1<html> 2<head> 3<title>pcrepattern specification</title> 4</head> 5<body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB"> 6<h1>pcrepattern man page</h1> 7<p> 8Return to the <a href="index.html">PCRE index page</a>. 9</p> 10<p> 11This page is part of the PCRE HTML documentation. It was generated automatically 12from the original man page. If there is any nonsense in it, please consult the 13man page, in case the conversion went wrong. 14<br> 15<ul> 16<li><a name="TOC1" href="#SEC1">PCRE REGULAR EXPRESSION DETAILS</a> 17<li><a name="TOC2" href="#SEC2">SPECIAL START-OF-PATTERN ITEMS</a> 18<li><a name="TOC3" href="#SEC3">EBCDIC CHARACTER CODES</a> 19<li><a name="TOC4" href="#SEC4">CHARACTERS AND METACHARACTERS</a> 20<li><a name="TOC5" href="#SEC5">BACKSLASH</a> 21<li><a name="TOC6" href="#SEC6">CIRCUMFLEX AND DOLLAR</a> 22<li><a name="TOC7" href="#SEC7">FULL STOP (PERIOD, DOT) AND \N</a> 23<li><a name="TOC8" href="#SEC8">MATCHING A SINGLE DATA UNIT</a> 24<li><a name="TOC9" href="#SEC9">SQUARE BRACKETS AND CHARACTER CLASSES</a> 25<li><a name="TOC10" href="#SEC10">POSIX CHARACTER CLASSES</a> 26<li><a name="TOC11" href="#SEC11">COMPATIBILITY FEATURE FOR WORD BOUNDARIES</a> 27<li><a name="TOC12" href="#SEC12">VERTICAL BAR</a> 28<li><a name="TOC13" href="#SEC13">INTERNAL OPTION SETTING</a> 29<li><a name="TOC14" href="#SEC14">SUBPATTERNS</a> 30<li><a name="TOC15" href="#SEC15">DUPLICATE SUBPATTERN NUMBERS</a> 31<li><a name="TOC16" href="#SEC16">NAMED SUBPATTERNS</a> 32<li><a name="TOC17" href="#SEC17">REPETITION</a> 33<li><a name="TOC18" href="#SEC18">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a> 34<li><a name="TOC19" href="#SEC19">BACK REFERENCES</a> 35<li><a name="TOC20" href="#SEC20">ASSERTIONS</a> 36<li><a name="TOC21" href="#SEC21">CONDITIONAL SUBPATTERNS</a> 37<li><a name="TOC22" href="#SEC22">COMMENTS</a> 38<li><a name="TOC23" href="#SEC23">RECURSIVE PATTERNS</a> 39<li><a name="TOC24" href="#SEC24">SUBPATTERNS AS SUBROUTINES</a> 40<li><a name="TOC25" href="#SEC25">ONIGURUMA SUBROUTINE SYNTAX</a> 41<li><a name="TOC26" href="#SEC26">CALLOUTS</a> 42<li><a name="TOC27" href="#SEC27">BACKTRACKING CONTROL</a> 43<li><a name="TOC28" href="#SEC28">SEE ALSO</a> 44<li><a name="TOC29" href="#SEC29">AUTHOR</a> 45<li><a name="TOC30" href="#SEC30">REVISION</a> 46</ul> 47<br><a name="SEC1" href="#TOC1">PCRE REGULAR EXPRESSION DETAILS</a><br> 48<P> 49The syntax and semantics of the regular expressions that are supported by PCRE 50are described in detail below. There is a quick-reference syntax summary in the 51<a href="pcresyntax.html"><b>pcresyntax</b></a> 52page. PCRE tries to match Perl syntax and semantics as closely as it can. PCRE 53also supports some alternative regular expression syntax (which does not 54conflict with the Perl syntax) in order to provide some compatibility with 55regular expressions in Python, .NET, and Oniguruma. 56</P> 57<P> 58Perl's regular expressions are described in its own documentation, and 59regular expressions in general are covered in a number of books, some of which 60have copious examples. Jeffrey Friedl's "Mastering Regular Expressions", 61published by O'Reilly, covers regular expressions in great detail. This 62description of PCRE's regular expressions is intended as reference material. 63</P> 64<P> 65This document discusses the patterns that are supported by PCRE when one its 66main matching functions, <b>pcre_exec()</b> (8-bit) or <b>pcre[16|32]_exec()</b> 67(16- or 32-bit), is used. PCRE also has alternative matching functions, 68<b>pcre_dfa_exec()</b> and <b>pcre[16|32_dfa_exec()</b>, which match using a 69different algorithm that is not Perl-compatible. Some of the features discussed 70below are not available when DFA matching is used. The advantages and 71disadvantages of the alternative functions, and how they differ from the normal 72functions, are discussed in the 73<a href="pcrematching.html"><b>pcrematching</b></a> 74page. 75</P> 76<br><a name="SEC2" href="#TOC1">SPECIAL START-OF-PATTERN ITEMS</a><br> 77<P> 78A number of options that can be passed to <b>pcre_compile()</b> can also be set 79by special items at the start of a pattern. These are not Perl-compatible, but 80are provided to make these options accessible to pattern writers who are not 81able to change the program that processes the pattern. Any number of these 82items may appear, but they must all be together right at the start of the 83pattern string, and the letters must be in upper case. 84</P> 85<br><b> 86UTF support 87</b><br> 88<P> 89The original operation of PCRE was on strings of one-byte characters. However, 90there is now also support for UTF-8 strings in the original library, an 91extra library that supports 16-bit and UTF-16 character strings, and a 92third library that supports 32-bit and UTF-32 character strings. To use these 93features, PCRE must be built to include appropriate support. When using UTF 94strings you must either call the compiling function with the PCRE_UTF8, 95PCRE_UTF16, or PCRE_UTF32 option, or the pattern must start with one of 96these special sequences: 97<pre> 98 (*UTF8) 99 (*UTF16) 100 (*UTF32) 101 (*UTF) 102</pre> 103(*UTF) is a generic sequence that can be used with any of the libraries. 104Starting a pattern with such a sequence is equivalent to setting the relevant 105option. How setting a UTF mode affects pattern matching is mentioned in several 106places below. There is also a summary of features in the 107<a href="pcreunicode.html"><b>pcreunicode</b></a> 108page. 109</P> 110<P> 111Some applications that allow their users to supply patterns may wish to 112restrict them to non-UTF data for security reasons. If the PCRE_NEVER_UTF 113option is set at compile time, (*UTF) etc. are not allowed, and their 114appearance causes an error. 115</P> 116<br><b> 117Unicode property support 118</b><br> 119<P> 120Another special sequence that may appear at the start of a pattern is (*UCP). 121This has the same effect as setting the PCRE_UCP option: it causes sequences 122such as \d and \w to use Unicode properties to determine character types, 123instead of recognizing only characters with codes less than 128 via a lookup 124table. 125</P> 126<br><b> 127Disabling auto-possessification 128</b><br> 129<P> 130If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting 131the PCRE_NO_AUTO_POSSESS option at compile time. This stops PCRE from making 132quantifiers possessive when what follows cannot match the repeated item. For 133example, by default a+b is treated as a++b. For more details, see the 134<a href="pcreapi.html"><b>pcreapi</b></a> 135documentation. 136</P> 137<br><b> 138Disabling start-up optimizations 139</b><br> 140<P> 141If a pattern starts with (*NO_START_OPT), it has the same effect as setting the 142PCRE_NO_START_OPTIMIZE option either at compile or matching time. This disables 143several optimizations for quickly reaching "no match" results. For more 144details, see the 145<a href="pcreapi.html"><b>pcreapi</b></a> 146documentation. 147<a name="newlines"></a></P> 148<br><b> 149Newline conventions 150</b><br> 151<P> 152PCRE supports five different conventions for indicating line breaks in 153strings: a single CR (carriage return) character, a single LF (linefeed) 154character, the two-character sequence CRLF, any of the three preceding, or any 155Unicode newline sequence. The 156<a href="pcreapi.html"><b>pcreapi</b></a> 157page has 158<a href="pcreapi.html#newlines">further discussion</a> 159about newlines, and shows how to set the newline convention in the 160<i>options</i> arguments for the compiling and matching functions. 161</P> 162<P> 163It is also possible to specify a newline convention by starting a pattern 164string with one of the following five sequences: 165<pre> 166 (*CR) carriage return 167 (*LF) linefeed 168 (*CRLF) carriage return, followed by linefeed 169 (*ANYCRLF) any of the three above 170 (*ANY) all Unicode newline sequences 171</pre> 172These override the default and the options given to the compiling function. For 173example, on a Unix system where LF is the default newline sequence, the pattern 174<pre> 175 (*CR)a.b 176</pre> 177changes the convention to CR. That pattern matches "a\nb" because LF is no 178longer a newline. If more than one of these settings is present, the last one 179is used. 180</P> 181<P> 182The newline convention affects where the circumflex and dollar assertions are 183true. It also affects the interpretation of the dot metacharacter when 184PCRE_DOTALL is not set, and the behaviour of \N. However, it does not affect 185what the \R escape sequence matches. By default, this is any Unicode newline 186sequence, for Perl compatibility. However, this can be changed; see the 187description of \R in the section entitled 188<a href="#newlineseq">"Newline sequences"</a> 189below. A change of \R setting can be combined with a change of newline 190convention. 191</P> 192<br><b> 193Setting match and recursion limits 194</b><br> 195<P> 196The caller of <b>pcre_exec()</b> can set a limit on the number of times the 197internal <b>match()</b> function is called and on the maximum depth of 198recursive calls. These facilities are provided to catch runaway matches that 199are provoked by patterns with huge matching trees (a typical example is a 200pattern with nested unlimited repeats) and to avoid running out of system stack 201by too much recursion. When one of these limits is reached, <b>pcre_exec()</b> 202gives an error return. The limits can also be set by items at the start of the 203pattern of the form 204<pre> 205 (*LIMIT_MATCH=d) 206 (*LIMIT_RECURSION=d) 207</pre> 208where d is any number of decimal digits. However, the value of the setting must 209be less than the value set (or defaulted) by the caller of <b>pcre_exec()</b> 210for it to have any effect. In other words, the pattern writer can lower the 211limits set by the programmer, but not raise them. If there is more than one 212setting of one of these limits, the lower value is used. 213</P> 214<br><a name="SEC3" href="#TOC1">EBCDIC CHARACTER CODES</a><br> 215<P> 216PCRE can be compiled to run in an environment that uses EBCDIC as its character 217code rather than ASCII or Unicode (typically a mainframe system). In the 218sections below, character code values are ASCII or Unicode; in an EBCDIC 219environment these characters may have different code values, and there are no 220code points greater than 255. 221</P> 222<br><a name="SEC4" href="#TOC1">CHARACTERS AND METACHARACTERS</a><br> 223<P> 224A regular expression is a pattern that is matched against a subject string from 225left to right. Most characters stand for themselves in a pattern, and match the 226corresponding characters in the subject. As a trivial example, the pattern 227<pre> 228 The quick brown fox 229</pre> 230matches a portion of a subject string that is identical to itself. When 231caseless matching is specified (the PCRE_CASELESS option), letters are matched 232independently of case. In a UTF mode, PCRE always understands the concept of 233case for characters whose values are less than 128, so caseless matching is 234always possible. For characters with higher values, the concept of case is 235supported if PCRE is compiled with Unicode property support, but not otherwise. 236If you want to use caseless matching for characters 128 and above, you must 237ensure that PCRE is compiled with Unicode property support as well as with 238UTF support. 239</P> 240<P> 241The power of regular expressions comes from the ability to include alternatives 242and repetitions in the pattern. These are encoded in the pattern by the use of 243<i>metacharacters</i>, which do not stand for themselves but instead are 244interpreted in some special way. 245</P> 246<P> 247There are two different sets of metacharacters: those that are recognized 248anywhere in the pattern except within square brackets, and those that are 249recognized within square brackets. Outside square brackets, the metacharacters 250are as follows: 251<pre> 252 \ general escape character with several uses 253 ^ assert start of string (or line, in multiline mode) 254 $ assert end of string (or line, in multiline mode) 255 . match any character except newline (by default) 256 [ start character class definition 257 | start of alternative branch 258 ( start subpattern 259 ) end subpattern 260 ? extends the meaning of ( 261 also 0 or 1 quantifier 262 also quantifier minimizer 263 * 0 or more quantifier 264 + 1 or more quantifier 265 also "possessive quantifier" 266 { start min/max quantifier 267</pre> 268Part of a pattern that is in square brackets is called a "character class". In 269a character class the only metacharacters are: 270<pre> 271 \ general escape character 272 ^ negate the class, but only if the first character 273 - indicates character range 274 [ POSIX character class (only if followed by POSIX syntax) 275 ] terminates the character class 276</pre> 277The following sections describe the use of each of the metacharacters. 278</P> 279<br><a name="SEC5" href="#TOC1">BACKSLASH</a><br> 280<P> 281The backslash character has several uses. Firstly, if it is followed by a 282character that is not a number or a letter, it takes away any special meaning 283that character may have. This use of backslash as an escape character applies 284both inside and outside character classes. 285</P> 286<P> 287For example, if you want to match a * character, you write \* in the pattern. 288This escaping action applies whether or not the following character would 289otherwise be interpreted as a metacharacter, so it is always safe to precede a 290non-alphanumeric with backslash to specify that it stands for itself. In 291particular, if you want to match a backslash, you write \\. 292</P> 293<P> 294In a UTF mode, only ASCII numbers and letters have any special meaning after a 295backslash. All other characters (in particular, those whose codepoints are 296greater than 127) are treated as literals. 297</P> 298<P> 299If a pattern is compiled with the PCRE_EXTENDED option, most white space in the 300pattern (other than in a character class), and characters between a # outside a 301character class and the next newline, inclusive, are ignored. An escaping 302backslash can be used to include a white space or # character as part of the 303pattern. 304</P> 305<P> 306If you want to remove the special meaning from a sequence of characters, you 307can do so by putting them between \Q and \E. This is different from Perl in 308that $ and @ are handled as literals in \Q...\E sequences in PCRE, whereas in 309Perl, $ and @ cause variable interpolation. Note the following examples: 310<pre> 311 Pattern PCRE matches Perl matches 312 313 \Qabc$xyz\E abc$xyz abc followed by the contents of $xyz 314 \Qabc\$xyz\E abc\$xyz abc\$xyz 315 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz 316</pre> 317The \Q...\E sequence is recognized both inside and outside character classes. 318An isolated \E that is not preceded by \Q is ignored. If \Q is not followed 319by \E later in the pattern, the literal interpretation continues to the end of 320the pattern (that is, \E is assumed at the end). If the isolated \Q is inside 321a character class, this causes an error, because the character class is not 322terminated. 323<a name="digitsafterbackslash"></a></P> 324<br><b> 325Non-printing characters 326</b><br> 327<P> 328A second use of backslash provides a way of encoding non-printing characters 329in patterns in a visible manner. There is no restriction on the appearance of 330non-printing characters, apart from the binary zero that terminates a pattern, 331but when a pattern is being prepared by text editing, it is often easier to use 332one of the following escape sequences than the binary character it represents: 333<pre> 334 \a alarm, that is, the BEL character (hex 07) 335 \cx "control-x", where x is any ASCII character 336 \e escape (hex 1B) 337 \f form feed (hex 0C) 338 \n linefeed (hex 0A) 339 \r carriage return (hex 0D) 340 \t tab (hex 09) 341 \0dd character with octal code 0dd 342 \ddd character with octal code ddd, or back reference 343 \o{ddd..} character with octal code ddd.. 344 \xhh character with hex code hh 345 \x{hhh..} character with hex code hhh.. (non-JavaScript mode) 346 \uhhhh character with hex code hhhh (JavaScript mode only) 347</pre> 348The precise effect of \cx on ASCII characters is as follows: if x is a lower 349case letter, it is converted to upper case. Then bit 6 of the character (hex 35040) is inverted. Thus \cA to \cZ become hex 01 to hex 1A (A is 41, Z is 5A), 351but \c{ becomes hex 3B ({ is 7B), and \c; becomes hex 7B (; is 3B). If the 352data item (byte or 16-bit value) following \c has a value greater than 127, a 353compile-time error occurs. This locks out non-ASCII characters in all modes. 354</P> 355<P> 356The \c facility was designed for use with ASCII characters, but with the 357extension to Unicode it is even less useful than it once was. It is, however, 358recognized when PCRE is compiled in EBCDIC mode, where data items are always 359bytes. In this mode, all values are valid after \c. If the next character is a 360lower case letter, it is converted to upper case. Then the 0xc0 bits of the 361byte are inverted. Thus \cA becomes hex 01, as in ASCII (A is C1), but because 362the EBCDIC letters are disjoint, \cZ becomes hex 29 (Z is E9), and other 363characters also generate different values. 364</P> 365<P> 366After \0 up to two further octal digits are read. If there are fewer than two 367digits, just those that are present are used. Thus the sequence \0\x\07 368specifies two binary zeros followed by a BEL character (code value 7). Make 369sure you supply two digits after the initial zero if the pattern character that 370follows is itself an octal digit. 371</P> 372<P> 373The escape \o must be followed by a sequence of octal digits, enclosed in 374braces. An error occurs if this is not the case. This escape is a recent 375addition to Perl; it provides way of specifying character code points as octal 376numbers greater than 0777, and it also allows octal numbers and back references 377to be unambiguously specified. 378</P> 379<P> 380For greater clarity and unambiguity, it is best to avoid following \ by a 381digit greater than zero. Instead, use \o{} or \x{} to specify character 382numbers, and \g{} to specify back references. The following paragraphs 383describe the old, ambiguous syntax. 