the literal characters "8" and "9", and otherwise re-reads up to three
379    octal digits following the backslash, using them to generate a data character.
380    Any subsequent digits stand for themselves. For example:
381    .sp
382      \e040   is another way of writing an ASCII space
383    .\" JOIN
384      \e40    is the same, provided there are fewer than 40
385              previous capturing subpatterns              previous capturing subpatterns
386    \\7     is always a back reference    \e7     is always a back reference
387    \\11    might be a back reference, or another way of  .\" JOIN
388      \e11    might be a back reference, or another way of
389              writing a tab              writing a tab
390    \\011   is always a tab    \e011   is always a tab
391    \\0113  is a tab followed by the character "3"    \e0113  is a tab followed by the character "3"
392    \\113   might be a back reference, otherwise the  .\" JOIN
393      \e113   might be a back reference, otherwise the
394              character with octal code 113              character with octal code 113
395    \\377   might be a back reference, otherwise  .\" JOIN
396              the byte consisting entirely of 1 bits    \e377   might be a back reference, otherwise
397    \\81    is either a back reference, or a binary zero              the value 255 (decimal)
398              followed by the two characters "8" and "1"  .\" JOIN
399      \e81    is either a back reference, or the two
400  Note that octal values of 100 or greater must not be introduced by a leading              characters "8" and "1"
401  zero, because no more than three octal digits are ever read.  .sp
402    Note that octal values of 100 or greater that are specified using this syntax
403  All the sequences that define a single byte value or a single UTF-8 character  must not be introduced by a leading zero, because no more than three octal
404  (in UTF-8 mode) can be used both inside and outside character classes. In  digits are ever read.
405  addition, inside a character class, the sequence \\b is interpreted as the  .P
406  backspace character (hex 08). Outside a character class it has a different  By default, after \ex that is not followed by {, from zero to two hexadecimal
407  meaning (see below).  digits are read (letters can be in upper or lower case). Any number of
408    hexadecimal digits may appear between \ex{ and }. If a character other than
409  The third use of backslash is for specifying generic character types:  a hexadecimal digit appears between \ex{ and }, or if there is no terminating
410    }, an error occurs.
411    \\d     any decimal digit  .P
412    \\D     any character that is not a decimal digit  If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \ex is
413    \\s     any whitespace character  as just described only when it is followed by two hexadecimal digits.
414    \\S     any character that is not a whitespace character  Otherwise, it matches a literal "x" character. In JavaScript mode, support for
415    \\w     any "word" character  code points greater than 256 is provided by \eu, which must be followed by
416    \\W     any "non-word" character  four hexadecimal digits; otherwise it matches a literal "u" character.
417    .P
418  Each pair of escape sequences partitions the complete set of characters into  Characters whose value is less than 256 can be defined by either of the two
419  two disjoint sets. Any given character matches one, and only one, of each pair.  syntaxes for \ex (or by \eu in JavaScript mode). There is no difference in the
420    way they are handled. For example, \exdc is exactly the same as \ex{dc} (or
421  In UTF-8 mode, characters with values greater than 255 never match \\d, \\s, or  \eu00dc in JavaScript mode).
422  \\w, and always match \\D, \\S, and \\W.  .
423    .
424  For compatibility with Perl, \\s does not match the VT character (code 11).  .SS "Constraints on character values"
425  This makes it different from the the POSIX "space" class. The \\s characters  .rs
426  are HT (9), LF (10), FF (12), CR (13), and space (32).  .sp
427    Characters that are specified using octal or hexadecimal numbers are
428  A "word" character is any letter or digit or the underscore character, that is,  limited to certain values, as follows:
429  any character which can be part of a Perl "word". The definition of letters and  .sp
430  digits is controlled by PCRE's character tables, and may vary if locale-    8-bit non-UTF mode    less than 0x100
431  specific matching is taking place (see    8-bit UTF-8 mode      less than 0x10ffff and a valid codepoint
432      16-bit non-UTF mode   less than 0x10000
433      16-bit UTF-16 mode    less than 0x10ffff and a valid codepoint
434      32-bit non-UTF mode   less than 0x100000000
435      32-bit UTF-32 mode    less than 0x10ffff and a valid codepoint
436    .sp
437    Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called
438    "surrogate" codepoints), and 0xffef.
439    .
440    .
441    .SS "Escape sequences in character classes"
442    .rs
443    .sp
444    All the sequences that define a single character value can be used both inside
445    and outside character classes. In addition, inside a character class, \eb is
446    interpreted as the backspace character (hex 08).
447    .P
448    \eN is not allowed in a character class. \eB, \eR, and \eX are not special
449    inside a character class. Like other unrecognized escape sequences, they are
450    treated as the literal characters "B", "R", and "X" by default, but cause an
451    error if the PCRE_EXTRA option is set. Outside a character class, these
452    sequences have different meanings.
453    .
454    .
455    .SS "Unsupported escape sequences"
456    .rs
457    .sp
458    In Perl, the sequences \el, \eL, \eu, and \eU are recognized by its string
459    handler and used to modify the case of following characters. By default, PCRE
460    does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT
461    option is set, \eU matches a "U" character, and \eu can be used to define a
462    character by code point, as described in the previous section.
463    .
464    .
465    .SS "Absolute and relative back references"
466    .rs
467    .sp
468    The sequence \eg followed by an unsigned or a negative number, optionally
469    enclosed in braces, is an absolute or relative back reference. A named back
470    reference can be coded as \eg{name}. Back references are discussed
471    .\" HTML <a href="#backreferences">
472    .\" </a>
473    later,
474    .\"
475    following the discussion of
476    .\" HTML <a href="#subpattern">
477    .\" </a>
478    parenthesized subpatterns.
479    .\"
480    .
481    .
482    .SS "Absolute and relative subroutine calls"
483    .rs
484    .sp
485    For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
486    a number enclosed either in angle brackets or single quotes, is an alternative
487    syntax for referencing a subpattern as a "subroutine". Details are discussed
488    .\" HTML <a href="#onigurumasubroutines">
489    .\" </a>
490    later.
491    .\"
492    Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
493    synonymous. The former is a back reference; the latter is a
494    .\" HTML <a href="#subpatternsassubroutines">
495    .\" </a>
496    subroutine
497    .\"
498    call.
499    .
500    .
501    .\" HTML <a name="genericchartypes"></a>
502    .SS "Generic character types"
503    .rs
504    .sp
505    Another use of backslash is for specifying generic character types:
506    .sp
507      \ed     any decimal digit
508      \eD     any character that is not a decimal digit
509      \eh     any horizontal white space character
510      \eH     any character that is not a horizontal white space character
511      \es     any white space character
512      \eS     any character that is not a white space character
513      \ev     any vertical white space character
514      \eV     any character that is not a vertical white space character
515      \ew     any "word" character
516      \eW     any "non-word" character
517    .sp
518    There is also the single sequence \eN, which matches a non-newline character.
519    This is the same as
520    .\" HTML <a href="#fullstopdot">
521    .\" </a>
522    the "." metacharacter
523    .\"
524    when PCRE_DOTALL is not set. Perl also uses \eN to match characters by name;
525    PCRE does not support this.
526    .P
527    Each pair of lower and upper case escape sequences partitions the complete set
528    of characters into two disjoint sets. Any given character matches one, and only
529    one, of each pair. The sequences can appear both inside and outside character
530    classes. They each match one character of the appropriate type. If the current
531    matching point is at the end of the subject string, all of them fail, because
532    there is no character to match.
533    .P
534    For compatibility with Perl, \es did not used to match the VT character (code
535    11), which made it different from the the POSIX "space" class. However, Perl
536    added VT at release 5.18, and PCRE followed suit at release 8.34. The default
537    \es characters are now HT (9), LF (10), VT (11), FF (12), CR (13), and space
538    (32), which are defined as white space in the "C" locale. This list may vary if
539    locale-specific matching is taking place; in particular, in some locales the
540    "non-breaking space" character (\exA0) is recognized as white space.
541    .P
542    A "word" character is an underscore or any character that is a letter or digit.
543    By default, the definition of letters and digits is controlled by PCRE's
544    low-valued character tables, and may vary if locale-specific matching is taking
545    place (see
546  .\" HTML <a href="pcreapi.html#localesupport">  .\" HTML <a href="pcreapi.html#localesupport">
547  .\" </a>  .\" </a>
548  "Locale support"  "Locale support"
549  .\"  .\"
550  in the  in the
551  .\" HREF  .\" HREF
552  \fBpcreapi\fR  \fBpcreapi\fP
553  .\"  .\"
554  page). For example, in the "fr" (French) locale, some character codes greater  page). For example, in a French locale such as "fr_FR" in Unix-like systems,
555  than 128 are used for accented letters, and these are matched by \\w.  or "french" in Windows, some character codes greater than 127 are used for
556    accented letters, and these are then matched by \ew. The use of locales with
557  These character type sequences can appear both inside and outside character  Unicode is discouraged.
558  classes. They each match one character of the appropriate type. If the current  .P
559  matching point is at the end of the subject string, all of them fail, since  By default, characters whose code points are greater than 127 never match \ed,
560  there is no character to match.  \es, or \ew, and always match \eD, \eS, and \eW, although this may vary for
561    characters in the range 128-255 when locale-specific matching is happening.
562  The fourth use of backslash is for certain simple assertions. An assertion  These escape sequences retain their original meanings from before Unicode
563    support was available, mainly for efficiency reasons. If PCRE is compiled with
564    Unicode property support, and the PCRE_UCP option is set, the behaviour is
565    changed so that Unicode properties are used to determine character types, as
566    follows:
567    .sp
568      \ed  any character that matches \ep{Nd} (decimal digit)
569      \es  any character that matches \ep{Z} or \eh or \ev
570      \ew  any character that matches \ep{L} or \ep{N}, plus underscore
571    .sp
572    The upper case escapes match the inverse sets of characters. Note that \ed
573    matches only decimal digits, whereas \ew matches any Unicode digit, as well as
574    any Unicode letter, and underscore. Note also that PCRE_UCP affects \eb, and
575    \eB because they are defined in terms of \ew and \eW. Matching these sequences
576    is noticeably slower when PCRE_UCP is set.
577    .P
578    The sequences \eh, \eH, \ev, and \eV are features that were added to Perl at
579    release 5.10. In contrast to the other sequences, which match only ASCII
580    characters by default, these always match certain high-valued code points,
581    whether or not PCRE_UCP is set. The horizontal space characters are:
582    .sp
583      U+0009     Horizontal tab (HT)
584      U+0020     Space
585      U+00A0     Non-break space
586      U+1680     Ogham space mark
587      U+180E     Mongolian vowel separator
588      U+2000     En quad
589      U+2001     Em quad
590      U+2002     En space
591      U+2003     Em space
592      U+2004     Three-per-em space
593      U+2005     Four-per-em space
594      U+2006     Six-per-em space
595      U+2007     Figure space
596      U+2008     Punctuation space
597      U+2009     Thin space
598      U+200A     Hair space
599      U+202F     Narrow no-break space
600      U+205F     Medium mathematical space
601      U+3000     Ideographic space
602    .sp
603    The vertical space characters are:
604    .sp
605      U+000A     Linefeed (LF)
606      U+000B     Vertical tab (VT)
607      U+000C     Form feed (FF)
608      U+000D     Carriage return (CR)
609      U+0085     Next line (NEL)
610      U+2028     Line separator
611      U+2029     Paragraph separator
612    .sp
613    In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are
614    relevant.
615    .
616    .
617    .\" HTML <a name="newlineseq"></a>
618    .SS "Newline sequences"
619    .rs
620    .sp
621    Outside a character class, by default, the escape sequence \eR matches any
622    Unicode newline sequence. In 8-bit non-UTF-8 mode \eR is equivalent to the
623    following:
624    .sp
625      (?>\er\en|\en|\ex0b|\ef|\er|\ex85)
626    .sp
627    This is an example of an "atomic group", details of which are given
628    .\" HTML <a href="#atomicgroup">
629    .\" </a>
630    below.
631    .\"
632    This particular group matches either the two-character sequence CR followed by
633    LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab,
634    U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next
635    line, U+0085). The two-character sequence is treated as a single unit that
636    cannot be split.
637    .P
638    In other modes, two additional characters whose codepoints are greater than 255
639    are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
640    Unicode character property support is not needed for these characters to be
641    recognized.
642    .P
643    It is possible to restrict \eR to match only CR, LF, or CRLF (instead of the
644    complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF
645    either at compile time or when the pattern is matched. (BSR is an abbrevation
646    for "backslash R".) This can be made the default when PCRE is built; if this is
647    the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option.
648    It is also possible to specify these settings by starting a pattern string with
649    one of the following sequences:
650    .sp
651      (*BSR_ANYCRLF)   CR, LF, or CRLF only
652      (*BSR_UNICODE)   any Unicode newline sequence
653    .sp
654    These override the default and the options given to the compiling function, but
655    they can themselves be overridden by options given to a matching function. Note
656    that these special settings, which are not Perl-compatible, are recognized only
657    at the very start of a pattern, and that they must be in upper case. If more
658    than one of them is present, the last one is used. They can be combined with a
659    change of newline convention; for example, a pattern can start with:
660    .sp
661      (*ANY)(*BSR_ANYCRLF)
662    .sp
663    They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or
664    (*UCP) special sequences. Inside a character class, \eR is treated as an
665    unrecognized escape sequence, and so matches the letter "R" by default, but
666    causes an error if PCRE_EXTRA is set.
667    .
668    .
669    .\" HTML <a name="uniextseq"></a>
670    .SS Unicode character properties
671    .rs
672    .sp
673    When PCRE is built with Unicode character property support, three additional
674    escape sequences that match characters with specific properties are available.
675    When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing
676    characters whose codepoints are less than 256, but they do work in this mode.
677    The extra escape sequences are:
678    .sp
679      \ep{\fIxx\fP}   a character with the \fIxx\fP property
680      \eP{\fIxx\fP}   a character without the \fIxx\fP property
681      \eX       a Unicode extended grapheme cluster
682    .sp
683    The property names represented by \fIxx\fP above are limited to the Unicode
684    script names, the general category properties, "Any", which matches any
685    character (including newline), and some special PCRE properties (described
686    in the
687    .\" HTML <a href="#extraprops">
688    .\" </a>
689    next section).
690    .\"
691    Other Perl properties such as "InMusicalSymbols" are not currently supported by
692    PCRE. Note that \eP{Any} does not match any characters, so always causes a
693    match failure.
694    .P
695    Sets of Unicode characters are defined as belonging to certain scripts. A
696    character from one of these sets can be matched using a script name. For
697    example:
698    .sp
699      \ep{Greek}
700      \eP{Han}
701    .sp
702    Those that are not part of an identified script are lumped together as
703    "Common". The current list of scripts is:
704    .P
705    Arabic,
706    Armenian,
707    Avestan,
708    Balinese,
709    Bamum,
710    Batak,
711    Bengali,
712    Bopomofo,
713    Brahmi,
714    Braille,
715    Buginese,
716    Buhid,
717    Canadian_Aboriginal,
718    Carian,
719    Chakma,
720    Cham,
721    Cherokee,
722    Common,
723    Coptic,
724    Cuneiform,
725    Cypriot,
726    Cyrillic,
727    Deseret,
728    Devanagari,
729    Egyptian_Hieroglyphs,
730    Ethiopic,
731    Georgian,
732    Glagolitic,
733    Gothic,
734    Greek,
735    Gujarati,
736    Gurmukhi,
737    Han,
738    Hangul,
739    Hanunoo,
740    Hebrew,
741    Hiragana,
742    Imperial_Aramaic,
743    Inherited,
744    Inscriptional_Pahlavi,
745    Inscriptional_Parthian,
746    Javanese,
747    Kaithi,
748    Kannada,
749    Katakana,
750    Kayah_Li,
751    Kharoshthi,
752    Khmer,
753    Lao,
754    Latin,
755    Lepcha,
756    Limbu,
757    Linear_B,
758    Lisu,
759    Lycian,
760    Lydian,
761    Malayalam,
762    Mandaic,
763    Meetei_Mayek,
764    Meroitic_Cursive,
765    Meroitic_Hieroglyphs,
766    Miao,
767    Mongolian,
768    Myanmar,
769    New_Tai_Lue,
770    Nko,
771    Ogham,
772    Old_Italic,
773    Old_Persian,
774    Old_South_Arabian,
775    Old_Turkic,
776    Ol_Chiki,
777    Oriya,
778    Osmanya,
779    Phags_Pa,
780    Phoenician,
781    Rejang,
782    Runic,
783    Samaritan,
784    Saurashtra,
785    Sharada,
786    Shavian,
787    Sinhala,
788    Sora_Sompeng,
789    Sundanese,
790    Syloti_Nagri,
791    Syriac,
792    Tagalog,
793    Tagbanwa,
794    Tai_Le,
795    Tai_Tham,
796    Tai_Viet,
797    Takri,
798    Tamil,
799    Telugu,
800    Thaana,
801    Thai,
802    Tibetan,
803    Tifinagh,
804    Ugaritic,
805    Vai,
806    Yi.
807    .P
808    Each character has exactly one Unicode general category property, specified by
809    a two-letter abbreviation. For compatibility with Perl, negation can be
810    specified by including a circumflex between the opening brace and the property
811    name. For example, \ep{^Lu} is the same as \eP{Lu}.
812    .P
813    If only one letter is specified with \ep or \eP, it includes all the general
814    category properties that start with that letter. In this case, in the absence
815    of negation, the curly brackets in the escape sequence are optional; these two
816    examples have the same effect:
817    .sp
818      \ep{L}
819      \epL
820    .sp
821    The following general category property codes are supported:
822    .sp
823      C     Other
824      Cc    Control
825      Cf    Format
826      Cn    Unassigned
827      Co    Private use
828      Cs    Surrogate
829    .sp
830      L     Letter
831      Ll    Lower case letter
832      Lm    Modifier letter
833      Lo    Other letter
834      Lt    Title case letter
835      Lu    Upper case letter
836    .sp
837      M     Mark
838      Mc    Spacing mark
839      Me    Enclosing mark
840      Mn    Non-spacing mark
841    .sp
842      N     Number
843      Nd    Decimal number
844      Nl    Letter number
845      No    Other number
846    .sp
847      P     Punctuation
848      Pc    Connector punctuation
849      Pd    Dash punctuation
850      Pe    Close punctuation
851      Pf    Final punctuation
852      Pi    Initial punctuation
853      Po    Other punctuation
854      Ps    Open punctuation
855    .sp
856      S     Symbol
857      Sc    Currency symbol
858      Sk    Modifier symbol
859      Sm    Mathematical symbol
860      So    Other symbol
861    .sp
862      Z     Separator
863      Zl    Line separator
864      Zp    Paragraph separator
865      Zs    Space separator
866    .sp
867    The special property L& is also supported: it matches a character that has
868    the Lu, Ll, or Lt property, in other words, a letter that is not classified as
869    a modifier or "other".
870    .P
871    The Cs (Surrogate) property applies only to characters in the range U+D800 to
872    U+DFFF. Such characters are not valid in Unicode strings and so
873    cannot be tested by PCRE, unless UTF validity checking has been turned off
874    (see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and
875    PCRE_NO_UTF32_CHECK in the
876    .\" HREF
877    \fBpcreapi\fP
878    .\"
879    page). Perl does not support the Cs property.
880    .P
881    The long synonyms for property names that Perl supports (such as \ep{Letter})
882    are not supported by PCRE, nor is it permitted to prefix any of these
883    properties with "Is".
