Tristan Matthews | 0461646 | 2013-11-14 16:09:34 -0500 | [diff] [blame] | 1 | ----------------------------------------------------------------------------- |
| 2 | This file contains a concatenation of the PCRE man pages, converted to plain |
| 3 | text format for ease of searching with a text editor, or for use on systems |
| 4 | that do not have a man page processor. The small individual files that give |
| 5 | synopses of each function in the library have not been included. Neither has |
| 6 | the pcredemo program. There are separate text files for the pcregrep and |
| 7 | pcretest commands. |
| 8 | ----------------------------------------------------------------------------- |
| 9 | |
| 10 | |
| 11 | PCRE(3) PCRE(3) |
| 12 | |
| 13 | |
| 14 | NAME |
| 15 | PCRE - Perl-compatible regular expressions |
| 16 | |
| 17 | |
| 18 | INTRODUCTION |
| 19 | |
| 20 | The PCRE library is a set of functions that implement regular expres- |
| 21 | sion pattern matching using the same syntax and semantics as Perl, with |
| 22 | just a few differences. Some features that appeared in Python and PCRE |
| 23 | before they appeared in Perl are also available using the Python syn- |
| 24 | tax, there is some support for one or two .NET and Oniguruma syntax |
| 25 | items, and there is an option for requesting some minor changes that |
| 26 | give better JavaScript compatibility. |
| 27 | |
| 28 | The current implementation of PCRE corresponds approximately with Perl |
| 29 | 5.12, including support for UTF-8 encoded strings and Unicode general |
| 30 | category properties. However, UTF-8 and Unicode support has to be |
| 31 | explicitly enabled; it is not the default. The Unicode tables corre- |
| 32 | spond to Unicode release 6.0.0. |
| 33 | |
| 34 | In addition to the Perl-compatible matching function, PCRE contains an |
| 35 | alternative function that matches the same compiled patterns in a dif- |
| 36 | ferent way. In certain circumstances, the alternative function has some |
| 37 | advantages. For a discussion of the two matching algorithms, see the |
| 38 | pcrematching page. |
| 39 | |
| 40 | PCRE is written in C and released as a C library. A number of people |
| 41 | have written wrappers and interfaces of various kinds. In particular, |
| 42 | Google Inc. have provided a comprehensive C++ wrapper. This is now |
| 43 | included as part of the PCRE distribution. The pcrecpp page has details |
| 44 | of this interface. Other people's contributions can be found in the |
| 45 | Contrib directory at the primary FTP site, which is: |
| 46 | |
| 47 | ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre |
| 48 | |
| 49 | Details of exactly which Perl regular expression features are and are |
| 50 | not supported by PCRE are given in separate documents. See the pcrepat- |
| 51 | tern and pcrecompat pages. There is a syntax summary in the pcresyntax |
| 52 | page. |
| 53 | |
| 54 | Some features of PCRE can be included, excluded, or changed when the |
| 55 | library is built. The pcre_config() function makes it possible for a |
| 56 | client to discover which features are available. The features them- |
| 57 | selves are described in the pcrebuild page. Documentation about build- |
| 58 | ing PCRE for various operating systems can be found in the README and |
| 59 | NON-UNIX-USE files in the source distribution. |
| 60 | |
| 61 | The library contains a number of undocumented internal functions and |
| 62 | data tables that are used by more than one of the exported external |
| 63 | functions, but which are not intended for use by external callers. |
| 64 | Their names all begin with "_pcre_", which hopefully will not provoke |
| 65 | any name clashes. In some environments, it is possible to control which |
| 66 | external symbols are exported when a shared library is built, and in |
| 67 | these cases the undocumented symbols are not exported. |
| 68 | |
| 69 | |
| 70 | USER DOCUMENTATION |
| 71 | |
| 72 | The user documentation for PCRE comprises a number of different sec- |
| 73 | tions. In the "man" format, each of these is a separate "man page". In |
| 74 | the HTML format, each is a separate page, linked from the index page. |
| 75 | In the plain text format, all the sections, except the pcredemo sec- |
| 76 | tion, are concatenated, for ease of searching. The sections are as fol- |
| 77 | lows: |
| 78 | |
| 79 | pcre this document |
| 80 | pcre-config show PCRE installation configuration information |
| 81 | pcreapi details of PCRE's native C API |
| 82 | pcrebuild options for building PCRE |
| 83 | pcrecallout details of the callout feature |
| 84 | pcrecompat discussion of Perl compatibility |
| 85 | pcrecpp details of the C++ wrapper |
| 86 | pcredemo a demonstration C program that uses PCRE |
| 87 | pcregrep description of the pcregrep command |
| 88 | pcrejit discussion of the just-in-time optimization support |
| 89 | pcrelimits details of size and other limits |
| 90 | pcrematching discussion of the two matching algorithms |
| 91 | pcrepartial details of the partial matching facility |
| 92 | pcrepattern syntax and semantics of supported |
| 93 | regular expressions |
| 94 | pcreperform discussion of performance issues |
| 95 | pcreposix the POSIX-compatible C API |
| 96 | pcreprecompile details of saving and re-using precompiled patterns |
| 97 | pcresample discussion of the pcredemo program |
| 98 | pcrestack discussion of stack usage |
| 99 | pcresyntax quick syntax reference |
| 100 | pcretest description of the pcretest testing command |
| 101 | pcreunicode discussion of Unicode and UTF-8 support |
| 102 | |
| 103 | In addition, in the "man" and HTML formats, there is a short page for |
| 104 | each C library function, listing its arguments and results. |
| 105 | |
| 106 | |
| 107 | AUTHOR |
| 108 | |
| 109 | Philip Hazel |
| 110 | University Computing Service |
| 111 | Cambridge CB2 3QH, England. |
| 112 | |
| 113 | Putting an actual email address here seems to have been a spam magnet, |
| 114 | so I've taken it away. If you want to email me, use my two initials, |
| 115 | followed by the two digits 10, at the domain cam.ac.uk. |
| 116 | |
| 117 | |
| 118 | REVISION |
| 119 | |
| 120 | Last updated: 24 August 2011 |
| 121 | Copyright (c) 1997-2011 University of Cambridge. |
| 122 | ------------------------------------------------------------------------------ |
| 123 | |
| 124 | |
| 125 | PCREBUILD(3) PCREBUILD(3) |
| 126 | |
| 127 | |
| 128 | NAME |
| 129 | PCRE - Perl-compatible regular expressions |
| 130 | |
| 131 | |
| 132 | PCRE BUILD-TIME OPTIONS |
| 133 | |
| 134 | This document describes the optional features of PCRE that can be |
| 135 | selected when the library is compiled. It assumes use of the configure |
| 136 | script, where the optional features are selected or deselected by pro- |
| 137 | viding options to configure before running the make command. However, |
| 138 | the same options can be selected in both Unix-like and non-Unix-like |
| 139 | environments using the GUI facility of cmake-gui if you are using CMake |
| 140 | instead of configure to build PCRE. |
| 141 | |
| 142 | There is a lot more information about building PCRE in non-Unix-like |
| 143 | environments in the file called NON_UNIX_USE, which is part of the PCRE |
| 144 | distribution. You should consult this file as well as the README file |
| 145 | if you are building in a non-Unix-like environment. |
| 146 | |
| 147 | The complete list of options for configure (which includes the standard |
| 148 | ones such as the selection of the installation directory) can be |
| 149 | obtained by running |
| 150 | |
| 151 | ./configure --help |
| 152 | |
| 153 | The following sections include descriptions of options whose names |
| 154 | begin with --enable or --disable. These settings specify changes to the |
| 155 | defaults for the configure command. Because of the way that configure |
| 156 | works, --enable and --disable always come in pairs, so the complemen- |
| 157 | tary option always exists as well, but as it specifies the default, it |
| 158 | is not described. |
| 159 | |
| 160 | |
| 161 | BUILDING SHARED AND STATIC LIBRARIES |
| 162 | |
| 163 | The PCRE building process uses libtool to build both shared and static |
| 164 | Unix libraries by default. You can suppress one of these by adding one |
| 165 | of |
| 166 | |
| 167 | --disable-shared |
| 168 | --disable-static |
| 169 | |
| 170 | to the configure command, as required. |
| 171 | |
| 172 | |
| 173 | C++ SUPPORT |
| 174 | |
| 175 | By default, the configure script will search for a C++ compiler and C++ |
| 176 | header files. If it finds them, it automatically builds the C++ wrapper |
| 177 | library for PCRE. You can disable this by adding |
| 178 | |
| 179 | --disable-cpp |
| 180 | |
| 181 | to the configure command. |
| 182 | |
| 183 | |
| 184 | UTF-8 SUPPORT |
| 185 | |
| 186 | To build PCRE with support for UTF-8 Unicode character strings, add |
| 187 | |
| 188 | --enable-utf8 |
| 189 | |
| 190 | to the configure command. Of itself, this does not make PCRE treat |
| 191 | strings as UTF-8. As well as compiling PCRE with this option, you also |
| 192 | have have to set the PCRE_UTF8 option when you call the pcre_compile() |
| 193 | or pcre_compile2() functions. |
| 194 | |
| 195 | If you set --enable-utf8 when compiling in an EBCDIC environment, PCRE |
| 196 | expects its input to be either ASCII or UTF-8 (depending on the runtime |
| 197 | option). It is not possible to support both EBCDIC and UTF-8 codes in |
| 198 | the same version of the library. Consequently, --enable-utf8 and |
| 199 | --enable-ebcdic are mutually exclusive. |
| 200 | |
| 201 | |
| 202 | UNICODE CHARACTER PROPERTY SUPPORT |
| 203 | |
| 204 | UTF-8 support allows PCRE to process character values greater than 255 |
| 205 | in the strings that it handles. On its own, however, it does not pro- |
| 206 | vide any facilities for accessing the properties of such characters. If |
| 207 | you want to be able to use the pattern escapes \P, \p, and \X, which |
| 208 | refer to Unicode character properties, you must add |
| 209 | |
| 210 | --enable-unicode-properties |
| 211 | |
| 212 | to the configure command. This implies UTF-8 support, even if you have |
| 213 | not explicitly requested it. |
| 214 | |
| 215 | Including Unicode property support adds around 30K of tables to the |
| 216 | PCRE library. Only the general category properties such as Lu and Nd |
| 217 | are supported. Details are given in the pcrepattern documentation. |
| 218 | |
| 219 | |
| 220 | JUST-IN-TIME COMPILER SUPPORT |
| 221 | |
| 222 | Just-in-time compiler support is included in the build by specifying |
| 223 | |
| 224 | --enable-jit |
| 225 | |
| 226 | This support is available only for certain hardware architectures. If |
| 227 | this option is set for an unsupported architecture, a compile time |
| 228 | error occurs. See the pcrejit documentation for a discussion of JIT |
| 229 | usage. When JIT support is enabled, pcregrep automatically makes use of |
| 230 | it, unless you add |
| 231 | |
| 232 | --disable-pcregrep-jit |
| 233 | |
| 234 | to the "configure" command. |
| 235 | |
| 236 | |
| 237 | CODE VALUE OF NEWLINE |
| 238 | |
| 239 | By default, PCRE interprets the linefeed (LF) character as indicating |
| 240 | the end of a line. This is the normal newline character on Unix-like |
| 241 | systems. You can compile PCRE to use carriage return (CR) instead, by |
| 242 | adding |
| 243 | |
| 244 | --enable-newline-is-cr |
| 245 | |
| 246 | to the configure command. There is also a --enable-newline-is-lf |
| 247 | option, which explicitly specifies linefeed as the newline character. |
| 248 | |
| 249 | Alternatively, you can specify that line endings are to be indicated by |
| 250 | the two character sequence CRLF. If you want this, add |
| 251 | |
| 252 | --enable-newline-is-crlf |
| 253 | |
| 254 | to the configure command. There is a fourth option, specified by |
| 255 | |
| 256 | --enable-newline-is-anycrlf |
| 257 | |
| 258 | which causes PCRE to recognize any of the three sequences CR, LF, or |
| 259 | CRLF as indicating a line ending. Finally, a fifth option, specified by |
| 260 | |
| 261 | --enable-newline-is-any |
| 262 | |
| 263 | causes PCRE to recognize any Unicode newline sequence. |
| 264 | |
| 265 | Whatever line ending convention is selected when PCRE is built can be |
| 266 | overridden when the library functions are called. At build time it is |
| 267 | conventional to use the standard for your operating system. |
| 268 | |
| 269 | |
| 270 | WHAT \R MATCHES |
| 271 | |
| 272 | By default, the sequence \R in a pattern matches any Unicode newline |
| 273 | sequence, whatever has been selected as the line ending sequence. If |
| 274 | you specify |
| 275 | |
| 276 | --enable-bsr-anycrlf |
| 277 | |
| 278 | the default is changed so that \R matches only CR, LF, or CRLF. What- |
| 279 | ever is selected when PCRE is built can be overridden when the library |
| 280 | functions are called. |
| 281 | |
| 282 | |
| 283 | POSIX MALLOC USAGE |
| 284 | |
| 285 | When PCRE is called through the POSIX interface (see the pcreposix doc- |
| 286 | umentation), additional working storage is required for holding the |
| 287 | pointers to capturing substrings, because PCRE requires three integers |
| 288 | per substring, whereas the POSIX interface provides only two. If the |
| 289 | number of expected substrings is small, the wrapper function uses space |
| 290 | on the stack, because this is faster than using malloc() for each call. |
| 291 | The default threshold above which the stack is no longer used is 10; it |
| 292 | can be changed by adding a setting such as |
| 293 | |
| 294 | --with-posix-malloc-threshold=20 |
| 295 | |
| 296 | to the configure command. |
| 297 | |
| 298 | |
| 299 | HANDLING VERY LARGE PATTERNS |
| 300 | |
| 301 | Within a compiled pattern, offset values are used to point from one |
| 302 | part to another (for example, from an opening parenthesis to an alter- |
| 303 | nation metacharacter). By default, two-byte values are used for these |
| 304 | offsets, leading to a maximum size for a compiled pattern of around |
| 305 | 64K. This is sufficient to handle all but the most gigantic patterns. |
| 306 | Nevertheless, some people do want to process truyl enormous patterns, |
| 307 | so it is possible to compile PCRE to use three-byte or four-byte off- |
| 308 | sets by adding a setting such as |
| 309 | |
| 310 | --with-link-size=3 |
| 311 | |
| 312 | to the configure command. The value given must be 2, 3, or 4. Using |
| 313 | longer offsets slows down the operation of PCRE because it has to load |
| 314 | additional bytes when handling them. |
| 315 | |
| 316 | |
| 317 | AVOIDING EXCESSIVE STACK USAGE |
| 318 | |
| 319 | When matching with the pcre_exec() function, PCRE implements backtrack- |
| 320 | ing by making recursive calls to an internal function called match(). |
| 321 | In environments where the size of the stack is limited, this can se- |
| 322 | verely limit PCRE's operation. (The Unix environment does not usually |
| 323 | suffer from this problem, but it may sometimes be necessary to increase |
| 324 | the maximum stack size. There is a discussion in the pcrestack docu- |
| 325 | mentation.) An alternative approach to recursion that uses memory from |
| 326 | the heap to remember data, instead of using recursive function calls, |
| 327 | has been implemented to work round the problem of limited stack size. |
| 328 | If you want to build a version of PCRE that works this way, add |
| 329 | |
| 330 | --disable-stack-for-recursion |
| 331 | |
| 332 | to the configure command. With this configuration, PCRE will use the |
| 333 | pcre_stack_malloc and pcre_stack_free variables to call memory manage- |
| 334 | ment functions. By default these point to malloc() and free(), but you |
| 335 | can replace the pointers so that your own functions are used instead. |
| 336 | |
| 337 | Separate functions are provided rather than using pcre_malloc and |
| 338 | pcre_free because the usage is very predictable: the block sizes |
| 339 | requested are always the same, and the blocks are always freed in |
| 340 | reverse order. A calling program might be able to implement optimized |
| 341 | functions that perform better than malloc() and free(). PCRE runs |
| 342 | noticeably more slowly when built in this way. This option affects only |
| 343 | the pcre_exec() function; it is not relevant for pcre_dfa_exec(). |
| 344 | |
| 345 | |
| 346 | LIMITING PCRE RESOURCE USAGE |
| 347 | |
| 348 | Internally, PCRE has a function called match(), which it calls repeat- |
| 349 | edly (sometimes recursively) when matching a pattern with the |
| 350 | pcre_exec() function. By controlling the maximum number of times this |
| 351 | function may be called during a single matching operation, a limit can |
| 352 | be placed on the resources used by a single call to pcre_exec(). The |
| 353 | limit can be changed at run time, as described in the pcreapi documen- |
| 354 | tation. The default is 10 million, but this can be changed by adding a |
| 355 | setting such as |
| 356 | |
| 357 | --with-match-limit=500000 |
| 358 | |
| 359 | to the configure command. This setting has no effect on the |
| 360 | pcre_dfa_exec() matching function. |
| 361 | |
| 362 | In some environments it is desirable to limit the depth of recursive |
| 363 | calls of match() more strictly than the total number of calls, in order |
| 364 | to restrict the maximum amount of stack (or heap, if --disable-stack- |
| 365 | for-recursion is specified) that is used. A second limit controls this; |
| 366 | it defaults to the value that is set for --with-match-limit, which |
| 367 | imposes no additional constraints. However, you can set a lower limit |
| 368 | by adding, for example, |
| 369 | |
| 370 | --with-match-limit-recursion=10000 |
| 371 | |
| 372 | to the configure command. This value can also be overridden at run |
| 373 | time. |
| 374 | |
| 375 | |
| 376 | CREATING CHARACTER TABLES AT BUILD TIME |
| 377 | |
| 378 | PCRE uses fixed tables for processing characters whose code values are |
| 379 | less than 256. By default, PCRE is built with a set of tables that are |
| 380 | distributed in the file pcre_chartables.c.dist. These tables are for |
| 381 | ASCII codes only. If you add |
| 382 | |
| 383 | --enable-rebuild-chartables |
| 384 | |
| 385 | to the configure command, the distributed tables are no longer used. |
| 386 | Instead, a program called dftables is compiled and run. This outputs |
| 387 | the source for new set of tables, created in the default locale of your |
| 388 | C runtime system. (This method of replacing the tables does not work if |
| 389 | you are cross compiling, because dftables is run on the local host. If |
| 390 | you need to create alternative tables when cross compiling, you will |
| 391 | have to do so "by hand".) |
| 392 | |
| 393 | |
| 394 | USING EBCDIC CODE |
| 395 | |
| 396 | PCRE assumes by default that it will run in an environment where the |
| 397 | character code is ASCII (or Unicode, which is a superset of ASCII). |
| 398 | This is the case for most computer operating systems. PCRE can, how- |
| 399 | ever, be compiled to run in an EBCDIC environment by adding |
| 400 | |
| 401 | --enable-ebcdic |
| 402 | |
| 403 | to the configure command. This setting implies --enable-rebuild-charta- |
| 404 | bles. You should only use it if you know that you are in an EBCDIC |
| 405 | environment (for example, an IBM mainframe operating system). The |
| 406 | --enable-ebcdic option is incompatible with --enable-utf8. |
| 407 | |
| 408 | |
| 409 | PCREGREP OPTIONS FOR COMPRESSED FILE SUPPORT |
| 410 | |
| 411 | By default, pcregrep reads all files as plain text. You can build it so |
| 412 | that it recognizes files whose names end in .gz or .bz2, and reads them |
| 413 | with libz or libbz2, respectively, by adding one or both of |
| 414 | |
| 415 | --enable-pcregrep-libz |
| 416 | --enable-pcregrep-libbz2 |
| 417 | |
| 418 | to the configure command. These options naturally require that the rel- |
| 419 | evant libraries are installed on your system. Configuration will fail |
| 420 | if they are not. |
| 421 | |
| 422 | |
| 423 | PCREGREP BUFFER SIZE |
| 424 | |
| 425 | pcregrep uses an internal buffer to hold a "window" on the file it is |
| 426 | scanning, in order to be able to output "before" and "after" lines when |
| 427 | it finds a match. The size of the buffer is controlled by a parameter |
| 428 | whose default value is 20K. The buffer itself is three times this size, |
| 429 | but because of the way it is used for holding "before" lines, the long- |
| 430 | est line that is guaranteed to be processable is the parameter size. |
| 431 | You can change the default parameter value by adding, for example, |
| 432 | |
| 433 | --with-pcregrep-bufsize=50K |
| 434 | |
| 435 | to the configure command. The caller of pcregrep can, however, override |
| 436 | this value by specifying a run-time option. |
| 437 | |
| 438 | |
| 439 | PCRETEST OPTION FOR LIBREADLINE SUPPORT |
| 440 | |
| 441 | If you add |
| 442 | |
| 443 | --enable-pcretest-libreadline |
| 444 | |
| 445 | to the configure command, pcretest is linked with the libreadline |
| 446 | library, and when its input is from a terminal, it reads it using the |
| 447 | readline() function. This provides line-editing and history facilities. |
| 448 | Note that libreadline is GPL-licensed, so if you distribute a binary of |
| 449 | pcretest linked in this way, there may be licensing issues. |
| 450 | |
| 451 | Setting this option causes the -lreadline option to be added to the |
| 452 | pcretest build. In many operating environments with a sytem-installed |
| 453 | libreadline this is sufficient. However, in some environments (e.g. if |
| 454 | an unmodified distribution version of readline is in use), some extra |
| 455 | configuration may be necessary. The INSTALL file for libreadline says |
| 456 | this: |
| 457 | |
| 458 | "Readline uses the termcap functions, but does not link with the |
| 459 | termcap or curses library itself, allowing applications which link |
| 460 | with readline the to choose an appropriate library." |
| 461 | |
| 462 | If your environment has not been set up so that an appropriate library |
| 463 | is automatically included, you may need to add something like |
| 464 | |
| 465 | LIBS="-ncurses" |
| 466 | |
| 467 | immediately before the configure command. |
| 468 | |
| 469 | |
| 470 | SEE ALSO |
| 471 | |
| 472 | pcreapi(3), pcre_config(3). |
| 473 | |
| 474 | |
| 475 | AUTHOR |
| 476 | |
| 477 | Philip Hazel |
| 478 | University Computing Service |
| 479 | Cambridge CB2 3QH, England. |
| 480 | |
| 481 | |
| 482 | REVISION |
| 483 | |
| 484 | Last updated: 06 September 2011 |
| 485 | Copyright (c) 1997-2011 University of Cambridge. |
| 486 | ------------------------------------------------------------------------------ |
| 487 | |
| 488 | |
| 489 | PCREMATCHING(3) PCREMATCHING(3) |
| 490 | |
| 491 | |
| 492 | NAME |
| 493 | PCRE - Perl-compatible regular expressions |
| 494 | |
| 495 | |
| 496 | PCRE MATCHING ALGORITHMS |
| 497 | |
| 498 | This document describes the two different algorithms that are available |
| 499 | in PCRE for matching a compiled regular expression against a given sub- |
| 500 | ject string. The "standard" algorithm is the one provided by the |
| 501 | pcre_exec() function. This works in the same was as Perl's matching |
| 502 | function, and provides a Perl-compatible matching operation. |
| 503 | |
| 504 | An alternative algorithm is provided by the pcre_dfa_exec() function; |
| 505 | this operates in a different way, and is not Perl-compatible. It has |
| 506 | advantages and disadvantages compared with the standard algorithm, and |
| 507 | these are described below. |
| 508 | |
| 509 | When there is only one possible way in which a given subject string can |
| 510 | match a pattern, the two algorithms give the same answer. A difference |
| 511 | arises, however, when there are multiple possibilities. For example, if |
| 512 | the pattern |
| 513 | |
| 514 | ^<.*> |
| 515 | |
| 516 | is matched against the string |
| 517 | |
| 518 | <something> <something else> <something further> |
| 519 | |
| 520 | there are three possible answers. The standard algorithm finds only one |
| 521 | of them, whereas the alternative algorithm finds all three. |
| 522 | |
| 523 | |
| 524 | REGULAR EXPRESSIONS AS TREES |
| 525 | |
| 526 | The set of strings that are matched by a regular expression can be rep- |
| 527 | resented as a tree structure. An unlimited repetition in the pattern |
| 528 | makes the tree of infinite size, but it is still a tree. Matching the |
| 529 | pattern to a given subject string (from a given starting point) can be |
| 530 | thought of as a search of the tree. There are two ways to search a |
| 531 | tree: depth-first and breadth-first, and these correspond to the two |
| 532 | matching algorithms provided by PCRE. |
| 533 | |
| 534 | |
| 535 | THE STANDARD MATCHING ALGORITHM |
| 536 | |
| 537 | In the terminology of Jeffrey Friedl's book "Mastering Regular Expres- |
| 538 | sions", the standard algorithm is an "NFA algorithm". It conducts a |
| 539 | depth-first search of the pattern tree. That is, it proceeds along a |
| 540 | single path through the tree, checking that the subject matches what is |
| 541 | required. When there is a mismatch, the algorithm tries any alterna- |
| 542 | tives at the current point, and if they all fail, it backs up to the |
| 543 | previous branch point in the tree, and tries the next alternative |
| 544 | branch at that level. This often involves backing up (moving to the |
| 545 | left) in the subject string as well. The order in which repetition |
| 546 | branches are tried is controlled by the greedy or ungreedy nature of |
| 547 | the quantifier. |
| 548 | |
| 549 | If a leaf node is reached, a matching string has been found, and at |
| 550 | that point the algorithm stops. Thus, if there is more than one possi- |
| 551 | ble match, this algorithm returns the first one that it finds. Whether |
| 552 | this is the shortest, the longest, or some intermediate length depends |
| 553 | on the way the greedy and ungreedy repetition quantifiers are specified |
| 554 | in the pattern. |
| 555 | |
| 556 | Because it ends up with a single path through the tree, it is rela- |
| 557 | tively straightforward for this algorithm to keep track of the sub- |
| 558 | strings that are matched by portions of the pattern in parentheses. |
| 559 | This provides support for capturing parentheses and back references. |
| 560 | |
| 561 | |
| 562 | THE ALTERNATIVE MATCHING ALGORITHM |
| 563 | |
| 564 | This algorithm conducts a breadth-first search of the tree. Starting |
| 565 | from the first matching point in the subject, it scans the subject |
| 566 | string from left to right, once, character by character, and as it does |
| 567 | this, it remembers all the paths through the tree that represent valid |
| 568 | matches. In Friedl's terminology, this is a kind of "DFA algorithm", |
| 569 | though it is not implemented as a traditional finite state machine (it |
| 570 | keeps multiple states active simultaneously). |
| 571 | |
| 572 | Although the general principle of this matching algorithm is that it |
| 573 | scans the subject string only once, without backtracking, there is one |
| 574 | exception: when a lookaround assertion is encountered, the characters |
| 575 | following or preceding the current point have to be independently |
| 576 | inspected. |
| 577 | |
| 578 | The scan continues until either the end of the subject is reached, or |
| 579 | there are no more unterminated paths. At this point, terminated paths |
| 580 | represent the different matching possibilities (if there are none, the |
| 581 | match has failed). Thus, if there is more than one possible match, |
| 582 | this algorithm finds all of them, and in particular, it finds the long- |
| 583 | est. The matches are returned in decreasing order of length. There is |
| 584 | an option to stop the algorithm after the first match (which is neces- |
| 585 | sarily the shortest) is found. |
| 586 | |
| 587 | Note that all the matches that are found start at the same point in the |
| 588 | subject. If the pattern |
| 589 | |
| 590 | cat(er(pillar)?)? |
| 591 | |
| 592 | is matched against the string "the caterpillar catchment", the result |
| 593 | will be the three strings "caterpillar", "cater", and "cat" that start |
| 594 | at the fifth character of the subject. The algorithm does not automati- |
| 595 | cally move on to find matches that start at later positions. |
| 596 | |
| 597 | There are a number of features of PCRE regular expressions that are not |
| 598 | supported by the alternative matching algorithm. They are as follows: |
| 599 | |
| 600 | 1. Because the algorithm finds all possible matches, the greedy or |
| 601 | ungreedy nature of repetition quantifiers is not relevant. Greedy and |
| 602 | ungreedy quantifiers are treated in exactly the same way. However, pos- |
| 603 | sessive quantifiers can make a difference when what follows could also |
| 604 | match what is quantified, for example in a pattern like this: |
| 605 | |
| 606 | ^a++\w! |
| 607 | |
| 608 | This pattern matches "aaab!" but not "aaa!", which would be matched by |
| 609 | a non-possessive quantifier. Similarly, if an atomic group is present, |
| 610 | it is matched as if it were a standalone pattern at the current point, |
| 611 | and the longest match is then "locked in" for the rest of the overall |
| 612 | pattern. |
| 613 | |
| 614 | 2. When dealing with multiple paths through the tree simultaneously, it |
| 615 | is not straightforward to keep track of captured substrings for the |
| 616 | different matching possibilities, and PCRE's implementation of this |
| 617 | algorithm does not attempt to do this. This means that no captured sub- |
| 618 | strings are available. |
| 619 | |
| 620 | 3. Because no substrings are captured, back references within the pat- |
| 621 | tern are not supported, and cause errors if encountered. |
| 622 | |
| 623 | 4. For the same reason, conditional expressions that use a backrefer- |
| 624 | ence as the condition or test for a specific group recursion are not |
| 625 | supported. |
| 626 | |
| 627 | 5. Because many paths through the tree may be active, the \K escape |
| 628 | sequence, which resets the start of the match when encountered (but may |
| 629 | be on some paths and not on others), is not supported. It causes an |
| 630 | error if encountered. |
| 631 | |
| 632 | 6. Callouts are supported, but the value of the capture_top field is |
| 633 | always 1, and the value of the capture_last field is always -1. |
| 634 | |
| 635 | 7. The \C escape sequence, which (in the standard algorithm) matches a |
| 636 | single byte, even in UTF-8 mode, is not supported in UTF-8 mode, |
| 637 | because the alternative algorithm moves through the subject string one |
| 638 | character at a time, for all active paths through the tree. |
| 639 | |
| 640 | 8. Except for (*FAIL), the backtracking control verbs such as (*PRUNE) |
| 641 | are not supported. (*FAIL) is supported, and behaves like a failing |
| 642 | negative assertion. |
| 643 | |
| 644 | |
| 645 | ADVANTAGES OF THE ALTERNATIVE ALGORITHM |
| 646 | |
| 647 | Using the alternative matching algorithm provides the following advan- |
| 648 | tages: |
| 649 | |
| 650 | 1. All possible matches (at a single point in the subject) are automat- |
| 651 | ically found, and in particular, the longest match is found. To find |
| 652 | more than one match using the standard algorithm, you have to do kludgy |
| 653 | things with callouts. |
| 654 | |
| 655 | 2. Because the alternative algorithm scans the subject string just |
| 656 | once, and never needs to backtrack, it is possible to pass very long |
| 657 | subject strings to the matching function in several pieces, checking |
| 658 | for partial matching each time. Although it is possible to do multi- |
| 659 | segment matching using the standard algorithm (pcre_exec()), by retain- |
| 660 | ing partially matched substrings, it is more complicated. The pcrepar- |
| 661 | tial documentation gives details of partial matching and discusses |
| 662 | multi-segment matching. |
| 663 | |
| 664 | |
| 665 | DISADVANTAGES OF THE ALTERNATIVE ALGORITHM |
| 666 | |
| 667 | The alternative algorithm suffers from a number of disadvantages: |
| 668 | |
| 669 | 1. It is substantially slower than the standard algorithm. This is |
| 670 | partly because it has to search for all possible matches, but is also |
| 671 | because it is less susceptible to optimization. |
| 672 | |
| 673 | 2. Capturing parentheses and back references are not supported. |
| 674 | |
| 675 | 3. Although atomic groups are supported, their use does not provide the |
| 676 | performance advantage that it does for the standard algorithm. |
| 677 | |
| 678 | |
| 679 | AUTHOR |
| 680 | |
| 681 | Philip Hazel |
| 682 | University Computing Service |
| 683 | Cambridge CB2 3QH, England. |
| 684 | |
| 685 | |
| 686 | REVISION |
| 687 | |
| 688 | Last updated: 19 November 2011 |
| 689 | Copyright (c) 1997-2010 University of Cambridge. |
| 690 | ------------------------------------------------------------------------------ |
| 691 | |
| 692 | |
| 693 | PCREAPI(3) PCREAPI(3) |
| 694 | |
| 695 | |
| 696 | NAME |
| 697 | PCRE - Perl-compatible regular expressions |
| 698 | |
| 699 | |
| 700 | PCRE NATIVE API BASIC FUNCTIONS |
| 701 | |
| 702 | #include <pcre.h> |
| 703 | |
| 704 | pcre *pcre_compile(const char *pattern, int options, |
| 705 | const char **errptr, int *erroffset, |
| 706 | const unsigned char *tableptr); |
| 707 | |
| 708 | pcre *pcre_compile2(const char *pattern, int options, |
| 709 | int *errorcodeptr, |
| 710 | const char **errptr, int *erroffset, |
| 711 | const unsigned char *tableptr); |
| 712 | |
| 713 | pcre_extra *pcre_study(const pcre *code, int options, |
| 714 | const char **errptr); |
| 715 | |
| 716 | void pcre_free_study(pcre_extra *extra); |
| 717 | |
| 718 | int pcre_exec(const pcre *code, const pcre_extra *extra, |
| 719 | const char *subject, int length, int startoffset, |
| 720 | int options, int *ovector, int ovecsize); |
| 721 | |
| 722 | |
| 723 | PCRE NATIVE API AUXILIARY FUNCTIONS |
| 724 | |
| 725 | pcre_jit_stack *pcre_jit_stack_alloc(int startsize, int maxsize); |
| 726 | |
| 727 | void pcre_jit_stack_free(pcre_jit_stack *stack); |
| 728 | |
| 729 | void pcre_assign_jit_stack(pcre_extra *extra, |
| 730 | pcre_jit_callback callback, void *data); |
| 731 | |
| 732 | int pcre_dfa_exec(const pcre *code, const pcre_extra *extra, |
| 733 | const char *subject, int length, int startoffset, |
| 734 | int options, int *ovector, int ovecsize, |
| 735 | int *workspace, int wscount); |
| 736 | |
| 737 | int pcre_copy_named_substring(const pcre *code, |
| 738 | const char *subject, int *ovector, |
| 739 | int stringcount, const char *stringname, |
| 740 | char *buffer, int buffersize); |
| 741 | |
| 742 | int pcre_copy_substring(const char *subject, int *ovector, |
| 743 | int stringcount, int stringnumber, char *buffer, |
| 744 | int buffersize); |
| 745 | |
| 746 | int pcre_get_named_substring(const pcre *code, |
| 747 | const char *subject, int *ovector, |
| 748 | int stringcount, const char *stringname, |
| 749 | const char **stringptr); |
| 750 | |
| 751 | int pcre_get_stringnumber(const pcre *code, |
| 752 | const char *name); |
| 753 | |
| 754 | int pcre_get_stringtable_entries(const pcre *code, |
| 755 | const char *name, char **first, char **last); |
| 756 | |
| 757 | int pcre_get_substring(const char *subject, int *ovector, |
| 758 | int stringcount, int stringnumber, |
| 759 | const char **stringptr); |
| 760 | |
| 761 | int pcre_get_substring_list(const char *subject, |
| 762 | int *ovector, int stringcount, const char ***listptr); |
| 763 | |
| 764 | void pcre_free_substring(const char *stringptr); |
| 765 | |
| 766 | void pcre_free_substring_list(const char **stringptr); |
| 767 | |
| 768 | const unsigned char *pcre_maketables(void); |
| 769 | |
| 770 | int pcre_fullinfo(const pcre *code, const pcre_extra *extra, |
| 771 | int what, void *where); |
| 772 | |
| 773 | int pcre_info(const pcre *code, int *optptr, int *firstcharptr); |
| 774 | |
| 775 | int pcre_refcount(pcre *code, int adjust); |
| 776 | |
| 777 | int pcre_config(int what, void *where); |
| 778 | |
| 779 | char *pcre_version(void); |
| 780 | |
| 781 | |
| 782 | PCRE NATIVE API INDIRECTED FUNCTIONS |
| 783 | |
| 784 | void *(*pcre_malloc)(size_t); |
| 785 | |
| 786 | void (*pcre_free)(void *); |
| 787 | |
| 788 | void *(*pcre_stack_malloc)(size_t); |
| 789 | |
| 790 | void (*pcre_stack_free)(void *); |
| 791 | |
| 792 | int (*pcre_callout)(pcre_callout_block *); |
| 793 | |
| 794 | |
| 795 | PCRE API OVERVIEW |
| 796 | |
| 797 | PCRE has its own native API, which is described in this document. There |
| 798 | are also some wrapper functions that correspond to the POSIX regular |
| 799 | expression API, but they do not give access to all the functionality. |
| 800 | They are described in the pcreposix documentation. Both of these APIs |
| 801 | define a set of C function calls. A C++ wrapper is also distributed |
| 802 | with PCRE. It is documented in the pcrecpp page. |
| 803 | |
| 804 | The native API C function prototypes are defined in the header file |
| 805 | pcre.h, and on Unix systems the library itself is called libpcre. It |
| 806 | can normally be accessed by adding -lpcre to the command for linking an |
| 807 | application that uses PCRE. The header file defines the macros |
| 808 | PCRE_MAJOR and PCRE_MINOR to contain the major and minor release num- |
| 809 | bers for the library. Applications can use these to include support |
| 810 | for different releases of PCRE. |
| 811 | |
| 812 | In a Windows environment, if you want to statically link an application |