384</P> 385<P> 386The handling of a backslash followed by a digit other than 0 is complicated, 387and Perl has changed in recent releases, causing PCRE also to change. Outside a 388character class, PCRE reads the digit and any following digits as a decimal 389number. If the number is less than 8, or if there have been at least that many 390previous capturing left parentheses in the expression, the entire sequence is 391taken as a <i>back reference</i>. A description of how this works is given 392<a href="#backreferences">later,</a> 393following the discussion of 394<a href="#subpattern">parenthesized subpatterns.</a> 395</P> 396<P> 397Inside a character class, or if the decimal number following \ is greater than 3987 and there have not been that many capturing subpatterns, PCRE handles \8 and 399\9 as the literal characters "8" and "9", and otherwise re-reads up to three 400octal digits following the backslash, using them to generate a data character. 401Any subsequent digits stand for themselves. For example: 402<pre> 403 \040 is another way of writing an ASCII space 404 \40 is the same, provided there are fewer than 40 previous capturing subpatterns 405 \7 is always a back reference 406 \11 might be a back reference, or another way of writing a tab 407 \011 is always a tab 408 \0113 is a tab followed by the character "3" 409 \113 might be a back reference, otherwise the character with octal code 113 410 \377 might be a back reference, otherwise the value 255 (decimal) 411 \81 is either a back reference, or the two characters "8" and "1" 412</pre> 413Note that octal values of 100 or greater that are specified using this syntax 414must not be introduced by a leading zero, because no more than three octal 415digits are ever read. 416</P> 417<P> 418By default, after \x that is not followed by {, from zero to two hexadecimal 419digits are read (letters can be in upper or lower case). Any number of 420hexadecimal digits may appear between \x{ and }. If a character other than 421a hexadecimal digit appears between \x{ and }, or if there is no terminating 422}, an error occurs. 423</P> 424<P> 425If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x is 426as just described only when it is followed by two hexadecimal digits. 427Otherwise, it matches a literal "x" character. In JavaScript mode, support for 428code points greater than 256 is provided by \u, which must be followed by 429four hexadecimal digits; otherwise it matches a literal "u" character. 430</P> 431<P> 432Characters whose value is less than 256 can be defined by either of the two 433syntaxes for \x (or by \u in JavaScript mode). There is no difference in the 434way they are handled. For example, \xdc is exactly the same as \x{dc} (or 435\u00dc in JavaScript mode). 436</P> 437<br><b> 438Constraints on character values 439</b><br> 440<P> 441Characters that are specified using octal or hexadecimal numbers are 442limited to certain values, as follows: 443<pre> 444 8-bit non-UTF mode less than 0x100 445 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint 446 16-bit non-UTF mode less than 0x10000 447 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint 448 32-bit non-UTF mode less than 0x100000000 449 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint 450</pre> 451Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-called 452"surrogate" codepoints), and 0xffef. 453</P> 454<br><b> 455Escape sequences in character classes 456</b><br> 457<P> 458All the sequences that define a single character value can be used both inside 459and outside character classes. In addition, inside a character class, \b is 460interpreted as the backspace character (hex 08). 461</P> 462<P> 463\N is not allowed in a character class. \B, \R, and \X are not special 464inside a character class. Like other unrecognized escape sequences, they are 465treated as the literal characters "B", "R", and "X" by default, but cause an 466error if the PCRE_EXTRA option is set. Outside a character class, these 467sequences have different meanings. 468</P> 469<br><b> 470Unsupported escape sequences 471</b><br> 472<P> 473In Perl, the sequences \l, \L, \u, and \U are recognized by its string 474handler and used to modify the case of following characters. By default, PCRE 475does not support these escape sequences. However, if the PCRE_JAVASCRIPT_COMPAT 476option is set, \U matches a "U" character, and \u can be used to define a 477character by code point, as described in the previous section. 478</P> 479<br><b> 480Absolute and relative back references 481</b><br> 482<P> 483The sequence \g followed by an unsigned or a negative number, optionally 484enclosed in braces, is an absolute or relative back reference. A named back 485reference can be coded as \g{name}. Back references are discussed 486<a href="#backreferences">later,</a> 487following the discussion of 488<a href="#subpattern">parenthesized subpatterns.</a> 489</P> 490<br><b> 491Absolute and relative subroutine calls 492</b><br> 493<P> 494For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or 495a number enclosed either in angle brackets or single quotes, is an alternative 496syntax for referencing a subpattern as a "subroutine". Details are discussed 497<a href="#onigurumasubroutines">later.</a> 498Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are <i>not</i> 499synonymous. The former is a back reference; the latter is a 500<a href="#subpatternsassubroutines">subroutine</a> 501call. 502<a name="genericchartypes"></a></P> 503<br><b> 504Generic character types 505</b><br> 506<P> 507Another use of backslash is for specifying generic character types: 508<pre> 509 \d any decimal digit 510 \D any character that is not a decimal digit 511 \h any horizontal white space character 512 \H any character that is not a horizontal white space character 513 \s any white space character 514 \S any character that is not a white space character 515 \v any vertical white space character 516 \V any character that is not a vertical white space character 517 \w any "word" character 518 \W any "non-word" character 519</pre> 520There is also the single sequence \N, which matches a non-newline character. 521This is the same as 522<a href="#fullstopdot">the "." metacharacter</a> 523when PCRE_DOTALL is not set. Perl also uses \N to match characters by name; 524PCRE does not support this. 525</P> 526<P> 527Each pair of lower and upper case escape sequences partitions the complete set 528of characters into two disjoint sets. Any given character matches one, and only 529one, of each pair. The sequences can appear both inside and outside character 530classes. They each match one character of the appropriate type. If the current 531matching point is at the end of the subject string, all of them fail, because 532there is no character to match. 533</P> 534<P> 535For compatibility with Perl, \s did not used to match the VT character (code 53611), which made it different from the the POSIX "space" class. However, Perl 537added VT at release 5.18, and PCRE followed suit at release 8.34. The default 538\s characters are now HT (9), LF (10), VT (11), FF (12), CR (13), and space 539(32), which are defined as white space in the "C" locale. This list may vary if 540locale-specific matching is taking place. For example, in some locales the 541"non-breaking space" character (\xA0) is recognized as white space, and in 542others the VT character is not. 543</P> 544<P> 545A "word" character is an underscore or any character that is a letter or digit. 546By default, the definition of letters and digits is controlled by PCRE's 547low-valued character tables, and may vary if locale-specific matching is taking 548place (see 549<a href="pcreapi.html#localesupport">"Locale support"</a> 550in the 551<a href="pcreapi.html"><b>pcreapi</b></a> 552page). For example, in a French locale such as "fr_FR" in Unix-like systems, 553or "french" in Windows, some character codes greater than 127 are used for 554accented letters, and these are then matched by \w. The use of locales with 555Unicode is discouraged. 556</P> 557<P> 558By default, characters whose code points are greater than 127 never match \d, 559\s, or \w, and always match \D, \S, and \W, although this may vary for 560characters in the range 128-255 when locale-specific matching is happening. 561These escape sequences retain their original meanings from before Unicode 562support was available, mainly for efficiency reasons. If PCRE is compiled with 563Unicode property support, and the PCRE_UCP option is set, the behaviour is 564changed so that Unicode properties are used to determine character types, as 565follows: 566<pre> 567 \d any character that matches \p{Nd} (decimal digit) 568 \s any character that matches \p{Z} or \h or \v 569 \w any character that matches \p{L} or \p{N}, plus underscore 570</pre> 571The upper case escapes match the inverse sets of characters. Note that \d 572matches only decimal digits, whereas \w matches any Unicode digit, as well as 573any Unicode letter, and underscore. Note also that PCRE_UCP affects \b, and 574\B because they are defined in terms of \w and \W. Matching these sequences 575is noticeably slower when PCRE_UCP is set. 576</P> 577<P> 578The sequences \h, \H, \v, and \V are features that were added to Perl at 579release 5.10. In contrast to the other sequences, which match only ASCII 580characters by default, these always match certain high-valued code points, 581whether or not PCRE_UCP is set. The horizontal space characters are: 582<pre> 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</pre> 603The vertical space characters are: 604<pre> 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</pre> 613In 8-bit, non-UTF-8 mode, only the characters with codepoints less than 256 are 614relevant. 615<a name="newlineseq"></a></P> 616<br><b> 617Newline sequences 618</b><br> 619<P> 620Outside a character class, by default, the escape sequence \R matches any 621Unicode newline sequence. In 8-bit non-UTF-8 mode \R is equivalent to the 622following: 623<pre> 624 (?>\r\n|\n|\x0b|\f|\r|\x85) 625</pre> 626This is an example of an "atomic group", details of which are given 627<a href="#atomicgroup">below.</a> 628This particular group matches either the two-character sequence CR followed by 629LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab, 630U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next 631line, U+0085). The two-character sequence is treated as a single unit that 632cannot be split. 633</P> 634<P> 635In other modes, two additional characters whose codepoints are greater than 255 636are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029). 637Unicode character property support is not needed for these characters to be 638recognized. 639</P> 640<P> 641It is possible to restrict \R to match only CR, LF, or CRLF (instead of the 642complete set of Unicode line endings) by setting the option PCRE_BSR_ANYCRLF 643either at compile time or when the pattern is matched. (BSR is an abbrevation 644for "backslash R".) This can be made the default when PCRE is built; if this is 645the case, the other behaviour can be requested via the PCRE_BSR_UNICODE option. 646It is also possible to specify these settings by starting a pattern string with 647one of the following sequences: 648<pre> 649 (*BSR_ANYCRLF) CR, LF, or CRLF only 650 (*BSR_UNICODE) any Unicode newline sequence 651</pre> 652These override the default and the options given to the compiling function, but 653they can themselves be overridden by options given to a matching function. Note 654that these special settings, which are not Perl-compatible, are recognized only 655at the very start of a pattern, and that they must be in upper case. If more 656than one of them is present, the last one is used. They can be combined with a 657change of newline convention; for example, a pattern can start with: 658<pre> 659 (*ANY)(*BSR_ANYCRLF) 660</pre> 661They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF) or 662(*UCP) special sequences. Inside a character class, \R is treated as an 663unrecognized escape sequence, and so matches the letter "R" by default, but 664causes an error if PCRE_EXTRA is set. 665<a name="uniextseq"></a></P> 666<br><b> 667Unicode character properties 668</b><br> 669<P> 670When PCRE is built with Unicode character property support, three additional 671escape sequences that match characters with specific properties are available. 672When in 8-bit non-UTF-8 mode, these sequences are of course limited to testing 673characters whose codepoints are less than 256, but they do work in this mode. 674The extra escape sequences are: 675<pre> 676 \p{<i>xx</i>} a character with the <i>xx</i> property 677 \P{<i>xx</i>} a character without the <i>xx</i> property 678 \X a Unicode extended grapheme cluster 679</pre> 680The property names represented by <i>xx</i> above are limited to the Unicode 681script names, the general category properties, "Any", which matches any 682character (including newline), and some special PCRE properties (described 683in the 684<a href="#extraprops">next section).</a> 685Other Perl properties such as "InMusicalSymbols" are not currently supported by 686PCRE. Note that \P{Any} does not match any characters, so always causes a 687match failure. 688</P> 689<P> 690Sets of Unicode characters are defined as belonging to certain scripts. A 691character from one of these sets can be matched using a script name. For 692example: 693<pre> 694 \p{Greek} 695 \P{Han} 696</pre> 697Those that are not part of an identified script are lumped together as 698"Common". The current list of scripts is: 699</P> 700<P> 701Arabic, 702Armenian, 703Avestan, 704Balinese, 705Bamum, 706Bassa_Vah, 707Batak, 708Bengali, 709Bopomofo, 710Brahmi, 711Braille, 712Buginese, 713Buhid, 714Canadian_Aboriginal, 715Carian, 716Caucasian_Albanian, 717Chakma, 718Cham, 719Cherokee, 720Common, 721Coptic, 722Cuneiform, 723Cypriot, 724Cyrillic, 725Deseret, 726Devanagari, 727Duployan, 728Egyptian_Hieroglyphs, 729Elbasan, 730Ethiopic, 731Georgian, 732Glagolitic, 733Gothic, 734Grantha, 735Greek, 736Gujarati, 737Gurmukhi, 738Han, 739Hangul, 740Hanunoo, 741Hebrew, 742Hiragana, 743Imperial_Aramaic, 744Inherited, 745Inscriptional_Pahlavi, 746Inscriptional_Parthian, 747Javanese, 748Kaithi, 749Kannada, 750Katakana, 751Kayah_Li, 752Kharoshthi, 753Khmer, 754Khojki, 755Khudawadi, 756Lao, 757Latin, 758Lepcha, 759Limbu, 760Linear_A, 761Linear_B, 762Lisu, 763Lycian, 764Lydian, 765Mahajani, 766Malayalam, 767Mandaic, 768Manichaean, 769Meetei_Mayek, 770Mende_Kikakui, 771Meroitic_Cursive, 772Meroitic_Hieroglyphs, 773Miao, 774Modi, 775Mongolian, 776Mro, 777Myanmar, 778Nabataean, 779New_Tai_Lue, 780Nko, 781Ogham, 782Ol_Chiki, 783Old_Italic, 784Old_North_Arabian, 785Old_Permic, 786Old_Persian, 787Old_South_Arabian, 788Old_Turkic, 789Oriya, 790Osmanya, 791Pahawh_Hmong, 792Palmyrene, 793Pau_Cin_Hau, 794Phags_Pa, 795Phoenician, 796Psalter_Pahlavi, 797Rejang, 798Runic, 799Samaritan, 800Saurashtra, 801Sharada, 802Shavian, 803Siddham, 804Sinhala, 805Sora_Sompeng, 806Sundanese, 807Syloti_Nagri, 808Syriac, 809Tagalog, 810Tagbanwa, 811Tai_Le, 812Tai_Tham, 813Tai_Viet, 814Takri, 815Tamil, 816Telugu, 817Thaana, 818Thai, 819Tibetan, 820Tifinagh, 821Tirhuta, 822Ugaritic, 823Vai, 824Warang_Citi, 825Yi. 826</P> 827<P> 828Each character has exactly one Unicode general category property, specified by 829a two-letter abbreviation. For compatibility with Perl, negation can be 830specified by including a circumflex between the opening brace and the property 831name. For example, \p{^Lu} is the same as \P{Lu}. 832</P> 833<P> 834If only one letter is specified with \p or \P, it includes all the general 835category properties that start with that letter. In this case, in the absence 836of negation, the curly brackets in the escape sequence are optional; these two 837examples have the same effect: 838<pre> 839 \p{L} 840 \pL 841</pre> 842The following general category property codes are supported: 843<pre> 844 C Other 845 Cc Control 846 Cf Format 847 Cn Unassigned 848 Co Private use 849 Cs Surrogate 850 851 L Letter 852 Ll Lower case letter 853 Lm Modifier letter 854 Lo Other letter 855 Lt Title case letter 856 Lu Upper case letter 857 858 M Mark 859 Mc Spacing mark 860 Me Enclosing mark 861 Mn Non-spacing mark 862 863 N Number 864 Nd Decimal number 865 Nl Letter number 866 No Other number 867 868 P Punctuation 869 Pc Connector punctuation 870 Pd Dash punctuation 871 Pe Close punctuation 872 Pf Final punctuation 873 Pi Initial punctuation 874 Po Other punctuation 875 Ps Open punctuation 876 877 S Symbol 878 Sc Currency symbol 879 Sk Modifier symbol 880 Sm Mathematical symbol 881 So Other symbol 882 883 Z Separator 884 Zl Line separator 885 Zp Paragraph separator 886 Zs Space separator 887</pre> 888The special property L& is also supported: it matches a character that has 889the Lu, Ll, or Lt property, in other words, a letter that is not classified as 890a modifier or "other". 891</P> 892<P> 893The Cs (Surrogate) property applies only to characters in the range U+D800 to 894U+DFFF. Such characters are not valid in Unicode strings and so 895cannot be tested by PCRE, unless UTF validity checking has been turned off 896(see the discussion of PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK and 897PCRE_NO_UTF32_CHECK in the 898<a href="pcreapi.html"><b>pcreapi</b></a> 899page). Perl does not support the Cs property. 900</P> 901<P> 902The long synonyms for property names that Perl supports (such as \p{Letter}) 903are not supported by PCRE, nor is it permitted to prefix any of these 904properties with "Is". 905</P> 906<P> 907No character that is in the Unicode table has the Cn (unassigned) property. 908Instead, this property is assumed for any code point that is not in the 909Unicode table. 910</P> 911<P> 912Specifying caseless matching does not affect these escape sequences. For 913example, \p{Lu} always matches only upper case letters. This is different from 914the behaviour of current versions of Perl. 915</P> 916<P> 917Matching characters by Unicode property is not fast, because PCRE has to do a 918multistage table lookup in order to find a character's property. That is why 919the traditional escape sequences such as \d and \w do not use Unicode 920properties in PCRE by default, though you can make them do so by setting the 921PCRE_UCP option or by starting the pattern with (*UCP). 922</P> 923<br><b> 924Extended grapheme clusters 925</b><br> 926<P> 927The \X escape matches any number of Unicode characters that form an "extended 928grapheme cluster", and treats the sequence as an atomic group 929<a href="#atomicgroup">(see below).</a> 930Up to and including release 8.31, PCRE matched an earlier, simpler definition 931that was equivalent to 932<pre> 933 (?