| 1 | This file contains the PCRE man page that describes the regular expressions |
| 2 | supported by PCRE version 7.2. Note that not all of the features are relevant |
| 3 | in the context of Exim. In particular, the version of PCRE that is compiled |
| 4 | with Exim does not include UTF-8 support, there is no mechanism for changing |
| 5 | the options with which the PCRE functions are called, and features such as |
| 6 | callout are not accessible. |
| 7 | ----------------------------------------------------------------------------- |
| 8 | |
| 9 | PCREPATTERN(3) PCREPATTERN(3) |
| 10 | |
| 11 | |
| 12 | NAME |
| 13 | PCRE - Perl-compatible regular expressions |
| 14 | |
| 15 | |
| 16 | PCRE REGULAR EXPRESSION DETAILS |
| 17 | |
| 18 | The syntax and semantics of the regular expressions supported by PCRE |
| 19 | are described below. Regular expressions are also described in the Perl |
| 20 | documentation and in a number of books, some of which have copious |
| 21 | examples. Jeffrey Friedl's "Mastering Regular Expressions", published |
| 22 | by O'Reilly, covers regular expressions in great detail. This descrip- |
| 23 | tion of PCRE's regular expressions is intended as reference material. |
| 24 | |
| 25 | The original operation of PCRE was on strings of one-byte characters. |
| 26 | However, there is now also support for UTF-8 character strings. To use |
| 27 | this, you must build PCRE to include UTF-8 support, and then call |
| 28 | pcre_compile() with the PCRE_UTF8 option. How this affects pattern |
| 29 | matching is mentioned in several places below. There is also a summary |
| 30 | of UTF-8 features in the section on UTF-8 support in the main pcre |
| 31 | page. |
| 32 | |
| 33 | The remainder of this document discusses the patterns that are sup- |
| 34 | ported by PCRE when its main matching function, pcre_exec(), is used. |
| 35 | From release 6.0, PCRE offers a second matching function, |
| 36 | pcre_dfa_exec(), which matches using a different algorithm that is not |
| 37 | Perl-compatible. Some of the features discussed below are not available |
| 38 | when pcre_dfa_exec() is used. The advantages and disadvantages of the |
| 39 | alternative function, and how it differs from the normal function, are |
| 40 | discussed in the pcrematching page. |
| 41 | |
| 42 | |
| 43 | CHARACTERS AND METACHARACTERS |
| 44 | |
| 45 | A regular expression is a pattern that is matched against a subject |
| 46 | string from left to right. Most characters stand for themselves in a |
| 47 | pattern, and match the corresponding characters in the subject. As a |
| 48 | trivial example, the pattern |
| 49 | |
| 50 | The quick brown fox |
| 51 | |
| 52 | matches a portion of a subject string that is identical to itself. When |
| 53 | caseless matching is specified (the PCRE_CASELESS option), letters are |
| 54 | matched independently of case. In UTF-8 mode, PCRE always understands |
| 55 | the concept of case for characters whose values are less than 128, so |
| 56 | caseless matching is always possible. For characters with higher val- |
| 57 | ues, the concept of case is supported if PCRE is compiled with Unicode |
| 58 | property support, but not otherwise. If you want to use caseless |
| 59 | matching for characters 128 and above, you must ensure that PCRE is |
| 60 | compiled with Unicode property support as well as with UTF-8 support. |
| 61 | |
| 62 | The power of regular expressions comes from the ability to include |
| 63 | alternatives and repetitions in the pattern. These are encoded in the |
| 64 | pattern by the use of metacharacters, which do not stand for themselves |
| 65 | but instead are interpreted in some special way. |
| 66 | |
| 67 | There are two different sets of metacharacters: those that are recog- |
| 68 | nized anywhere in the pattern except within square brackets, and those |
| 69 | that are recognized within square brackets. Outside square brackets, |
| 70 | the metacharacters are as follows: |
| 71 | |
| 72 | \ general escape character with several uses |
| 73 | ^ assert start of string (or line, in multiline mode) |
| 74 | $ assert end of string (or line, in multiline mode) |
| 75 | . match any character except newline (by default) |
| 76 | [ start character class definition |
| 77 | | start of alternative branch |
| 78 | ( start subpattern |
| 79 | ) end subpattern |
| 80 | ? extends the meaning of ( |
| 81 | also 0 or 1 quantifier |
| 82 | also quantifier minimizer |
| 83 | * 0 or more quantifier |
| 84 | + 1 or more quantifier |
| 85 | also "possessive quantifier" |
| 86 | { start min/max quantifier |
| 87 | |
| 88 | Part of a pattern that is in square brackets is called a "character |
| 89 | class". In a character class the only metacharacters are: |
| 90 | |
| 91 | \ general escape character |
| 92 | ^ negate the class, but only if the first character |
| 93 | - indicates character range |
| 94 | [ POSIX character class (only if followed by POSIX |
| 95 | syntax) |
| 96 | ] terminates the character class |
| 97 | |
| 98 | The following sections describe the use of each of the metacharacters. |
| 99 | |
| 100 | |
| 101 | BACKSLASH |
| 102 | |
| 103 | The backslash character has several uses. Firstly, if it is followed by |
| 104 | a non-alphanumeric character, it takes away any special meaning that |
| 105 | character may have. This use of backslash as an escape character |
| 106 | applies both inside and outside character classes. |
| 107 | |
| 108 | For example, if you want to match a * character, you write \* in the |
| 109 | pattern. This escaping action applies whether or not the following |
| 110 | character would otherwise be interpreted as a metacharacter, so it is |
| 111 | always safe to precede a non-alphanumeric with backslash to specify |
| 112 | that it stands for itself. In particular, if you want to match a back- |
| 113 | slash, you write \\. |
| 114 | |
| 115 | If a pattern is compiled with the PCRE_EXTENDED option, whitespace in |
| 116 | the pattern (other than in a character class) and characters between a |
| 117 | # outside a character class and the next newline are ignored. An escap- |
| 118 | ing backslash can be used to include a whitespace or # character as |
| 119 | part of the pattern. |
| 120 | |
| 121 | If you want to remove the special meaning from a sequence of charac- |
| 122 | ters, you can do so by putting them between \Q and \E. This is differ- |
| 123 | ent from Perl in that $ and @ are handled as literals in \Q...\E |
| 124 | sequences in PCRE, whereas in Perl, $ and @ cause variable interpola- |
| 125 | tion. Note the following examples: |
| 126 | |
| 127 | Pattern PCRE matches Perl matches |
| 128 | |
| 129 | \Qabc$xyz\E abc$xyz abc followed by the |
| 130 | contents of $xyz |
| 131 | \Qabc\$xyz\E abc\$xyz abc\$xyz |
| 132 | \Qabc\E\$\Qxyz\E abc$xyz abc$xyz |
| 133 | |
| 134 | The \Q...\E sequence is recognized both inside and outside character |
| 135 | classes. |
| 136 | |
| 137 | Non-printing characters |
| 138 | |
| 139 | A second use of backslash provides a way of encoding non-printing char- |
| 140 | acters in patterns in a visible manner. There is no restriction on the |
| 141 | appearance of non-printing characters, apart from the binary zero that |
| 142 | terminates a pattern, but when a pattern is being prepared by text |
| 143 | editing, it is usually easier to use one of the following escape |
| 144 | sequences than the binary character it represents: |
| 145 | |
| 146 | \a alarm, that is, the BEL character (hex 07) |
| 147 | \cx "control-x", where x is any character |
| 148 | \e escape (hex 1B) |
| 149 | \f formfeed (hex 0C) |
| 150 | \n newline (hex 0A) |
| 151 | \r carriage return (hex 0D) |
| 152 | \t tab (hex 09) |
| 153 | \ddd character with octal code ddd, or backreference |
| 154 | \xhh character with hex code hh |
| 155 | \x{hhh..} character with hex code hhh.. |
| 156 | |
| 157 | The precise effect of \cx is as follows: if x is a lower case letter, |
| 158 | it is converted to upper case. Then bit 6 of the character (hex 40) is |
| 159 | inverted. Thus \cz becomes hex 1A, but \c{ becomes hex 3B, while \c; |
| 160 | becomes hex 7B. |
| 161 | |
| 162 | After \x, from zero to two hexadecimal digits are read (letters can be |
| 163 | in upper or lower case). Any number of hexadecimal digits may appear |
| 164 | between \x{ and }, but the value of the character code must be less |
| 165 | than 256 in non-UTF-8 mode, and less than 2**31 in UTF-8 mode (that is, |
| 166 | the maximum hexadecimal value is 7FFFFFFF). If characters other than |
| 167 | hexadecimal digits appear between \x{ and }, or if there is no termi- |
| 168 | nating }, this form of escape is not recognized. Instead, the initial |
| 169 | \x will be interpreted as a basic hexadecimal escape, with no following |
| 170 | digits, giving a character whose value is zero. |
| 171 | |
| 172 | Characters whose value is less than 256 can be defined by either of the |
| 173 | two syntaxes for \x. There is no difference in the way they are han- |
| 174 | dled. For example, \xdc is exactly the same as \x{dc}. |
| 175 | |
| 176 | After \0 up to two further octal digits are read. If there are fewer |
| 177 | than two digits, just those that are present are used. Thus the |
| 178 | sequence \0\x\07 specifies two binary zeros followed by a BEL character |
| 179 | (code value 7). Make sure you supply two digits after the initial zero |
| 180 | if the pattern character that follows is itself an octal digit. |
| 181 | |
| 182 | The handling of a backslash followed by a digit other than 0 is compli- |
| 183 | cated. Outside a character class, PCRE reads it and any following dig- |
| 184 | its as a decimal number. If the number is less than 10, or if there |
| 185 | have been at least that many previous capturing left parentheses in the |
| 186 | expression, the entire sequence is taken as a back reference. A |
| 187 | description of how this works is given later, following the discussion |
| 188 | of parenthesized subpatterns. |
| 189 | |
| 190 | Inside a character class, or if the decimal number is greater than 9 |
| 191 | and there have not been that many capturing subpatterns, PCRE re-reads |
| 192 | up to three octal digits following the backslash, and uses them to gen- |
| 193 | erate a data character. Any subsequent digits stand for themselves. In |
| 194 | non-UTF-8 mode, the value of a character specified in octal must be |
| 195 | less than \400. In UTF-8 mode, values up to \777 are permitted. For |
| 196 | example: |
| 197 | |
| 198 | \040 is another way of writing a space |
| 199 | \40 is the same, provided there are fewer than 40 |
| 200 | previous capturing subpatterns |
| 201 | \7 is always a back reference |
| 202 | \11 might be a back reference, or another way of |
| 203 | writing a tab |
| 204 | \011 is always a tab |
| 205 | \0113 is a tab followed by the character "3" |
| 206 | \113 might be a back reference, otherwise the |
| 207 | character with octal code 113 |
| 208 | \377 might be a back reference, otherwise |
| 209 | the byte consisting entirely of 1 bits |
| 210 | \81 is either a back reference, or a binary zero |
| 211 | followed by the two characters "8" and "1" |
| 212 | |
| 213 | Note that octal values of 100 or greater must not be introduced by a |
| 214 | leading zero, because no more than three octal digits are ever read. |
| 215 | |
| 216 | All the sequences that define a single character value can be used both |
| 217 | inside and outside character classes. In addition, inside a character |
| 218 | class, the sequence \b is interpreted as the backspace character (hex |
| 219 | 08), and the sequences \R and \X are interpreted as the characters "R" |
| 220 | and "X", respectively. Outside a character class, these sequences have |
| 221 | different meanings (see below). |