| 813 | program against a non-dll pcre.a file, you must define PCRE_STATIC |
| 814 | before including pcre.h or pcrecpp.h, because otherwise the pcre_mal- |
| 815 | loc() and pcre_free() exported functions will be declared |
| 816 | __declspec(dllimport), with unwanted results. |
| 817 | |
| 818 | The functions pcre_compile(), pcre_compile2(), pcre_study(), and |
| 819 | pcre_exec() are used for compiling and matching regular expressions in |
| 820 | a Perl-compatible manner. A sample program that demonstrates the sim- |
| 821 | plest way of using them is provided in the file called pcredemo.c in |
| 822 | the PCRE source distribution. A listing of this program is given in the |
| 823 | pcredemo documentation, and the pcresample documentation describes how |
| 824 | to compile and run it. |
| 825 | |
| 826 | Just-in-time compiler support is an optional feature of PCRE that can |
| 827 | be built in appropriate hardware environments. It greatly speeds up the |
| 828 | matching performance of many patterns. Simple programs can easily |
| 829 | request that it be used if available, by setting an option that is |
| 830 | ignored when it is not relevant. More complicated programs might need |
| 831 | to make use of the functions pcre_jit_stack_alloc(), |
| 832 | pcre_jit_stack_free(), and pcre_assign_jit_stack() in order to control |
| 833 | the JIT code's memory usage. These functions are discussed in the |
| 834 | pcrejit documentation. |
| 835 | |
| 836 | A second matching function, pcre_dfa_exec(), which is not Perl-compati- |
| 837 | ble, is also provided. This uses a different algorithm for the match- |
| 838 | ing. The alternative algorithm finds all possible matches (at a given |
| 839 | point in the subject), and scans the subject just once (unless there |
| 840 | are lookbehind assertions). However, this algorithm does not return |
| 841 | captured substrings. A description of the two matching algorithms and |
| 842 | their advantages and disadvantages is given in the pcrematching docu- |
| 843 | mentation. |
| 844 | |
| 845 | In addition to the main compiling and matching functions, there are |
| 846 | convenience functions for extracting captured substrings from a subject |
| 847 | string that is matched by pcre_exec(). They are: |
| 848 | |
| 849 | pcre_copy_substring() |
| 850 | pcre_copy_named_substring() |
| 851 | pcre_get_substring() |
| 852 | pcre_get_named_substring() |
| 853 | pcre_get_substring_list() |
| 854 | pcre_get_stringnumber() |
| 855 | pcre_get_stringtable_entries() |
| 856 | |
| 857 | pcre_free_substring() and pcre_free_substring_list() are also provided, |
| 858 | to free the memory used for extracted strings. |
| 859 | |
| 860 | The function pcre_maketables() is used to build a set of character |
| 861 | tables in the current locale for passing to pcre_compile(), |
| 862 | pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is |
| 863 | provided for specialist use. Most commonly, no special tables are |
| 864 | passed, in which case internal tables that are generated when PCRE is |
| 865 | built are used. |
| 866 | |
| 867 | The function pcre_fullinfo() is used to find out information about a |
| 868 | compiled pattern; pcre_info() is an obsolete version that returns only |
| 869 | some of the available information, but is retained for backwards com- |
| 870 | patibility. The function pcre_version() returns a pointer to a string |
| 871 | containing the version of PCRE and its date of release. |
| 872 | |
| 873 | The function pcre_refcount() maintains a reference count in a data |
| 874 | block containing a compiled pattern. This is provided for the benefit |
| 875 | of object-oriented applications. |
| 876 | |
| 877 | The global variables pcre_malloc and pcre_free initially contain the |
| 878 | entry points of the standard malloc() and free() functions, respec- |
| 879 | tively. PCRE calls the memory management functions via these variables, |
| 880 | so a calling program can replace them if it wishes to intercept the |
| 881 | calls. This should be done before calling any PCRE functions. |
| 882 | |
| 883 | The global variables pcre_stack_malloc and pcre_stack_free are also |
| 884 | indirections to memory management functions. These special functions |
| 885 | are used only when PCRE is compiled to use the heap for remembering |
| 886 | data, instead of recursive function calls, when running the pcre_exec() |
| 887 | function. See the pcrebuild documentation for details of how to do |
| 888 | this. It is a non-standard way of building PCRE, for use in environ- |
| 889 | ments that have limited stacks. Because of the greater use of memory |
| 890 | management, it runs more slowly. Separate functions are provided so |
| 891 | that special-purpose external code can be used for this case. When |
| 892 | used, these functions are always called in a stack-like manner (last |
| 893 | obtained, first freed), and always for memory blocks of the same size. |
| 894 | There is a discussion about PCRE's stack usage in the pcrestack docu- |
| 895 | mentation. |
| 896 | |
| 897 | The global variable pcre_callout initially contains NULL. It can be set |
| 898 | by the caller to a "callout" function, which PCRE will then call at |
| 899 | specified points during a matching operation. Details are given in the |
| 900 | pcrecallout documentation. |
| 901 | |
| 902 | |
| 903 | NEWLINES |
| 904 | |
| 905 | PCRE supports five different conventions for indicating line breaks in |
| 906 | strings: a single CR (carriage return) character, a single LF (line- |
| 907 | feed) character, the two-character sequence CRLF, any of the three pre- |
| 908 | ceding, or any Unicode newline sequence. The Unicode newline sequences |
| 909 | are the three just mentioned, plus the single characters VT (vertical |
| 910 | tab, U+000B), FF (formfeed, U+000C), NEL (next line, U+0085), LS (line |
| 911 | separator, U+2028), and PS (paragraph separator, U+2029). |
| 912 | |
| 913 | Each of the first three conventions is used by at least one operating |
| 914 | system as its standard newline sequence. When PCRE is built, a default |
| 915 | can be specified. The default default is LF, which is the Unix stan- |
| 916 | dard. When PCRE is run, the default can be overridden, either when a |
| 917 | pattern is compiled, or when it is matched. |
| 918 | |
| 919 | At compile time, the newline convention can be specified by the options |
| 920 | argument of pcre_compile(), or it can be specified by special text at |
| 921 | the start of the pattern itself; this overrides any other settings. See |
| 922 | the pcrepattern page for details of the special character sequences. |
| 923 | |
| 924 | In the PCRE documentation the word "newline" is used to mean "the char- |
| 925 | acter or pair of characters that indicate a line break". The choice of |
| 926 | newline convention affects the handling of the dot, circumflex, and |
| 927 | dollar metacharacters, the handling of #-comments in /x mode, and, when |
| 928 | CRLF is a recognized line ending sequence, the match position advance- |
| 929 | ment for a non-anchored pattern. There is more detail about this in the |
| 930 | section on pcre_exec() options below. |
| 931 | |
| 932 | The choice of newline convention does not affect the interpretation of |
| 933 | the \n or \r escape sequences, nor does it affect what \R matches, |
| 934 | which is controlled in a similar way, but by separate options. |
| 935 | |
| 936 | |
| 937 | MULTITHREADING |
| 938 | |
| 939 | The PCRE functions can be used in multi-threading applications, with |
| 940 | the proviso that the memory management functions pointed to by |
| 941 | pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the |
| 942 | callout function pointed to by pcre_callout, are shared by all threads. |
| 943 | |
| 944 | The compiled form of a regular expression is not altered during match- |
| 945 | ing, so the same compiled pattern can safely be used by several threads |
| 946 | at once. |
| 947 | |
| 948 | If the just-in-time optimization feature is being used, it needs sepa- |
| 949 | rate memory stack areas for each thread. See the pcrejit documentation |
| 950 | for more details. |
| 951 | |
| 952 | |
| 953 | SAVING PRECOMPILED PATTERNS FOR LATER USE |
| 954 | |
| 955 | The compiled form of a regular expression can be saved and re-used at a |
| 956 | later time, possibly by a different program, and even on a host other |
| 957 | than the one on which it was compiled. Details are given in the |
| 958 | pcreprecompile documentation. However, compiling a regular expression |
| 959 | with one version of PCRE for use with a different version is not guar- |
| 960 | anteed to work and may cause crashes. |
| 961 | |
| 962 | |
| 963 | CHECKING BUILD-TIME OPTIONS |
| 964 | |
| 965 | int pcre_config(int what, void *where); |
| 966 | |
| 967 | The function pcre_config() makes it possible for a PCRE client to dis- |
| 968 | cover which optional features have been compiled into the PCRE library. |
| 969 | The pcrebuild documentation has more details about these optional fea- |
| 970 | tures. |
| 971 | |
| 972 | The first argument for pcre_config() is an integer, specifying which |
| 973 | information is required; the second argument is a pointer to a variable |
| 974 | into which the information is placed. The following information is |
| 975 | available: |
| 976 | |
| 977 | PCRE_CONFIG_UTF8 |
| 978 | |
| 979 | The output is an integer that is set to one if UTF-8 support is avail- |
| 980 | able; otherwise it is set to zero. |
| 981 | |
| 982 | PCRE_CONFIG_UNICODE_PROPERTIES |
| 983 | |
| 984 | The output is an integer that is set to one if support for Unicode |
| 985 | character properties is available; otherwise it is set to zero. |
| 986 | |
| 987 | PCRE_CONFIG_JIT |
| 988 | |
| 989 | The output is an integer that is set to one if support for just-in-time |
| 990 | compiling is available; otherwise it is set to zero. |
| 991 | |
| 992 | PCRE_CONFIG_NEWLINE |
| 993 | |
| 994 | The output is an integer whose value specifies the default character |
| 995 | sequence that is recognized as meaning "newline". The four values that |
| 996 | are supported are: 10 for LF, 13 for CR, 3338 for CRLF, -2 for ANYCRLF, |
| 997 | and -1 for ANY. Though they are derived from ASCII, the same values |
| 998 | are returned in EBCDIC environments. The default should normally corre- |
| 999 | spond to the standard sequence for your operating system. |
| 1000 | |
| 1001 | PCRE_CONFIG_BSR |
| 1002 | |
| 1003 | The output is an integer whose value indicates what character sequences |
| 1004 | the \R escape sequence matches by default. A value of 0 means that \R |
| 1005 | matches any Unicode line ending sequence; a value of 1 means that \R |
| 1006 | matches only CR, LF, or CRLF. The default can be overridden when a pat- |
| 1007 | tern is compiled or matched. |
| 1008 | |
| 1009 | PCRE_CONFIG_LINK_SIZE |
| 1010 | |
| 1011 | The output is an integer that contains the number of bytes used for |
| 1012 | internal linkage in compiled regular expressions. The value is 2, 3, or |
| 1013 | 4. Larger values allow larger regular expressions to be compiled, at |
| 1014 | the expense of slower matching. The default value of 2 is sufficient |
| 1015 | for all but the most massive patterns, since it allows the compiled |
| 1016 | pattern to be up to 64K in size. |
| 1017 | |
| 1018 | PCRE_CONFIG_POSIX_MALLOC_THRESHOLD |
| 1019 | |
| 1020 | The output is an integer that contains the threshold above which the |
| 1021 | POSIX interface uses malloc() for output vectors. Further details are |
| 1022 | given in the pcreposix documentation. |
| 1023 | |
| 1024 | PCRE_CONFIG_MATCH_LIMIT |
| 1025 | |
| 1026 | The output is a long integer that gives the default limit for the num- |
| 1027 | ber of internal matching function calls in a pcre_exec() execution. |
| 1028 | Further details are given with pcre_exec() below. |
| 1029 | |
| 1030 | PCRE_CONFIG_MATCH_LIMIT_RECURSION |
| 1031 | |
| 1032 | The output is a long integer that gives the default limit for the depth |
| 1033 | of recursion when calling the internal matching function in a |
| 1034 | pcre_exec() execution. Further details are given with pcre_exec() |
| 1035 | below. |
| 1036 | |
| 1037 | PCRE_CONFIG_STACKRECURSE |
| 1038 | |
| 1039 | The output is an integer that is set to one if internal recursion when |
| 1040 | running pcre_exec() is implemented by recursive function calls that use |
| 1041 | the stack to remember their state. This is the usual way that PCRE is |
| 1042 | compiled. The output is zero if PCRE was compiled to use blocks of data |
| 1043 | on the heap instead of recursive function calls. In this case, |
| 1044 | pcre_stack_malloc and pcre_stack_free are called to manage memory |
| 1045 | blocks on the heap, thus avoiding the use of the stack. |
| 1046 | |
| 1047 | |
| 1048 | COMPILING A PATTERN |
| 1049 | |
| 1050 | pcre *pcre_compile(const char *pattern, int options, |
| 1051 | const char **errptr, int *erroffset, |
| 1052 | const unsigned char *tableptr); |
| 1053 | |
| 1054 | pcre *pcre_compile2(const char *pattern, int options, |
| 1055 | int *errorcodeptr, |
| 1056 | const char **errptr, int *erroffset, |
| 1057 | const unsigned char *tableptr); |
| 1058 | |
| 1059 | Either of the functions pcre_compile() or pcre_compile2() can be called |
| 1060 | to compile a pattern into an internal form. The only difference between |
| 1061 | the two interfaces is that pcre_compile2() has an additional argument, |
| 1062 | errorcodeptr, via which a numerical error code can be returned. To |
| 1063 | avoid too much repetition, we refer just to pcre_compile() below, but |
| 1064 | the information applies equally to pcre_compile2(). |
| 1065 | |
| 1066 | The pattern is a C string terminated by a binary zero, and is passed in |
| 1067 | the pattern argument. A pointer to a single block of memory that is |
| 1068 | obtained via pcre_malloc is returned. This contains the compiled code |
| 1069 | and related data. The pcre type is defined for the returned block; this |
| 1070 | is a typedef for a structure whose contents are not externally defined. |
| 1071 | It is up to the caller to free the memory (via pcre_free) when it is no |
| 1072 | longer required. |
| 1073 | |
| 1074 | Although the compiled code of a PCRE regex is relocatable, that is, it |
| 1075 | does not depend on memory location, the complete pcre data block is not |
| 1076 | fully relocatable, because it may contain a copy of the tableptr argu- |
| 1077 | ment, which is an address (see below). |
| 1078 | |
| 1079 | The options argument contains various bit settings that affect the com- |
| 1080 | pilation. It should be zero if no options are required. The available |
| 1081 | options are described below. Some of them (in particular, those that |
| 1082 | are compatible with Perl, but some others as well) can also be set and |
| 1083 | unset from within the pattern (see the detailed description in the |
| 1084 | pcrepattern documentation). For those options that can be different in |
| 1085 | different parts of the pattern, the contents of the options argument |
| 1086 | specifies their settings at the start of compilation and execution. The |
| 1087 | PCRE_ANCHORED, PCRE_BSR_xxx, PCRE_NEWLINE_xxx, PCRE_NO_UTF8_CHECK, and |
| 1088 | PCRE_NO_START_OPT options can be set at the time of matching as well as |
| 1089 | at compile time. |
| 1090 | |
| 1091 | If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise, |
| 1092 | if compilation of a pattern fails, pcre_compile() returns NULL, and |
| 1093 | sets the variable pointed to by errptr to point to a textual error mes- |
| 1094 | sage. This is a static string that is part of the library. You must not |
| 1095 | try to free it. Normally, the offset from the start of the pattern to |
| 1096 | the byte that was being processed when the error was discovered is |
| 1097 | placed in the variable pointed to by erroffset, which must not be NULL |
| 1098 | (if it is, an immediate error is given). However, for an invalid UTF-8 |
| 1099 | string, the offset is that of the first byte of the failing character. |
| 1100 | Also, some errors are not detected until checks are carried out when |
| 1101 | the whole pattern has been scanned; in these cases the offset passed |
| 1102 | back is the length of the pattern. |
| 1103 | |
| 1104 | Note that the offset is in bytes, not characters, even in UTF-8 mode. |
| 1105 | It may sometimes point into the middle of a UTF-8 character. |
| 1106 | |
| 1107 | If pcre_compile2() is used instead of pcre_compile(), and the error- |
| 1108 | codeptr argument is not NULL, a non-zero error code number is returned |
| 1109 | via this argument in the event of an error. This is in addition to the |
| 1110 | textual error message. Error codes and messages are listed below. |
| 1111 | |
| 1112 | If the final argument, tableptr, is NULL, PCRE uses a default set of |
| 1113 | character tables that are built when PCRE is compiled, using the |
| 1114 | default C locale. Otherwise, tableptr must be an address that is the |
| 1115 | result of a call to pcre_maketables(). This value is stored with the |
| 1116 | compiled pattern, and used again by pcre_exec(), unless another table |
| 1117 | pointer is passed to it. For more discussion, see the section on locale |
| 1118 | support below. |
| 1119 | |
| 1120 | This code fragment shows a typical straightforward call to pcre_com- |
| 1121 | pile(): |
| 1122 | |
| 1123 | pcre *re; |
| 1124 | const char *error; |
| 1125 | int erroffset; |
| 1126 | re = pcre_compile( |
| 1127 | "^A.*Z", /* the pattern */ |
| 1128 | 0, /* default options */ |
| 1129 | &error, /* for error message */ |
| 1130 | &erroffset, /* for error offset */ |
| 1131 | NULL); /* use default character tables */ |
| 1132 | |
| 1133 | The following names for option bits are defined in the pcre.h header |
| 1134 | file: |
| 1135 | |
| 1136 | PCRE_ANCHORED |
| 1137 | |
| 1138 | If this bit is set, the pattern is forced to be "anchored", that is, it |
| 1139 | is constrained to match only at the first matching point in the string |
| 1140 | that is being searched (the "subject string"). This effect can also be |
| 1141 | achieved by appropriate constructs in the pattern itself, which is the |
| 1142 | only way to do it in Perl. |
| 1143 | |
| 1144 | PCRE_AUTO_CALLOUT |
| 1145 | |
| 1146 | If this bit is set, pcre_compile() automatically inserts callout items, |
| 1147 | all with number 255, before each pattern item. For discussion of the |
| 1148 | callout facility, see the pcrecallout documentation. |
| 1149 | |
| 1150 | PCRE_BSR_ANYCRLF |
| 1151 | PCRE_BSR_UNICODE |
| 1152 | |
| 1153 | These options (which are mutually exclusive) control what the \R escape |
| 1154 | sequence matches. The choice is either to match only CR, LF, or CRLF, |
| 1155 | or to match any Unicode newline sequence. The default is specified when |
| 1156 | PCRE is built. It can be overridden from within the pattern, or by set- |
| 1157 | ting an option when a compiled pattern is matched. |
| 1158 | |
| 1159 | PCRE_CASELESS |
| 1160 | |
| 1161 | If this bit is set, letters in the pattern match both upper and lower |
| 1162 | case letters. It is equivalent to Perl's /i option, and it can be |
| 1163 | changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE |
| 1164 | always understands the concept of case for characters whose values are |
| 1165 | less than 128, so caseless matching is always possible. For characters |
| 1166 | with higher values, the concept of case is supported if PCRE is com- |
| 1167 | piled with Unicode property support, but not otherwise. If you want to |
| 1168 | use caseless matching for characters 128 and above, you must ensure |
| 1169 | that PCRE is compiled with Unicode property support as well as with |
| 1170 | UTF-8 support. |
| 1171 | |
| 1172 | PCRE_DOLLAR_ENDONLY |
| 1173 | |
| 1174 | If this bit is set, a dollar metacharacter in the pattern matches only |
| 1175 | at the end of the subject string. Without this option, a dollar also |
| 1176 | matches immediately before a newline at the end of the string (but not |
| 1177 | before any other newlines). The PCRE_DOLLAR_ENDONLY option is ignored |
| 1178 | if PCRE_MULTILINE is set. There is no equivalent to this option in |
| 1179 | Perl, and no way to set it within a pattern. |
| 1180 | |
| 1181 | PCRE_DOTALL |
| 1182 | |
| 1183 | If this bit is set, a dot metacharacter in the pattern matches a char- |
| 1184 | acter of any value, including one that indicates a newline. However, it |
| 1185 | only ever matches one character, even if newlines are coded as CRLF. |
| 1186 | Without this option, a dot does not match when the current position is |
| 1187 | at a newline. This option is equivalent to Perl's /s option, and it can |
| 1188 | be changed within a pattern by a (?s) option setting. A negative class |
| 1189 | such as [^a] always matches newline characters, independent of the set- |
| 1190 | ting of this option. |
| 1191 | |
| 1192 | PCRE_DUPNAMES |
| 1193 | |
| 1194 | If this bit is set, names used to identify capturing subpatterns need |
| 1195 | not be unique. This can be helpful for certain types of pattern when it |
| 1196 | is known that only one instance of the named subpattern can ever be |
| 1197 | matched. There are more details of named subpatterns below; see also |
| 1198 | the pcrepattern documentation. |
| 1199 | |
| 1200 | PCRE_EXTENDED |
| 1201 | |
| 1202 | If this bit is set, whitespace data characters in the pattern are |
| 1203 | totally ignored except when escaped or inside a character class. White- |
| 1204 | space does not include the VT character (code 11). In addition, charac- |
| 1205 | ters between an unescaped # outside a character class and the next new- |
| 1206 | line, inclusive, are also ignored. This is equivalent to Perl's /x |
| 1207 | option, and it can be changed within a pattern by a (?x) option set- |
| 1208 | ting. |
| 1209 | |
| 1210 | Which characters are interpreted as newlines is controlled by the |
| 1211 | options passed to pcre_compile() or by a special sequence at the start |
| 1212 | of the pattern, as described in the section entitled "Newline conven- |
| 1213 | tions" in the pcrepattern documentation. Note that the end of this type |
| 1214 | of comment is a literal newline sequence in the pattern; escape |
| 1215 | sequences that happen to represent a newline do not count. |
| 1216 | |
| 1217 | This option makes it possible to include comments inside complicated |
| 1218 | patterns. Note, however, that this applies only to data characters. |
| 1219 | Whitespace characters may never appear within special character |
| 1220 | sequences in a pattern, for example within the sequence (?( that intro- |
| 1221 | duces a conditional subpattern. |
| 1222 | |
| 1223 | PCRE_EXTRA |
| 1224 | |
| 1225 | This option was invented in order to turn on additional functionality |
| 1226 | of PCRE that is incompatible with Perl, but it is currently of very |
| 1227 | little use. When set, any backslash in a pattern that is followed by a |
| 1228 | letter that has no special meaning causes an error, thus reserving |
| 1229 | these combinations for future expansion. By default, as in Perl, a |
| 1230 | backslash followed by a letter with no special meaning is treated as a |
| 1231 | literal. (Perl can, however, be persuaded to give an error for this, by |
| 1232 | running it with the -w option.) There are at present no other features |
| 1233 | controlled by this option. It can also be set by a (?X) option setting |
| 1234 | within a pattern. |
| 1235 | |
| 1236 | PCRE_FIRSTLINE |
| 1237 | |
| 1238 | If this option is set, an unanchored pattern is required to match |
| 1239 | before or at the first newline in the subject string, though the |
| 1240 | matched text may continue over the newline. |
| 1241 | |
| 1242 | PCRE_JAVASCRIPT_COMPAT |
| 1243 | |
| 1244 | If this option is set, PCRE's behaviour is changed in some ways so that |
| 1245 | it is compatible with JavaScript rather than Perl. The changes are as |
| 1246 | follows: |
| 1247 | |
| 1248 | (1) A lone closing square bracket in a pattern causes a compile-time |
| 1249 | error, because this is illegal in JavaScript (by default it is treated |
| 1250 | as a data character). Thus, the pattern AB]CD becomes illegal when this |
| 1251 | option is set. |
| 1252 | |
| 1253 | (2) At run time, a back reference to an unset subpattern group matches |
| 1254 | an empty string (by default this causes the current matching alterna- |
| 1255 | tive to fail). A pattern such as (\1)(a) succeeds when this option is |
| 1256 | set (assuming it can find an "a" in the subject), whereas it fails by |
| 1257 | default, for Perl compatibility. |
| 1258 | |
| 1259 | (3) \U matches an upper case "U" character; by default \U causes a com- |
| 1260 | pile time error (Perl uses \U to upper case subsequent characters). |
| 1261 | |
| 1262 | (4) \u matches a lower case "u" character unless it is followed by four |
| 1263 | hexadecimal digits, in which case the hexadecimal number defines the |
| 1264 | code point to match. By default, \u causes a compile time error (Perl |
| 1265 | uses it to upper case the following character). |
| 1266 | |
| 1267 | (5) \x matches a lower case "x" character unless it is followed by two |
| 1268 | hexadecimal digits, in which case the hexadecimal number defines the |
| 1269 | code point to match. By default, as in Perl, a hexadecimal number is |
| 1270 | always expected after \x, but it may have zero, one, or two digits (so, |
| 1271 | for example, \xz matches a binary zero character followed by z). |
| 1272 | |
| 1273 | PCRE_MULTILINE |
| 1274 | |
| 1275 | By default, PCRE treats the subject string as consisting of a single |
| 1276 | line of characters (even if it actually contains newlines). The "start |
| 1277 | of line" metacharacter (^) matches only at the start of the string, |
| 1278 | while the "end of line" metacharacter ($) matches only at the end of |
| 1279 | the string, or before a terminating newline (unless PCRE_DOLLAR_ENDONLY |
| 1280 | is set). This is the same as Perl. |
| 1281 | |
| 1282 | When PCRE_MULTILINE it is set, the "start of line" and "end of line" |
| 1283 | constructs match immediately following or immediately before internal |
| 1284 | newlines in the subject string, respectively, as well as at the very |
| 1285 | start and end. This is equivalent to Perl's /m option, and it can be |
| 1286 | changed within a pattern by a (?m) option setting. If there are no new- |
| 1287 | lines in a subject string, or no occurrences of ^ or $ in a pattern, |
| 1288 | setting PCRE_MULTILINE has no effect. |
| 1289 | |
| 1290 | PCRE_NEWLINE_CR |
| 1291 | PCRE_NEWLINE_LF |
| 1292 | PCRE_NEWLINE_CRLF |
| 1293 | PCRE_NEWLINE_ANYCRLF |
| 1294 | PCRE_NEWLINE_ANY |
| 1295 | |
| 1296 | These options override the default newline definition that was chosen |
| 1297 | when PCRE was built. Setting the first or the second specifies that a |
| 1298 | newline is indicated by a single character (CR or LF, respectively). |
| 1299 | Setting PCRE_NEWLINE_CRLF specifies that a newline is indicated by the |
| 1300 | two-character CRLF sequence. Setting PCRE_NEWLINE_ANYCRLF specifies |
| 1301 | that any of the three preceding sequences should be recognized. Setting |
| 1302 | PCRE_NEWLINE_ANY specifies that any Unicode newline sequence should be |
| 1303 | recognized. The Unicode newline sequences are the three just mentioned, |
| 1304 | plus the single characters VT (vertical tab, U+000B), FF (formfeed, |
| 1305 | U+000C), NEL (next line, U+0085), LS (line separator, U+2028), and PS |
| 1306 | (paragraph separator, U+2029). The last two are recognized only in |
| 1307 | UTF-8 mode. |
| 1308 | |
| 1309 | The newline setting in the options word uses three bits that are |
| 1310 | treated as a number, giving eight possibilities. Currently only six are |
| 1311 | used (default plus the five values above). This means that if you set |
| 1312 | more than one newline option, the combination may or may not be sensi- |
| 1313 | ble. For example, PCRE_NEWLINE_CR with PCRE_NEWLINE_LF is equivalent to |
| 1314 | PCRE_NEWLINE_CRLF, but other combinations may yield unused numbers and |
| 1315 | cause an error. |
| 1316 | |
| 1317 | The only time that a line break in a pattern is specially recognized |
| 1318 | when compiling is when PCRE_EXTENDED is set. CR and LF are whitespace |
| 1319 | characters, and so are ignored in this mode. Also, an unescaped # out- |
| 1320 | side a character class indicates a comment that lasts until after the |
| 1321 | next line break sequence. In other circumstances, line break sequences |
| 1322 | in patterns are treated as literal data. |
| 1323 | |
| 1324 | The newline option that is set at compile time becomes the default that |
| 1325 | is used for pcre_exec() and pcre_dfa_exec(), but it can be overridden. |
| 1326 | |
| 1327 | PCRE_NO_AUTO_CAPTURE |
| 1328 | |
| 1329 | If this option is set, it disables the use of numbered capturing paren- |
| 1330 | theses in the pattern. Any opening parenthesis that is not followed by |
| 1331 | ? behaves as if it were followed by ?: but named parentheses can still |
| 1332 | be used for capturing (and they acquire numbers in the usual way). |
| 1333 | There is no equivalent of this option in Perl. |
| 1334 | |
| 1335 | NO_START_OPTIMIZE |
| 1336 | |
| 1337 | This is an option that acts at matching time; that is, it is really an |
| 1338 | option for pcre_exec() or pcre_dfa_exec(). If it is set at compile |
| 1339 | time, it is remembered with the compiled pattern and assumed at match- |
| 1340 | ing time. For details see the discussion of PCRE_NO_START_OPTIMIZE |
| 1341 | below. |
| 1342 | |
| 1343 | PCRE_UCP |
| 1344 | |
| 1345 | This option changes the way PCRE processes \B, \b, \D, \d, \S, \s, \W, |
| 1346 | \w, and some of the POSIX character classes. By default, only ASCII |
| 1347 | characters are recognized, but if PCRE_UCP is set, Unicode properties |
| 1348 | are used instead to classify characters. More details are given in the |
| 1349 | section on generic character types in the pcrepattern page. If you set |
| 1350 | PCRE_UCP, matching one of the items it affects takes much longer. The |
| 1351 | option is available only if PCRE has been compiled with Unicode prop- |
| 1352 | erty support. |
| 1353 | |
| 1354 | PCRE_UNGREEDY |
| 1355 | |
| 1356 | This option inverts the "greediness" of the quantifiers so that they |
| 1357 | are not greedy by default, but become greedy if followed by "?". It is |
| 1358 | not compatible with Perl. It can also be set by a (?U) option setting |
| 1359 | within the pattern. |
| 1360 | |
| 1361 | PCRE_UTF8 |
| 1362 | |
| 1363 | This option causes PCRE to regard both the pattern and the subject as |
| 1364 | strings of UTF-8 characters instead of single-byte character strings. |
| 1365 | However, it is available only when PCRE is built to include UTF-8 sup- |
| 1366 | port. If not, the use of this option provokes an error. Details of how |
| 1367 | this option changes the behaviour of PCRE are given in the pcreunicode |
| 1368 | page. |
| 1369 | |
| 1370 | PCRE_NO_UTF8_CHECK |
| 1371 | |
| 1372 | When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is |
| 1373 | automatically checked. There is a discussion about the validity of |
| 1374 | UTF-8 strings in the main pcre page. If an invalid UTF-8 sequence of |
| 1375 | bytes is found, pcre_compile() returns an error. If you already know |
| 1376 | that your pattern is valid, and you want to skip this check for perfor- |
| 1377 | mance reasons, you can set the PCRE_NO_UTF8_CHECK option. When it is |
| 1378 | set, the effect of passing an invalid UTF-8 string as a pattern is |
| 1379 | undefined. It may cause your program to crash. Note that this option |
| 1380 | can also be passed to pcre_exec() and pcre_dfa_exec(), to suppress the |
| 1381 | UTF-8 validity checking of subject strings. |
| 1382 | |
| 1383 | |
| 1384 | COMPILATION ERROR CODES |
| 1385 | |
| 1386 | The following table lists the error codes than may be returned by |
| 1387 | pcre_compile2(), along with the error messages that may be returned by |
| 1388 | both compiling functions. As PCRE has developed, some error codes have |
| 1389 | fallen out of use. To avoid confusion, they have not been re-used. |
| 1390 | |
| 1391 | 0 no error |
| 1392 | 1 \ at end of pattern |
| 1393 | 2 \c at end of pattern |
| 1394 | 3 unrecognized character follows \ |
| 1395 | 4 numbers out of order in {} quantifier |
| 1396 | 5 number too big in {} quantifier |
| 1397 | 6 missing terminating ] for character class |
| 1398 | 7 invalid escape sequence in character class |
| 1399 | 8 range out of order in character class |
| 1400 | 9 nothing to repeat |
| 1401 | 10 [this code is not in use] |
| 1402 | 11 internal error: unexpected repeat |
| 1403 | 12 unrecognized character after (? or (?- |
| 1404 | 13 POSIX named classes are supported only within a class |
| 1405 | 14 missing ) |
| 1406 | 15 reference to non-existent subpattern |
| 1407 | 16 erroffset passed as NULL |
| 1408 | 17 unknown option bit(s) set |
| 1409 | 18 missing ) after comment |
| 1410 | 19 [this code is not in use] |
| 1411 | 20 regular expression is too large |
| 1412 | 21 failed to get memory |
| 1413 | 22 unmatched parentheses |
| 1414 | 23 internal error: code overflow |
| 1415 | 24 unrecognized character after (?< |
| 1416 | 25 lookbehind assertion is not fixed length |
| 1417 | 26 malformed number or name after (?( |
| 1418 | 27 conditional group contains more than two branches |
| 1419 | 28 assertion expected after (?( |
| 1420 | 29 (?R or (?[+-]digits must be followed by ) |
| 1421 | 30 unknown POSIX class name |
| 1422 | 31 POSIX collating elements are not supported |
| 1423 | 32 this version of PCRE is not compiled with PCRE_UTF8 support |
| 1424 | 33 [this code is not in use] |
| 1425 | 34 character value in \x{...} sequence is too large |
| 1426 | 35 invalid condition (?(0) |
| 1427 | 36 \C not allowed in lookbehind assertion |
| 1428 | 37 PCRE does not support \L, \l, \N{name}, \U, or \u |
| 1429 | 38 number after (?C is > 255 |
| 1430 | 39 closing ) for (?C expected |
| 1431 | 40 recursive call could loop indefinitely |
| 1432 | 41 unrecognized character after (?P |
| 1433 | 42 syntax error in subpattern name (missing terminator) |
| 1434 | 43 two named subpatterns have the same name |
| 1435 | 44 invalid UTF-8 string |
| 1436 | 45 support for \P, \p, and \X has not been compiled |
| 1437 | 46 malformed \P or \p sequence |
| 1438 | 47 unknown property name after \P or \p |
| 1439 | 48 subpattern name is too long (maximum 32 characters) |
| 1440 | 49 too many named subpatterns (maximum 10000) |
| 1441 | 50 [this code is not in use] |
| 1442 | 51 octal value is greater than \377 (not in UTF-8 mode) |
| 1443 | 52 internal error: overran compiling workspace |
| 1444 | 53 internal error: previously-checked referenced subpattern |
| 1445 | not found |
| 1446 | 54 DEFINE group contains more than one branch |
| 1447 | 55 repeating a DEFINE group is not allowed |
| 1448 | 56 inconsistent NEWLINE options |
| 1449 | 57 \g is not followed by a braced, angle-bracketed, or quoted |
| 1450 | name/number or by a plain number |
| 1451 | 58 a numbered reference must not be zero |
| 1452 | 59 an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT) |
| 1453 | 60 (*VERB) not recognized |
| 1454 | 61 number is too big |
| 1455 | 62 subpattern name expected |
| 1456 | 63 digit expected after (?+ |
| 1457 | 64 ] is an invalid data character in JavaScript compatibility mode |
| 1458 | 65 different names for subpatterns of the same number are |
| 1459 | not allowed |
| 1460 | 66 (*MARK) must have an argument |
| 1461 | 67 this version of PCRE is not compiled with PCRE_UCP support |
| 1462 | 68 \c must be followed by an ASCII character |
| 1463 | 69 \k is not followed by a braced, angle-bracketed, or quoted name |
| 1464 | |
| 1465 | The numbers 32 and 10000 in errors 48 and 49 are defaults; different |
| 1466 | values may be used if the limits were changed when PCRE was built. |
| 1467 | |
| 1468 | |
| 1469 | STUDYING A PATTERN |
| 1470 | |
| 1471 | pcre_extra *pcre_study(const pcre *code, int options |
| 1472 | const char **errptr); |
| 1473 | |
| 1474 | If a compiled pattern is going to be used several times, it is worth |
| 1475 | spending more time analyzing it in order to speed up the time taken for |
| 1476 | matching. The function pcre_study() takes a pointer to a compiled pat- |
| 1477 | tern as its first argument. If studying the pattern produces additional |
| 1478 | information that will help speed up matching, pcre_study() returns a |
| 1479 | pointer to a pcre_extra block, in which the study_data field points to |
| 1480 | the results of the study. |
| 1481 | |
| 1482 | The returned value from pcre_study() can be passed directly to |
| 1483 | pcre_exec() or pcre_dfa_exec(). However, a pcre_extra block also con- |
| 1484 | tains other fields that can be set by the caller before the block is |
| 1485 | passed; these are described below in the section on matching a pattern. |
| 1486 | |
| 1487 | If studying the pattern does not produce any useful information, |
| 1488 | pcre_study() returns NULL. In that circumstance, if the calling program |
| 1489 | wants to pass any of the other fields to pcre_exec() or |
| 1490 | pcre_dfa_exec(), it must set up its own pcre_extra block. |
| 1491 | |
| 1492 | The second argument of pcre_study() contains option bits. There is only |
| 1493 | one option: PCRE_STUDY_JIT_COMPILE. If this is set, and the just-in- |
| 1494 | time compiler is available, the pattern is further compiled into |
| 1495 | machine code that executes much faster than the pcre_exec() matching |
| 1496 | function. If the just-in-time compiler is not available, this option is |
| 1497 | ignored. All other bits in the options argument must be zero. |
| 1498 | |
| 1499 | JIT compilation is a heavyweight optimization. It can take some time |