>\PM\pM*) 934</pre> 935That is, it matched a character without the "mark" property, followed by zero 936or more characters with the "mark" property. Characters with the "mark" 937property are typically non-spacing accents that affect the preceding character. 938</P> 939<P> 940This simple definition was extended in Unicode to include more complicated 941kinds of composite character by giving each character a grapheme breaking 942property, and creating rules that use these properties to define the boundaries 943of extended grapheme clusters. In releases of PCRE later than 8.31, \X matches 944one of these clusters. 945</P> 946<P> 947\X always matches at least one character. Then it decides whether to add 948additional characters according to the following rules for ending a cluster: 949</P> 950<P> 9511. End at the end of the subject string. 952</P> 953<P> 9542. Do not end between CR and LF; otherwise end after any control character. 955</P> 956<P> 9573. Do not break Hangul (a Korean script) syllable sequences. Hangul characters 958are of five types: L, V, T, LV, and LVT. An L character may be followed by an 959L, V, LV, or LVT character; an LV or V character may be followed by a V or T 960character; an LVT or T character may be follwed only by a T character. 961</P> 962<P> 9634. Do not end before extending characters or spacing marks. Characters with 964the "mark" property always have the "extend" grapheme breaking property. 965</P> 966<P> 9675. Do not end after prepend characters. 968</P> 969<P> 9706. Otherwise, end the cluster. 971<a name="extraprops"></a></P> 972<br><b> 973PCRE's additional properties 974</b><br> 975<P> 976As well as the standard Unicode properties described above, PCRE supports four 977more that make it possible to convert traditional escape sequences such as \w 978and \s to use Unicode properties. PCRE uses these non-standard, non-Perl 979properties internally when PCRE_UCP is set. However, they may also be used 980explicitly. These properties are: 981<pre> 982 Xan Any alphanumeric character 983 Xps Any POSIX space character 984 Xsp Any Perl space character 985 Xwd Any Perl "word" character 986</pre> 987Xan matches characters that have either the L (letter) or the N (number) 988property. Xps matches the characters tab, linefeed, vertical tab, form feed, or 989carriage return, and any other character that has the Z (separator) property. 990Xsp is the same as Xps; it used to exclude vertical tab, for Perl 991compatibility, but Perl changed, and so PCRE followed at release 8.34. Xwd 992matches the same characters as Xan, plus underscore. 993</P> 994<P> 995There is another non-standard property, Xuc, which matches any character that 996can be represented by a Universal Character Name in C++ and other programming 997languages. These are the characters $, @, ` (grave accent), and all characters 998with Unicode code points greater than or equal to U+00A0, except for the 999surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are 1000excluded. (Universal Character Names are of the form \uHHHH or \UHHHHHHHH 1001where H is a hexadecimal digit. Note that the Xuc property does not match these 1002sequences but the characters that they represent.) 1003<a name="resetmatchstart"></a></P> 1004<br><b> 1005Resetting the match start 1006</b><br> 1007<P> 1008The escape sequence \K causes any previously matched characters not to be 1009included in the final matched sequence. For example, the pattern: 1010<pre> 1011 foo\Kbar 1012</pre> 1013matches "foobar", but reports that it has matched "bar". This feature is 1014similar to a lookbehind assertion 1015<a href="#lookbehind">(described below).</a> 1016However, in this case, the part of the subject before the real match does not 1017have to be of fixed length, as lookbehind assertions do. The use of \K does 1018not interfere with the setting of 1019<a href="#subpattern">captured substrings.</a> 1020For example, when the pattern 1021<pre> 1022 (foo)\Kbar 1023</pre> 1024matches "foobar", the first substring is still set to "foo". 1025</P> 1026<P> 1027Perl documents that the use of \K within assertions is "not well defined". In 1028PCRE, \K is acted upon when it occurs inside positive assertions, but is 1029ignored in negative assertions. Note that when a pattern such as (?=ab\K) 1030matches, the reported start of the match can be greater than the end of the 1031match. 1032<a name="smallassertions"></a></P> 1033<br><b> 1034Simple assertions 1035</b><br> 1036<P> 1037The final use of backslash is for certain simple assertions. An assertion 1038specifies a condition that has to be met at a particular point in a match, 1039without consuming any characters from the subject string. The use of 1040subpatterns for more complicated assertions is described 1041<a href="#bigassertions">below.</a> 1042The backslashed assertions are: 1043<pre> 1044 \b matches at a word boundary 1045 \B matches when not at a word boundary 1046 \A matches at the start of the subject 1047 \Z matches at the end of the subject 1048 also matches before a newline at the end of the subject 1049 \z matches only at the end of the subject 1050 \G matches at the first matching position in the subject 1051</pre> 1052Inside a character class, \b has a different meaning; it matches the backspace 1053character. If any other of these assertions appears in a character class, by 1054default it matches the corresponding literal character (for example, \B 1055matches the letter B). However, if the PCRE_EXTRA option is set, an "invalid 1056escape sequence" error is generated instead. 1057</P> 1058<P> 1059A word boundary is a position in the subject string where the current character 1060and the previous character do not both match \w or \W (i.e. one matches 1061\w and the other matches \W), or the start or end of the string if the 1062first or last character matches \w, respectively. In a UTF mode, the meanings 1063of \w and \W can be changed by setting the PCRE_UCP option. When this is 1064done, it also affects \b and \B. Neither PCRE nor Perl has a separate "start 1065of word" or "end of word" metasequence. However, whatever follows \b normally 1066determines which it is. For example, the fragment \ba matches "a" at the start 1067of a word. 1068</P> 1069<P> 1070The \A, \Z, and \z assertions differ from the traditional circumflex and 1071dollar (described in the next section) in that they only ever match at the very 1072start and end of the subject string, whatever options are set. Thus, they are 1073independent of multiline mode. These three assertions are not affected by the 1074PCRE_NOTBOL or PCRE_NOTEOL options, which affect only the behaviour of the 1075circumflex and dollar metacharacters. However, if the <i>startoffset</i> 1076argument of <b>pcre_exec()</b> is non-zero, indicating that matching is to start 1077at a point other than the beginning of the subject, \A can never match. The 1078difference between \Z and \z is that \Z matches before a newline at the end 1079of the string as well as at the very end, whereas \z matches only at the end. 1080</P> 1081<P> 1082The \G assertion is true only when the current matching position is at the 1083start point of the match, as specified by the <i>startoffset</i> argument of 1084<b>pcre_exec()</b>. It differs from \A when the value of <i>startoffset</i> is 1085non-zero. By calling <b>pcre_exec()</b> multiple times with appropriate 1086arguments, you can mimic Perl's /g option, and it is in this kind of 1087implementation where \G can be useful. 1088</P> 1089<P> 1090Note, however, that PCRE's interpretation of \G, as the start of the current 1091match, is subtly different from Perl's, which defines it as the end of the 1092previous match. In Perl, these can be different when the previously matched 1093string was empty. Because PCRE does just one match at a time, it cannot 1094reproduce this behaviour. 1095</P> 1096<P> 1097If all the alternatives of a pattern begin with \G, the expression is anchored 1098to the starting match position, and the "anchored" flag is set in the compiled 1099regular expression. 1100</P> 1101<br><a name="SEC6" href="#TOC1">CIRCUMFLEX AND DOLLAR</a><br> 1102<P> 1103The circumflex and dollar metacharacters are zero-width assertions. That is, 1104they test for a particular condition being true without consuming any 1105characters from the subject string. 1106</P> 1107<P> 1108Outside a character class, in the default matching mode, the circumflex 1109character is an assertion that is true only if the current matching point is at 1110the start of the subject string. If the <i>startoffset</i> argument of 1111<b>pcre_exec()</b> is non-zero, circumflex can never match if the PCRE_MULTILINE 1112option is unset. Inside a character class, circumflex has an entirely different 1113meaning 1114<a href="#characterclass">(see below).</a> 1115</P> 1116<P> 1117Circumflex need not be the first character of the pattern if a number of 1118alternatives are involved, but it should be the first thing in each alternative 1119in which it appears if the pattern is ever to match that branch. If all 1120possible alternatives start with a circumflex, that is, if the pattern is 1121constrained to match only at the start of the subject, it is said to be an 1122"anchored" pattern. (There are also other constructs that can cause a pattern 1123to be anchored.) 1124</P> 1125<P> 1126The dollar character is an assertion that is true only if the current matching 1127point is at the end of the subject string, or immediately before a newline at 1128the end of the string (by default). Note, however, that it does not actually 1129match the newline. Dollar need not be the last character of the pattern if a 1130number of alternatives are involved, but it should be the last item in any 1131branch in which it appears. Dollar has no special meaning in a character class. 1132</P> 1133<P> 1134The meaning of dollar can be changed so that it matches only at the very end of 1135the string, by setting the PCRE_DOLLAR_ENDONLY option at compile time. This 1136does not affect the \Z assertion. 1137</P> 1138<P> 1139The meanings of the circumflex and dollar characters are changed if the 1140PCRE_MULTILINE option is set. When this is the case, a circumflex matches 1141immediately after internal newlines as well as at the start of the subject 1142string. It does not match after a newline that ends the string. A dollar 1143matches before any newlines in the string, as well as at the very end, when 1144PCRE_MULTILINE is set. When newline is specified as the two-character 1145sequence CRLF, isolated CR and LF characters do not indicate newlines. 1146</P> 1147<P> 1148For example, the pattern /^abc$/ matches the subject string "def\nabc" (where 1149\n represents a newline) in multiline mode, but not otherwise. Consequently, 1150patterns that are anchored in single line mode because all branches start with 1151^ are not anchored in multiline mode, and a match for circumflex is possible 1152when the <i>startoffset</i> argument of <b>pcre_exec()</b> is non-zero. The 1153PCRE_DOLLAR_ENDONLY option is ignored if PCRE_MULTILINE is set. 1154</P> 1155<P> 1156Note that the sequences \A, \Z, and \z can be used to match the start and 1157end of the subject in both modes, and if all branches of a pattern start with 1158\A it is always anchored, whether or not PCRE_MULTILINE is set. 1159<a name="fullstopdot"></a></P> 1160<br><a name="SEC7" href="#TOC1">FULL STOP (PERIOD, DOT) AND \N</a><br> 1161<P> 1162Outside a character class, a dot in the pattern matches any one character in 1163the subject string except (by default) a character that signifies the end of a 1164line. 1165</P> 1166<P> 1167When a line ending is defined as a single character, dot never matches that 1168character; when the two-character sequence CRLF is used, dot does not match CR 1169if it is immediately followed by LF, but otherwise it matches all characters 1170(including isolated CRs and LFs). When any Unicode line endings are being 1171recognized, dot does not match CR or LF or any of the other line ending 1172characters. 1173</P> 1174<P> 1175The behaviour of dot with regard to newlines can be changed. If the PCRE_DOTALL 1176option is set, a dot matches any one character, without exception. If the 1177two-character sequence CRLF is present in the subject string, it takes two dots 1178to match it. 1179</P> 1180<P> 1181The handling of dot is entirely independent of the handling of circumflex and 1182dollar, the only relationship being that they both involve newlines. Dot has no 1183special meaning in a character class. 1184</P> 1185<P> 1186The escape sequence \N behaves like a dot, except that it is not affected by 1187the PCRE_DOTALL option. In other words, it matches any character except one 1188that signifies the end of a line. Perl also uses \N to match characters by 1189name; PCRE does not support this. 1190</P> 1191<br><a name="SEC8" href="#TOC1">MATCHING A SINGLE DATA UNIT</a><br> 1192<P> 1193Outside a character class, the escape sequence \C matches any one data unit, 1194whether or not a UTF mode is set. In the 8-bit library, one data unit is one 1195byte; in the 16-bit library it is a 16-bit unit; in the 32-bit library it is 1196a 32-bit unit. Unlike a dot, \C always 1197matches line-ending characters. The feature is provided in Perl in order to 1198match individual bytes in UTF-8 mode, but it is unclear how it can usefully be 1199used. Because \C breaks up characters into individual data units, matching one 1200unit with \C in a UTF mode means that the rest of the string may start with a 1201malformed UTF character. This has undefined results, because PCRE assumes that 1202it is dealing with valid UTF strings (and by default it checks this at the 1203start of processing unless the PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or 1204PCRE_NO_UTF32_CHECK option is used). 1205</P> 1206<P> 1207PCRE does not allow \C to appear in lookbehind assertions 1208<a href="#lookbehind">(described below)</a> 1209in a UTF mode, because this would make it impossible to calculate the length of 1210the lookbehind. 1211</P> 1212<P> 1213In general, the \C escape sequence is best avoided. However, one 1214way of using it that avoids the problem of malformed UTF characters is to use a 1215lookahead to check the length of the next character, as in this pattern, which 1216could be used with a UTF-8 string (ignore white space and line breaks): 1217<pre> 1218 (?| (?=[\x00-\x7f])(\C) | 1219 (?=[\x80-\x{7ff}])(\C)(\C) | 1220 (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) | 1221 (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C)) 1222</pre> 1223A group that starts with (?| resets the capturing parentheses numbers in each 1224alternative (see 1225<a href="#dupsubpatternnumber">"Duplicate Subpattern Numbers"</a> 1226below). The assertions at the start of each branch check the next UTF-8 1227character for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The 1228character's individual bytes are then captured by the appropriate number of 1229groups. 1230<a name="characterclass"></a></P> 1231<br><a name="SEC9" href="#TOC1">SQUARE BRACKETS AND CHARACTER CLASSES</a><br> 1232<P> 1233An opening square bracket introduces a character class, terminated by a closing 1234square bracket. A closing square bracket on its own is not special by default. 1235However, if the PCRE_JAVASCRIPT_COMPAT option is set, a lone closing square 1236bracket causes a compile-time error. If a closing square bracket is required as 1237a member of the class, it should be the first data character in the class 1238(after an initial circumflex, if present) or escaped with a backslash. 1239</P> 1240<P> 1241A character class matches a single character in the subject. In a UTF mode, the 1242character may be more than one data unit long. A matched character must be in 1243the set of characters defined by the class, unless the first character in the 1244class definition is a circumflex, in which case the subject character must not 1245be in the set defined by the class. If a circumflex is actually required as a 1246member of the class, ensure it is not the first character, or escape it with a 1247backslash. 1248</P> 1249<P> 1250For example, the character class [aeiou] matches any lower case vowel, while 1251[^aeiou] matches any character that is not a lower case vowel. Note that a 1252circumflex is just a convenient notation for specifying the characters that 1253are in the class by enumerating those that are not. A class that starts with a 1254circumflex is not an assertion; it still consumes a character from the subject 1255string, and therefore it fails if the current pointer is at the end of the 1256string. 1257</P> 1258<P> 1259In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255 (0xffff) 1260can be included in a class as a literal string of data units, or by using the 1261\x{ escaping mechanism. 1262</P> 1263<P> 1264When caseless matching is set, any letters in a class represent both their 1265upper case and lower case versions, so for example, a caseless [aeiou] matches 1266"A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a 1267caseful version would. In a UTF mode, PCRE always understands the concept of 1268case for characters whose values are less than 128, so caseless matching is 1269always possible. For characters with higher values, the concept of case is 1270supported if PCRE is compiled with Unicode property support, but not otherwise. 1271If you want to use caseless matching in a UTF mode for characters 128 and 1272above, you must ensure that PCRE is compiled with Unicode property support as 1273well as with UTF support. 1274</P> 1275<P> 1276Characters that might indicate line breaks are never treated in any special way 1277when matching character classes, whatever line-ending sequence is in use, and 1278whatever setting of the PCRE_DOTALL and PCRE_MULTILINE options is used. A class 1279such as [^a] always matches one of these characters. 1280</P> 1281<P> 1282The minus (hyphen) character can be used to specify a range of characters in a 1283character class. For example, [d-m] matches any letter between d and m, 1284inclusive. If a minus character is required in a class, it must be escaped with 1285a backslash or appear in a position where it cannot be interpreted as 1286indicating a range, typically as the first or last character in the class, or 1287immediately after a range. For example, [b-d-z] matches letters in the range b 1288to d, a hyphen character, or z. 1289</P> 1290<P> 1291It is not possible to have the literal character "]" as the end character of a 1292range. A pattern such as [W-]46] is interpreted as a class of two characters 1293("W" and "-") followed by a literal string "46]", so it would match "W46]" or 1294"-46]". However, if the "]" is escaped with a backslash it is interpreted as 1295the end of range, so [W-\]46] is interpreted as a class containing a range 1296followed by two other characters. The octal or hexadecimal representation of 1297"]" can also be used to end a range. 