884    .P
885    No character that is in the Unicode table has the Cn (unassigned) property.
886    Instead, this property is assumed for any code point that is not in the
887    Unicode table.
888    .P
889    Specifying caseless matching does not affect these escape sequences. For
890    example, \ep{Lu} always matches only upper case letters. This is different from
891    the behaviour of current versions of Perl.
892    .P
893    Matching characters by Unicode property is not fast, because PCRE has to do a
894    multistage table lookup in order to find a character's property. That is why
895    the traditional escape sequences such as \ed and \ew do not use Unicode
896    properties in PCRE by default, though you can make them do so by setting the
897    PCRE_UCP option or by starting the pattern with (*UCP).
898    .
899    .
900    .SS Extended grapheme clusters
901    .rs
902    .sp
903    The \eX escape matches any number of Unicode characters that form an "extended
904    grapheme cluster", and treats the sequence as an atomic group
905    .\" HTML <a href="#atomicgroup">
906    .\" </a>
907    (see below).
908    .\"
909    Up to and including release 8.31, PCRE matched an earlier, simpler definition
910    that was equivalent to
911    .sp
912      (?>\ePM\epM*)
913    .sp
914    That is, it matched a character without the "mark" property, followed by zero
915    or more characters with the "mark" property. Characters with the "mark"
916    property are typically non-spacing accents that affect the preceding character.
917    .P
918    This simple definition was extended in Unicode to include more complicated
919    kinds of composite character by giving each character a grapheme breaking
920    property, and creating rules that use these properties to define the boundaries
921    of extended grapheme clusters. In releases of PCRE later than 8.31, \eX matches
922    one of these clusters.
923    .P
924    \eX always matches at least one character. Then it decides whether to add
925    additional characters according to the following rules for ending a cluster:
926    .P
927    1. End at the end of the subject string.
928    .P
929    2. Do not end between CR and LF; otherwise end after any control character.
930    .P
931    3. Do not break Hangul (a Korean script) syllable sequences. Hangul characters
932    are of five types: L, V, T, LV, and LVT. An L character may be followed by an
933    L, V, LV, or LVT character; an LV or V character may be followed by a V or T
934    character; an LVT or T character may be follwed only by a T character.
935    .P
936    4. Do not end before extending characters or spacing marks. Characters with
937    the "mark" property always have the "extend" grapheme breaking property.
938    .P
939    5. Do not end after prepend characters.
940    .P
941    6. Otherwise, end the cluster.
942    .
943    .
944    .\" HTML <a name="extraprops"></a>
945    .SS PCRE's additional properties
946    .rs
947    .sp
948    As well as the standard Unicode properties described above, PCRE supports four
949    more that make it possible to convert traditional escape sequences such as \ew
950    and \es to use Unicode properties. PCRE uses these non-standard, non-Perl
951    properties internally when PCRE_UCP is set. However, they may also be used
952    explicitly. These properties are:
953    .sp
954      Xan   Any alphanumeric character
955      Xps   Any POSIX space character
956      Xsp   Any Perl space character
957      Xwd   Any Perl "word" character
958    .sp
959    Xan matches characters that have either the L (letter) or the N (number)
960    property. Xps matches the characters tab, linefeed, vertical tab, form feed, or
961    carriage return, and any other character that has the Z (separator) property.
962    Xsp is the same as Xps; it used to exclude vertical tab, for Perl
963    compatibility, but Perl changed, and so PCRE followed at release 8.34. Xwd
964    matches the same characters as Xan, plus underscore.
965    .P
966    There is another non-standard property, Xuc, which matches any character that
967    can be represented by a Universal Character Name in C++ and other programming
968    languages. These are the characters $, @, ` (grave accent), and all characters
969    with Unicode code points greater than or equal to U+00A0, except for the
970    surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are
971    excluded. (Universal Character Names are of the form \euHHHH or \eUHHHHHHHH
972    where H is a hexadecimal digit. Note that the Xuc property does not match these
973    sequences but the characters that they represent.)
974    .
975    .
976    .\" HTML <a name="resetmatchstart"></a>
977    .SS "Resetting the match start"
978    .rs
979    .sp
980    The escape sequence \eK causes any previously matched characters not to be
981    included in the final matched sequence. For example, the pattern:
982    .sp
983      foo\eKbar
984    .sp
985    matches "foobar", but reports that it has matched "bar". This feature is
986    similar to a lookbehind assertion
987    .\" HTML <a href="#lookbehind">
988    .\" </a>
989    (described below).
990    .\"
991    However, in this case, the part of the subject before the real match does not
992    have to be of fixed length, as lookbehind assertions do. The use of \eK does
993    not interfere with the setting of
994    .\" HTML <a href="#subpattern">
995    .\" </a>
996    captured substrings.
997    .\"
998    For example, when the pattern
999    .sp
1000      (foo)\eKbar
1001    .sp
1002    matches "foobar", the first substring is still set to "foo".
1003    .P
1004    Perl documents that the use of \eK within assertions is "not well defined". In
1005    PCRE, \eK is acted upon when it occurs inside positive assertions, but is
1006    ignored in negative assertions.
1007    .
1008    .
1009    .\" HTML <a name="smallassertions"></a>
1010    .SS "Simple assertions"
1011    .rs
1012    .sp
1013    The final use of backslash is for certain simple assertions. An assertion
1014  specifies a condition that has to be met at a particular point in a match,  specifies a condition that has to be met at a particular point in a match,
1015  without consuming any characters from the subject string. The use of  without consuming any characters from the subject string. The use of
1016  subpatterns for more complicated assertions is described below. The backslashed  subpatterns for more complicated assertions is described
1017  assertions are  .\" HTML <a href="#bigassertions">
1018    .\" </a>
1019    \\b     matches at a word boundary  below.
1020    \\B     matches when not at a word boundary  .\"
1021    \\A     matches at start of subject  The backslashed assertions are:
1022    \\Z     matches at end of subject or before newline at end  .sp
1023    \\z     matches at end of subject    \eb     matches at a word boundary
1024    \\G     matches at first matching position in subject    \eB     matches when not at a word boundary
1025      \eA     matches at the start of the subject
1026  These assertions may not appear in character classes (but note that \\b has a    \eZ     matches at the end of the subject
1027  different meaning, namely the backspace character, inside a character class).            also matches before a newline at the end of the subject
1028      \ez     matches only at the end of the subject
1029      \eG     matches at the first matching position in the subject
1030    .sp
1031    Inside a character class, \eb has a different meaning; it matches the backspace
1032    character. If any other of these assertions appears in a character class, by
1033    default it matches the corresponding literal character (for example, \eB
1034    matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid
1035    escape sequence" error is generated instead.
1036    .P
1037  A word boundary is a position in the subject string where the current character  A word boundary is a position in the subject string where the current character
1038  and the previous character do not both match \\w or \\W (i.e. one matches  and the previous character do not both match \ew or \eW (i.e. one matches
1039  \\w and the other matches \\W), or the start or end of the string if the  \ew and the other matches \eW), or the start or end of the string if the
1040  first or last character matches \\w, respectively.  first or last character matches \ew, respectively. In a UTF mode, the meanings
1041    of \ew and \eW can be changed by setting the PCRE_UCP option. When this is
1042  The \\A, \\Z, and \\z assertions differ from the traditional circumflex and  done, it also affects \eb and \eB. Neither PCRE nor Perl has a separate "start
1043  dollar (described below) in that they only ever match at the very start and end  of word" or "end of word" metasequence. However, whatever follows \eb normally
1044  of the subject string, whatever options are set. Thus, they are independent of  determines which it is. For example, the fragment \eba matches "a" at the start
1045  multiline mode.  of a word.
1046    .P
1047  They are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options. If the  The \eA, \eZ, and \ez assertions differ from the traditional circumflex and
1048  \fIstartoffset\fR argument of \fBpcre_exec()\fR is non-zero, indicating that  dollar (described in the next section) in that they only ever match at the very
1049  matching is to start at a point other than the beginning of the subject, \\A  start and end of the subject string, whatever options are set. Thus, they are
1050  can never match. The difference between \\Z and \\z is that \\Z matches before  independent of multiline mode. These three assertions are not affected by the
1051  a newline that is the last character of the string as well as at the end of the  PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the
1052  string, whereas \\z matches only at the end.  circumflex and dollar metacharacters. However, if the \fIstartoffset\fP
1053    argument of \fBpcre_exec()\fP is non-zero, indicating that matching is to start
1054  The \\G assertion is true only when the current matching position is at the  at a point other than the beginning of the subject, \eA can never match. The
1055  start point of the match, as specified by the \fIstartoffset\fR argument of  difference between \eZ and \ez is that \eZ matches before a newline at the end
1056  \fBpcre_exec()\fR. It differs from \\A when the value of \fIstartoffset\fR is  of the string as well as at the very end, whereas \ez matches only at the end.
1057  non-zero. By calling \fBpcre_exec()\fR multiple times with appropriate  .P
1058    The \eG assertion is true only when the current matching position is at the
1059    start point of the match, as specified by the \fIstartoffset\fP argument of
1060    \fBpcre_exec()\fP. It differs from \eA when the value of \fIstartoffset\fP is
1061    non-zero. By calling \fBpcre_exec()\fP multiple times with appropriate
1062  arguments, you can mimic Perl's /g option, and it is in this kind of  arguments, you can mimic Perl's /g option, and it is in this kind of
1063  implementation where \\G can be useful.  implementation where \eG can be useful.
1064    .P
1065  Note, however, that PCRE's interpretation of \\G, as the start of the current  Note, however, that PCRE's interpretation of \eG, as the start of the current
1066  match, is subtly different from Perl's, which defines it as the end of the  match, is subtly different from Perl's, which defines it as the end of the
1067  previous match. In Perl, these can be different when the previously matched  previous match. In Perl, these can be different when the previously matched
1068  string was empty. Because PCRE does just one match at a time, it cannot  string was empty. Because PCRE does just one match at a time, it cannot
1069  reproduce this behaviour.  reproduce this behaviour.
1070    .P
1071  If all the alternatives of a pattern begin with \\G, the expression is anchored  If all the alternatives of a pattern begin with \eG, the expression is anchored
1072  to the starting match position, and the "anchored" flag is set in the compiled  to the starting match position, and the "anchored" flag is set in the compiled
1073  regular expression.  regular expression.
1074    .
1075  .SH CIRCUMFLEX AND DOLLAR  .
1076    .SH "CIRCUMFLEX AND DOLLAR"
1077  .rs  .rs
1078  .sp  .sp
1079    The circumflex and dollar metacharacters are zero-width assertions. That is,
1080    they test for a particular condition being true without consuming any
1081    characters from the subject string.
1082    .P
1083  Outside a character class, in the default matching mode, the circumflex  Outside a character class, in the default matching mode, the circumflex
1084  character is an assertion which is true only if the current matching point is  character is an assertion that is true only if the current matching point is at
1085  at the start of the subject string. If the \fIstartoffset\fR argument of  the start of the subject string. If the \fIstartoffset\fP argument of
1086  \fBpcre_exec()\fR is non-zero, circumflex can never match if the PCRE_MULTILINE  \fBpcre_exec()\fP is non-zero, circumflex can never match if the PCRE_MULTILINE
1087  option is unset. Inside a character class, circumflex has an entirely different  option is unset. Inside a character class, circumflex has an entirely different
1088  meaning (see below).  meaning
1089    .\" HTML <a href="#characterclass">
1090    .\" </a>
1091    (see below).
1092    .\"
1093    .P
1094  Circumflex need not be the first character of the pattern if a number of  Circumflex need not be the first character of the pattern if a number of
1095  alternatives are involved, but it should be the first thing in each alternative  alternatives are involved, but it should be the first thing in each alternative
1096  in which it appears if the pattern is ever to match that branch. If all  in which it appears if the pattern is ever to match that branch. If all
# Line 290  possible alternatives start with a circu Line 1098  possible alternatives start with a circu
1098  constrained to match only at the start of the subject, it is said to be an  constrained to match only at the start of the subject, it is said to be an
1099  "anchored" pattern. (There are also other constructs that can cause a pattern  "anchored" pattern. (There are also other constructs that can cause a pattern
1100  to be anchored.)  to be anchored.)
1101    .P
1102  A dollar character is an assertion which is true only if the current matching  The dollar character is an assertion that is true only if the current matching
1103  point is at the end of the subject string, or immediately before a newline  point is at the end of the subject string, or immediately before a newline at
1104  character that is the last character in the string (by default). Dollar need  the end of the string (by default). Note, however, that it does not actually
1105  not be the last character of the pattern if a number of alternatives are  match the newline. Dollar need not be the last character of the pattern if a
1106  involved, but it should be the last item in any branch in which it appears.  number of alternatives are involved, but it should be the last item in any
1107  Dollar has no special meaning in a character class.  branch in which it appears. Dollar has no special meaning in a character class.
1108    .P
1109  The meaning of dollar can be changed so that it matches only at the very end of  The meaning of dollar can be changed so that it matches only at the very end of
1110  the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This  the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This
1111  does not affect the \\Z assertion.  does not affect the \eZ assertion.
1112    .P
1113  The meanings of the circumflex and dollar characters are changed if the  The meanings of the circumflex and dollar characters are changed if the
1114  PCRE_MULTILINE option is set. When this is the case, they match immediately  PCRE_MULTILINE option is set. When this is the case, a circumflex matches
1115  after and immediately before an internal newline character, respectively, in  immediately after internal newlines as well as at the start of the subject
1116  addition to matching at the start and end of the subject string. For example,  string. It does not match after a newline that ends the string. A dollar
1117  the pattern /^abc$/ matches the subject string "def\\nabc" in multiline mode,  matches before any newlines in the string, as well as at the very end, when
1118  but not otherwise. Consequently, patterns that are anchored in single line mode  PCRE_MULTILINE is set. When newline is specified as the two-character
1119  because all branches start with ^ are not anchored in multiline mode, and a  sequence CRLF, isolated CR and LF characters do not indicate newlines.
1120  match for circumflex is possible when the \fIstartoffset\fR argument of  .P
1121  \fBpcre_exec()\fR is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if  For example, the pattern /^abc$/ matches the subject string "def\enabc" (where
1122  PCRE_MULTILINE is set.  \en represents a newline) in multiline mode, but not otherwise. Consequently,
1123    patterns that are anchored in single line mode because all branches start with
1124  Note that the sequences \\A, \\Z, and \\z can be used to match the start and  ^ are not anchored in multiline mode, and a match for circumflex is possible
1125    when the \fIstartoffset\fP argument of \fBpcre_exec()\fP is non-zero. The
1126    PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set.
1127    .P
1128    Note that the sequences \eA, \eZ, and \ez can be used to match the start and
1129  end of the subject in both modes, and if all branches of a pattern start with  end of the subject in both modes, and if all branches of a pattern start with
1130  \\A it is always anchored, whether PCRE_MULTILINE is set or not.  \eA it is always anchored, whether or not PCRE_MULTILINE is set.
1131    .
1132  .SH FULL STOP (PERIOD, DOT)  .
1133    .\" HTML <a name="fullstopdot"></a>
1134    .SH "FULL STOP (PERIOD, DOT) AND \eN"
1135  .rs  .rs
1136  .sp  .sp
1137  Outside a character class, a dot in the pattern matches any one character in  Outside a character class, a dot in the pattern matches any one character in
1138  the subject, including a non-printing character, but not (by default) newline.  the subject string except (by default) a character that signifies the end of a
1139  In UTF-8 mode, a dot matches any UTF-8 character, which might be more than one  line.
1140  byte long, except (by default) for newline. If the PCRE_DOTALL option is set,  .P
1141  dots match newlines as well. The handling of dot is entirely independent of the  When a line ending is defined as a single character, dot never matches that
1142  handling of circumflex and dollar, the only relationship being that they both  character; when the two-character sequence CRLF is used, dot does not match CR
1143  involve newline characters. Dot has no special meaning in a character class.  if it is immediately followed by LF, but otherwise it matches all characters
1144    (including isolated CRs and LFs). When any Unicode line endings are being
1145  .SH MATCHING A SINGLE BYTE  recognized, dot does not match CR or LF or any of the other line ending
1146  .rs  characters.
1147  .sp  .P
1148  Outside a character class, the escape sequence \\C matches any one byte, both  The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL
1149  in and out of UTF-8 mode. Unlike a dot, it always matches a newline. The  option is set, a dot matches any one character, without exception. If the
1150  feature is provided in Perl in order to match individual bytes in UTF-8 mode.  two-character sequence CRLF is present in the subject string, it takes two dots
1151  Because it breaks up UTF-8 characters into individual bytes, what remains in  to match it.
1152  the string may be a malformed UTF-8 string. For this reason it is best avoided.  .P
1153    The handling of dot is entirely independent of the handling of circumflex and
1154  PCRE does not allow \\C to appear in lookbehind assertions (see below), because  dollar, the only relationship being that they both involve newlines. Dot has no
1155  in UTF-8 mode it makes it impossible to calculate the length of the lookbehind.  special meaning in a character class.
1156    .P
1157  .SH SQUARE BRACKETS  The escape sequence \eN behaves like a dot, except that it is not affected by
1158    the PCRE_DOTALL option. In other words, it matches any character except one
1159    that signifies the end of a line. Perl also uses \eN to match characters by
1160    name; PCRE does not support this.
1161    .
1162    .
1163    .SH "MATCHING A SINGLE DATA UNIT"
1164    .rs
1165    .sp
1166    Outside a character class, the escape sequence \eC matches any one data unit,
1167    whether or not a UTF mode is set. In the 8-bit library, one data unit is one
1168    byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is
1169    a 32-bit unit. Unlike a dot, \eC always
1170    matches line-ending characters. The feature is provided in Perl in order to
1171    match individual bytes in UTF-8 mode, but it is unclear how it can usefully be
1172    used. Because \eC breaks up characters into individual data units, matching one
1173    unit with \eC in a UTF mode means that the rest of the string may start with a
1174    malformed UTF character. This has undefined results, because PCRE assumes that
1175    it is dealing with valid UTF strings (and by default it checks this at the
1176    start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or
1177    PCRE_NO_UTF32_CHECK option is used).
1178    .P
1179    PCRE does not allow \eC to appear in lookbehind assertions
1180    .\" HTML <a href="#lookbehind">
1181    .\" </a>
1182    (described below)
1183    .\"
1184    in a UTF mode, because this would make it impossible to calculate the length of
1185    the lookbehind.
1186    .P
1187    In general, the \eC escape sequence is best avoided. However, one
1188    way of using it that avoids the problem of malformed UTF characters is to use a
1189    lookahead to check the length of the next character, as in this pattern, which
1190    could be used with a UTF-8 string (ignore white space and line breaks):
1191    .sp
1192      (?| (?=[\ex00-\ex7f])(\eC) |
1193          (?=[\ex80-\ex{7ff}])(\eC)(\eC) |
1194          (?=[\ex{800}-\ex{ffff}])(\eC)(\eC)(\eC) |
1195          (?=[\ex{10000}-\ex{1fffff}])(\eC)(\eC)(\eC)(\eC))
1196    .sp
1197    A group that starts with (?| resets the capturing parentheses numbers in each
1198    alternative (see
1199    .\" HTML <a href="#dupsubpatternnumber">
1200    .\" </a>
1201    "Duplicate Subpattern Numbers"
1202    .\"
1203    below). The assertions at the start of each branch check the next UTF-8
1204    character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
1205    character's individual bytes are then captured by the appropriate number of
1206    groups.
1207    .
1208    .
1209    .\" HTML <a name="characterclass"></a>
1210    .SH "SQUARE BRACKETS AND CHARACTER CLASSES"
1211  .rs  .rs
1212  .sp  .sp
1213  An opening square bracket introduces a character class, terminated by a closing  An opening square bracket introduces a character class, terminated by a closing
1214  square bracket. A closing square bracket on its own is not special. If a  square bracket. A closing square bracket on its own is not special by default.