| 1500 | for patterns to be analyzed, and for one-off matches and simple pat- |
| 1501 | terns the benefit of faster execution might be offset by a much slower |
| 1502 | study time. Not all patterns can be optimized by the JIT compiler. For |
| 1503 | those that cannot be handled, matching automatically falls back to the |
| 1504 | pcre_exec() interpreter. For more details, see the pcrejit documenta- |
| 1505 | tion. |
| 1506 | |
| 1507 | The third argument for pcre_study() is a pointer for an error message. |
| 1508 | If studying succeeds (even if no data is returned), the variable it |
| 1509 | points to is set to NULL. Otherwise it is set to point to a textual |
| 1510 | error message. This is a static string that is part of the library. You |
| 1511 | must not try to free it. You should test the error pointer for NULL |
| 1512 | after calling pcre_study(), to be sure that it has run successfully. |
| 1513 | |
| 1514 | When you are finished with a pattern, you can free the memory used for |
| 1515 | the study data by calling pcre_free_study(). This function was added to |
| 1516 | the API for release 8.20. For earlier versions, the memory could be |
| 1517 | freed with pcre_free(), just like the pattern itself. This will still |
| 1518 | work in cases where PCRE_STUDY_JIT_COMPILE is not used, but it is |
| 1519 | advisable to change to the new function when convenient. |
| 1520 | |
| 1521 | This is a typical way in which pcre_study() is used (except that in a |
| 1522 | real application there should be tests for errors): |
| 1523 | |
| 1524 | int rc; |
| 1525 | pcre *re; |
| 1526 | pcre_extra *sd; |
| 1527 | re = pcre_compile("pattern", 0, &error, &erroroffset, NULL); |
| 1528 | sd = pcre_study( |
| 1529 | re, /* result of pcre_compile() */ |
| 1530 | 0, /* no options */ |
| 1531 | &error); /* set to NULL or points to a message */ |
| 1532 | rc = pcre_exec( /* see below for details of pcre_exec() options */ |
| 1533 | re, sd, "subject", 7, 0, 0, ovector, 30); |
| 1534 | ... |
| 1535 | pcre_free_study(sd); |
| 1536 | pcre_free(re); |
| 1537 | |
| 1538 | Studying a pattern does two things: first, a lower bound for the length |
| 1539 | of subject string that is needed to match the pattern is computed. This |
| 1540 | does not mean that there are any strings of that length that match, but |
| 1541 | it does guarantee that no shorter strings match. The value is used by |
| 1542 | pcre_exec() and pcre_dfa_exec() to avoid wasting time by trying to |
| 1543 | match strings that are shorter than the lower bound. You can find out |
| 1544 | the value in a calling program via the pcre_fullinfo() function. |
| 1545 | |
| 1546 | Studying a pattern is also useful for non-anchored patterns that do not |
| 1547 | have a single fixed starting character. A bitmap of possible starting |
| 1548 | bytes is created. This speeds up finding a position in the subject at |
| 1549 | which to start matching. |
| 1550 | |
| 1551 | These two optimizations apply to both pcre_exec() and pcre_dfa_exec(). |
| 1552 | However, they are not used by pcre_exec() if pcre_study() is called |
| 1553 | with the PCRE_STUDY_JIT_COMPILE option, and just-in-time compiling is |
| 1554 | successful. The optimizations can be disabled by setting the |
| 1555 | PCRE_NO_START_OPTIMIZE option when calling pcre_exec() or |
| 1556 | pcre_dfa_exec(). You might want to do this if your pattern contains |
| 1557 | callouts or (*MARK) (which cannot be handled by the JIT compiler), and |
| 1558 | you want to make use of these facilities in cases where matching fails. |
| 1559 | See the discussion of PCRE_NO_START_OPTIMIZE below. |
| 1560 | |
| 1561 | |
| 1562 | LOCALE SUPPORT |
| 1563 | |
| 1564 | PCRE handles caseless matching, and determines whether characters are |
| 1565 | letters, digits, or whatever, by reference to a set of tables, indexed |
| 1566 | by character value. When running in UTF-8 mode, this applies only to |
| 1567 | characters with codes less than 128. By default, higher-valued codes |
| 1568 | never match escapes such as \w or \d, but they can be tested with \p if |
| 1569 | PCRE is built with Unicode character property support. Alternatively, |
| 1570 | the PCRE_UCP option can be set at compile time; this causes \w and |
| 1571 | friends to use Unicode property support instead of built-in tables. The |
| 1572 | use of locales with Unicode is discouraged. If you are handling charac- |
| 1573 | ters with codes greater than 128, you should either use UTF-8 and Uni- |
| 1574 | code, or use locales, but not try to mix the two. |
| 1575 | |
| 1576 | PCRE contains an internal set of tables that are used when the final |
| 1577 | argument of pcre_compile() is NULL. These are sufficient for many |
| 1578 | applications. Normally, the internal tables recognize only ASCII char- |
| 1579 | acters. However, when PCRE is built, it is possible to cause the inter- |
| 1580 | nal tables to be rebuilt in the default "C" locale of the local system, |
| 1581 | which may cause them to be different. |
| 1582 | |
| 1583 | The internal tables can always be overridden by tables supplied by the |
| 1584 | application that calls PCRE. These may be created in a different locale |
| 1585 | from the default. As more and more applications change to using Uni- |
| 1586 | code, the need for this locale support is expected to die away. |
| 1587 | |
| 1588 | External tables are built by calling the pcre_maketables() function, |
| 1589 | which has no arguments, in the relevant locale. The result can then be |
| 1590 | passed to pcre_compile() or pcre_exec() as often as necessary. For |
| 1591 | example, to build and use tables that are appropriate for the French |
| 1592 | locale (where accented characters with values greater than 128 are |
| 1593 | treated as letters), the following code could be used: |
| 1594 | |
| 1595 | setlocale(LC_CTYPE, "fr_FR"); |
| 1596 | tables = pcre_maketables(); |
| 1597 | re = pcre_compile(..., tables); |
| 1598 | |
| 1599 | The locale name "fr_FR" is used on Linux and other Unix-like systems; |
| 1600 | if you are using Windows, the name for the French locale is "french". |
| 1601 | |
| 1602 | When pcre_maketables() runs, the tables are built in memory that is |
| 1603 | obtained via pcre_malloc. It is the caller's responsibility to ensure |
| 1604 | that the memory containing the tables remains available for as long as |
| 1605 | it is needed. |
| 1606 | |
| 1607 | The pointer that is passed to pcre_compile() is saved with the compiled |
| 1608 | pattern, and the same tables are used via this pointer by pcre_study() |
| 1609 | and normally also by pcre_exec(). Thus, by default, for any single pat- |
| 1610 | tern, compilation, studying and matching all happen in the same locale, |
| 1611 | but different patterns can be compiled in different locales. |
| 1612 | |
| 1613 | It is possible to pass a table pointer or NULL (indicating the use of |
| 1614 | the internal tables) to pcre_exec(). Although not intended for this |
| 1615 | purpose, this facility could be used to match a pattern in a different |
| 1616 | locale from the one in which it was compiled. Passing table pointers at |
| 1617 | run time is discussed below in the section on matching a pattern. |
| 1618 | |
| 1619 | |
| 1620 | INFORMATION ABOUT A PATTERN |
| 1621 | |
| 1622 | int pcre_fullinfo(const pcre *code, const pcre_extra *extra, |
| 1623 | int what, void *where); |
| 1624 | |
| 1625 | The pcre_fullinfo() function returns information about a compiled pat- |
| 1626 | tern. It replaces the obsolete pcre_info() function, which is neverthe- |
| 1627 | less retained for backwards compability (and is documented below). |
| 1628 | |
| 1629 | The first argument for pcre_fullinfo() is a pointer to the compiled |
| 1630 | pattern. The second argument is the result of pcre_study(), or NULL if |
| 1631 | the pattern was not studied. The third argument specifies which piece |
| 1632 | of information is required, and the fourth argument is a pointer to a |
| 1633 | variable to receive the data. The yield of the function is zero for |
| 1634 | success, or one of the following negative numbers: |
| 1635 | |
| 1636 | PCRE_ERROR_NULL the argument code was NULL |
| 1637 | the argument where was NULL |
| 1638 | PCRE_ERROR_BADMAGIC the "magic number" was not found |
| 1639 | PCRE_ERROR_BADOPTION the value of what was invalid |
| 1640 | |
| 1641 | The "magic number" is placed at the start of each compiled pattern as |
| 1642 | an simple check against passing an arbitrary memory pointer. Here is a |
| 1643 | typical call of pcre_fullinfo(), to obtain the length of the compiled |
| 1644 | pattern: |
| 1645 | |
| 1646 | int rc; |
| 1647 | size_t length; |
| 1648 | rc = pcre_fullinfo( |
| 1649 | re, /* result of pcre_compile() */ |
| 1650 | sd, /* result of pcre_study(), or NULL */ |
| 1651 | PCRE_INFO_SIZE, /* what is required */ |
| 1652 | &length); /* where to put the data */ |
| 1653 | |
| 1654 | The possible values for the third argument are defined in pcre.h, and |
| 1655 | are as follows: |
| 1656 | |
| 1657 | PCRE_INFO_BACKREFMAX |
| 1658 | |
| 1659 | Return the number of the highest back reference in the pattern. The |
| 1660 | fourth argument should point to an int variable. Zero is returned if |
| 1661 | there are no back references. |
| 1662 | |
| 1663 | PCRE_INFO_CAPTURECOUNT |
| 1664 | |
| 1665 | Return the number of capturing subpatterns in the pattern. The fourth |
| 1666 | argument should point to an int variable. |
| 1667 | |
| 1668 | PCRE_INFO_DEFAULT_TABLES |
| 1669 | |
| 1670 | Return a pointer to the internal default character tables within PCRE. |
| 1671 | The fourth argument should point to an unsigned char * variable. This |
| 1672 | information call is provided for internal use by the pcre_study() func- |
| 1673 | tion. External callers can cause PCRE to use its internal tables by |
| 1674 | passing a NULL table pointer. |
| 1675 | |
| 1676 | PCRE_INFO_FIRSTBYTE |
| 1677 | |
| 1678 | Return information about the first byte of any matched string, for a |
| 1679 | non-anchored pattern. The fourth argument should point to an int vari- |
| 1680 | able. (This option used to be called PCRE_INFO_FIRSTCHAR; the old name |
| 1681 | is still recognized for backwards compatibility.) |
| 1682 | |
| 1683 | If there is a fixed first byte, for example, from a pattern such as |
| 1684 | (cat|cow|coyote), its value is returned. Otherwise, if either |
| 1685 | |
| 1686 | (a) the pattern was compiled with the PCRE_MULTILINE option, and every |
| 1687 | branch starts with "^", or |
| 1688 | |
| 1689 | (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not |
| 1690 | set (if it were set, the pattern would be anchored), |
| 1691 | |
| 1692 | -1 is returned, indicating that the pattern matches only at the start |
| 1693 | of a subject string or after any newline within the string. Otherwise |
| 1694 | -2 is returned. For anchored patterns, -2 is returned. |
| 1695 | |
| 1696 | PCRE_INFO_FIRSTTABLE |
| 1697 | |
| 1698 | If the pattern was studied, and this resulted in the construction of a |
| 1699 | 256-bit table indicating a fixed set of bytes for the first byte in any |
| 1700 | matching string, a pointer to the table is returned. Otherwise NULL is |
| 1701 | returned. The fourth argument should point to an unsigned char * vari- |
| 1702 | able. |
| 1703 | |
| 1704 | PCRE_INFO_HASCRORLF |
| 1705 | |
| 1706 | Return 1 if the pattern contains any explicit matches for CR or LF |
| 1707 | characters, otherwise 0. The fourth argument should point to an int |
| 1708 | variable. An explicit match is either a literal CR or LF character, or |
| 1709 | \r or \n. |
| 1710 | |
| 1711 | PCRE_INFO_JCHANGED |
| 1712 | |
| 1713 | Return 1 if the (?J) or (?-J) option setting is used in the pattern, |
| 1714 | otherwise 0. The fourth argument should point to an int variable. (?J) |
| 1715 | and (?-J) set and unset the local PCRE_DUPNAMES option, respectively. |
| 1716 | |
| 1717 | PCRE_INFO_JIT |
| 1718 | |
| 1719 | Return 1 if the pattern was studied with the PCRE_STUDY_JIT_COMPILE |
| 1720 | option, and just-in-time compiling was successful. The fourth argument |
| 1721 | should point to an int variable. A return value of 0 means that JIT |
| 1722 | support is not available in this version of PCRE, or that the pattern |
| 1723 | was not studied with the PCRE_STUDY_JIT_COMPILE option, or that the JIT |
| 1724 | compiler could not handle this particular pattern. See the pcrejit doc- |
| 1725 | umentation for details of what can and cannot be handled. |
| 1726 | |
| 1727 | PCRE_INFO_JITSIZE |
| 1728 | |
| 1729 | If the pattern was successfully studied with the PCRE_STUDY_JIT_COMPILE |
| 1730 | option, return the size of the JIT compiled code, otherwise return |
| 1731 | zero. The fourth argument should point to a size_t variable. |
| 1732 | |
| 1733 | PCRE_INFO_LASTLITERAL |
| 1734 | |
| 1735 | Return the value of the rightmost literal byte that must exist in any |
| 1736 | matched string, other than at its start, if such a byte has been |
| 1737 | recorded. The fourth argument should point to an int variable. If there |
| 1738 | is no such byte, -1 is returned. For anchored patterns, a last literal |
| 1739 | byte is recorded only if it follows something of variable length. For |
| 1740 | example, for the pattern /^a\d+z\d+/ the returned value is "z", but for |
| 1741 | /^a\dz\d/ the returned value is -1. |
| 1742 | |
| 1743 | PCRE_INFO_MINLENGTH |
| 1744 | |
| 1745 | If the pattern was studied and a minimum length for matching subject |
| 1746 | strings was computed, its value is returned. Otherwise the returned |
| 1747 | value is -1. The value is a number of characters, not bytes (this may |
| 1748 | be relevant in UTF-8 mode). The fourth argument should point to an int |
| 1749 | variable. A non-negative value is a lower bound to the length of any |
| 1750 | matching string. There may not be any strings of that length that do |
| 1751 | actually match, but every string that does match is at least that long. |
| 1752 | |
| 1753 | PCRE_INFO_NAMECOUNT |
| 1754 | PCRE_INFO_NAMEENTRYSIZE |
| 1755 | PCRE_INFO_NAMETABLE |
| 1756 | |
| 1757 | PCRE supports the use of named as well as numbered capturing parenthe- |
| 1758 | ses. The names are just an additional way of identifying the parenthe- |
| 1759 | ses, which still acquire numbers. Several convenience functions such as |
| 1760 | pcre_get_named_substring() are provided for extracting captured sub- |
| 1761 | strings by name. It is also possible to extract the data directly, by |
| 1762 | first converting the name to a number in order to access the correct |
| 1763 | pointers in the output vector (described with pcre_exec() below). To do |
| 1764 | the conversion, you need to use the name-to-number map, which is |
| 1765 | described by these three values. |
| 1766 | |
| 1767 | The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT |
| 1768 | gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size |
| 1769 | of each entry; both of these return an int value. The entry size |
| 1770 | depends on the length of the longest name. PCRE_INFO_NAMETABLE returns |
| 1771 | a pointer to the first entry of the table (a pointer to char). The |
| 1772 | first two bytes of each entry are the number of the capturing parenthe- |
| 1773 | sis, most significant byte first. The rest of the entry is the corre- |
| 1774 | sponding name, zero terminated. |
| 1775 | |
| 1776 | The names are in alphabetical order. Duplicate names may appear if (?| |
| 1777 | is used to create multiple groups with the same number, as described in |
| 1778 | the section on duplicate subpattern numbers in the pcrepattern page. |
| 1779 | Duplicate names for subpatterns with different numbers are permitted |
| 1780 | only if PCRE_DUPNAMES is set. In all cases of duplicate names, they |
| 1781 | appear in the table in the order in which they were found in the pat- |
| 1782 | tern. In the absence of (?| this is the order of increasing number; |
| 1783 | when (?| is used this is not necessarily the case because later subpat- |
| 1784 | terns may have lower numbers. |
| 1785 | |
| 1786 | As a simple example of the name/number table, consider the following |
| 1787 | pattern (assume PCRE_EXTENDED is set, so white space - including new- |
| 1788 | lines - is ignored): |
| 1789 | |
| 1790 | (?<date> (?<year>(\d\d)?\d\d) - |
| 1791 | (?<month>\d\d) - (?<day>\d\d) ) |
| 1792 | |
| 1793 | There are four named subpatterns, so the table has four entries, and |
| 1794 | each entry in the table is eight bytes long. The table is as follows, |
| 1795 | with non-printing bytes shows in hexadecimal, and undefined bytes shown |
| 1796 | as ??: |
| 1797 | |
| 1798 | 00 01 d a t e 00 ?? |
| 1799 | 00 05 d a y 00 ?? ?? |
| 1800 | 00 04 m o n t h 00 |
| 1801 | 00 02 y e a r 00 ?? |
| 1802 | |
| 1803 | When writing code to extract data from named subpatterns using the |
| 1804 | name-to-number map, remember that the length of the entries is likely |
| 1805 | to be different for each compiled pattern. |
| 1806 | |
| 1807 | PCRE_INFO_OKPARTIAL |
| 1808 | |
| 1809 | Return 1 if the pattern can be used for partial matching with |
| 1810 | pcre_exec(), otherwise 0. The fourth argument should point to an int |
| 1811 | variable. From release 8.00, this always returns 1, because the |
| 1812 | restrictions that previously applied to partial matching have been |
| 1813 | lifted. The pcrepartial documentation gives details of partial match- |
| 1814 | ing. |
| 1815 | |
| 1816 | PCRE_INFO_OPTIONS |
| 1817 | |
| 1818 | Return a copy of the options with which the pattern was compiled. The |
| 1819 | fourth argument should point to an unsigned long int variable. These |
| 1820 | option bits are those specified in the call to pcre_compile(), modified |
| 1821 | by any top-level option settings at the start of the pattern itself. In |
| 1822 | other words, they are the options that will be in force when matching |
| 1823 | starts. For example, if the pattern /(?im)abc(?-i)d/ is compiled with |
| 1824 | the PCRE_EXTENDED option, the result is PCRE_CASELESS, PCRE_MULTILINE, |
| 1825 | and PCRE_EXTENDED. |
| 1826 | |
| 1827 | A pattern is automatically anchored by PCRE if all of its top-level |
| 1828 | alternatives begin with one of the following: |
| 1829 | |
| 1830 | ^ unless PCRE_MULTILINE is set |
| 1831 | \A always |
| 1832 | \G always |
| 1833 | .* if PCRE_DOTALL is set and there are no back |
| 1834 | references to the subpattern in which .* appears |
| 1835 | |
| 1836 | For such patterns, the PCRE_ANCHORED bit is set in the options returned |
| 1837 | by pcre_fullinfo(). |
| 1838 | |
| 1839 | PCRE_INFO_SIZE |
| 1840 | |
| 1841 | Return the size of the compiled pattern. The fourth argument should |
| 1842 | point to a size_t variable. This value does not include the size of the |
| 1843 | pcre structure that is returned by pcre_compile(). The value that is |
| 1844 | passed as the argument to pcre_malloc() when pcre_compile() is getting |
| 1845 | memory in which to place the compiled data is the value returned by |
| 1846 | this option plus the size of the pcre structure. Studying a compiled |
| 1847 | pattern, with or without JIT, does not alter the value returned by this |
| 1848 | option. |
| 1849 | |
| 1850 | PCRE_INFO_STUDYSIZE |
| 1851 | |
| 1852 | Return the size of the data block pointed to by the study_data field in |
| 1853 | a pcre_extra block. If pcre_extra is NULL, or there is no study data, |
| 1854 | zero is returned. The fourth argument should point to a size_t vari- |
| 1855 | able. The study_data field is set by pcre_study() to record informa- |
| 1856 | tion that will speed up matching (see the section entitled "Studying a |
| 1857 | pattern" above). The format of the study_data block is private, but its |
| 1858 | length is made available via this option so that it can be saved and |
| 1859 | restored (see the pcreprecompile documentation for details). |
| 1860 | |
| 1861 | |
| 1862 | OBSOLETE INFO FUNCTION |
| 1863 | |
| 1864 | int pcre_info(const pcre *code, int *optptr, int *firstcharptr); |
| 1865 | |
| 1866 | The pcre_info() function is now obsolete because its interface is too |
| 1867 | restrictive to return all the available data about a compiled pattern. |
| 1868 | New programs should use pcre_fullinfo() instead. The yield of |
| 1869 | pcre_info() is the number of capturing subpatterns, or one of the fol- |
| 1870 | lowing negative numbers: |
| 1871 | |
| 1872 | PCRE_ERROR_NULL the argument code was NULL |
| 1873 | PCRE_ERROR_BADMAGIC the "magic number" was not found |
| 1874 | |
| 1875 | If the optptr argument is not NULL, a copy of the options with which |
| 1876 | the pattern was compiled is placed in the integer it points to (see |
| 1877 | PCRE_INFO_OPTIONS above). |
| 1878 | |
| 1879 | If the pattern is not anchored and the firstcharptr argument is not |
| 1880 | NULL, it is used to pass back information about the first character of |
| 1881 | any matched string (see PCRE_INFO_FIRSTBYTE above). |
| 1882 | |
| 1883 | |
| 1884 | REFERENCE COUNTS |
| 1885 | |
| 1886 | int pcre_refcount(pcre *code, int adjust); |
| 1887 | |
| 1888 | The pcre_refcount() function is used to maintain a reference count in |
| 1889 | the data block that contains a compiled pattern. It is provided for the |
| 1890 | benefit of applications that operate in an object-oriented manner, |
| 1891 | where different parts of the application may be using the same compiled |
| 1892 | pattern, but you want to free the block when they are all done. |
| 1893 | |
| 1894 | When a pattern is compiled, the reference count field is initialized to |
| 1895 | zero. It is changed only by calling this function, whose action is to |
| 1896 | add the adjust value (which may be positive or negative) to it. The |
| 1897 | yield of the function is the new value. However, the value of the count |
| 1898 | is constrained to lie between 0 and 65535, inclusive. If the new value |
| 1899 | is outside these limits, it is forced to the appropriate limit value. |
| 1900 | |
| 1901 | Except when it is zero, the reference count is not correctly preserved |
| 1902 | if a pattern is compiled on one host and then transferred to a host |
| 1903 | whose byte-order is different. (This seems a highly unlikely scenario.) |
| 1904 | |
| 1905 | |
| 1906 | MATCHING A PATTERN: THE TRADITIONAL FUNCTION |
| 1907 | |
| 1908 | int pcre_exec(const pcre *code, const pcre_extra *extra, |
| 1909 | const char *subject, int length, int startoffset, |
| 1910 | int options, int *ovector, int ovecsize); |
| 1911 | |
| 1912 | The function pcre_exec() is called to match a subject string against a |
| 1913 | compiled pattern, which is passed in the code argument. If the pattern |
| 1914 | was studied, the result of the study should be passed in the extra |
| 1915 | argument. You can call pcre_exec() with the same code and extra argu- |
| 1916 | ments as many times as you like, in order to match different subject |
| 1917 | strings with the same pattern. |
| 1918 | |
| 1919 | This function is the main matching facility of the library, and it |
| 1920 | operates in a Perl-like manner. For specialist use there is also an |
| 1921 | alternative matching function, which is described below in the section |
| 1922 | about the pcre_dfa_exec() function. |
| 1923 | |
| 1924 | In most applications, the pattern will have been compiled (and option- |
| 1925 | ally studied) in the same process that calls pcre_exec(). However, it |
| 1926 | is possible to save compiled patterns and study data, and then use them |
| 1927 | later in different processes, possibly even on different hosts. For a |
| 1928 | discussion about this, see the pcreprecompile documentation. |
| 1929 | |
| 1930 | Here is an example of a simple call to pcre_exec(): |
| 1931 | |
| 1932 | int rc; |
| 1933 | int ovector[30]; |
| 1934 | rc = pcre_exec( |
| 1935 | re, /* result of pcre_compile() */ |
| 1936 | NULL, /* we didn't study the pattern */ |
| 1937 | "some string", /* the subject string */ |
| 1938 | 11, /* the length of the subject string */ |
| 1939 | 0, /* start at offset 0 in the subject */ |
| 1940 | 0, /* default options */ |
| 1941 | ovector, /* vector of integers for substring information */ |
| 1942 | 30); /* number of elements (NOT size in bytes) */ |
| 1943 | |
| 1944 | Extra data for pcre_exec() |
| 1945 | |
| 1946 | If the extra argument is not NULL, it must point to a pcre_extra data |
| 1947 | block. The pcre_study() function returns such a block (when it doesn't |
| 1948 | return NULL), but you can also create one for yourself, and pass addi- |
| 1949 | tional information in it. The pcre_extra block contains the following |
| 1950 | fields (not necessarily in this order): |
| 1951 | |
| 1952 | unsigned long int flags; |
| 1953 | void *study_data; |
| 1954 | void *executable_jit; |
| 1955 | unsigned long int match_limit; |
| 1956 | unsigned long int match_limit_recursion; |
| 1957 | void *callout_data; |
| 1958 | const unsigned char *tables; |
| 1959 | unsigned char **mark; |
| 1960 | |
| 1961 | The flags field is a bitmap that specifies which of the other fields |
| 1962 | are set. The flag bits are: |
| 1963 | |
| 1964 | PCRE_EXTRA_STUDY_DATA |
| 1965 | PCRE_EXTRA_EXECUTABLE_JIT |
| 1966 | PCRE_EXTRA_MATCH_LIMIT |
| 1967 | PCRE_EXTRA_MATCH_LIMIT_RECURSION |
| 1968 | PCRE_EXTRA_CALLOUT_DATA |
| 1969 | PCRE_EXTRA_TABLES |
| 1970 | PCRE_EXTRA_MARK |
| 1971 | |
| 1972 | Other flag bits should be set to zero. The study_data field and some- |
| 1973 | times the executable_jit field are set in the pcre_extra block that is |
| 1974 | returned by pcre_study(), together with the appropriate flag bits. You |
| 1975 | should not set these yourself, but you may add to the block by setting |
| 1976 | the other fields and their corresponding flag bits. |
| 1977 | |
| 1978 | The match_limit field provides a means of preventing PCRE from using up |
| 1979 | a vast amount of resources when running patterns that are not going to |
| 1980 | match, but which have a very large number of possibilities in their |
| 1981 | search trees. The classic example is a pattern that uses nested unlim- |
| 1982 | ited repeats. |
| 1983 | |
| 1984 | Internally, pcre_exec() uses a function called match(), which it calls |
| 1985 | repeatedly (sometimes recursively). The limit set by match_limit is |
| 1986 | imposed on the number of times this function is called during a match, |
| 1987 | which has the effect of limiting the amount of backtracking that can |
| 1988 | take place. For patterns that are not anchored, the count restarts from |
| 1989 | zero for each position in the subject string. |
| 1990 | |
| 1991 | When pcre_exec() is called with a pattern that was successfully studied |
| 1992 | with the PCRE_STUDY_JIT_COMPILE option, the way that the matching is |
| 1993 | executed is entirely different. However, there is still the possibility |
| 1994 | of runaway matching that goes on for a very long time, and so the |
| 1995 | match_limit value is also used in this case (but in a different way) to |
| 1996 | limit how long the matching can continue. |
| 1997 | |
| 1998 | The default value for the limit can be set when PCRE is built; the |
| 1999 | default default is 10 million, which handles all but the most extreme |
| 2000 | cases. You can override the default by suppling pcre_exec() with a |
| 2001 | pcre_extra block in which match_limit is set, and |
| 2002 | PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is |
| 2003 | exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT. |
| 2004 | |
| 2005 | The match_limit_recursion field is similar to match_limit, but instead |
| 2006 | of limiting the total number of times that match() is called, it limits |
| 2007 | the depth of recursion. The recursion depth is a smaller number than |
| 2008 | the total number of calls, because not all calls to match() are recur- |
| 2009 | sive. This limit is of use only if it is set smaller than match_limit. |
| 2010 | |
| 2011 | Limiting the recursion depth limits the amount of machine stack that |
| 2012 | can be used, or, when PCRE has been compiled to use memory on the heap |
| 2013 | instead of the stack, the amount of heap memory that can be used. This |
| 2014 | limit is not relevant, and is ignored, if the pattern was successfully |
| 2015 | studied with PCRE_STUDY_JIT_COMPILE. |
| 2016 | |
| 2017 | The default value for match_limit_recursion can be set when PCRE is |
| 2018 | built; the default default is the same value as the default for |
| 2019 | match_limit. You can override the default by suppling pcre_exec() with |
| 2020 | a pcre_extra block in which match_limit_recursion is set, and |
| 2021 | PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the |
| 2022 | limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT. |
| 2023 | |
| 2024 | The callout_data field is used in conjunction with the "callout" fea- |
| 2025 | ture, and is described in the pcrecallout documentation. |
| 2026 | |
| 2027 | The tables field is used to pass a character tables pointer to |
| 2028 | pcre_exec(); this overrides the value that is stored with the compiled |
| 2029 | pattern. A non-NULL value is stored with the compiled pattern only if |
| 2030 | custom tables were supplied to pcre_compile() via its tableptr argu- |
| 2031 | ment. If NULL is passed to pcre_exec() using this mechanism, it forces |
| 2032 | PCRE's internal tables to be used. This facility is helpful when re- |
| 2033 | using patterns that have been saved after compiling with an external |
| 2034 | set of tables, because the external tables might be at a different |
| 2035 | address when pcre_exec() is called. See the pcreprecompile documenta- |
| 2036 | tion for a discussion of saving compiled patterns for later use. |
| 2037 | |
| 2038 | If PCRE_EXTRA_MARK is set in the flags field, the mark field must be |
| 2039 | set to point to a char * variable. If the pattern contains any back- |
| 2040 | tracking control verbs such as (*MARK:NAME), and the execution ends up |
| 2041 | with a name to pass back, a pointer to the name string (zero termi- |
| 2042 | nated) is placed in the variable pointed to by the mark field. The |
| 2043 | names are within the compiled pattern; if you wish to retain such a |
| 2044 | name you must copy it before freeing the memory of a compiled pattern. |
| 2045 | If there is no name to pass back, the variable pointed to by the mark |
| 2046 | field set to NULL. For details of the backtracking control verbs, see |
| 2047 | the section entitled "Backtracking control" in the pcrepattern documen- |
| 2048 | tation. |
| 2049 | |
| 2050 | Option bits for pcre_exec() |
| 2051 | |
| 2052 | The unused bits of the options argument for pcre_exec() must be zero. |
| 2053 | The only bits that may be set are PCRE_ANCHORED, PCRE_NEWLINE_xxx, |
| 2054 | PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, |
| 2055 | PCRE_NO_START_OPTIMIZE, PCRE_NO_UTF8_CHECK, PCRE_PARTIAL_SOFT, and |
| 2056 | PCRE_PARTIAL_HARD. |
| 2057 | |
| 2058 | If the pattern was successfully studied with the PCRE_STUDY_JIT_COMPILE |
| 2059 | option, the only supported options for JIT execution are |
| 2060 | PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, and |
| 2061 | PCRE_NOTEMPTY_ATSTART. Note in particular that partial matching is not |
| 2062 | supported. If an unsupported option is used, JIT execution is disabled |
| 2063 | and the normal interpretive code in pcre_exec() is run. |
| 2064 | |
| 2065 | PCRE_ANCHORED |
| 2066 | |
| 2067 | The PCRE_ANCHORED option limits pcre_exec() to matching at the first |
| 2068 | matching position. If a pattern was compiled with PCRE_ANCHORED, or |
| 2069 | turned out to be anchored by virtue of its contents, it cannot be made |
| 2070 | unachored at matching time. |
| 2071 | |
| 2072 | PCRE_BSR_ANYCRLF |
| 2073 | PCRE_BSR_UNICODE |
| 2074 | |
| 2075 | These options (which are mutually exclusive) control what the \R escape |
| 2076 | sequence matches. The choice is either to match only CR, LF, or CRLF, |
| 2077 | or to match any Unicode newline sequence. These options override the |
| 2078 | choice that was made or defaulted when the pattern was compiled. |
| 2079 | |
| 2080 | PCRE_NEWLINE_CR |
| 2081 | PCRE_NEWLINE_LF |
| 2082 | PCRE_NEWLINE_CRLF |
| 2083 | PCRE_NEWLINE_ANYCRLF |
| 2084 | PCRE_NEWLINE_ANY |
| 2085 | |
| 2086 | These options override the newline definition that was chosen or |
| 2087 | defaulted when the pattern was compiled. For details, see the descrip- |
| 2088 | tion of pcre_compile() above. During matching, the newline choice |
| 2089 | affects the behaviour of the dot, circumflex, and dollar metacharac- |
| 2090 | ters. It may also alter the way the match position is advanced after a |
| 2091 | match failure for an unanchored pattern. |
| 2092 | |
| 2093 | When PCRE_NEWLINE_CRLF, PCRE_NEWLINE_ANYCRLF, or PCRE_NEWLINE_ANY is |
| 2094 | set, and a match attempt for an unanchored pattern fails when the cur- |
| 2095 | rent position is at a CRLF sequence, and the pattern contains no |
| 2096 | explicit matches for CR or LF characters, the match position is |
| 2097 | advanced by two characters instead of one, in other words, to after the |
| 2098 | CRLF. |
| 2099 | |
| 2100 | The above rule is a compromise that makes the most common cases work as |
| 2101 | expected. For example, if the pattern is .+A (and the PCRE_DOTALL |
| 2102 | option is not set), it does not match the string "\r\nA" because, after |
| 2103 | failing at the start, it skips both the CR and the LF before retrying. |
| 2104 | However, the pattern [\r\n]A does match that string, because it con- |
| 2105 | tains an explicit CR or LF reference, and so advances only by one char- |
| 2106 | acter after the first failure. |
| 2107 | |
| 2108 | An explicit match for CR of LF is either a literal appearance of one of |
| 2109 | those characters, or one of the \r or \n escape sequences. Implicit |
| 2110 | matches such as [^X] do not count, nor does \s (which includes CR and |
| 2111 | LF in the characters that it matches). |
| 2112 | |
| 2113 | Notwithstanding the above, anomalous effects may still occur when CRLF |
| 2114 | is a valid newline sequence and explicit \r or \n escapes appear in the |
| 2115 | pattern. |
| 2116 | |
| 2117 | PCRE_NOTBOL |
| 2118 | |
| 2119 | This option specifies that first character of the subject string is not |
| 2120 | the beginning of a line, so the circumflex metacharacter should not |
| 2121 | match before it. Setting this without PCRE_MULTILINE (at compile time) |
| 2122 | causes circumflex never to match. This option affects only the behav- |
| 2123 | iour of the circumflex metacharacter. It does not affect \A. |
| 2124 | |
| 2125 | PCRE_NOTEOL |
| 2126 | |
| 2127 | This option specifies that the end of the subject string is not the end |
| 2128 | of a line, so the dollar metacharacter should not match it nor (except |
| 2129 | in multiline mode) a newline immediately before it. Setting this with- |
| 2130 | out PCRE_MULTILINE (at compile time) causes dollar never to match. This |
| 2131 | option affects only the behaviour of the dollar metacharacter. It does |
| 2132 | not affect \Z or \z. |
| 2133 | |
| 2134 | PCRE_NOTEMPTY |
| 2135 | |
| 2136 | An empty string is not considered to be a valid match if this option is |
| 2137 | set. If there are alternatives in the pattern, they are tried. If all |
| 2138 | the alternatives match the empty string, the entire match fails. For |
| 2139 | example, if the pattern |
| 2140 | |
| 2141 | a?b? |
| 2142 | |
| 2143 | is applied to a string not beginning with "a" or "b", it matches an |
| 2144 | empty string at the start of the subject. With PCRE_NOTEMPTY set, this |
| 2145 | match is not valid, so PCRE searches further into the string for occur- |
| 2146 | rences of "a" or "b". |
| 2147 | |
| 2148 | PCRE_NOTEMPTY_ATSTART |
| 2149 | |
| 2150 | This is like PCRE_NOTEMPTY, except that an empty string match that is |
| 2151 | not at the start of the subject is permitted. If the pattern is |
| 2152 | anchored, such a match can occur only if the pattern contains \K. |
| 2153 | |
| 2154 | Perl has no direct equivalent of PCRE_NOTEMPTY or |
| 2155 | PCRE_NOTEMPTY_ATSTART, but it does make a special case of a pattern |
| 2156 | match of the empty string within its split() function, and when using |
| 2157 | the /g modifier. It is possible to emulate Perl's behaviour after |
| 2158 | matching a null string by first trying the match again at the same off- |
| 2159 | set with PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED, and then if that |
| 2160 | fails, by advancing the starting offset (see below) and trying an ordi- |
| 2161 | nary match again. There is some code that demonstrates how to do this |
| 2162 | in the pcredemo sample program. In the most general case, you have to |
| 2163 | check to see if the newline convention recognizes CRLF as a newline, |
| 2164 | and if so, and the current character is CR followed by LF, advance the |
| 2165 | starting offset by two characters instead of one. |
| 2166 | |
| 2167 | PCRE_NO_START_OPTIMIZE |
| 2168 | |
| 2169 | There are a number of optimizations that pcre_exec() uses at the start |
| 2170 | of a match, in order to speed up the process. For example, if it is |
| 2171 | known that an unanchored match must start with a specific character, it |
| 2172 | searches the subject for that character, and fails immediately if it |
| 2173 | cannot find it, without actually running the main matching function. |
| 2174 | This means that a special item such as (*COMMIT) at the start of a pat- |
| 2175 | tern is not considered until after a suitable starting point for the |
| 2176 | match has been found. When callouts or (*MARK) items are in use, these |
| 2177 | "start-up" optimizations can cause them to be skipped if the pattern is |
| 2178 | never actually used. The start-up optimizations are in effect a pre- |