1298</P> 1299<P> 1300An error is generated if a POSIX character class (see below) or an escape 1301sequence other than one that defines a single character appears at a point 1302where a range ending character is expected. For example, [z-\xff] is valid, 1303but [A-\d] and [A-[:digit:]] are not. 1304</P> 1305<P> 1306Ranges operate in the collating sequence of character values. They can also be 1307used for characters specified numerically, for example [\000-\037]. Ranges 1308can include any characters that are valid for the current mode. 1309</P> 1310<P> 1311If a range that includes letters is used when caseless matching is set, it 1312matches the letters in either case. For example, [W-c] is equivalent to 1313[][\\^_`wxyzabc], matched caselessly, and in a non-UTF mode, if character 1314tables for a French locale are in use, [\xc8-\xcb] matches accented E 1315characters in both cases. In UTF modes, PCRE supports the concept of case for 1316characters with values greater than 128 only when it is compiled with Unicode 1317property support. 1318</P> 1319<P> 1320The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, 1321\V, \w, and \W may appear in a character class, and add the characters that 1322they match to the class. For example, [\dABCDEF] matches any hexadecimal 1323digit. In UTF modes, the PCRE_UCP option affects the meanings of \d, \s, \w 1324and their upper case partners, just as it does when they appear outside a 1325character class, as described in the section entitled 1326<a href="#genericchartypes">"Generic character types"</a> 1327above. The escape sequence \b has a different meaning inside a character 1328class; it matches the backspace character. The sequences \B, \N, \R, and \X 1329are not special inside a character class. Like any other unrecognized escape 1330sequences, they are treated as the literal characters "B", "N", "R", and "X" by 1331default, but cause an error if the PCRE_EXTRA option is set. 1332</P> 1333<P> 1334A circumflex can conveniently be used with the upper case character types to 1335specify a more restricted set of characters than the matching lower case type. 1336For example, the class [^\W_] matches any letter or digit, but not underscore, 1337whereas [\w] includes underscore. A positive character class should be read as 1338"something OR something OR ..." and a negative class as "NOT something AND NOT 1339something AND NOT ...". 1340</P> 1341<P> 1342The only metacharacters that are recognized in character classes are backslash, 1343hyphen (only where it can be interpreted as specifying a range), circumflex 1344(only at the start), opening square bracket (only when it can be interpreted as 1345introducing a POSIX class name, or for a special compatibility feature - see 1346the next two sections), and the terminating closing square bracket. However, 1347escaping other non-alphanumeric characters does no harm. 1348</P> 1349<br><a name="SEC10" href="#TOC1">POSIX CHARACTER CLASSES</a><br> 1350<P> 1351Perl supports the POSIX notation for character classes. This uses names 1352enclosed by [: and :] within the enclosing square brackets. PCRE also supports 1353this notation. For example, 1354<pre> 1355 [01[:alpha:]%] 1356</pre> 1357matches "0", "1", any alphabetic character, or "%". The supported class names 1358are: 1359<pre> 1360 alnum letters and digits 1361 alpha letters 1362 ascii character codes 0 - 127 1363 blank space or tab only 1364 cntrl control characters 1365 digit decimal digits (same as \d) 1366 graph printing characters, excluding space 1367 lower lower case letters 1368 print printing characters, including space 1369 punct printing characters, excluding letters and digits and space 1370 space white space (the same as \s from PCRE 8.34) 1371 upper upper case letters 1372 word "word" characters (same as \w) 1373 xdigit hexadecimal digits 1374</pre> 1375The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), 1376and space (32). If locale-specific matching is taking place, the list of space 1377characters may be different; there may be fewer or more of them. "Space" used 1378to be different to \s, which did not include VT, for Perl compatibility. 1379However, Perl changed at release 5.18, and PCRE followed at release 8.34. 1380"Space" and \s now match the same set of characters. 1381</P> 1382<P> 1383The name "word" is a Perl extension, and "blank" is a GNU extension from Perl 13845.8. Another Perl extension is negation, which is indicated by a ^ character 1385after the colon. For example, 1386<pre> 1387 [12[:^digit:]] 1388</pre> 1389matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the POSIX 1390syntax [.ch.] and [=ch=] where "ch" is a "collating element", but these are not 1391supported, and an error is given if they are encountered. 1392</P> 1393<P> 1394By default, characters with values greater than 128 do not match any of the 1395POSIX character classes. However, if the PCRE_UCP option is passed to 1396<b>pcre_compile()</b>, some of the classes are changed so that Unicode character 1397properties are used. This is achieved by replacing certain POSIX classes by 1398other sequences, as follows: 1399<pre> 1400 [:alnum:] becomes \p{Xan} 1401 [:alpha:] becomes \p{L} 1402 [:blank:] becomes \h 1403 [:digit:] becomes \p{Nd} 1404 [:lower:] becomes \p{Ll} 1405 [:space:] becomes \p{Xps} 1406 [:upper:] becomes \p{Lu} 1407 [:word:] becomes \p{Xwd} 1408</pre> 1409Negated versions, such as [:^alpha:] use \P instead of \p. Three other POSIX 1410classes are handled specially in UCP mode: 1411</P> 1412<P> 1413[:graph:] 1414This matches characters that have glyphs that mark the page when printed. In 1415Unicode property terms, it matches all characters with the L, M, N, P, S, or Cf 1416properties, except for: 1417<pre> 1418 U+061C Arabic Letter Mark 1419 U+180E Mongolian Vowel Separator 1420 U+2066 - U+2069 Various "isolate"s 1421 1422</PRE> 1423</P> 1424<P> 1425[:print:] 1426This matches the same characters as [:graph:] plus space characters that are 1427not controls, that is, characters with the Zs property. 1428</P> 1429<P> 1430[:punct:] 1431This matches all characters that have the Unicode P (punctuation) property, 1432plus those characters whose code points are less than 128 that have the S 1433(Symbol) property. 1434</P> 1435<P> 1436The other POSIX classes are unchanged, and match only characters with code 1437points less than 128. 1438</P> 1439<br><a name="SEC11" href="#TOC1">COMPATIBILITY FEATURE FOR WORD BOUNDARIES</a><br> 1440<P> 1441In the POSIX.2 compliant library that was included in 4.4BSD Unix, the ugly 1442syntax [[:<:]] and [[:>:]] is used for matching "start of word" and "end of 1443word". PCRE treats these items as follows: 1444<pre> 1445 [[:<:]] is converted to \b(?=\w) 1446 [[:>:]] is converted to \b(?<=\w) 1447</pre> 1448Only these exact character sequences are recognized. A sequence such as 1449[a[:<:]b] provokes error for an unrecognized POSIX class name. This support is 1450not compatible with Perl. It is provided to help migrations from other 1451environments, and is best not used in any new patterns. Note that \b matches 1452at the start and the end of a word (see 1453<a href="#smallassertions">"Simple assertions"</a> 1454above), and in a Perl-style pattern the preceding or following character 1455normally shows which is wanted, without the need for the assertions that are 1456used above in order to give exactly the POSIX behaviour. 1457</P> 1458<br><a name="SEC12" href="#TOC1">VERTICAL BAR</a><br> 1459<P> 1460Vertical bar characters are used to separate alternative patterns. For example, 1461the pattern 1462<pre> 1463 gilbert|sullivan 1464</pre> 1465matches either "gilbert" or "sullivan". Any number of alternatives may appear, 1466and an empty alternative is permitted (matching the empty string). The matching 1467process tries each alternative in turn, from left to right, and the first one 1468that succeeds is used. If the alternatives are within a subpattern 1469<a href="#subpattern">(defined below),</a> 1470"succeeds" means matching the rest of the main pattern as well as the 1471alternative in the subpattern. 1472</P> 1473<br><a name="SEC13" href="#TOC1">INTERNAL OPTION SETTING</a><br> 1474<P> 1475The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and 1476PCRE_EXTENDED options (which are Perl-compatible) can be changed from within 1477the pattern by a sequence of Perl option letters enclosed between "(?" and ")". 1478The option letters are 1479<pre> 1480 i for PCRE_CASELESS 1481 m for PCRE_MULTILINE 1482 s for PCRE_DOTALL 1483 x for PCRE_EXTENDED 1484</pre> 1485For example, (?im) sets caseless, multiline matching. It is also possible to 1486unset these options by preceding the letter with a hyphen, and a combined 1487setting and unsetting such as (?im-sx), which sets PCRE_CASELESS and 1488PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, is also 1489permitted. If a letter appears both before and after the hyphen, the option is 1490unset. 1491</P> 1492<P> 1493The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA can be 1494changed in the same way as the Perl-compatible options by using the characters 1495J, U and X respectively. 1496</P> 1497<P> 1498When one of these option changes occurs at top level (that is, not inside 1499subpattern parentheses), the change applies to the remainder of the pattern 1500that follows. If the change is placed right at the start of a pattern, PCRE 1501extracts it into the global options (and it will therefore show up in data 1502extracted by the <b>pcre_fullinfo()</b> function). 1503</P> 1504<P> 1505An option change within a subpattern (see below for a description of 1506subpatterns) affects only that part of the subpattern that follows it, so 1507<pre> 1508 (a(?i)b)c 1509</pre> 1510matches abc and aBc and no other strings (assuming PCRE_CASELESS is not used). 1511By this means, options can be made to have different settings in different 1512parts of the pattern. Any changes made in one alternative do carry on 1513into subsequent branches within the same subpattern. For example, 1514<pre> 1515 (a(?i)b|c) 1516</pre> 1517matches "ab", "aB", "c", and "C", even though when matching "C" the first 1518branch is abandoned before the option setting. This is because the effects of 1519option settings happen at compile time. There would be some very weird 1520behaviour otherwise. 1521</P> 1522<P> 1523<b>Note:</b> There are other PCRE-specific options that can be set by the 1524application when the compiling or matching functions are called. In some cases 1525the pattern can contain special leading sequences such as (*CRLF) to override 1526what the application has set or what has been defaulted. Details are given in 1527the section entitled 1528<a href="#newlineseq">"Newline sequences"</a> 1529above. There are also the (*UTF8), (*UTF16),(*UTF32), and (*UCP) leading 1530sequences that can be used to set UTF and Unicode property modes; they are 1531equivalent to setting the PCRE_UTF8, PCRE_UTF16, PCRE_UTF32 and the PCRE_UCP 1532options, respectively. The (*UTF) sequence is a generic version that can be 1533used with any of the libraries. However, the application can set the 1534PCRE_NEVER_UTF option, which locks out the use of the (*UTF) sequences. 1535<a name="subpattern"></a></P> 1536<br><a name="SEC14" href="#TOC1">SUBPATTERNS</a><br> 1537<P> 1538Subpatterns are delimited by parentheses (round brackets), which can be nested. 1539Turning part of a pattern into a subpattern does two things: 1540<br> 1541<br> 15421. It localizes a set of alternatives. For example, the pattern 1543<pre> 1544 cat(aract|erpillar|) 1545</pre> 1546matches "cataract", "caterpillar", or "cat". Without the parentheses, it would 1547match "cataract", "erpillar" or an empty string. 1548<br> 1549<br> 15502. It sets up the subpattern as a capturing subpattern. This means that, when 1551the whole pattern matches, that portion of the subject string that matched the 1552subpattern is passed back to the caller via the <i>ovector</i> argument of the 1553matching function. (This applies only to the traditional matching functions; 1554the DFA matching functions do not support capturing.) 1555</P> 1556<P> 1557Opening parentheses are counted from left to right (starting from 1) to obtain 1558numbers for the capturing subpatterns. For example, if the string "the red 1559king" is matched against the pattern 1560<pre> 1561 the ((red|white) (king|queen)) 1562</pre> 1563the captured substrings are "red king", "red", and "king", and are numbered 1, 15642, and 3, respectively. 1565</P> 1566<P> 1567The fact that plain parentheses fulfil two functions is not always helpful. 1568There are often times when a grouping subpattern is required without a 1569capturing requirement. If an opening parenthesis is followed by a question mark 1570and a colon, the subpattern does not do any capturing, and is not counted when 1571computing the number of any subsequent capturing subpatterns. For example, if 1572the string "the white queen" is matched against the pattern 1573<pre> 1574 the ((?:red|white) (king|queen)) 1575</pre> 1576the captured substrings are "white queen" and "queen", and are numbered 1 and 15772. The maximum number of capturing subpatterns is 65535. 1578</P> 1579<P> 1580As a convenient shorthand, if any option settings are required at the start of 1581a non-capturing subpattern, the option letters may appear between the "?" and 1582the ":". Thus the two patterns 1583<pre> 1584 (?i:saturday|sunday) 1585 (?:(?i)saturday|sunday) 1586</pre> 1587match exactly the same set of strings. Because alternative branches are tried 1588from left to right, and options are not reset until the end of the subpattern 1589is reached, an option setting in one branch does affect subsequent branches, so 1590the above patterns match "SUNDAY" as well as "Saturday". 1591<a name="dupsubpatternnumber"></a></P> 1592<br><a name="SEC15" href="#TOC1">DUPLICATE SUBPATTERN NUMBERS</a><br> 1593<P> 1594Perl 5.10 introduced a feature whereby each alternative in a subpattern uses 1595the same numbers for its capturing parentheses. Such a subpattern starts with 1596(?| and is itself a non-capturing subpattern. For example, consider this 1597pattern: 1598<pre> 1599 (?|(Sat)ur|(Sun))day 1600</pre> 1601Because the two alternatives are inside a (?| group, both sets of capturing 1602parentheses are numbered one. Thus, when the pattern matches, you can look 1603at captured substring number one, whichever alternative matched. This construct 1604is useful when you want to capture part, but not all, of one of a number of 1605alternatives. Inside a (?| group, parentheses are numbered as usual, but the 1606number is reset at the start of each branch. The numbers of any capturing 1607parentheses that follow the subpattern start after the highest number used in 1608any branch. The following example is taken from the Perl documentation. The 1609numbers underneath show in which buffer the captured content will be stored. 1610<pre> 1611 # before ---------------branch-reset----------- after 1612 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x 1613 # 1 2 2 3 2 3 4 1614</pre> 1615A back reference to a numbered subpattern uses the most recent value that is 1616set for that number by any subpattern. The following pattern matches "abcabc" 1617or "defdef": 1618<pre> 1619 /(?|(abc)|(def))\1/ 1620</pre> 1621In contrast, a subroutine call to a numbered subpattern always refers to the 1622first one in the pattern with the given number. The following pattern matches 1623"abcabc" or "defabc": 1624<pre> 1625 /(?|(abc)|(def))(?1)/ 1626</pre> 1627If a 1628<a href="#conditions">condition test</a> 1629for a subpattern's having matched refers to a non-unique number, the test is 1630true if any of the subpatterns of that number have matched. 1631</P> 1632<P> 1633An alternative approach to using this "branch reset" feature is to use 1634duplicate named subpatterns, as described in the next section. 1635</P> 1636<br><a name="SEC16" href="#TOC1">NAMED SUBPATTERNS</a><br> 1637<P> 1638Identifying capturing parentheses by number is simple, but it can be very hard 1639to keep track of the numbers in complicated regular expressions. Furthermore, 1640if an expression is modified, the numbers may change. To help with this 1641difficulty, PCRE supports the naming of subpatterns. This feature was not 1642added to Perl until release 5.10. Python had the feature earlier, and PCRE 1643introduced it at release 4.0, using the Python syntax. PCRE now supports both 1644the Perl and the Python syntax. Perl allows identically numbered subpatterns to 1645have different names, but PCRE does not. 1646</P> 1647<P> 1648In PCRE, a subpattern can be named in one of three ways: (?<name>...) or 1649(?'name'...) as in Perl, or (?P<name>...) as in Python. References to capturing 1650parentheses from other parts of the pattern, such as 1651<a href="#backreferences">back references,</a> 1652<a href="#recursion">recursion,</a> 1653and 1654<a href="#conditions">conditions,</a> 1655can be made by name as well as by number. 1656</P> 1657<P> 1658Names consist of up to 32 alphanumeric characters and underscores, but must 1659start with a non-digit. Named capturing parentheses are still allocated numbers 1660as well as names, exactly as if the names were not present. The PCRE API 1661provides function calls for extracting the name-to-number translation table 1662from a compiled pattern. There is also a convenience function for extracting a 1663captured substring by name. 1664</P> 1665<P> 1666By default, a name must be unique within a pattern, but it is possible to relax 1667this constraint by setting the PCRE_DUPNAMES option at compile time. (Duplicate 1668names are also always permitted for subpatterns with the same number, set up as 1669described in the previous section.) Duplicate names can be useful for patterns 1670where only one instance of the named parentheses can match. Suppose you want to 1671match the name of a weekday, either as a 3-letter abbreviation or as the full 1672name, and in both cases you want to extract the abbreviation. This pattern 1673(ignoring the line breaks) does the job: 1674<pre> 1675 (?