| 222 | |
| 223 | Absolute and relative back references |
| 224 | |
| 225 | The sequence \g followed by a positive or negative number, optionally |
| 226 | enclosed in braces, is an absolute or relative back reference. A named |
| 227 | back reference can be coded as \g{name}. Back references are discussed |
| 228 | later, following the discussion of parenthesized subpatterns. |
| 229 | |
| 230 | Generic character types |
| 231 | |
| 232 | Another use of backslash is for specifying generic character types. The |
| 233 | following are always recognized: |
| 234 | |
| 235 | \d any decimal digit |
| 236 | \D any character that is not a decimal digit |
| 237 | \h any horizontal whitespace character |
| 238 | \H any character that is not a horizontal whitespace character |
| 239 | \s any whitespace character |
| 240 | \S any character that is not a whitespace character |
| 241 | \v any vertical whitespace character |
| 242 | \V any character that is not a vertical whitespace character |
| 243 | \w any "word" character |
| 244 | \W any "non-word" character |
| 245 | |
| 246 | Each pair of escape sequences partitions the complete set of characters |
| 247 | into two disjoint sets. Any given character matches one, and only one, |
| 248 | of each pair. |
| 249 | |
| 250 | These character type sequences can appear both inside and outside char- |
| 251 | acter classes. They each match one character of the appropriate type. |
| 252 | If the current matching point is at the end of the subject string, all |
| 253 | of them fail, since there is no character to match. |
| 254 | |
| 255 | For compatibility with Perl, \s does not match the VT character (code |
| 256 | 11). This makes it different from the the POSIX "space" class. The \s |
| 257 | characters are HT (9), LF (10), FF (12), CR (13), and space (32). If |
| 258 | "use locale;" is included in a Perl script, \s may match the VT charac- |
| 259 | ter. In PCRE, it never does. |
| 260 | |
| 261 | In UTF-8 mode, characters with values greater than 128 never match \d, |
| 262 | \s, or \w, and always match \D, \S, and \W. This is true even when Uni- |
| 263 | code character property support is available. These sequences retain |
| 264 | their original meanings from before UTF-8 support was available, mainly |
| 265 | for efficiency reasons. |
| 266 | |
| 267 | The sequences \h, \H, \v, and \V are Perl 5.10 features. In contrast to |
| 268 | the other sequences, these do match certain high-valued codepoints in |
| 269 | UTF-8 mode. The horizontal space characters are: |
| 270 | |
| 271 | U+0009 Horizontal tab |
| 272 | U+0020 Space |
| 273 | U+00A0 Non-break space |
| 274 | U+1680 Ogham space mark |
| 275 | U+180E Mongolian vowel separator |
| 276 | U+2000 En quad |
| 277 | U+2001 Em quad |
| 278 | U+2002 En space |
| 279 | U+2003 Em space |
| 280 | U+2004 Three-per-em space |
| 281 | U+2005 Four-per-em space |
| 282 | U+2006 Six-per-em space |
| 283 | U+2007 Figure space |
| 284 | U+2008 Punctuation space |
| 285 | U+2009 Thin space |
| 286 | U+200A Hair space |
| 287 | U+202F Narrow no-break space |
| 288 | U+205F Medium mathematical space |
| 289 | U+3000 Ideographic space |
| 290 | |
| 291 | The vertical space characters are: |
| 292 | |
| 293 | U+000A Linefeed |
| 294 | U+000B Vertical tab |
| 295 | U+000C Formfeed |
| 296 | U+000D Carriage return |
| 297 | U+0085 Next line |
| 298 | U+2028 Line separator |
| 299 | U+2029 Paragraph separator |
| 300 | |
| 301 | A "word" character is an underscore or any character less than 256 that |
| 302 | is a letter or digit. The definition of letters and digits is con- |
| 303 | trolled by PCRE's low-valued character tables, and may vary if locale- |
| 304 | specific matching is taking place (see "Locale support" in the pcreapi |
| 305 | page). For example, in a French locale such as "fr_FR" in Unix-like |
| 306 | systems, or "french" in Windows, some character codes greater than 128 |
| 307 | are used for accented letters, and these are matched by \w. The use of |
| 308 | locales with Unicode is discouraged. |
| 309 | |
| 310 | Newline sequences |
| 311 | |
| 312 | Outside a character class, the escape sequence \R matches any Unicode |
| 313 | newline sequence. This is a Perl 5.10 feature. In non-UTF-8 mode \R is |
| 314 | equivalent to the following: |
| 315 | |
| 316 | (?>\r\n|\n|\x0b|\f|\r|\x85) |
| 317 | |
| 318 | This is an example of an "atomic group", details of which are given |
| 319 | below. This particular group matches either the two-character sequence |
| 320 | CR followed by LF, or one of the single characters LF (linefeed, |
| 321 | U+000A), VT (vertical tab, U+000B), FF (formfeed, U+000C), CR (carriage |
| 322 | return, U+000D), or NEL (next line, U+0085). The two-character sequence |
| 323 | is treated as a single unit that cannot be split. |
| 324 | |
| 325 | In UTF-8 mode, two additional characters whose codepoints are greater |
| 326 | than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa- |
| 327 | rator, U+2029). Unicode character property support is not needed for |
| 328 | these characters to be recognized. |
| 329 | |
| 330 | Inside a character class, \R matches the letter "R". |
| 331 | |
| 332 | Unicode character properties |
| 333 | |
| 334 | When PCRE is built with Unicode character property support, three addi- |
| 335 | tional escape sequences that match characters with specific properties |
| 336 | are available. When not in UTF-8 mode, these sequences are of course |
| 337 | limited to testing characters whose codepoints are less than 256, but |
| 338 | they do work in this mode. The extra escape sequences are: |
| 339 | |
| 340 | \p{xx} a character with the xx property |
| 341 | \P{xx} a character without the xx property |
| 342 | \X an extended Unicode sequence |
| 343 | |
| 344 | The property names represented by xx above are limited to the Unicode |
| 345 | script names, the general category properties, and "Any", which matches |
| 346 | any character (including newline). Other properties such as "InMusical- |
| 347 | Symbols" are not currently supported by PCRE. Note that \P{Any} does |
| 348 | not match any characters, so always causes a match failure. |
| 349 | |
| 350 | Sets of Unicode characters are defined as belonging to certain scripts. |
| 351 | A character from one of these sets can be matched using a script name. |
| 352 | For example: |
| 353 | |
| 354 | \p{Greek} |
| 355 | \P{Han} |
| 356 | |
| 357 | Those that are not part of an identified script are lumped together as |
| 358 | "Common". The current list of scripts is: |
| 359 | |
| 360 | Arabic, Armenian, Balinese, Bengali, Bopomofo, Braille, Buginese, |
| 361 | Buhid, Canadian_Aboriginal, Cherokee, Common, Coptic, Cuneiform, |
| 362 | Cypriot, Cyrillic, Deseret, Devanagari, Ethiopic, Georgian, Glagolitic, |
| 363 | Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hira- |
| 364 | gana, Inherited, Kannada, Katakana, Kharoshthi, Khmer, Lao, Latin, |
| 365 | Limbu, Linear_B, Malayalam, Mongolian, Myanmar, New_Tai_Lue, Nko, |
| 366 | Ogham, Old_Italic, Old_Persian, Oriya, Osmanya, Phags_Pa, Phoenician, |
| 367 | Runic, Shavian, Sinhala, Syloti_Nagri, Syriac, Tagalog, Tagbanwa, |
| 368 | Tai_Le, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, Ugaritic, Yi. |
| 369 | |
| 370 | Each character has exactly one general category property, specified by |
| 371 | a two-letter abbreviation. For compatibility with Perl, negation can be |
| 372 | specified by including a circumflex between the opening brace and the |
| 373 | property name. For example, \p{^Lu} is the same as \P{Lu}. |
| 374 | |
| 375 | If only one letter is specified with \p or \P, it includes all the gen- |
| 376 | eral category properties that start with that letter. In this case, in |
| 377 | the absence of negation, the curly brackets in the escape sequence are |
| 378 | optional; these two examples have the same effect: |
| 379 | |
| 380 | \p{L} |
| 381 | \pL |
| 382 | |
| 383 | The following general category property codes are supported: |
| 384 | |
| 385 | C Other |
| 386 | Cc Control |
| 387 | Cf Format |
| 388 | Cn Unassigned |
| 389 | Co Private use |
| 390 | Cs Surrogate |
| 391 | |
| 392 | L Letter |
| 393 | Ll Lower case letter |
| 394 | Lm Modifier letter |
| 395 | Lo Other letter |
| 396 | Lt Title case letter |
| 397 | Lu Upper case letter |
| 398 | |
| 399 | M Mark |
| 400 | Mc Spacing mark |
| 401 | Me Enclosing mark |
| 402 | Mn Non-spacing mark |
| 403 | |
| 404 | N Number |
| 405 | Nd Decimal number |
| 406 | Nl Letter number |
| 407 | No Other number |
| 408 | |
| 409 | P Punctuation |
| 410 | Pc Connector punctuation |
| 411 | Pd Dash punctuation |
| 412 | Pe Close punctuation |
| 413 | Pf Final punctuation |
| 414 | Pi Initial punctuation |
| 415 | Po Other punctuation |
| 416 | Ps Open punctuation |
| 417 | |
| 418 | S Symbol |
| 419 | Sc Currency symbol |
| 420 | Sk Modifier symbol |
| 421 | Sm Mathematical symbol |
| 422 | So Other symbol |
| 423 | |
| 424 | Z Separator |
| 425 | Zl Line separator |
| 426 | Zp Paragraph separator |
| 427 | Zs Space separator |
| 428 | |
| 429 | The special property L& is also supported: it matches a character that |
| 430 | has the Lu, Ll, or Lt property, in other words, a letter that is not |
| 431 | classified as a modifier or "other". |
| 432 | |
| 433 | The long synonyms for these properties that Perl supports (such as |
| 434 | \p{Letter}) are not supported by PCRE, nor is it permitted to prefix |
| 435 | any of these properties with "Is". |
| 436 | |
| 437 | No character that is in the Unicode table has the Cn (unassigned) prop- |
| 438 | erty. Instead, this property is assumed for any code point that is not |
| 439 | in the Unicode table. |
| 440 | |
| 441 | Specifying caseless matching does not affect these escape sequences. |
| 442 | For example, \p{Lu} always matches only upper case letters. |
| 443 | |
| 444 | The \X escape matches any number of Unicode characters that form an |
| 445 | extended Unicode sequence. \X is equivalent to |
| 446 | |
| 447 | (?>\PM\pM*) |
| 448 | |
| 449 | That is, it matches a character without the "mark" property, followed |
| 450 | by zero or more characters with the "mark" property, and treats the |
| 451 | sequence as an atomic group (see below). Characters with the "mark" |
| 452 | property are typically accents that affect the preceding character. |
| 453 | None of them have codepoints less than 256, so in non-UTF-8 mode \X |
| 454 | matches any one character. |
| 455 | |
| 456 | Matching characters by Unicode property is not fast, because PCRE has |
| 457 | to search a structure that contains data for over fifteen thousand |
| 458 | characters. That is why the traditional escape sequences such as \d and |
| 459 | \w do not use Unicode properties in PCRE. |
| 460 | |
| 461 | Resetting the match start |
| 462 | |
| 463 | The escape sequence \K, which is a Perl 5.10 feature, causes any previ- |
| 464 | ously matched characters not to be included in the final matched |
| 465 | sequence. For example, the pattern: |
| 466 | |
| 467 | foo\Kbar |
| 468 | |
| 469 | matches "foobar", but reports that it has matched "bar". This feature |
| 470 | is similar to a lookbehind assertion (described below). However, in |
| 471 | this case, the part of the subject before the real match does not have |
| 472 | to be of fixed length, as lookbehind assertions do. The use of \K does |
| 473 | not interfere with the setting of captured substrings. For example, |
| 474 | when the pattern |
| 475 | |
| 476 | (foo)\Kbar |
| 477 | |
| 478 | matches "foobar", the first substring is still set to "foo". |
| 479 | |
| 480 | Simple assertions |
| 481 | |
| 482 | The final use of backslash is for certain simple assertions. An asser- |
| 483 | tion specifies a condition that has to be met at a particular point in |