| 2179 | scan of the subject that takes place before the pattern is run. |
| 2180 | |
| 2181 | The PCRE_NO_START_OPTIMIZE option disables the start-up optimizations, |
| 2182 | possibly causing performance to suffer, but ensuring that in cases |
| 2183 | where the result is "no match", the callouts do occur, and that items |
| 2184 | such as (*COMMIT) and (*MARK) are considered at every possible starting |
| 2185 | position in the subject string. If PCRE_NO_START_OPTIMIZE is set at |
| 2186 | compile time, it cannot be unset at matching time. |
| 2187 | |
| 2188 | Setting PCRE_NO_START_OPTIMIZE can change the outcome of a matching |
| 2189 | operation. Consider the pattern |
| 2190 | |
| 2191 | (*COMMIT)ABC |
| 2192 | |
| 2193 | When this is compiled, PCRE records the fact that a match must start |
| 2194 | with the character "A". Suppose the subject string is "DEFABC". The |
| 2195 | start-up optimization scans along the subject, finds "A" and runs the |
| 2196 | first match attempt from there. The (*COMMIT) item means that the pat- |
| 2197 | tern must match the current starting position, which in this case, it |
| 2198 | does. However, if the same match is run with PCRE_NO_START_OPTIMIZE |
| 2199 | set, the initial scan along the subject string does not happen. The |
| 2200 | first match attempt is run starting from "D" and when this fails, |
| 2201 | (*COMMIT) prevents any further matches being tried, so the overall |
| 2202 | result is "no match". If the pattern is studied, more start-up opti- |
| 2203 | mizations may be used. For example, a minimum length for the subject |
| 2204 | may be recorded. Consider the pattern |
| 2205 | |
| 2206 | (*MARK:A)(X|Y) |
| 2207 | |
| 2208 | The minimum length for a match is one character. If the subject is |
| 2209 | "ABC", there will be attempts to match "ABC", "BC", "C", and then |
| 2210 | finally an empty string. If the pattern is studied, the final attempt |
| 2211 | does not take place, because PCRE knows that the subject is too short, |
| 2212 | and so the (*MARK) is never encountered. In this case, studying the |
| 2213 | pattern does not affect the overall match result, which is still "no |
| 2214 | match", but it does affect the auxiliary information that is returned. |
| 2215 | |
| 2216 | PCRE_NO_UTF8_CHECK |
| 2217 | |
| 2218 | When PCRE_UTF8 is set at compile time, the validity of the subject as a |
| 2219 | UTF-8 string is automatically checked when pcre_exec() is subsequently |
| 2220 | called. The value of startoffset is also checked to ensure that it |
| 2221 | points to the start of a UTF-8 character. There is a discussion about |
| 2222 | the validity of UTF-8 strings in the section on UTF-8 support in the |
| 2223 | main pcre page. If an invalid UTF-8 sequence of bytes is found, |
| 2224 | pcre_exec() returns the error PCRE_ERROR_BADUTF8 or, if PCRE_PAR- |
| 2225 | TIAL_HARD is set and the problem is a truncated UTF-8 character at the |
| 2226 | end of the subject, PCRE_ERROR_SHORTUTF8. In both cases, information |
| 2227 | about the precise nature of the error may also be returned (see the |
| 2228 | descriptions of these errors in the section entitled Error return val- |
| 2229 | ues from pcre_exec() below). If startoffset contains a value that does |
| 2230 | not point to the start of a UTF-8 character (or to the end of the sub- |
| 2231 | ject), PCRE_ERROR_BADUTF8_OFFSET is returned. |
| 2232 | |
| 2233 | If you already know that your subject is valid, and you want to skip |
| 2234 | these checks for performance reasons, you can set the |
| 2235 | PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to |
| 2236 | do this for the second and subsequent calls to pcre_exec() if you are |
| 2237 | making repeated calls to find all the matches in a single subject |
| 2238 | string. However, you should be sure that the value of startoffset |
| 2239 | points to the start of a UTF-8 character (or the end of the subject). |
| 2240 | When PCRE_NO_UTF8_CHECK is set, the effect of passing an invalid UTF-8 |
| 2241 | string as a subject or an invalid value of startoffset is undefined. |
| 2242 | Your program may crash. |
| 2243 | |
| 2244 | PCRE_PARTIAL_HARD |
| 2245 | PCRE_PARTIAL_SOFT |
| 2246 | |
| 2247 | These options turn on the partial matching feature. For backwards com- |
| 2248 | patibility, PCRE_PARTIAL is a synonym for PCRE_PARTIAL_SOFT. A partial |
| 2249 | match occurs if the end of the subject string is reached successfully, |
| 2250 | but there are not enough subject characters to complete the match. If |
| 2251 | this happens when PCRE_PARTIAL_SOFT (but not PCRE_PARTIAL_HARD) is set, |
| 2252 | matching continues by testing any remaining alternatives. Only if no |
| 2253 | complete match can be found is PCRE_ERROR_PARTIAL returned instead of |
| 2254 | PCRE_ERROR_NOMATCH. In other words, PCRE_PARTIAL_SOFT says that the |
| 2255 | caller is prepared to handle a partial match, but only if no complete |
| 2256 | match can be found. |
| 2257 | |
| 2258 | If PCRE_PARTIAL_HARD is set, it overrides PCRE_PARTIAL_SOFT. In this |
| 2259 | case, if a partial match is found, pcre_exec() immediately returns |
| 2260 | PCRE_ERROR_PARTIAL, without considering any other alternatives. In |
| 2261 | other words, when PCRE_PARTIAL_HARD is set, a partial match is consid- |
| 2262 | ered to be more important that an alternative complete match. |
| 2263 | |
| 2264 | In both cases, the portion of the string that was inspected when the |
| 2265 | partial match was found is set as the first matching string. There is a |
| 2266 | more detailed discussion of partial and multi-segment matching, with |
| 2267 | examples, in the pcrepartial documentation. |
| 2268 | |
| 2269 | The string to be matched by pcre_exec() |
| 2270 | |
| 2271 | The subject string is passed to pcre_exec() as a pointer in subject, a |
| 2272 | length (in bytes) in length, and a starting byte offset in startoffset. |
| 2273 | If this is negative or greater than the length of the subject, |
| 2274 | pcre_exec() returns PCRE_ERROR_BADOFFSET. When the starting offset is |
| 2275 | zero, the search for a match starts at the beginning of the subject, |
| 2276 | and this is by far the most common case. In UTF-8 mode, the byte offset |
| 2277 | must point to the start of a UTF-8 character (or the end of the sub- |
| 2278 | ject). Unlike the pattern string, the subject may contain binary zero |
| 2279 | bytes. |
| 2280 | |
| 2281 | A non-zero starting offset is useful when searching for another match |
| 2282 | in the same subject by calling pcre_exec() again after a previous suc- |
| 2283 | cess. Setting startoffset differs from just passing over a shortened |
| 2284 | string and setting PCRE_NOTBOL in the case of a pattern that begins |
| 2285 | with any kind of lookbehind. For example, consider the pattern |
| 2286 | |
| 2287 | \Biss\B |
| 2288 | |
| 2289 | which finds occurrences of "iss" in the middle of words. (\B matches |
| 2290 | only if the current position in the subject is not a word boundary.) |
| 2291 | When applied to the string "Mississipi" the first call to pcre_exec() |
| 2292 | finds the first occurrence. If pcre_exec() is called again with just |
| 2293 | the remainder of the subject, namely "issipi", it does not match, |
| 2294 | because \B is always false at the start of the subject, which is deemed |
| 2295 | to be a word boundary. However, if pcre_exec() is passed the entire |
| 2296 | string again, but with startoffset set to 4, it finds the second occur- |
| 2297 | rence of "iss" because it is able to look behind the starting point to |
| 2298 | discover that it is preceded by a letter. |
| 2299 | |
| 2300 | Finding all the matches in a subject is tricky when the pattern can |
| 2301 | match an empty string. It is possible to emulate Perl's /g behaviour by |
| 2302 | first trying the match again at the same offset, with the |
| 2303 | PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED options, and then if that |
| 2304 | fails, advancing the starting offset and trying an ordinary match |
| 2305 | again. There is some code that demonstrates how to do this in the pcre- |
| 2306 | demo sample program. In the most general case, you have to check to see |
| 2307 | if the newline convention recognizes CRLF as a newline, and if so, and |
| 2308 | the current character is CR followed by LF, advance the starting offset |
| 2309 | by two characters instead of one. |
| 2310 | |
| 2311 | If a non-zero starting offset is passed when the pattern is anchored, |
| 2312 | one attempt to match at the given offset is made. This can only succeed |
| 2313 | if the pattern does not require the match to be at the start of the |
| 2314 | subject. |
| 2315 | |
| 2316 | How pcre_exec() returns captured substrings |
| 2317 | |
| 2318 | In general, a pattern matches a certain portion of the subject, and in |
| 2319 | addition, further substrings from the subject may be picked out by |
| 2320 | parts of the pattern. Following the usage in Jeffrey Friedl's book, |
| 2321 | this is called "capturing" in what follows, and the phrase "capturing |
| 2322 | subpattern" is used for a fragment of a pattern that picks out a sub- |
| 2323 | string. PCRE supports several other kinds of parenthesized subpattern |
| 2324 | that do not cause substrings to be captured. |
| 2325 | |
| 2326 | Captured substrings are returned to the caller via a vector of integers |
| 2327 | whose address is passed in ovector. The number of elements in the vec- |
| 2328 | tor is passed in ovecsize, which must be a non-negative number. Note: |
| 2329 | this argument is NOT the size of ovector in bytes. |
| 2330 | |
| 2331 | The first two-thirds of the vector is used to pass back captured sub- |
| 2332 | strings, each substring using a pair of integers. The remaining third |
| 2333 | of the vector is used as workspace by pcre_exec() while matching cap- |
| 2334 | turing subpatterns, and is not available for passing back information. |
| 2335 | The number passed in ovecsize should always be a multiple of three. If |
| 2336 | it is not, it is rounded down. |
| 2337 | |
| 2338 | When a match is successful, information about captured substrings is |
| 2339 | returned in pairs of integers, starting at the beginning of ovector, |
| 2340 | and continuing up to two-thirds of its length at the most. The first |
| 2341 | element of each pair is set to the byte offset of the first character |
| 2342 | in a substring, and the second is set to the byte offset of the first |
| 2343 | character after the end of a substring. Note: these values are always |
| 2344 | byte offsets, even in UTF-8 mode. They are not character counts. |
| 2345 | |
| 2346 | The first pair of integers, ovector[0] and ovector[1], identify the |
| 2347 | portion of the subject string matched by the entire pattern. The next |
| 2348 | pair is used for the first capturing subpattern, and so on. The value |
| 2349 | returned by pcre_exec() is one more than the highest numbered pair that |
| 2350 | has been set. For example, if two substrings have been captured, the |
| 2351 | returned value is 3. If there are no capturing subpatterns, the return |
| 2352 | value from a successful match is 1, indicating that just the first pair |
| 2353 | of offsets has been set. |
| 2354 | |
| 2355 | If a capturing subpattern is matched repeatedly, it is the last portion |
| 2356 | of the string that it matched that is returned. |
| 2357 | |
| 2358 | If the vector is too small to hold all the captured substring offsets, |
| 2359 | it is used as far as possible (up to two-thirds of its length), and the |
| 2360 | function returns a value of zero. If neither the actual string matched |
| 2361 | not any captured substrings are of interest, pcre_exec() may be called |
| 2362 | with ovector passed as NULL and ovecsize as zero. However, if the pat- |
| 2363 | tern contains back references and the ovector is not big enough to |
| 2364 | remember the related substrings, PCRE has to get additional memory for |
| 2365 | use during matching. Thus it is usually advisable to supply an ovector |
| 2366 | of reasonable size. |
| 2367 | |
| 2368 | There are some cases where zero is returned (indicating vector over- |
| 2369 | flow) when in fact the vector is exactly the right size for the final |
| 2370 | match. For example, consider the pattern |
| 2371 | |
| 2372 | (a)(?:(b)c|bd) |
| 2373 | |
| 2374 | If a vector of 6 elements (allowing for only 1 captured substring) is |
| 2375 | given with subject string "abd", pcre_exec() will try to set the second |
| 2376 | captured string, thereby recording a vector overflow, before failing to |
| 2377 | match "c" and backing up to try the second alternative. The zero |
| 2378 | return, however, does correctly indicate that the maximum number of |
| 2379 | slots (namely 2) have been filled. In similar cases where there is tem- |
| 2380 | porary overflow, but the final number of used slots is actually less |
| 2381 | than the maximum, a non-zero value is returned. |
| 2382 | |
| 2383 | The pcre_fullinfo() function can be used to find out how many capturing |
| 2384 | subpatterns there are in a compiled pattern. The smallest size for |
| 2385 | ovector that will allow for n captured substrings, in addition to the |
| 2386 | offsets of the substring matched by the whole pattern, is (n+1)*3. |
| 2387 | |
| 2388 | It is possible for capturing subpattern number n+1 to match some part |
| 2389 | of the subject when subpattern n has not been used at all. For example, |
| 2390 | if the string "abc" is matched against the pattern (a|(z))(bc) the |
| 2391 | return from the function is 4, and subpatterns 1 and 3 are matched, but |
| 2392 | 2 is not. When this happens, both values in the offset pairs corre- |
| 2393 | sponding to unused subpatterns are set to -1. |
| 2394 | |
| 2395 | Offset values that correspond to unused subpatterns at the end of the |
| 2396 | expression are also set to -1. For example, if the string "abc" is |
| 2397 | matched against the pattern (abc)(x(yz)?)? subpatterns 2 and 3 are not |
| 2398 | matched. The return from the function is 2, because the highest used |
| 2399 | capturing subpattern number is 1, and the offsets for for the second |
| 2400 | and third capturing subpatterns (assuming the vector is large enough, |
| 2401 | of course) are set to -1. |
| 2402 | |
| 2403 | Note: Elements in the first two-thirds of ovector that do not corre- |
| 2404 | spond to capturing parentheses in the pattern are never changed. That |
| 2405 | is, if a pattern contains n capturing parentheses, no more than ovec- |
| 2406 | tor[0] to ovector[2n+1] are set by pcre_exec(). The other elements (in |
| 2407 | the first two-thirds) retain whatever values they previously had. |
| 2408 | |
| 2409 | Some convenience functions are provided for extracting the captured |
| 2410 | substrings as separate strings. These are described below. |
| 2411 | |
| 2412 | Error return values from pcre_exec() |
| 2413 | |
| 2414 | If pcre_exec() fails, it returns a negative number. The following are |
| 2415 | defined in the header file: |
| 2416 | |
| 2417 | PCRE_ERROR_NOMATCH (-1) |
| 2418 | |
| 2419 | The subject string did not match the pattern. |
| 2420 | |
| 2421 | PCRE_ERROR_NULL (-2) |
| 2422 | |
| 2423 | Either code or subject was passed as NULL, or ovector was NULL and |
| 2424 | ovecsize was not zero. |
| 2425 | |
| 2426 | PCRE_ERROR_BADOPTION (-3) |
| 2427 | |
| 2428 | An unrecognized bit was set in the options argument. |
| 2429 | |
| 2430 | PCRE_ERROR_BADMAGIC (-4) |
| 2431 | |
| 2432 | PCRE stores a 4-byte "magic number" at the start of the compiled code, |
| 2433 | to catch the case when it is passed a junk pointer and to detect when a |
| 2434 | pattern that was compiled in an environment of one endianness is run in |
| 2435 | an environment with the other endianness. This is the error that PCRE |
| 2436 | gives when the magic number is not present. |
| 2437 | |
| 2438 | PCRE_ERROR_UNKNOWN_OPCODE (-5) |
| 2439 | |
| 2440 | While running the pattern match, an unknown item was encountered in the |
| 2441 | compiled pattern. This error could be caused by a bug in PCRE or by |
| 2442 | overwriting of the compiled pattern. |
| 2443 | |
| 2444 | PCRE_ERROR_NOMEMORY (-6) |
| 2445 | |
| 2446 | If a pattern contains back references, but the ovector that is passed |
| 2447 | to pcre_exec() is not big enough to remember the referenced substrings, |
| 2448 | PCRE gets a block of memory at the start of matching to use for this |
| 2449 | purpose. If the call via pcre_malloc() fails, this error is given. The |
| 2450 | memory is automatically freed at the end of matching. |
| 2451 | |
| 2452 | This error is also given if pcre_stack_malloc() fails in pcre_exec(). |
| 2453 | This can happen only when PCRE has been compiled with --disable-stack- |
| 2454 | for-recursion. |
| 2455 | |
| 2456 | PCRE_ERROR_NOSUBSTRING (-7) |
| 2457 | |
| 2458 | This error is used by the pcre_copy_substring(), pcre_get_substring(), |
| 2459 | and pcre_get_substring_list() functions (see below). It is never |
| 2460 | returned by pcre_exec(). |
| 2461 | |
| 2462 | PCRE_ERROR_MATCHLIMIT (-8) |
| 2463 | |
| 2464 | The backtracking limit, as specified by the match_limit field in a |
| 2465 | pcre_extra structure (or defaulted) was reached. See the description |
| 2466 | above. |
| 2467 | |
| 2468 | PCRE_ERROR_CALLOUT (-9) |
| 2469 | |
| 2470 | This error is never generated by pcre_exec() itself. It is provided for |
| 2471 | use by callout functions that want to yield a distinctive error code. |
| 2472 | See the pcrecallout documentation for details. |
| 2473 | |
| 2474 | PCRE_ERROR_BADUTF8 (-10) |
| 2475 | |
| 2476 | A string that contains an invalid UTF-8 byte sequence was passed as a |
| 2477 | subject, and the PCRE_NO_UTF8_CHECK option was not set. If the size of |
| 2478 | the output vector (ovecsize) is at least 2, the byte offset to the |
| 2479 | start of the the invalid UTF-8 character is placed in the first ele- |
| 2480 | ment, and a reason code is placed in the second element. The reason |
| 2481 | codes are listed in the following section. For backward compatibility, |
| 2482 | if PCRE_PARTIAL_HARD is set and the problem is a truncated UTF-8 char- |
| 2483 | acter at the end of the subject (reason codes 1 to 5), |
| 2484 | PCRE_ERROR_SHORTUTF8 is returned instead of PCRE_ERROR_BADUTF8. |
| 2485 | |
| 2486 | PCRE_ERROR_BADUTF8_OFFSET (-11) |
| 2487 | |
| 2488 | The UTF-8 byte sequence that was passed as a subject was checked and |
| 2489 | found to be valid (the PCRE_NO_UTF8_CHECK option was not set), but the |
| 2490 | value of startoffset did not point to the beginning of a UTF-8 charac- |
| 2491 | ter or the end of the subject. |
| 2492 | |
| 2493 | PCRE_ERROR_PARTIAL (-12) |
| 2494 | |
| 2495 | The subject string did not match, but it did match partially. See the |
| 2496 | pcrepartial documentation for details of partial matching. |
| 2497 | |
| 2498 | PCRE_ERROR_BADPARTIAL (-13) |
| 2499 | |
| 2500 | This code is no longer in use. It was formerly returned when the |
| 2501 | PCRE_PARTIAL option was used with a compiled pattern containing items |
| 2502 | that were not supported for partial matching. From release 8.00 |
| 2503 | onwards, there are no restrictions on partial matching. |
| 2504 | |
| 2505 | PCRE_ERROR_INTERNAL (-14) |
| 2506 | |
| 2507 | An unexpected internal error has occurred. This error could be caused |
| 2508 | by a bug in PCRE or by overwriting of the compiled pattern. |
| 2509 | |
| 2510 | PCRE_ERROR_BADCOUNT (-15) |
| 2511 | |
| 2512 | This error is given if the value of the ovecsize argument is negative. |
| 2513 | |
| 2514 | PCRE_ERROR_RECURSIONLIMIT (-21) |
| 2515 | |
| 2516 | The internal recursion limit, as specified by the match_limit_recursion |
| 2517 | field in a pcre_extra structure (or defaulted) was reached. See the |
| 2518 | description above. |
| 2519 | |
| 2520 | PCRE_ERROR_BADNEWLINE (-23) |
| 2521 | |
| 2522 | An invalid combination of PCRE_NEWLINE_xxx options was given. |
| 2523 | |
| 2524 | PCRE_ERROR_BADOFFSET (-24) |
| 2525 | |
| 2526 | The value of startoffset was negative or greater than the length of the |
| 2527 | subject, that is, the value in length. |
| 2528 | |
| 2529 | PCRE_ERROR_SHORTUTF8 (-25) |
| 2530 | |
| 2531 | This error is returned instead of PCRE_ERROR_BADUTF8 when the subject |
| 2532 | string ends with a truncated UTF-8 character and the PCRE_PARTIAL_HARD |
| 2533 | option is set. Information about the failure is returned as for |
| 2534 | PCRE_ERROR_BADUTF8. It is in fact sufficient to detect this case, but |
| 2535 | this special error code for PCRE_PARTIAL_HARD precedes the implementa- |
| 2536 | tion of returned information; it is retained for backwards compatibil- |
| 2537 | ity. |
| 2538 | |
| 2539 | PCRE_ERROR_RECURSELOOP (-26) |
| 2540 | |
| 2541 | This error is returned when pcre_exec() detects a recursion loop within |
| 2542 | the pattern. Specifically, it means that either the whole pattern or a |
| 2543 | subpattern has been called recursively for the second time at the same |
| 2544 | position in the subject string. Some simple patterns that might do this |
| 2545 | are detected and faulted at compile time, but more complicated cases, |
| 2546 | in particular mutual recursions between two different subpatterns, can- |
| 2547 | not be detected until run time. |
| 2548 | |
| 2549 | PCRE_ERROR_JIT_STACKLIMIT (-27) |
| 2550 | |
| 2551 | This error is returned when a pattern that was successfully studied |
| 2552 | using the PCRE_STUDY_JIT_COMPILE option is being matched, but the mem- |
| 2553 | ory available for the just-in-time processing stack is not large |
| 2554 | enough. See the pcrejit documentation for more details. |
| 2555 | |
| 2556 | Error numbers -16 to -20 and -22 are not used by pcre_exec(). |
| 2557 | |
| 2558 | Reason codes for invalid UTF-8 strings |
| 2559 | |
| 2560 | When pcre_exec() returns either PCRE_ERROR_BADUTF8 or PCRE_ERROR_SHORT- |
| 2561 | UTF8, and the size of the output vector (ovecsize) is at least 2, the |
| 2562 | offset of the start of the invalid UTF-8 character is placed in the |
| 2563 | first output vector element (ovector[0]) and a reason code is placed in |
| 2564 | the second element (ovector[1]). The reason codes are given names in |
| 2565 | the pcre.h header file: |
| 2566 | |
| 2567 | PCRE_UTF8_ERR1 |
| 2568 | PCRE_UTF8_ERR2 |
| 2569 | PCRE_UTF8_ERR3 |
| 2570 | PCRE_UTF8_ERR4 |
| 2571 | PCRE_UTF8_ERR5 |
| 2572 | |
| 2573 | The string ends with a truncated UTF-8 character; the code specifies |
| 2574 | how many bytes are missing (1 to 5). Although RFC 3629 restricts UTF-8 |
| 2575 | characters to be no longer than 4 bytes, the encoding scheme (origi- |
| 2576 | nally defined by RFC 2279) allows for up to 6 bytes, and this is |
| 2577 | checked first; hence the possibility of 4 or 5 missing bytes. |
| 2578 | |
| 2579 | PCRE_UTF8_ERR6 |
| 2580 | PCRE_UTF8_ERR7 |
| 2581 | PCRE_UTF8_ERR8 |
| 2582 | PCRE_UTF8_ERR9 |
| 2583 | PCRE_UTF8_ERR10 |
| 2584 | |
| 2585 | The two most significant bits of the 2nd, 3rd, 4th, 5th, or 6th byte of |
| 2586 | the character do not have the binary value 0b10 (that is, either the |
| 2587 | most significant bit is 0, or the next bit is 1). |
| 2588 | |
| 2589 | PCRE_UTF8_ERR11 |
| 2590 | PCRE_UTF8_ERR12 |
| 2591 | |
| 2592 | A character that is valid by the RFC 2279 rules is either 5 or 6 bytes |
| 2593 | long; these code points are excluded by RFC 3629. |
| 2594 | |
| 2595 | PCRE_UTF8_ERR13 |
| 2596 | |
| 2597 | A 4-byte character has a value greater than 0x10fff; these code points |
| 2598 | are excluded by RFC 3629. |
| 2599 | |
| 2600 | PCRE_UTF8_ERR14 |
| 2601 | |
| 2602 | A 3-byte character has a value in the range 0xd800 to 0xdfff; this |
| 2603 | range of code points are reserved by RFC 3629 for use with UTF-16, and |
| 2604 | so are excluded from UTF-8. |
| 2605 | |
| 2606 | PCRE_UTF8_ERR15 |
| 2607 | PCRE_UTF8_ERR16 |
| 2608 | PCRE_UTF8_ERR17 |
| 2609 | PCRE_UTF8_ERR18 |
| 2610 | PCRE_UTF8_ERR19 |
| 2611 | |
| 2612 | A 2-, 3-, 4-, 5-, or 6-byte character is "overlong", that is, it codes |
| 2613 | for a value that can be represented by fewer bytes, which is invalid. |
| 2614 | For example, the two bytes 0xc0, 0xae give the value 0x2e, whose cor- |
| 2615 | rect coding uses just one byte. |
| 2616 | |
| 2617 | PCRE_UTF8_ERR20 |
| 2618 | |
| 2619 | The two most significant bits of the first byte of a character have the |
| 2620 | binary value 0b10 (that is, the most significant bit is 1 and the sec- |
| 2621 | ond is 0). Such a byte can only validly occur as the second or subse- |
| 2622 | quent byte of a multi-byte character. |
| 2623 | |
| 2624 | PCRE_UTF8_ERR21 |
| 2625 | |
| 2626 | The first byte of a character has the value 0xfe or 0xff. These values |
| 2627 | can never occur in a valid UTF-8 string. |
| 2628 | |
| 2629 | |
| 2630 | EXTRACTING CAPTURED SUBSTRINGS BY NUMBER |
| 2631 | |
| 2632 | int pcre_copy_substring(const char *subject, int *ovector, |
| 2633 | int stringcount, int stringnumber, char *buffer, |
| 2634 | int buffersize); |
| 2635 | |
| 2636 | int pcre_get_substring(const char *subject, int *ovector, |
| 2637 | int stringcount, int stringnumber, |
| 2638 | const char **stringptr); |
| 2639 | |
| 2640 | int pcre_get_substring_list(const char *subject, |
| 2641 | int *ovector, int stringcount, const char ***listptr); |
| 2642 | |
| 2643 | Captured substrings can be accessed directly by using the offsets |
| 2644 | returned by pcre_exec() in ovector. For convenience, the functions |
| 2645 | pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub- |
| 2646 | string_list() are provided for extracting captured substrings as new, |
| 2647 | separate, zero-terminated strings. These functions identify substrings |
| 2648 | by number. The next section describes functions for extracting named |
| 2649 | substrings. |
| 2650 | |
| 2651 | A substring that contains a binary zero is correctly extracted and has |
| 2652 | a further zero added on the end, but the result is not, of course, a C |
| 2653 | string. However, you can process such a string by referring to the |
| 2654 | length that is returned by pcre_copy_substring() and pcre_get_sub- |
| 2655 | string(). Unfortunately, the interface to pcre_get_substring_list() is |
| 2656 | not adequate for handling strings containing binary zeros, because the |
| 2657 | end of the final string is not independently indicated. |
| 2658 | |
| 2659 | The first three arguments are the same for all three of these func- |
| 2660 | tions: subject is the subject string that has just been successfully |
| 2661 | matched, ovector is a pointer to the vector of integer offsets that was |
| 2662 | passed to pcre_exec(), and stringcount is the number of substrings that |
| 2663 | were captured by the match, including the substring that matched the |
| 2664 | entire regular expression. This is the value returned by pcre_exec() if |
| 2665 | it is greater than zero. If pcre_exec() returned zero, indicating that |
| 2666 | it ran out of space in ovector, the value passed as stringcount should |
| 2667 | be the number of elements in the vector divided by three. |
| 2668 | |
| 2669 | The functions pcre_copy_substring() and pcre_get_substring() extract a |
| 2670 | single substring, whose number is given as stringnumber. A value of |
| 2671 | zero extracts the substring that matched the entire pattern, whereas |
| 2672 | higher values extract the captured substrings. For pcre_copy_sub- |
| 2673 | string(), the string is placed in buffer, whose length is given by |
| 2674 | buffersize, while for pcre_get_substring() a new block of memory is |
| 2675 | obtained via pcre_malloc, and its address is returned via stringptr. |
| 2676 | The yield of the function is the length of the string, not including |
| 2677 | the terminating zero, or one of these error codes: |
| 2678 | |
| 2679 | PCRE_ERROR_NOMEMORY (-6) |
| 2680 | |
| 2681 | The buffer was too small for pcre_copy_substring(), or the attempt to |
| 2682 | get memory failed for pcre_get_substring(). |
| 2683 | |
| 2684 | PCRE_ERROR_NOSUBSTRING (-7) |
| 2685 | |
| 2686 | There is no substring whose number is stringnumber. |
| 2687 | |
| 2688 | The pcre_get_substring_list() function extracts all available sub- |
| 2689 | strings and builds a list of pointers to them. All this is done in a |
| 2690 | single block of memory that is obtained via pcre_malloc. The address of |
| 2691 | the memory block is returned via listptr, which is also the start of |
| 2692 | the list of string pointers. The end of the list is marked by a NULL |
| 2693 | pointer. The yield of the function is zero if all went well, or the |
| 2694 | error code |
| 2695 | |
| 2696 | PCRE_ERROR_NOMEMORY (-6) |
| 2697 | |
| 2698 | if the attempt to get the memory block failed. |
| 2699 | |
| 2700 | When any of these functions encounter a substring that is unset, which |
| 2701 | can happen when capturing subpattern number n+1 matches some part of |
| 2702 | the subject, but subpattern n has not been used at all, they return an |
| 2703 | empty string. This can be distinguished from a genuine zero-length sub- |
| 2704 | string by inspecting the appropriate offset in ovector, which is nega- |
| 2705 | tive for unset substrings. |
| 2706 | |
| 2707 | The two convenience functions pcre_free_substring() and pcre_free_sub- |
| 2708 | string_list() can be used to free the memory returned by a previous |
| 2709 | call of pcre_get_substring() or pcre_get_substring_list(), respec- |
| 2710 | tively. They do nothing more than call the function pointed to by |
| 2711 | pcre_free, which of course could be called directly from a C program. |
| 2712 | However, PCRE is used in some situations where it is linked via a spe- |
| 2713 | cial interface to another programming language that cannot use |
| 2714 | pcre_free directly; it is for these cases that the functions are pro- |
| 2715 | vided. |
| 2716 | |
| 2717 | |
| 2718 | EXTRACTING CAPTURED SUBSTRINGS BY NAME |
| 2719 | |
| 2720 | int pcre_get_stringnumber(const pcre *code, |
| 2721 | const char *name); |
| 2722 | |
| 2723 | int pcre_copy_named_substring(const pcre *code, |
| 2724 | const char *subject, int *ovector, |
| 2725 | int stringcount, const char *stringname, |
| 2726 | char *buffer, int buffersize); |
| 2727 | |
| 2728 | int pcre_get_named_substring(const pcre *code, |
| 2729 | const char *subject, int *ovector, |
| 2730 | int stringcount, const char *stringname, |
| 2731 | const char **stringptr); |
| 2732 | |
| 2733 | To extract a substring by name, you first have to find associated num- |
| 2734 | ber. For example, for this pattern |
| 2735 | |
| 2736 | (a+)b(?<xxx>\d+)... |
| 2737 | |
| 2738 | the number of the subpattern called "xxx" is 2. If the name is known to |
| 2739 | be unique (PCRE_DUPNAMES was not set), you can find the number from the |
| 2740 | name by calling pcre_get_stringnumber(). The first argument is the com- |
| 2741 | piled pattern, and the second is the name. The yield of the function is |
| 2742 | the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if there is no |
| 2743 | subpattern of that name. |
| 2744 | |
| 2745 | Given the number, you can extract the substring directly, or use one of |
| 2746 | the functions described in the previous section. For convenience, there |
| 2747 | are also two functions that do the whole job. |
| 2748 | |
| 2749 | Most of the arguments of pcre_copy_named_substring() and |
| 2750 | pcre_get_named_substring() are the same as those for the similarly |
| 2751 | named functions that extract by number. As these are described in the |
| 2752 | previous section, they are not re-described here. There are just two |
| 2753 | differences: |
| 2754 | |
| 2755 | First, instead of a substring number, a substring name is given. Sec- |
| 2756 | ond, there is an extra argument, given at the start, which is a pointer |
| 2757 | to the compiled pattern. This is needed in order to gain access to the |
| 2758 | name-to-number translation table. |
| 2759 | |
| 2760 | These functions call pcre_get_stringnumber(), and if it succeeds, they |
| 2761 | then call pcre_copy_substring() or pcre_get_substring(), as appropri- |
| 2762 | ate. NOTE: If PCRE_DUPNAMES is set and there are duplicate names, the |
| 2763 | behaviour may not be what you want (see the next section). |
| 2764 | |
| 2765 | Warning: If the pattern uses the (?| feature to set up multiple subpat- |
| 2766 | terns with the same number, as described in the section on duplicate |
| 2767 | subpattern numbers in the pcrepattern page, you cannot use names to |
| 2768 | distinguish the different subpatterns, because names are not included |
| 2769 | in the compiled code. The matching process uses only numbers. For this |
| 2770 | reason, the use of different names for subpatterns of the same number |
| 2771 | causes an error at compile time. |
| 2772 | |
| 2773 | |
| 2774 | DUPLICATE SUBPATTERN NAMES |
| 2775 | |
| 2776 | int pcre_get_stringtable_entries(const pcre *code, |
| 2777 | const char *name, char **first, char **last); |
| 2778 | |
| 2779 | When a pattern is compiled with the PCRE_DUPNAMES option, names for |
| 2780 | subpatterns are not required to be unique. (Duplicate names are always |
| 2781 | allowed for subpatterns with the same number, created by using the (?| |
| 2782 | feature. Indeed, if such subpatterns are named, they are required to |
| 2783 | use the same names.) |
| 2784 | |
| 2785 | Normally, patterns with duplicate names are such that in any one match, |
| 2786 | only one of the named subpatterns participates. An example is shown in |
| 2787 | the pcrepattern documentation. |
| 2788 | |
| 2789 | When duplicates are present, pcre_copy_named_substring() and |
| 2790 | pcre_get_named_substring() return the first substring corresponding to |
| 2791 | the given name that is set. If none are set, PCRE_ERROR_NOSUBSTRING |
| 2792 | (-7) is returned; no data is returned. The pcre_get_stringnumber() |
| 2793 | function returns one of the numbers that are associated with the name, |
| 2794 | but it is not defined which it is. |
| 2795 | |
| 2796 | If you want to get full details of all captured substrings for a given |
| 2797 | name, you must use the pcre_get_stringtable_entries() function. The |
| 2798 | first argument is the compiled pattern, and the second is the name. The |
| 2799 | third and fourth are pointers to variables which are updated by the |
| 2800 | function. After it has run, they point to the first and last entries in |
| 2801 | the name-to-number table for the given name. The function itself |
| 2802 | returns the length of each entry, or PCRE_ERROR_NOSUBSTRING (-7) if |
| 2803 | there are none. The format of the table is described above in the sec- |
| 2804 | tion entitled Information about a pattern above. Given all the rele- |
| 2805 | vant entries for the name, you can extract each of their numbers, and |
| 2806 | hence the captured data, if any. |
| 2807 | |
| 2808 | |
| 2809 | FINDING ALL POSSIBLE MATCHES |
| 2810 | |
| 2811 | The traditional matching function uses a similar algorithm to Perl, |
| 2812 | which stops when it finds the first match, starting at a given point in |
| 2813 | the subject. If you want to find all possible matches, or the longest |
| 2814 | possible match, consider using the alternative matching function (see |
| 2815 | below) instead. If you cannot use the alternative function, but still |
| 2816 | need to find all possible matches, you can kludge it up by making use |
| 2817 | of the callout facility, which is described in the pcrecallout documen- |
| 2818 | tation. |
| 2819 | |
| 2820 | What you have to do is to insert a callout right at the end of the pat- |
| 2821 | tern. When your callout function is called, extract and save the cur- |
| 2822 | rent matched substring. Then return 1, which forces pcre_exec() to |
| 2823 | backtrack and try other alternatives. Ultimately, when it runs out of |
| 2824 | matches, pcre_exec() will yield PCRE_ERROR_NOMATCH. |
| 2825 | |
| 2826 | |
| 2827 | MATCHING A PATTERN: THE ALTERNATIVE FUNCTION |
| 2828 | |
| 2829 | int pcre_dfa_exec(const pcre *code, const pcre_extra *extra, |
| 2830 | const char *subject, int length, int startoffset, |
| 2831 | int options, int *ovector, int ovecsize, |
| 2832 | int *workspace, int wscount); |
| 2833 | |
| 2834 | The function pcre_dfa_exec() is called to match a subject string |
| 2835 | against a compiled pattern, using a matching algorithm that scans the |
| 2836 | subject string just once, and does not backtrack. This has different |
| 2837 | characteristics to the normal algorithm, and is not compatible with |
| 2838 | Perl. Some of the features of PCRE patterns are not supported. Never- |
| 2839 | theless, there are times when this kind of matching can be useful. For |
| 2840 | a discussion of the two matching algorithms, and a list of features |
| 2841 | that pcre_dfa_exec() does not support, see the pcrematching documenta- |
| 2842 | tion. |
| 2843 | |
| 2844 | The arguments for the pcre_dfa_exec() function are the same as for |
| 2845 | pcre_exec(), plus two extras. The ovector argument is used in a differ- |
| 2846 | ent way, and this is described below. The other common arguments are |
| 2847 | used in the same way as for pcre_exec(), so their description is not |
| 2848 | repeated here. |
| 2849 | |
| 2850 | The two additional arguments provide workspace for the function. The |
| 2851 | workspace vector should contain at least 20 elements. It is used for |
| 2852 | keeping track of multiple paths through the pattern tree. More |
| 2853 | workspace will be needed for patterns and subjects where there are a |
| 2854 | lot of potential matches. |
| 2855 | |
| 2856 | Here is an example of a simple call to pcre_dfa_exec(): |
| 2857 | |
| 2858 | int rc; |
| 2859 | int ovector[10]; |
| 2860 | int wspace[20]; |
| 2861 | rc = pcre_dfa_exec( |
| 2862 | re, /* result of pcre_compile() */ |
| 2863 | NULL, /* we didn't study the pattern */ |
| 2864 | "some string", /* the subject string */ |