<DN>Mon|Fri|Sun)(?:day)?| 1676 (?<DN>Tue)(?:sday)?| 1677 (?<DN>Wed)(?:nesday)?| 1678 (?<DN>Thu)(?:rsday)?| 1679 (?<DN>Sat)(?:urday)? 1680</pre> 1681There are five capturing substrings, but only one is ever set after a match. 1682(An alternative way of solving this problem is to use a "branch reset" 1683subpattern, as described in the previous section.) 1684</P> 1685<P> 1686The convenience function for extracting the data by name returns the substring 1687for the first (and in this example, the only) subpattern of that name that 1688matched. This saves searching to find which numbered subpattern it was. 1689</P> 1690<P> 1691If you make a back reference to a non-unique named subpattern from elsewhere in 1692the pattern, the subpatterns to which the name refers are checked in the order 1693in which they appear in the overall pattern. The first one that is set is used 1694for the reference. For example, this pattern matches both "foofoo" and 1695"barbar" but not "foobar" or "barfoo": 1696<pre> 1697 (?:(?<n>foo)|(?<n>bar))\k<n> 1698 1699</PRE> 1700</P> 1701<P> 1702If you make a subroutine call to a non-unique named subpattern, the one that 1703corresponds to the first occurrence of the name is used. In the absence of 1704duplicate numbers (see the previous section) this is the one with the lowest 1705number. 1706</P> 1707<P> 1708If you use a named reference in a condition 1709test (see the 1710<a href="#conditions">section about conditions</a> 1711below), either to check whether a subpattern has matched, or to check for 1712recursion, all subpatterns with the same name are tested. If the condition is 1713true for any one of them, the overall condition is true. This is the same 1714behaviour as testing by number. For further details of the interfaces for 1715handling named subpatterns, see the 1716<a href="pcreapi.html"><b>pcreapi</b></a> 1717documentation. 1718</P> 1719<P> 1720<b>Warning:</b> You cannot use different names to distinguish between two 1721subpatterns with the same number because PCRE uses only the numbers when 1722matching. For this reason, an error is given at compile time if different names 1723are given to subpatterns with the same number. However, you can always give the 1724same name to subpatterns with the same number, even when PCRE_DUPNAMES is not 1725set. 1726</P> 1727<br><a name="SEC17" href="#TOC1">REPETITION</a><br> 1728<P> 1729Repetition is specified by quantifiers, which can follow any of the following 1730items: 1731<pre> 1732 a literal data character 1733 the dot metacharacter 1734 the \C escape sequence 1735 the \X escape sequence 1736 the \R escape sequence 1737 an escape such as \d or \pL that matches a single character 1738 a character class 1739 a back reference (see next section) 1740 a parenthesized subpattern (including assertions) 1741 a subroutine call to a subpattern (recursive or otherwise) 1742</pre> 1743The general repetition quantifier specifies a minimum and maximum number of 1744permitted matches, by giving the two numbers in curly brackets (braces), 1745separated by a comma. The numbers must be less than 65536, and the first must 1746be less than or equal to the second. For example: 1747<pre> 1748 z{2,4} 1749</pre> 1750matches "zz", "zzz", or "zzzz". A closing brace on its own is not a special 1751character. If the second number is omitted, but the comma is present, there is 1752no upper limit; if the second number and the comma are both omitted, the 1753quantifier specifies an exact number of required matches. Thus 1754<pre> 1755 [aeiou]{3,} 1756</pre> 1757matches at least 3 successive vowels, but may match many more, while 1758<pre> 1759 \d{8} 1760</pre> 1761matches exactly 8 digits. An opening curly bracket that appears in a position 1762where a quantifier is not allowed, or one that does not match the syntax of a 1763quantifier, is taken as a literal character. For example, {,6} is not a 1764quantifier, but a literal string of four characters. 1765</P> 1766<P> 1767In UTF modes, quantifiers apply to characters rather than to individual data 1768units. Thus, for example, \x{100}{2} matches two characters, each of 1769which is represented by a two-byte sequence in a UTF-8 string. Similarly, 1770\X{3} matches three Unicode extended grapheme clusters, each of which may be 1771several data units long (and they may be of different lengths). 1772</P> 1773<P> 1774The quantifier {0} is permitted, causing the expression to behave as if the 1775previous item and the quantifier were not present. This may be useful for 1776subpatterns that are referenced as 1777<a href="#subpatternsassubroutines">subroutines</a> 1778from elsewhere in the pattern (but see also the section entitled 1779<a href="#subdefine">"Defining subpatterns for use by reference only"</a> 1780below). Items other than subpatterns that have a {0} quantifier are omitted 1781from the compiled pattern. 1782</P> 1783<P> 1784For convenience, the three most common quantifiers have single-character 1785abbreviations: 1786<pre> 1787 * is equivalent to {0,} 1788 + is equivalent to {1,} 1789 ? is equivalent to {0,1} 1790</pre> 1791It is possible to construct infinite loops by following a subpattern that can 1792match no characters with a quantifier that has no upper limit, for example: 1793<pre> 1794 (a?)* 1795</pre> 1796Earlier versions of Perl and PCRE used to give an error at compile time for 1797such patterns. However, because there are cases where this can be useful, such 1798patterns are now accepted, but if any repetition of the subpattern does in fact 1799match no characters, the loop is forcibly broken. 1800</P> 1801<P> 1802By default, the quantifiers are "greedy", that is, they match as much as 1803possible (up to the maximum number of permitted times), without causing the 1804rest of the pattern to fail. The classic example of where this gives problems 1805is in trying to match comments in C programs. These appear between /* and */ 1806and within the comment, individual * and / characters may appear. An attempt to 1807match C comments by applying the pattern 1808<pre> 1809 /\*.*\*/ 1810</pre> 1811to the string 1812<pre> 1813 /* first comment */ not comment /* second comment */ 1814</pre> 1815fails, because it matches the entire string owing to the greediness of the .* 1816item. 1817</P> 1818<P> 1819However, if a quantifier is followed by a question mark, it ceases to be 1820greedy, and instead matches the minimum number of times possible, so the 1821pattern 1822<pre> 1823 /\*.*?\*/ 1824</pre> 1825does the right thing with the C comments. The meaning of the various 1826quantifiers is not otherwise changed, just the preferred number of matches. 1827Do not confuse this use of question mark with its use as a quantifier in its 1828own right. Because it has two uses, it can sometimes appear doubled, as in 1829<pre> 1830 \d??\d 1831</pre> 1832which matches one digit by preference, but can match two if that is the only 1833way the rest of the pattern matches. 1834</P> 1835<P> 1836If the PCRE_UNGREEDY option is set (an option that is not available in Perl), 1837the quantifiers are not greedy by default, but individual ones can be made 1838greedy by following them with a question mark. In other words, it inverts the 1839default behaviour. 1840</P> 1841<P> 1842When a parenthesized subpattern is quantified with a minimum repeat count that 1843is greater than 1 or with a limited maximum, more memory is required for the 1844compiled pattern, in proportion to the size of the minimum or maximum. 1845</P> 1846<P> 1847If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equivalent 1848to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is 1849implicitly anchored, because whatever follows will be tried against every 1850character position in the subject string, so there is no point in retrying the 1851overall match at any position after the first. PCRE normally treats such a 1852pattern as though it were preceded by \A. 1853</P> 1854<P> 1855In cases where it is known that the subject string contains no newlines, it is 1856worth setting PCRE_DOTALL in order to obtain this optimization, or 1857alternatively using ^ to indicate anchoring explicitly. 1858</P> 1859<P> 1860However, there are some cases where the optimization cannot be used. When .* 1861is inside capturing parentheses that are the subject of a back reference 1862elsewhere in the pattern, a match at the start may fail where a later one 1863succeeds. Consider, for example: 1864<pre> 1865 (.*)abc\1 1866</pre> 1867If the subject is "xyz123abc123" the match point is the fourth character. For 1868this reason, such a pattern is not implicitly anchored. 1869</P> 1870<P> 1871Another case where implicit anchoring is not applied is when the leading .* is 1872inside an atomic group. Once again, a match at the start may fail where a later 1873one succeeds. Consider this pattern: 1874<pre> 1875 (?>.*?a)b 1876</pre> 1877It matches "ab" in the subject "aab". The use of the backtracking control verbs 1878(*PRUNE) and (*SKIP) also disable this optimization. 1879</P> 1880<P> 1881When a capturing subpattern is repeated, the value captured is the substring 1882that matched the final iteration. For example, after 1883<pre> 1884 (tweedle[dume]{3}\s*)+ 1885</pre> 1886has matched "tweedledum tweedledee" the value of the captured substring is 1887"tweedledee". However, if there are nested capturing subpatterns, the 1888corresponding captured values may have been set in previous iterations. For 1889example, after 1890<pre> 1891 /(a|(b))+/ 1892</pre> 1893matches "aba" the value of the second captured substring is "b". 1894<a name="atomicgroup"></a></P> 1895<br><a name="SEC18" href="#TOC1">ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS</a><br> 1896<P> 1897With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy") 1898repetition, failure of what follows normally causes the repeated item to be 1899re-evaluated to see if a different number of repeats allows the rest of the 1900pattern to match. Sometimes it is useful to prevent this, either to change the 1901nature of the match, or to cause it fail earlier than it otherwise might, when 1902the author of the pattern knows there is no point in carrying on. 1903</P> 1904<P> 1905Consider, for example, the pattern \d+foo when applied to the subject line 1906<pre> 1907 123456bar 1908</pre> 1909After matching all 6 digits and then failing to match "foo", the normal 1910action of the matcher is to try again with only 5 digits matching the \d+ 1911item, and then with 4, and so on, before ultimately failing. "Atomic grouping" 1912(a term taken from Jeffrey Friedl's book) provides the means for specifying 1913that once a subpattern has matched, it is not to be re-evaluated in this way. 1914</P> 1915<P> 1916If we use atomic grouping for the previous example, the matcher gives up 1917immediately on failing to match "foo" the first time. The notation is a kind of 1918special parenthesis, starting with (?> as in this example: 1919<pre> 1920 (?>\d+)foo 1921</pre> 1922This kind of parenthesis "locks up" the part of the pattern it contains once 1923it has matched, and a failure further into the pattern is prevented from 1924backtracking into it. Backtracking past it to previous items, however, works as 1925normal. 1926</P> 1927<P> 1928An alternative description is that a subpattern of this type matches the string 1929of characters that an identical standalone pattern would match, if anchored at 1930the current point in the subject string. 1931</P> 1932<P> 1933Atomic grouping subpatterns are not capturing subpatterns. Simple cases such as 1934the above example can be thought of as a maximizing repeat that must swallow 1935everything it can. So, while both \d+ and \d+? are prepared to adjust the 1936number of digits they match in order to make the rest of the pattern match, 1937(?>\d+) can only match an entire sequence of digits. 1938</P> 1939<P> 1940Atomic groups in general can of course contain arbitrarily complicated 1941subpatterns, and can be nested. However, when the subpattern for an atomic 1942group is just a single repeated item, as in the example above, a simpler 1943notation, called a "possessive quantifier" can be used. This consists of an 1944additional + character following a quantifier. Using this notation, the 1945previous example can be rewritten as 1946<pre> 1947 \d++foo 1948</pre> 1949Note that a possessive quantifier can be used with an entire group, for 1950example: 1951<pre> 1952 (abc|xyz){2,3}+ 1953</pre> 1954Possessive quantifiers are always greedy; the setting of the PCRE_UNGREEDY 1955option is ignored. They are a convenient notation for the simpler forms of 1956atomic group. However, there is no difference in the meaning of a possessive 1957quantifier and the equivalent atomic group, though there may be a performance 1958difference; possessive quantifiers should be slightly faster. 1959</P> 1960<P> 1961The possessive quantifier syntax is an extension to the Perl 5.8 syntax. 1962Jeffrey Friedl originated the idea (and the name) in the first edition of his 1963book. Mike McCloskey liked it, so implemented it when he built Sun's Java 1964package, and PCRE copied it from there. It ultimately found its way into Perl 1965at release 5.10. 1966</P> 1967<P> 1968PCRE has an optimization that automatically "possessifies" certain simple 1969pattern constructs. For example, the sequence A+B is treated as A++B because 1970there is no point in backtracking into a sequence of A's when B must follow. 1971</P> 1972<P> 1973When a pattern contains an unlimited repeat inside a subpattern that can itself 1974be repeated an unlimited number of times, the use of an atomic group is the 1975only way to avoid some failing matches taking a very long time indeed. The 1976pattern 1977<pre> 1978 (\D+|<\d+>)*[!?] 1979</pre> 1980matches an unlimited number of substrings that either consist of non-digits, or 1981digits enclosed in <>, followed by either ! or ?. When it matches, it runs 1982quickly. However, if it is applied to 1983<pre> 1984 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa 1985</pre> 1986it takes a long time before reporting failure. This is because the string can 1987be divided between the internal \D+ repeat and the external * repeat in a 1988large number of ways, and all have to be tried. (The example uses [!?] rather 1989than a single character at the end, because both PCRE and Perl have an 1990optimization that allows for fast failure when a single character is used. They 1991remember the last single character that is required for a match, and fail early 1992if it is not present in the string.) If the pattern is changed so that it uses 1993an atomic group, like this: 1994<pre> 1995 ((?>\D+)|<\d+>)*[!?] 1996</pre> 1997sequences of non-digits cannot be broken, and failure happens quickly. 1998<a name="backreferences"></a></P> 1999<br><a name="SEC19" href="#TOC1">BACK REFERENCES</a><br> 2000<P> 2001Outside a character class, a backslash followed by a digit greater than 0 (and 2002possibly further digits) is a back reference to a capturing subpattern earlier 2003(that is, to its left) in the pattern, provided there have been that many 2004previous capturing left parentheses. 2005</P> 2006<P> 2007However, if the decimal number following the backslash is less than 10, it is 2008always taken as a back reference, and causes an error only if there are not 2009that many capturing left parentheses in the entire pattern. In other words, the 2010parentheses that are referenced need not be to the left of the reference for 2011numbers less than 10. A "forward back reference" of this type can make sense 2012when a repetition is involved and the subpattern to the right has participated 2013in an earlier iteration. 2014</P> 2015<P> 2016It is not possible to have a numerical "forward back reference" to a subpattern 2017whose number is 10 or more using this syntax because a sequence such as \50 is 2018interpreted as a character defined in octal. See the subsection entitled 2019"Non-printing characters" 2020<a href="#digitsafterbackslash">above</a> 2021for further details of the handling of digits following a backslash. There is 2022no such problem when named parentheses are used. A back reference to any 2023subpattern is possible using named parentheses (see below). 2024</P> 2025<P> 2026Another way of avoiding the ambiguity inherent in the use of digits following a 2027backslash is to use the \g escape sequence. This escape must be followed by an 2028unsigned number or a negative number, optionally enclosed in braces. These 2029examples are all identical: 2030<pre> 2031 (ring), \1 2032 (ring), \g1 2033 (ring), \g{1} 2034</pre> 2035An unsigned number specifies an absolute reference without the ambiguity that 2036is present in the older syntax. It is also useful when literal digits follow 2037the reference. A negative number is a relative reference. Consider this 2038example: 2039<pre> 2040 (abc(def)ghi)\g{-1} 2041</pre> 2042The sequence \g{-1} is a reference to the most recently started capturing 2043subpattern before \g, that is, is it equivalent to \2 in this example. 2044Similarly, \g{-2} would be equivalent to \1. The use of relative references 2045can be helpful in long patterns, and also in patterns that are created by 2046joining together fragments that contain references within themselves. 2047</P> 2048<P> 2049A back reference matches whatever actually matched the capturing subpattern in 2050the current subject string, rather than anything matching the subpattern 2051itself (see 2052<a href="#subpatternsassubroutines">"Subpatterns as subroutines"</a> 2053below for a way of doing that). So the pattern 2054<pre> 2055 (sens|respons)e and \1ibility 2056</pre> 2057matches "sense and sensibility" and "response and responsibility", but not 2058"sense and responsibility". If caseful matching is in force at the time of the 2059back reference, the case of letters is relevant. For example, 2060<pre> 2061 ((?i)rah)\s+\1 2062</pre> 2063matches "rah rah" and "RAH RAH", but not "RAH rah", even though the original 2064capturing subpattern is matched caselessly. 2065</P> 2066<P> 2067There are several different ways of writing back references to named 2068subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or 2069\k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's unified 2070back reference syntax, in which \g can be used for both numeric and named 2071references, is also supported. We could rewrite the above example in any of 2072the following ways: 2073<pre> 2074 (?