| 484 | a match, without consuming any characters from the subject string. The |
| 485 | use of subpatterns for more complicated assertions is described below. |
| 486 | The backslashed assertions are: |
| 487 | |
| 488 | \b matches at a word boundary |
| 489 | \B matches when not at a word boundary |
| 490 | \A matches at the start of the subject |
| 491 | \Z matches at the end of the subject |
| 492 | also matches before a newline at the end of the subject |
| 493 | \z matches only at the end of the subject |
| 494 | \G matches at the first matching position in the subject |
| 495 | |
| 496 | These assertions may not appear in character classes (but note that \b |
| 497 | has a different meaning, namely the backspace character, inside a char- |
| 498 | acter class). |
| 499 | |
| 500 | A word boundary is a position in the subject string where the current |
| 501 | character and the previous character do not both match \w or \W (i.e. |
| 502 | one matches \w and the other matches \W), or the start or end of the |
| 503 | string if the first or last character matches \w, respectively. |
| 504 | |
| 505 | The \A, \Z, and \z assertions differ from the traditional circumflex |
| 506 | and dollar (described in the next section) in that they only ever match |
| 507 | at the very start and end of the subject string, whatever options are |
| 508 | set. Thus, they are independent of multiline mode. These three asser- |
| 509 | tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which |
| 510 | affect only the behaviour of the circumflex and dollar metacharacters. |
| 511 | However, if the startoffset argument of pcre_exec() is non-zero, indi- |
| 512 | cating that matching is to start at a point other than the beginning of |
| 513 | the subject, \A can never match. The difference between \Z and \z is |
| 514 | that \Z matches before a newline at the end of the string as well as at |
| 515 | the very end, whereas \z matches only at the end. |
| 516 | |
| 517 | The \G assertion is true only when the current matching position is at |
| 518 | the start point of the match, as specified by the startoffset argument |
| 519 | of pcre_exec(). It differs from \A when the value of startoffset is |
| 520 | non-zero. By calling pcre_exec() multiple times with appropriate argu- |
| 521 | ments, you can mimic Perl's /g option, and it is in this kind of imple- |
| 522 | mentation where \G can be useful. |
| 523 | |
| 524 | Note, however, that PCRE's interpretation of \G, as the start of the |
| 525 | current match, is subtly different from Perl's, which defines it as the |
| 526 | end of the previous match. In Perl, these can be different when the |
| 527 | previously matched string was empty. Because PCRE does just one match |
| 528 | at a time, it cannot reproduce this behaviour. |
| 529 | |
| 530 | If all the alternatives of a pattern begin with \G, the expression is |
| 531 | anchored to the starting match position, and the "anchored" flag is set |
| 532 | in the compiled regular expression. |
| 533 | |
| 534 | |
| 535 | CIRCUMFLEX AND DOLLAR |
| 536 | |
| 537 | Outside a character class, in the default matching mode, the circumflex |
| 538 | character is an assertion that is true only if the current matching |
| 539 | point is at the start of the subject string. If the startoffset argu- |
| 540 | ment of pcre_exec() is non-zero, circumflex can never match if the |
| 541 | PCRE_MULTILINE option is unset. Inside a character class, circumflex |
| 542 | has an entirely different meaning (see below). |
| 543 | |
| 544 | Circumflex need not be the first character of the pattern if a number |
| 545 | of alternatives are involved, but it should be the first thing in each |
| 546 | alternative in which it appears if the pattern is ever to match that |
| 547 | branch. If all possible alternatives start with a circumflex, that is, |
| 548 | if the pattern is constrained to match only at the start of the sub- |
| 549 | ject, it is said to be an "anchored" pattern. (There are also other |
| 550 | constructs that can cause a pattern to be anchored.) |
| 551 | |
| 552 | A dollar character is an assertion that is true only if the current |
| 553 | matching point is at the end of the subject string, or immediately |
| 554 | before a newline at the end of the string (by default). Dollar need not |
| 555 | be the last character of the pattern if a number of alternatives are |
| 556 | involved, but it should be the last item in any branch in which it |
| 557 | appears. Dollar has no special meaning in a character class. |
| 558 | |
| 559 | The meaning of dollar can be changed so that it matches only at the |
| 560 | very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at |
| 561 | compile time. This does not affect the \Z assertion. |
| 562 | |
| 563 | The meanings of the circumflex and dollar characters are changed if the |
| 564 | PCRE_MULTILINE option is set. When this is the case, a circumflex |
| 565 | matches immediately after internal newlines as well as at the start of |
| 566 | the subject string. It does not match after a newline that ends the |
| 567 | string. A dollar matches before any newlines in the string, as well as |
| 568 | at the very end, when PCRE_MULTILINE is set. When newline is specified |
| 569 | as the two-character sequence CRLF, isolated CR and LF characters do |
| 570 | not indicate newlines. |
| 571 | |
| 572 | For example, the pattern /^abc$/ matches the subject string "def\nabc" |
| 573 | (where \n represents a newline) in multiline mode, but not otherwise. |
| 574 | Consequently, patterns that are anchored in single line mode because |
| 575 | all branches start with ^ are not anchored in multiline mode, and a |
| 576 | match for circumflex is possible when the startoffset argument of |
| 577 | pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if |
| 578 | PCRE_MULTILINE is set. |
| 579 | |
| 580 | Note that the sequences \A, \Z, and \z can be used to match the start |
| 581 | and end of the subject in both modes, and if all branches of a pattern |
| 582 | start with \A it is always anchored, whether or not PCRE_MULTILINE is |
| 583 | set. |
| 584 | |
| 585 | |
| 586 | FULL STOP (PERIOD, DOT) |
| 587 | |
| 588 | Outside a character class, a dot in the pattern matches any one charac- |
| 589 | ter in the subject string except (by default) a character that signi- |
| 590 | fies the end of a line. In UTF-8 mode, the matched character may be |
| 591 | more than one byte long. |
| 592 | |
| 593 | When a line ending is defined as a single character, dot never matches |
| 594 | that character; when the two-character sequence CRLF is used, dot does |
| 595 | not match CR if it is immediately followed by LF, but otherwise it |
| 596 | matches all characters (including isolated CRs and LFs). When any Uni- |
| 597 | code line endings are being recognized, dot does not match CR or LF or |
| 598 | any of the other line ending characters. |
| 599 | |
| 600 | The behaviour of dot with regard to newlines can be changed. If the |
| 601 | PCRE_DOTALL option is set, a dot matches any one character, without |
| 602 | exception. If the two-character sequence CRLF is present in the subject |
| 603 | string, it takes two dots to match it. |
| 604 | |
| 605 | The handling of dot is entirely independent of the handling of circum- |
| 606 | flex and dollar, the only relationship being that they both involve |
| 607 | newlines. Dot has no special meaning in a character class. |
| 608 | |
| 609 | |
| 610 | MATCHING A SINGLE BYTE |
| 611 | |
| 612 | Outside a character class, the escape sequence \C matches any one byte, |
| 613 | both in and out of UTF-8 mode. Unlike a dot, it always matches any |
| 614 | line-ending characters. The feature is provided in Perl in order to |
| 615 | match individual bytes in UTF-8 mode. Because it breaks up UTF-8 char- |
| 616 | acters into individual bytes, what remains in the string may be a mal- |
| 617 | formed UTF-8 string. For this reason, the \C escape sequence is best |
| 618 | avoided. |
| 619 | |
| 620 | PCRE does not allow \C to appear in lookbehind assertions (described |
| 621 | below), because in UTF-8 mode this would make it impossible to calcu- |
| 622 | late the length of the lookbehind. |
| 623 | |
| 624 | |
| 625 | SQUARE BRACKETS AND CHARACTER CLASSES |
| 626 | |
| 627 | An opening square bracket introduces a character class, terminated by a |
| 628 | closing square bracket. A closing square bracket on its own is not spe- |
| 629 | cial. If a closing square bracket is required as a member of the class, |
| 630 | it should be the first data character in the class (after an initial |
| 631 | circumflex, if present) or escaped with a backslash. |
| 632 | |
| 633 | A character class matches a single character in the subject. In UTF-8 |
| 634 | mode, the character may occupy more than one byte. A matched character |
| 635 | must be in the set of characters defined by the class, unless the first |
| 636 | character in the class definition is a circumflex, in which case the |
| 637 | subject character must not be in the set defined by the class. If a |
| 638 | circumflex is actually required as a member of the class, ensure it is |
| 639 | not the first character, or escape it with a backslash. |
| 640 | |
| 641 | For example, the character class [aeiou] matches any lower case vowel, |
| 642 | while [^aeiou] matches any character that is not a lower case vowel. |
| 643 | Note that a circumflex is just a convenient notation for specifying the |
| 644 | characters that are in the class by enumerating those that are not. A |
| 645 | class that starts with a circumflex is not an assertion: it still con- |
| 646 | sumes a character from the subject string, and therefore it fails if |
| 647 | the current pointer is at the end of the string. |
| 648 | |
| 649 | In UTF-8 mode, characters with values greater than 255 can be included |
| 650 | in a class as a literal string of bytes, or by using the \x{ escaping |
| 651 | mechanism. |
| 652 | |
| 653 | When caseless matching is set, any letters in a class represent both |
| 654 | their upper case and lower case versions, so for example, a caseless |
| 655 | [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not |
| 656 | match "A", whereas a caseful version would. In UTF-8 mode, PCRE always |
| 657 | understands the concept of case for characters whose values are less |
| 658 | than 128, so caseless matching is always possible. For characters with |
| 659 | higher values, the concept of case is supported if PCRE is compiled |
| 660 | with Unicode property support, but not otherwise. If you want to use |
| 661 | caseless matching for characters 128 and above, you must ensure that |
| 662 | PCRE is compiled with Unicode property support as well as with UTF-8 |
| 663 | support. |
| 664 | |
| 665 | Characters that might indicate line breaks are never treated in any |
| 666 | special way when matching character classes, whatever line-ending |
| 667 | sequence is in use, and whatever setting of the PCRE_DOTALL and |
| 668 | PCRE_MULTILINE options is used. A class such as [^a] always matches one |
| 669 | of these characters. |
| 670 | |
| 671 | The minus (hyphen) character can be used to specify a range of charac- |
| 672 | ters in a character class. For example, [d-m] matches any letter |
| 673 | between d and m, inclusive. If a minus character is required in a |
| 674 | class, it must be escaped with a backslash or appear in a position |
| 675 | where it cannot be interpreted as indicating a range, typically as the |
| 676 | first or last character in the class. |
| 677 | |
| 678 | It is not possible to have the literal character "]" as the end charac- |
| 679 | ter of a range. A pattern such as [W-]46] is interpreted as a class of |