| 2865 | 11, /* the length of the subject string */ |
| 2866 | 0, /* start at offset 0 in the subject */ |
| 2867 | 0, /* default options */ |
| 2868 | ovector, /* vector of integers for substring information */ |
| 2869 | 10, /* number of elements (NOT size in bytes) */ |
| 2870 | wspace, /* working space vector */ |
| 2871 | 20); /* number of elements (NOT size in bytes) */ |
| 2872 | |
| 2873 | Option bits for pcre_dfa_exec() |
| 2874 | |
| 2875 | The unused bits of the options argument for pcre_dfa_exec() must be |
| 2876 | zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NEW- |
| 2877 | LINE_xxx, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, |
| 2878 | PCRE_NOTEMPTY_ATSTART, PCRE_NO_UTF8_CHECK, PCRE_BSR_ANYCRLF, |
| 2879 | PCRE_BSR_UNICODE, PCRE_NO_START_OPTIMIZE, PCRE_PARTIAL_HARD, PCRE_PAR- |
| 2880 | TIAL_SOFT, PCRE_DFA_SHORTEST, and PCRE_DFA_RESTART. All but the last |
| 2881 | four of these are exactly the same as for pcre_exec(), so their |
| 2882 | description is not repeated here. |
| 2883 | |
| 2884 | PCRE_PARTIAL_HARD |
| 2885 | PCRE_PARTIAL_SOFT |
| 2886 | |
| 2887 | These have the same general effect as they do for pcre_exec(), but the |
| 2888 | details are slightly different. When PCRE_PARTIAL_HARD is set for |
| 2889 | pcre_dfa_exec(), it returns PCRE_ERROR_PARTIAL if the end of the sub- |
| 2890 | ject is reached and there is still at least one matching possibility |
| 2891 | that requires additional characters. This happens even if some complete |
| 2892 | matches have also been found. When PCRE_PARTIAL_SOFT is set, the return |
| 2893 | code PCRE_ERROR_NOMATCH is converted into PCRE_ERROR_PARTIAL if the end |
| 2894 | of the subject is reached, there have been no complete matches, but |
| 2895 | there is still at least one matching possibility. The portion of the |
| 2896 | string that was inspected when the longest partial match was found is |
| 2897 | set as the first matching string in both cases. There is a more |
| 2898 | detailed discussion of partial and multi-segment matching, with exam- |
| 2899 | ples, in the pcrepartial documentation. |
| 2900 | |
| 2901 | PCRE_DFA_SHORTEST |
| 2902 | |
| 2903 | Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to |
| 2904 | stop as soon as it has found one match. Because of the way the alterna- |
| 2905 | tive algorithm works, this is necessarily the shortest possible match |
| 2906 | at the first possible matching point in the subject string. |
| 2907 | |
| 2908 | PCRE_DFA_RESTART |
| 2909 | |
| 2910 | When pcre_dfa_exec() returns a partial match, it is possible to call it |
| 2911 | again, with additional subject characters, and have it continue with |
| 2912 | the same match. The PCRE_DFA_RESTART option requests this action; when |
| 2913 | it is set, the workspace and wscount options must reference the same |
| 2914 | vector as before because data about the match so far is left in them |
| 2915 | after a partial match. There is more discussion of this facility in the |
| 2916 | pcrepartial documentation. |
| 2917 | |
| 2918 | Successful returns from pcre_dfa_exec() |
| 2919 | |
| 2920 | When pcre_dfa_exec() succeeds, it may have matched more than one sub- |
| 2921 | string in the subject. Note, however, that all the matches from one run |
| 2922 | of the function start at the same point in the subject. The shorter |
| 2923 | matches are all initial substrings of the longer matches. For example, |
| 2924 | if the pattern |
| 2925 | |
| 2926 | <.*> |
| 2927 | |
| 2928 | is matched against the string |
| 2929 | |
| 2930 | This is <something> <something else> <something further> no more |
| 2931 | |
| 2932 | the three matched strings are |
| 2933 | |
| 2934 | <something> |
| 2935 | <something> <something else> |
| 2936 | <something> <something else> <something further> |
| 2937 | |
| 2938 | On success, the yield of the function is a number greater than zero, |
| 2939 | which is the number of matched substrings. The substrings themselves |
| 2940 | are returned in ovector. Each string uses two elements; the first is |
| 2941 | the offset to the start, and the second is the offset to the end. In |
| 2942 | fact, all the strings have the same start offset. (Space could have |
| 2943 | been saved by giving this only once, but it was decided to retain some |
| 2944 | compatibility with the way pcre_exec() returns data, even though the |
| 2945 | meaning of the strings is different.) |
| 2946 | |
| 2947 | The strings are returned in reverse order of length; that is, the long- |
| 2948 | est matching string is given first. If there were too many matches to |
| 2949 | fit into ovector, the yield of the function is zero, and the vector is |
| 2950 | filled with the longest matches. Unlike pcre_exec(), pcre_dfa_exec() |
| 2951 | can use the entire ovector for returning matched strings. |
| 2952 | |
| 2953 | Error returns from pcre_dfa_exec() |
| 2954 | |
| 2955 | The pcre_dfa_exec() function returns a negative number when it fails. |
| 2956 | Many of the errors are the same as for pcre_exec(), and these are |
| 2957 | described above. There are in addition the following errors that are |
| 2958 | specific to pcre_dfa_exec(): |
| 2959 | |
| 2960 | PCRE_ERROR_DFA_UITEM (-16) |
| 2961 | |
| 2962 | This return is given if pcre_dfa_exec() encounters an item in the pat- |
| 2963 | tern that it does not support, for instance, the use of \C or a back |
| 2964 | reference. |
| 2965 | |
| 2966 | PCRE_ERROR_DFA_UCOND (-17) |
| 2967 | |
| 2968 | This return is given if pcre_dfa_exec() encounters a condition item |
| 2969 | that uses a back reference for the condition, or a test for recursion |
| 2970 | in a specific group. These are not supported. |
| 2971 | |
| 2972 | PCRE_ERROR_DFA_UMLIMIT (-18) |
| 2973 | |
| 2974 | This return is given if pcre_dfa_exec() is called with an extra block |
| 2975 | that contains a setting of the match_limit or match_limit_recursion |
| 2976 | fields. This is not supported (these fields are meaningless for DFA |
| 2977 | matching). |
| 2978 | |
| 2979 | PCRE_ERROR_DFA_WSSIZE (-19) |
| 2980 | |
| 2981 | This return is given if pcre_dfa_exec() runs out of space in the |
| 2982 | workspace vector. |
| 2983 | |
| 2984 | PCRE_ERROR_DFA_RECURSE (-20) |
| 2985 | |
| 2986 | When a recursive subpattern is processed, the matching function calls |
| 2987 | itself recursively, using private vectors for ovector and workspace. |
| 2988 | This error is given if the output vector is not large enough. This |
| 2989 | should be extremely rare, as a vector of size 1000 is used. |
| 2990 | |
| 2991 | |
| 2992 | SEE ALSO |
| 2993 | |
| 2994 | pcrebuild(3), pcrecallout(3), pcrecpp(3)(3), pcrematching(3), pcrepar- |
| 2995 | tial(3), pcreposix(3), pcreprecompile(3), pcresample(3), pcrestack(3). |
| 2996 | |
| 2997 | |
| 2998 | AUTHOR |
| 2999 | |
| 3000 | Philip Hazel |
| 3001 | University Computing Service |
| 3002 | Cambridge CB2 3QH, England. |
| 3003 | |
| 3004 | |
| 3005 | REVISION |
| 3006 | |
| 3007 | Last updated: 02 December 2011 |
| 3008 | Copyright (c) 1997-2011 University of Cambridge. |
| 3009 | ------------------------------------------------------------------------------ |
| 3010 | |
| 3011 | |
| 3012 | PCRECALLOUT(3) PCRECALLOUT(3) |
| 3013 | |
| 3014 | |
| 3015 | NAME |
| 3016 | PCRE - Perl-compatible regular expressions |
| 3017 | |
| 3018 | |
| 3019 | PCRE CALLOUTS |
| 3020 | |
| 3021 | int (*pcre_callout)(pcre_callout_block *); |
| 3022 | |
| 3023 | PCRE provides a feature called "callout", which is a means of temporar- |
| 3024 | ily passing control to the caller of PCRE in the middle of pattern |
| 3025 | matching. The caller of PCRE provides an external function by putting |
| 3026 | its entry point in the global variable pcre_callout. By default, this |
| 3027 | variable contains NULL, which disables all calling out. |
| 3028 | |
| 3029 | Within a regular expression, (?C) indicates the points at which the |
| 3030 | external function is to be called. Different callout points can be |
| 3031 | identified by putting a number less than 256 after the letter C. The |
| 3032 | default value is zero. For example, this pattern has two callout |
| 3033 | points: |
| 3034 | |
| 3035 | (?C1)abc(?C2)def |
| 3036 | |
| 3037 | If the PCRE_AUTO_CALLOUT option bit is set when pcre_compile() or |
| 3038 | pcre_compile2() is called, PCRE automatically inserts callouts, all |
| 3039 | with number 255, before each item in the pattern. For example, if |
| 3040 | PCRE_AUTO_CALLOUT is used with the pattern |
| 3041 | |
| 3042 | A(\d{2}|--) |
| 3043 | |
| 3044 | it is processed as if it were |
| 3045 | |
| 3046 | (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255) |
| 3047 | |
| 3048 | Notice that there is a callout before and after each parenthesis and |
| 3049 | alternation bar. Automatic callouts can be used for tracking the |
| 3050 | progress of pattern matching. The pcretest command has an option that |
| 3051 | sets automatic callouts; when it is used, the output indicates how the |
| 3052 | pattern is matched. This is useful information when you are trying to |
| 3053 | optimize the performance of a particular pattern. |
| 3054 | |
| 3055 | The use of callouts in a pattern makes it ineligible for optimization |
| 3056 | by the just-in-time compiler. Studying such a pattern with the |
| 3057 | PCRE_STUDY_JIT_COMPILE option always fails. |
| 3058 | |
| 3059 | |
| 3060 | MISSING CALLOUTS |
| 3061 | |
| 3062 | You should be aware that, because of optimizations in the way PCRE |
| 3063 | matches patterns by default, callouts sometimes do not happen. For |
| 3064 | example, if the pattern is |
| 3065 | |
| 3066 | ab(?C4)cd |
| 3067 | |
| 3068 | PCRE knows that any matching string must contain the letter "d". If the |
| 3069 | subject string is "abyz", the lack of "d" means that matching doesn't |
| 3070 | ever start, and the callout is never reached. However, with "abyd", |
| 3071 | though the result is still no match, the callout is obeyed. |
| 3072 | |
| 3073 | If the pattern is studied, PCRE knows the minimum length of a matching |
| 3074 | string, and will immediately give a "no match" return without actually |
| 3075 | running a match if the subject is not long enough, or, for unanchored |
| 3076 | patterns, if it has been scanned far enough. |
| 3077 | |
| 3078 | You can disable these optimizations by passing the PCRE_NO_START_OPTI- |
| 3079 | MIZE option to pcre_compile(), pcre_exec(), or pcre_dfa_exec(), or by |
| 3080 | starting the pattern with (*NO_START_OPT). This slows down the matching |
| 3081 | process, but does ensure that callouts such as the example above are |
| 3082 | obeyed. |
| 3083 | |
| 3084 | |
| 3085 | THE CALLOUT INTERFACE |
| 3086 | |
| 3087 | During matching, when PCRE reaches a callout point, the external func- |
| 3088 | tion defined by pcre_callout is called (if it is set). This applies to |
| 3089 | both the pcre_exec() and the pcre_dfa_exec() matching functions. The |
| 3090 | only argument to the callout function is a pointer to a pcre_callout |
| 3091 | block. This structure contains the following fields: |
| 3092 | |
| 3093 | int version; |
| 3094 | int callout_number; |
| 3095 | int *offset_vector; |
| 3096 | const char *subject; |
| 3097 | int subject_length; |
| 3098 | int start_match; |
| 3099 | int current_position; |
| 3100 | int capture_top; |
| 3101 | int capture_last; |
| 3102 | void *callout_data; |
| 3103 | int pattern_position; |
| 3104 | int next_item_length; |
| 3105 | const unsigned char *mark; |
| 3106 | |
| 3107 | The version field is an integer containing the version number of the |
| 3108 | block format. The initial version was 0; the current version is 2. The |
| 3109 | version number will change again in future if additional fields are |
| 3110 | added, but the intention is never to remove any of the existing fields. |
| 3111 | |
| 3112 | The callout_number field contains the number of the callout, as com- |
| 3113 | piled into the pattern (that is, the number after ?C for manual call- |
| 3114 | outs, and 255 for automatically generated callouts). |
| 3115 | |
| 3116 | The offset_vector field is a pointer to the vector of offsets that was |
| 3117 | passed by the caller to pcre_exec() or pcre_dfa_exec(). When |
| 3118 | pcre_exec() is used, the contents can be inspected in order to extract |
| 3119 | substrings that have been matched so far, in the same way as for |
| 3120 | extracting substrings after a match has completed. For pcre_dfa_exec() |
| 3121 | this field is not useful. |
| 3122 | |
| 3123 | The subject and subject_length fields contain copies of the values that |
| 3124 | were passed to pcre_exec(). |
| 3125 | |
| 3126 | The start_match field normally contains the offset within the subject |
| 3127 | at which the current match attempt started. However, if the escape |
| 3128 | sequence \K has been encountered, this value is changed to reflect the |
| 3129 | modified starting point. If the pattern is not anchored, the callout |
| 3130 | function may be called several times from the same point in the pattern |
| 3131 | for different starting points in the subject. |
| 3132 | |
| 3133 | The current_position field contains the offset within the subject of |
| 3134 | the current match pointer. |
| 3135 | |
| 3136 | When the pcre_exec() function is used, the capture_top field contains |
| 3137 | one more than the number of the highest numbered captured substring so |
| 3138 | far. If no substrings have been captured, the value of capture_top is |
| 3139 | one. This is always the case when pcre_dfa_exec() is used, because it |
| 3140 | does not support captured substrings. |
| 3141 | |
| 3142 | The capture_last field contains the number of the most recently cap- |
| 3143 | tured substring. If no substrings have been captured, its value is -1. |
| 3144 | This is always the case when pcre_dfa_exec() is used. |
| 3145 | |
| 3146 | The callout_data field contains a value that is passed to pcre_exec() |
| 3147 | or pcre_dfa_exec() specifically so that it can be passed back in call- |
| 3148 | outs. It is passed in the pcre_callout field of the pcre_extra data |
| 3149 | structure. If no such data was passed, the value of callout_data in a |
| 3150 | pcre_callout block is NULL. There is a description of the pcre_extra |
| 3151 | structure in the pcreapi documentation. |
| 3152 | |
| 3153 | The pattern_position field is present from version 1 of the pcre_call- |
| 3154 | out structure. It contains the offset to the next item to be matched in |
| 3155 | the pattern string. |
| 3156 | |
| 3157 | The next_item_length field is present from version 1 of the pcre_call- |
| 3158 | out structure. It contains the length of the next item to be matched in |
| 3159 | the pattern string. When the callout immediately precedes an alterna- |
| 3160 | tion bar, a closing parenthesis, or the end of the pattern, the length |
| 3161 | is zero. When the callout precedes an opening parenthesis, the length |
| 3162 | is that of the entire subpattern. |
| 3163 | |
| 3164 | The pattern_position and next_item_length fields are intended to help |
| 3165 | in distinguishing between different automatic callouts, which all have |
| 3166 | the same callout number. However, they are set for all callouts. |
| 3167 | |
| 3168 | The mark field is present from version 2 of the pcre_callout structure. |
| 3169 | In callouts from pcre_exec() it contains a pointer to the zero-termi- |
| 3170 | nated name of the most recently passed (*MARK), (*PRUNE), or (*THEN) |
| 3171 | item in the match, or NULL if no such items have been passed. Instances |
| 3172 | of (*PRUNE) or (*THEN) without a name do not obliterate a previous |
| 3173 | (*MARK). In callouts from pcre_dfa_exec() this field always contains |
| 3174 | NULL. |
| 3175 | |
| 3176 | |
| 3177 | RETURN VALUES |
| 3178 | |
| 3179 | The external callout function returns an integer to PCRE. If the value |
| 3180 | is zero, matching proceeds as normal. If the value is greater than |
| 3181 | zero, matching fails at the current point, but the testing of other |
| 3182 | matching possibilities goes ahead, just as if a lookahead assertion had |
| 3183 | failed. If the value is less than zero, the match is abandoned, and |
| 3184 | pcre_exec() or pcre_dfa_exec() returns the negative value. |
| 3185 | |
| 3186 | Negative values should normally be chosen from the set of |
| 3187 | PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan- |
| 3188 | dard "no match" failure. The error number PCRE_ERROR_CALLOUT is |
| 3189 | reserved for use by callout functions; it will never be used by PCRE |
| 3190 | itself. |
| 3191 | |
| 3192 | |
| 3193 | AUTHOR |
| 3194 | |
| 3195 | Philip Hazel |
| 3196 | University Computing Service |
| 3197 | Cambridge CB2 3QH, England. |
| 3198 | |
| 3199 | |
| 3200 | REVISION |
| 3201 | |
| 3202 | Last updated: 30 November 2011 |
| 3203 | Copyright (c) 1997-2011 University of Cambridge. |
| 3204 | ------------------------------------------------------------------------------ |
| 3205 | |
| 3206 | |
| 3207 | PCRECOMPAT(3) PCRECOMPAT(3) |
| 3208 | |
| 3209 | |
| 3210 | NAME |
| 3211 | PCRE - Perl-compatible regular expressions |
| 3212 | |
| 3213 | |
| 3214 | DIFFERENCES BETWEEN PCRE AND PERL |
| 3215 | |
| 3216 | This document describes the differences in the ways that PCRE and Perl |
| 3217 | handle regular expressions. The differences described here are with |
| 3218 | respect to Perl versions 5.10 and above. |
| 3219 | |
| 3220 | 1. PCRE has only a subset of Perl's UTF-8 and Unicode support. Details |
| 3221 | of what it does have are given in the pcreunicode page. |
| 3222 | |
| 3223 | 2. PCRE allows repeat quantifiers only on parenthesized assertions, but |
| 3224 | they do not mean what you might think. For example, (?!a){3} does not |
| 3225 | assert that the next three characters are not "a". It just asserts that |
| 3226 | the next character is not "a" three times (in principle: PCRE optimizes |
| 3227 | this to run the assertion just once). Perl allows repeat quantifiers on |
| 3228 | other assertions such as \b, but these do not seem to have any use. |
| 3229 | |
| 3230 | 3. Capturing subpatterns that occur inside negative lookahead asser- |
| 3231 | tions are counted, but their entries in the offsets vector are never |
| 3232 | set. Perl sets its numerical variables from any such patterns that are |
| 3233 | matched before the assertion fails to match something (thereby succeed- |
| 3234 | ing), but only if the negative lookahead assertion contains just one |
| 3235 | branch. |
| 3236 | |
| 3237 | 4. Though binary zero characters are supported in the subject string, |
| 3238 | they are not allowed in a pattern string because it is passed as a nor- |
| 3239 | mal C string, terminated by zero. The escape sequence \0 can be used in |
| 3240 | the pattern to represent a binary zero. |
| 3241 | |
| 3242 | 5. The following Perl escape sequences are not supported: \l, \u, \L, |
| 3243 | \U, and \N when followed by a character name or Unicode value. (\N on |
| 3244 | its own, matching a non-newline character, is supported.) In fact these |
| 3245 | are implemented by Perl's general string-handling and are not part of |
| 3246 | its pattern matching engine. If any of these are encountered by PCRE, |
| 3247 | an error is generated by default. However, if the PCRE_JAVASCRIPT_COM- |
| 3248 | PAT option is set, \U and \u are interpreted as JavaScript interprets |
| 3249 | them. |
| 3250 | |
| 3251 | 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE |
| 3252 | is built with Unicode character property support. The properties that |
| 3253 | can be tested with \p and \P are limited to the general category prop- |
| 3254 | erties such as Lu and Nd, script names such as Greek or Han, and the |
| 3255 | derived properties Any and L&. PCRE does support the Cs (surrogate) |
| 3256 | property, which Perl does not; the Perl documentation says "Because |
| 3257 | Perl hides the need for the user to understand the internal representa- |
| 3258 | tion of Unicode characters, there is no need to implement the somewhat |
| 3259 | messy concept of surrogates." |
| 3260 | |
| 3261 | 7. PCRE implements a simpler version of \X than Perl, which changed to |
| 3262 | make \X match what Unicode calls an "extended grapheme cluster". This |
| 3263 | is more complicated than an extended Unicode sequence, which is what |
| 3264 | PCRE matches. |
| 3265 | |
| 3266 | 8. PCRE does support the \Q...\E escape for quoting substrings. Charac- |
| 3267 | ters in between are treated as literals. This is slightly different |
| 3268 | from Perl in that $ and @ are also handled as literals inside the |
| 3269 | quotes. In Perl, they cause variable interpolation (but of course PCRE |
| 3270 | does not have variables). Note the following examples: |
| 3271 | |
| 3272 | Pattern PCRE matches Perl matches |
| 3273 | |
| 3274 | \Qabc$xyz\E abc$xyz abc followed by the |
| 3275 | contents of $xyz |
| 3276 | \Qabc\$xyz\E abc\$xyz abc\$xyz |
| 3277 | \Qabc\E\$\Qxyz\E abc$xyz abc$xyz |
| 3278 | |
| 3279 | The \Q...\E sequence is recognized both inside and outside character |
| 3280 | classes. |
| 3281 | |
| 3282 | 9. Fairly obviously, PCRE does not support the (?{code}) and (??{code}) |
| 3283 | constructions. However, there is support for recursive patterns. This |
| 3284 | is not available in Perl 5.8, but it is in Perl 5.10. Also, the PCRE |
| 3285 | "callout" feature allows an external function to be called during pat- |
| 3286 | tern matching. See the pcrecallout documentation for details. |
| 3287 | |
| 3288 | 10. Subpatterns that are called as subroutines (whether or not recur- |
| 3289 | sively) are always treated as atomic groups in PCRE. This is like |
| 3290 | Python, but unlike Perl. Captured values that are set outside a sub- |
| 3291 | routine call can be reference from inside in PCRE, but not in Perl. |
| 3292 | There is a discussion that explains these differences in more detail in |
| 3293 | the section on recursion differences from Perl in the pcrepattern page. |
| 3294 | |
| 3295 | 11. If (*THEN) is present in a group that is called as a subroutine, |
| 3296 | its action is limited to that group, even if the group does not contain |
| 3297 | any | characters. |
| 3298 | |
| 3299 | 12. There are some differences that are concerned with the settings of |
| 3300 | captured strings when part of a pattern is repeated. For example, |
| 3301 | matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2 |
| 3302 | unset, but in PCRE it is set to "b". |
| 3303 | |
| 3304 | 13. PCRE's handling of duplicate subpattern numbers and duplicate sub- |
| 3305 | pattern names is not as general as Perl's. This is a consequence of the |
| 3306 | fact the PCRE works internally just with numbers, using an external ta- |
| 3307 | ble to translate between numbers and names. In particular, a pattern |
| 3308 | such as (?|(?<a>A)|(?<b)B), where the two capturing parentheses have |
| 3309 | the same number but different names, is not supported, and causes an |
| 3310 | error at compile time. If it were allowed, it would not be possible to |
| 3311 | distinguish which parentheses matched, because both names map to cap- |
| 3312 | turing subpattern number 1. To avoid this confusing situation, an error |
| 3313 | is given at compile time. |
| 3314 | |
| 3315 | 14. Perl recognizes comments in some places that PCRE does not, for |
| 3316 | example, between the ( and ? at the start of a subpattern. If the /x |
| 3317 | modifier is set, Perl allows whitespace between ( and ? but PCRE never |
| 3318 | does, even if the PCRE_EXTENDED option is set. |
| 3319 | |
| 3320 | 15. PCRE provides some extensions to the Perl regular expression facil- |
| 3321 | ities. Perl 5.10 includes new features that are not in earlier ver- |
| 3322 | sions of Perl, some of which (such as named parentheses) have been in |
| 3323 | PCRE for some time. This list is with respect to Perl 5.10: |
| 3324 | |
| 3325 | (a) Although lookbehind assertions in PCRE must match fixed length |
| 3326 | strings, each alternative branch of a lookbehind assertion can match a |
| 3327 | different length of string. Perl requires them all to have the same |
| 3328 | length. |
| 3329 | |
| 3330 | (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $ |
| 3331 | meta-character matches only at the very end of the string. |
| 3332 | |
| 3333 | (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe- |
| 3334 | cial meaning is faulted. Otherwise, like Perl, the backslash is quietly |
| 3335 | ignored. (Perl can be made to issue a warning.) |
| 3336 | |
| 3337 | (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti- |
| 3338 | fiers is inverted, that is, by default they are not greedy, but if fol- |
| 3339 | lowed by a question mark they are. |
| 3340 | |
| 3341 | (e) PCRE_ANCHORED can be used at matching time to force a pattern to be |
| 3342 | tried only at the first matching position in the subject string. |
| 3343 | |
| 3344 | (f) The PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, |
| 3345 | and PCRE_NO_AUTO_CAPTURE options for pcre_exec() have no Perl equiva- |
| 3346 | lents. |
| 3347 | |
| 3348 | (g) The \R escape sequence can be restricted to match only CR, LF, or |
| 3349 | CRLF by the PCRE_BSR_ANYCRLF option. |
| 3350 | |
| 3351 | (h) The callout facility is PCRE-specific. |
| 3352 | |
| 3353 | (i) The partial matching facility is PCRE-specific. |
| 3354 | |
| 3355 | (j) Patterns compiled by PCRE can be saved and re-used at a later time, |
| 3356 | even on different hosts that have the other endianness. However, this |
| 3357 | does not apply to optimized data created by the just-in-time compiler. |
| 3358 | |
| 3359 | (k) The alternative matching function (pcre_dfa_exec()) matches in a |
| 3360 | different way and is not Perl-compatible. |
| 3361 | |
| 3362 | (l) PCRE recognizes some special sequences such as (*CR) at the start |
| 3363 | of a pattern that set overall options that cannot be changed within the |
| 3364 | pattern. |
| 3365 | |
| 3366 | |
| 3367 | AUTHOR |
| 3368 | |
| 3369 | Philip Hazel |
| 3370 | University Computing Service |
| 3371 | Cambridge CB2 3QH, England. |
| 3372 | |
| 3373 | |
| 3374 | REVISION |
| 3375 | |
| 3376 | Last updated: 14 November 2011 |
| 3377 | Copyright (c) 1997-2011 University of Cambridge. |
| 3378 | ------------------------------------------------------------------------------ |
| 3379 | |
| 3380 | |
| 3381 | PCREPATTERN(3) PCREPATTERN(3) |
| 3382 | |
| 3383 | |
| 3384 | NAME |
| 3385 | PCRE - Perl-compatible regular expressions |
| 3386 | |
| 3387 | |
| 3388 | PCRE REGULAR EXPRESSION DETAILS |
| 3389 | |
| 3390 | The syntax and semantics of the regular expressions that are supported |
| 3391 | by PCRE are described in detail below. There is a quick-reference syn- |
| 3392 | tax summary in the pcresyntax page. PCRE tries to match Perl syntax and |
| 3393 | semantics as closely as it can. PCRE also supports some alternative |
| 3394 | regular expression syntax (which does not conflict with the Perl syn- |
| 3395 | tax) in order to provide some compatibility with regular expressions in |
| 3396 | Python, .NET, and Oniguruma. |
| 3397 | |
| 3398 | Perl's regular expressions are described in its own documentation, and |
| 3399 | regular expressions in general are covered in a number of books, some |
| 3400 | of which have copious examples. Jeffrey Friedl's "Mastering Regular |
| 3401 | Expressions", published by O'Reilly, covers regular expressions in |
| 3402 | great detail. This description of PCRE's regular expressions is |
| 3403 | intended as reference material. |
| 3404 | |
| 3405 | The original operation of PCRE was on strings of one-byte characters. |
| 3406 | However, there is now also support for UTF-8 character strings. To use |
| 3407 | this, PCRE must be built to include UTF-8 support, and you must call |
| 3408 | pcre_compile() or pcre_compile2() with the PCRE_UTF8 option. There is |
| 3409 | also a special sequence that can be given at the start of a pattern: |
| 3410 | |
| 3411 | (*UTF8) |
| 3412 | |
| 3413 | Starting a pattern with this sequence is equivalent to setting the |
| 3414 | PCRE_UTF8 option. This feature is not Perl-compatible. How setting |
| 3415 | UTF-8 mode affects pattern matching is mentioned in several places |
| 3416 | below. There is also a summary of UTF-8 features in the pcreunicode |
| 3417 | page. |
| 3418 | |
| 3419 | Another special sequence that may appear at the start of a pattern or |
| 3420 | in combination with (*UTF8) is: |
| 3421 | |
| 3422 | (*UCP) |
| 3423 | |
| 3424 | This has the same effect as setting the PCRE_UCP option: it causes |
| 3425 | sequences such as \d and \w to use Unicode properties to determine |
| 3426 | character types, instead of recognizing only characters with codes less |
| 3427 | than 128 via a lookup table. |
| 3428 | |
| 3429 | If a pattern starts with (*NO_START_OPT), it has the same effect as |
| 3430 | setting the PCRE_NO_START_OPTIMIZE option either at compile or matching |
| 3431 | time. There are also some more of these special sequences that are con- |
| 3432 | cerned with the handling of newlines; they are described below. |
| 3433 | |
| 3434 | The remainder of this document discusses the patterns that are sup- |
| 3435 | ported by PCRE when its main matching function, pcre_exec(), is used. |
| 3436 | From release 6.0, PCRE offers a second matching function, |
| 3437 | pcre_dfa_exec(), which matches using a different algorithm that is not |
| 3438 | Perl-compatible. Some of the features discussed below are not available |
| 3439 | when pcre_dfa_exec() is used. The advantages and disadvantages of the |
| 3440 | alternative function, and how it differs from the normal function, are |
| 3441 | discussed in the pcrematching page. |
| 3442 | |
| 3443 | |
| 3444 | NEWLINE CONVENTIONS |
| 3445 | |
| 3446 | PCRE supports five different conventions for indicating line breaks in |
| 3447 | strings: a single CR (carriage return) character, a single LF (line- |
| 3448 | feed) character, the two-character sequence CRLF, any of the three pre- |
| 3449 | ceding, or any Unicode newline sequence. The pcreapi page has further |
| 3450 | discussion about newlines, and shows how to set the newline convention |
| 3451 | in the options arguments for the compiling and matching functions. |
| 3452 | |
| 3453 | It is also possible to specify a newline convention by starting a pat- |
| 3454 | tern string with one of the following five sequences: |
| 3455 | |
| 3456 | (*CR) carriage return |
| 3457 | (*LF) linefeed |
| 3458 | (*CRLF) carriage return, followed by linefeed |
| 3459 | (*ANYCRLF) any of the three above |
| 3460 | (*ANY) all Unicode newline sequences |
| 3461 | |
| 3462 | These override the default and the options given to pcre_compile() or |
| 3463 | pcre_compile2(). For example, on a Unix system where LF is the default |
| 3464 | newline sequence, the pattern |
| 3465 | |
| 3466 | (*CR)a.b |
| 3467 | |
| 3468 | changes the convention to CR. That pattern matches "a\nb" because LF is |
| 3469 | no longer a newline. Note that these special settings, which are not |
| 3470 | Perl-compatible, are recognized only at the very start of a pattern, |
| 3471 | and that they must be in upper case. If more than one of them is |
| 3472 | present, the last one is used. |
| 3473 | |
| 3474 | The newline convention affects the interpretation of the dot metachar- |
| 3475 | acter when PCRE_DOTALL is not set, and also the behaviour of \N. How- |
| 3476 | ever, it does not affect what the \R escape sequence matches. By |
| 3477 | default, this is any Unicode newline sequence, for Perl compatibility. |
| 3478 | However, this can be changed; see the description of \R in the section |
| 3479 | entitled "Newline sequences" below. A change of \R setting can be com- |
| 3480 | bined with a change of newline convention. |
| 3481 | |
| 3482 | |
| 3483 | CHARACTERS AND METACHARACTERS |
| 3484 | |
| 3485 | A regular expression is a pattern that is matched against a subject |
| 3486 | string from left to right. Most characters stand for themselves in a |
| 3487 | pattern, and match the corresponding characters in the subject. As a |
| 3488 | trivial example, the pattern |
| 3489 | |
| 3490 | The quick brown fox |
| 3491 | |
| 3492 | matches a portion of a subject string that is identical to itself. When |
| 3493 | caseless matching is specified (the PCRE_CASELESS option), letters are |
| 3494 | matched independently of case. In UTF-8 mode, PCRE always understands |
| 3495 | the concept of case for characters whose values are less than 128, so |
| 3496 | caseless matching is always possible. For characters with higher val- |
| 3497 | ues, the concept of case is supported if PCRE is compiled with Unicode |
| 3498 | property support, but not otherwise. If you want to use caseless |
| 3499 | matching for characters 128 and above, you must ensure that PCRE is |
| 3500 | compiled with Unicode property support as well as with UTF-8 support. |
| 3501 | |
| 3502 | The power of regular expressions comes from the ability to include |
| 3503 | alternatives and repetitions in the pattern. These are encoded in the |
| 3504 | pattern by the use of metacharacters, which do not stand for themselves |
| 3505 | but instead are interpreted in some special way. |
| 3506 | |
| 3507 | There are two different sets of metacharacters: those that are recog- |
| 3508 | nized anywhere in the pattern except within square brackets, and those |
| 3509 | that are recognized within square brackets. Outside square brackets, |
| 3510 | the metacharacters are as follows: |
| 3511 | |
| 3512 | \ general escape character with several uses |
| 3513 | ^ assert start of string (or line, in multiline mode) |
| 3514 | $ assert end of string (or line, in multiline mode) |
| 3515 | . match any character except newline (by default) |
| 3516 | [ start character class definition |
| 3517 | | start of alternative branch |
| 3518 | ( start subpattern |
| 3519 | ) end subpattern |
| 3520 | ? extends the meaning of ( |
| 3521 | also 0 or 1 quantifier |
| 3522 | also quantifier minimizer |
| 3523 | * 0 or more quantifier |
| 3524 | + 1 or more quantifier |
| 3525 | also "possessive quantifier" |
| 3526 | { start min/max quantifier |
| 3527 | |
| 3528 | Part of a pattern that is in square brackets is called a "character |
| 3529 | class". In a character class the only metacharacters are: |
| 3530 | |
| 3531 | \ general escape character |
| 3532 | ^ negate the class, but only if the first character |
| 3533 | - indicates character range |
| 3534 | [ POSIX character class (only if followed by POSIX |
| 3535 | syntax) |
| 3536 | ] terminates the character class |
| 3537 | |
| 3538 | The following sections describe the use of each of the metacharacters. |
| 3539 | |
| 3540 | |
| 3541 | BACKSLASH |
| 3542 | |
| 3543 | The backslash character has several uses. Firstly, if it is followed by |
| 3544 | a character that is not a number or a letter, it takes away any special |
| 3545 | meaning that character may have. This use of backslash as an escape |
| 3546 | character applies both inside and outside character classes. |
| 3547 | |
| 3548 | For example, if you want to match a * character, you write \* in the |
| 3549 | pattern. This escaping action applies whether or not the following |
| 3550 | character would otherwise be interpreted as a metacharacter, so it is |
| 3551 | always safe to precede a non-alphanumeric with backslash to specify |
| 3552 | that it stands for itself. In particular, if you want to match a back- |
| 3553 | slash, you write \\. |
| 3554 | |
| 3555 | In UTF-8 mode, only ASCII numbers and letters have any special meaning |
| 3556 | after a backslash. All other characters (in particular, those whose |
| 3557 | codepoints are greater than 127) are treated as literals. |
| 3558 | |
| 3559 | If a pattern is compiled with the PCRE_EXTENDED option, whitespace in |
| 3560 | the pattern (other than in a character class) and characters between a |
| 3561 | # outside a character class and the next newline are ignored. An escap- |
| 3562 | ing backslash can be used to include a whitespace or # character as |
| 3563 | part of the pattern. |
| 3564 | |
| 3565 | If you want to remove the special meaning from a sequence of charac- |
| 3566 | ters, you can do so by putting them between \Q and \E. This is differ- |
| 3567 | ent from Perl in that $ and @ are handled as literals in \Q...\E |
| 3568 | sequences in PCRE, whereas in Perl, $ and @ cause variable interpola- |
| 3569 | tion. Note the following examples: |
| 3570 | |
| 3571 | Pattern PCRE matches Perl matches |
| 3572 | |
| 3573 | \Qabc$xyz\E abc$xyz abc followed by the |
| 3574 | contents of $xyz |
| 3575 | \Qabc\$xyz\E abc\$xyz abc\$xyz |
| 3576 | \Qabc\E\$\Qxyz\E abc$xyz abc$xyz |
| 3577 | |
| 3578 | The \Q...\E sequence is recognized both inside and outside character |
| 3579 | classes. An isolated \E that is not preceded by \Q is ignored. If \Q |
| 3580 | is not followed by \E later in the pattern, the literal interpretation |
| 3581 | continues to the end of the pattern (that is, \E is assumed at the |
| 3582 | end). If the isolated \Q is inside a character class, this causes an |
| 3583 | error, because the character class is not terminated. |
| 3584 | |
| 3585 | Non-printing characters |
| 3586 | |
| 3587 | A second use of backslash provides a way of encoding non-printing char- |
| 3588 | acters in patterns in a visible manner. There is no restriction on the |
| 3589 | appearance of non-printing characters, apart from the binary zero that |
| 3590 | terminates a pattern, but when a pattern is being prepared by text |
| 3591 | editing, it is often easier to use one of the following escape |
| 3592 | sequences than the binary character it represents: |
| 3593 | |
| 3594 | \a alarm, that is, the BEL character (hex 07) |
| 3595 | \cx "control-x", where x is any ASCII character |
| 3596 | \e escape (hex 1B) |
| 3597 | \f formfeed (hex 0C) |