<p1>(?i)rah)\s+\k<p1> 2075 (?'p1'(?i)rah)\s+\k{p1} 2076 (?P<p1>(?i)rah)\s+(?P=p1) 2077 (?<p1>(?i)rah)\s+\g{p1} 2078</pre> 2079A subpattern that is referenced by name may appear in the pattern before or 2080after the reference. 2081</P> 2082<P> 2083There may be more than one back reference to the same subpattern. If a 2084subpattern has not actually been used in a particular match, any back 2085references to it always fail by default. For example, the pattern 2086<pre> 2087 (a|(bc))\2 2088</pre> 2089always fails if it starts to match "a" rather than "bc". However, if the 2090PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back reference to an 2091unset value matches an empty string. 2092</P> 2093<P> 2094Because there may be many capturing parentheses in a pattern, all digits 2095following a backslash are taken as part of a potential back reference number. 2096If the pattern continues with a digit character, some delimiter must be used to 2097terminate the back reference. If the PCRE_EXTENDED option is set, this can be 2098white space. Otherwise, the \g{ syntax or an empty comment (see 2099<a href="#comments">"Comments"</a> 2100below) can be used. 2101</P> 2102<br><b> 2103Recursive back references 2104</b><br> 2105<P> 2106A back reference that occurs inside the parentheses to which it refers fails 2107when the subpattern is first used, so, for example, (a\1) never matches. 2108However, such references can be useful inside repeated subpatterns. For 2109example, the pattern 2110<pre> 2111 (a|b\1)+ 2112</pre> 2113matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of 2114the subpattern, the back reference matches the character string corresponding 2115to the previous iteration. In order for this to work, the pattern must be such 2116that the first iteration does not need to match the back reference. This can be 2117done using alternation, as in the example above, or by a quantifier with a 2118minimum of zero. 2119</P> 2120<P> 2121Back references of this type cause the group that they reference to be treated 2122as an 2123<a href="#atomicgroup">atomic group.</a> 2124Once the whole group has been matched, a subsequent matching failure cannot 2125cause backtracking into the middle of the group. 2126<a name="bigassertions"></a></P> 2127<br><a name="SEC20" href="#TOC1">ASSERTIONS</a><br> 2128<P> 2129An assertion is a test on the characters following or preceding the current 2130matching point that does not actually consume any characters. The simple 2131assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are described 2132<a href="#smallassertions">above.</a> 2133</P> 2134<P> 2135More complicated assertions are coded as subpatterns. There are two kinds: 2136those that look ahead of the current position in the subject string, and those 2137that look behind it. An assertion subpattern is matched in the normal way, 2138except that it does not cause the current matching position to be changed. 2139</P> 2140<P> 2141Assertion subpatterns are not capturing subpatterns. If such an assertion 2142contains capturing subpatterns within it, these are counted for the purposes of 2143numbering the capturing subpatterns in the whole pattern. However, substring 2144capturing is carried out only for positive assertions. (Perl sometimes, but not 2145always, does do capturing in negative assertions.) 2146</P> 2147<P> 2148For compatibility with Perl, assertion subpatterns may be repeated; though 2149it makes no sense to assert the same thing several times, the side effect of 2150capturing parentheses may occasionally be useful. In practice, there only three 2151cases: 2152<br> 2153<br> 2154(1) If the quantifier is {0}, the assertion is never obeyed during matching. 2155However, it may contain internal capturing parenthesized groups that are called 2156from elsewhere via the 2157<a href="#subpatternsassubroutines">subroutine mechanism.</a> 2158<br> 2159<br> 2160(2) If quantifier is {0,n} where n is greater than zero, it is treated as if it 2161were {0,1}. At run time, the rest of the pattern match is tried with and 2162without the assertion, the order depending on the greediness of the quantifier. 2163<br> 2164<br> 2165(3) If the minimum repetition is greater than zero, the quantifier is ignored. 2166The assertion is obeyed just once when encountered during matching. 2167</P> 2168<br><b> 2169Lookahead assertions 2170</b><br> 2171<P> 2172Lookahead assertions start with (?= for positive assertions and (?! for 2173negative assertions. For example, 2174<pre> 2175 \w+(?=;) 2176</pre> 2177matches a word followed by a semicolon, but does not include the semicolon in 2178the match, and 2179<pre> 2180 foo(?!bar) 2181</pre> 2182matches any occurrence of "foo" that is not followed by "bar". Note that the 2183apparently similar pattern 2184<pre> 2185 (?!foo)bar 2186</pre> 2187does not find an occurrence of "bar" that is preceded by something other than 2188"foo"; it finds any occurrence of "bar" whatsoever, because the assertion 2189(?!foo) is always true when the next three characters are "bar". A 2190lookbehind assertion is needed to achieve the other effect. 2191</P> 2192<P> 2193If you want to force a matching failure at some point in a pattern, the most 2194convenient way to do it is with (?!) because an empty string always matches, so 2195an assertion that requires there not to be an empty string must always fail. 2196The backtracking control verb (*FAIL) or (*F) is a synonym for (?!). 2197<a name="lookbehind"></a></P> 2198<br><b> 2199Lookbehind assertions 2200</b><br> 2201<P> 2202Lookbehind assertions start with (?<= for positive assertions and (?<! for 2203negative assertions. For example, 2204<pre> 2205 (?<!foo)bar 2206</pre> 2207does find an occurrence of "bar" that is not preceded by "foo". The contents of 2208a lookbehind assertion are restricted such that all the strings it matches must 2209have a fixed length. However, if there are several top-level alternatives, they 2210do not all have to have the same fixed length. Thus 2211<pre> 2212 (?<=bullock|donkey) 2213</pre> 2214is permitted, but 2215<pre> 2216 (?<!dogs?|cats?) 2217</pre> 2218causes an error at compile time. Branches that match different length strings 2219are permitted only at the top level of a lookbehind assertion. This is an 2220extension compared with Perl, which requires all branches to match the same 2221length of string. An assertion such as 2222<pre> 2223 (?<=ab(c|de)) 2224</pre> 2225is not permitted, because its single top-level branch can match two different 2226lengths, but it is acceptable to PCRE if rewritten to use two top-level 2227branches: 2228<pre> 2229 (?<=abc|abde) 2230</pre> 2231In some cases, the escape sequence \K 2232<a href="#resetmatchstart">(see above)</a> 2233can be used instead of a lookbehind assertion to get round the fixed-length 2234restriction. 2235</P> 2236<P> 2237The implementation of lookbehind assertions is, for each alternative, to 2238temporarily move the current position back by the fixed length and then try to 2239match. If there are insufficient characters before the current position, the 2240assertion fails. 2241</P> 2242<P> 2243In a UTF mode, PCRE does not allow the \C escape (which matches a single data 2244unit even in a UTF mode) to appear in lookbehind assertions, because it makes 2245it impossible to calculate the length of the lookbehind. The \X and \R 2246escapes, which can match different numbers of data units, are also not 2247permitted. 2248</P> 2249<P> 2250<a href="#subpatternsassubroutines">"Subroutine"</a> 2251calls (see below) such as (?2) or (?&X) are permitted in lookbehinds, as long 2252as the subpattern matches a fixed-length string. 2253<a href="#recursion">Recursion,</a> 2254however, is not supported. 2255</P> 2256<P> 2257Possessive quantifiers can be used in conjunction with lookbehind assertions to 2258specify efficient matching of fixed-length strings at the end of subject 2259strings. Consider a simple pattern such as 2260<pre> 2261 abcd$ 2262</pre> 2263when applied to a long string that does not match. Because matching proceeds 2264from left to right, PCRE will look for each "a" in the subject and then see if 2265what follows matches the rest of the pattern. If the pattern is specified as 2266<pre> 2267 ^.*abcd$ 2268</pre> 2269the initial .* matches the entire string at first, but when this fails (because 2270there is no following "a"), it backtracks to match all but the last character, 2271then all but the last two characters, and so on. Once again the search for "a" 2272covers the entire string, from right to left, so we are no better off. However, 2273if the pattern is written as 2274<pre> 2275 ^.*+(?<=abcd) 2276</pre> 2277there can be no backtracking for the .*+ item; it can match only the entire 2278string. The subsequent lookbehind assertion does a single test on the last four 2279characters. If it fails, the match fails immediately. For long strings, this 2280approach makes a significant difference to the processing time. 2281</P> 2282<br><b> 2283Using multiple assertions 2284</b><br> 2285<P> 2286Several assertions (of any sort) may occur in succession. For example, 2287<pre> 2288 (?<=\d{3})(?<!999)foo 2289</pre> 2290matches "foo" preceded by three digits that are not "999". Notice that each of 2291the assertions is applied independently at the same point in the subject 2292string. First there is a check that the previous three characters are all 2293digits, and then there is a check that the same three characters are not "999". 2294This pattern does <i>not</i> match "foo" preceded by six characters, the first 2295of which are digits and the last three of which are not "999". For example, it 2296doesn't match "123abcfoo". A pattern to do that is 2297<pre> 2298 (?<=\d{3}...)(?<!999)foo 2299</pre> 2300This time the first assertion looks at the preceding six characters, checking 2301that the first three are digits, and then the second assertion checks that the 2302preceding three characters are not "999". 2303</P> 2304<P> 2305Assertions can be nested in any combination. For example, 2306<pre> 2307 (?<=(?<!foo)bar)baz 2308</pre> 2309matches an occurrence of "baz" that is preceded by "bar" which in turn is not 2310preceded by "foo", while 2311<pre> 2312 (?<=\d{3}(?!999)...)foo 2313</pre> 2314is another pattern that matches "foo" preceded by three digits and any three 2315characters that are not "999". 2316<a name="conditions"></a></P> 2317<br><a name="SEC21" href="#TOC1">CONDITIONAL SUBPATTERNS</a><br> 2318<P> 2319It is possible to cause the matching process to obey a subpattern 2320conditionally or to choose between two alternative subpatterns, depending on 2321the result of an assertion, or whether a specific capturing subpattern has 2322already been matched. The two possible forms of conditional subpattern are: 2323<pre> 2324 (?(condition)yes-pattern) 2325 (?(condition)yes-pattern|no-pattern) 2326</pre> 2327If the condition is satisfied, the yes-pattern is used; otherwise the 2328no-pattern (if present) is used. If there are more than two alternatives in the 2329subpattern, a compile-time error occurs. Each of the two alternatives may 2330itself contain nested subpatterns of any form, including conditional 2331subpatterns; the restriction to two alternatives applies only at the level of 2332the condition. This pattern fragment is an example where the alternatives are 2333complex: 2334<pre> 2335 (?(1) (A|B|C) | (D | (?(2)E|F) | E) ) 2336 2337</PRE> 2338</P> 2339<P> 2340There are four kinds of condition: references to subpatterns, references to 2341recursion, a pseudo-condition called DEFINE, and assertions. 2342</P> 2343<br><b> 2344Checking for a used subpattern by number 2345</b><br> 2346<P> 2347If the text between the parentheses consists of a sequence of digits, the 2348condition is true if a capturing subpattern of that number has previously 2349matched. If there is more than one capturing subpattern with the same number 2350(see the earlier 2351<a href="#recursion">section about duplicate subpattern numbers),</a> 2352the condition is true if any of them have matched. An alternative notation is 2353to precede the digits with a plus or minus sign. In this case, the subpattern 2354number is relative rather than absolute. The most recently opened parentheses 2355can be referenced by (?(-1), the next most recent by (?(-2), and so on. Inside 2356loops it can also make sense to refer to subsequent groups. The next 2357parentheses to be opened can be referenced as (?(+1), and so on. (The value 2358zero in any of these forms is not used; it provokes a compile-time error.) 2359</P> 2360<P> 2361Consider the following pattern, which contains non-significant white space to 2362make it more readable (assume the PCRE_EXTENDED option) and to divide it into 2363three parts for ease of discussion: 2364<pre> 2365 ( \( )? [^()]+ (?(1) \) ) 2366</pre> 2367The first part matches an optional opening parenthesis, and if that 2368character is present, sets it as the first captured substring. The second part 2369matches one or more characters that are not parentheses. The third part is a 2370conditional subpattern that tests whether or not the first set of parentheses 2371matched. If they did, that is, if subject started with an opening parenthesis, 2372the condition is true, and so the yes-pattern is executed and a closing 2373parenthesis is required. Otherwise, since no-pattern is not present, the 2374subpattern matches nothing. In other words, this pattern matches a sequence of 2375non-parentheses, optionally enclosed in parentheses. 2376</P> 2377<P> 2378If you were embedding this pattern in a larger one, you could use a relative 2379reference: 2380<pre> 2381 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ... 2382</pre> 2383This makes the fragment independent of the parentheses in the larger pattern. 2384</P> 2385<br><b> 2386Checking for a used subpattern by name 2387</b><br> 2388<P> 2389Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used 2390subpattern by name. For compatibility with earlier versions of PCRE, which had 2391this facility before Perl, the syntax (?(name)...) is also recognized. 2392</P> 2393<P> 2394Rewriting the above example to use a named subpattern gives this: 2395<pre> 2396 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) ) 2397</pre> 2398If the name used in a condition of this kind is a duplicate, the test is 2399applied to all subpatterns of the same name, and is true if any one of them has 2400matched. 2401</P> 2402<br><b> 2403Checking for pattern recursion 2404</b><br> 2405<P> 2406If the condition is the string (R), and there is no subpattern with the name R, 2407the condition is true if a recursive call to the whole pattern or any 2408subpattern has been made. If digits or a name preceded by ampersand follow the 2409letter R, for example: 2410<pre> 2411 (?(R3)...) or (?(R&name)...) 2412</pre> 2413the condition is true if the most recent recursion is into a subpattern whose 2414number or name is given. This condition does not check the entire recursion 2415stack. If the name used in a condition of this kind is a duplicate, the test is 2416applied to all subpatterns of the same name, and is true if any one of them is 2417the most recent recursion. 2418</P> 2419<P> 2420At "top level", all these recursion test conditions are false. 2421<a href="#recursion">The syntax for recursive patterns</a> 2422is described below. 2423<a name="subdefine"></a></P> 2424<br><b> 2425Defining subpatterns for use by reference only 2426</b><br> 2427<P> 2428If the condition is the string (DEFINE), and there is no subpattern with the 2429name DEFINE, the condition is always false. In this case, there may be only one 2430alternative in the subpattern. It is always skipped if control reaches this 2431point in the pattern; the idea of DEFINE is that it can be used to define 2432subroutines that can be referenced from elsewhere. (The use of 2433<a href="#subpatternsassubroutines">subroutines</a> 2434is described below.) For example, a pattern to match an IPv4 address such as 2435"192.168.23.245" could be written like this (ignore white space and line 2436breaks): 2437<pre> 2438 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) ) 2439 \b (?&byte) (\.(?&byte)){3} \b 2440</pre> 2441The first part of the pattern is a DEFINE group inside which a another group 2442named "byte" is defined. This matches an individual component of an IPv4 2443address (a number less than 256). When matching takes place, this part of the 2444pattern is skipped because DEFINE acts like a false condition. The rest of the 2445pattern uses references to the named group to match the four dot-separated 2446components of an IPv4 address, insisting on a word boundary at each end. 2447</P> 2448<br><b> 2449Assertion conditions 2450</b><br> 2451<P> 2452If the condition is not in any of the above formats, it must be an assertion. 2453This may be a positive or negative lookahead or lookbehind assertion. Consider 2454this pattern, again containing non-significant white space, and with the two 2455alternatives on the second line: 2456<pre> 2457 (?(?=[^a-z]*[a-z]) 2458 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} ) 2459</pre> 2460The condition is a positive lookahead assertion that matches an optional 2461sequence of non-letters followed by a letter. In other words, it tests for the 2462presence of at least one letter in the subject. If a letter is found, the 2463subject is matched against the first alternative; otherwise it is matched 2464against the second. This pattern matches strings in one of the two forms 2465dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits. 2466<a name="comments"></a></P> 2467<br><a name="SEC22" href="#TOC1">COMMENTS</a><br> 2468<P> 2469There are two ways of including comments in patterns that are processed by 2470PCRE. In both cases, the start of the comment must not be in a character class, 2471nor in the middle of any other sequence of related characters such as (?: or a 2472subpattern name or number. The characters that make up a comment play no part 2473in the pattern matching. 2474</P> 2475<P> 2476The sequence (?