| 680 | two characters ("W" and "-") followed by a literal string "46]", so it |
| 681 | would match "W46]" or "-46]". However, if the "]" is escaped with a |
| 682 | backslash it is interpreted as the end of range, so [W-\]46] is inter- |
| 683 | preted as a class containing a range followed by two other characters. |
| 684 | The octal or hexadecimal representation of "]" can also be used to end |
| 685 | a range. |
| 686 | |
| 687 | Ranges operate in the collating sequence of character values. They can |
| 688 | also be used for characters specified numerically, for example |
| 689 | [\000-\037]. In UTF-8 mode, ranges can include characters whose values |
| 690 | are greater than 255, for example [\x{100}-\x{2ff}]. |
| 691 | |
| 692 | If a range that includes letters is used when caseless matching is set, |
| 693 | it matches the letters in either case. For example, [W-c] is equivalent |
| 694 | to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if |
| 695 | character tables for a French locale are in use, [\xc8-\xcb] matches |
| 696 | accented E characters in both cases. In UTF-8 mode, PCRE supports the |
| 697 | concept of case for characters with values greater than 128 only when |
| 698 | it is compiled with Unicode property support. |
| 699 | |
| 700 | The character types \d, \D, \p, \P, \s, \S, \w, and \W may also appear |
| 701 | in a character class, and add the characters that they match to the |
| 702 | class. For example, [\dABCDEF] matches any hexadecimal digit. A circum- |
| 703 | flex can conveniently be used with the upper case character types to |
| 704 | specify a more restricted set of characters than the matching lower |
| 705 | case type. For example, the class [^\W_] matches any letter or digit, |
| 706 | but not underscore. |
| 707 | |
| 708 | The only metacharacters that are recognized in character classes are |
| 709 | backslash, hyphen (only where it can be interpreted as specifying a |
| 710 | range), circumflex (only at the start), opening square bracket (only |
| 711 | when it can be interpreted as introducing a POSIX class name - see the |
| 712 | next section), and the terminating closing square bracket. However, |
| 713 | escaping other non-alphanumeric characters does no harm. |
| 714 | |
| 715 | |
| 716 | POSIX CHARACTER CLASSES |
| 717 | |
| 718 | Perl supports the POSIX notation for character classes. This uses names |
| 719 | enclosed by [: and :] within the enclosing square brackets. PCRE also |
| 720 | supports this notation. For example, |
| 721 | |
| 722 | [01[:alpha:]%] |
| 723 | |
| 724 | matches "0", "1", any alphabetic character, or "%". The supported class |
| 725 | names are |
| 726 | |
| 727 | alnum letters and digits |
| 728 | alpha letters |
| 729 | ascii character codes 0 - 127 |
| 730 | blank space or tab only |
| 731 | cntrl control characters |
| 732 | digit decimal digits (same as \d) |
| 733 | graph printing characters, excluding space |
| 734 | lower lower case letters |
| 735 | print printing characters, including space |
| 736 | punct printing characters, excluding letters and digits |
| 737 | space white space (not quite the same as \s) |
| 738 | upper upper case letters |
| 739 | word "word" characters (same as \w) |
| 740 | xdigit hexadecimal digits |
| 741 | |
| 742 | The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), |
| 743 | and space (32). Notice that this list includes the VT character (code |
| 744 | 11). This makes "space" different to \s, which does not include VT (for |
| 745 | Perl compatibility). |
| 746 | |
| 747 | The name "word" is a Perl extension, and "blank" is a GNU extension |
| 748 | from Perl 5.8. Another Perl extension is negation, which is indicated |
| 749 | by a ^ character after the colon. For example, |
| 750 | |
| 751 | [12[:^digit:]] |
| 752 | |
| 753 | matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the |
| 754 | POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but |
| 755 | these are not supported, and an error is given if they are encountered. |
| 756 | |
| 757 | In UTF-8 mode, characters with values greater than 128 do not match any |
| 758 | of the POSIX character classes. |
| 759 | |
| 760 | |
| 761 | VERTICAL BAR |
| 762 | |
| 763 | Vertical bar characters are used to separate alternative patterns. For |
| 764 | example, the pattern |
| 765 | |
| 766 | gilbert|sullivan |
| 767 | |
| 768 | matches either "gilbert" or "sullivan". Any number of alternatives may |
| 769 | appear, and an empty alternative is permitted (matching the empty |
| 770 | string). The matching process tries each alternative in turn, from left |
| 771 | to right, and the first one that succeeds is used. If the alternatives |
| 772 | are within a subpattern (defined below), "succeeds" means matching the |
| 773 | rest of the main pattern as well as the alternative in the subpattern. |
| 774 | |
| 775 | |
| 776 | INTERNAL OPTION SETTING |
| 777 | |
| 778 | The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and |
| 779 | PCRE_EXTENDED options can be changed from within the pattern by a |
| 780 | sequence of Perl option letters enclosed between "(?" and ")". The |
| 781 | option letters are |
| 782 | |
| 783 | i for PCRE_CASELESS |
| 784 | m for PCRE_MULTILINE |
| 785 | s for PCRE_DOTALL |
| 786 | x for PCRE_EXTENDED |
| 787 | |
| 788 | For example, (?im) sets caseless, multiline matching. It is also possi- |
| 789 | ble to unset these options by preceding the letter with a hyphen, and a |
| 790 | combined setting and unsetting such as (?im-sx), which sets PCRE_CASE- |
| 791 | LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, |
| 792 | is also permitted. If a letter appears both before and after the |
| 793 | hyphen, the option is unset. |
| 794 | |
| 795 | When an option change occurs at top level (that is, not inside subpat- |
| 796 | tern parentheses), the change applies to the remainder of the pattern |
| 797 | that follows. If the change is placed right at the start of a pattern, |
| 798 | PCRE extracts it into the global options (and it will therefore show up |
| 799 | in data extracted by the pcre_fullinfo() function). |
| 800 | |
| 801 | An option change within a subpattern (see below for a description of |
| 802 | subpatterns) affects only that part of the current pattern that follows |
| 803 | it, so |
| 804 | |
| 805 | (a(?i)b)c |
| 806 | |
| 807 | matches abc and aBc and no other strings (assuming PCRE_CASELESS is not |
| 808 | used). By this means, options can be made to have different settings |
| 809 | in different parts of the pattern. Any changes made in one alternative |
| 810 | do carry on into subsequent branches within the same subpattern. For |
| 811 | example, |
| 812 | |
| 813 | (a(?i)b|c) |
| 814 | |
| 815 | matches "ab", "aB", "c", and "C", even though when matching "C" the |
| 816 | first branch is abandoned before the option setting. This is because |
| 817 | the effects of option settings happen at compile time. There would be |
| 818 | some very weird behaviour otherwise. |
| 819 | |
| 820 | The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA |
| 821 | can be changed in the same way as the Perl-compatible options by using |
| 822 | the characters J, U and X respectively. |
| 823 | |
| 824 | |
| 825 | SUBPATTERNS |
| 826 | |
| 827 | Subpatterns are delimited by parentheses (round brackets), which can be |
| 828 | nested. Turning part of a pattern into a subpattern does two things: |
| 829 | |
| 830 | 1. It localizes a set of alternatives. For example, the pattern |
| 831 | |
| 832 | cat(aract|erpillar|) |
| 833 | |
| 834 | matches one of the words "cat", "cataract", or "caterpillar". Without |
| 835 | the parentheses, it would match "cataract", "erpillar" or an empty |
| 836 | string. |
| 837 | |
| 838 | 2. It sets up the subpattern as a capturing subpattern. This means |
| 839 | that, when the whole pattern matches, that portion of the subject |
| 840 | string that matched the subpattern is passed back to the caller via the |
| 841 | ovector argument of pcre_exec(). Opening parentheses are counted from |
| 842 | left to right (starting from 1) to obtain numbers for the capturing |
| 843 | subpatterns. |
| 844 | |
| 845 | For example, if the string "the red king" is matched against the pat- |
| 846 | tern |
| 847 | |
| 848 | the ((red|white) (king|queen)) |
| 849 | |
| 850 | the captured substrings are "red king", "red", and "king", and are num- |
| 851 | bered 1, 2, and 3, respectively. |
| 852 | |
| 853 | The fact that plain parentheses fulfil two functions is not always |
| 854 | helpful. There are often times when a grouping subpattern is required |
| 855 | without a capturing requirement. If an opening parenthesis is followed |
| 856 | by a question mark and a colon, the subpattern does not do any captur- |
| 857 | ing, and is not counted when computing the number of any subsequent |
| 858 | capturing subpatterns. For example, if the string "the white queen" is |
| 859 | matched against the pattern |
| 860 | |
| 861 | the ((?:red|white) (king|queen)) |
| 862 | |
| 863 | the captured substrings are "white queen" and "queen", and are numbered |
| 864 | 1 and 2. The maximum number of capturing subpatterns is 65535. |
| 865 | |
| 866 | As a convenient shorthand, if any option settings are required at the |
| 867 | start of a non-capturing subpattern, the option letters may appear |
| 868 | between the "?" and the ":". Thus the two patterns |
| 869 | |
| 870 | (?i:saturday|sunday) |
| 871 | (?:(?i)saturday|sunday) |
| 872 | |
| 873 | match exactly the same set of strings. Because alternative branches are |
| 874 | tried from left to right, and options are not reset until the end of |
| 875 | the subpattern is reached, an option setting in one branch does affect |
| 876 | subsequent branches, so the above patterns match "SUNDAY" as well as |
| 877 | "Saturday". |
| 878 | |
| 879 | |
| 880 | DUPLICATE SUBPATTERN NUMBERS |
| 881 | |
| 882 | Perl 5.10 introduced a feature whereby each alternative in a subpattern |
| 883 | uses the same numbers for its capturing parentheses. Such a subpattern |
| 884 | starts with (?| and is itself a non-capturing subpattern. For example, |
| 885 | consider this pattern: |
| 886 | |
| 887 | (?|(Sat)ur|(Sun))day |
| 888 | |
| 889 | Because the two alternatives are inside a (?| group, both sets of cap- |
| 890 | turing parentheses are numbered one. Thus, when the pattern matches, |
| 891 | you can look at captured substring number one, whichever alternative |
| 892 | matched. This construct is useful when you want to capture part, but |
| 893 | not all, of one of a number of alternatives. Inside a (?| group, paren- |
| 894 | theses are numbered as usual, but the number is reset at the start of |
| 895 | each branch. The numbers of any capturing buffers that follow the sub- |
| 896 | pattern start after the highest number used in any branch. The follow- |
| 897 | ing example is taken from the Perl documentation. The numbers under- |
| 898 | neath show in which buffer the captured content will be stored. |
| 899 | |
| 900 | # before ---------------branch-reset----------- after |
| 901 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x |
| 902 | # 1 2 2 3 2 3 4 |
| 903 | |
| 904 | A backreference or a recursive call to a numbered subpattern always |
| 905 | refers to the first one in the pattern with the given number. |
| 906 | |
| 907 | An alternative approach to using this "branch reset" feature is to use |
| 908 | duplicate named subpatterns, as described in the next section. |
| 909 | |
| 910 | |
| 911 | NAMED SUBPATTERNS |
| 912 | |
| 913 | Identifying capturing parentheses by number is simple, but it can be |
| 914 | very hard to keep track of the numbers in complicated regular expres- |
| 915 | sions. Furthermore, if an expression is modified, the numbers may |