| 3598 | \n linefeed (hex 0A) |
| 3599 | \r carriage return (hex 0D) |
| 3600 | \t tab (hex 09) |
| 3601 | \ddd character with octal code ddd, or back reference |
| 3602 | \xhh character with hex code hh |
| 3603 | \x{hhh..} character with hex code hhh.. (non-JavaScript mode) |
| 3604 | \uhhhh character with hex code hhhh (JavaScript mode only) |
| 3605 | |
| 3606 | The precise effect of \cx is as follows: if x is a lower case letter, |
| 3607 | it is converted to upper case. Then bit 6 of the character (hex 40) is |
| 3608 | inverted. Thus \cz becomes hex 1A (z is 7A), but \c{ becomes hex 3B ({ |
| 3609 | is 7B), while \c; becomes hex 7B (; is 3B). If the byte following \c |
| 3610 | has a value greater than 127, a compile-time error occurs. This locks |
| 3611 | out non-ASCII characters in both byte mode and UTF-8 mode. (When PCRE |
| 3612 | is compiled in EBCDIC mode, all byte values are valid. A lower case |
| 3613 | letter is converted to upper case, and then the 0xc0 bits are flipped.) |
| 3614 | |
| 3615 | By default, after \x, from zero to two hexadecimal digits are read |
| 3616 | (letters can be in upper or lower case). Any number of hexadecimal dig- |
| 3617 | its may appear between \x{ and }, but the value of the character code |
| 3618 | must be less than 256 in non-UTF-8 mode, and less than 2**31 in UTF-8 |
| 3619 | mode. That is, the maximum value in hexadecimal is 7FFFFFFF. Note that |
| 3620 | this is bigger than the largest Unicode code point, which is 10FFFF. |
| 3621 | |
| 3622 | If characters other than hexadecimal digits appear between \x{ and }, |
| 3623 | or if there is no terminating }, this form of escape is not recognized. |
| 3624 | Instead, the initial \x will be interpreted as a basic hexadecimal |
| 3625 | escape, with no following digits, giving a character whose value is |
| 3626 | zero. |
| 3627 | |
| 3628 | If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x |
| 3629 | is as just described only when it is followed by two hexadecimal dig- |
| 3630 | its. Otherwise, it matches a literal "x" character. In JavaScript |
| 3631 | mode, support for code points greater than 256 is provided by \u, which |
| 3632 | must be followed by four hexadecimal digits; otherwise it matches a |
| 3633 | literal "u" character. |
| 3634 | |
| 3635 | Characters whose value is less than 256 can be defined by either of the |
| 3636 | two syntaxes for \x (or by \u in JavaScript mode). There is no differ- |
| 3637 | ence in the way they are handled. For example, \xdc is exactly the same |
| 3638 | as \x{dc} (or \u00dc in JavaScript mode). |
| 3639 | |
| 3640 | After \0 up to two further octal digits are read. If there are fewer |
| 3641 | than two digits, just those that are present are used. Thus the |
| 3642 | sequence \0\x\07 specifies two binary zeros followed by a BEL character |
| 3643 | (code value 7). Make sure you supply two digits after the initial zero |
| 3644 | if the pattern character that follows is itself an octal digit. |
| 3645 | |
| 3646 | The handling of a backslash followed by a digit other than 0 is compli- |
| 3647 | cated. Outside a character class, PCRE reads it and any following dig- |
| 3648 | its as a decimal number. If the number is less than 10, or if there |
| 3649 | have been at least that many previous capturing left parentheses in the |
| 3650 | expression, the entire sequence is taken as a back reference. A |
| 3651 | description of how this works is given later, following the discussion |
| 3652 | of parenthesized subpatterns. |
| 3653 | |
| 3654 | Inside a character class, or if the decimal number is greater than 9 |
| 3655 | and there have not been that many capturing subpatterns, PCRE re-reads |
| 3656 | up to three octal digits following the backslash, and uses them to gen- |
| 3657 | erate a data character. Any subsequent digits stand for themselves. In |
| 3658 | non-UTF-8 mode, the value of a character specified in octal must be |
| 3659 | less than \400. In UTF-8 mode, values up to \777 are permitted. For |
| 3660 | example: |
| 3661 | |
| 3662 | \040 is another way of writing a space |
| 3663 | \40 is the same, provided there are fewer than 40 |
| 3664 | previous capturing subpatterns |
| 3665 | \7 is always a back reference |
| 3666 | \11 might be a back reference, or another way of |
| 3667 | writing a tab |
| 3668 | \011 is always a tab |
| 3669 | \0113 is a tab followed by the character "3" |
| 3670 | \113 might be a back reference, otherwise the |
| 3671 | character with octal code 113 |
| 3672 | \377 might be a back reference, otherwise |
| 3673 | the byte consisting entirely of 1 bits |
| 3674 | \81 is either a back reference, or a binary zero |
| 3675 | followed by the two characters "8" and "1" |
| 3676 | |
| 3677 | Note that octal values of 100 or greater must not be introduced by a |
| 3678 | leading zero, because no more than three octal digits are ever read. |
| 3679 | |
| 3680 | All the sequences that define a single character value can be used both |
| 3681 | inside and outside character classes. In addition, inside a character |
| 3682 | class, \b is interpreted as the backspace character (hex 08). |
| 3683 | |
| 3684 | \N is not allowed in a character class. \B, \R, and \X are not special |
| 3685 | inside a character class. Like other unrecognized escape sequences, |
| 3686 | they are treated as the literal characters "B", "R", and "X" by |
| 3687 | default, but cause an error if the PCRE_EXTRA option is set. Outside a |
| 3688 | character class, these sequences have different meanings. |
| 3689 | |
| 3690 | Unsupported escape sequences |
| 3691 | |
| 3692 | In Perl, the sequences \l, \L, \u, and \U are recognized by its string |
| 3693 | handler and used to modify the case of following characters. By |
| 3694 | default, PCRE does not support these escape sequences. However, if the |
| 3695 | PCRE_JAVASCRIPT_COMPAT option is set, \U matches a "U" character, and |
| 3696 | \u can be used to define a character by code point, as described in the |
| 3697 | previous section. |
| 3698 | |
| 3699 | Absolute and relative back references |
| 3700 | |
| 3701 | The sequence \g followed by an unsigned or a negative number, option- |
| 3702 | ally enclosed in braces, is an absolute or relative back reference. A |
| 3703 | named back reference can be coded as \g{name}. Back references are dis- |
| 3704 | cussed later, following the discussion of parenthesized subpatterns. |
| 3705 | |
| 3706 | Absolute and relative subroutine calls |
| 3707 | |
| 3708 | For compatibility with Oniguruma, the non-Perl syntax \g followed by a |
| 3709 | name or a number enclosed either in angle brackets or single quotes, is |
| 3710 | an alternative syntax for referencing a subpattern as a "subroutine". |
| 3711 | Details are discussed later. Note that \g{...} (Perl syntax) and |
| 3712 | \g<...> (Oniguruma syntax) are not synonymous. The former is a back |
| 3713 | reference; the latter is a subroutine call. |
| 3714 | |
| 3715 | Generic character types |
| 3716 | |
| 3717 | Another use of backslash is for specifying generic character types: |
| 3718 | |
| 3719 | \d any decimal digit |
| 3720 | \D any character that is not a decimal digit |
| 3721 | \h any horizontal whitespace character |
| 3722 | \H any character that is not a horizontal whitespace character |
| 3723 | \s any whitespace character |
| 3724 | \S any character that is not a whitespace character |
| 3725 | \v any vertical whitespace character |
| 3726 | \V any character that is not a vertical whitespace character |
| 3727 | \w any "word" character |
| 3728 | \W any "non-word" character |
| 3729 | |
| 3730 | There is also the single sequence \N, which matches a non-newline char- |
| 3731 | acter. This is the same as the "." metacharacter when PCRE_DOTALL is |
| 3732 | not set. Perl also uses \N to match characters by name; PCRE does not |
| 3733 | support this. |
| 3734 | |
| 3735 | Each pair of lower and upper case escape sequences partitions the com- |
| 3736 | plete set of characters into two disjoint sets. Any given character |
| 3737 | matches one, and only one, of each pair. The sequences can appear both |
| 3738 | inside and outside character classes. They each match one character of |
| 3739 | the appropriate type. If the current matching point is at the end of |
| 3740 | the subject string, all of them fail, because there is no character to |
| 3741 | match. |
| 3742 | |
| 3743 | For compatibility with Perl, \s does not match the VT character (code |
| 3744 | 11). This makes it different from the the POSIX "space" class. The \s |
| 3745 | characters are HT (9), LF (10), FF (12), CR (13), and space (32). If |
| 3746 | "use locale;" is included in a Perl script, \s may match the VT charac- |
| 3747 | ter. In PCRE, it never does. |
| 3748 | |
| 3749 | A "word" character is an underscore or any character that is a letter |
| 3750 | or digit. By default, the definition of letters and digits is con- |
| 3751 | trolled by PCRE's low-valued character tables, and may vary if locale- |
| 3752 | specific matching is taking place (see "Locale support" in the pcreapi |
| 3753 | page). For example, in a French locale such as "fr_FR" in Unix-like |
| 3754 | systems, or "french" in Windows, some character codes greater than 128 |
| 3755 | are used for accented letters, and these are then matched by \w. The |
| 3756 | use of locales with Unicode is discouraged. |
| 3757 | |
| 3758 | By default, in UTF-8 mode, characters with values greater than 128 |
| 3759 | never match \d, \s, or \w, and always match \D, \S, and \W. These |
| 3760 | sequences retain their original meanings from before UTF-8 support was |
| 3761 | available, mainly for efficiency reasons. However, if PCRE is compiled |
| 3762 | with Unicode property support, and the PCRE_UCP option is set, the be- |
| 3763 | haviour is changed so that Unicode properties are used to determine |
| 3764 | character types, as follows: |
| 3765 | |
| 3766 | \d any character that \p{Nd} matches (decimal digit) |
| 3767 | \s any character that \p{Z} matches, plus HT, LF, FF, CR |
| 3768 | \w any character that \p{L} or \p{N} matches, plus underscore |
| 3769 | |
| 3770 | The upper case escapes match the inverse sets of characters. Note that |
| 3771 | \d matches only decimal digits, whereas \w matches any Unicode digit, |
| 3772 | as well as any Unicode letter, and underscore. Note also that PCRE_UCP |
| 3773 | affects \b, and \B because they are defined in terms of \w and \W. |
| 3774 | Matching these sequences is noticeably slower when PCRE_UCP is set. |
| 3775 | |
| 3776 | The sequences \h, \H, \v, and \V are features that were added to Perl |
| 3777 | at release 5.10. In contrast to the other sequences, which match only |
| 3778 | ASCII characters by default, these always match certain high-valued |
| 3779 | codepoints in UTF-8 mode, whether or not PCRE_UCP is set. The horizon- |
| 3780 | tal space characters are: |
| 3781 | |
| 3782 | U+0009 Horizontal tab |
| 3783 | U+0020 Space |
| 3784 | U+00A0 Non-break space |
| 3785 | U+1680 Ogham space mark |
| 3786 | U+180E Mongolian vowel separator |
| 3787 | U+2000 En quad |
| 3788 | U+2001 Em quad |
| 3789 | U+2002 En space |
| 3790 | U+2003 Em space |
| 3791 | U+2004 Three-per-em space |
| 3792 | U+2005 Four-per-em space |
| 3793 | U+2006 Six-per-em space |
| 3794 | U+2007 Figure space |
| 3795 | U+2008 Punctuation space |
| 3796 | U+2009 Thin space |
| 3797 | U+200A Hair space |
| 3798 | U+202F Narrow no-break space |
| 3799 | U+205F Medium mathematical space |
| 3800 | U+3000 Ideographic space |
| 3801 | |
| 3802 | The vertical space characters are: |
| 3803 | |
| 3804 | U+000A Linefeed |
| 3805 | U+000B Vertical tab |
| 3806 | U+000C Formfeed |
| 3807 | U+000D Carriage return |
| 3808 | U+0085 Next line |
| 3809 | U+2028 Line separator |
| 3810 | U+2029 Paragraph separator |
| 3811 | |
| 3812 | Newline sequences |
| 3813 | |
| 3814 | Outside a character class, by default, the escape sequence \R matches |
| 3815 | any Unicode newline sequence. In non-UTF-8 mode \R is equivalent to the |
| 3816 | following: |
| 3817 | |
| 3818 | (?>\r\n|\n|\x0b|\f|\r|\x85) |
| 3819 | |
| 3820 | This is an example of an "atomic group", details of which are given |
| 3821 | below. This particular group matches either the two-character sequence |
| 3822 | CR followed by LF, or one of the single characters LF (linefeed, |
| 3823 | U+000A), VT (vertical tab, U+000B), FF (formfeed, U+000C), CR (carriage |
| 3824 | return, U+000D), or NEL (next line, U+0085). The two-character sequence |
| 3825 | is treated as a single unit that cannot be split. |
| 3826 | |
| 3827 | In UTF-8 mode, two additional characters whose codepoints are greater |
| 3828 | than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa- |
| 3829 | rator, U+2029). Unicode character property support is not needed for |
| 3830 | these characters to be recognized. |
| 3831 | |
| 3832 | It is possible to restrict \R to match only CR, LF, or CRLF (instead of |
| 3833 | the complete set of Unicode line endings) by setting the option |
| 3834 | PCRE_BSR_ANYCRLF either at compile time or when the pattern is matched. |
| 3835 | (BSR is an abbrevation for "backslash R".) This can be made the default |
| 3836 | when PCRE is built; if this is the case, the other behaviour can be |
| 3837 | requested via the PCRE_BSR_UNICODE option. It is also possible to |
| 3838 | specify these settings by starting a pattern string with one of the |
| 3839 | following sequences: |
| 3840 | |
| 3841 | (*BSR_ANYCRLF) CR, LF, or CRLF only |
| 3842 | (*BSR_UNICODE) any Unicode newline sequence |
| 3843 | |
| 3844 | These override the default and the options given to pcre_compile() or |
| 3845 | pcre_compile2(), but they can be overridden by options given to |
| 3846 | pcre_exec() or pcre_dfa_exec(). Note that these special settings, which |
| 3847 | are not Perl-compatible, are recognized only at the very start of a |
| 3848 | pattern, and that they must be in upper case. If more than one of them |
| 3849 | is present, the last one is used. They can be combined with a change of |
| 3850 | newline convention; for example, a pattern can start with: |
| 3851 | |
| 3852 | (*ANY)(*BSR_ANYCRLF) |
| 3853 | |
| 3854 | They can also be combined with the (*UTF8) or (*UCP) special sequences. |
| 3855 | Inside a character class, \R is treated as an unrecognized escape |
| 3856 | sequence, and so matches the letter "R" by default, but causes an error |
| 3857 | if PCRE_EXTRA is set. |
| 3858 | |
| 3859 | Unicode character properties |
| 3860 | |
| 3861 | When PCRE is built with Unicode character property support, three addi- |
| 3862 | tional escape sequences that match characters with specific properties |
| 3863 | are available. When not in UTF-8 mode, these sequences are of course |
| 3864 | limited to testing characters whose codepoints are less than 256, but |
| 3865 | they do work in this mode. The extra escape sequences are: |
| 3866 | |
| 3867 | \p{xx} a character with the xx property |
| 3868 | \P{xx} a character without the xx property |
| 3869 | \X an extended Unicode sequence |
| 3870 | |
| 3871 | The property names represented by xx above are limited to the Unicode |
| 3872 | script names, the general category properties, "Any", which matches any |
| 3873 | character (including newline), and some special PCRE properties |
| 3874 | (described in the next section). Other Perl properties such as "InMu- |
| 3875 | sicalSymbols" are not currently supported by PCRE. Note that \P{Any} |
| 3876 | does not match any characters, so always causes a match failure. |
| 3877 | |
| 3878 | Sets of Unicode characters are defined as belonging to certain scripts. |
| 3879 | A character from one of these sets can be matched using a script name. |
| 3880 | For example: |
| 3881 | |
| 3882 | \p{Greek} |
| 3883 | \P{Han} |
| 3884 | |
| 3885 | Those that are not part of an identified script are lumped together as |
| 3886 | "Common". The current list of scripts is: |
| 3887 | |
| 3888 | Arabic, Armenian, Avestan, Balinese, Bamum, Bengali, Bopomofo, Braille, |
| 3889 | Buginese, Buhid, Canadian_Aboriginal, Carian, Cham, Cherokee, Common, |
| 3890 | Coptic, Cuneiform, Cypriot, Cyrillic, Deseret, Devanagari, Egyp- |
| 3891 | tian_Hieroglyphs, Ethiopic, Georgian, Glagolitic, Gothic, Greek, |
| 3892 | Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana, Impe- |
| 3893 | rial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscriptional_Parthian, |
| 3894 | Javanese, Kaithi, Kannada, Katakana, Kayah_Li, Kharoshthi, Khmer, Lao, |
| 3895 | Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian, Lydian, Malayalam, |
| 3896 | Meetei_Mayek, Mongolian, Myanmar, New_Tai_Lue, Nko, Ogham, Old_Italic, |
| 3897 | Old_Persian, Old_South_Arabian, Old_Turkic, Ol_Chiki, Oriya, Osmanya, |
| 3898 | Phags_Pa, Phoenician, Rejang, Runic, Samaritan, Saurashtra, Shavian, |
| 3899 | Sinhala, Sundanese, Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le, |
| 3900 | Tai_Tham, Tai_Viet, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, |
| 3901 | Ugaritic, Vai, Yi. |
| 3902 | |
| 3903 | Each character has exactly one Unicode general category property, spec- |
| 3904 | ified by a two-letter abbreviation. For compatibility with Perl, nega- |
| 3905 | tion can be specified by including a circumflex between the opening |
| 3906 | brace and the property name. For example, \p{^Lu} is the same as |
| 3907 | \P{Lu}. |
| 3908 | |
| 3909 | If only one letter is specified with \p or \P, it includes all the gen- |
| 3910 | eral category properties that start with that letter. In this case, in |
| 3911 | the absence of negation, the curly brackets in the escape sequence are |
| 3912 | optional; these two examples have the same effect: |
| 3913 | |
| 3914 | \p{L} |
| 3915 | \pL |
| 3916 | |
| 3917 | The following general category property codes are supported: |
| 3918 | |
| 3919 | C Other |
| 3920 | Cc Control |
| 3921 | Cf Format |
| 3922 | Cn Unassigned |
| 3923 | Co Private use |
| 3924 | Cs Surrogate |
| 3925 | |
| 3926 | L Letter |
| 3927 | Ll Lower case letter |
| 3928 | Lm Modifier letter |
| 3929 | Lo Other letter |
| 3930 | Lt Title case letter |
| 3931 | Lu Upper case letter |
| 3932 | |
| 3933 | M Mark |
| 3934 | Mc Spacing mark |
| 3935 | Me Enclosing mark |
| 3936 | Mn Non-spacing mark |
| 3937 | |
| 3938 | N Number |
| 3939 | Nd Decimal number |
| 3940 | Nl Letter number |
| 3941 | No Other number |
| 3942 | |
| 3943 | P Punctuation |
| 3944 | Pc Connector punctuation |
| 3945 | Pd Dash punctuation |
| 3946 | Pe Close punctuation |
| 3947 | Pf Final punctuation |
| 3948 | Pi Initial punctuation |
| 3949 | Po Other punctuation |
| 3950 | Ps Open punctuation |
| 3951 | |
| 3952 | S Symbol |
| 3953 | Sc Currency symbol |
| 3954 | Sk Modifier symbol |
| 3955 | Sm Mathematical symbol |
| 3956 | So Other symbol |
| 3957 | |
| 3958 | Z Separator |
| 3959 | Zl Line separator |
| 3960 | Zp Paragraph separator |
| 3961 | Zs Space separator |
| 3962 | |
| 3963 | The special property L& is also supported: it matches a character that |
| 3964 | has the Lu, Ll, or Lt property, in other words, a letter that is not |
| 3965 | classified as a modifier or "other". |
| 3966 | |
| 3967 | The Cs (Surrogate) property applies only to characters in the range |
| 3968 | U+D800 to U+DFFF. Such characters are not valid in UTF-8 strings (see |
| 3969 | RFC 3629) and so cannot be tested by PCRE, unless UTF-8 validity check- |
| 3970 | ing has been turned off (see the discussion of PCRE_NO_UTF8_CHECK in |
| 3971 | the pcreapi page). Perl does not support the Cs property. |
| 3972 | |
| 3973 | The long synonyms for property names that Perl supports (such as |
| 3974 | \p{Letter}) are not supported by PCRE, nor is it permitted to prefix |
| 3975 | any of these properties with "Is". |
| 3976 | |
| 3977 | No character that is in the Unicode table has the Cn (unassigned) prop- |
| 3978 | erty. Instead, this property is assumed for any code point that is not |
| 3979 | in the Unicode table. |
| 3980 | |
| 3981 | Specifying caseless matching does not affect these escape sequences. |
| 3982 | For example, \p{Lu} always matches only upper case letters. |
| 3983 | |
| 3984 | The \X escape matches any number of Unicode characters that form an |
| 3985 | extended Unicode sequence. \X is equivalent to |
| 3986 | |
| 3987 | (?>\PM\pM*) |
| 3988 | |
| 3989 | That is, it matches a character without the "mark" property, followed |
| 3990 | by zero or more characters with the "mark" property, and treats the |
| 3991 | sequence as an atomic group (see below). Characters with the "mark" |
| 3992 | property are typically accents that affect the preceding character. |
| 3993 | None of them have codepoints less than 256, so in non-UTF-8 mode \X |
| 3994 | matches any one character. |
| 3995 | |
| 3996 | Note that recent versions of Perl have changed \X to match what Unicode |
| 3997 | calls an "extended grapheme cluster", which has a more complicated def- |
| 3998 | inition. |
| 3999 | |
| 4000 | Matching characters by Unicode property is not fast, because PCRE has |
| 4001 | to search a structure that contains data for over fifteen thousand |
| 4002 | characters. That is why the traditional escape sequences such as \d and |
| 4003 | \w do not use Unicode properties in PCRE by default, though you can |
| 4004 | make them do so by setting the PCRE_UCP option for pcre_compile() or by |
| 4005 | starting the pattern with (*UCP). |
| 4006 | |
| 4007 | PCRE's additional properties |
| 4008 | |
| 4009 | As well as the standard Unicode properties described in the previous |
| 4010 | section, PCRE supports four more that make it possible to convert tra- |
| 4011 | ditional escape sequences such as \w and \s and POSIX character classes |
| 4012 | to use Unicode properties. PCRE uses these non-standard, non-Perl prop- |
| 4013 | erties internally when PCRE_UCP is set. They are: |
| 4014 | |
| 4015 | Xan Any alphanumeric character |
| 4016 | Xps Any POSIX space character |
| 4017 | Xsp Any Perl space character |
| 4018 | Xwd Any Perl "word" character |
| 4019 | |
| 4020 | Xan matches characters that have either the L (letter) or the N (num- |
| 4021 | ber) property. Xps matches the characters tab, linefeed, vertical tab, |
| 4022 | formfeed, or carriage return, and any other character that has the Z |
| 4023 | (separator) property. Xsp is the same as Xps, except that vertical tab |
| 4024 | is excluded. Xwd matches the same characters as Xan, plus underscore. |
| 4025 | |
| 4026 | Resetting the match start |
| 4027 | |
| 4028 | The escape sequence \K causes any previously matched characters not to |
| 4029 | be included in the final matched sequence. For example, the pattern: |
| 4030 | |
| 4031 | foo\Kbar |
| 4032 | |
| 4033 | matches "foobar", but reports that it has matched "bar". This feature |
| 4034 | is similar to a lookbehind assertion (described below). However, in |
| 4035 | this case, the part of the subject before the real match does not have |
| 4036 | to be of fixed length, as lookbehind assertions do. The use of \K does |
| 4037 | not interfere with the setting of captured substrings. For example, |
| 4038 | when the pattern |
| 4039 | |
| 4040 | (foo)\Kbar |
| 4041 | |
| 4042 | matches "foobar", the first substring is still set to "foo". |
| 4043 | |
| 4044 | Perl documents that the use of \K within assertions is "not well |
| 4045 | defined". In PCRE, \K is acted upon when it occurs inside positive |
| 4046 | assertions, but is ignored in negative assertions. |
| 4047 | |
| 4048 | Simple assertions |
| 4049 | |
| 4050 | The final use of backslash is for certain simple assertions. An asser- |
| 4051 | tion specifies a condition that has to be met at a particular point in |
| 4052 | a match, without consuming any characters from the subject string. The |
| 4053 | use of subpatterns for more complicated assertions is described below. |
| 4054 | The backslashed assertions are: |
| 4055 | |
| 4056 | \b matches at a word boundary |
| 4057 | \B matches when not at a word boundary |
| 4058 | \A matches at the start of the subject |
| 4059 | \Z matches at the end of the subject |
| 4060 | also matches before a newline at the end of the subject |
| 4061 | \z matches only at the end of the subject |
| 4062 | \G matches at the first matching position in the subject |
| 4063 | |
| 4064 | Inside a character class, \b has a different meaning; it matches the |
| 4065 | backspace character. If any other of these assertions appears in a |
| 4066 | character class, by default it matches the corresponding literal char- |
| 4067 | acter (for example, \B matches the letter B). However, if the |
| 4068 | PCRE_EXTRA option is set, an "invalid escape sequence" error is gener- |
| 4069 | ated instead. |
| 4070 | |
| 4071 | A word boundary is a position in the subject string where the current |
| 4072 | character and the previous character do not both match \w or \W (i.e. |
| 4073 | one matches \w and the other matches \W), or the start or end of the |
| 4074 | string if the first or last character matches \w, respectively. In |
| 4075 | UTF-8 mode, the meanings of \w and \W can be changed by setting the |
| 4076 | PCRE_UCP option. When this is done, it also affects \b and \B. Neither |
| 4077 | PCRE nor Perl has a separate "start of word" or "end of word" metase- |
| 4078 | quence. However, whatever follows \b normally determines which it is. |
| 4079 | For example, the fragment \ba matches "a" at the start of a word. |
| 4080 | |
| 4081 | The \A, \Z, and \z assertions differ from the traditional circumflex |
| 4082 | and dollar (described in the next section) in that they only ever match |
| 4083 | at the very start and end of the subject string, whatever options are |
| 4084 | set. Thus, they are independent of multiline mode. These three asser- |
| 4085 | tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which |
| 4086 | affect only the behaviour of the circumflex and dollar metacharacters. |
| 4087 | However, if the startoffset argument of pcre_exec() is non-zero, indi- |
| 4088 | cating that matching is to start at a point other than the beginning of |
| 4089 | the subject, \A can never match. The difference between \Z and \z is |
| 4090 | that \Z matches before a newline at the end of the string as well as at |
| 4091 | the very end, whereas \z matches only at the end. |
| 4092 | |
| 4093 | The \G assertion is true only when the current matching position is at |
| 4094 | the start point of the match, as specified by the startoffset argument |
| 4095 | of pcre_exec(). It differs from \A when the value of startoffset is |
| 4096 | non-zero. By calling pcre_exec() multiple times with appropriate argu- |
| 4097 | ments, you can mimic Perl's /g option, and it is in this kind of imple- |
| 4098 | mentation where \G can be useful. |
| 4099 | |
| 4100 | Note, however, that PCRE's interpretation of \G, as the start of the |
| 4101 | current match, is subtly different from Perl's, which defines it as the |
| 4102 | end of the previous match. In Perl, these can be different when the |
| 4103 | previously matched string was empty. Because PCRE does just one match |
| 4104 | at a time, it cannot reproduce this behaviour. |
| 4105 | |
| 4106 | If all the alternatives of a pattern begin with \G, the expression is |
| 4107 | anchored to the starting match position, and the "anchored" flag is set |
| 4108 | in the compiled regular expression. |
| 4109 | |
| 4110 | |
| 4111 | CIRCUMFLEX AND DOLLAR |
| 4112 | |
| 4113 | Outside a character class, in the default matching mode, the circumflex |
| 4114 | character is an assertion that is true only if the current matching |
| 4115 | point is at the start of the subject string. If the startoffset argu- |
| 4116 | ment of pcre_exec() is non-zero, circumflex can never match if the |
| 4117 | PCRE_MULTILINE option is unset. Inside a character class, circumflex |
| 4118 | has an entirely different meaning (see below). |
| 4119 | |
| 4120 | Circumflex need not be the first character of the pattern if a number |
| 4121 | of alternatives are involved, but it should be the first thing in each |
| 4122 | alternative in which it appears if the pattern is ever to match that |
| 4123 | branch. If all possible alternatives start with a circumflex, that is, |
| 4124 | if the pattern is constrained to match only at the start of the sub- |
| 4125 | ject, it is said to be an "anchored" pattern. (There are also other |
| 4126 | constructs that can cause a pattern to be anchored.) |
| 4127 | |
| 4128 | A dollar character is an assertion that is true only if the current |
| 4129 | matching point is at the end of the subject string, or immediately |
| 4130 | before a newline at the end of the string (by default). Dollar need not |
| 4131 | be the last character of the pattern if a number of alternatives are |
| 4132 | involved, but it should be the last item in any branch in which it |
| 4133 | appears. Dollar has no special meaning in a character class. |
| 4134 | |
| 4135 | The meaning of dollar can be changed so that it matches only at the |
| 4136 | very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at |
| 4137 | compile time. This does not affect the \Z assertion. |
| 4138 | |
| 4139 | The meanings of the circumflex and dollar characters are changed if the |
| 4140 | PCRE_MULTILINE option is set. When this is the case, a circumflex |
| 4141 | matches immediately after internal newlines as well as at the start of |
| 4142 | the subject string. It does not match after a newline that ends the |
| 4143 | string. A dollar matches before any newlines in the string, as well as |
| 4144 | at the very end, when PCRE_MULTILINE is set. When newline is specified |
| 4145 | as the two-character sequence CRLF, isolated CR and LF characters do |
| 4146 | not indicate newlines. |
| 4147 | |
| 4148 | For example, the pattern /^abc$/ matches the subject string "def\nabc" |
| 4149 | (where \n represents a newline) in multiline mode, but not otherwise. |
| 4150 | Consequently, patterns that are anchored in single line mode because |
| 4151 | all branches start with ^ are not anchored in multiline mode, and a |
| 4152 | match for circumflex is possible when the startoffset argument of |
| 4153 | pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if |
| 4154 | PCRE_MULTILINE is set. |
| 4155 | |
| 4156 | Note that the sequences \A, \Z, and \z can be used to match the start |
| 4157 | and end of the subject in both modes, and if all branches of a pattern |
| 4158 | start with \A it is always anchored, whether or not PCRE_MULTILINE is |
| 4159 | set. |
| 4160 | |
| 4161 | |
| 4162 | FULL STOP (PERIOD, DOT) AND \N |
| 4163 | |
| 4164 | Outside a character class, a dot in the pattern matches any one charac- |
| 4165 | ter in the subject string except (by default) a character that signi- |
| 4166 | fies the end of a line. In UTF-8 mode, the matched character may be |
| 4167 | more than one byte long. |
| 4168 | |
| 4169 | When a line ending is defined as a single character, dot never matches |
| 4170 | that character; when the two-character sequence CRLF is used, dot does |
| 4171 | not match CR if it is immediately followed by LF, but otherwise it |
| 4172 | matches all characters (including isolated CRs and LFs). When any Uni- |
| 4173 | code line endings are being recognized, dot does not match CR or LF or |
| 4174 | any of the other line ending characters. |
| 4175 | |
| 4176 | The behaviour of dot with regard to newlines can be changed. If the |
| 4177 | PCRE_DOTALL option is set, a dot matches any one character, without |
| 4178 | exception. If the two-character sequence CRLF is present in the subject |
| 4179 | string, it takes two dots to match it. |
| 4180 | |
| 4181 | The handling of dot is entirely independent of the handling of circum- |
| 4182 | flex and dollar, the only relationship being that they both involve |
| 4183 | newlines. Dot has no special meaning in a character class. |
| 4184 | |
| 4185 | The escape sequence \N behaves like a dot, except that it is not |
| 4186 | affected by the PCRE_DOTALL option. In other words, it matches any |
| 4187 | character except one that signifies the end of a line. Perl also uses |
| 4188 | \N to match characters by name; PCRE does not support this. |
| 4189 | |
| 4190 | |
| 4191 | MATCHING A SINGLE BYTE |
| 4192 | |
| 4193 | Outside a character class, the escape sequence \C matches any one byte, |
| 4194 | both in and out of UTF-8 mode. Unlike a dot, it always matches line- |
| 4195 | ending characters. The feature is provided in Perl in order to match |
| 4196 | individual bytes in UTF-8 mode, but it is unclear how it can usefully |
| 4197 | be used. Because \C breaks up characters into individual bytes, match- |
| 4198 | ing one byte with \C in UTF-8 mode means that the rest of the string |
| 4199 | may start with a malformed UTF-8 character. This has undefined results, |
| 4200 | because PCRE assumes that it is dealing with valid UTF-8 strings (and |
| 4201 | by default it checks this at the start of processing unless the |
| 4202 | PCRE_NO_UTF8_CHECK option is used). |
| 4203 | |
| 4204 | PCRE does not allow \C to appear in lookbehind assertions (described |
| 4205 | below) in UTF-8 mode, because this would make it impossible to calcu- |
| 4206 | late the length of the lookbehind. |
| 4207 | |
| 4208 | In general, the \C escape sequence is best avoided in UTF-8 mode. How- |
| 4209 | ever, one way of using it that avoids the problem of malformed UTF-8 |
| 4210 | characters is to use a lookahead to check the length of the next char- |
| 4211 | acter, as in this pattern (ignore white space and line breaks): |
| 4212 | |
| 4213 | (?| (?=[\x00-\x7f])(\C) | |
| 4214 | (?=[\x80-\x{7ff}])(\C)(\C) | |
| 4215 | (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) | |
| 4216 | (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C)) |
| 4217 | |
| 4218 | A group that starts with (?| resets the capturing parentheses numbers |
| 4219 | in each alternative (see "Duplicate Subpattern Numbers" below). The |
| 4220 | assertions at the start of each branch check the next UTF-8 character |
| 4221 | for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The |
| 4222 | character's individual bytes are then captured by the appropriate num- |
| 4223 | ber of groups. |
| 4224 | |
| 4225 | |
| 4226 | SQUARE BRACKETS AND CHARACTER CLASSES |
| 4227 | |
| 4228 | An opening square bracket introduces a character class, terminated by a |
| 4229 | closing square bracket. A closing square bracket on its own is not spe- |
| 4230 | cial by default. However, if the PCRE_JAVASCRIPT_COMPAT option is set, |
| 4231 | a lone closing square bracket causes a compile-time error. If a closing |
| 4232 | square bracket is required as a member of the class, it should be the |
| 4233 | first data character in the class (after an initial circumflex, if |
| 4234 | present) or escaped with a backslash. |
| 4235 | |
| 4236 | A character class matches a single character in the subject. In UTF-8 |
| 4237 | mode, the character may be more than one byte long. A matched character |
| 4238 | must be in the set of characters defined by the class, unless the first |
| 4239 | character in the class definition is a circumflex, in which case the |
| 4240 | subject character must not be in the set defined by the class. If a |
| 4241 | circumflex is actually required as a member of the class, ensure it is |
| 4242 | not the first character, or escape it with a backslash. |
| 4243 | |
| 4244 | For example, the character class [aeiou] matches any lower case vowel, |
| 4245 | while [^aeiou] matches any character that is not a lower case vowel. |
| 4246 | Note that a circumflex is just a convenient notation for specifying the |
| 4247 | characters that are in the class by enumerating those that are not. A |
| 4248 | class that starts with a circumflex is not an assertion; it still con- |
| 4249 | sumes a character from the subject string, and therefore it fails if |
| 4250 | the current pointer is at the end of the string. |
| 4251 | |
| 4252 | In UTF-8 mode, characters with values greater than 255 can be included |
| 4253 | in a class as a literal string of bytes, or by using the \x{ escaping |
| 4254 | mechanism. |
| 4255 | |
| 4256 | When caseless matching is set, any letters in a class represent both |
| 4257 | their upper case and lower case versions, so for example, a caseless |
| 4258 | [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not |
| 4259 | match "A", whereas a caseful version would. In UTF-8 mode, PCRE always |
| 4260 | understands the concept of case for characters whose values are less |
| 4261 | than 128, so caseless matching is always possible. For characters with |
| 4262 | higher values, the concept of case is supported if PCRE is compiled |
| 4263 | with Unicode property support, but not otherwise. If you want to use |
| 4264 | caseless matching in UTF8-mode for characters 128 and above, you must |
| 4265 | ensure that PCRE is compiled with Unicode property support as well as |
| 4266 | with UTF-8 support. |
| 4267 | |
| 4268 | Characters that might indicate line breaks are never treated in any |
| 4269 | special way when matching character classes, whatever line-ending |
| 4270 | sequence is in use, and whatever setting of the PCRE_DOTALL and |
| 4271 | PCRE_MULTILINE options is used. A class such as [^a] always matches one |
| 4272 | of these characters. |
| 4273 | |
| 4274 | The minus (hyphen) character can be used to specify a range of charac- |
| 4275 | ters in a character class. For example, [d-m] matches any letter |
| 4276 | between d and m, inclusive. If a minus character is required in a |
| 4277 | class, it must be escaped with a backslash or appear in a position |
| 4278 | where it cannot be interpreted as indicating a range, typically as the |
| 4279 | first or last character in the class. |
| 4280 | |
| 4281 | It is not possible to have the literal character "]" as the end charac- |