1215  closing square bracket is required as a member of the class, it should be the  However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square
1216  first data character in the class (after an initial circumflex, if present) or  bracket causes a compile-time error. If a closing square bracket is required as
1217  escaped with a backslash.  a member of the class, it should be the first data character in the class
1218    (after an initial circumflex, if present) or escaped with a backslash.
1219  A character class matches a single character in the subject. In UTF-8 mode, the  .P
1220  character may occupy more than one byte. A matched character must be in the set  A character class matches a single character in the subject. In a UTF mode, the
1221  of characters defined by the class, unless the first character in the class  character may be more than one data unit long. A matched character must be in
1222  definition is a circumflex, in which case the subject character must not be in  the set of characters defined by the class, unless the first character in the
1223  the set defined by the class. If a circumflex is actually required as a member  class definition is a circumflex, in which case the subject character must not
1224  of the class, ensure it is not the first character, or escape it with a  be in the set defined by the class. If a circumflex is actually required as a
1225    member of the class, ensure it is not the first character, or escape it with a
1226  backslash.  backslash.
1227    .P
1228  For example, the character class [aeiou] matches any lower case vowel, while  For example, the character class [aeiou] matches any lower case vowel, while
1229  [^aeiou] matches any character that is not a lower case vowel. Note that a  [^aeiou] matches any character that is not a lower case vowel. Note that a
1230  circumflex is just a convenient notation for specifying the characters which  circumflex is just a convenient notation for specifying the characters that
1231  are in the class by enumerating those that are not. It is not an assertion: it  are in the class by enumerating those that are not. A class that starts with a
1232  still consumes a character from the subject string, and fails if the current  circumflex is not an assertion; it still consumes a character from the subject
1233  pointer is at the end of the string.  string, and therefore it fails if the current pointer is at the end of the
1234    string.
1235  In UTF-8 mode, characters with values greater than 255 can be included in a  .P
1236  class as a literal string of bytes, or by using the \\x{ escaping mechanism.  In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff)
1237    can be included in a class as a literal string of data units, or by using the
1238    \ex{ escaping mechanism.
1239    .P
1240  When caseless matching is set, any letters in a class represent both their  When caseless matching is set, any letters in a class represent both their
1241  upper case and lower case versions, so for example, a caseless [aeiou] matches  upper case and lower case versions, so for example, a caseless [aeiou] matches
1242  "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a  "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a
1243  caseful version would. PCRE does not support the concept of case for characters  caseful version would. In a UTF mode, PCRE always understands the concept of
1244  with values greater than 255.  case for characters whose values are less than 128, so caseless matching is
1245    always possible. For characters with higher values, the concept of case is
1246  The newline character is never treated in any special way in character classes,  supported if PCRE is compiled with Unicode property support, but not otherwise.
1247  whatever the setting of the PCRE_DOTALL or PCRE_MULTILINE options is. A class  If you want to use caseless matching in a UTF mode for characters 128 and
1248  such as [^a] will always match a newline.  above, you must ensure that PCRE is compiled with Unicode property support as
1249    well as with UTF support.
1250    .P
1251    Characters that might indicate line breaks are never treated in any special way
1252    when matching character classes, whatever line-ending sequence is in use, and
1253    whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class
1254    such as [^a] always matches one of these characters.
1255    .P
1256  The minus (hyphen) character can be used to specify a range of characters in a  The minus (hyphen) character can be used to specify a range of characters in a
1257  character class. For example, [d-m] matches any letter between d and m,  character class. For example, [d-m] matches any letter between d and m,
1258  inclusive. If a minus character is required in a class, it must be escaped with  inclusive. If a minus character is required in a class, it must be escaped with
1259  a backslash or appear in a position where it cannot be interpreted as  a backslash or appear in a position where it cannot be interpreted as
1260  indicating a range, typically as the first or last character in the class.  indicating a range, typically as the first or last character in the class, or
1261    immediately after a range. For example, [b-d-z] matches letters in the range b
1262    to d, a hyphen character, or z.
1263    .P
1264  It is not possible to have the literal character "]" as the end character of a  It is not possible to have the literal character "]" as the end character of a
1265  range. A pattern such as [W-]46] is interpreted as a class of two characters  range. A pattern such as [W-]46] is interpreted as a class of two characters
1266  ("W" and "-") followed by a literal string "46]", so it would match "W46]" or  ("W" and "-") followed by a literal string "46]", so it would match "W46]" or
1267  "-46]". However, if the "]" is escaped with a backslash it is interpreted as  "-46]". However, if the "]" is escaped with a backslash it is interpreted as
1268  the end of range, so [W-\\]46] is interpreted as a single class containing a  the end of range, so [W-\e]46] is interpreted as a class containing a range
1269  range followed by two separate characters. The octal or hexadecimal  followed by two other characters. The octal or hexadecimal representation of
1270  representation of "]" can also be used to end a range.  "]" can also be used to end a range.
1271    .P
1272    An error is generated if a POSIX character class (see below) or an escape
1273    sequence other than one that defines a single character appears at a point
1274    where a range ending character is expected. For example, [z-\exff] is valid,
1275    but [A-\ed] and [A-[:digit:]] are not.
1276    .P
1277  Ranges operate in the collating sequence of character values. They can also be  Ranges operate in the collating sequence of character values. They can also be
1278  used for characters specified numerically, for example [\\000-\\037]. In UTF-8  used for characters specified numerically, for example [\e000-\e037]. Ranges
1279  mode, ranges can include characters whose values are greater than 255, for  can include any characters that are valid for the current mode.
1280  example [\\x{100}-\\x{2ff}].  .P
   
1281  If a range that includes letters is used when caseless matching is set, it  If a range that includes letters is used when caseless matching is set, it
1282  matches the letters in either case. For example, [W-c] is equivalent to  matches the letters in either case. For example, [W-c] is equivalent to
1283  [][\\^_`wxyzabc], matched caselessly, and if character tables for the "fr"  [][\e\e^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character
1284  locale are in use, [\\xc8-\\xcb] matches accented E characters in both cases.  tables for a French locale are in use, [\exc8-\excb] matches accented E
1285    characters in both cases. In UTF modes, PCRE supports the concept of case for
1286  The character types \\d, \\D, \\s, \\S, \\w, and \\W may also appear in a  characters with values greater than 128 only when it is compiled with Unicode
1287  character class, and add the characters that they match to the class. For  property support.
1288  example, [\\dABCDEF] matches any hexadecimal digit. A circumflex can  .P
1289  conveniently be used with the upper case character types to specify a more  The character escape sequences \ed, \eD, \eh, \eH, \ep, \eP, \es, \eS, \ev,
1290  restricted set of characters than the matching lower case type. For example,  \eV, \ew, and \eW may appear in a character class, and add the characters that
1291  the class [^\\W_] matches any letter or digit, but not underscore.  they match to the class. For example, [\edABCDEF] matches any hexadecimal
1292    digit. In UTF modes, the PCRE_UCP option affects the meanings of \ed, \es, \ew
1293  All non-alphameric characters other than \\, -, ^ (at the start) and the  and their upper case partners, just as it does when they appear outside a
1294  terminating ] are non-special in character classes, but it does no harm if they  character class, as described in the section entitled
1295  are escaped.  .\" HTML <a href="#genericchartypes">
1296    .\" </a>
1297  .SH POSIX CHARACTER CLASSES  "Generic character types"
1298    .\"
1299    above. The escape sequence \eb has a different meaning inside a character
1300    class; it matches the backspace character. The sequences \eB, \eN, \eR, and \eX
1301    are not special inside a character class. Like any other unrecognized escape
1302    sequences, they are treated as the literal characters "B", "N", "R", and "X" by
1303    default, but cause an error if the PCRE_EXTRA option is set.
1304    .P
1305    A circumflex can conveniently be used with the upper case character types to
1306    specify a more restricted set of characters than the matching lower case type.
1307    For example, the class [^\eW_] matches any letter or digit, but not underscore,
1308    whereas [\ew] includes underscore. A positive character class should be read as
1309    "something OR something OR ..." and a negative class as "NOT something AND NOT
1310    something AND NOT ...".
1311    .P
1312    The only metacharacters that are recognized in character classes are backslash,
1313    hyphen (only where it can be interpreted as specifying a range), circumflex
1314    (only at the start), opening square bracket (only when it can be interpreted as
1315    introducing a POSIX class name - see the next section), and the terminating
1316    closing square bracket. However, escaping other non-alphanumeric characters
1317    does no harm.
1318    .
1319    .
1320    .SH "POSIX CHARACTER CLASSES"
1321  .rs  .rs
1322  .sp  .sp
1323  Perl supports the POSIX notation for character classes, which uses names  Perl supports the POSIX notation for character classes. This uses names
1324  enclosed by [: and :] within the enclosing square brackets. PCRE also supports  enclosed by [: and :] within the enclosing square brackets. PCRE also supports
1325  this notation. For example,  this notation. For example,
1326    .sp
1327    [01[:alpha:]%]    [01[:alpha:]%]
1328    .sp
1329  matches "0", "1", any alphabetic character, or "%". The supported class names  matches "0", "1", any alphabetic character, or "%". The supported class names
1330  are  are:
1331    .sp
1332    alnum    letters and digits    alnum    letters and digits
1333    alpha    letters    alpha    letters
1334    ascii    character codes 0 - 127    ascii    character codes 0 - 127
1335    blank    space or tab only    blank    space or tab only
1336    cntrl    control characters    cntrl    control characters
1337    digit    decimal digits (same as \\d)    digit    decimal digits (same as \ed)
1338    graph    printing characters, excluding space    graph    printing characters, excluding space
1339    lower    lower case letters    lower    lower case letters
1340    print    printing characters, including space    print    printing characters, including space
1341    punct    printing characters, excluding letters and digits    punct    printing characters, excluding letters and digits and space
1342    space    white space (not quite the same as \\s)    space    white space (the same as \es from PCRE 8.34)
1343    upper    upper case letters    upper    upper case letters
1344    word     "word" characters (same as \\w)    word     "word" characters (same as \ew)
1345    xdigit   hexadecimal digits    xdigit   hexadecimal digits
1346    .sp
1347  The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and  The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
1348  space (32). Notice that this list includes the VT character (code 11). This  and space (32). If locale-specific matching is taking place, there may be
1349  makes "space" different to \\s, which does not include VT (for Perl  additional space characters. "Space" used to be different to \es, which did not
1350  compatibility).  include VT, for Perl compatibility. However, Perl changed at release 5.18, and
1351    PCRE followed at release 8.34. "Space" and \es now match the same set of
1352    characters.
1353    .P
1354  The name "word" is a Perl extension, and "blank" is a GNU extension from Perl  The name "word" is a Perl extension, and "blank" is a GNU extension from Perl
1355  5.8. Another Perl extension is negation, which is indicated by a ^ character  5.8. Another Perl extension is negation, which is indicated by a ^ character
1356  after the colon. For example,  after the colon. For example,
1357    .sp
1358    [12[:^digit:]]    [12[:^digit:]]
1359    .sp
1360  matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX  matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX
1361  syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not  syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not
1362  supported, and an error is given if they are encountered.  supported, and an error is given if they are encountered.
1363    .P
1364  In UTF-8 mode, characters with values greater than 255 do not match any of  By default, characters with values greater than 128 do not match any of the
1365  the POSIX character classes.  POSIX character classes. However, if the PCRE_UCP option is passed to
1366    \fBpcre_compile()\fP, some of the classes are changed so that Unicode character
1367  .SH VERTICAL BAR  properties are used. This is achieved by replacing certain POSIX classes by
1368    other sequences, as follows:
1369    .sp
1370      [:alnum:]  becomes  \ep{Xan}
1371      [:alpha:]  becomes  \ep{L}
1372      [:blank:]  becomes  \eh
1373      [:digit:]  becomes  \ep{Nd}
1374      [:lower:]  becomes  \ep{Ll}
1375      [:space:]  becomes  \ep{Xps}
1376      [:upper:]  becomes  \ep{Lu}
1377      [:word:]   becomes  \ep{Xwd}
1378    .sp
1379    Negated versions, such as [:^alpha:] use \eP instead of \ep. Three other POSIX
1380    classes are handled specially in UCP mode:
1381    .TP 10
1382    [:graph:]
1383    This matches characters that have glyphs that mark the page when printed. In
1384    Unicode property terms, it matches all characters with the L, M, N, P, S, or Cf
1385    properties, except for:
1386    .sp
1387      U+061C           Arabic Letter Mark
1388      U+180E           Mongolian Vowel Separator
1389      U+2066 - U+2069  Various "isolate"s
1390    .sp
1391    .TP 10
1392    [:print:]
1393    This matches the same characters as [:graph:] plus space characters that are
1394    not controls, that is, characters with the Zs property.
1395    .TP 10
1396    [:punct:]
1397    This matches all characters that have the Unicode P (punctuation) property,
1398    plus those characters whose code points are less than 128 that have the S
1399    (Symbol) property.
1400    .P
1401    The other POSIX classes are unchanged, and match only characters with code
1402    points less than 128.
1403    .
1404    .
1405    .SH "VERTICAL BAR"
1406  .rs  .rs
1407  .sp  .sp
1408  Vertical bar characters are used to separate alternative patterns. For example,  Vertical bar characters are used to separate alternative patterns. For example,
1409  the pattern  the pattern
1410    .sp
1411    gilbert|sullivan    gilbert|sullivan
1412    .sp
1413  matches either "gilbert" or "sullivan". Any number of alternatives may appear,  matches either "gilbert" or "sullivan". Any number of alternatives may appear,
1414  and an empty alternative is permitted (matching the empty string).  and an empty alternative is permitted (matching the empty string). The matching
1415  The matching process tries each alternative in turn, from left to right,  process tries each alternative in turn, from left to right, and the first one
1416  and the first one that succeeds is used. If the alternatives are within a  that succeeds is used. If the alternatives are within a subpattern
1417  subpattern (defined below), "succeeds" means matching the rest of the main  .\" HTML <a href="#subpattern">
1418  pattern as well as the alternative in the subpattern.  .\" </a>
1419    (defined below),
1420  .SH INTERNAL OPTION SETTING  .\"
1421    "succeeds" means matching the rest of the main pattern as well as the
1422    alternative in the subpattern.
1423    .
1424    .
1425    .SH "INTERNAL OPTION SETTING"
1426  .rs  .rs
1427  .sp  .sp
1428  The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and  The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
1429  PCRE_EXTENDED options can be changed from within the pattern by a sequence of  PCRE_EXTENDED options (which are Perl-compatible) can be changed from within
1430  Perl option letters enclosed between "(?" and ")". The option letters are  the pattern by a sequence of Perl option letters enclosed between "(?" and ")".
1431    The option letters are
1432    .sp
1433    i  for PCRE_CASELESS    i  for PCRE_CASELESS
1434    m  for PCRE_MULTILINE    m  for PCRE_MULTILINE
1435    s  for PCRE_DOTALL    s  for PCRE_DOTALL
1436    x  for PCRE_EXTENDED    x  for PCRE_EXTENDED
1437    .sp
1438  For example, (?im) sets caseless, multiline matching. It is also possible to  For example, (?im) sets caseless, multiline matching. It is also possible to
1439  unset these options by preceding the letter with a hyphen, and a combined  unset these options by preceding the letter with a hyphen, and a combined
1440  setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and  setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and
1441  PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also  PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also
1442  permitted. If a letter appears both before and after the hyphen, the option is  permitted. If a letter appears both before and after the hyphen, the option is
1443  unset.  unset.
1444    .P
1445  When an option change occurs at top level (that is, not inside subpattern  The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be
1446  parentheses), the change applies to the remainder of the pattern that follows.  changed in the same way as the Perl-compatible options by using the characters
1447  If the change is placed right at the start of a pattern, PCRE extracts it into  J, U and X respectively.
1448  the global options (and it will therefore show up in data extracted by the  .P
1449  \fBpcre_fullinfo()\fR function).  When one of these option changes occurs at top level (that is, not inside
1450    subpattern parentheses), the change applies to the remainder of the pattern
1451  An option change within a subpattern affects only that part of the current  that follows. If the change is placed right at the start of a pattern, PCRE
1452  pattern that follows it, so  extracts it into the global options (and it will therefore show up in data
1453    extracted by the \fBpcre_fullinfo()\fP function).
1454    .P
1455    An option change within a subpattern (see below for a description of
1456    subpatterns) affects only that part of the subpattern that follows it, so
1457    .sp
1458    (a(?i)b)c    (a(?i)b)c
1459    .sp
1460  matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).  matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used).
1461  By this means, options can be made to have different settings in different  By this means, options can be made to have different settings in different
1462  parts of the pattern. Any changes made in one alternative do carry on  parts of the pattern. Any changes made in one alternative do carry on
1463  into subsequent branches within the same subpattern. For example,  into subsequent branches within the same subpattern. For example,
1464    .sp
1465    (a(?i)b|c)    (a(?i)b|c)
1466    .sp
1467  matches "ab", "aB", "c", and "C", even though when matching "C" the first  matches "ab", "aB", "c", and "C", even though when matching "C" the first
1468  branch is abandoned before the option setting. This is because the effects of  branch is abandoned before the option setting. This is because the effects of
1469  option settings happen at compile time. There would be some very weird  option settings happen at compile time. There would be some very weird
1470  behaviour otherwise.  behaviour otherwise.
1471    .P
1472  The PCRE-specific options PCRE_UNGREEDY and PCRE_EXTRA can be changed in the  \fBNote:\fP There are other PCRE-specific options that can be set by the
1473  same way as the Perl-compatible options by using the characters U and X  application when the compiling or matching functions are called. In some cases
1474  respectively. The (?X) flag setting is special in that it must always occur  the pattern can contain special leading sequences such as (*CRLF) to override
1475  earlier in the pattern than any of the additional features it turns on, even  what the application has set or what has been defaulted. Details are given in
1476  when it is at top level. It is best put at the start.  the section entitled
1477    .\" HTML <a href="#newlineseq">
1478    .\" </a>
1479    "Newline sequences"
1480    .\"
1481    above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading
1482    sequences that can be used to set UTF and Unicode property modes; they are
1483    equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP
1484    options, respectively. The (*UTF) sequence is a generic version that can be
1485    used with any of the libraries. However, the application can set the
1486    PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences.
1487    .
1488    .
1489    .\" HTML <a name="subpattern"></a>
1490  .SH SUBPATTERNS  .SH SUBPATTERNS
1491  .rs  .rs
1492  .sp  .sp
1493  Subpatterns are delimited by parentheses (round brackets), which can be nested.  Subpatterns are delimited by parentheses (round brackets), which can be nested.
1494  Marking part of a pattern as a subpattern does two things:  Turning part of a pattern into a subpattern does two things:
1495    .sp
1496  1. It localizes a set of alternatives. For example, the pattern  1. It localizes a set of alternatives. For example, the pattern
1497    .sp
1498    cat(aract|erpillar|)    cat(aract|erpillar|)
1499    .sp
1500  matches one of the words "cat", "cataract", or "caterpillar". Without the  matches "cataract", "caterpillar", or "cat". Without the parentheses, it would
1501  parentheses, it would match "cataract", "erpillar" or the empty string.  match "cataract", "erpillar" or an empty string.