# marks the start of a comment that continues up to the next 2477closing parenthesis. Nested parentheses are not permitted. If the PCRE_EXTENDED 2478option is set, an unescaped # character also introduces a comment, which in 2479this case continues to immediately after the next newline character or 2480character sequence in the pattern. Which characters are interpreted as newlines 2481is controlled by the options passed to a compiling function or by a special 2482sequence at the start of the pattern, as described in the section entitled 2483<a href="#newlines">"Newline conventions"</a> 2484above. Note that the end of this type of comment is a literal newline sequence 2485in the pattern; escape sequences that happen to represent a newline do not 2486count. For example, consider this pattern when PCRE_EXTENDED is set, and the 2487default newline convention is in force: 2488<pre> 2489 abc #comment \n still comment 2490</pre> 2491On encountering the # character, <b>pcre_compile()</b> skips along, looking for 2492a newline in the pattern. The sequence \n is still literal at this stage, so 2493it does not terminate the comment. Only an actual character with the code value 24940x0a (the default newline) does so. 2495<a name="recursion"></a></P> 2496<br><a name="SEC23" href="#TOC1">RECURSIVE PATTERNS</a><br> 2497<P> 2498Consider the problem of matching a string in parentheses, allowing for 2499unlimited nested parentheses. Without the use of recursion, the best that can 2500be done is to use a pattern that matches up to some fixed depth of nesting. It 2501is not possible to handle an arbitrary nesting depth. 2502</P> 2503<P> 2504For some time, Perl has provided a facility that allows regular expressions to 2505recurse (amongst other things). It does this by interpolating Perl code in the 2506expression at run time, and the code can refer to the expression itself. A Perl 2507pattern using code interpolation to solve the parentheses problem can be 2508created like this: 2509<pre> 2510 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x; 2511</pre> 2512The (?p{...}) item interpolates Perl code at run time, and in this case refers 2513recursively to the pattern in which it appears. 2514</P> 2515<P> 2516Obviously, PCRE cannot support the interpolation of Perl code. Instead, it 2517supports special syntax for recursion of the entire pattern, and also for 2518individual subpattern recursion. After its introduction in PCRE and Python, 2519this kind of recursion was subsequently introduced into Perl at release 5.10. 2520</P> 2521<P> 2522A special item that consists of (? followed by a number greater than zero and a 2523closing parenthesis is a recursive subroutine call of the subpattern of the 2524given number, provided that it occurs inside that subpattern. (If not, it is a 2525<a href="#subpatternsassubroutines">non-recursive subroutine</a> 2526call, which is described in the next section.) The special item (?R) or (?0) is 2527a recursive call of the entire regular expression. 2528</P> 2529<P> 2530This PCRE pattern solves the nested parentheses problem (assume the 2531PCRE_EXTENDED option is set so that white space is ignored): 2532<pre> 2533 \( ( [^()]++ | (?R) )* \) 2534</pre> 2535First it matches an opening parenthesis. Then it matches any number of 2536substrings which can either be a sequence of non-parentheses, or a recursive 2537match of the pattern itself (that is, a correctly parenthesized substring). 2538Finally there is a closing parenthesis. Note the use of a possessive quantifier 2539to avoid backtracking into sequences of non-parentheses. 2540</P> 2541<P> 2542If this were part of a larger pattern, you would not want to recurse the entire 2543pattern, so instead you could use this: 2544<pre> 2545 ( \( ( [^()]++ | (?1) )* \) ) 2546</pre> 2547We have put the pattern into parentheses, and caused the recursion to refer to 2548them instead of the whole pattern. 2549</P> 2550<P> 2551In a larger pattern, keeping track of parenthesis numbers can be tricky. This 2552is made easier by the use of relative references. Instead of (?1) in the 2553pattern above you can write (?-2) to refer to the second most recently opened 2554parentheses preceding the recursion. In other words, a negative number counts 2555capturing parentheses leftwards from the point at which it is encountered. 2556</P> 2557<P> 2558It is also possible to refer to subsequently opened parentheses, by writing 2559references such as (?+2). However, these cannot be recursive because the 2560reference is not inside the parentheses that are referenced. They are always 2561<a href="#subpatternsassubroutines">non-recursive subroutine</a> 2562calls, as described in the next section. 2563</P> 2564<P> 2565An alternative approach is to use named parentheses instead. The Perl syntax 2566for this is (?&name); PCRE's earlier syntax (?P>name) is also supported. We 2567could rewrite the above example as follows: 2568<pre> 2569 (?<pn> \( ( [^()]++ | (?&pn) )* \) ) 2570</pre> 2571If there is more than one subpattern with the same name, the earliest one is 2572used. 2573</P> 2574<P> 2575This particular example pattern that we have been looking at contains nested 2576unlimited repeats, and so the use of a possessive quantifier for matching 2577strings of non-parentheses is important when applying the pattern to strings 2578that do not match. For example, when this pattern is applied to 2579<pre> 2580 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa() 2581</pre> 2582it yields "no match" quickly. However, if a possessive quantifier is not used, 2583the match runs for a very long time indeed because there are so many different 2584ways the + and * repeats can carve up the subject, and all have to be tested 2585before failure can be reported. 2586</P> 2587<P> 2588At the end of a match, the values of capturing parentheses are those from 2589the outermost level. If you want to obtain intermediate values, a callout 2590function can be used (see below and the 2591<a href="pcrecallout.html"><b>pcrecallout</b></a> 2592documentation). If the pattern above is matched against 2593<pre> 2594 (ab(cd)ef) 2595</pre> 2596the value for the inner capturing parentheses (numbered 2) is "ef", which is 2597the last value taken on at the top level. If a capturing subpattern is not 2598matched at the top level, its final captured value is unset, even if it was 2599(temporarily) set at a deeper level during the matching process. 2600</P> 2601<P> 2602If there are more than 15 capturing parentheses in a pattern, PCRE has to 2603obtain extra memory to store data during a recursion, which it does by using 2604<b>pcre_malloc</b>, freeing it via <b>pcre_free</b> afterwards. If no memory can 2605be obtained, the match fails with the PCRE_ERROR_NOMEMORY error. 2606</P> 2607<P> 2608Do not confuse the (?R) item with the condition (R), which tests for recursion. 2609Consider this pattern, which matches text in angle brackets, allowing for 2610arbitrary nesting. Only digits are allowed in nested brackets (that is, when 2611recursing), whereas any characters are permitted at the outer level. 2612<pre> 2613 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * > 2614</pre> 2615In this pattern, (?(R) is the start of a conditional subpattern, with two 2616different alternatives for the recursive and non-recursive cases. The (?R) item 2617is the actual recursive call. 2618<a name="recursiondifference"></a></P> 2619<br><b> 2620Differences in recursion processing between PCRE and Perl 2621</b><br> 2622<P> 2623Recursion processing in PCRE differs from Perl in two important ways. In PCRE 2624(like Python, but unlike Perl), a recursive subpattern call is always treated 2625as an atomic group. That is, once it has matched some of the subject string, it 2626is never re-entered, even if it contains untried alternatives and there is a 2627subsequent matching failure. This can be illustrated by the following pattern, 2628which purports to match a palindromic string that contains an odd number of 2629characters (for example, "a", "aba", "abcba", "abcdcba"): 2630<pre> 2631 ^(.|(.)(?1)\2)$ 2632</pre> 2633The idea is that it either matches a single character, or two identical 2634characters surrounding a sub-palindrome. In Perl, this pattern works; in PCRE 2635it does not if the pattern is longer than three characters. Consider the 2636subject string "abcba": 2637</P> 2638<P> 2639At the top level, the first character is matched, but as it is not at the end 2640of the string, the first alternative fails; the second alternative is taken 2641and the recursion kicks in. The recursive call to subpattern 1 successfully 2642matches the next character ("b"). (Note that the beginning and end of line 2643tests are not part of the recursion). 2644</P> 2645<P> 2646Back at the top level, the next character ("c") is compared with what 2647subpattern 2 matched, which was "a". This fails. Because the recursion is 2648treated as an atomic group, there are now no backtracking points, and so the 2649entire match fails. (Perl is able, at this point, to re-enter the recursion and 2650try the second alternative.) However, if the pattern is written with the 2651alternatives in the other order, things are different: 2652<pre> 2653 ^((.)(?1)\2|.)$ 2654</pre> 2655This time, the recursing alternative is tried first, and continues to recurse 2656until it runs out of characters, at which point the recursion fails. But this 2657time we do have another alternative to try at the higher level. That is the big 2658difference: in the previous case the remaining alternative is at a deeper 2659recursion level, which PCRE cannot use. 2660</P> 2661<P> 2662To change the pattern so that it matches all palindromic strings, not just 2663those with an odd number of characters, it is tempting to change the pattern to 2664this: 2665<pre> 2666 ^((.)(?1)\2|.?)$ 2667</pre> 2668Again, this works in Perl, but not in PCRE, and for the same reason. When a 2669deeper recursion has matched a single character, it cannot be entered again in 2670order to match an empty string. The solution is to separate the two cases, and 2671write out the odd and even cases as alternatives at the higher level: 2672<pre> 2673 ^(?:((.)(?1)\2|)|((.)(?3)\4|.)) 2674</pre> 2675If you want to match typical palindromic phrases, the pattern has to ignore all 2676non-word characters, which can be done like this: 2677<pre> 2678 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$ 2679</pre> 2680If run with the PCRE_CASELESS option, this pattern matches phrases such as "A 2681man, a plan, a canal: Panama!" and it works well in both PCRE and Perl. Note 2682the use of the possessive quantifier *+ to avoid backtracking into sequences of 2683non-word characters. Without this, PCRE takes a great deal longer (ten times or 2684more) to match typical phrases, and Perl takes so long that you think it has 2685gone into a loop. 2686</P> 2687<P> 2688<b>WARNING</b>: The palindrome-matching patterns above work only if the subject 2689string does not start with a palindrome that is shorter than the entire string. 2690For example, although "abcba" is correctly matched, if the subject is "ababa", 2691PCRE finds the palindrome "aba" at the start, then fails at top level because 2692the end of the string does not follow. Once again, it cannot jump back into the 2693recursion to try other alternatives, so the entire match fails. 2694</P> 2695<P> 2696The second way in which PCRE and Perl differ in their recursion processing is 2697in the handling of captured values. In Perl, when a subpattern is called 2698recursively or as a subpattern (see the next section), it has no access to any 2699values that were captured outside the recursion, whereas in PCRE these values 2700can be referenced. Consider this pattern: 2701<pre> 2702 ^(.)(\1|a(?2)) 2703</pre> 2704In PCRE, this pattern matches "bab". The first capturing parentheses match "b", 2705then in the second group, when the back reference \1 fails to match "b", the 2706second alternative matches "a" and then recurses. In the recursion, \1 does 2707now match "b" and so the whole match succeeds. In Perl, the pattern fails to 2708match because inside the recursive call \1 cannot access the externally set 2709value. 2710<a name="subpatternsassubroutines"></a></P> 2711<br><a name="SEC24" href="#TOC1">SUBPATTERNS AS SUBROUTINES</a><br> 2712<P> 2713If the syntax for a recursive subpattern call (either by number or by 2714name) is used outside the parentheses to which it refers, it operates like a 2715subroutine in a programming language. The called subpattern may be defined 2716before or after the reference. A numbered reference can be absolute or 2717relative, as in these examples: 2718<pre> 2719 (...(absolute)...)...(?2)... 2720 (...(relative)...)...(?-1)... 2721 (...(?+1)...(relative)... 2722</pre> 2723An earlier example pointed out that the pattern 2724<pre> 2725 (sens|respons)e and \1ibility 2726</pre> 2727matches "sense and sensibility" and "response and responsibility", but not 2728"sense and responsibility". If instead the pattern 2729<pre> 2730 (sens|respons)e and (?1)ibility 2731</pre> 2732is used, it does match "sense and responsibility" as well as the other two 2733strings. Another example is given in the discussion of DEFINE above. 2734</P> 2735<P> 2736All subroutine calls, whether recursive or not, are always treated as atomic 2737groups. That is, once a subroutine has matched some of the subject string, it 2738is never re-entered, even if it contains untried alternatives and there is a 2739subsequent matching failure. Any capturing parentheses that are set during the 2740subroutine call revert to their previous values afterwards. 2741</P> 2742<P> 2743Processing options such as case-independence are fixed when a subpattern is 2744defined, so if it is used as a subroutine, such options cannot be changed for 2745different calls. For example, consider this pattern: 2746<pre> 2747 (abc)(?i:(?-1)) 2748</pre> 2749It matches "abcabc". It does not match "abcABC" because the change of 2750processing option does not affect the called subpattern. 2751<a name="onigurumasubroutines"></a></P> 2752<br><a name="SEC25" href="#TOC1">ONIGURUMA SUBROUTINE SYNTAX</a><br> 2753<P> 2754For compatibility with Oniguruma, the non-Perl syntax \g followed by a name or 2755a number enclosed either in angle brackets or single quotes, is an alternative 2756syntax for referencing a subpattern as a subroutine, possibly recursively. Here 2757are two of the examples used above, rewritten using this syntax: 2758<pre> 2759 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) ) 2760 (sens|respons)e and \g'1'ibility 2761</pre> 2762PCRE supports an extension to Oniguruma: if a number is preceded by a 2763plus or a minus sign it is taken as a relative reference. For example: 2764<pre> 2765 (abc)(?i:\g<-1>) 2766</pre> 2767Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are <i>not</i> 2768synonymous. The former is a back reference; the latter is a subroutine call. 2769</P> 2770<br><a name="SEC26" href="#TOC1">CALLOUTS</a><br> 2771<P> 2772Perl has a feature whereby using the sequence (?{...}) causes arbitrary Perl 2773code to be obeyed in the middle of matching a regular expression. This makes it 2774possible, amongst other things, to extract different substrings that match the 2775same pair of parentheses when there is a repetition. 2776</P> 2777<P> 2778PCRE provides a similar feature, but of course it cannot obey arbitrary Perl 2779code. The feature is called "callout". The caller of PCRE provides an external 2780function by putting its entry point in the global variable <i>pcre_callout</i> 2781(8-bit library) or <i>pcre[16|32]_callout</i> (16-bit or 32-bit library). 2782By default, this variable contains NULL, which disables all calling out. 2783</P> 2784<P> 2785Within a regular expression, (?C) indicates the points at which the external 2786function is to be called. If you want to identify different callout points, you 2787can put a number less than 256 after the letter C. The default value is zero. 2788For example, this pattern has two callout points: 2789<pre> 2790 (?C1)abc(?C2)def 2791</pre> 2792If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, callouts are 2793automatically installed before each item in the pattern. They are all numbered 2794255. If there is a conditional group in the pattern whose condition is an 2795assertion, an additional callout is inserted just before the condition. An 2796explicit callout may also be set at this position, as in this example: 2797<pre> 2798 (?(?C9)(?=a)abc|def) 2799</pre> 2800Note that this applies only to assertion conditions, not to other types of 2801condition. 2802</P> 2803<P> 2804During matching, when PCRE reaches a callout point, the external function is 2805called. It is provided with the number of the callout, the position in the 2806pattern, and, optionally, one item of data originally supplied by the caller of 2807the matching function. The callout function may cause matching to proceed, to 2808backtrack, or to fail altogether. 2809</P> 2810<P> 2811By default, PCRE implements a number of optimizations at compile time and 2812matching time, and one side-effect is that sometimes callouts are skipped. If 2813you need all possible callouts to happen, you need to set options that disable 2814the relevant optimizations. More details, and a complete description of the 2815interface to the callout function, are given in the 2816<a href="pcrecallout.html"><b>pcrecallout</b></a> 2817documentation. 2818<a name="backtrackcontrol"></a></P> 2819<br><a name="SEC27" href="#TOC1">BACKTRACKING CONTROL</a><br> 2820<P> 2821Perl 5.10 introduced a number of "Special Backtracking Control Verbs", which 2822are still described in the Perl documentation as "experimental and subject to 2823change or removal in a future version of Perl". It goes on to say: "Their usage 2824in production code should be noted to avoid problems during upgrades." The same 2825remarks apply to the PCRE features described in this section. 2826</P> 2827<P> 2828The new verbs make use of what was previously invalid syntax: an opening 2829parenthesis followed by an asterisk. They are generally of the form 2830(*VERB) or (*VERB:NAME). Some may take either form, possibly behaving 2831differently depending on whether or not a name is present. A name is any 2832sequence of characters that does not include a closing parenthesis. The maximum 2833length of name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit 2834libraries. If the name is empty, that is, if the closing parenthesis 2835immediately follows the colon, the effect is as if the colon were not there. 2836Any number of these verbs may occur in a pattern. 2837</P> 2838<P> 2839Since these verbs are specifically related to backtracking, most of them can be 2840used only when the pattern is to be matched using one of the traditional 2841matching functions, because these use a backtracking algorithm. With the 2842exception of (*FAIL), which behaves like a failing negative assertion, the 2843backtracking control verbs cause an error if encountered by a DFA matching 2844function. 