| 4282 | ter of a range. A pattern such as [W-]46] is interpreted as a class of |
| 4283 | two characters ("W" and "-") followed by a literal string "46]", so it |
| 4284 | would match "W46]" or "-46]". However, if the "]" is escaped with a |
| 4285 | backslash it is interpreted as the end of range, so [W-\]46] is inter- |
| 4286 | preted as a class containing a range followed by two other characters. |
| 4287 | The octal or hexadecimal representation of "]" can also be used to end |
| 4288 | a range. |
| 4289 | |
| 4290 | Ranges operate in the collating sequence of character values. They can |
| 4291 | also be used for characters specified numerically, for example |
| 4292 | [\000-\037]. In UTF-8 mode, ranges can include characters whose values |
| 4293 | are greater than 255, for example [\x{100}-\x{2ff}]. |
| 4294 | |
| 4295 | If a range that includes letters is used when caseless matching is set, |
| 4296 | it matches the letters in either case. For example, [W-c] is equivalent |
| 4297 | to [][\\^_`wxyzabc], matched caselessly, and in non-UTF-8 mode, if |
| 4298 | character tables for a French locale are in use, [\xc8-\xcb] matches |
| 4299 | accented E characters in both cases. In UTF-8 mode, PCRE supports the |
| 4300 | concept of case for characters with values greater than 128 only when |
| 4301 | it is compiled with Unicode property support. |
| 4302 | |
| 4303 | The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, \V, |
| 4304 | \w, and \W may appear in a character class, and add the characters that |
| 4305 | they match to the class. For example, [\dABCDEF] matches any hexadeci- |
| 4306 | mal digit. In UTF-8 mode, the PCRE_UCP option affects the meanings of |
| 4307 | \d, \s, \w and their upper case partners, just as it does when they |
| 4308 | appear outside a character class, as described in the section entitled |
| 4309 | "Generic character types" above. The escape sequence \b has a different |
| 4310 | meaning inside a character class; it matches the backspace character. |
| 4311 | The sequences \B, \N, \R, and \X are not special inside a character |
| 4312 | class. Like any other unrecognized escape sequences, they are treated |
| 4313 | as the literal characters "B", "N", "R", and "X" by default, but cause |
| 4314 | an error if the PCRE_EXTRA option is set. |
| 4315 | |
| 4316 | A circumflex can conveniently be used with the upper case character |
| 4317 | types to specify a more restricted set of characters than the matching |
| 4318 | lower case type. For example, the class [^\W_] matches any letter or |
| 4319 | digit, but not underscore, whereas [\w] includes underscore. A positive |
| 4320 | character class should be read as "something OR something OR ..." and a |
| 4321 | negative class as "NOT something AND NOT something AND NOT ...". |
| 4322 | |
| 4323 | The only metacharacters that are recognized in character classes are |
| 4324 | backslash, hyphen (only where it can be interpreted as specifying a |
| 4325 | range), circumflex (only at the start), opening square bracket (only |
| 4326 | when it can be interpreted as introducing a POSIX class name - see the |
| 4327 | next section), and the terminating closing square bracket. However, |
| 4328 | escaping other non-alphanumeric characters does no harm. |
| 4329 | |
| 4330 | |
| 4331 | POSIX CHARACTER CLASSES |
| 4332 | |
| 4333 | Perl supports the POSIX notation for character classes. This uses names |
| 4334 | enclosed by [: and :] within the enclosing square brackets. PCRE also |
| 4335 | supports this notation. For example, |
| 4336 | |
| 4337 | [01[:alpha:]%] |
| 4338 | |
| 4339 | matches "0", "1", any alphabetic character, or "%". The supported class |
| 4340 | names are: |
| 4341 | |
| 4342 | alnum letters and digits |
| 4343 | alpha letters |
| 4344 | ascii character codes 0 - 127 |
| 4345 | blank space or tab only |
| 4346 | cntrl control characters |
| 4347 | digit decimal digits (same as \d) |
| 4348 | graph printing characters, excluding space |
| 4349 | lower lower case letters |
| 4350 | print printing characters, including space |
| 4351 | punct printing characters, excluding letters and digits and space |
| 4352 | space white space (not quite the same as \s) |
| 4353 | upper upper case letters |
| 4354 | word "word" characters (same as \w) |
| 4355 | xdigit hexadecimal digits |
| 4356 | |
| 4357 | The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), |
| 4358 | and space (32). Notice that this list includes the VT character (code |
| 4359 | 11). This makes "space" different to \s, which does not include VT (for |
| 4360 | Perl compatibility). |
| 4361 | |
| 4362 | The name "word" is a Perl extension, and "blank" is a GNU extension |
| 4363 | from Perl 5.8. Another Perl extension is negation, which is indicated |
| 4364 | by a ^ character after the colon. For example, |
| 4365 | |
| 4366 | [12[:^digit:]] |
| 4367 | |
| 4368 | matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the |
| 4369 | POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but |
| 4370 | these are not supported, and an error is given if they are encountered. |
| 4371 | |
| 4372 | By default, in UTF-8 mode, characters with values greater than 128 do |
| 4373 | not match any of the POSIX character classes. However, if the PCRE_UCP |
| 4374 | option is passed to pcre_compile(), some of the classes are changed so |
| 4375 | that Unicode character properties are used. This is achieved by replac- |
| 4376 | ing the POSIX classes by other sequences, as follows: |
| 4377 | |
| 4378 | [:alnum:] becomes \p{Xan} |
| 4379 | [:alpha:] becomes \p{L} |
| 4380 | [:blank:] becomes \h |
| 4381 | [:digit:] becomes \p{Nd} |
| 4382 | [:lower:] becomes \p{Ll} |
| 4383 | [:space:] becomes \p{Xps} |
| 4384 | [:upper:] becomes \p{Lu} |
| 4385 | [:word:] becomes \p{Xwd} |
| 4386 | |
| 4387 | Negated versions, such as [:^alpha:] use \P instead of \p. The other |
| 4388 | POSIX classes are unchanged, and match only characters with code points |
| 4389 | less than 128. |
| 4390 | |
| 4391 | |
| 4392 | VERTICAL BAR |
| 4393 | |
| 4394 | Vertical bar characters are used to separate alternative patterns. For |
| 4395 | example, the pattern |
| 4396 | |
| 4397 | gilbert|sullivan |
| 4398 | |
| 4399 | matches either "gilbert" or "sullivan". Any number of alternatives may |
| 4400 | appear, and an empty alternative is permitted (matching the empty |
| 4401 | string). The matching process tries each alternative in turn, from left |
| 4402 | to right, and the first one that succeeds is used. If the alternatives |
| 4403 | are within a subpattern (defined below), "succeeds" means matching the |
| 4404 | rest of the main pattern as well as the alternative in the subpattern. |
| 4405 | |
| 4406 | |
| 4407 | INTERNAL OPTION SETTING |
| 4408 | |
| 4409 | The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and |
| 4410 | PCRE_EXTENDED options (which are Perl-compatible) can be changed from |
| 4411 | within the pattern by a sequence of Perl option letters enclosed |
| 4412 | between "(?" and ")". The option letters are |
| 4413 | |
| 4414 | i for PCRE_CASELESS |
| 4415 | m for PCRE_MULTILINE |
| 4416 | s for PCRE_DOTALL |
| 4417 | x for PCRE_EXTENDED |
| 4418 | |
| 4419 | For example, (?im) sets caseless, multiline matching. It is also possi- |
| 4420 | ble to unset these options by preceding the letter with a hyphen, and a |
| 4421 | combined setting and unsetting such as (?im-sx), which sets PCRE_CASE- |
| 4422 | LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED, |
| 4423 | is also permitted. If a letter appears both before and after the |
| 4424 | hyphen, the option is unset. |
| 4425 | |
| 4426 | The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA |
| 4427 | can be changed in the same way as the Perl-compatible options by using |
| 4428 | the characters J, U and X respectively. |
| 4429 | |
| 4430 | When one of these option changes occurs at top level (that is, not |
| 4431 | inside subpattern parentheses), the change applies to the remainder of |
| 4432 | the pattern that follows. If the change is placed right at the start of |
| 4433 | a pattern, PCRE extracts it into the global options (and it will there- |
| 4434 | fore show up in data extracted by the pcre_fullinfo() function). |
| 4435 | |
| 4436 | An option change within a subpattern (see below for a description of |
| 4437 | subpatterns) affects only that part of the subpattern that follows it, |
| 4438 | so |
| 4439 | |
| 4440 | (a(?i)b)c |
| 4441 | |
| 4442 | matches abc and aBc and no other strings (assuming PCRE_CASELESS is not |
| 4443 | used). By this means, options can be made to have different settings |
| 4444 | in different parts of the pattern. Any changes made in one alternative |
| 4445 | do carry on into subsequent branches within the same subpattern. For |
| 4446 | example, |
| 4447 | |
| 4448 | (a(?i)b|c) |
| 4449 | |
| 4450 | matches "ab", "aB", "c", and "C", even though when matching "C" the |
| 4451 | first branch is abandoned before the option setting. This is because |
| 4452 | the effects of option settings happen at compile time. There would be |
| 4453 | some very weird behaviour otherwise. |
| 4454 | |
| 4455 | Note: There are other PCRE-specific options that can be set by the |
| 4456 | application when the compile or match functions are called. In some |
| 4457 | cases the pattern can contain special leading sequences such as (*CRLF) |
| 4458 | to override what the application has set or what has been defaulted. |
| 4459 | Details are given in the section entitled "Newline sequences" above. |
| 4460 | There are also the (*UTF8) and (*UCP) leading sequences that can be |
| 4461 | used to set UTF-8 and Unicode property modes; they are equivalent to |
| 4462 | setting the PCRE_UTF8 and the PCRE_UCP options, respectively. |
| 4463 | |
| 4464 | |
| 4465 | SUBPATTERNS |
| 4466 | |
| 4467 | Subpatterns are delimited by parentheses (round brackets), which can be |
| 4468 | nested. Turning part of a pattern into a subpattern does two things: |
| 4469 | |
| 4470 | 1. It localizes a set of alternatives. For example, the pattern |
| 4471 | |
| 4472 | cat(aract|erpillar|) |
| 4473 | |
| 4474 | matches "cataract", "caterpillar", or "cat". Without the parentheses, |
| 4475 | it would match "cataract", "erpillar" or an empty string. |
| 4476 | |
| 4477 | 2. It sets up the subpattern as a capturing subpattern. This means |
| 4478 | that, when the whole pattern matches, that portion of the subject |
| 4479 | string that matched the subpattern is passed back to the caller via the |
| 4480 | ovector argument of pcre_exec(). Opening parentheses are counted from |
| 4481 | left to right (starting from 1) to obtain numbers for the capturing |
| 4482 | subpatterns. For example, if the string "the red king" is matched |
| 4483 | against the pattern |
| 4484 | |
| 4485 | the ((red|white) (king|queen)) |
| 4486 | |
| 4487 | the captured substrings are "red king", "red", and "king", and are num- |
| 4488 | bered 1, 2, and 3, respectively. |
| 4489 | |
| 4490 | The fact that plain parentheses fulfil two functions is not always |
| 4491 | helpful. There are often times when a grouping subpattern is required |
| 4492 | without a capturing requirement. If an opening parenthesis is followed |
| 4493 | by a question mark and a colon, the subpattern does not do any captur- |
| 4494 | ing, and is not counted when computing the number of any subsequent |
| 4495 | capturing subpatterns. For example, if the string "the white queen" is |
| 4496 | matched against the pattern |
| 4497 | |
| 4498 | the ((?:red|white) (king|queen)) |
| 4499 | |
| 4500 | the captured substrings are "white queen" and "queen", and are numbered |
| 4501 | 1 and 2. The maximum number of capturing subpatterns is 65535. |
| 4502 | |
| 4503 | As a convenient shorthand, if any option settings are required at the |
| 4504 | start of a non-capturing subpattern, the option letters may appear |
| 4505 | between the "?" and the ":". Thus the two patterns |
| 4506 | |
| 4507 | (?i:saturday|sunday) |
| 4508 | (?:(?i)saturday|sunday) |
| 4509 | |
| 4510 | match exactly the same set of strings. Because alternative branches are |
| 4511 | tried from left to right, and options are not reset until the end of |
| 4512 | the subpattern is reached, an option setting in one branch does affect |
| 4513 | subsequent branches, so the above patterns match "SUNDAY" as well as |
| 4514 | "Saturday". |
| 4515 | |
| 4516 | |
| 4517 | DUPLICATE SUBPATTERN NUMBERS |
| 4518 | |
| 4519 | Perl 5.10 introduced a feature whereby each alternative in a subpattern |
| 4520 | uses the same numbers for its capturing parentheses. Such a subpattern |
| 4521 | starts with (?| and is itself a non-capturing subpattern. For example, |
| 4522 | consider this pattern: |
| 4523 | |
| 4524 | (?|(Sat)ur|(Sun))day |
| 4525 | |
| 4526 | Because the two alternatives are inside a (?| group, both sets of cap- |
| 4527 | turing parentheses are numbered one. Thus, when the pattern matches, |
| 4528 | you can look at captured substring number one, whichever alternative |
| 4529 | matched. This construct is useful when you want to capture part, but |
| 4530 | not all, of one of a number of alternatives. Inside a (?| group, paren- |
| 4531 | theses are numbered as usual, but the number is reset at the start of |
| 4532 | each branch. The numbers of any capturing parentheses that follow the |
| 4533 | subpattern start after the highest number used in any branch. The fol- |
| 4534 | lowing example is taken from the Perl documentation. The numbers under- |
| 4535 | neath show in which buffer the captured content will be stored. |
| 4536 | |
| 4537 | # before ---------------branch-reset----------- after |
| 4538 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x |
| 4539 | # 1 2 2 3 2 3 4 |
| 4540 | |
| 4541 | A back reference to a numbered subpattern uses the most recent value |
| 4542 | that is set for that number by any subpattern. The following pattern |
| 4543 | matches "abcabc" or "defdef": |
| 4544 | |
| 4545 | /(?|(abc)|(def))\1/ |
| 4546 | |
| 4547 | In contrast, a subroutine call to a numbered subpattern always refers |
| 4548 | to the first one in the pattern with the given number. The following |
| 4549 | pattern matches "abcabc" or "defabc": |
| 4550 | |
| 4551 | /(?|(abc)|(def))(?1)/ |
| 4552 | |
| 4553 | If a condition test for a subpattern's having matched refers to a non- |
| 4554 | unique number, the test is true if any of the subpatterns of that num- |
| 4555 | ber have matched. |
| 4556 | |
| 4557 | An alternative approach to using this "branch reset" feature is to use |
| 4558 | duplicate named subpatterns, as described in the next section. |
| 4559 | |
| 4560 | |
| 4561 | NAMED SUBPATTERNS |
| 4562 | |
| 4563 | Identifying capturing parentheses by number is simple, but it can be |
| 4564 | very hard to keep track of the numbers in complicated regular expres- |
| 4565 | sions. Furthermore, if an expression is modified, the numbers may |
| 4566 | change. To help with this difficulty, PCRE supports the naming of sub- |
| 4567 | patterns. This feature was not added to Perl until release 5.10. Python |
| 4568 | had the feature earlier, and PCRE introduced it at release 4.0, using |
| 4569 | the Python syntax. PCRE now supports both the Perl and the Python syn- |
| 4570 | tax. Perl allows identically numbered subpatterns to have different |
| 4571 | names, but PCRE does not. |
| 4572 | |
| 4573 | In PCRE, a subpattern can be named in one of three ways: (?<name>...) |
| 4574 | or (?'name'...) as in Perl, or (?P<name>...) as in Python. References |
| 4575 | to capturing parentheses from other parts of the pattern, such as back |
| 4576 | references, recursion, and conditions, can be made by name as well as |
| 4577 | by number. |
| 4578 | |
| 4579 | Names consist of up to 32 alphanumeric characters and underscores. |
| 4580 | Named capturing parentheses are still allocated numbers as well as |
| 4581 | names, exactly as if the names were not present. The PCRE API provides |
| 4582 | function calls for extracting the name-to-number translation table from |
| 4583 | a compiled pattern. There is also a convenience function for extracting |
| 4584 | a captured substring by name. |
| 4585 | |
| 4586 | By default, a name must be unique within a pattern, but it is possible |
| 4587 | to relax this constraint by setting the PCRE_DUPNAMES option at compile |
| 4588 | time. (Duplicate names are also always permitted for subpatterns with |
| 4589 | the same number, set up as described in the previous section.) Dupli- |
| 4590 | cate names can be useful for patterns where only one instance of the |
| 4591 | named parentheses can match. Suppose you want to match the name of a |
| 4592 | weekday, either as a 3-letter abbreviation or as the full name, and in |
| 4593 | both cases you want to extract the abbreviation. This pattern (ignoring |
| 4594 | the line breaks) does the job: |
| 4595 | |
| 4596 | (?<DN>Mon|Fri|Sun)(?:day)?| |
| 4597 | (?<DN>Tue)(?:sday)?| |
| 4598 | (?<DN>Wed)(?:nesday)?| |
| 4599 | (?<DN>Thu)(?:rsday)?| |
| 4600 | (?<DN>Sat)(?:urday)? |
| 4601 | |
| 4602 | There are five capturing substrings, but only one is ever set after a |
| 4603 | match. (An alternative way of solving this problem is to use a "branch |
| 4604 | reset" subpattern, as described in the previous section.) |
| 4605 | |
| 4606 | The convenience function for extracting the data by name returns the |
| 4607 | substring for the first (and in this example, the only) subpattern of |
| 4608 | that name that matched. This saves searching to find which numbered |
| 4609 | subpattern it was. |
| 4610 | |
| 4611 | If you make a back reference to a non-unique named subpattern from |
| 4612 | elsewhere in the pattern, the one that corresponds to the first occur- |
| 4613 | rence of the name is used. In the absence of duplicate numbers (see the |
| 4614 | previous section) this is the one with the lowest number. If you use a |
| 4615 | named reference in a condition test (see the section about conditions |
| 4616 | below), either to check whether a subpattern has matched, or to check |
| 4617 | for recursion, all subpatterns with the same name are tested. If the |
| 4618 | condition is true for any one of them, the overall condition is true. |
| 4619 | This is the same behaviour as testing by number. For further details of |
| 4620 | the interfaces for handling named subpatterns, see the pcreapi documen- |
| 4621 | tation. |
| 4622 | |
| 4623 | Warning: You cannot use different names to distinguish between two sub- |
| 4624 | patterns with the same number because PCRE uses only the numbers when |
| 4625 | matching. For this reason, an error is given at compile time if differ- |
| 4626 | ent names are given to subpatterns with the same number. However, you |
| 4627 | can give the same name to subpatterns with the same number, even when |
| 4628 | PCRE_DUPNAMES is not set. |
| 4629 | |
| 4630 | |
| 4631 | REPETITION |
| 4632 | |
| 4633 | Repetition is specified by quantifiers, which can follow any of the |
| 4634 | following items: |
| 4635 | |
| 4636 | a literal data character |
| 4637 | the dot metacharacter |
| 4638 | the \C escape sequence |
| 4639 | the \X escape sequence (in UTF-8 mode with Unicode properties) |
| 4640 | the \R escape sequence |
| 4641 | an escape such as \d or \pL that matches a single character |
| 4642 | a character class |
| 4643 | a back reference (see next section) |
| 4644 | a parenthesized subpattern (including assertions) |
| 4645 | a subroutine call to a subpattern (recursive or otherwise) |
| 4646 | |
| 4647 | The general repetition quantifier specifies a minimum and maximum num- |
| 4648 | ber of permitted matches, by giving the two numbers in curly brackets |
| 4649 | (braces), separated by a comma. The numbers must be less than 65536, |
| 4650 | and the first must be less than or equal to the second. For example: |
| 4651 | |
| 4652 | z{2,4} |
| 4653 | |
| 4654 | matches "zz", "zzz", or "zzzz". A closing brace on its own is not a |
| 4655 | special character. If the second number is omitted, but the comma is |
| 4656 | present, there is no upper limit; if the second number and the comma |
| 4657 | are both omitted, the quantifier specifies an exact number of required |
| 4658 | matches. Thus |
| 4659 | |
| 4660 | [aeiou]{3,} |
| 4661 | |
| 4662 | matches at least 3 successive vowels, but may match many more, while |
| 4663 | |
| 4664 | \d{8} |
| 4665 | |
| 4666 | matches exactly 8 digits. An opening curly bracket that appears in a |
| 4667 | position where a quantifier is not allowed, or one that does not match |
| 4668 | the syntax of a quantifier, is taken as a literal character. For exam- |
| 4669 | ple, {,6} is not a quantifier, but a literal string of four characters. |
| 4670 | |
| 4671 | In UTF-8 mode, quantifiers apply to UTF-8 characters rather than to |
| 4672 | individual bytes. Thus, for example, \x{100}{2} matches two UTF-8 char- |
| 4673 | acters, each of which is represented by a two-byte sequence. Similarly, |
| 4674 | when Unicode property support is available, \X{3} matches three Unicode |
| 4675 | extended sequences, each of which may be several bytes long (and they |
| 4676 | may be of different lengths). |
| 4677 | |
| 4678 | The quantifier {0} is permitted, causing the expression to behave as if |
| 4679 | the previous item and the quantifier were not present. This may be use- |
| 4680 | ful for subpatterns that are referenced as subroutines from elsewhere |
| 4681 | in the pattern (but see also the section entitled "Defining subpatterns |
| 4682 | for use by reference only" below). Items other than subpatterns that |
| 4683 | have a {0} quantifier are omitted from the compiled pattern. |
| 4684 | |
| 4685 | For convenience, the three most common quantifiers have single-charac- |
| 4686 | ter abbreviations: |
| 4687 | |
| 4688 | * is equivalent to {0,} |
| 4689 | + is equivalent to {1,} |
| 4690 | ? is equivalent to {0,1} |
| 4691 | |
| 4692 | It is possible to construct infinite loops by following a subpattern |
| 4693 | that can match no characters with a quantifier that has no upper limit, |
| 4694 | for example: |
| 4695 | |
| 4696 | (a?)* |
| 4697 | |
| 4698 | Earlier versions of Perl and PCRE used to give an error at compile time |
| 4699 | for such patterns. However, because there are cases where this can be |
| 4700 | useful, such patterns are now accepted, but if any repetition of the |
| 4701 | subpattern does in fact match no characters, the loop is forcibly bro- |
| 4702 | ken. |
| 4703 | |
| 4704 | By default, the quantifiers are "greedy", that is, they match as much |
| 4705 | as possible (up to the maximum number of permitted times), without |
| 4706 | causing the rest of the pattern to fail. The classic example of where |
| 4707 | this gives problems is in trying to match comments in C programs. These |
| 4708 | appear between /* and */ and within the comment, individual * and / |
| 4709 | characters may appear. An attempt to match C comments by applying the |
| 4710 | pattern |
| 4711 | |
| 4712 | /\*.*\*/ |
| 4713 | |
| 4714 | to the string |
| 4715 | |
| 4716 | /* first comment */ not comment /* second comment */ |
| 4717 | |
| 4718 | fails, because it matches the entire string owing to the greediness of |
| 4719 | the .* item. |
| 4720 | |
| 4721 | However, if a quantifier is followed by a question mark, it ceases to |
| 4722 | be greedy, and instead matches the minimum number of times possible, so |
| 4723 | the pattern |
| 4724 | |
| 4725 | /\*.*?\*/ |
| 4726 | |
| 4727 | does the right thing with the C comments. The meaning of the various |
| 4728 | quantifiers is not otherwise changed, just the preferred number of |
| 4729 | matches. Do not confuse this use of question mark with its use as a |
| 4730 | quantifier in its own right. Because it has two uses, it can sometimes |
| 4731 | appear doubled, as in |
| 4732 | |
| 4733 | \d??\d |
| 4734 | |
| 4735 | which matches one digit by preference, but can match two if that is the |
| 4736 | only way the rest of the pattern matches. |
| 4737 | |
| 4738 | If the PCRE_UNGREEDY option is set (an option that is not available in |
| 4739 | Perl), the quantifiers are not greedy by default, but individual ones |
| 4740 | can be made greedy by following them with a question mark. In other |
| 4741 | words, it inverts the default behaviour. |
| 4742 | |
| 4743 | When a parenthesized subpattern is quantified with a minimum repeat |
| 4744 | count that is greater than 1 or with a limited maximum, more memory is |
| 4745 | required for the compiled pattern, in proportion to the size of the |
| 4746 | minimum or maximum. |
| 4747 | |
| 4748 | If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv- |
| 4749 | alent to Perl's /s) is set, thus allowing the dot to match newlines, |
| 4750 | the pattern is implicitly anchored, because whatever follows will be |
| 4751 | tried against every character position in the subject string, so there |
| 4752 | is no point in retrying the overall match at any position after the |
| 4753 | first. PCRE normally treats such a pattern as though it were preceded |
| 4754 | by \A. |
| 4755 | |
| 4756 | In cases where it is known that the subject string contains no new- |
| 4757 | lines, it is worth setting PCRE_DOTALL in order to obtain this opti- |
| 4758 | mization, or alternatively using ^ to indicate anchoring explicitly. |
| 4759 | |
| 4760 | However, there is one situation where the optimization cannot be used. |
| 4761 | When .* is inside capturing parentheses that are the subject of a back |
| 4762 | reference elsewhere in the pattern, a match at the start may fail where |
| 4763 | a later one succeeds. Consider, for example: |
| 4764 | |
| 4765 | (.*)abc\1 |
| 4766 | |
| 4767 | If the subject is "xyz123abc123" the match point is the fourth charac- |
| 4768 | ter. For this reason, such a pattern is not implicitly anchored. |
| 4769 | |
| 4770 | When a capturing subpattern is repeated, the value captured is the sub- |
| 4771 | string that matched the final iteration. For example, after |
| 4772 | |
| 4773 | (tweedle[dume]{3}\s*)+ |
| 4774 | |
| 4775 | has matched "tweedledum tweedledee" the value of the captured substring |
| 4776 | is "tweedledee". However, if there are nested capturing subpatterns, |
| 4777 | the corresponding captured values may have been set in previous itera- |
| 4778 | tions. For example, after |
| 4779 | |
| 4780 | /(a|(b))+/ |
| 4781 | |
| 4782 | matches "aba" the value of the second captured substring is "b". |
| 4783 | |
| 4784 | |
| 4785 | ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS |
| 4786 | |
| 4787 | With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy") |
| 4788 | repetition, failure of what follows normally causes the repeated item |
| 4789 | to be re-evaluated to see if a different number of repeats allows the |
| 4790 | rest of the pattern to match. Sometimes it is useful to prevent this, |
| 4791 | either to change the nature of the match, or to cause it fail earlier |
| 4792 | than it otherwise might, when the author of the pattern knows there is |
| 4793 | no point in carrying on. |
| 4794 | |
| 4795 | Consider, for example, the pattern \d+foo when applied to the subject |
| 4796 | line |
| 4797 | |
| 4798 | 123456bar |
| 4799 | |
| 4800 | After matching all 6 digits and then failing to match "foo", the normal |
| 4801 | action of the matcher is to try again with only 5 digits matching the |
| 4802 | \d+ item, and then with 4, and so on, before ultimately failing. |
| 4803 | "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides |
| 4804 | the means for specifying that once a subpattern has matched, it is not |
| 4805 | to be re-evaluated in this way. |
| 4806 | |
| 4807 | If we use atomic grouping for the previous example, the matcher gives |
| 4808 | up immediately on failing to match "foo" the first time. The notation |
| 4809 | is a kind of special parenthesis, starting with (?> as in this example: |
| 4810 | |
| 4811 | (?>\d+)foo |
| 4812 | |
| 4813 | This kind of parenthesis "locks up" the part of the pattern it con- |
| 4814 | tains once it has matched, and a failure further into the pattern is |
| 4815 | prevented from backtracking into it. Backtracking past it to previous |
| 4816 | items, however, works as normal. |
| 4817 | |
| 4818 | An alternative description is that a subpattern of this type matches |
| 4819 | the string of characters that an identical standalone pattern would |
| 4820 | match, if anchored at the current point in the subject string. |
| 4821 | |
| 4822 | Atomic grouping subpatterns are not capturing subpatterns. Simple cases |
| 4823 | such as the above example can be thought of as a maximizing repeat that |
| 4824 | must swallow everything it can. So, while both \d+ and \d+? are pre- |
| 4825 | pared to adjust the number of digits they match in order to make the |
| 4826 | rest of the pattern match, (?>\d+) can only match an entire sequence of |
| 4827 | digits. |
| 4828 | |
| 4829 | Atomic groups in general can of course contain arbitrarily complicated |
| 4830 | subpatterns, and can be nested. However, when the subpattern for an |
| 4831 | atomic group is just a single repeated item, as in the example above, a |
| 4832 | simpler notation, called a "possessive quantifier" can be used. This |
| 4833 | consists of an additional + character following a quantifier. Using |
| 4834 | this notation, the previous example can be rewritten as |
| 4835 | |
| 4836 | \d++foo |
| 4837 | |
| 4838 | Note that a possessive quantifier can be used with an entire group, for |
| 4839 | example: |
| 4840 | |
| 4841 | (abc|xyz){2,3}+ |
| 4842 | |
| 4843 | Possessive quantifiers are always greedy; the setting of the |
| 4844 | PCRE_UNGREEDY option is ignored. They are a convenient notation for the |
| 4845 | simpler forms of atomic group. However, there is no difference in the |
| 4846 | meaning of a possessive quantifier and the equivalent atomic group, |
| 4847 | though there may be a performance difference; possessive quantifiers |
| 4848 | should be slightly faster. |
| 4849 | |
| 4850 | The possessive quantifier syntax is an extension to the Perl 5.8 syn- |
| 4851 | tax. Jeffrey Friedl originated the idea (and the name) in the first |
| 4852 | edition of his book. Mike McCloskey liked it, so implemented it when he |
| 4853 | built Sun's Java package, and PCRE copied it from there. It ultimately |
| 4854 | found its way into Perl at release 5.10. |
| 4855 | |
| 4856 | PCRE has an optimization that automatically "possessifies" certain sim- |
| 4857 | ple pattern constructs. For example, the sequence A+B is treated as |
| 4858 | A++B because there is no point in backtracking into a sequence of A's |
| 4859 | when B must follow. |
| 4860 | |
| 4861 | When a pattern contains an unlimited repeat inside a subpattern that |
| 4862 | can itself be repeated an unlimited number of times, the use of an |
| 4863 | atomic group is the only way to avoid some failing matches taking a |
| 4864 | very long time indeed. The pattern |
| 4865 | |
| 4866 | (\D+|<\d+>)*[!?] |
| 4867 | |
| 4868 | matches an unlimited number of substrings that either consist of non- |
| 4869 | digits, or digits enclosed in <>, followed by either ! or ?. When it |
| 4870 | matches, it runs quickly. However, if it is applied to |
| 4871 | |
| 4872 | aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa |
| 4873 | |
| 4874 | it takes a long time before reporting failure. This is because the |
| 4875 | string can be divided between the internal \D+ repeat and the external |
| 4876 | * repeat in a large number of ways, and all have to be tried. (The |
| 4877 | example uses [!?] rather than a single character at the end, because |
| 4878 | both PCRE and Perl have an optimization that allows for fast failure |
| 4879 | when a single character is used. They remember the last single charac- |
| 4880 | ter that is required for a match, and fail early if it is not present |
| 4881 | in the string.) If the pattern is changed so that it uses an atomic |
| 4882 | group, like this: |
| 4883 | |
| 4884 | ((?>\D+)|<\d+>)*[!?] |
| 4885 | |
| 4886 | sequences of non-digits cannot be broken, and failure happens quickly. |
| 4887 | |
| 4888 | |
| 4889 | BACK REFERENCES |
| 4890 | |
| 4891 | Outside a character class, a backslash followed by a digit greater than |
| 4892 | 0 (and possibly further digits) is a back reference to a capturing sub- |
| 4893 | pattern earlier (that is, to its left) in the pattern, provided there |
| 4894 | have been that many previous capturing left parentheses. |
| 4895 | |
| 4896 | However, if the decimal number following the backslash is less than 10, |
| 4897 | it is always taken as a back reference, and causes an error only if |
| 4898 | there are not that many capturing left parentheses in the entire pat- |
| 4899 | tern. In other words, the parentheses that are referenced need not be |
| 4900 | to the left of the reference for numbers less than 10. A "forward back |
| 4901 | reference" of this type can make sense when a repetition is involved |
| 4902 | and the subpattern to the right has participated in an earlier itera- |
| 4903 | tion. |
| 4904 | |
| 4905 | It is not possible to have a numerical "forward back reference" to a |
| 4906 | subpattern whose number is 10 or more using this syntax because a |
| 4907 | sequence such as \50 is interpreted as a character defined in octal. |
| 4908 | See the subsection entitled "Non-printing characters" above for further |
| 4909 | details of the handling of digits following a backslash. There is no |
| 4910 | such problem when named parentheses are used. A back reference to any |
| 4911 | subpattern is possible using named parentheses (see below). |
| 4912 | |
| 4913 | Another way of avoiding the ambiguity inherent in the use of digits |
| 4914 | following a backslash is to use the \g escape sequence. This escape |
| 4915 | must be followed by an unsigned number or a negative number, optionally |
| 4916 | enclosed in braces. These examples are all identical: |
| 4917 | |
| 4918 | (ring), \1 |
| 4919 | (ring), \g1 |
| 4920 | (ring), \g{1} |
| 4921 | |
| 4922 | An unsigned number specifies an absolute reference without the ambigu- |
| 4923 | ity that is present in the older syntax. It is also useful when literal |
| 4924 | digits follow the reference. A negative number is a relative reference. |
| 4925 | Consider this example: |
| 4926 | |
| 4927 | (abc(def)ghi)\g{-1} |
| 4928 | |
| 4929 | The sequence \g{-1} is a reference to the most recently started captur- |
| 4930 | ing subpattern before \g, that is, is it equivalent to \2 in this exam- |
| 4931 | ple. Similarly, \g{-2} would be equivalent to \1. The use of relative |
| 4932 | references can be helpful in long patterns, and also in patterns that |
| 4933 | are created by joining together fragments that contain references |
| 4934 | within themselves. |
| 4935 | |
| 4936 | A back reference matches whatever actually matched the capturing sub- |
| 4937 | pattern in the current subject string, rather than anything matching |
| 4938 | the subpattern itself (see "Subpatterns as subroutines" below for a way |
| 4939 | of doing that). So the pattern |
| 4940 | |
| 4941 | (sens|respons)e and \1ibility |
| 4942 | |
| 4943 | matches "sense and sensibility" and "response and responsibility", but |
| 4944 | not "sense and responsibility". If caseful matching is in force at the |
| 4945 | time of the back reference, the case of letters is relevant. For exam- |
| 4946 | ple, |
| 4947 | |
| 4948 | ((?i)rah)\s+\1 |
| 4949 | |
| 4950 | matches "rah rah" and "RAH RAH", but not "RAH rah", even though the |
| 4951 | original capturing subpattern is matched caselessly. |
| 4952 | |
| 4953 | There are several different ways of writing back references to named |
| 4954 | subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or |
| 4955 | \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's |
| 4956 | unified back reference syntax, in which \g can be used for both numeric |
| 4957 | and named references, is also supported. We could rewrite the above |
| 4958 | example in any of the following ways: |
| 4959 | |
| 4960 | (?<p1>(?i)rah)\s+\k<p1> |
| 4961 | (?'p1'(?i)rah)\s+\k{p1} |
| 4962 | (?P<p1>(?i)rah)\s+(?P=p1) |
| 4963 | (?<p1>(?i)rah)\s+\g{p1} |
| 4964 | |
| 4965 | A subpattern that is referenced by name may appear in the pattern |
| 4966 | before or after the reference. |
| 4967 | |
| 4968 | There may be more than one back reference to the same subpattern. If a |
| 4969 | subpattern has not actually been used in a particular match, any back |
| 4970 | references to it always fail by default. For example, the pattern |
| 4971 | |
| 4972 | (a|(bc))\2 |
| 4973 | |
| 4974 | always fails if it starts to match "a" rather than "bc". However, if |
| 4975 | the PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back refer- |
| 4976 | ence to an unset value matches an empty string. |
| 4977 | |
| 4978 | Because there may be many capturing parentheses in a pattern, all dig- |
| 4979 | its following a backslash are taken as part of a potential back refer- |
| 4980 | ence number. If the pattern continues with a digit character, some |
| 4981 | delimiter must be used to terminate the back reference. If the |
| 4982 | PCRE_EXTENDED option is set, this can be whitespace. Otherwise, the \g{ |
| 4983 | syntax or an empty comment (see "Comments" below) can be used. |
| 4984 | |
| 4985 | Recursive back references |
| 4986 | |
| 4987 | A back reference that occurs inside the parentheses to which it refers |
| 4988 | fails when the subpattern is first used, so, for example, (a\1) never |
| 4989 | matches. However, such references can be useful inside repeated sub- |
| 4990 | patterns. For example, the pattern |
| 4991 | |
| 4992 | (a|b\1)+ |
| 4993 | |
| 4994 | matches any number of "a"s and also "aba", "ababbaa" etc. At each iter- |
| 4995 | ation of the subpattern, the back reference matches the character |