| 916 | change. To help with this difficulty, PCRE supports the naming of sub- |
| 917 | patterns. This feature was not added to Perl until release 5.10. Python |
| 918 | had the feature earlier, and PCRE introduced it at release 4.0, using |
| 919 | the Python syntax. PCRE now supports both the Perl and the Python syn- |
| 920 | tax. |
| 921 | |
| 922 | In PCRE, a subpattern can be named in one of three ways: (?<name>...) |
| 923 | or (?'name'...) as in Perl, or (?P<name>...) as in Python. References |
| 924 | to capturing parentheses from other parts of the pattern, such as back- |
| 925 | references, recursion, and conditions, can be made by name as well as |
| 926 | by number. |
| 927 | |
| 928 | Names consist of up to 32 alphanumeric characters and underscores. |
| 929 | Named capturing parentheses are still allocated numbers as well as |
| 930 | names, exactly as if the names were not present. The PCRE API provides |
| 931 | function calls for extracting the name-to-number translation table from |
| 932 | a compiled pattern. There is also a convenience function for extracting |
| 933 | a captured substring by name. |
| 934 | |
| 935 | By default, a name must be unique within a pattern, but it is possible |
| 936 | to relax this constraint by setting the PCRE_DUPNAMES option at compile |
| 937 | time. This can be useful for patterns where only one instance of the |
| 938 | named parentheses can match. Suppose you want to match the name of a |
| 939 | weekday, either as a 3-letter abbreviation or as the full name, and in |
| 940 | both cases you want to extract the abbreviation. This pattern (ignoring |
| 941 | the line breaks) does the job: |
| 942 | |
| 943 | (?<DN>Mon|Fri|Sun)(?:day)?| |
| 944 | (?<DN>Tue)(?:sday)?| |
| 945 | (?<DN>Wed)(?:nesday)?| |
| 946 | (?<DN>Thu)(?:rsday)?| |
| 947 | (?<DN>Sat)(?:urday)? |
| 948 | |
| 949 | There are five capturing substrings, but only one is ever set after a |
| 950 | match. (An alternative way of solving this problem is to use a "branch |
| 951 | reset" subpattern, as described in the previous section.) |
| 952 | |
| 953 | The convenience function for extracting the data by name returns the |
| 954 | substring for the first (and in this example, the only) subpattern of |
| 955 | that name that matched. This saves searching to find which numbered |
| 956 | subpattern it was. If you make a reference to a non-unique named sub- |
| 957 | pattern from elsewhere in the pattern, the one that corresponds to the |
| 958 | lowest number is used. For further details of the interfaces for han- |
| 959 | dling named subpatterns, see the pcreapi documentation. |
| 960 | |
| 961 | |
| 962 | REPETITION |
| 963 | |
| 964 | Repetition is specified by quantifiers, which can follow any of the |
| 965 | following items: |
| 966 | |
| 967 | a literal data character |
| 968 | the dot metacharacter |
| 969 | the \C escape sequence |
| 970 | the \X escape sequence (in UTF-8 mode with Unicode properties) |
| 971 | the \R escape sequence |
| 972 | an escape such as \d that matches a single character |
| 973 | a character class |
| 974 | a back reference (see next section) |
| 975 | a parenthesized subpattern (unless it is an assertion) |
| 976 | |
| 977 | The general repetition quantifier specifies a minimum and maximum num- |
| 978 | ber of permitted matches, by giving the two numbers in curly brackets |
| 979 | (braces), separated by a comma. The numbers must be less than 65536, |
| 980 | and the first must be less than or equal to the second. For example: |
| 981 | |
| 982 | z{2,4} |
| 983 | |
| 984 | matches "zz", "zzz", or "zzzz". A closing brace on its own is not a |
| 985 | special character. If the second number is omitted, but the comma is |
| 986 | present, there is no upper limit; if the second number and the comma |
| 987 | are both omitted, the quantifier specifies an exact number of required |
| 988 | matches. Thus |
| 989 | |
| 990 | [aeiou]{3,} |
| 991 | |
| 992 | matches at least 3 successive vowels, but may match many more, while |
| 993 | |
| 994 | \d{8} |
| 995 | |
| 996 | matches exactly 8 digits. An opening curly bracket that appears in a |
| 997 | position where a quantifier is not allowed, or one that does not match |
| 998 | the syntax of a quantifier, is taken as a literal character. For exam- |
| 999 | ple, {,6} is not a quantifier, but a literal string of four characters. |
| 1000 | |
| 1001 | In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to |
| 1002 | individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char- |
| 1003 | acters, each of which is represented by a two-byte sequence. Similarly, |
| 1004 | when Unicode property support is available, \X{3} matches three Unicode |
| 1005 | extended sequences, each of which may be several bytes long (and they |
| 1006 | may be of different lengths). |
| 1007 | |
| 1008 | The quantifier {0} is permitted, causing the expression to behave as if |
| 1009 | the previous item and the quantifier were not present. |
| 1010 | |
| 1011 | For convenience, the three most common quantifiers have single-charac- |
| 1012 | ter abbreviations: |
| 1013 | |
| 1014 | * is equivalent to {0,} |
| 1015 | + is equivalent to {1,} |
| 1016 | ? is equivalent to {0,1} |
| 1017 | |
| 1018 | It is possible to construct infinite loops by following a subpattern |
| 1019 | that can match no characters with a quantifier that has no upper limit, |
| 1020 | for example: |
| 1021 | |
| 1022 | (a?)* |
| 1023 | |
| 1024 | Earlier versions of Perl and PCRE used to give an error at compile time |
| 1025 | for such patterns. However, because there are cases where this can be |
| 1026 | useful, such patterns are now accepted, but if any repetition of the |
| 1027 | subpattern does in fact match no characters, the loop is forcibly bro- |
| 1028 | ken. |
| 1029 | |
| 1030 | By default, the quantifiers are "greedy", that is, they match as much |
| 1031 | as possible (up to the maximum number of permitted times), without |
| 1032 | causing the rest of the pattern to fail. The classic example of where |
| 1033 | this gives problems is in trying to match comments in C programs. These |
| 1034 | appear between /* and */ and within the comment, individual * and / |
| 1035 | characters may appear. An attempt to match C comments by applying the |
| 1036 | pattern |
| 1037 | |
| 1038 | /\*.*\*/ |
| 1039 | |
| 1040 | to the string |
| 1041 | |
| 1042 | /* first comment */ not comment /* second comment */ |
| 1043 | |
| 1044 | fails, because it matches the entire string owing to the greediness of |
| 1045 | the .* item. |
| 1046 | |
| 1047 | However, if a quantifier is followed by a question mark, it ceases to |
| 1048 | be greedy, and instead matches the minimum number of times possible, so |
| 1049 | the pattern |
| 1050 | |
| 1051 | /\*.*?\*/ |
| 1052 | |
| 1053 | does the right thing with the C comments. The meaning of the various |
| 1054 | quantifiers is not otherwise changed, just the preferred number of |
| 1055 | matches. Do not confuse this use of question mark with its use as a |
| 1056 | quantifier in its own right. Because it has two uses, it can sometimes |
| 1057 | appear doubled, as in |
| 1058 | |
| 1059 | \d??\d |
| 1060 | |
| 1061 | which matches one digit by preference, but can match two if that is the |
| 1062 | only way the rest of the pattern matches. |
| 1063 | |
| 1064 | If the PCRE_UNGREEDY option is set (an option that is not available in |
| 1065 | Perl), the quantifiers are not greedy by default, but individual ones |
| 1066 | can be made greedy by following them with a question mark. In other |
| 1067 | words, it inverts the default behaviour. |
| 1068 | |
| 1069 | When a parenthesized subpattern is quantified with a minimum repeat |
| 1070 | count that is greater than 1 or with a limited maximum, more memory is |
| 1071 | required for the compiled pattern, in proportion to the size of the |
| 1072 | minimum or maximum. |
| 1073 | |
| 1074 | If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv- |
| 1075 | alent to Perl's /s) is set, thus allowing the dot to match newlines, |
| 1076 | the pattern is implicitly anchored, because whatever follows will be |
| 1077 | tried against every character position in the subject string, so there |
| 1078 | is no point in retrying the overall match at any position after the |
| 1079 | first. PCRE normally treats such a pattern as though it were preceded |
| 1080 | by \A. |
| 1081 | |
| 1082 | In cases where it is known that the subject string contains no new- |
| 1083 | lines, it is worth setting PCRE_DOTALL in order to obtain this opti- |
| 1084 | mization, or alternatively using ^ to indicate anchoring explicitly. |
| 1085 | |
| 1086 | However, there is one situation where the optimization cannot be used. |
| 1087 | When .* is inside capturing parentheses that are the subject of a |
| 1088 | backreference elsewhere in the pattern, a match at the start may fail |
| 1089 | where a later one succeeds. Consider, for example: |
| 1090 | |
| 1091 | (.*)abc\1 |
| 1092 | |
| 1093 | If the subject is "xyz123abc123" the match point is the fourth charac- |
| 1094 | ter. For this reason, such a pattern is not implicitly anchored. |
| 1095 | |
| 1096 | When a capturing subpattern is repeated, the value captured is the sub- |
| 1097 | string that matched the final iteration. For example, after |
| 1098 | |
| 1099 | (tweedle[dume]{3}\s*)+ |
| 1100 | |
| 1101 | has matched "tweedledum tweedledee" the value of the captured substring |
| 1102 | is "tweedledee". However, if there are nested capturing subpatterns, |
| 1103 | the corresponding captured values may have been set in previous itera- |
| 1104 | tions. For example, after |
| 1105 | |
| 1106 | /(a|(b))+/ |
| 1107 | |
| 1108 | matches "aba" the value of the second captured substring is "b". |
| 1109 | |
| 1110 | |
| 1111 | ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS |
| 1112 | |
| 1113 | With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy") |
| 1114 | repetition, failure of what follows normally causes the repeated item |
| 1115 | to be re-evaluated to see if a different number of repeats allows the |
| 1116 | rest of the pattern to match. Sometimes it is useful to prevent this, |
| 1117 | either to change the nature of the match, or to cause it fail earlier |
| 1118 | than it otherwise might, when the author of the pattern knows there is |
| 1119 | no point in carrying on. |
| 1120 | |
| 1121 | Consider, for example, the pattern \d+foo when applied to the subject |
| 1122 | line |
| 1123 | |
| 1124 | 123456bar |
| 1125 | |
| 1126 | After matching all 6 digits and then failing to match "foo", the normal |
| 1127 | action of the matcher is to try again with only 5 digits matching the |
| 1128 | \d+ item, and then with 4, and so on, before ultimately failing. |
| 1129 | "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides |
| 1130 | the means for specifying that once a subpattern has matched, it is not |
| 1131 | to be re-evaluated in this way. |
| 1132 | |
| 1133 | If we use atomic grouping for the previous example, the matcher gives |
| 1134 | up immediately on failing to match "foo" the first time. The notation |
| 1135 | is a kind of special parenthesis, starting with (?> as in this example: |
| 1136 | |
| 1137 | (?>\d+)foo |
| 1138 | |
| 1139 | This kind of parenthesis "locks up" the part of the pattern it con- |
| 1140 | tains once it has matched, and a failure further into the pattern is |
| 1141 | prevented from backtracking into it. Backtracking past it to previous |
| 1142 | items, however, works as normal. |
| 1143 | |
| 1144 | An alternative description is that a subpattern of this type matches |
| 1145 | the string of characters that an identical standalone pattern would |