1502    .sp
1503  2. It sets up the subpattern as a capturing subpattern (as defined above).  2. It sets up the subpattern as a capturing subpattern. This means that, when
1504  When the whole pattern matches, that portion of the subject string that matched  the whole pattern matches, that portion of the subject string that matched the
1505  the subpattern is passed back to the caller via the \fIovector\fR argument of  subpattern is passed back to the caller via the \fIovector\fP argument of the
1506  \fBpcre_exec()\fR. Opening parentheses are counted from left to right (starting  matching function. (This applies only to the traditional matching functions;
1507  from 1) to obtain the numbers of the capturing subpatterns.  the DFA matching functions do not support capturing.)
1508    .P
1509  For example, if the string "the red king" is matched against the pattern  Opening parentheses are counted from left to right (starting from 1) to obtain
1510    numbers for the capturing subpatterns. For example, if the string "the red
1511    king" is matched against the pattern
1512    .sp
1513    the ((red|white) (king|queen))    the ((red|white) (king|queen))
1514    .sp
1515  the captured substrings are "red king", "red", and "king", and are numbered 1,  the captured substrings are "red king", "red", and "king", and are numbered 1,
1516  2, and 3, respectively.  2, and 3, respectively.
1517    .P
1518  The fact that plain parentheses fulfil two functions is not always helpful.  The fact that plain parentheses fulfil two functions is not always helpful.
1519  There are often times when a grouping subpattern is required without a  There are often times when a grouping subpattern is required without a
1520  capturing requirement. If an opening parenthesis is followed by a question mark  capturing requirement. If an opening parenthesis is followed by a question mark
1521  and a colon, the subpattern does not do any capturing, and is not counted when  and a colon, the subpattern does not do any capturing, and is not counted when
1522  computing the number of any subsequent capturing subpatterns. For example, if  computing the number of any subsequent capturing subpatterns. For example, if
1523  the string "the white queen" is matched against the pattern  the string "the white queen" is matched against the pattern
1524    .sp
1525    the ((?:red|white) (king|queen))    the ((?:red|white) (king|queen))
1526    .sp
1527  the captured substrings are "white queen" and "queen", and are numbered 1 and  the captured substrings are "white queen" and "queen", and are numbered 1 and
1528  2. The maximum number of capturing subpatterns is 65535, and the maximum depth  2. The maximum number of capturing subpatterns is 65535.
1529  of nesting of all subpatterns, both capturing and non-capturing, is 200.  .P
   
1530  As a convenient shorthand, if any option settings are required at the start of  As a convenient shorthand, if any option settings are required at the start of
1531  a non-capturing subpattern, the option letters may appear between the "?" and  a non-capturing subpattern, the option letters may appear between the "?" and
1532  the ":". Thus the two patterns  the ":". Thus the two patterns
1533    .sp
1534    (?i:saturday|sunday)    (?i:saturday|sunday)
1535    (?:(?i)saturday|sunday)    (?:(?i)saturday|sunday)
1536    .sp
1537  match exactly the same set of strings. Because alternative branches are tried  match exactly the same set of strings. Because alternative branches are tried
1538  from left to right, and options are not reset until the end of the subpattern  from left to right, and options are not reset until the end of the subpattern
1539  is reached, an option setting in one branch does affect subsequent branches, so  is reached, an option setting in one branch does affect subsequent branches, so
1540  the above patterns match "SUNDAY" as well as "Saturday".  the above patterns match "SUNDAY" as well as "Saturday".
1541    .
1542  .SH NAMED SUBPATTERNS  .
1543    .\" HTML <a name="dupsubpatternnumber"></a>
1544    .SH "DUPLICATE SUBPATTERN NUMBERS"
1545    .rs
1546    .sp
1547    Perl 5.10 introduced a feature whereby each alternative in a subpattern uses
1548    the same numbers for its capturing parentheses. Such a subpattern starts with
1549    (?| and is itself a non-capturing subpattern. For example, consider this
1550    pattern:
1551    .sp
1552      (?|(Sat)ur|(Sun))day
1553    .sp
1554    Because the two alternatives are inside a (?| group, both sets of capturing
1555    parentheses are numbered one. Thus, when the pattern matches, you can look
1556    at captured substring number one, whichever alternative matched. This construct
1557    is useful when you want to capture part, but not all, of one of a number of
1558    alternatives. Inside a (?| group, parentheses are numbered as usual, but the
1559    number is reset at the start of each branch. The numbers of any capturing
1560    parentheses that follow the subpattern start after the highest number used in
1561    any branch. The following example is taken from the Perl documentation. The
1562    numbers underneath show in which buffer the captured content will be stored.
1563    .sp
1564      # before  ---------------branch-reset----------- after
1565      / ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
1566      # 1            2         2  3        2     3     4
1567    .sp
1568    A back reference to a numbered subpattern uses the most recent value that is
1569    set for that number by any subpattern. The following pattern matches "abcabc"
1570    or "defdef":
1571    .sp
1572      /(?|(abc)|(def))\e1/
1573    .sp
1574    In contrast, a subroutine call to a numbered subpattern always refers to the
1575    first one in the pattern with the given number. The following pattern matches
1576    "abcabc" or "defabc":
1577    .sp
1578      /(?|(abc)|(def))(?1)/
1579    .sp
1580    If a
1581    .\" HTML <a href="#conditions">
1582    .\" </a>
1583    condition test
1584    .\"
1585    for a subpattern's having matched refers to a non-unique number, the test is
1586    true if any of the subpatterns of that number have matched.
1587    .P
1588    An alternative approach to using this "branch reset" feature is to use
1589    duplicate named subpatterns, as described in the next section.
1590    .
1591    .
1592    .SH "NAMED SUBPATTERNS"
1593  .rs  .rs
1594  .sp  .sp
1595  Identifying capturing parentheses by number is simple, but it can be very hard  Identifying capturing parentheses by number is simple, but it can be very hard
1596  to keep track of the numbers in complicated regular expressions. Furthermore,  to keep track of the numbers in complicated regular expressions. Furthermore,
1597  if an expression is modified, the numbers may change. To help with the  if an expression is modified, the numbers may change. To help with this
1598  difficulty, PCRE supports the naming of subpatterns, something that Perl does  difficulty, PCRE supports the naming of subpatterns. This feature was not
1599  not provide. The Python syntax (?P<name>...) is used. Names consist of  added to Perl until release 5.10. Python had the feature earlier, and PCRE
1600  alphanumeric characters and underscores, and must be unique within a pattern.  introduced it at release 4.0, using the Python syntax. PCRE now supports both
1601    the Perl and the Python syntax. Perl allows identically numbered subpatterns to
1602  Named capturing parentheses are still allocated numbers as well as names. The  have different names, but PCRE does not.
1603  PCRE API provides function calls for extracting the name-to-number translation  .P
1604  table from a compiled pattern. For further details see the  In PCRE, a subpattern can be named in one of three ways: (?<name>...) or
1605    (?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing
1606    parentheses from other parts of the pattern, such as
1607    .\" HTML <a href="#backreferences">
1608    .\" </a>
1609    back references,
1610    .\"
1611    .\" HTML <a href="#recursion">
1612    .\" </a>
1613    recursion,
1614    .\"
1615    and
1616    .\" HTML <a href="#conditions">
1617    .\" </a>
1618    conditions,
1619    .\"
1620    can be made by name as well as by number.
1621    .P
1622    Names consist of up to 32 alphanumeric characters and underscores, but must
1623    start with a non-digit. Named capturing parentheses are still allocated numbers
1624    as well as names, exactly as if the names were not present. The PCRE API
1625    provides function calls for extracting the name-to-number translation table
1626    from a compiled pattern. There is also a convenience function for extracting a
1627    captured substring by name.
1628    .P
1629    By default, a name must be unique within a pattern, but it is possible to relax
1630    this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate
1631    names are also always permitted for subpatterns with the same number, set up as
1632    described in the previous section.) Duplicate names can be useful for patterns
1633    where only one instance of the named parentheses can match. Suppose you want to
1634    match the name of a weekday, either as a 3-letter abbreviation or as the full
1635    name, and in both cases you want to extract the abbreviation. This pattern
1636    (ignoring the line breaks) does the job:
1637    .sp
1638      (?<DN>Mon|Fri|Sun)(?:day)?|
1639      (?<DN>Tue)(?:sday)?|
1640      (?<DN>Wed)(?:nesday)?|
1641      (?<DN>Thu)(?:rsday)?|
1642      (?<DN>Sat)(?:urday)?
1643    .sp
1644    There are five capturing substrings, but only one is ever set after a match.
1645    (An alternative way of solving this problem is to use a "branch reset"
1646    subpattern, as described in the previous section.)
1647    .P
1648    The convenience function for extracting the data by name returns the substring
1649    for the first (and in this example, the only) subpattern of that name that
1650    matched. This saves searching to find which numbered subpattern it was.
1651    .P
1652    If you make a back reference to a non-unique named subpattern from elsewhere in
1653    the pattern, the subpatterns to which the name refers are checked in the order
1654    in which they appear in the overall pattern. The first one that is set is used
1655    for the reference. For example, this pattern matches both "foofoo" and
1656    "barbar" but not "foobar" or "barfoo":
1657    .sp
1658      (?:(?<n>foo)|(?<n>bar))\k<n>
1659    .sp
1660    .P
1661    If you make a subroutine call to a non-unique named subpattern, the one that
1662    corresponds to the first occurrence of the name is used. In the absence of
1663    duplicate numbers (see the previous section) this is the one with the lowest
1664    number.
1665    .P
1666    If you use a named reference in a condition
1667    test (see the
1668    .\"
1669    .\" HTML <a href="#conditions">
1670    .\" </a>
1671    section about conditions
1672    .\"
1673    below), either to check whether a subpattern has matched, or to check for
1674    recursion, all subpatterns with the same name are tested. If the condition is
1675    true for any one of them, the overall condition is true. This is the same
1676    behaviour as testing by number. For further details of the interfaces for
1677    handling named subpatterns, see the
1678  .\" HREF  .\" HREF
1679  \fBpcreapi\fR  \fBpcreapi\fP
1680  .\"  .\"
1681  documentation.  documentation.
1682    .P
1683    \fBWarning:\fP You cannot use different names to distinguish between two
1684    subpatterns with the same number because PCRE uses only the numbers when
1685    matching. For this reason, an error is given at compile time if different names
1686    are given to subpatterns with the same number. However, you can always give the
1687    same name to subpatterns with the same number, even when PCRE_DUPNAMES is not
1688    set.
1689    .
1690    .
1691  .SH REPETITION  .SH REPETITION
1692  .rs  .rs
1693  .sp  .sp
1694  Repetition is specified by quantifiers, which can follow any of the following  Repetition is specified by quantifiers, which can follow any of the following
1695  items:  items:
1696    .sp
1697    a literal data character    a literal data character
1698    the . metacharacter    the dot metacharacter
1699    the \\C escape sequence    the \eC escape sequence
1700    escapes such as \\d that match single characters    the \eX escape sequence
1701      the \eR escape sequence
1702      an escape such as \ed or \epL that matches a single character
1703    a character class    a character class
1704    a back reference (see next section)    a back reference (see next section)
1705    a parenthesized subpattern (unless it is an assertion)    a parenthesized subpattern (including assertions)
1706      a subroutine call to a subpattern (recursive or otherwise)
1707    .sp
1708  The general repetition quantifier specifies a minimum and maximum number of  The general repetition quantifier specifies a minimum and maximum number of
1709  permitted matches, by giving the two numbers in curly brackets (braces),  permitted matches, by giving the two numbers in curly brackets (braces),
1710  separated by a comma. The numbers must be less than 65536, and the first must  separated by a comma. The numbers must be less than 65536, and the first must
1711  be less than or equal to the second. For example:  be less than or equal to the second. For example:
1712    .sp
1713    z{2,4}    z{2,4}
1714    .sp
1715  matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special  matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special
1716  character. If the second number is omitted, but the comma is present, there is  character. If the second number is omitted, but the comma is present, there is
1717  no upper limit; if the second number and the comma are both omitted, the  no upper limit; if the second number and the comma are both omitted, the
1718  quantifier specifies an exact number of required matches. Thus  quantifier specifies an exact number of required matches. Thus
1719    .sp
1720    [aeiou]{3,}    [aeiou]{3,}
1721    .sp
1722  matches at least 3 successive vowels, but may match many more, while  matches at least 3 successive vowels, but may match many more, while
1723    .sp
1724    \\d{8}    \ed{8}
1725    .sp
1726  matches exactly 8 digits. An opening curly bracket that appears in a position  matches exactly 8 digits. An opening curly bracket that appears in a position
1727  where a quantifier is not allowed, or one that does not match the syntax of a  where a quantifier is not allowed, or one that does not match the syntax of a
1728  quantifier, is taken as a literal character. For example, {,6} is not a  quantifier, is taken as a literal character. For example, {,6} is not a
1729  quantifier, but a literal string of four characters.  quantifier, but a literal string of four characters.
1730    .P
1731  In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to individual  In UTF modes, quantifiers apply to characters rather than to individual data
1732  bytes. Thus, for example, \\x{100}{2} matches two UTF-8 characters, each of  units. Thus, for example, \ex{100}{2} matches two characters, each of
1733  which is represented by a two-byte sequence.  which is represented by a two-byte sequence in a UTF-8 string. Similarly,
1734    \eX{3} matches three Unicode extended grapheme clusters, each of which may be
1735    several data units long (and they may be of different lengths).
1736    .P
1737  The quantifier {0} is permitted, causing the expression to behave as if the  The quantifier {0} is permitted, causing the expression to behave as if the
1738  previous item and the quantifier were not present.  previous item and the quantifier were not present. This may be useful for
1739    subpatterns that are referenced as
1740  For convenience (and historical compatibility) the three most common  .\" HTML <a href="#subpatternsassubroutines">
1741  quantifiers have single-character abbreviations:  .\" </a>
1742    subroutines
1743    .\"
1744    from elsewhere in the pattern (but see also the section entitled
1745    .\" HTML <a href="#subdefine">
1746    .\" </a>
1747    "Defining subpatterns for use by reference only"
1748    .\"
1749    below). Items other than subpatterns that have a {0} quantifier are omitted
1750    from the compiled pattern.
1751    .P
1752    For convenience, the three most common quantifiers have single-character
1753    abbreviations:
1754    .sp
1755    *    is equivalent to {0,}    *    is equivalent to {0,}
1756    +    is equivalent to {1,}    +    is equivalent to {1,}
1757    ?    is equivalent to {0,1}    ?    is equivalent to {0,1}
1758    .sp
1759  It is possible to construct infinite loops by following a subpattern that can  It is possible to construct infinite loops by following a subpattern that can
1760  match no characters with a quantifier that has no upper limit, for example:  match no characters with a quantifier that has no upper limit, for example:
1761    .sp
1762    (a?)*    (a?)*
1763    .sp
1764  Earlier versions of Perl and PCRE used to give an error at compile time for  Earlier versions of Perl and PCRE used to give an error at compile time for
1765  such patterns. However, because there are cases where this can be useful, such  such patterns. However, because there are cases where this can be useful, such
1766  patterns are now accepted, but if any repetition of the subpattern does in fact  patterns are now accepted, but if any repetition of the subpattern does in fact
1767  match no characters, the loop is forcibly broken.  match no characters, the loop is forcibly broken.
1768    .P
1769  By default, the quantifiers are "greedy", that is, they match as much as  By default, the quantifiers are "greedy", that is, they match as much as
1770  possible (up to the maximum number of permitted times), without causing the  possible (up to the maximum number of permitted times), without causing the
1771  rest of the pattern to fail. The classic example of where this gives problems  rest of the pattern to fail. The classic example of where this gives problems
1772  is in trying to match comments in C programs. These appear between the  is in trying to match comments in C programs. These appear between /* and */
1773  sequences /* and */ and within the sequence, individual * and / characters may  and within the comment, individual * and / characters may appear. An attempt to
1774  appear. An attempt to match C comments by applying the pattern  match C comments by applying the pattern
1775    .sp
1776    /\\*.*\\*/    /\e*.*\e*/
1777    .sp
1778  to the string  to the string
1779    .sp
1780    /* first command */  not comment  /* second comment */    /* first comment */  not comment  /* second comment */
1781    .sp
1782  fails, because it matches the entire string owing to the greediness of the .*  fails, because it matches the entire string owing to the greediness of the .*
1783  item.  item.
1784    .P
1785  However, if a quantifier is followed by a question mark, it ceases to be  However, if a quantifier is followed by a question mark, it ceases to be
1786  greedy, and instead matches the minimum number of times possible, so the  greedy, and instead matches the minimum number of times possible, so the
1787  pattern  pattern
1788    .sp
1789    /\\*.*?\\*/    /\e*.*?\e*/
1790    .sp
1791  does the right thing with the C comments. The meaning of the various  does the right thing with the C comments. The meaning of the various
1792  quantifiers is not otherwise changed, just the preferred number of matches.  quantifiers is not otherwise changed, just the preferred number of matches.
1793  Do not confuse this use of question mark with its use as a quantifier in its  Do not confuse this use of question mark with its use as a quantifier in its
1794  own right. Because it has two uses, it can sometimes appear doubled, as in  own right. Because it has two uses, it can sometimes appear doubled, as in
1795    .sp
1796    \\d??\\d    \ed??\ed
1797    .sp
1798  which matches one digit by preference, but can match two if that is the only  which matches one digit by preference, but can match two if that is the only
1799  way the rest of the pattern matches.  way the rest of the pattern matches.
1800    .P
1801  If the PCRE_UNGREEDY option is set (an option which is not available in Perl),  If the PCRE_UNGREEDY option is set (an option that is not available in Perl),
1802  the quantifiers are not greedy by default, but individual ones can be made  the quantifiers are not greedy by default, but individual ones can be made
1803  greedy by following them with a question mark. In other words, it inverts the  greedy by following them with a question mark. In other words, it inverts the
1804  default behaviour.  default behaviour.
1805    .P
1806  When a parenthesized subpattern is quantified with a minimum repeat count that  When a parenthesized subpattern is quantified with a minimum repeat count that
1807  is greater than 1 or with a limited maximum, more store is required for the  is greater than 1 or with a limited maximum, more memory is required for the
1808  compiled pattern, in proportion to the size of the minimum or maximum.  compiled pattern, in proportion to the size of the minimum or maximum.
1809    .P
1810  If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent  If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent
1811  to Perl's /s) is set, thus allowing the . to match newlines, the pattern is  to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is
1812  implicitly anchored, because whatever follows will be tried against every  implicitly anchored, because whatever follows will be tried against every
1813  character position in the subject string, so there is no point in retrying the  character position in the subject string, so there is no point in retrying the
1814  overall match at any position after the first. PCRE normally treats such a  overall match at any position after the first. PCRE normally treats such a
1815  pattern as though it were preceded by \\A.  pattern as though it were preceded by \eA.
1816    .P
1817  In cases where it is known that the subject string contains no newlines, it is  In cases where it is known that the subject string contains no newlines, it is
1818  worth setting PCRE_DOTALL in order to obtain this optimization, or  worth setting PCRE_DOTALL in order to obtain this optimization, or
1819  alternatively using ^ to indicate anchoring explicitly.  alternatively using ^ to indicate anchoring explicitly.
1820    .P
1821  However, there is one situation where the optimization cannot be used. When .*  However, there are some cases where the optimization cannot be used. When .*
1822  is inside capturing parentheses that are the subject of a backreference  is inside capturing parentheses that are the subject of a back reference
1823  elsewhere in the pattern, a match at the start may fail, and a later one  elsewhere in the pattern, a match at the start may fail where a later one
1824  succeed. Consider, for example:  succeeds. Consider, for example:
1825    .sp
1826    (.*)abc\\1    (.*)abc\e1
1827    .sp
1828  If the subject is "xyz123abc123" the match point is the fourth character. For  If the subject is "xyz123abc123" the match point is the fourth character. For
1829  this reason, such a pattern is not implicitly anchored.  this reason, such a pattern is not implicitly anchored.