2845</P> 2846<P> 2847The behaviour of these verbs in 2848<a href="#btrepeat">repeated groups,</a> 2849<a href="#btassert">assertions,</a> 2850and in 2851<a href="#btsub">subpatterns called as subroutines</a> 2852(whether or not recursively) is documented below. 2853<a name="nooptimize"></a></P> 2854<br><b> 2855Optimizations that affect backtracking verbs 2856</b><br> 2857<P> 2858PCRE contains some optimizations that are used to speed up matching by running 2859some checks at the start of each match attempt. For example, it may know the 2860minimum length of matching subject, or that a particular character must be 2861present. When one of these optimizations bypasses the running of a match, any 2862included backtracking verbs will not, of course, be processed. You can suppress 2863the start-of-match optimizations by setting the PCRE_NO_START_OPTIMIZE option 2864when calling <b>pcre_compile()</b> or <b>pcre_exec()</b>, or by starting the 2865pattern with (*NO_START_OPT). There is more discussion of this option in the 2866section entitled 2867<a href="pcreapi.html#execoptions">"Option bits for <b>pcre_exec()</b>"</a> 2868in the 2869<a href="pcreapi.html"><b>pcreapi</b></a> 2870documentation. 2871</P> 2872<P> 2873Experiments with Perl suggest that it too has similar optimizations, sometimes 2874leading to anomalous results. 2875</P> 2876<br><b> 2877Verbs that act immediately 2878</b><br> 2879<P> 2880The following verbs act as soon as they are encountered. They may not be 2881followed by a name. 2882<pre> 2883 (*ACCEPT) 2884</pre> 2885This verb causes the match to end successfully, skipping the remainder of the 2886pattern. However, when it is inside a subpattern that is called as a 2887subroutine, only that subpattern is ended successfully. Matching then continues 2888at the outer level. If (*ACCEPT) in triggered in a positive assertion, the 2889assertion succeeds; in a negative assertion, the assertion fails. 2890</P> 2891<P> 2892If (*ACCEPT) is inside capturing parentheses, the data so far is captured. For 2893example: 2894<pre> 2895 A((?:A|B(*ACCEPT)|C)D) 2896</pre> 2897This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is captured by 2898the outer parentheses. 2899<pre> 2900 (*FAIL) or (*F) 2901</pre> 2902This verb causes a matching failure, forcing backtracking to occur. It is 2903equivalent to (?!) but easier to read. The Perl documentation notes that it is 2904probably useful only when combined with (?{}) or (??{}). Those are, of course, 2905Perl features that are not present in PCRE. The nearest equivalent is the 2906callout feature, as for example in this pattern: 2907<pre> 2908 a+(?C)(*FAIL) 2909</pre> 2910A match with the string "aaaa" always fails, but the callout is taken before 2911each backtrack happens (in this example, 10 times). 2912</P> 2913<br><b> 2914Recording which path was taken 2915</b><br> 2916<P> 2917There is one verb whose main purpose is to track how a match was arrived at, 2918though it also has a secondary use in conjunction with advancing the match 2919starting point (see (*SKIP) below). 2920<pre> 2921 (*MARK:NAME) or (*:NAME) 2922</pre> 2923A name is always required with this verb. There may be as many instances of 2924(*MARK) as you like in a pattern, and their names do not have to be unique. 2925</P> 2926<P> 2927When a match succeeds, the name of the last-encountered (*MARK:NAME), 2928(*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the 2929caller as described in the section entitled 2930<a href="pcreapi.html#extradata">"Extra data for <b>pcre_exec()</b>"</a> 2931in the 2932<a href="pcreapi.html"><b>pcreapi</b></a> 2933documentation. Here is an example of <b>pcretest</b> output, where the /K 2934modifier requests the retrieval and outputting of (*MARK) data: 2935<pre> 2936 re> /X(*MARK:A)Y|X(*MARK:B)Z/K 2937 data> XY 2938 0: XY 2939 MK: A 2940 XZ 2941 0: XZ 2942 MK: B 2943</pre> 2944The (*MARK) name is tagged with "MK:" in this output, and in this example it 2945indicates which of the two alternatives matched. This is a more efficient way 2946of obtaining this information than putting each alternative in its own 2947capturing parentheses. 2948</P> 2949<P> 2950If a verb with a name is encountered in a positive assertion that is true, the 2951name is recorded and passed back if it is the last-encountered. This does not 2952happen for negative assertions or failing positive assertions. 2953</P> 2954<P> 2955After a partial match or a failed match, the last encountered name in the 2956entire match process is returned. For example: 2957<pre> 2958 re> /X(*MARK:A)Y|X(*MARK:B)Z/K 2959 data> XP 2960 No match, mark = B 2961</pre> 2962Note that in this unanchored example the mark is retained from the match 2963attempt that started at the letter "X" in the subject. Subsequent match 2964attempts starting at "P" and then with an empty string do not get as far as the 2965(*MARK) item, but nevertheless do not reset it. 2966</P> 2967<P> 2968If you are interested in (*MARK) values after failed matches, you should 2969probably set the PCRE_NO_START_OPTIMIZE option 2970<a href="#nooptimize">(see above)</a> 2971to ensure that the match is always attempted. 2972</P> 2973<br><b> 2974Verbs that act after backtracking 2975</b><br> 2976<P> 2977The following verbs do nothing when they are encountered. Matching continues 2978with what follows, but if there is no subsequent match, causing a backtrack to 2979the verb, a failure is forced. That is, backtracking cannot pass to the left of 2980the verb. However, when one of these verbs appears inside an atomic group or an 2981assertion that is true, its effect is confined to that group, because once the 2982group has been matched, there is never any backtracking into it. In this 2983situation, backtracking can "jump back" to the left of the entire atomic group 2984or assertion. (Remember also, as stated above, that this localization also 2985applies in subroutine calls.) 2986</P> 2987<P> 2988These verbs differ in exactly what kind of failure occurs when backtracking 2989reaches them. The behaviour described below is what happens when the verb is 2990not in a subroutine or an assertion. Subsequent sections cover these special 2991cases. 2992<pre> 2993 (*COMMIT) 2994</pre> 2995This verb, which may not be followed by a name, causes the whole match to fail 2996outright if there is a later matching failure that causes backtracking to reach 2997it. Even if the pattern is unanchored, no further attempts to find a match by 2998advancing the starting point take place. If (*COMMIT) is the only backtracking 2999verb that is encountered, once it has been passed <b>pcre_exec()</b> is 3000committed to finding a match at the current starting point, or not at all. For 3001example: 3002<pre> 3003 a+(*COMMIT)b 3004</pre> 3005This matches "xxaab" but not "aacaab". It can be thought of as a kind of 3006dynamic anchor, or "I've started, so I must finish." The name of the most 3007recently passed (*MARK) in the path is passed back when (*COMMIT) forces a 3008match failure. 3009</P> 3010<P> 3011If there is more than one backtracking verb in a pattern, a different one that 3012follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a 3013match does not always guarantee that a match must be at this starting point. 3014</P> 3015<P> 3016Note that (*COMMIT) at the start of a pattern is not the same as an anchor, 3017unless PCRE's start-of-match optimizations are turned off, as shown in this 3018output from <b>pcretest</b>: 3019<pre> 3020 re> /(*COMMIT)abc/ 3021 data> xyzabc 3022 0: abc 3023 data> xyzabc\Y 3024 No match 3025</pre> 3026For this pattern, PCRE knows that any match must start with "a", so the 3027optimization skips along the subject to "a" before applying the pattern to the 3028first set of data. The match attempt then succeeds. In the second set of data, 3029the escape sequence \Y is interpreted by the <b>pcretest</b> program. It causes 3030the PCRE_NO_START_OPTIMIZE option to be set when <b>pcre_exec()</b> is called. 3031This disables the optimization that skips along to the first character. The 3032pattern is now applied starting at "x", and so the (*COMMIT) causes the match 3033to fail without trying any other starting points. 3034<pre> 3035 (*PRUNE) or (*PRUNE:NAME) 3036</pre> 3037This verb causes the match to fail at the current starting position in the 3038subject if there is a later matching failure that causes backtracking to reach 3039it. If the pattern is unanchored, the normal "bumpalong" advance to the next 3040starting character then happens. Backtracking can occur as usual to the left of 3041(*PRUNE), before it is reached, or when matching to the right of (*PRUNE), but 3042if there is no match to the right, backtracking cannot cross (*PRUNE). In 3043simple cases, the use of (*PRUNE) is just an alternative to an atomic group or 3044possessive quantifier, but there are some uses of (*PRUNE) that cannot be 3045expressed in any other way. In an anchored pattern (*PRUNE) has the same effect 3046as (*COMMIT). 3047</P> 3048<P> 3049The behaviour of (*PRUNE:NAME) is the not the same as (*MARK:NAME)(*PRUNE). 3050It is like (*MARK:NAME) in that the name is remembered for passing back to the 3051caller. However, (*SKIP:NAME) searches only for names set with (*MARK). 3052<pre> 3053 (*SKIP) 3054</pre> 3055This verb, when given without a name, is like (*PRUNE), except that if the 3056pattern is unanchored, the "bumpalong" advance is not to the next character, 3057but to the position in the subject where (*SKIP) was encountered. (*SKIP) 3058signifies that whatever text was matched leading up to it cannot be part of a 3059successful match. Consider: 3060<pre> 3061 a+(*SKIP)b 3062</pre> 3063If the subject is "aaaac...", after the first match attempt fails (starting at 3064the first character in the string), the starting point skips on to start the 3065next attempt at "c". Note that a possessive quantifer does not have the same 3066effect as this example; although it would suppress backtracking during the 3067first match attempt, the second attempt would start at the second character 3068instead of skipping on to "c". 3069<pre> 3070 (*SKIP:NAME) 3071</pre> 3072When (*SKIP) has an associated name, its behaviour is modified. When it is 3073triggered, the previous path through the pattern is searched for the most 3074recent (*MARK) that has the same name. If one is found, the "bumpalong" advance 3075is to the subject position that corresponds to that (*MARK) instead of to where 3076(*SKIP) was encountered. If no (*MARK) with a matching name is found, the 3077(*SKIP) is ignored. 3078</P> 3079<P> 3080Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores 3081names that are set by (*PRUNE:NAME) or (*THEN:NAME). 3082<pre> 3083 (*THEN) or (*THEN:NAME) 3084</pre> 3085This verb causes a skip to the next innermost alternative when backtracking 3086reaches it. That is, it cancels any further backtracking within the current 3087alternative. Its name comes from the observation that it can be used for a 3088pattern-based if-then-else block: 3089<pre> 3090 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ... 3091</pre> 3092If the COND1 pattern matches, FOO is tried (and possibly further items after 3093the end of the group if FOO succeeds); on failure, the matcher skips to the 3094second alternative and tries COND2, without backtracking into COND1. If that 3095succeeds and BAR fails, COND3 is tried. If subsequently BAZ fails, there are no 3096more alternatives, so there is a backtrack to whatever came before the entire 3097group. If (*THEN) is not inside an alternation, it acts like (*PRUNE). 3098</P> 3099<P> 3100The behaviour of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN). 3101It is like (*MARK:NAME) in that the name is remembered for passing back to the 3102caller. However, (*SKIP:NAME) searches only for names set with (*MARK). 3103</P> 3104<P> 3105A subpattern that does not contain a | character is just a part of the 3106enclosing alternative; it is not a nested alternation with only one 3107alternative. The effect of (*THEN) extends beyond such a subpattern to the 3108enclosing alternative. Consider this pattern, where A, B, etc. are complex 3109pattern fragments that do not contain any | characters at this level: 3110<pre> 3111 A (B(*THEN)C) | D 3112</pre> 3113If A and B are matched, but there is a failure in C, matching does not 3114backtrack into A; instead it moves to the next alternative, that is, D. 3115However, if the subpattern containing (*THEN) is given an alternative, it 3116behaves differently: 3117<pre> 3118 A (B(*THEN)C | (*FAIL)) | D 3119</pre> 3120The effect of (*THEN) is now confined to the inner subpattern. After a failure 3121in C, matching moves to (*FAIL), which causes the whole subpattern to fail 3122because there are no more alternatives to try. In this case, matching does now 3123backtrack into A. 3124</P> 3125<P> 3126Note that a conditional subpattern is not considered as having two 3127alternatives, because only one is ever used. In other words, the | character in 3128a conditional subpattern has a different meaning. Ignoring white space, 3129consider: 3130<pre> 3131 ^.*? (?(?=a) a | b(*THEN)c ) 3132</pre> 3133If the subject is "ba", this pattern does not match. Because .*? is ungreedy, 3134it initially matches zero characters. The condition (?=a) then fails, the 3135character "b" is matched, but "c" is not. At this point, matching does not 3136backtrack to .*? as might perhaps be expected from the presence of the | 3137character. The conditional subpattern is part of the single alternative that 3138comprises the whole pattern, and so the match fails. (If there was a backtrack 3139into .*?, allowing it to match "b", the match would succeed.) 3140</P> 3141<P> 3142The verbs just described provide four different "strengths" of control when 3143subsequent matching fails. (*THEN) is the weakest, carrying on the match at the 3144next alternative. (*PRUNE) comes next, failing the match at the current 3145starting position, but allowing an advance to the next character (for an 3146unanchored pattern). (*SKIP) is similar, except that the advance may be more 3147than one character. (*COMMIT) is the strongest, causing the entire match to 3148fail. 3149</P> 3150<br><b> 3151More than one backtracking verb 3152</b><br> 3153<P> 3154If more than one backtracking verb is present in a pattern, the one that is 3155backtracked onto first acts. For example, consider this pattern, where A, B, 3156etc. are complex pattern fragments: 3157<pre> 3158 (A(*COMMIT)B(*THEN)C|ABD) 3159</pre> 3160If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to 3161fail. However, if A and B match, but C fails, the backtrack to (*THEN) causes 3162the next alternative (ABD) to be tried. This behaviour is consistent, but is 3163not always the same as Perl's. It means that if two or more backtracking verbs 3164appear in succession, all the the last of them has no effect. Consider this 3165example: 3166<pre> 3167 ...(*COMMIT)(*PRUNE)... 3168</pre> 3169If there is a matching failure to the right, backtracking onto (*PRUNE) causes 3170it to be triggered, and its action is taken. There can never be a backtrack 3171onto (*COMMIT). 3172<a name="btrepeat"></a></P> 3173<br><b> 3174Backtracking verbs in repeated groups 3175</b><br> 3176<P> 3177PCRE differs from Perl in its handling of backtracking verbs in repeated 3178groups. For example, consider: 3179<pre> 3180 /(a(*COMMIT)b)+ac/ 3181</pre> 3182If the subject is "abac", Perl matches, but PCRE fails because the (*COMMIT) in 3183the second repeat of the group acts. 3184<a name="btassert"></a></P> 3185<br><b> 3186Backtracking verbs in assertions 3187</b><br> 3188<P> 3189(*FAIL) in an assertion has its normal effect: it forces an immediate backtrack. 3190</P> 3191<P> 3192(*ACCEPT) in a positive assertion causes the assertion to succeed without any 3193further processing. In a negative assertion, (*ACCEPT) causes the assertion to 3194fail without any further processing. 3195</P> 3196<P> 3197The other backtracking verbs are not treated specially if they appear in a 3198positive assertion. In particular, (*THEN) skips to the next alternative in the 3199innermost enclosing group that has alternations, whether or not this is within 3200the assertion. 3201</P> 3202<P> 3203Negative assertions are, however, different, in order to ensure that changing a 3204positive assertion into a negative assertion changes its result. Backtracking 3205into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative assertion to be true, 3206without considering any further alternative branches in the assertion. 3207Backtracking into (*THEN) causes it to skip to the next enclosing alternative 3208within the assertion (the normal behaviour), but if the assertion does not have 3209such an alternative, (*THEN) behaves like (*PRUNE). 3210<a name="btsub"></a></P> 3211<br><b> 3212Backtracking verbs in subroutines 3213</b><br> 3214<P> 3215These behaviours occur whether or not the subpattern is called recursively. 3216Perl's treatment of subroutines is different in some cases. 3217</P> 3218<P> 3219(*FAIL) in a subpattern called as a subroutine has its normal effect: it forces 3220an immediate backtrack. 3221</P> 3222<P> 3223(*ACCEPT) in a subpattern called as a subroutine causes the subroutine match to 3224succeed without any further processing. Matching then continues after the 3225subroutine call. 3226</P> 3227<P> 3228(*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine cause 3229the subroutine match to fail. 3230</P> 3231<P> 3232(*THEN) skips to the next alternative in the innermost enclosing group within 3233the subpattern that has alternatives. If there is no such group within the 3234subpattern, (*THEN) causes the subroutine match to fail. 3235</P> 3236<br><a name="SEC28" href="#TOC1">SEE ALSO</a><br> 3237<P> 3238<b>pcreapi</b>(3), <b>pcrecallout</b>(3), <b>pcrematching</b>(3), 3239<b>pcresyntax</b>(3), <b>pcre</b>(3), <b>pcre16(3)</b>, <b>pcre32(3)</b>. 3240</P> 3241<br><a name="SEC29" href="#TOC1">AUTHOR</a><br> 3242<P> 3243Philip Hazel 3244<br> 3245University Computing Service 3246<br> 3247Cambridge CB2 3QH, England. 3248<br> 3249</P> 3250<br><a name="SEC30" href="#TOC1">REVISION</a><br> 3251<P> 3252Last updated: 08 January 2014 3253<br> 3254Copyright © 1997-2014 University of Cambridge. 3255<br> 3256<p> 3257Return to the <a href="index.html">PCRE index page</a>. 3258</p> 3259