| 1146 | match, if anchored at the current point in the subject string. |
| 1147 | |
| 1148 | Atomic grouping subpatterns are not capturing subpatterns. Simple cases |
| 1149 | such as the above example can be thought of as a maximizing repeat that |
| 1150 | must swallow everything it can. So, while both \d+ and \d+? are pre- |
| 1151 | pared to adjust the number of digits they match in order to make the |
| 1152 | rest of the pattern match, (?>\d+) can only match an entire sequence of |
| 1153 | digits. |
| 1154 | |
| 1155 | Atomic groups in general can of course contain arbitrarily complicated |
| 1156 | subpatterns, and can be nested. However, when the subpattern for an |
| 1157 | atomic group is just a single repeated item, as in the example above, a |
| 1158 | simpler notation, called a "possessive quantifier" can be used. This |
| 1159 | consists of an additional + character following a quantifier. Using |
| 1160 | this notation, the previous example can be rewritten as |
| 1161 | |
| 1162 | \d++foo |
| 1163 | |
| 1164 | Possessive quantifiers are always greedy; the setting of the |
| 1165 | PCRE_UNGREEDY option is ignored. They are a convenient notation for the |
| 1166 | simpler forms of atomic group. However, there is no difference in the |
| 1167 | meaning of a possessive quantifier and the equivalent atomic group, |
| 1168 | though there may be a performance difference; possessive quantifiers |
| 1169 | should be slightly faster. |
| 1170 | |
| 1171 | The possessive quantifier syntax is an extension to the Perl 5.8 syn- |
| 1172 | tax. Jeffrey Friedl originated the idea (and the name) in the first |
| 1173 | edition of his book. Mike McCloskey liked it, so implemented it when he |
| 1174 | built Sun's Java package, and PCRE copied it from there. It ultimately |
| 1175 | found its way into Perl at release 5.10. |
| 1176 | |
| 1177 | PCRE has an optimization that automatically "possessifies" certain sim- |
| 1178 | ple pattern constructs. For example, the sequence A+B is treated as |
| 1179 | A++B because there is no point in backtracking into a sequence of A's |
| 1180 | when B must follow. |
| 1181 | |
| 1182 | When a pattern contains an unlimited repeat inside a subpattern that |
| 1183 | can itself be repeated an unlimited number of times, the use of an |
| 1184 | atomic group is the only way to avoid some failing matches taking a |
| 1185 | very long time indeed. The pattern |
| 1186 | |
| 1187 | (\D+|<\d+>)*[!?] |
| 1188 | |
| 1189 | matches an unlimited number of substrings that either consist of non- |
| 1190 | digits, or digits enclosed in <>, followed by either ! or ?. When it |
| 1191 | matches, it runs quickly. However, if it is applied to |
| 1192 | |
| 1193 | aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa |
| 1194 | |
| 1195 | it takes a long time before reporting failure. This is because the |
| 1196 | string can be divided between the internal \D+ repeat and the external |
| 1197 | * repeat in a large number of ways, and all have to be tried. (The |
| 1198 | example uses [!?] rather than a single character at the end, because |
| 1199 | both PCRE and Perl have an optimization that allows for fast failure |
| 1200 | when a single character is used. They remember the last single charac- |
| 1201 | ter that is required for a match, and fail early if it is not present |
| 1202 | in the string.) If the pattern is changed so that it uses an atomic |
| 1203 | group, like this: |
| 1204 | |
| 1205 | ((?>\D+)|<\d+>)*[!?] |
| 1206 | |
| 1207 | sequences of non-digits cannot be broken, and failure happens quickly. |
| 1208 | |
| 1209 | |
| 1210 | BACK REFERENCES |
| 1211 | |
| 1212 | Outside a character class, a backslash followed by a digit greater than |
| 1213 | 0 (and possibly further digits) is a back reference to a capturing sub- |
| 1214 | pattern earlier (that is, to its left) in the pattern, provided there |
| 1215 | have been that many previous capturing left parentheses. |
| 1216 | |
| 1217 | However, if the decimal number following the backslash is less than 10, |
| 1218 | it is always taken as a back reference, and causes an error only if |
| 1219 | there are not that many capturing left parentheses in the entire pat- |
| 1220 | tern. In other words, the parentheses that are referenced need not be |
| 1221 | to the left of the reference for numbers less than 10. A "forward back |
| 1222 | reference" of this type can make sense when a repetition is involved |
| 1223 | and the subpattern to the right has participated in an earlier itera- |
| 1224 | tion. |
| 1225 | |
| 1226 | It is not possible to have a numerical "forward back reference" to a |
| 1227 | subpattern whose number is 10 or more using this syntax because a |
| 1228 | sequence such as \50 is interpreted as a character defined in octal. |
| 1229 | See the subsection entitled "Non-printing characters" above for further |
| 1230 | details of the handling of digits following a backslash. There is no |
| 1231 | such problem when named parentheses are used. A back reference to any |
| 1232 | subpattern is possible using named parentheses (see below). |
| 1233 | |
| 1234 | Another way of avoiding the ambiguity inherent in the use of digits |
| 1235 | following a backslash is to use the \g escape sequence, which is a fea- |
| 1236 | ture introduced in Perl 5.10. This escape must be followed by a posi- |
| 1237 | tive or a negative number, optionally enclosed in braces. These exam- |
| 1238 | ples are all identical: |
| 1239 | |
| 1240 | (ring), \1 |
| 1241 | (ring), \g1 |
| 1242 | (ring), \g{1} |
| 1243 | |
| 1244 | A positive number specifies an absolute reference without the ambiguity |
| 1245 | that is present in the older syntax. It is also useful when literal |
| 1246 | digits follow the reference. A negative number is a relative reference. |
| 1247 | Consider this example: |
| 1248 | |
| 1249 | (abc(def)ghi)\g{-1} |
| 1250 | |
| 1251 | The sequence \g{-1} is a reference to the most recently started captur- |
| 1252 | ing subpattern before \g, that is, is it equivalent to \2. Similarly, |
| 1253 | \g{-2} would be equivalent to \1. The use of relative references can be |
| 1254 | helpful in long patterns, and also in patterns that are created by |
| 1255 | joining together fragments that contain references within themselves. |
| 1256 | |
| 1257 | A back reference matches whatever actually matched the capturing sub- |
| 1258 | pattern in the current subject string, rather than anything matching |
| 1259 | the subpattern itself (see "Subpatterns as subroutines" below for a way |
| 1260 | of doing that). So the pattern |
| 1261 | |
| 1262 | (sens|respons)e and \1ibility |
| 1263 | |
| 1264 | matches "sense and sensibility" and "response and responsibility", but |
| 1265 | not "sense and responsibility". If caseful matching is in force at the |
| 1266 | time of the back reference, the case of letters is relevant. For exam- |
| 1267 | ple, |
| 1268 | |
| 1269 | ((?i)rah)\s+\1 |
| 1270 | |
| 1271 | matches "rah rah" and "RAH RAH", but not "RAH rah", even though the |
| 1272 | original capturing subpattern is matched caselessly. |
| 1273 | |
| 1274 | There are several different ways of writing back references to named |
| 1275 | subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or |
| 1276 | \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's |
| 1277 | unified back reference syntax, in which \g can be used for both numeric |
| 1278 | and named references, is also supported. We could rewrite the above |
| 1279 | example in any of the following ways: |
| 1280 | |
| 1281 | (?<p1>(?i)rah)\s+\k<p1> |
| 1282 | (?'p1'(?i)rah)\s+\k{p1} |
| 1283 | (?P<p1>(?i)rah)\s+(?P=p1) |
| 1284 | (?<p1>(?i)rah)\s+\g{p1} |
| 1285 | |
| 1286 | A subpattern that is referenced by name may appear in the pattern |
| 1287 | before or after the reference. |
| 1288 | |
| 1289 | There may be more than one back reference to the same subpattern. If a |
| 1290 | subpattern has not actually been used in a particular match, any back |
| 1291 | references to it always fail. For example, the pattern |
| 1292 | |
| 1293 | (a|(bc))\2 |
| 1294 | |
| 1295 | always fails if it starts to match "a" rather than "bc". Because there |
| 1296 | may be many capturing parentheses in a pattern, all digits following |
| 1297 | the backslash are taken as part of a potential back reference number. |
| 1298 | If the pattern continues with a digit character, some delimiter must be |
| 1299 | used to terminate the back reference. If the PCRE_EXTENDED option is |
| 1300 | set, this can be whitespace. Otherwise an empty comment (see "Com- |
| 1301 | ments" below) can be used. |
| 1302 | |
| 1303 | A back reference that occurs inside the parentheses to which it refers |
| 1304 | fails when the subpattern is first used, so, for example, (a\1) never |
| 1305 | matches. However, such references can be useful inside repeated sub- |
| 1306 | patterns. For example, the pattern |
| 1307 | |
| 1308 | (a|b\1)+ |
| 1309 | |
| 1310 | matches any number of "a"s and also "aba", "ababbaa" etc. At each iter- |
| 1311 | ation of the subpattern, the back reference matches the character |
| 1312 | string corresponding to the previous iteration. In order for this to |
| 1313 | work, the pattern must be such that the first iteration does not need |
| 1314 | to match the back reference. This can be done using alternation, as in |
| 1315 | the example above, or by a quantifier with a minimum of zero. |
| 1316 | |
| 1317 | |
| 1318 | ASSERTIONS |
| 1319 | |
| 1320 | An assertion is a test on the characters following or preceding the |
| 1321 | current matching point that does not actually consume any characters. |
| 1322 | The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are |
| 1323 | described above. |
| 1324 | |
| 1325 | More complicated assertions are coded as subpatterns. There are two |
| 1326 | kinds: those that look ahead of the current position in the subject |
| 1327 | string, and those that look behind it. An assertion subpattern is |
| 1328 | matched in the normal way, except that it does not cause the current |
| 1329 | matching position to be changed. |
| 1330 | |
| 1331 | Assertion subpatterns are not capturing subpatterns, and may not be |
| 1332 | repeated, because it makes no sense to assert the same thing several |
| 1333 | times. If any kind of assertion contains capturing subpatterns within |
| 1334 | it, these are counted for the purposes of numbering the capturing sub- |
| 1335 | patterns in the whole pattern. However, substring capturing is carried |
| 1336 | out only for positive assertions, because it does not make sense for |
| 1337 | negative assertions. |
| 1338 | |
| 1339 | Lookahead assertions |
| 1340 | |
| 1341 | Lookahead assertions start with (?= for positive assertions and (?! for |
| 1342 | negative assertions. For example, |
| 1343 | |
| 1344 | \w+(?=;) |
| 1345 | |
| 1346 | matches a word followed by a semicolon, but does not include the semi- |
| 1347 | colon in the match, and |
| 1348 | |
| 1349 | foo(?!bar) |
| 1350 | |
| 1351 | matches any occurrence of "foo" that is not followed by "bar". Note |
| 1352 | that the apparently similar pattern |
| 1353 | |
| 1354 | (?!foo)bar |
| 1355 | |
| 1356 | does not find an occurrence of "bar" that is preceded by something |
| 1357 | other than "foo"; it finds any occurrence of "bar" whatsoever, because |
| 1358 | the assertion (?!foo) is always true when the next three characters are |
| 1359 | "bar". A lookbehind assertion is needed to achieve the other effect. |
| 1360 | |
| 1361 | If you want to force a matching failure at some point in a pattern, the |
| 1362 | most convenient way to do it is with (?!) because an empty string |
| 1363 | always matches, so an assertion that requires there not to be an empty |
| 1364 | string must always fail. |