1830    .P
1831    Another case where implicit anchoring is not applied is when the leading .* is
1832    inside an atomic group. Once again, a match at the start may fail where a later
1833    one succeeds. Consider this pattern:
1834    .sp
1835      (?>.*?a)b
1836    .sp
1837    It matches "ab" in the subject "aab". The use of the backtracking control verbs
1838    (*PRUNE) and (*SKIP) also disable this optimization.
1839    .P
1840  When a capturing subpattern is repeated, the value captured is the substring  When a capturing subpattern is repeated, the value captured is the substring
1841  that matched the final iteration. For example, after  that matched the final iteration. For example, after
1842    .sp
1843    (tweedle[dume]{3}\\s*)+    (tweedle[dume]{3}\es*)+
1844    .sp
1845  has matched "tweedledum tweedledee" the value of the captured substring is  has matched "tweedledum tweedledee" the value of the captured substring is
1846  "tweedledee". However, if there are nested capturing subpatterns, the  "tweedledee". However, if there are nested capturing subpatterns, the
1847  corresponding captured values may have been set in previous iterations. For  corresponding captured values may have been set in previous iterations. For
1848  example, after  example, after
1849    .sp
1850    /(a|(b))+/    /(a|(b))+/
1851    .sp
1852  matches "aba" the value of the second captured substring is "b".  matches "aba" the value of the second captured substring is "b".
1853    .
1854  .SH ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS  .
1855    .\" HTML <a name="atomicgroup"></a>
1856    .SH "ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS"
1857  .rs  .rs
1858  .sp  .sp
1859  With both maximizing and minimizing repetition, failure of what follows  With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
1860  normally causes the repeated item to be re-evaluated to see if a different  repetition, failure of what follows normally causes the repeated item to be
1861  number of repeats allows the rest of the pattern to match. Sometimes it is  re-evaluated to see if a different number of repeats allows the rest of the
1862  useful to prevent this, either to change the nature of the match, or to cause  pattern to match. Sometimes it is useful to prevent this, either to change the
1863  it fail earlier than it otherwise might, when the author of the pattern knows  nature of the match, or to cause it fail earlier than it otherwise might, when
1864  there is no point in carrying on.  the author of the pattern knows there is no point in carrying on.
1865    .P
1866  Consider, for example, the pattern \\d+foo when applied to the subject line  Consider, for example, the pattern \ed+foo when applied to the subject line
1867    .sp
1868    123456bar    123456bar
1869    .sp
1870  After matching all 6 digits and then failing to match "foo", the normal  After matching all 6 digits and then failing to match "foo", the normal
1871  action of the matcher is to try again with only 5 digits matching the \\d+  action of the matcher is to try again with only 5 digits matching the \ed+
1872  item, and then with 4, and so on, before ultimately failing. "Atomic grouping"  item, and then with 4, and so on, before ultimately failing. "Atomic grouping"
1873  (a term taken from Jeffrey Friedl's book) provides the means for specifying  (a term taken from Jeffrey Friedl's book) provides the means for specifying
1874  that once a subpattern has matched, it is not to be re-evaluated in this way.  that once a subpattern has matched, it is not to be re-evaluated in this way.
1875    .P
1876  If we use atomic grouping for the previous example, the matcher would give up  If we use atomic grouping for the previous example, the matcher gives up
1877  immediately on failing to match "foo" the first time. The notation is a kind of  immediately on failing to match "foo" the first time. The notation is a kind of
1878  special parenthesis, starting with (?> as in this example:  special parenthesis, starting with (?> as in this example:
1879    .sp
1880    (?>\\d+)bar    (?>\ed+)foo
1881    .sp
1882  This kind of parenthesis "locks up" the  part of the pattern it contains once  This kind of parenthesis "locks up" the  part of the pattern it contains once
1883  it has matched, and a failure further into the pattern is prevented from  it has matched, and a failure further into the pattern is prevented from
1884  backtracking into it. Backtracking past it to previous items, however, works as  backtracking into it. Backtracking past it to previous items, however, works as
1885  normal.  normal.
1886    .P
1887  An alternative description is that a subpattern of this type matches the string  An alternative description is that a subpattern of this type matches the string
1888  of characters that an identical standalone pattern would match, if anchored at  of characters that an identical standalone pattern would match, if anchored at
1889  the current point in the subject string.  the current point in the subject string.
1890    .P
1891  Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as  Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as
1892  the above example can be thought of as a maximizing repeat that must swallow  the above example can be thought of as a maximizing repeat that must swallow
1893  everything it can. So, while both \\d+ and \\d+? are prepared to adjust the  everything it can. So, while both \ed+ and \ed+? are prepared to adjust the
1894  number of digits they match in order to make the rest of the pattern match,  number of digits they match in order to make the rest of the pattern match,
1895  (?>\\d+) can only match an entire sequence of digits.  (?>\ed+) can only match an entire sequence of digits.
1896    .P
1897  Atomic groups in general can of course contain arbitrarily complicated  Atomic groups in general can of course contain arbitrarily complicated
1898  subpatterns, and can be nested. However, when the subpattern for an atomic  subpatterns, and can be nested. However, when the subpattern for an atomic
1899  group is just a single repeated item, as in the example above, a simpler  group is just a single repeated item, as in the example above, a simpler
1900  notation, called a "possessive quantifier" can be used. This consists of an  notation, called a "possessive quantifier" can be used. This consists of an
1901  additional + character following a quantifier. Using this notation, the  additional + character following a quantifier. Using this notation, the
1902  previous example can be rewritten as  previous example can be rewritten as
1903    .sp
1904    \\d++bar    \ed++foo
1905    .sp
1906    Note that a possessive quantifier can be used with an entire group, for
1907    example:
1908    .sp
1909      (abc|xyz){2,3}+
1910    .sp
1911  Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY  Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY
1912  option is ignored. They are a convenient notation for the simpler forms of  option is ignored. They are a convenient notation for the simpler forms of
1913  atomic group. However, there is no difference in the meaning or processing of a  atomic group. However, there is no difference in the meaning of a possessive
1914  possessive quantifier and the equivalent atomic group.  quantifier and the equivalent atomic group, though there may be a performance
1915    difference; possessive quantifiers should be slightly faster.
1916  The possessive quantifier syntax is an extension to the Perl syntax. It  .P
1917  originates in Sun's Java package.  The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
1918    Jeffrey Friedl originated the idea (and the name) in the first edition of his
1919    book. Mike McCloskey liked it, so implemented it when he built Sun's Java
1920    package, and PCRE copied it from there. It ultimately found its way into Perl
1921    at release 5.10.
1922    .P
1923    PCRE has an optimization that automatically "possessifies" certain simple
1924    pattern constructs. For example, the sequence A+B is treated as A++B because
1925    there is no point in backtracking into a sequence of A's when B must follow.
1926    .P
1927  When a pattern contains an unlimited repeat inside a subpattern that can itself  When a pattern contains an unlimited repeat inside a subpattern that can itself
1928  be repeated an unlimited number of times, the use of an atomic group is the  be repeated an unlimited number of times, the use of an atomic group is the
1929  only way to avoid some failing matches taking a very long time indeed. The  only way to avoid some failing matches taking a very long time indeed. The
1930  pattern  pattern
1931    .sp
1932    (\\D+|<\\d+>)*[!?]    (\eD+|<\ed+>)*[!?]
1933    .sp
1934  matches an unlimited number of substrings that either consist of non-digits, or  matches an unlimited number of substrings that either consist of non-digits, or
1935  digits enclosed in <>, followed by either ! or ?. When it matches, it runs  digits enclosed in <>, followed by either ! or ?. When it matches, it runs
1936  quickly. However, if it is applied to  quickly. However, if it is applied to
1937    .sp
1938    aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa    aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1939    .sp
1940  it takes a long time before reporting failure. This is because the string can  it takes a long time before reporting failure. This is because the string can
1941  be divided between the two repeats in a large number of ways, and all have to  be divided between the internal \eD+ repeat and the external * repeat in a
1942  be tried. (The example used [!?] rather than a single character at the end,  large number of ways, and all have to be tried. (The example uses [!?] rather
1943  because both PCRE and Perl have an optimization that allows for fast failure  than a single character at the end, because both PCRE and Perl have an
1944  when a single character is used. They remember the last single character that  optimization that allows for fast failure when a single character is used. They
1945  is required for a match, and fail early if it is not present in the string.)  remember the last single character that is required for a match, and fail early
1946  If the pattern is changed to  if it is not present in the string.) If the pattern is changed so that it uses
1947    an atomic group, like this:
1948    ((?>\\D+)|<\\d+>)*[!?]  .sp
1949      ((?>\eD+)|<\ed+>)*[!?]
1950    .sp
1951  sequences of non-digits cannot be broken, and failure happens quickly.  sequences of non-digits cannot be broken, and failure happens quickly.
1952    .
1953  .SH BACK REFERENCES  .
1954    .\" HTML <a name="backreferences"></a>
1955    .SH "BACK REFERENCES"
1956  .rs  .rs
1957  .sp  .sp
1958  Outside a character class, a backslash followed by a digit greater than 0 (and  Outside a character class, a backslash followed by a digit greater than 0 (and
1959  possibly further digits) is a back reference to a capturing subpattern earlier  possibly further digits) is a back reference to a capturing subpattern earlier
1960  (that is, to its left) in the pattern, provided there have been that many  (that is, to its left) in the pattern, provided there have been that many
1961  previous capturing left parentheses.  previous capturing left parentheses.
1962    .P
1963  However, if the decimal number following the backslash is less than 10, it is  However, if the decimal number following the backslash is less than 10, it is
1964  always taken as a back reference, and causes an error only if there are not  always taken as a back reference, and causes an error only if there are not
1965  that many capturing left parentheses in the entire pattern. In other words, the  that many capturing left parentheses in the entire pattern. In other words, the
1966  parentheses that are referenced need not be to the left of the reference for  parentheses that are referenced need not be to the left of the reference for
1967  numbers less than 10. See the section entitled "Backslash" above for further  numbers less than 10. A "forward back reference" of this type can make sense
1968  details of the handling of digits following a backslash.  when a repetition is involved and the subpattern to the right has participated
1969    in an earlier iteration.
1970    .P
1971    It is not possible to have a numerical "forward back reference" to a subpattern
1972    whose number is 10 or more using this syntax because a sequence such as \e50 is
1973    interpreted as a character defined in octal. See the subsection entitled
1974    "Non-printing characters"
1975    .\" HTML <a href="#digitsafterbackslash">
1976    .\" </a>
1977    above
1978    .\"
1979    for further details of the handling of digits following a backslash. There is
1980    no such problem when named parentheses are used. A back reference to any
1981    subpattern is possible using named parentheses (see below).
1982    .P
1983    Another way of avoiding the ambiguity inherent in the use of digits following a
1984    backslash is to use the \eg escape sequence. This escape must be followed by an
1985    unsigned number or a negative number, optionally enclosed in braces. These
1986    examples are all identical:
1987    .sp
1988      (ring), \e1
1989      (ring), \eg1
1990      (ring), \eg{1}
1991    .sp
1992    An unsigned number specifies an absolute reference without the ambiguity that
1993    is present in the older syntax. It is also useful when literal digits follow
1994    the reference. A negative number is a relative reference. Consider this
1995    example:
1996    .sp
1997      (abc(def)ghi)\eg{-1}
1998    .sp
1999    The sequence \eg{-1} is a reference to the most recently started capturing
2000    subpattern before \eg, that is, is it equivalent to \e2 in this example.
2001    Similarly, \eg{-2} would be equivalent to \e1. The use of relative references
2002    can be helpful in long patterns, and also in patterns that are created by
2003    joining together fragments that contain references within themselves.
2004    .P
2005  A back reference matches whatever actually matched the capturing subpattern in  A back reference matches whatever actually matched the capturing subpattern in
2006  the current subject string, rather than anything matching the subpattern  the current subject string, rather than anything matching the subpattern
2007  itself (see  itself (see
# Line 832  itself (see Line 2010  itself (see
2010  "Subpatterns as subroutines"  "Subpatterns as subroutines"
2011  .\"  .\"
2012  below for a way of doing that). So the pattern  below for a way of doing that). So the pattern
2013    .sp
2014    (sens|respons)e and \\1ibility    (sens|respons)e and \e1ibility
2015    .sp
2016  matches "sense and sensibility" and "response and responsibility", but not  matches "sense and sensibility" and "response and responsibility", but not
2017  "sense and responsibility". If caseful matching is in force at the time of the  "sense and responsibility". If caseful matching is in force at the time of the
2018  back reference, the case of letters is relevant. For example,  back reference, the case of letters is relevant. For example,
2019    .sp
2020    ((?i)rah)\\s+\\1    ((?i)rah)\es+\e1
2021    .sp
2022  matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original  matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original
2023  capturing subpattern is matched caselessly.  capturing subpattern is matched caselessly.
2024    .P
2025  Back references to named subpatterns use the Python syntax (?P=name). We could  There are several different ways of writing back references to named
2026  rewrite the above example as follows:  subpatterns. The .NET syntax \ek{name} and the Perl syntax \ek<name> or
2027    \ek'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified
2028    (?<p1>(?i)rah)\\s+(?P=p1)  back reference syntax, in which \eg can be used for both numeric and named
2029    references, is also supported. We could rewrite the above example in any of
2030    the following ways:
2031    .sp
2032      (?<p1>(?i)rah)\es+\ek<p1>
2033      (?'p1'(?i)rah)\es+\ek{p1}
2034      (?P<p1>(?i)rah)\es+(?P=p1)
2035      (?<p1>(?i)rah)\es+\eg{p1}
2036    .sp
2037    A subpattern that is referenced by name may appear in the pattern before or
2038    after the reference.
2039    .P
2040  There may be more than one back reference to the same subpattern. If a  There may be more than one back reference to the same subpattern. If a
2041  subpattern has not actually been used in a particular match, any back  subpattern has not actually been used in a particular match, any back
2042  references to it always fail. For example, the pattern  references to it always fail by default. For example, the pattern
2043    .sp
2044    (a|(bc))\\2    (a|(bc))\e2
2045    .sp
2046  always fails if it starts to match "a" rather than "bc". Because there may be  always fails if it starts to match "a" rather than "bc". However, if the
2047  many capturing parentheses in a pattern, all digits following the backslash are  PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an
2048  taken as part of a potential back reference number. If the pattern continues  unset value matches an empty string.
2049  with a digit character, some delimiter must be used to terminate the back  .P
2050  reference. If the PCRE_EXTENDED option is set, this can be whitespace.  Because there may be many capturing parentheses in a pattern, all digits
2051  Otherwise an empty comment can be used.  following a backslash are taken as part of a potential back reference number.
2052    If the pattern continues with a digit character, some delimiter must be used to
2053    terminate the back reference. If the PCRE_EXTENDED option is set, this can be
2054    white space. Otherwise, the \eg{ syntax or an empty comment (see
2055    .\" HTML <a href="#comments">
2056    .\" </a>
2057    "Comments"
2058    .\"
2059    below) can be used.
2060    .
2061    .SS "Recursive back references"
2062    .rs
2063    .sp
2064  A back reference that occurs inside the parentheses to which it refers fails  A back reference that occurs inside the parentheses to which it refers fails
2065  when the subpattern is first used, so, for example, (a\\1) never matches.  when the subpattern is first used, so, for example, (a\e1) never matches.
2066  However, such references can be useful inside repeated subpatterns. For  However, such references can be useful inside repeated subpatterns. For
2067  example, the pattern  example, the pattern
2068    .sp
2069    (a|b\\1)+    (a|b\e1)+
2070    .sp
2071  matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of  matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of
2072  the subpattern, the back reference matches the character string corresponding  the subpattern, the back reference matches the character string corresponding
2073  to the previous iteration. In order for this to work, the pattern must be such  to the previous iteration. In order for this to work, the pattern must be such
2074  that the first iteration does not need to match the back reference. This can be  that the first iteration does not need to match the back reference. This can be
2075  done using alternation, as in the example above, or by a quantifier with a  done using alternation, as in the example above, or by a quantifier with a
2076  minimum of zero.  minimum of zero.
2077    .P
2078    Back references of this type cause the group that they reference to be treated
2079    as an
2080    .\" HTML <a href="#atomicgroup">
2081    .\" </a>
2082    atomic group.
2083    .\"
2084    Once the whole group has been matched, a subsequent matching failure cannot
2085    cause backtracking into the middle of the group.
2086    .
2087    .
2088    .\" HTML <a name="bigassertions"></a>
2089  .SH ASSERTIONS  .SH ASSERTIONS
2090  .rs  .rs
2091  .sp  .sp
2092  An assertion is a test on the characters following or preceding the current  An assertion is a test on the characters following or preceding the current
2093  matching point that does not actually consume any characters. The simple  matching point that does not actually consume any characters. The simple
2094  assertions coded as \\b, \\B, \\A, \\G, \\Z, \\z, ^ and $ are described above.  assertions coded as \eb, \eB, \eA, \eG, \eZ, \ez, ^ and $ are described
2095    .\" HTML <a href="#smallassertions">
2096    .\" </a>
2097    above.
2098    .\"
2099    .P
2100  More complicated assertions are coded as subpatterns. There are two kinds:  More complicated assertions are coded as subpatterns. There are two kinds:
2101  those that look ahead of the current position in the subject string, and those  those that look ahead of the current position in the subject string, and those
2102  that look behind it.  that look behind it. An assertion subpattern is matched in the normal way,
2103    except that it does not cause the current matching position to be changed.
2104  An assertion subpattern is matched in the normal way, except that it does not  .P
2105  cause the current matching position to be changed. Lookahead assertions start  Assertion subpatterns are not capturing subpatterns. If such an assertion
2106  with (?= for positive assertions and (?! for negative assertions. For example,  contains capturing subpatterns within it, these are counted for the purposes of
2107    numbering the capturing subpatterns in the whole pattern. However, substring
2108    \\w+(?=;)  capturing is carried out only for positive assertions. (Perl sometimes, but not
2109    always, does do capturing in negative assertions.)
2110    .P
2111    For compatibility with Perl, assertion subpatterns may be repeated; though
2112    it makes no sense to assert the same thing several times, the side effect of
2113    capturing parentheses may occasionally be useful. In practice, there only three
2114    cases:
2115    .sp
2116    (1) If the quantifier is {0}, the assertion is never obeyed during matching.
2117    However, it may contain internal capturing parenthesized groups that are called
2118    from elsewhere via the
2119    .\" HTML <a href="#subpatternsassubroutines">
2120    .\" </a>
2121    subroutine mechanism.
2122    .\"
2123    .sp
2124    (2) If quantifier is {0,n} where n is greater than zero, it is treated as if it
2125    were {0,1}. At run time, the rest of the pattern match is tried with and
2126    without the assertion, the order depending on the greediness of the quantifier.
2127    .sp
2128    (3) If the minimum repetition is greater than zero, the quantifier is ignored.
2129    The assertion is obeyed just once when encountered during matching.
2130    .
2131    .
2132    .SS "Lookahead assertions"
2133    .rs
2134    .sp
2135    Lookahead assertions start with (?= for positive assertions and (?! for
2136    negative assertions. For example,
2137    .sp
2138      \ew+(?=;)
2139    .sp
2140  matches a word followed by a semicolon, but does not include the semicolon in  matches a word followed by a semicolon, but does not include the semicolon in
2141  the match, and  the match, and
2142    .sp
2143    foo(?!bar)    foo(?!bar)
2144    .sp
2145  matches any occurrence of "foo" that is not followed by "bar". Note that the  matches any occurrence of "foo" that is not followed by "bar". Note that the
2146  apparently similar pattern  apparently similar pattern
2147    .sp
2148    (?!foo)bar    (?!foo)bar
2149    .sp
2150  does not find an occurrence of "bar" that is preceded by something other than  does not find an occurrence of "bar" that is preceded by something other than
2151  "foo"; it finds any occurrence of "bar" whatsoever, because the assertion  "foo"; it finds any occurrence of "bar" whatsoever, because the assertion
2152  (?!foo) is always true when the next three characters are "bar". A  (?!foo) is always true when the next three characters are "bar". A
2153  lookbehind assertion is needed to achieve this effect.  lookbehind assertion is needed to achieve the other effect.
2154    .P
2155  If you want to force a matching failure at some point in a pattern, the most  If you want to force a matching failure at some point in a pattern, the most
2156  convenient way to do it is with (?!) because an empty string always matches, so  convenient way to do it is with (?!) because an empty string always matches, so
2157  an assertion that requires there not to be an empty string must always fail.  an assertion that requires there not to be an empty string must always fail.
2158    The backtracking control verb (*FAIL) or (*F) is a synonym for (?!).
2159    .
2160    .
2161    .\" HTML <a name="lookbehind"></a>
2162    .SS "Lookbehind assertions"
2163    .rs
2164    .sp
2165  Lookbehind assertions start with (?<= for positive assertions and (?<! for  Lookbehind assertions start with (?<= for positive assertions and (?<! for
2166  negative assertions. For example,  negative assertions. For example,
2167    .sp
2168    (?<!foo)bar    (?<!foo)bar
2169    .sp
2170  does find an occurrence of "bar" that is not preceded by "foo". The contents of  does find an occurrence of "bar" that is not preceded by "foo". The contents of
2171  a lookbehind assertion are restricted such that all the strings it matches must  a lookbehind assertion are restricted such that all the strings it matches must
2172  have a fixed length. However, if there are several alternatives, they do not  have a fixed length. However, if there are several top-level alternatives, they
2173  all have to have the same fixed length. Thus  do not all have to have the same fixed length. Thus
2174    .sp
2175    (?<=bullock|donkey)    (?<=bullock|donkey)
2176    .sp
2177  is permitted, but  is permitted, but
2178    .sp
2179    (?<!dogs?|cats?)    (?<!dogs?|cats?)
2180    .sp
2181  causes an error at compile time. Branches that match different length strings  causes an error at compile time. Branches that match different length strings
2182  are permitted only at the top level of a lookbehind assertion. This is an  are permitted only at the top level of a lookbehind assertion. This is an
2183  extension compared with Perl (at least for 5.8), which requires all branches to  extension compared with Perl, which requires all branches to match the same
2184  match the same length of string. An assertion such as  length of string. An assertion such as
2185    .sp
2186    (?<=ab(c|de))    (?<=ab(c|de))
2187    .sp
2188  is not permitted, because its single top-level branch can match two different  is not permitted, because its single top-level branch can match two different
2189  lengths, but it is acceptable if rewritten to use two top-level branches:  lengths, but it is acceptable to PCRE if rewritten to use two top-level
2190    branches:
2191    .sp
2192    (?<=abc|abde)    (?<=abc|abde)
2193    .sp
2194    In some cases, the escape sequence \eK
2195    .\" HTML <a href="#resetmatchstart">
2196    .\" </a>
2197    (see above)
2198    .\"
2199    can be used instead of a lookbehind assertion to get round the fixed-length
2200    restriction.
2201    .P
2202  The implementation of lookbehind assertions is, for each alternative, to  The implementation of lookbehind assertions is, for each alternative, to
2203  temporarily move the current position back by the fixed width and then try to  temporarily move the current position back by the fixed length and then try to
2204  match. If there are insufficient characters before the current position, the  match. If there are insufficient characters before the current position, the
2205  match is deemed to fail.  assertion fails.
2206    .P
2207  PCRE does not allow the \\C escape (which matches a single byte in UTF-8 mode)  In a UTF mode, PCRE does not allow the \eC escape (which matches a single data
2208  to appear in lookbehind assertions, because it makes it impossible to calculate  unit even in a UTF mode) to appear in lookbehind assertions, because it makes
2209  the length of the lookbehind.  it impossible to calculate the length of the lookbehind. The \eX and \eR
2210    escapes, which can match different numbers of data units, are also not
2211  Atomic groups can be used in conjunction with lookbehind assertions to specify  permitted.
2212  efficient matching at the end of the subject string. Consider a simple pattern  .P
2213  such as  .\" HTML <a href="#subpatternsassubroutines">
2214    .\" </a>
2215    "Subroutine"
2216    .\"
2217    calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long
2218    as the subpattern matches a fixed-length string.
2219    .\" HTML <a href="#recursion">
2220    .\" </a>
2221    Recursion,
2222    .\"
2223    however, is not supported.
2224    .P
2225    Possessive quantifiers can be used in conjunction with lookbehind assertions to
2226    specify efficient matching of fixed-length strings at the end of subject
2227    strings. Consider a simple pattern such as
2228    .sp
2229    abcd$    abcd$
2230    .sp
2231  when applied to a long string that does not match. Because matching proceeds  when applied to a long string that does not match. Because matching proceeds
2232  from left to right, PCRE will look for each "a" in the subject and then see if  from left to right, PCRE will look for each "a" in the subject and then see if
2233  what follows matches the rest of the pattern. If the pattern is specified as  what follows matches the rest of the pattern. If the pattern is specified as
2234    .sp
2235    ^.*abcd$    ^.*abcd$
2236    .sp
2237  the initial .* matches the entire string at first, but when this fails (because  the initial .* matches the entire string at first, but when this fails (because
2238  there is no following "a"), it backtracks to match all but the last character,  there is no following "a"), it backtracks to match all but the last character,
2239  then all but the last two characters, and so on. Once again the search for "a"  then all but the last two characters, and so on. Once again the search for "a"
2240  covers the entire string, from right to left, so we are no better off. However,  covers the entire string, from right to left, so we are no better off. However,
2241  if the pattern is written as  if the pattern is written as
2242    .sp
   ^(?>.*)(?<=abcd)  
   
 or, equivalently,  
   
2243    ^.*+(?<=abcd)    ^.*+(?<=abcd)
2244    .sp
2245  there can be no backtracking for the .* item; it can match only the entire  there can be no backtracking for the .*+ item; it can match only the entire
2246  string. The subsequent lookbehind assertion does a single test on the last four  string. The subsequent lookbehind assertion does a single test on the last four
2247  characters. If it fails, the match fails immediately. For long strings, this  characters. If it fails, the match fails immediately. For long strings, this
2248  approach makes a significant difference to the processing time.  approach makes a significant difference to the processing time.
2249    .
2250    .
2251    .SS "Using multiple assertions"
2252    .rs
2253    .sp
2254  Several assertions (of any sort) may occur in succession. For example,  Several assertions (of any sort) may occur in succession. For example,
2255    .sp
2256    (?<=\\d{3})(?<!999)foo    (?<=\ed{3})(?<!999)foo
2257    .sp
2258  matches "foo" preceded by three digits that are not "999". Notice that each of  matches "foo" preceded by three digits that are not "999". Notice that each of
2259  the assertions is applied independently at the same point in the subject  the assertions is applied independently at the same point in the subject
2260  string. First there is a check that the previous three characters are all  string. First there is a check that the previous three characters are all
2261  digits, and then there is a check that the same three characters are not "999".  digits, and then there is a check that the same three characters are not "999".
2262  This pattern does \fInot\fR match "foo" preceded by six characters, the first  This pattern does \fInot\fP match "foo" preceded by six characters, the first
2263  of which are digits and the last three of which are not "999". For example, it  of which are digits and the last three of which are not "999". For example, it
2264  doesn't match "123abcfoo". A pattern to do that is  doesn't match "123abcfoo". A pattern to do that is
2265    .sp
2266    (?<=\\d{3}...)(?<!999)foo    (?<=\ed{3}...)(?<!999)foo
2267    .sp
2268  This time the first assertion looks at the preceding six characters, checking  This time the first assertion looks at the preceding six characters, checking
2269  that the first three are digits, and then the second assertion checks that the  that the first three are digits, and then the second assertion checks that the
2270  preceding three characters are not "999".  preceding three characters are not "999".
2271    .P
2272  Assertions can be nested in any combination. For example,  Assertions can be nested in any combination. For example,
2273    .sp
2274    (?<=(?<!foo)bar)baz    (?<=(?<!foo)bar)baz
2275    .sp
2276  matches an occurrence of "baz" that is preceded by "bar" which in turn is not  matches an occurrence of "baz" that is preceded by "bar" which in turn is not
2277  preceded by "foo", while  preceded by "foo", while
2278    .sp
2279    (?<=\\d{3}(?!999)...)foo    (?<=\ed{3}(?!999)...)foo
2280    .sp
2281  is another pattern which matches "foo" preceded by three digits and any three  is another pattern that matches "foo" preceded by three digits and any three
2282  characters that are not "999".  characters that are not "999".
2283    .
2284  Assertion subpatterns are not capturing subpatterns, and may not be repeated,  .
2285  because it makes no sense to assert the same thing several times. If any kind  .\" HTML <a name="conditions"></a>
2286  of assertion contains capturing subpatterns within it, these are counted for  .SH "CONDITIONAL SUBPATTERNS"
 the purposes of numbering the capturing subpatterns in the whole pattern.  
 However, substring capturing is carried out only for positive assertions,  
 because it does not make sense for negative assertions.  
   
 .SH CONDITIONAL SUBPATTERNS  
2287  .rs  .rs
2288  .sp  .sp
2289  It is possible to cause the matching process to obey a subpattern  It is possible to cause the matching process to obey a subpattern
2290  conditionally or to choose between two alternative subpatterns, depending on  conditionally or to choose between two alternative subpatterns, depending on
2291  the result of an assertion, or whether a previous capturing subpattern matched  the result of an assertion, or whether a specific capturing subpattern has
2292  or not. The two possible forms of conditional subpattern are  already been matched. The two possible forms of conditional subpattern are:
2293    .sp
2294    (?(condition)yes-pattern)    (?(condition)yes-pattern)
2295    (?(condition)yes-pattern|no-pattern)    (?(condition)yes-pattern|no-pattern)
2296    .sp
2297  If the condition is satisfied, the yes-pattern is used; otherwise the  If the condition is satisfied, the yes-pattern is used; otherwise the
2298  no-pattern (if present) is used. If there are more than two alternatives in the  no-pattern (if present) is used. If there are more than two alternatives in the
2299  subpattern, a compile-time error occurs.  subpattern, a compile-time error occurs. Each of the two alternatives may
2300    itself contain nested subpatterns of any form, including conditional
2301  There are three kinds of condition. If the text between the parentheses  subpatterns; the restriction to two alternatives applies only at the level of
2302  consists of a sequence of digits, the condition is satisfied if the capturing  the condition. This pattern fragment is an example where the alternatives are
2303  subpattern of that number has previously matched. The number must be greater  complex:
2304  than zero. Consider the following pattern, which contains non-significant white  .sp
2305  space to make it more readable (assume the PCRE_EXTENDED option) and to divide    (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
2306  it into three parts for ease of discussion:  .sp
2307    .P
2308    ( \\( )?    [^()]+    (?(1) \\) )  There are four kinds of condition: references to subpatterns, references to
2309    recursion, a pseudo-condition called DEFINE, and assertions.
2310    .
2311    .SS "Checking for a used subpattern by number"
2312    .rs
2313    .sp
2314    If the text between the parentheses consists of a sequence of digits, the
2315    condition is true if a capturing subpattern of that number has previously
2316    matched. If there is more than one capturing subpattern with the same number
2317    (see the earlier
2318    .\"
2319    .\" HTML <a href="#recursion">
2320    .\" </a>
2321    section about duplicate subpattern numbers),
2322    .\"
2323    the condition is true if any of them have matched. An alternative notation is
2324    to precede the digits with a plus or minus sign. In this case, the subpattern
2325    number is relative rather than absolute. The most recently opened parentheses
2326    can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
2327    loops it can also make sense to refer to subsequent groups. The next
2328    parentheses to be opened can be referenced as (?(+1), and so on. (The value
2329    zero in any of these forms is not used; it provokes a compile-time error.)
2330    .P
2331    Consider the following pattern, which contains non-significant white space to
2332    make it more readable (assume the PCRE_EXTENDED option) and to divide it into
2333    three parts for ease of discussion:
2334    .sp
2335      ( \e( )?    [^()]+    (?(1) \e) )
2336    .sp
2337  The first part matches an optional opening parenthesis, and if that  The first part matches an optional opening parenthesis, and if that
2338  character is present, sets it as the first captured substring. The second part  character is present, sets it as the first captured substring. The second part
2339  matches one or more characters that are not parentheses. The third part is a  matches one or more characters that are not parentheses. The third part is a
2340  conditional subpattern that tests whether the first set of parentheses matched  conditional subpattern that tests whether or not the first set of parentheses
2341  or not. If they did, that is, if subject started with an opening parenthesis,  matched. If they did, that is, if subject started with an opening parenthesis,
2342  the condition is true, and so the yes-pattern is executed and a closing  the condition is true, and so the yes-pattern is executed and a closing
2343  parenthesis is required. Otherwise, since no-pattern is not present, the  parenthesis is required. Otherwise, since no-pattern is not present, the
2344  subpattern matches nothing. In other words, this pattern matches a sequence of  subpattern matches nothing. In other words, this pattern matches a sequence of
2345  non-parentheses, optionally enclosed in parentheses.  non-parentheses, optionally enclosed in parentheses.
2346    .P
2347  If the condition is the string (R), it is satisfied if a recursive call to the  If you were embedding this pattern in a larger one, you could use a relative
2348  pattern or subpattern has been made. At "top level", the condition is false.  reference:
2349  This is a PCRE extension. Recursive patterns are described in the next section.  .sp
2350      ...other stuff... ( \e( )?    [^()]+    (?(-1) \e) ) ...
2351  If the condition is not a sequence of digits or (R), it must be an assertion.  .sp
2352    This makes the fragment independent of the parentheses in the larger pattern.
2353    .
2354    .SS "Checking for a used subpattern by name"
2355    .rs
2356    .sp
2357    Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used
2358    subpattern by name. For compatibility with earlier versions of PCRE, which had
2359    this facility before Perl, the syntax (?(name)...) is also recognized.
2360    .P
2361    Rewriting the above example to use a named subpattern gives this:
2362    .sp
2363      (?<OPEN> \e( )?    [^()]+    (?(<OPEN>) \e) )
2364    .sp
2365    If the name used in a condition of this kind is a duplicate, the test is
2366    applied to all subpatterns of the same name, and is true if any one of them has
2367    matched.
2368    .
2369    .SS "Checking for pattern recursion"
2370    .rs
2371    .sp
2372    If the condition is the string (R), and there is no subpattern with the name R,
2373    the condition is true if a recursive call to the whole pattern or any
2374    subpattern has been made. If digits or a name preceded by ampersand follow the
2375    letter R, for example:
2376    .sp
2377      (?(R3)...) or (?(R&name)...)
2378    .sp
2379    the condition is true if the most recent recursion is into a subpattern whose
2380    number or name is given. This condition does not check the entire recursion
2381    stack. If the name used in a condition of this kind is a duplicate, the test is
2382    applied to all subpatterns of the same name, and is true if any one of them is
2383    the most recent recursion.
2384    .P
2385    At "top level", all these recursion test conditions are false.
2386    .\" HTML <a href="#recursion">
2387    .\" </a>
2388    The syntax for recursive patterns
2389    .\"
2390    is described below.
2391    .
2392    .\" HTML <a name="subdefine"></a>
2393    .SS "Defining subpatterns for use by reference only"
2394    .rs
2395    .sp
2396    If the condition is the string (DEFINE), and there is no subpattern with the
2397    name DEFINE, the condition is always false. In this case, there may be only one
2398    alternative in the subpattern. It is always skipped if control reaches this
2399    point in the pattern; the idea of DEFINE is that it can be used to define
2400    subroutines that can be referenced from elsewhere. (The use of
2401    .\" HTML <a href="#subpatternsassubroutines">
2402    .\" </a>
2403    subroutines
2404    .\"
2405    is described below.) For example, a pattern to match an IPv4 address such as
2406    "192.168.23.245" could be written like this (ignore white space and line
2407    breaks):
2408    .sp
2409      (?(DEFINE) (?<byte> 2[0-4]\ed | 25[0-5] | 1\ed\ed | [1-9]?\ed) )
2410      \eb (?&byte) (\e.(?&byte)){3} \eb
2411    .sp
2412    The first part of the pattern is a DEFINE group inside which a another group
2413    named "byte" is defined. This matches an individual component of an IPv4
2414    address (a number less than 256). When matching takes place, this part of the
2415    pattern is skipped because DEFINE acts like a false condition. The rest of the
2416    pattern uses references to the named group to match the four dot-separated
2417    components of an IPv4 address, insisting on a word boundary at each end.
2418    .
2419    .SS "Assertion conditions"
2420    .rs
2421    .sp
2422    If the condition is not in any of the above formats, it must be an assertion.
2423  This may be a positive or negative lookahead or lookbehind assertion. Consider  This may be a positive or negative lookahead or lookbehind assertion. Consider
2424  this pattern, again containing non-significant white space, and with the two  this pattern, again containing non-significant white space, and with the two
2425  alternatives on the second line:  alternatives on the second line:
2426    .sp
2427    (?(?=[^a-z]*[a-z])    (?(?=[^a-z]*[a-z])
2428    \\d{2}-[a-z]{3}-\\d{2}  |  \\d{2}-\\d{2}-\\d{2} )    \ed{2}-[a-z]{3}-\ed{2}  |  \ed{2}-\ed{2}-\ed{2} )
2429    .sp
2430  The condition is a positive lookahead assertion that matches an optional  The condition is a positive lookahead assertion that matches an optional
2431  sequence of non-letters followed by a letter. In other words, it tests for the  sequence of non-letters followed by a letter. In other words, it tests for the
2432  presence of at least one letter in the subject. If a letter is found, the  presence of at least one letter in the subject. If a letter is found, the
2433  subject is matched against the first alternative; otherwise it is matched  subject is matched against the first alternative; otherwise it is matched
2434  against the second. This pattern matches strings in one of the two forms  against the second. This pattern matches strings in one of the two forms
2435  dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.  dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.
2436    .
2437    .
2438    .\" HTML <a name="comments"></a>
2439  .SH COMMENTS  .SH COMMENTS
2440  .rs  .rs
2441  .sp  .sp
2442  The sequence (?# marks the start of a comment which continues up to the next  There are two ways of including comments in patterns that are processed by
2443  closing parenthesis. Nested parentheses are not permitted. The characters  PCRE. In both cases, the start of the comment must not be in a character class,
2444  that make up a comment play no part in the pattern matching at all.  nor in the middle of any other sequence of related characters such as (?: or a
2445    subpattern name or number. The characters that make up a comment play no part
2446  If the PCRE_EXTENDED option is set, an unescaped # character outside a  in the pattern matching.
2447  character class introduces a comment that continues up to the next newline  .P
2448  character in the pattern.  The sequence (?# marks the start of a comment that continues up to the next
2449    closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED
2450  .SH RECURSIVE PATTERNS  option is set, an unescaped # character also introduces a comment, which in
2451    this case continues to immediately after the next newline character or
2452    character sequence in the pattern. Which characters are interpreted as newlines
2453    is controlled by the options passed to a compiling function or by a special
2454    sequence at the start of the pattern, as described in the section entitled
2455    .\" HTML <a href="#newlines">
2456    .\" </a>
2457    "Newline conventions"
2458    .\"
2459    above. Note that the end of this type of comment is a literal newline sequence
2460    in the pattern; escape sequences that happen to represent a newline do not
2461    count. For example, consider this pattern when PCRE_EXTENDED is set, and the
2462    default newline convention is in force:
2463    .sp
2464      abc #comment \en still comment
2465    .sp
2466    On encountering the # character, \fBpcre_compile()\fP skips along, looking for
2467    a newline in the pattern. The sequence \en is still literal at this stage, so
2468    it does not terminate the comment. Only an actual character with the code value
2469    0x0a (the default newline) does so.