| 1365 | |
| 1366 | Lookbehind assertions |
| 1367 | |
| 1368 | Lookbehind assertions start with (?<= for positive assertions and (?<! |
| 1369 | for negative assertions. For example, |
| 1370 | |
| 1371 | (?<!foo)bar |
| 1372 | |
| 1373 | does find an occurrence of "bar" that is not preceded by "foo". The |
| 1374 | contents of a lookbehind assertion are restricted such that all the |
| 1375 | strings it matches must have a fixed length. However, if there are sev- |
| 1376 | eral top-level alternatives, they do not all have to have the same |
| 1377 | fixed length. Thus |
| 1378 | |
| 1379 | (?<=bullock|donkey) |
| 1380 | |
| 1381 | is permitted, but |
| 1382 | |
| 1383 | (?<!dogs?|cats?) |
| 1384 | |
| 1385 | causes an error at compile time. Branches that match different length |
| 1386 | strings are permitted only at the top level of a lookbehind assertion. |
| 1387 | This is an extension compared with Perl (at least for 5.8), which |
| 1388 | requires all branches to match the same length of string. An assertion |
| 1389 | such as |
| 1390 | |
| 1391 | (?<=ab(c|de)) |
| 1392 | |
| 1393 | is not permitted, because its single top-level branch can match two |
| 1394 | different lengths, but it is acceptable if rewritten to use two top- |
| 1395 | level branches: |
| 1396 | |
| 1397 | (?<=abc|abde) |
| 1398 | |
| 1399 | In some cases, the Perl 5.10 escape sequence \K (see above) can be used |
| 1400 | instead of a lookbehind assertion; this is not restricted to a fixed- |
| 1401 | length. |
| 1402 | |
| 1403 | The implementation of lookbehind assertions is, for each alternative, |
| 1404 | to temporarily move the current position back by the fixed length and |
| 1405 | then try to match. If there are insufficient characters before the cur- |
| 1406 | rent position, the assertion fails. |
| 1407 | |
| 1408 | PCRE does not allow the \C escape (which matches a single byte in UTF-8 |
| 1409 | mode) to appear in lookbehind assertions, because it makes it impossi- |
| 1410 | ble to calculate the length of the lookbehind. The \X and \R escapes, |
| 1411 | which can match different numbers of bytes, are also not permitted. |
| 1412 | |
| 1413 | Possessive quantifiers can be used in conjunction with lookbehind |
| 1414 | assertions to specify efficient matching at the end of the subject |
| 1415 | string. Consider a simple pattern such as |
| 1416 | |
| 1417 | abcd$ |
| 1418 | |
| 1419 | when applied to a long string that does not match. Because matching |
| 1420 | proceeds from left to right, PCRE will look for each "a" in the subject |
| 1421 | and then see if what follows matches the rest of the pattern. If the |
| 1422 | pattern is specified as |
| 1423 | |
| 1424 | ^.*abcd$ |
| 1425 | |
| 1426 | the initial .* matches the entire string at first, but when this fails |
| 1427 | (because there is no following "a"), it backtracks to match all but the |
| 1428 | last character, then all but the last two characters, and so on. Once |
| 1429 | again the search for "a" covers the entire string, from right to left, |
| 1430 | so we are no better off. However, if the pattern is written as |
| 1431 | |
| 1432 | ^.*+(?<=abcd) |
| 1433 | |
| 1434 | there can be no backtracking for the .*+ item; it can match only the |
| 1435 | entire string. The subsequent lookbehind assertion does a single test |
| 1436 | on the last four characters. If it fails, the match fails immediately. |
| 1437 | For long strings, this approach makes a significant difference to the |
| 1438 | processing time. |
| 1439 | |
| 1440 | Using multiple assertions |
| 1441 | |
| 1442 | Several assertions (of any sort) may occur in succession. For example, |
| 1443 | |
| 1444 | (?<=\d{3})(?<!999)foo |
| 1445 | |
| 1446 | matches "foo" preceded by three digits that are not "999". Notice that |
| 1447 | each of the assertions is applied independently at the same point in |
| 1448 | the subject string. First there is a check that the previous three |
| 1449 | characters are all digits, and then there is a check that the same |
| 1450 | three characters are not "999". This pattern does not match "foo" pre- |
| 1451 | ceded by six characters, the first of which are digits and the last |
| 1452 | three of which are not "999". For example, it doesn't match "123abc- |
| 1453 | foo". A pattern to do that is |
| 1454 | |
| 1455 | (?<=\d{3}...)(?<!999)foo |
| 1456 | |
| 1457 | This time the first assertion looks at the preceding six characters, |
| 1458 | checking that the first three are digits, and then the second assertion |
| 1459 | checks that the preceding three characters are not "999". |
| 1460 | |
| 1461 | Assertions can be nested in any combination. For example, |
| 1462 | |
| 1463 | (?<=(?<!foo)bar)baz |
| 1464 | |
| 1465 | matches an occurrence of "baz" that is preceded by "bar" which in turn |
| 1466 | is not preceded by "foo", while |
| 1467 | |
| 1468 | (?<=\d{3}(?!999)...)foo |
| 1469 | |
| 1470 | is another pattern that matches "foo" preceded by three digits and any |
| 1471 | three characters that are not "999". |
| 1472 | |
| 1473 | |
| 1474 | CONDITIONAL SUBPATTERNS |
| 1475 | |
| 1476 | It is possible to cause the matching process to obey a subpattern con- |
| 1477 | ditionally or to choose between two alternative subpatterns, depending |
| 1478 | on the result of an assertion, or whether a previous capturing subpat- |
| 1479 | tern matched or not. The two possible forms of conditional subpattern |
| 1480 | are |
| 1481 | |
| 1482 | (?(condition)yes-pattern) |
| 1483 | (?(condition)yes-pattern|no-pattern) |
| 1484 | |
| 1485 | If the condition is satisfied, the yes-pattern is used; otherwise the |
| 1486 | no-pattern (if present) is used. If there are more than two alterna- |
| 1487 | tives in the subpattern, a compile-time error occurs. |
| 1488 | |
| 1489 | There are four kinds of condition: references to subpatterns, refer- |
| 1490 | ences to recursion, a pseudo-condition called DEFINE, and assertions. |
| 1491 | |
| 1492 | Checking for a used subpattern by number |
| 1493 | |
| 1494 | If the text between the parentheses consists of a sequence of digits, |
| 1495 | the condition is true if the capturing subpattern of that number has |
| 1496 | previously matched. An alternative notation is to precede the digits |
| 1497 | with a plus or minus sign. In this case, the subpattern number is rela- |
| 1498 | tive rather than absolute. The most recently opened parentheses can be |
| 1499 | referenced by (?(-1), the next most recent by (?(-2), and so on. In |
| 1500 | looping constructs it can also make sense to refer to subsequent groups |
| 1501 | with constructs such as (?(+2). |
| 1502 | |
| 1503 | Consider the following pattern, which contains non-significant white |
| 1504 | space to make it more readable (assume the PCRE_EXTENDED option) and to |
| 1505 | divide it into three parts for ease of discussion: |
| 1506 | |
| 1507 | ( \( )? [^()]+ (?(1) \) ) |
| 1508 | |
| 1509 | The first part matches an optional opening parenthesis, and if that |
| 1510 | character is present, sets it as the first captured substring. The sec- |
| 1511 | ond part matches one or more characters that are not parentheses. The |
| 1512 | third part is a conditional subpattern that tests whether the first set |
| 1513 | of parentheses matched or not. If they did, that is, if subject started |
| 1514 | with an opening parenthesis, the condition is true, and so the yes-pat- |
| 1515 | tern is executed and a closing parenthesis is required. Otherwise, |
| 1516 | since no-pattern is not present, the subpattern matches nothing. In |
| 1517 | other words, this pattern matches a sequence of non-parentheses, |
| 1518 | optionally enclosed in parentheses. |
| 1519 | |
| 1520 | If you were embedding this pattern in a larger one, you could use a |
| 1521 | relative reference: |
| 1522 | |
| 1523 | ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ... |
| 1524 | |
| 1525 | This makes the fragment independent of the parentheses in the larger |
| 1526 | pattern. |
| 1527 | |
| 1528 | Checking for a used subpattern by name |
| 1529 | |
| 1530 | Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a |
| 1531 | used subpattern by name. For compatibility with earlier versions of |
| 1532 | PCRE, which had this facility before Perl, the syntax (?(name)...) is |
| 1533 | also recognized. However, there is a possible ambiguity with this syn- |
| 1534 | tax, because subpattern names may consist entirely of digits. PCRE |
| 1535 | looks first for a named subpattern; if it cannot find one and the name |
| 1536 | consists entirely of digits, PCRE looks for a subpattern of that num- |
| 1537 | ber, which must be greater than zero. Using subpattern names that con- |
| 1538 | sist entirely of digits is not recommended. |
| 1539 | |
| 1540 | Rewriting the above example to use a named subpattern gives this: |
| 1541 | |
| 1542 | (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) ) |
| 1543 | |
| 1544 | |
| 1545 | Checking for pattern recursion |
| 1546 | |
| 1547 | If the condition is the string (R), and there is no subpattern with the |
| 1548 | name R, the condition is true if a recursive call to the whole pattern |
| 1549 | or any subpattern has been made. If digits or a name preceded by amper- |
| 1550 | sand follow the letter R, for example: |
| 1551 | |
| 1552 | (?(R3)...) or (?(R&name)...) |
| 1553 | |
| 1554 | the condition is true if the most recent recursion is into the subpat- |
| 1555 | tern whose number or name is given. This condition does not check the |
| 1556 | entire recursion stack. |
| 1557 | |
| 1558 | At "top level", all these recursion test conditions are false. Recur- |
| 1559 | sive patterns are described below. |
| 1560 | |
| 1561 | Defining subpatterns for use by reference only |
| 1562 | |
| 1563 | If the condition is the string (DEFINE), and there is no subpattern |
| 1564 | with the name DEFINE, the condition is always false. In this case, |
| 1565 | there may be only one alternative in the subpattern. It is always |
| 1566 | skipped if control reaches this point in the pattern; the idea of |
| 1567 | DEFINE is that it can be used to define "subroutines" that can be ref- |
| 1568 | erenced from elsewhere. (The use of "subroutines" is described below.) |
| 1569 | For example, a pattern to match an IPv4 address could be written like |
| 1570 | this (ignore whitespace and line breaks): |
| 1571 | |
| 1572 | (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) ) |
| 1573 | \b (?&byte) (\.(?&byte)){3} \b |
| 1574 | |
| 1575 | The first part of the pattern is a DEFINE group inside which a another |
| 1576 | group named "byte" is defined. This matches an individual component of |
| 1577 | an IPv4 address (a number less than 256). When matching takes place, |
| 1578 | this part of the pattern is skipped because DEFINE acts like a false |
| 1579 | condition. |
| 1580 | |
| 1581 | The rest of the pattern uses references to the named group to match the |
| 1582 | four dot-separated components of an IPv4 address, insisting on a word |
| 1583 | boundary at each end. |
| 1584 | |
| 1585 | Assertion conditions |
| 1586 | |
| 1587 | If the condition is not in any of the above formats, it must be an |
| 1588 | assertion. This may be a positive or negative lookahead or lookbehind |
| 1589 | assertion. Consider this pattern, again containing non-significant |
| 1590 | white space, and with the two alternatives on the second line: |
| 1591 | |
| 1592 | (?(?=[^a-z]*[a-z]) |
| 1593 | \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} ) |
| 1594 | |
| 1595 | The condition is a positive lookahead assertion that matches an |
| 1596 | optional sequence of non-letters followed by a letter. In other words, |
| 1597 | it tests for the presence of at least one letter in the subject. If a |