2470    .
2471    .
2472    .\" HTML <a name="recursion"></a>
2473    .SH "RECURSIVE PATTERNS"
2474  .rs  .rs
2475  .sp  .sp
2476  Consider the problem of matching a string in parentheses, allowing for  Consider the problem of matching a string in parentheses, allowing for
2477  unlimited nested parentheses. Without the use of recursion, the best that can  unlimited nested parentheses. Without the use of recursion, the best that can
2478  be done is to use a pattern that matches up to some fixed depth of nesting. It  be done is to use a pattern that matches up to some fixed depth of nesting. It
2479  is not possible to handle an arbitrary nesting depth. Perl has provided an  is not possible to handle an arbitrary nesting depth.
2480  experimental facility that allows regular expressions to recurse (amongst other  .P
2481  things). It does this by interpolating Perl code in the expression at run time,  For some time, Perl has provided a facility that allows regular expressions to
2482  and the code can refer to the expression itself. A Perl pattern to solve the  recurse (amongst other things). It does this by interpolating Perl code in the
2483  parentheses problem can be created like this:  expression at run time, and the code can refer to the expression itself. A Perl
2484    pattern using code interpolation to solve the parentheses problem can be
2485    $re = qr{\\( (?: (?>[^()]+) | (?p{$re}) )* \\)}x;  created like this:
2486    .sp
2487      $re = qr{\e( (?: (?>[^()]+) | (?p{$re}) )* \e)}x;
2488    .sp
2489  The (?p{...}) item interpolates Perl code at run time, and in this case refers  The (?p{...}) item interpolates Perl code at run time, and in this case refers
2490  recursively to the pattern in which it appears. Obviously, PCRE cannot support  recursively to the pattern in which it appears.
2491  the interpolation of Perl code. Instead, it supports some special syntax for  .P
2492  recursion of the entire pattern, and also for individual subpattern recursion.  Obviously, PCRE cannot support the interpolation of Perl code. Instead, it
2493    supports special syntax for recursion of the entire pattern, and also for
2494  The special item that consists of (? followed by a number greater than zero and  individual subpattern recursion. After its introduction in PCRE and Python,
2495  a closing parenthesis is a recursive call of the subpattern of the given  this kind of recursion was subsequently introduced into Perl at release 5.10.
2496  number, provided that it occurs inside that subpattern. (If not, it is a  .P
2497  "subroutine" call, which is described in the next section.) The special item  A special item that consists of (? followed by a number greater than zero and a
2498  (?R) is a recursive call of the entire regular expression.  closing parenthesis is a recursive subroutine call of the subpattern of the
2499    given number, provided that it occurs inside that subpattern. (If not, it is a
2500  For example, this PCRE pattern solves the nested parentheses problem (assume  .\" HTML <a href="#subpatternsassubroutines">
2501  the PCRE_EXTENDED option is set so that white space is ignored):  .\" </a>
2502    non-recursive subroutine
2503    \\( ( (?>[^()]+) | (?R) )* \\)  .\"
2504    call, which is described in the next section.) The special item (?R) or (?0) is
2505    a recursive call of the entire regular expression.
2506    .P
2507    This PCRE pattern solves the nested parentheses problem (assume the
2508    PCRE_EXTENDED option is set so that white space is ignored):
2509    .sp
2510      \e( ( [^()]++ | (?R) )* \e)
2511    .sp
2512  First it matches an opening parenthesis. Then it matches any number of  First it matches an opening parenthesis. Then it matches any number of
2513  substrings which can either be a sequence of non-parentheses, or a recursive  substrings which can either be a sequence of non-parentheses, or a recursive
2514  match of the pattern itself (that is a correctly parenthesized substring).  match of the pattern itself (that is, a correctly parenthesized substring).
2515  Finally there is a closing parenthesis.  Finally there is a closing parenthesis. Note the use of a possessive quantifier
2516    to avoid backtracking into sequences of non-parentheses.
2517    .P
2518  If this were part of a larger pattern, you would not want to recurse the entire  If this were part of a larger pattern, you would not want to recurse the entire
2519  pattern, so instead you could use this:  pattern, so instead you could use this:
2520    .sp
2521    ( \\( ( (?>[^()]+) | (?1) )* \\) )    ( \e( ( [^()]++ | (?1) )* \e) )
2522    .sp
2523  We have put the pattern into parentheses, and caused the recursion to refer to  We have put the pattern into parentheses, and caused the recursion to refer to
2524  them instead of the whole pattern. In a larger pattern, keeping track of  them instead of the whole pattern.
2525  parenthesis numbers can be tricky. It may be more convenient to use named  .P
2526  parentheses instead. For this, PCRE uses (?P>name), which is an extension to  In a larger pattern, keeping track of parenthesis numbers can be tricky. This
2527  the Python syntax that PCRE uses for named parentheses (Perl does not provide  is made easier by the use of relative references. Instead of (?1) in the
2528  named parentheses). We could rewrite the above example as follows:  pattern above you can write (?-2) to refer to the second most recently opened
2529    parentheses preceding the recursion. In other words, a negative number counts
2530    (?<pn> \\( ( (?>[^()]+) | (?P>pn) )* \\) )  capturing parentheses leftwards from the point at which it is encountered.
2531    .P
2532  This particular example pattern contains nested unlimited repeats, and so the  It is also possible to refer to subsequently opened parentheses, by writing
2533  use of atomic grouping for matching strings of non-parentheses is important  references such as (?+2). However, these cannot be recursive because the
2534  when applying the pattern to strings that do not match. For example, when this  reference is not inside the parentheses that are referenced. They are always
2535  pattern is applied to  .\" HTML <a href="#subpatternsassubroutines">
2536    .\" </a>
2537    non-recursive subroutine
2538    .\"
2539    calls, as described in the next section.
2540    .P
2541    An alternative approach is to use named parentheses instead. The Perl syntax
2542    for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We
2543    could rewrite the above example as follows:
2544    .sp
2545      (?<pn> \e( ( [^()]++ | (?&pn) )* \e) )
2546    .sp
2547    If there is more than one subpattern with the same name, the earliest one is
2548    used.
2549    .P
2550    This particular example pattern that we have been looking at contains nested
2551    unlimited repeats, and so the use of a possessive quantifier for matching
2552    strings of non-parentheses is important when applying the pattern to strings
2553    that do not match. For example, when this pattern is applied to
2554    .sp
2555    (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()    (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
2556    .sp
2557  it yields "no match" quickly. However, if atomic grouping is not used,  it yields "no match" quickly. However, if a possessive quantifier is not used,
2558  the match runs for a very long time indeed because there are so many different  the match runs for a very long time indeed because there are so many different
2559  ways the + and * repeats can carve up the subject, and all have to be tested  ways the + and * repeats can carve up the subject, and all have to be tested
2560  before failure can be reported.  before failure can be reported.
2561    .P
2562  At the end of a match, the values set for any capturing subpatterns are those  At the end of a match, the values of capturing parentheses are those from
2563  from the outermost level of the recursion at which the subpattern value is set.  the outermost level. If you want to obtain intermediate values, a callout
2564  If you want to obtain intermediate values, a callout function can be used (see  function can be used (see below and the
 below and the  
2565  .\" HREF  .\" HREF
2566  \fBpcrecallout\fR  \fBpcrecallout\fP
2567  .\"  .\"
2568  documentation). If the pattern above is matched against  documentation). If the pattern above is matched against
2569    .sp
2570    (ab(cd)ef)    (ab(cd)ef)
2571    .sp
2572  the value for the capturing parentheses is "ef", which is the last value taken  the value for the inner capturing parentheses (numbered 2) is "ef", which is
2573  on at the top level. If additional parentheses are added, giving  the last value taken on at the top level. If a capturing subpattern is not
2574    matched at the top level, its final captured value is unset, even if it was
2575    \\( ( ( (?>[^()]+) | (?R) )* ) \\)  (temporarily) set at a deeper level during the matching process.
2576       ^                        ^  .P
2577       ^                        ^  If there are more than 15 capturing parentheses in a pattern, PCRE has to
2578    obtain extra memory to store data during a recursion, which it does by using
2579  the string they capture is "ab(cd)ef", the contents of the top level  \fBpcre_malloc\fP, freeing it via \fBpcre_free\fP afterwards. If no memory can
2580  parentheses. If there are more than 15 capturing parentheses in a pattern, PCRE  be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
2581  has to obtain extra memory to store data during a recursion, which it does by  .P
 using \fBpcre_malloc\fR, freeing it via \fBpcre_free\fR afterwards. If no  
 memory can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.  
   
2582  Do not confuse the (?R) item with the condition (R), which tests for recursion.  Do not confuse the (?R) item with the condition (R), which tests for recursion.
2583  Consider this pattern, which matches text in angle brackets, allowing for  Consider this pattern, which matches text in angle brackets, allowing for
2584  arbitrary nesting. Only digits are allowed in nested brackets (that is, when  arbitrary nesting. Only digits are allowed in nested brackets (that is, when
2585  recursing), whereas any characters are permitted at the outer level.  recursing), whereas any characters are permitted at the outer level.
2586    .sp
2587    < (?: (?(R) \\d++  | [^<>]*+) | (?R)) * >    < (?: (?(R) \ed++  | [^<>]*+) | (?R)) * >
2588    .sp
2589  In this pattern, (?(R) is the start of a conditional subpattern, with two  In this pattern, (?(R) is the start of a conditional subpattern, with two
2590  different alternatives for the recursive and non-recursive cases. The (?R) item  different alternatives for the recursive and non-recursive cases. The (?R) item
2591  is the actual recursive call.  is the actual recursive call.
2592    .
2593    .
2594    .\" HTML <a name="recursiondifference"></a>
2595    .SS "Differences in recursion processing between PCRE and Perl"
2596    .rs
2597    .sp
2598    Recursion processing in PCRE differs from Perl in two important ways. In PCRE
2599    (like Python, but unlike Perl), a recursive subpattern call is always treated
2600    as an atomic group. That is, once it has matched some of the subject string, it
2601    is never re-entered, even if it contains untried alternatives and there is a
2602    subsequent matching failure. This can be illustrated by the following pattern,
2603    which purports to match a palindromic string that contains an odd number of
2604    characters (for example, "a", "aba", "abcba", "abcdcba"):
2605    .sp
2606      ^(.|(.)(?1)\e2)$
2607    .sp
2608    The idea is that it either matches a single character, or two identical
2609    characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE
2610    it does not if the pattern is longer than three characters. Consider the
2611    subject string "abcba":
2612    .P
2613    At the top level, the first character is matched, but as it is not at the end
2614    of the string, the first alternative fails; the second alternative is taken
2615    and the recursion kicks in. The recursive call to subpattern 1 successfully
2616    matches the next character ("b"). (Note that the beginning and end of line
2617    tests are not part of the recursion).
2618    .P
2619    Back at the top level, the next character ("c") is compared with what
2620    subpattern 2 matched, which was "a". This fails. Because the recursion is
2621    treated as an atomic group, there are now no backtracking points, and so the
2622    entire match fails. (Perl is able, at this point, to re-enter the recursion and
2623    try the second alternative.) However, if the pattern is written with the
2624    alternatives in the other order, things are different:
2625    .sp
2626      ^((.)(?1)\e2|.)$
2627    .sp
2628    This time, the recursing alternative is tried first, and continues to recurse
2629    until it runs out of characters, at which point the recursion fails. But this
2630    time we do have another alternative to try at the higher level. That is the big
2631    difference: in the previous case the remaining alternative is at a deeper
2632    recursion level, which PCRE cannot use.
2633    .P
2634    To change the pattern so that it matches all palindromic strings, not just
2635    those with an odd number of characters, it is tempting to change the pattern to
2636    this:
2637    .sp
2638      ^((.)(?1)\e2|.?)$
2639    .sp
2640    Again, this works in Perl, but not in PCRE, and for the same reason. When a
2641    deeper recursion has matched a single character, it cannot be entered again in
2642    order to match an empty string. The solution is to separate the two cases, and
2643    write out the odd and even cases as alternatives at the higher level:
2644    .sp
2645      ^(?:((.)(?1)\e2|)|((.)(?3)\e4|.))
2646    .sp
2647    If you want to match typical palindromic phrases, the pattern has to ignore all
2648    non-word characters, which can be done like this:
2649    .sp
2650      ^\eW*+(?:((.)\eW*+(?1)\eW*+\e2|)|((.)\eW*+(?3)\eW*+\e4|\eW*+.\eW*+))\eW*+$
2651    .sp
2652    If run with the PCRE_CASELESS option, this pattern matches phrases such as "A
2653    man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note
2654    the use of the possessive quantifier *+ to avoid backtracking into sequences of
2655    non-word characters. Without this, PCRE takes a great deal longer (ten times or
2656    more) to match typical phrases, and Perl takes so long that you think it has
2657    gone into a loop.
2658    .P
2659    \fBWARNING\fP: The palindrome-matching patterns above work only if the subject
2660    string does not start with a palindrome that is shorter than the entire string.
2661    For example, although "abcba" is correctly matched, if the subject is "ababa",
2662    PCRE finds the palindrome "aba" at the start, then fails at top level because
2663    the end of the string does not follow. Once again, it cannot jump back into the
2664    recursion to try other alternatives, so the entire match fails.
2665    .P
2666    The second way in which PCRE and Perl differ in their recursion processing is
2667    in the handling of captured values. In Perl, when a subpattern is called
2668    recursively or as a subpattern (see the next section), it has no access to any
2669    values that were captured outside the recursion, whereas in PCRE these values
2670    can be referenced. Consider this pattern:
2671    .sp
2672      ^(.)(\e1|a(?2))
2673    .sp
2674    In PCRE, this pattern matches "bab". The first capturing parentheses match "b",
2675    then in the second group, when the back reference \e1 fails to match "b", the
2676    second alternative matches "a" and then recurses. In the recursion, \e1 does
2677    now match "b" and so the whole match succeeds. In Perl, the pattern fails to
2678    match because inside the recursive call \e1 cannot access the externally set
2679    value.
2680    .
2681    .
2682  .\" HTML <a name="subpatternsassubroutines"></a>  .\" HTML <a name="subpatternsassubroutines"></a>
2683  .SH SUBPATTERNS AS SUBROUTINES  .SH "SUBPATTERNS AS SUBROUTINES"
2684  .rs  .rs
2685  .sp  .sp
2686  If the syntax for a recursive subpattern reference (either by number or by  If the syntax for a recursive subpattern call (either by number or by
2687  name) is used outside the parentheses to which it refers, it operates like a  name) is used outside the parentheses to which it refers, it operates like a
2688  subroutine in a programming language. An earlier example pointed out that the  subroutine in a programming language. The called subpattern may be defined
2689  pattern  before or after the reference. A numbered reference can be absolute or
2690    relative, as in these examples:
2691    (sens|respons)e and \\1ibility  .sp
2692      (...(absolute)...)...(?2)...
2693      (...(relative)...)...(?-1)...
2694      (...(?+1)...(relative)...
2695    .sp
2696    An earlier example pointed out that the pattern
2697    .sp
2698      (sens|respons)e and \e1ibility
2699    .sp
2700  matches "sense and sensibility" and "response and responsibility", but not  matches "sense and sensibility" and "response and responsibility", but not
2701  "sense and responsibility". If instead the pattern  "sense and responsibility". If instead the pattern
2702    .sp
2703    (sens|respons)e and (?1)ibility    (sens|respons)e and (?1)ibility
2704    .sp
2705  is used, it does match "sense and responsibility" as well as the other two  is used, it does match "sense and responsibility" as well as the other two
2706  strings. Such references must, however, follow the subpattern to which they  strings. Another example is given in the discussion of DEFINE above.
2707  refer.  .P
2708    All subroutine calls, whether recursive or not, are always treated as atomic
2709    groups. That is, once a subroutine has matched some of the subject string, it
2710    is never re-entered, even if it contains untried alternatives and there is a
2711    subsequent matching failure. Any capturing parentheses that are set during the
2712    subroutine call revert to their previous values afterwards.
2713    .P
2714    Processing options such as case-independence are fixed when a subpattern is
2715    defined, so if it is used as a subroutine, such options cannot be changed for
2716    different calls. For example, consider this pattern:
2717    .sp
2718      (abc)(?i:(?-1))
2719    .sp
2720    It matches "abcabc". It does not match "abcABC" because the change of
2721    processing option does not affect the called subpattern.
2722    .
2723    .
2724    .\" HTML <a name="onigurumasubroutines"></a>
2725    .SH "ONIGURUMA SUBROUTINE SYNTAX"
2726    .rs
2727    .sp
2728    For compatibility with Oniguruma, the non-Perl syntax \eg followed by a name or
2729    a number enclosed either in angle brackets or single quotes, is an alternative
2730    syntax for referencing a subpattern as a subroutine, possibly recursively. Here
2731    are two of the examples used above, rewritten using this syntax:
2732    .sp
2733      (?<pn> \e( ( (?>[^()]+) | \eg<pn> )* \e) )
2734      (sens|respons)e and \eg'1'ibility
2735    .sp
2736    PCRE supports an extension to Oniguruma: if a number is preceded by a
2737    plus or a minus sign it is taken as a relative reference. For example:
2738    .sp
2739      (abc)(?i:\eg<-1>)
2740    .sp
2741    Note that \eg{...} (Perl syntax) and \eg<...> (Oniguruma syntax) are \fInot\fP
2742    synonymous. The former is a back reference; the latter is a subroutine call.
2743    .
2744    .
2745  .SH CALLOUTS  .SH CALLOUTS
2746  .rs  .rs
2747  .sp  .sp
# Line 1201  Perl has a feature whereby using the seq Line 2749  Perl has a feature whereby using the seq
2749  code to be obeyed in the middle of matching a regular expression. This makes it  code to be obeyed in the middle of matching a regular expression. This makes it
2750  possible, amongst other things, to extract different substrings that match the  possible, amongst other things, to extract different substrings that match the
2751  same pair of parentheses when there is a repetition.  same pair of parentheses when there is a repetition.
2752    .P
2753  PCRE provides a similar feature, but of course it cannot obey arbitrary Perl  PCRE provides a similar feature, but of course it cannot obey arbitrary Perl
2754  code. The feature is called "callout". The caller of PCRE provides an external  code. The feature is called "callout". The caller of PCRE provides an external
2755  function by putting its entry point in the global variable \fIpcre_callout\fR.  function by putting its entry point in the global variable \fIpcre_callout\fP
2756    (8-bit library) or \fIpcre[16|32]_callout\fP (16-bit or 32-bit library).
2757  By default, this variable contains NULL, which disables all calling out.  By default, this variable contains NULL, which disables all calling out.
2758    .P
2759  Within a regular expression, (?C) indicates the points at which the external  Within a regular expression, (?C) indicates the points at which the external
2760  function is to be called. If you want to identify different callout points, you  function is to be called. If you want to identify different callout points, you
2761  can put a number less than 256 after the letter C. The default value is zero.  can put a number less than 256 after the letter C. The default value is zero.
2762  For example, this pattern has two callout points:  For example, this pattern has two callout points:
2763    .sp
2764    (?C1)\dabc(?C2)def    (?C1)abc(?C2)def
2765    .sp
2766  During matching, when PCRE reaches a callout point (and \fIpcre_callout\fR is  If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are
2767  set), the external function is called. It is provided with the number of the  automatically installed before each item in the pattern. They are all numbered
2768  callout, and, optionally, one item of data originally supplied by the caller of  255. If there is a conditional group in the pattern whose condition is an
2769  \fBpcre_exec()\fR. The callout function may cause matching to backtrack, or to  assertion, an additional callout is inserted just before the condition. An
2770  fail altogether. A complete description of the interface to the callout  explicit callout may also be set at this position, as in this example:
2771  function is given in the  .sp
2772      (?(?C9)(?=a)abc|def)
2773    .sp
2774    Note that this applies only to assertion conditions, not to other types of
2775    condition.
2776    .P
2777    During matching, when PCRE reaches a callout point, the external function is
2778    called. It is provided with the number of the callout, the position in the
2779    pattern, and, optionally, one item of data originally supplied by the caller of
2780    the matching function. The callout function may cause matching to proceed, to
2781    backtrack, or to fail altogether.
2782    .P
2783    By default, PCRE implements a number of optimizations at compile time and
2784    matching time, and one side-effect is that sometimes callouts are skipped. If
2785    you need all possible callouts to happen, you need to set options that disable
2786    the relevant optimizations. More details, and a complete description of the
2787    interface to the callout function, are given in the
2788    .\" HREF
2789    \fBpcrecallout\fP
2790    .\"
2791    documentation.
2792    .
2793    .
2794    .\" HTML <a name="backtrackcontrol"></a>
2795    .SH "BACKTRACKING CONTROL"
2796    .rs
2797    .sp
2798    Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which
2799    are still described in the Perl documentation as "experimental and subject to
2800    change or removal in a future version of Perl". It goes on to say: "Their usage
2801    in production code should be noted to avoid problems during upgrades." The same
2802    remarks apply to the PCRE features described in this section.
2803    .P
2804    The new verbs make use of what was previously invalid syntax: an opening
2805    parenthesis followed by an asterisk. They are generally of the form
2806    (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
2807    differently depending on whether or not a name is present. A name is any
2808    sequence of characters that does not include a closing parenthesis. The maximum
2809    length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
2810    libraries. If the name is empty, that is, if the closing parenthesis
2811    immediately follows the colon, the effect is as if the colon were not there.
2812    Any number of these verbs may occur in a pattern.
2813    .P
2814    Since these verbs are specifically related to backtracking, most of them can be
2815    used only when the pattern is to be matched using one of the traditional
2816    matching functions, because these use a backtracking algorithm. With the
2817    exception of (*FAIL), which behaves like a failing negative assertion, the
2818    backtracking control verbs cause an error if encountered by a DFA matching
2819    function.
2820    .P
2821    The behaviour of these verbs in
2822    .\" HTML <a href="#btrepeat">
2823    .\" </a>
2824    repeated groups,
2825    .\"
2826    .\" HTML <a href="#btassert">
2827    .\" </a>
2828    assertions,
2829    .\"
2830    and in
2831    .\" HTML <a href="#btsub">
2832    .\" </a>
2833    subpatterns called as subroutines
2834    .\"
2835    (whether or not recursively) is documented below.
2836    .
2837    .
2838    .\" HTML <a name="nooptimize"></a>
2839    .SS "Optimizations that affect backtracking verbs"
2840    .rs
2841    .sp
2842    PCRE contains some optimizations that are used to speed up matching by running
2843    some checks at the start of each match attempt. For example, it may know the
2844    minimum length of matching subject, or that a particular character must be
2845    present. When one of these optimizations bypasses the running of a match, any
2846    included backtracking verbs will not, of course, be processed. You can suppress
2847    the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option
2848    when calling \fBpcre_compile()\fP or \fBpcre_exec()\fP, or by starting the
2849    pattern with (*NO_START_OPT). There is more discussion of this option in the
2850    section entitled
2851    .\" HTML <a href="pcreapi.html#execoptions">
2852    .\" </a>
2853    "Option bits for \fBpcre_exec()\fP"
2854    .\"
2855    in the
2856  .\" HREF  .\" HREF
2857  \fBpcrecallout\fR  \fBpcreapi\fP
2858  .\"  .\"
2859  documentation.  documentation.
2860    .P
2861  .in 0  Experiments with Perl suggest that it too has similar optimizations, sometimes
2862  Last updated: 03 February 2003  leading to anomalous results.
2863  .br  .
2864  Copyright (c) 1997-2003 University of Cambridge.  .
2865    .SS "Verbs that act immediately"
2866    .rs
2867    .sp
2868    The following verbs act as soon as they are encountered. They may not be
2869    followed by a name.
2870    .sp
2871       (*ACCEPT)
2872    .sp
2873    This verb causes the match to end successfully, skipping the remainder of the
2874    pattern. However, when it is inside a subpattern that is called as a
2875    subroutine, only that subpattern is ended successfully. Matching then continues
2876    at the outer level. If (*ACCEPT) in triggered in a positive assertion, the
2877    assertion succeeds; in a negative assertion, the assertion fails.
2878    .P
2879    If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For
2880    example:
2881    .sp
2882      A((?:A|B(*ACCEPT)|C)D)
2883    .sp
2884    This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by
2885    the outer parentheses.
2886    .sp
2887      (*FAIL) or (*F)
2888    .sp
2889    This verb causes a matching failure, forcing backtracking to occur. It is
2890    equivalent to (?!) but easier to read. The Perl documentation notes that it is
2891    probably useful only when combined with (?{}) or (??{}). Those are, of course,
2892    Perl features that are not present in PCRE. The nearest equivalent is the
2893    callout feature, as for example in this pattern:
2894    .sp
2895      a+(?C)(*FAIL)
2896    .sp
2897    A match with the string "aaaa" always fails, but the callout is taken before
2898    each backtrack happens (in this example, 10 times).
2899    .
2900    .
2901    .SS "Recording which path was taken"
2902    .rs
2903    .sp
2904    There is one verb whose main purpose is to track how a match was arrived at,
2905    though it also has a secondary use in conjunction with advancing the match
2906    starting point (see (*SKIP) below).
2907    .sp
2908      (*MARK:NAME) or (*:NAME)
2909    .sp
2910    A name is always required with this verb. There may be as many instances of
2911    (*MARK) as you like in a pattern, and their names do not have to be unique.
2912    .P
2913    When a match succeeds, the name of the last-encountered (*MARK:NAME),
2914    (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
2915    caller as described in the section entitled
2916    .\" HTML <a href="pcreapi.html#extradata">
2917    .\" </a>
2918    "Extra data for \fBpcre_exec()\fP"
2919    .\"
2920    in the
2921    .\" HREF
2922    \fBpcreapi\fP
2923    .\"
2924    documentation. Here is an example of \fBpcretest\fP output, where the /K
2925    modifier requests the retrieval and outputting of (*MARK) data:
2926    .sp
2927        re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2928      data> XY
2929       0: XY
2930      MK: A
2931      XZ
2932       0: XZ
2933      MK: B
2934    .sp
2935    The (*MARK) name is tagged with "MK:" in this output, and in this example it
2936    indicates which of the two alternatives matched. This is a more efficient way
2937    of obtaining this information than putting each alternative in its own
2938    capturing parentheses.
2939    .P
2940    If a verb with a name is encountered in a positive assertion that is true, the
2941    name is recorded and passed back if it is the last-encountered. This does not
2942    happen for negative assertions or failing positive assertions.
2943    .P
2944    After a partial match or a failed match, the last encountered name in the
2945    entire match process is returned. For example:
2946    .sp
2947        re> /X(*MARK:A)Y|X(*MARK:B)Z/K
2948      data> XP
2949      No match, mark = B
2950    .sp
2951    Note that in this unanchored example the mark is retained from the match
2952    attempt that started at the letter "X" in the subject. Subsequent match
2953    attempts starting at "P" and then with an empty string do not get as far as the
2954    (*MARK) item, but nevertheless do not reset it.
2955    .P
2956    If you are interested in (*MARK) values after failed matches, you should
2957    probably set the PCRE_NO_START_OPTIMIZE option
2958    .\" HTML <a href="#nooptimize">
2959    .\" </a>
2960    (see above)
2961    .\"
2962    to ensure that the match is always attempted.
2963    .
2964    .
2965    .SS "Verbs that act after backtracking"
2966    .rs
2967    .sp
2968    The following verbs do nothing when they are encountered. Matching continues
2969    with what follows, but if there is no subsequent match, causing a backtrack to
2970    the verb, a failure is forced. That is, backtracking cannot pass to the left of
2971    the verb. However, when one of these verbs appears inside an atomic group or an
2972    assertion that is true, its effect is confined to that group, because once the
2973    group has been matched, there is never any backtracking into it. In this
2974    situation, backtracking can "jump back" to the left of the entire atomic group
2975    or assertion. (Remember also, as stated above, that this localization also
2976    applies in subroutine calls.)
2977    .P
2978    These verbs differ in exactly what kind of failure occurs when backtracking
2979    reaches them. The behaviour described below is what happens when the verb is
2980    not in a subroutine or an assertion. Subsequent sections cover these special
2981    cases.
2982    .sp
2983      (*COMMIT)
2984    .sp
2985    This verb, which may not be followed by a name, causes the whole match to fail
2986    outright if there is a later matching failure that causes backtracking to reach
2987    it. Even if the pattern is unanchored, no further attempts to find a match by
2988    advancing the starting point take place. If (*COMMIT) is the only backtracking
2989    verb that is encountered, once it has been passed \fBpcre_exec()\fP is
2990    committed to finding a match at the current starting point, or not at all. For
2991    example:
2992    .sp
2993      a+(*COMMIT)b
2994    .sp
2995    This matches "xxaab" but not "aacaab". It can be thought of as a kind of
2996    dynamic anchor, or "I've started, so I must finish." The name of the most
2997    recently passed (*MARK) in the path is passed back when (*COMMIT) forces a
2998    match failure.
2999    .P
3000    If there is more than one backtracking verb in a pattern, a different one that
3001    follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a
3002    match does not always guarantee that a match must be at this starting point.
3003    .P
3004    Note that (*COMMIT) at the start of a pattern is not the same as an anchor,
3005    unless PCRE's start-of-match optimizations are turned off, as shown in this
3006    \fBpcretest\fP example:
3007    .sp
3008        re> /(*COMMIT)abc/
3009      data> xyzabc
3010       0: abc
3011      xyzabc\eY
3012      No match
3013    .sp
3014    PCRE knows that any match must start with "a", so the optimization skips along
3015    the subject to "a" before running the first match attempt, which succeeds. When
3016    the optimization is disabled by the \eY escape in the second subject, the match
3017    starts at "x" and so the (*COMMIT) causes it to fail without trying any other
3018    starting points.
3019    .sp
3020      (*PRUNE) or (*PRUNE:NAME)
3021    .sp
3022    This verb causes the match to fail at the current starting position in the
3023    subject if there is a later matching failure that causes backtracking to reach
3024    it. If the pattern is unanchored, the normal "bumpalong" advance to the next
3025    starting character then happens. Backtracking can occur as usual to the left of
3026    (*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but
3027    if there is no match to the right, backtracking cannot cross (*PRUNE). In
3028    simple cases, the use of (*PRUNE) is just an alternative to an atomic group or
3029    possessive quantifier, but there are some uses of (*PRUNE) that cannot be
3030    expressed in any other way. In an anchored pattern (*PRUNE) has the same effect
3031    as (*COMMIT).
3032    .P
3033    The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE).
3034    It is like (*MARK:NAME) in that the name is remembered for passing back to the
3035    caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3036    .sp
3037      (*SKIP)
3038    .sp
3039    This verb, when given without a name, is like (*PRUNE), except that if the
3040    pattern is unanchored, the "bumpalong" advance is not to the next character,
3041    but to the position in the subject where (*SKIP) was encountered. (*SKIP)
3042    signifies that whatever text was matched leading up to it cannot be part of a
3043    successful match. Consider:
3044    .sp
3045      a+(*SKIP)b
3046    .sp
3047    If the subject is "aaaac...", after the first match attempt fails (starting at
3048    the first character in the string), the starting point skips on to start the
3049    next attempt at "c". Note that a possessive quantifer does not have the same
3050    effect as this example; although it would suppress backtracking during the
3051    first match attempt, the second attempt would start at the second character
3052    instead of skipping on to "c".
3053    .sp
3054      (*SKIP:NAME)
3055    .sp
3056    When (*SKIP) has an associated name, its behaviour is modified. When it is
3057    triggered, the previous path through the pattern is searched for the most
3058    recent (*MARK) that has the same name. If one is found, the "bumpalong" advance
3059    is to the subject position that corresponds to that (*MARK) instead of to where
3060    (*SKIP) was encountered. If no (*MARK) with a matching name is found, the
3061    (*SKIP) is ignored.
3062    .P
3063    Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores
3064    names that are set by (*PRUNE:NAME) or (*THEN:NAME).
3065    .sp
3066      (*THEN) or (*THEN:NAME)
3067    .sp
3068    This verb causes a skip to the next innermost alternative when backtracking
3069    reaches it. That is, it cancels any further backtracking within the current
3070    alternative. Its name comes from the observation that it can be used for a
3071    pattern-based if-then-else block:
3072    .sp
3073      ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
3074    .sp
3075    If the COND1 pattern matches, FOO is tried (and possibly further items after
3076    the end of the group if FOO succeeds); on failure, the matcher skips to the
3077    second alternative and tries COND2, without backtracking into COND1. If that
3078    succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no
3079    more alternatives, so there is a backtrack to whatever came before the entire
3080    group. If (*THEN) is not inside an alternation, it acts like (*PRUNE).
3081    .P
3082    The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
3083    It is like (*MARK:NAME) in that the name is remembered for passing back to the
3084    caller. However, (*SKIP:NAME) searches only for names set with (*MARK).
3085    .P
3086    A subpattern that does not contain a | character is just a part of the
3087    enclosing alternative; it is not a nested alternation with only one
3088    alternative. The effect of (*THEN) extends beyond such a subpattern to the
3089    enclosing alternative. Consider this pattern, where A, B, etc. are complex
3090    pattern fragments that do not contain any | characters at this level:
3091    .sp
3092      A (B(*THEN)C) | D
3093    .sp
3094    If A and B are matched, but there is a failure in C, matching does not
3095    backtrack into A; instead it moves to the next alternative, that is, D.
3096    However, if the subpattern containing (*THEN) is given an alternative, it
3097    behaves differently:
3098    .sp
3099      A (B(*THEN)C | (*FAIL)) | D
3100    .sp
3101    The effect of (*THEN) is now confined to the inner subpattern. After a failure
3102    in C, matching moves to (*FAIL), which causes the whole subpattern to fail
3103    because there are no more alternatives to try. In this case, matching does now
3104    backtrack into A.
3105    .P
3106    Note that a conditional subpattern is not considered as having two
3107    alternatives, because only one is ever used. In other words, the | character in
3108    a conditional subpattern has a different meaning. Ignoring white space,
3109    consider:
3110    .sp
3111      ^.*? (?(?=a) a | b(*THEN)c )
3112    .sp
3113    If the subject is "ba", this pattern does not match. Because .*? is ungreedy,
3114    it initially matches zero characters. The condition (?=a) then fails, the
3115    character "b" is matched, but "c" is not. At this point, matching does not
3116    backtrack to .*? as might perhaps be expected from the presence of the |
3117    character. The conditional subpattern is part of the single alternative that
3118    comprises the whole pattern, and so the match fails. (If there was a backtrack
3119    into .*?, allowing it to match "b", the match would succeed.)
3120    .P
3121    The verbs just described provide four different "strengths" of control when
3122    subsequent matching fails. (*THEN) is the weakest, carrying on the match at the
3123    next alternative. (*PRUNE) comes next, failing the match at the current
3124    starting position, but allowing an advance to the next character (for an
3125    unanchored pattern). (*SKIP) is similar, except that the advance may be more
3126    than one character. (*COMMIT) is the strongest, causing the entire match to
3127    fail.
3128    .
3129    .
3130    .SS "More than one backtracking verb"
3131    .rs
3132    .sp
3133    If more than one backtracking verb is present in a pattern, the one that is
3134    backtracked onto first acts. For example, consider this pattern, where A, B,
3135    etc. are complex pattern fragments:
3136    .sp
3137      (A(*COMMIT)B(*THEN)C|ABD)
3138    .sp
3139    If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to
3140    fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes
3141    the next alternative (ABD) to be tried. This behaviour is consistent, but is
3142    not always the same as Perl's. It means that if two or more backtracking verbs
3143    appear in succession, all the the last of them has no effect. Consider this
3144    example:
3145    .sp
3146      ...(*COMMIT)(*PRUNE)...
3147    .sp
3148    If there is a matching failure to the right, backtracking onto (*PRUNE) causes
3149    it to be triggered, and its action is taken. There can never be a backtrack
3150    onto (*COMMIT).
3151    .
3152    .
3153    .\" HTML <a name="btrepeat"></a>
3154    .SS "Backtracking verbs in repeated groups"
3155    .rs
3156    .sp
3157    PCRE differs from Perl in its handling of backtracking verbs in repeated
3158    groups. For example, consider:
3159    .sp
3160      /(a(*COMMIT)b)+ac/
3161    .sp
3162    If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in
3163    the second repeat of the group acts.
3164    .
3165    .
3166    .\" HTML <a name="btassert"></a>
3167    .SS "Backtracking verbs in assertions"
3168    .rs
3169    .sp
3170    (*FAIL) in an assertion has its normal effect: it forces an immediate backtrack.
3171    .P
3172    (*ACCEPT) in a positive assertion causes the assertion to succeed without any
3173    further processing. In a negative assertion, (*ACCEPT) causes the assertion to
3174    fail without any further processing.
3175    .P
3176    The other backtracking verbs are not treated specially if they appear in a
3177    positive assertion. In particular, (*THEN) skips to the next alternative in the
3178    innermost enclosing group that has alternations, whether or not this is within
3179    the assertion.
3180    .P
3181    Negative assertions are, however, different, in order to ensure that changing a
3182    positive assertion into a negative assertion changes its result. Backtracking
3183    into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true,
3184    without considering any further alternative branches in the assertion.
3185    Backtracking into (*THEN) causes it to skip to the next enclosing alternative
3186    within the assertion (the normal behaviour), but if the assertion does not have
3187    such an alternative, (*THEN) behaves like (*PRUNE).
3188    .
3189    .
3190    .\" HTML <a name="btsub"></a>
3191    .SS "Backtracking verbs in subroutines"
3192    .rs
3193    .sp
3194    These behaviours occur whether or not the subpattern is called recursively.
3195    Perl's treatment of subroutines is different in some cases.
3196    .P
3197    (*FAIL) in a subpattern called as a subroutine has its normal effect: it forces
3198    an immediate backtrack.
3199    .P
3200    (*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to
3201    succeed without any further processing. Matching then continues after the
3202    subroutine call.
3203    .P
3204    (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause
3205    the subroutine match to fail.
3206    .P
3207    (*THEN) skips to the next alternative in the innermost enclosing group within
3208    the subpattern that has alternatives. If there is no such group within the
3209    subpattern, (*THEN) causes the subroutine match to fail.
3210    .
3211