| 1598 | letter is found, the subject is matched against the first alternative; |
| 1599 | otherwise it is matched against the second. This pattern matches |
| 1600 | strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are |
| 1601 | letters and dd are digits. |
| 1602 | |
| 1603 | |
| 1604 | COMMENTS |
| 1605 | |
| 1606 | The sequence (?# marks the start of a comment that continues up to the |
| 1607 | next closing parenthesis. Nested parentheses are not permitted. The |
| 1608 | characters that make up a comment play no part in the pattern matching |
| 1609 | at all. |
| 1610 | |
| 1611 | If the PCRE_EXTENDED option is set, an unescaped # character outside a |
| 1612 | character class introduces a comment that continues to immediately |
| 1613 | after the next newline in the pattern. |
| 1614 | |
| 1615 | |
| 1616 | RECURSIVE PATTERNS |
| 1617 | |
| 1618 | Consider the problem of matching a string in parentheses, allowing for |
| 1619 | unlimited nested parentheses. Without the use of recursion, the best |
| 1620 | that can be done is to use a pattern that matches up to some fixed |
| 1621 | depth of nesting. It is not possible to handle an arbitrary nesting |
| 1622 | depth. |
| 1623 | |
| 1624 | For some time, Perl has provided a facility that allows regular expres- |
| 1625 | sions to recurse (amongst other things). It does this by interpolating |
| 1626 | Perl code in the expression at run time, and the code can refer to the |
| 1627 | expression itself. A Perl pattern using code interpolation to solve the |
| 1628 | parentheses problem can be created like this: |
| 1629 | |
| 1630 | $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x; |
| 1631 | |
| 1632 | The (?p{...}) item interpolates Perl code at run time, and in this case |
| 1633 | refers recursively to the pattern in which it appears. |
| 1634 | |
| 1635 | Obviously, PCRE cannot support the interpolation of Perl code. Instead, |
| 1636 | it supports special syntax for recursion of the entire pattern, and |
| 1637 | also for individual subpattern recursion. After its introduction in |
| 1638 | PCRE and Python, this kind of recursion was introduced into Perl at |
| 1639 | release 5.10. |
| 1640 | |
| 1641 | A special item that consists of (? followed by a number greater than |
| 1642 | zero and a closing parenthesis is a recursive call of the subpattern of |
| 1643 | the given number, provided that it occurs inside that subpattern. (If |
| 1644 | not, it is a "subroutine" call, which is described in the next sec- |
| 1645 | tion.) The special item (?R) or (?0) is a recursive call of the entire |
| 1646 | regular expression. |
| 1647 | |
| 1648 | In PCRE (like Python, but unlike Perl), a recursive subpattern call is |
| 1649 | always treated as an atomic group. That is, once it has matched some of |
| 1650 | the subject string, it is never re-entered, even if it contains untried |
| 1651 | alternatives and there is a subsequent matching failure. |
| 1652 | |
| 1653 | This PCRE pattern solves the nested parentheses problem (assume the |
| 1654 | PCRE_EXTENDED option is set so that white space is ignored): |
| 1655 | |
| 1656 | \( ( (?>[^()]+) | (?R) )* \) |
| 1657 | |
| 1658 | First it matches an opening parenthesis. Then it matches any number of |
| 1659 | substrings which can either be a sequence of non-parentheses, or a |
| 1660 | recursive match of the pattern itself (that is, a correctly parenthe- |
| 1661 | sized substring). Finally there is a closing parenthesis. |
| 1662 | |
| 1663 | If this were part of a larger pattern, you would not want to recurse |
| 1664 | the entire pattern, so instead you could use this: |
| 1665 | |
| 1666 | ( \( ( (?>[^()]+) | (?1) )* \) ) |
| 1667 | |
| 1668 | We have put the pattern into parentheses, and caused the recursion to |
| 1669 | refer to them instead of the whole pattern. |
| 1670 | |
| 1671 | In a larger pattern, keeping track of parenthesis numbers can be |
| 1672 | tricky. This is made easier by the use of relative references. (A Perl |
| 1673 | 5.10 feature.) Instead of (?1) in the pattern above you can write |
| 1674 | (?-2) to refer to the second most recently opened parentheses preceding |
| 1675 | the recursion. In other words, a negative number counts capturing |
| 1676 | parentheses leftwards from the point at which it is encountered. |
| 1677 | |
| 1678 | It is also possible to refer to subsequently opened parentheses, by |
| 1679 | writing references such as (?+2). However, these cannot be recursive |
| 1680 | because the reference is not inside the parentheses that are refer- |
| 1681 | enced. They are always "subroutine" calls, as described in the next |
| 1682 | section. |
| 1683 | |
| 1684 | An alternative approach is to use named parentheses instead. The Perl |
| 1685 | syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also |
| 1686 | supported. We could rewrite the above example as follows: |
| 1687 | |
| 1688 | (?<pn> \( ( (?>[^()]+) | (?&pn) )* \) ) |
| 1689 | |
| 1690 | If there is more than one subpattern with the same name, the earliest |
| 1691 | one is used. |
| 1692 | |
| 1693 | This particular example pattern that we have been looking at contains |
| 1694 | nested unlimited repeats, and so the use of atomic grouping for match- |
| 1695 | ing strings of non-parentheses is important when applying the pattern |
| 1696 | to strings that do not match. For example, when this pattern is applied |
| 1697 | to |
| 1698 | |
| 1699 | (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa() |
| 1700 | |
| 1701 | it yields "no match" quickly. However, if atomic grouping is not used, |
| 1702 | the match runs for a very long time indeed because there are so many |
| 1703 | different ways the + and * repeats can carve up the subject, and all |
| 1704 | have to be tested before failure can be reported. |
| 1705 | |
| 1706 | At the end of a match, the values set for any capturing subpatterns are |
| 1707 | those from the outermost level of the recursion at which the subpattern |
| 1708 | value is set. If you want to obtain intermediate values, a callout |
| 1709 | function can be used (see below and the pcrecallout documentation). If |
| 1710 | the pattern above is matched against |
| 1711 | |
| 1712 | (ab(cd)ef) |
| 1713 | |
| 1714 | the value for the capturing parentheses is "ef", which is the last |
| 1715 | value taken on at the top level. If additional parentheses are added, |
| 1716 | giving |
| 1717 | |
| 1718 | \( ( ( (?>[^()]+) | (?R) )* ) \) |
| 1719 | ^ ^ |
| 1720 | ^ ^ |
| 1721 | |
| 1722 | the string they capture is "ab(cd)ef", the contents of the top level |
| 1723 | parentheses. If there are more than 15 capturing parentheses in a pat- |
| 1724 | tern, PCRE has to obtain extra memory to store data during a recursion, |
| 1725 | which it does by using pcre_malloc, freeing it via pcre_free after- |
| 1726 | wards. If no memory can be obtained, the match fails with the |
| 1727 | PCRE_ERROR_NOMEMORY error. |
| 1728 | |
| 1729 | Do not confuse the (?R) item with the condition (R), which tests for |
| 1730 | recursion. Consider this pattern, which matches text in angle brack- |
| 1731 | ets, allowing for arbitrary nesting. Only digits are allowed in nested |
| 1732 | brackets (that is, when recursing), whereas any characters are permit- |
| 1733 | ted at the outer level. |
| 1734 | |
| 1735 | < (?: (?(R) \d++ | [^<>]*+) | (?R)) * > |
| 1736 | |
| 1737 | In this pattern, (?(R) is the start of a conditional subpattern, with |
| 1738 | two different alternatives for the recursive and non-recursive cases. |
| 1739 | The (?R) item is the actual recursive call. |
| 1740 | |
| 1741 | |
| 1742 | SUBPATTERNS AS SUBROUTINES |
| 1743 | |
| 1744 | If the syntax for a recursive subpattern reference (either by number or |
| 1745 | by name) is used outside the parentheses to which it refers, it oper- |
| 1746 | ates like a subroutine in a programming language. The "called" subpat- |
| 1747 | tern may be defined before or after the reference. A numbered reference |
| 1748 | can be absolute or relative, as in these examples: |
| 1749 | |
| 1750 | (...(absolute)...)...(?2)... |
| 1751 | (...(relative)...)...(?-1)... |
| 1752 | (...(?+1)...(relative)... |
| 1753 | |
| 1754 | An earlier example pointed out that the pattern |
| 1755 | |
| 1756 | (sens|respons)e and \1ibility |
| 1757 | |
| 1758 | matches "sense and sensibility" and "response and responsibility", but |
| 1759 | not "sense and responsibility". If instead the pattern |
| 1760 | |
| 1761 | (sens|respons)e and (?1)ibility |
| 1762 | |
| 1763 | is used, it does match "sense and responsibility" as well as the other |
| 1764 | two strings. Another example is given in the discussion of DEFINE |
| 1765 | above. |
| 1766 | |
| 1767 | Like recursive subpatterns, a "subroutine" call is always treated as an |
| 1768 | atomic group. That is, once it has matched some of the subject string, |
| 1769 | it is never re-entered, even if it contains untried alternatives and |
| 1770 | there is a subsequent matching failure. |
| 1771 | |
| 1772 | When a subpattern is used as a subroutine, processing options such as |
| 1773 | case-independence are fixed when the subpattern is defined. They cannot |
| 1774 | be changed for different calls. For example, consider this pattern: |
| 1775 | |
| 1776 | (abc)(?i:(?-1)) |
| 1777 | |
| 1778 | It matches "abcabc". It does not match "abcABC" because the change of |
| 1779 | processing option does not affect the called subpattern. |
| 1780 | |
| 1781 | |
| 1782 | CALLOUTS |
| 1783 | |
| 1784 | Perl has a feature whereby using the sequence (?{...}) causes arbitrary |
| 1785 | Perl code to be obeyed in the middle of matching a regular expression. |
| 1786 | This makes it possible, amongst other things, to extract different sub- |
| 1787 | strings that match the same pair of parentheses when there is a repeti- |
| 1788 | tion. |
| 1789 | |
| 1790 | PCRE provides a similar feature, but of course it cannot obey arbitrary |
| 1791 | Perl code. The feature is called "callout". The caller of PCRE provides |
| 1792 | an external function by putting its entry point in the global variable |
| 1793 | pcre_callout. By default, this variable contains NULL, which disables |
| 1794 | all calling out. |
| 1795 | |
| 1796 | Within a regular expression, (?C) indicates the points at which the |
| 1797 | external function is to be called. If you want to identify different |
| 1798 | callout points, you can put a number less than 256 after the letter C. |
| 1799 | The default value is zero. For example, this pattern has two callout |
| 1800 | points: |
| 1801 | |
| 1802 | (?C1)abc(?C2)def |
| 1803 | |
| 1804 | If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are |
| 1805 | automatically installed before each item in the pattern. They are all |
| 1806 | numbered 255. |
| 1807 | |
| 1808 | During matching, when PCRE reaches a callout point (and pcre_callout is |
| 1809 | set), the external function is called. It is provided with the number |
| 1810 | of the callout, the position in the pattern, and, optionally, one item |
| 1811 | of data originally supplied by the caller of pcre_exec(). The callout |
| 1812 | function may cause matching to proceed, to backtrack, or to fail alto- |
| 1813 | gether. A complete description of the interface to the callout function |
| 1814 | is given in the pcrecallout documentation. |
| 1815 | |
| 1816 | |
| 1817 | SEE ALSO |
| 1818 | |
| 1819 | pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3). |
| 1820 | |
| 1821 | |
| 1822 | AUTHOR |
| 1823 | |
| 1824 | Philip Hazel |
| 1825 | University Computing Service |
| 1826 | Cambridge CB2 3QH, England. |
| 1827 | |
| 1828 | |
| 1829 | REVISION |
| 1830 | |
| 1831 | Last updated: 19 June 2007 |
| 1832 | Copyright (c) 1997-2007 University of Cambridge. |