| 4996 | string corresponding to the previous iteration. In order for this to |
| 4997 | work, the pattern must be such that the first iteration does not need |
| 4998 | to match the back reference. This can be done using alternation, as in |
| 4999 | the example above, or by a quantifier with a minimum of zero. |
| 5000 | |
| 5001 | Back references of this type cause the group that they reference to be |
| 5002 | treated as an atomic group. Once the whole group has been matched, a |
| 5003 | subsequent matching failure cannot cause backtracking into the middle |
| 5004 | of the group. |
| 5005 | |
| 5006 | |
| 5007 | ASSERTIONS |
| 5008 | |
| 5009 | An assertion is a test on the characters following or preceding the |
| 5010 | current matching point that does not actually consume any characters. |
| 5011 | The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are |
| 5012 | described above. |
| 5013 | |
| 5014 | More complicated assertions are coded as subpatterns. There are two |
| 5015 | kinds: those that look ahead of the current position in the subject |
| 5016 | string, and those that look behind it. An assertion subpattern is |
| 5017 | matched in the normal way, except that it does not cause the current |
| 5018 | matching position to be changed. |
| 5019 | |
| 5020 | Assertion subpatterns are not capturing subpatterns. If such an asser- |
| 5021 | tion contains capturing subpatterns within it, these are counted for |
| 5022 | the purposes of numbering the capturing subpatterns in the whole pat- |
| 5023 | tern. However, substring capturing is carried out only for positive |
| 5024 | assertions, because it does not make sense for negative assertions. |
| 5025 | |
| 5026 | For compatibility with Perl, assertion subpatterns may be repeated; |
| 5027 | though it makes no sense to assert the same thing several times, the |
| 5028 | side effect of capturing parentheses may occasionally be useful. In |
| 5029 | practice, there only three cases: |
| 5030 | |
| 5031 | (1) If the quantifier is {0}, the assertion is never obeyed during |
| 5032 | matching. However, it may contain internal capturing parenthesized |
| 5033 | groups that are called from elsewhere via the subroutine mechanism. |
| 5034 | |
| 5035 | (2) If quantifier is {0,n} where n is greater than zero, it is treated |
| 5036 | as if it were {0,1}. At run time, the rest of the pattern match is |
| 5037 | tried with and without the assertion, the order depending on the greed- |
| 5038 | iness of the quantifier. |
| 5039 | |
| 5040 | (3) If the minimum repetition is greater than zero, the quantifier is |
| 5041 | ignored. The assertion is obeyed just once when encountered during |
| 5042 | matching. |
| 5043 | |
| 5044 | Lookahead assertions |
| 5045 | |
| 5046 | Lookahead assertions start with (?= for positive assertions and (?! for |
| 5047 | negative assertions. For example, |
| 5048 | |
| 5049 | \w+(?=;) |
| 5050 | |
| 5051 | matches a word followed by a semicolon, but does not include the semi- |
| 5052 | colon in the match, and |
| 5053 | |
| 5054 | foo(?!bar) |
| 5055 | |
| 5056 | matches any occurrence of "foo" that is not followed by "bar". Note |
| 5057 | that the apparently similar pattern |
| 5058 | |
| 5059 | (?!foo)bar |
| 5060 | |
| 5061 | does not find an occurrence of "bar" that is preceded by something |
| 5062 | other than "foo"; it finds any occurrence of "bar" whatsoever, because |
| 5063 | the assertion (?!foo) is always true when the next three characters are |
| 5064 | "bar". A lookbehind assertion is needed to achieve the other effect. |
| 5065 | |
| 5066 | If you want to force a matching failure at some point in a pattern, the |
| 5067 | most convenient way to do it is with (?!) because an empty string |
| 5068 | always matches, so an assertion that requires there not to be an empty |
| 5069 | string must always fail. The backtracking control verb (*FAIL) or (*F) |
| 5070 | is a synonym for (?!). |
| 5071 | |
| 5072 | Lookbehind assertions |
| 5073 | |
| 5074 | Lookbehind assertions start with (?<= for positive assertions and (?<! |
| 5075 | for negative assertions. For example, |
| 5076 | |
| 5077 | (?<!foo)bar |
| 5078 | |
| 5079 | does find an occurrence of "bar" that is not preceded by "foo". The |
| 5080 | contents of a lookbehind assertion are restricted such that all the |
| 5081 | strings it matches must have a fixed length. However, if there are sev- |
| 5082 | eral top-level alternatives, they do not all have to have the same |
| 5083 | fixed length. Thus |
| 5084 | |
| 5085 | (?<=bullock|donkey) |
| 5086 | |
| 5087 | is permitted, but |
| 5088 | |
| 5089 | (?<!dogs?|cats?) |
| 5090 | |
| 5091 | causes an error at compile time. Branches that match different length |
| 5092 | strings are permitted only at the top level of a lookbehind assertion. |
| 5093 | This is an extension compared with Perl, which requires all branches to |
| 5094 | match the same length of string. An assertion such as |
| 5095 | |
| 5096 | (?<=ab(c|de)) |
| 5097 | |
| 5098 | is not permitted, because its single top-level branch can match two |
| 5099 | different lengths, but it is acceptable to PCRE if rewritten to use two |
| 5100 | top-level branches: |
| 5101 | |
| 5102 | (?<=abc|abde) |
| 5103 | |
| 5104 | In some cases, the escape sequence \K (see above) can be used instead |
| 5105 | of a lookbehind assertion to get round the fixed-length restriction. |
| 5106 | |
| 5107 | The implementation of lookbehind assertions is, for each alternative, |
| 5108 | to temporarily move the current position back by the fixed length and |
| 5109 | then try to match. If there are insufficient characters before the cur- |
| 5110 | rent position, the assertion fails. |
| 5111 | |
| 5112 | In UTF-8 mode, PCRE does not allow the \C escape (which matches a sin- |
| 5113 | gle byte, even in UTF-8 mode) to appear in lookbehind assertions, |
| 5114 | because it makes it impossible to calculate the length of the lookbe- |
| 5115 | hind. The \X and \R escapes, which can match different numbers of |
| 5116 | bytes, are also not permitted. |
| 5117 | |
| 5118 | "Subroutine" calls (see below) such as (?2) or (?&X) are permitted in |
| 5119 | lookbehinds, as long as the subpattern matches a fixed-length string. |
| 5120 | Recursion, however, is not supported. |
| 5121 | |
| 5122 | Possessive quantifiers can be used in conjunction with lookbehind |
| 5123 | assertions to specify efficient matching of fixed-length strings at the |
| 5124 | end of subject strings. Consider a simple pattern such as |
| 5125 | |
| 5126 | abcd$ |
| 5127 | |
| 5128 | when applied to a long string that does not match. Because matching |
| 5129 | proceeds from left to right, PCRE will look for each "a" in the subject |
| 5130 | and then see if what follows matches the rest of the pattern. If the |
| 5131 | pattern is specified as |
| 5132 | |
| 5133 | ^.*abcd$ |
| 5134 | |
| 5135 | the initial .* matches the entire string at first, but when this fails |
| 5136 | (because there is no following "a"), it backtracks to match all but the |
| 5137 | last character, then all but the last two characters, and so on. Once |
| 5138 | again the search for "a" covers the entire string, from right to left, |
| 5139 | so we are no better off. However, if the pattern is written as |
| 5140 | |
| 5141 | ^.*+(?<=abcd) |
| 5142 | |
| 5143 | there can be no backtracking for the .*+ item; it can match only the |
| 5144 | entire string. The subsequent lookbehind assertion does a single test |
| 5145 | on the last four characters. If it fails, the match fails immediately. |
| 5146 | For long strings, this approach makes a significant difference to the |
| 5147 | processing time. |
| 5148 | |
| 5149 | Using multiple assertions |
| 5150 | |
| 5151 | Several assertions (of any sort) may occur in succession. For example, |
| 5152 | |
| 5153 | (?<=\d{3})(?<!999)foo |
| 5154 | |
| 5155 | matches "foo" preceded by three digits that are not "999". Notice that |
| 5156 | each of the assertions is applied independently at the same point in |
| 5157 | the subject string. First there is a check that the previous three |
| 5158 | characters are all digits, and then there is a check that the same |
| 5159 | three characters are not "999". This pattern does not match "foo" pre- |
| 5160 | ceded by six characters, the first of which are digits and the last |
| 5161 | three of which are not "999". For example, it doesn't match "123abc- |
| 5162 | foo". A pattern to do that is |
| 5163 | |
| 5164 | (?<=\d{3}...)(?<!999)foo |
| 5165 | |
| 5166 | This time the first assertion looks at the preceding six characters, |
| 5167 | checking that the first three are digits, and then the second assertion |
| 5168 | checks that the preceding three characters are not "999". |
| 5169 | |
| 5170 | Assertions can be nested in any combination. For example, |
| 5171 | |
| 5172 | (?<=(?<!foo)bar)baz |
| 5173 | |
| 5174 | matches an occurrence of "baz" that is preceded by "bar" which in turn |
| 5175 | is not preceded by "foo", while |
| 5176 | |
| 5177 | (?<=\d{3}(?!999)...)foo |
| 5178 | |
| 5179 | is another pattern that matches "foo" preceded by three digits and any |
| 5180 | three characters that are not "999". |
| 5181 | |
| 5182 | |
| 5183 | CONDITIONAL SUBPATTERNS |
| 5184 | |
| 5185 | It is possible to cause the matching process to obey a subpattern con- |
| 5186 | ditionally or to choose between two alternative subpatterns, depending |
| 5187 | on the result of an assertion, or whether a specific capturing subpat- |
| 5188 | tern has already been matched. The two possible forms of conditional |
| 5189 | subpattern are: |
| 5190 | |
| 5191 | (?(condition)yes-pattern) |
| 5192 | (?(condition)yes-pattern|no-pattern) |
| 5193 | |
| 5194 | If the condition is satisfied, the yes-pattern is used; otherwise the |
| 5195 | no-pattern (if present) is used. If there are more than two alterna- |
| 5196 | tives in the subpattern, a compile-time error occurs. Each of the two |
| 5197 | alternatives may itself contain nested subpatterns of any form, includ- |
| 5198 | ing conditional subpatterns; the restriction to two alternatives |
| 5199 | applies only at the level of the condition. This pattern fragment is an |
| 5200 | example where the alternatives are complex: |
| 5201 | |
| 5202 | (?(1) (A|B|C) | (D | (?(2)E|F) | E) ) |
| 5203 | |
| 5204 | |
| 5205 | There are four kinds of condition: references to subpatterns, refer- |
| 5206 | ences to recursion, a pseudo-condition called DEFINE, and assertions. |
| 5207 | |
| 5208 | Checking for a used subpattern by number |
| 5209 | |
| 5210 | If the text between the parentheses consists of a sequence of digits, |
| 5211 | the condition is true if a capturing subpattern of that number has pre- |
| 5212 | viously matched. If there is more than one capturing subpattern with |
| 5213 | the same number (see the earlier section about duplicate subpattern |
| 5214 | numbers), the condition is true if any of them have matched. An alter- |
| 5215 | native notation is to precede the digits with a plus or minus sign. In |
| 5216 | this case, the subpattern number is relative rather than absolute. The |
| 5217 | most recently opened parentheses can be referenced by (?(-1), the next |
| 5218 | most recent by (?(-2), and so on. Inside loops it can also make sense |
| 5219 | to refer to subsequent groups. The next parentheses to be opened can be |
| 5220 | referenced as (?(+1), and so on. (The value zero in any of these forms |
| 5221 | is not used; it provokes a compile-time error.) |
| 5222 | |
| 5223 | Consider the following pattern, which contains non-significant white |
| 5224 | space to make it more readable (assume the PCRE_EXTENDED option) and to |
| 5225 | divide it into three parts for ease of discussion: |
| 5226 | |
| 5227 | ( \( )? [^()]+ (?(1) \) ) |
| 5228 | |
| 5229 | The first part matches an optional opening parenthesis, and if that |
| 5230 | character is present, sets it as the first captured substring. The sec- |
| 5231 | ond part matches one or more characters that are not parentheses. The |
| 5232 | third part is a conditional subpattern that tests whether or not the |
| 5233 | first set of parentheses matched. If they did, that is, if subject |
| 5234 | started with an opening parenthesis, the condition is true, and so the |
| 5235 | yes-pattern is executed and a closing parenthesis is required. Other- |
| 5236 | wise, since no-pattern is not present, the subpattern matches nothing. |
| 5237 | In other words, this pattern matches a sequence of non-parentheses, |
| 5238 | optionally enclosed in parentheses. |
| 5239 | |
| 5240 | If you were embedding this pattern in a larger one, you could use a |
| 5241 | relative reference: |
| 5242 | |
| 5243 | ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ... |
| 5244 | |
| 5245 | This makes the fragment independent of the parentheses in the larger |
| 5246 | pattern. |
| 5247 | |
| 5248 | Checking for a used subpattern by name |
| 5249 | |
| 5250 | Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a |
| 5251 | used subpattern by name. For compatibility with earlier versions of |
| 5252 | PCRE, which had this facility before Perl, the syntax (?(name)...) is |
| 5253 | also recognized. However, there is a possible ambiguity with this syn- |
| 5254 | tax, because subpattern names may consist entirely of digits. PCRE |
| 5255 | looks first for a named subpattern; if it cannot find one and the name |
| 5256 | consists entirely of digits, PCRE looks for a subpattern of that num- |
| 5257 | ber, which must be greater than zero. Using subpattern names that con- |
| 5258 | sist entirely of digits is not recommended. |
| 5259 | |
| 5260 | Rewriting the above example to use a named subpattern gives this: |
| 5261 | |
| 5262 | (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) ) |
| 5263 | |
| 5264 | If the name used in a condition of this kind is a duplicate, the test |
| 5265 | is applied to all subpatterns of the same name, and is true if any one |
| 5266 | of them has matched. |
| 5267 | |
| 5268 | Checking for pattern recursion |
| 5269 | |
| 5270 | If the condition is the string (R), and there is no subpattern with the |
| 5271 | name R, the condition is true if a recursive call to the whole pattern |
| 5272 | or any subpattern has been made. If digits or a name preceded by amper- |
| 5273 | sand follow the letter R, for example: |
| 5274 | |
| 5275 | (?(R3)...) or (?(R&name)...) |
| 5276 | |
| 5277 | the condition is true if the most recent recursion is into a subpattern |
| 5278 | whose number or name is given. This condition does not check the entire |
| 5279 | recursion stack. If the name used in a condition of this kind is a |
| 5280 | duplicate, the test is applied to all subpatterns of the same name, and |
| 5281 | is true if any one of them is the most recent recursion. |
| 5282 | |
| 5283 | At "top level", all these recursion test conditions are false. The |
| 5284 | syntax for recursive patterns is described below. |
| 5285 | |
| 5286 | Defining subpatterns for use by reference only |
| 5287 | |
| 5288 | If the condition is the string (DEFINE), and there is no subpattern |
| 5289 | with the name DEFINE, the condition is always false. In this case, |
| 5290 | there may be only one alternative in the subpattern. It is always |
| 5291 | skipped if control reaches this point in the pattern; the idea of |
| 5292 | DEFINE is that it can be used to define subroutines that can be refer- |
| 5293 | enced from elsewhere. (The use of subroutines is described below.) For |
| 5294 | example, a pattern to match an IPv4 address such as "192.168.23.245" |
| 5295 | could be written like this (ignore whitespace and line breaks): |
| 5296 | |
| 5297 | (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) ) |
| 5298 | \b (?&byte) (\.(?&byte)){3} \b |
| 5299 | |
| 5300 | The first part of the pattern is a DEFINE group inside which a another |
| 5301 | group named "byte" is defined. This matches an individual component of |
| 5302 | an IPv4 address (a number less than 256). When matching takes place, |
| 5303 | this part of the pattern is skipped because DEFINE acts like a false |
| 5304 | condition. The rest of the pattern uses references to the named group |
| 5305 | to match the four dot-separated components of an IPv4 address, insist- |
| 5306 | ing on a word boundary at each end. |
| 5307 | |
| 5308 | Assertion conditions |
| 5309 | |
| 5310 | If the condition is not in any of the above formats, it must be an |
| 5311 | assertion. This may be a positive or negative lookahead or lookbehind |
| 5312 | assertion. Consider this pattern, again containing non-significant |
| 5313 | white space, and with the two alternatives on the second line: |
| 5314 | |
| 5315 | (?(?=[^a-z]*[a-z]) |
| 5316 | \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} ) |
| 5317 | |
| 5318 | The condition is a positive lookahead assertion that matches an |
| 5319 | optional sequence of non-letters followed by a letter. In other words, |
| 5320 | it tests for the presence of at least one letter in the subject. If a |
| 5321 | letter is found, the subject is matched against the first alternative; |
| 5322 | otherwise it is matched against the second. This pattern matches |
| 5323 | strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are |
| 5324 | letters and dd are digits. |
| 5325 | |
| 5326 | |
| 5327 | COMMENTS |
| 5328 | |
| 5329 | There are two ways of including comments in patterns that are processed |
| 5330 | by PCRE. In both cases, the start of the comment must not be in a char- |
| 5331 | acter class, nor in the middle of any other sequence of related charac- |
| 5332 | ters such as (?: or a subpattern name or number. The characters that |
| 5333 | make up a comment play no part in the pattern matching. |
| 5334 | |
| 5335 | The sequence (?# marks the start of a comment that continues up to the |
| 5336 | next closing parenthesis. Nested parentheses are not permitted. If the |
| 5337 | PCRE_EXTENDED option is set, an unescaped # character also introduces a |
| 5338 | comment, which in this case continues to immediately after the next |
| 5339 | newline character or character sequence in the pattern. Which charac- |
| 5340 | ters are interpreted as newlines is controlled by the options passed to |
| 5341 | pcre_compile() or by a special sequence at the start of the pattern, as |
| 5342 | described in the section entitled "Newline conventions" above. Note |
| 5343 | that the end of this type of comment is a literal newline sequence in |
| 5344 | the pattern; escape sequences that happen to represent a newline do not |
| 5345 | count. For example, consider this pattern when PCRE_EXTENDED is set, |
| 5346 | and the default newline convention is in force: |
| 5347 | |
| 5348 | abc #comment \n still comment |
| 5349 | |
| 5350 | On encountering the # character, pcre_compile() skips along, looking |
| 5351 | for a newline in the pattern. The sequence \n is still literal at this |
| 5352 | stage, so it does not terminate the comment. Only an actual character |
| 5353 | with the code value 0x0a (the default newline) does so. |
| 5354 | |
| 5355 | |
| 5356 | RECURSIVE PATTERNS |
| 5357 | |
| 5358 | Consider the problem of matching a string in parentheses, allowing for |
| 5359 | unlimited nested parentheses. Without the use of recursion, the best |
| 5360 | that can be done is to use a pattern that matches up to some fixed |
| 5361 | depth of nesting. It is not possible to handle an arbitrary nesting |
| 5362 | depth. |
| 5363 | |
| 5364 | For some time, Perl has provided a facility that allows regular expres- |
| 5365 | sions to recurse (amongst other things). It does this by interpolating |
| 5366 | Perl code in the expression at run time, and the code can refer to the |
| 5367 | expression itself. A Perl pattern using code interpolation to solve the |
| 5368 | parentheses problem can be created like this: |
| 5369 | |
| 5370 | $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x; |
| 5371 | |
| 5372 | The (?p{...}) item interpolates Perl code at run time, and in this case |
| 5373 | refers recursively to the pattern in which it appears. |
| 5374 | |
| 5375 | Obviously, PCRE cannot support the interpolation of Perl code. Instead, |
| 5376 | it supports special syntax for recursion of the entire pattern, and |
| 5377 | also for individual subpattern recursion. After its introduction in |
| 5378 | PCRE and Python, this kind of recursion was subsequently introduced |
| 5379 | into Perl at release 5.10. |
| 5380 | |
| 5381 | A special item that consists of (? followed by a number greater than |
| 5382 | zero and a closing parenthesis is a recursive subroutine call of the |
| 5383 | subpattern of the given number, provided that it occurs inside that |
| 5384 | subpattern. (If not, it is a non-recursive subroutine call, which is |
| 5385 | described in the next section.) The special item (?R) or (?0) is a |
| 5386 | recursive call of the entire regular expression. |
| 5387 | |
| 5388 | This PCRE pattern solves the nested parentheses problem (assume the |
| 5389 | PCRE_EXTENDED option is set so that white space is ignored): |
| 5390 | |
| 5391 | \( ( [^()]++ | (?R) )* \) |
| 5392 | |
| 5393 | First it matches an opening parenthesis. Then it matches any number of |
| 5394 | substrings which can either be a sequence of non-parentheses, or a |
| 5395 | recursive match of the pattern itself (that is, a correctly parenthe- |
| 5396 | sized substring). Finally there is a closing parenthesis. Note the use |
| 5397 | of a possessive quantifier to avoid backtracking into sequences of non- |
| 5398 | parentheses. |
| 5399 | |
| 5400 | If this were part of a larger pattern, you would not want to recurse |
| 5401 | the entire pattern, so instead you could use this: |
| 5402 | |
| 5403 | ( \( ( [^()]++ | (?1) )* \) ) |
| 5404 | |
| 5405 | We have put the pattern into parentheses, and caused the recursion to |
| 5406 | refer to them instead of the whole pattern. |
| 5407 | |
| 5408 | In a larger pattern, keeping track of parenthesis numbers can be |
| 5409 | tricky. This is made easier by the use of relative references. Instead |
| 5410 | of (?1) in the pattern above you can write (?-2) to refer to the second |
| 5411 | most recently opened parentheses preceding the recursion. In other |
| 5412 | words, a negative number counts capturing parentheses leftwards from |
| 5413 | the point at which it is encountered. |
| 5414 | |
| 5415 | It is also possible to refer to subsequently opened parentheses, by |
| 5416 | writing references such as (?+2). However, these cannot be recursive |
| 5417 | because the reference is not inside the parentheses that are refer- |
| 5418 | enced. They are always non-recursive subroutine calls, as described in |
| 5419 | the next section. |
| 5420 | |
| 5421 | An alternative approach is to use named parentheses instead. The Perl |
| 5422 | syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also |
| 5423 | supported. We could rewrite the above example as follows: |
| 5424 | |
| 5425 | (?<pn> \( ( [^()]++ | (?&pn) )* \) ) |
| 5426 | |
| 5427 | If there is more than one subpattern with the same name, the earliest |
| 5428 | one is used. |
| 5429 | |
| 5430 | This particular example pattern that we have been looking at contains |
| 5431 | nested unlimited repeats, and so the use of a possessive quantifier for |
| 5432 | matching strings of non-parentheses is important when applying the pat- |
| 5433 | tern to strings that do not match. For example, when this pattern is |
| 5434 | applied to |
| 5435 | |
| 5436 | (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa() |
| 5437 | |
| 5438 | it yields "no match" quickly. However, if a possessive quantifier is |
| 5439 | not used, the match runs for a very long time indeed because there are |
| 5440 | so many different ways the + and * repeats can carve up the subject, |
| 5441 | and all have to be tested before failure can be reported. |
| 5442 | |
| 5443 | At the end of a match, the values of capturing parentheses are those |
| 5444 | from the outermost level. If you want to obtain intermediate values, a |
| 5445 | callout function can be used (see below and the pcrecallout documenta- |
| 5446 | tion). If the pattern above is matched against |
| 5447 | |
| 5448 | (ab(cd)ef) |
| 5449 | |
| 5450 | the value for the inner capturing parentheses (numbered 2) is "ef", |
| 5451 | which is the last value taken on at the top level. If a capturing sub- |
| 5452 | pattern is not matched at the top level, its final captured value is |
| 5453 | unset, even if it was (temporarily) set at a deeper level during the |
| 5454 | matching process. |
| 5455 | |
| 5456 | If there are more than 15 capturing parentheses in a pattern, PCRE has |
| 5457 | to obtain extra memory to store data during a recursion, which it does |
| 5458 | by using pcre_malloc, freeing it via pcre_free afterwards. If no memory |
| 5459 | can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error. |
| 5460 | |
| 5461 | Do not confuse the (?R) item with the condition (R), which tests for |
| 5462 | recursion. Consider this pattern, which matches text in angle brack- |
| 5463 | ets, allowing for arbitrary nesting. Only digits are allowed in nested |
| 5464 | brackets (that is, when recursing), whereas any characters are permit- |
| 5465 | ted at the outer level. |
| 5466 | |
| 5467 | < (?: (?(R) \d++ | [^<>]*+) | (?R)) * > |
| 5468 | |
| 5469 | In this pattern, (?(R) is the start of a conditional subpattern, with |
| 5470 | two different alternatives for the recursive and non-recursive cases. |
| 5471 | The (?R) item is the actual recursive call. |
| 5472 | |
| 5473 | Differences in recursion processing between PCRE and Perl |
| 5474 | |
| 5475 | Recursion processing in PCRE differs from Perl in two important ways. |
| 5476 | In PCRE (like Python, but unlike Perl), a recursive subpattern call is |
| 5477 | always treated as an atomic group. That is, once it has matched some of |
| 5478 | the subject string, it is never re-entered, even if it contains untried |
| 5479 | alternatives and there is a subsequent matching failure. This can be |
| 5480 | illustrated by the following pattern, which purports to match a palin- |
| 5481 | dromic string that contains an odd number of characters (for example, |
| 5482 | "a", "aba", "abcba", "abcdcba"): |
| 5483 | |
| 5484 | ^(.|(.)(?1)\2)$ |
| 5485 | |
| 5486 | The idea is that it either matches a single character, or two identical |
| 5487 | characters surrounding a sub-palindrome. In Perl, this pattern works; |
| 5488 | in PCRE it does not if the pattern is longer than three characters. |
| 5489 | Consider the subject string "abcba": |
| 5490 | |
| 5491 | At the top level, the first character is matched, but as it is not at |
| 5492 | the end of the string, the first alternative fails; the second alterna- |
| 5493 | tive is taken and the recursion kicks in. The recursive call to subpat- |
| 5494 | tern 1 successfully matches the next character ("b"). (Note that the |
| 5495 | beginning and end of line tests are not part of the recursion). |
| 5496 | |
| 5497 | Back at the top level, the next character ("c") is compared with what |
| 5498 | subpattern 2 matched, which was "a". This fails. Because the recursion |
| 5499 | is treated as an atomic group, there are now no backtracking points, |
| 5500 | and so the entire match fails. (Perl is able, at this point, to re- |
| 5501 | enter the recursion and try the second alternative.) However, if the |
| 5502 | pattern is written with the alternatives in the other order, things are |
| 5503 | different: |
| 5504 | |
| 5505 | ^((.)(?1)\2|.)$ |
| 5506 | |
| 5507 | This time, the recursing alternative is tried first, and continues to |
| 5508 | recurse until it runs out of characters, at which point the recursion |
| 5509 | fails. But this time we do have another alternative to try at the |
| 5510 | higher level. That is the big difference: in the previous case the |
| 5511 | remaining alternative is at a deeper recursion level, which PCRE cannot |
| 5512 | use. |
| 5513 | |
| 5514 | To change the pattern so that it matches all palindromic strings, not |
| 5515 | just those with an odd number of characters, it is tempting to change |
| 5516 | the pattern to this: |
| 5517 | |
| 5518 | ^((.)(?1)\2|.?)$ |
| 5519 | |
| 5520 | Again, this works in Perl, but not in PCRE, and for the same reason. |
| 5521 | When a deeper recursion has matched a single character, it cannot be |
| 5522 | entered again in order to match an empty string. The solution is to |
| 5523 | separate the two cases, and write out the odd and even cases as alter- |
| 5524 | natives at the higher level: |
| 5525 | |
| 5526 | ^(?:((.)(?1)\2|)|((.)(?3)\4|.)) |
| 5527 | |
| 5528 | If you want to match typical palindromic phrases, the pattern has to |
| 5529 | ignore all non-word characters, which can be done like this: |
| 5530 | |
| 5531 | ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$ |
| 5532 | |
| 5533 | If run with the PCRE_CASELESS option, this pattern matches phrases such |
| 5534 | as "A man, a plan, a canal: Panama!" and it works well in both PCRE and |
| 5535 | Perl. Note the use of the possessive quantifier *+ to avoid backtrack- |
| 5536 | ing into sequences of non-word characters. Without this, PCRE takes a |
| 5537 | great deal longer (ten times or more) to match typical phrases, and |
| 5538 | Perl takes so long that you think it has gone into a loop. |
| 5539 | |
| 5540 | WARNING: The palindrome-matching patterns above work only if the sub- |
| 5541 | ject string does not start with a palindrome that is shorter than the |
| 5542 | entire string. For example, although "abcba" is correctly matched, if |
| 5543 | the subject is "ababa", PCRE finds the palindrome "aba" at the start, |
| 5544 | then fails at top level because the end of the string does not follow. |
| 5545 | Once again, it cannot jump back into the recursion to try other alter- |
| 5546 | natives, so the entire match fails. |
| 5547 | |
| 5548 | The second way in which PCRE and Perl differ in their recursion pro- |
| 5549 | cessing is in the handling of captured values. In Perl, when a subpat- |
| 5550 | tern is called recursively or as a subpattern (see the next section), |
| 5551 | it has no access to any values that were captured outside the recur- |
| 5552 | sion, whereas in PCRE these values can be referenced. Consider this |
| 5553 | pattern: |
| 5554 | |
| 5555 | ^(.)(\1|a(?2)) |
| 5556 | |
| 5557 | In PCRE, this pattern matches "bab". The first capturing parentheses |
| 5558 | match "b", then in the second group, when the back reference \1 fails |
| 5559 | to match "b", the second alternative matches "a" and then recurses. In |
| 5560 | the recursion, \1 does now match "b" and so the whole match succeeds. |
| 5561 | In Perl, the pattern fails to match because inside the recursive call |
| 5562 | \1 cannot access the externally set value. |
| 5563 | |
| 5564 | |
| 5565 | SUBPATTERNS AS SUBROUTINES |
| 5566 | |
| 5567 | If the syntax for a recursive subpattern call (either by number or by |
| 5568 | name) is used outside the parentheses to which it refers, it operates |
| 5569 | like a subroutine in a programming language. The called subpattern may |
| 5570 | be defined before or after the reference. A numbered reference can be |
| 5571 | absolute or relative, as in these examples: |
| 5572 | |
| 5573 | (...(absolute)...)...(?2)... |
| 5574 | (...(relative)...)...(?-1)... |
| 5575 | (...(?+1)...(relative)... |
| 5576 | |
| 5577 | An earlier example pointed out that the pattern |
| 5578 | |
| 5579 | (sens|respons)e and \1ibility |
| 5580 | |
| 5581 | matches "sense and sensibility" and "response and responsibility", but |
| 5582 | not "sense and responsibility". If instead the pattern |
| 5583 | |
| 5584 | (sens|respons)e and (?1)ibility |
| 5585 | |
| 5586 | is used, it does match "sense and responsibility" as well as the other |
| 5587 | two strings. Another example is given in the discussion of DEFINE |
| 5588 | above. |
| 5589 | |
| 5590 | All subroutine calls, whether recursive or not, are always treated as |
| 5591 | atomic groups. That is, once a subroutine has matched some of the sub- |
| 5592 | ject string, it is never re-entered, even if it contains untried alter- |
| 5593 | natives and there is a subsequent matching failure. Any capturing |
| 5594 | parentheses that are set during the subroutine call revert to their |
| 5595 | previous values afterwards. |
| 5596 | |
| 5597 | Processing options such as case-independence are fixed when a subpat- |
| 5598 | tern is defined, so if it is used as a subroutine, such options cannot |
| 5599 | be changed for different calls. For example, consider this pattern: |
| 5600 | |
| 5601 | (abc)(?i:(?-1)) |
| 5602 | |
| 5603 | It matches "abcabc". It does not match "abcABC" because the change of |
| 5604 | processing option does not affect the called subpattern. |
| 5605 | |
| 5606 | |
| 5607 | ONIGURUMA SUBROUTINE SYNTAX |
| 5608 | |
| 5609 | For compatibility with Oniguruma, the non-Perl syntax \g followed by a |
| 5610 | name or a number enclosed either in angle brackets or single quotes, is |
| 5611 | an alternative syntax for referencing a subpattern as a subroutine, |
| 5612 | possibly recursively. Here are two of the examples used above, rewrit- |
| 5613 | ten using this syntax: |
| 5614 | |
| 5615 | (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) ) |
| 5616 | (sens|respons)e and \g'1'ibility |
| 5617 | |
| 5618 | PCRE supports an extension to Oniguruma: if a number is preceded by a |
| 5619 | plus or a minus sign it is taken as a relative reference. For example: |
| 5620 | |
| 5621 | (abc)(?i:\g<-1>) |
| 5622 | |
| 5623 | Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not |
| 5624 | synonymous. The former is a back reference; the latter is a subroutine |
| 5625 | call. |
| 5626 | |
| 5627 | |
| 5628 | CALLOUTS |
| 5629 | |
| 5630 | Perl has a feature whereby using the sequence (?{...}) causes arbitrary |
| 5631 | Perl code to be obeyed in the middle of matching a regular expression. |
| 5632 | This makes it possible, amongst other things, to extract different sub- |
| 5633 | strings that match the same pair of parentheses when there is a repeti- |
| 5634 | tion. |
| 5635 | |
| 5636 | PCRE provides a similar feature, but of course it cannot obey arbitrary |
| 5637 | Perl code. The feature is called "callout". The caller of PCRE provides |
| 5638 | an external function by putting its entry point in the global variable |
| 5639 | pcre_callout. By default, this variable contains NULL, which disables |
| 5640 | all calling out. |
| 5641 | |
| 5642 | Within a regular expression, (?C) indicates the points at which the |
| 5643 | external function is to be called. If you want to identify different |
| 5644 | callout points, you can put a number less than 256 after the letter C. |
| 5645 | The default value is zero. For example, this pattern has two callout |
| 5646 | points: |
| 5647 | |
| 5648 | (?C1)abc(?C2)def |
| 5649 | |
| 5650 | If the PCRE_AUTO_CALLOUT flag is passed to pcre_compile(), callouts are |
| 5651 | automatically installed before each item in the pattern. They are all |
| 5652 | numbered 255. |
| 5653 | |
| 5654 | During matching, when PCRE reaches a callout point (and pcre_callout is |
| 5655 | set), the external function is called. It is provided with the number |
| 5656 | of the callout, the position in the pattern, and, optionally, one item |
| 5657 | of data originally supplied by the caller of pcre_exec(). The callout |
| 5658 | function may cause matching to proceed, to backtrack, or to fail alto- |
| 5659 | gether. A complete description of the interface to the callout function |
| 5660 | is given in the pcrecallout documentation. |
| 5661 | |
| 5662 | |
| 5663 | BACKTRACKING CONTROL |
| 5664 | |
| 5665 | Perl 5.10 introduced a number of "Special Backtracking Control Verbs", |
| 5666 | which are described in the Perl documentation as "experimental and sub- |
| 5667 | ject to change or removal in a future version of Perl". It goes on to |
| 5668 | say: "Their usage in production code should be noted to avoid problems |
| 5669 | during upgrades." The same remarks apply to the PCRE features described |
| 5670 | in this section. |
| 5671 | |
| 5672 | Since these verbs are specifically related to backtracking, most of |
| 5673 | them can be used only when the pattern is to be matched using |
| 5674 | pcre_exec(), which uses a backtracking algorithm. With the exception of |
| 5675 | (*FAIL), which behaves like a failing negative assertion, they cause an |
| 5676 | error if encountered by pcre_dfa_exec(). |
| 5677 | |
| 5678 | If any of these verbs are used in an assertion or in a subpattern that |
| 5679 | is called as a subroutine (whether or not recursively), their effect is |
| 5680 | confined to that subpattern; it does not extend to the surrounding pat- |
| 5681 | tern, with one exception: the name from a *(MARK), (*PRUNE), or (*THEN) |
| 5682 | that is encountered in a successful positive assertion is passed back |
| 5683 | when a match succeeds (compare capturing parentheses in assertions). |
| 5684 | Note that such subpatterns are processed as anchored at the point where |
| 5685 | they are tested. Note also that Perl's treatment of subroutines is dif- |
| 5686 | ferent in some cases. |
| 5687 | |
| 5688 | The new verbs make use of what was previously invalid syntax: an open- |
| 5689 | ing parenthesis followed by an asterisk. They are generally of the form |
| 5690 | (*VERB) or (*VERB:NAME). Some may take either form, with differing be- |
| 5691 | haviour, depending on whether or not an argument is present. A name is |
| 5692 | any sequence of characters that does not include a closing parenthesis. |
| 5693 | If the name is empty, that is, if the closing parenthesis immediately |
| 5694 | follows the colon, the effect is as if the colon were not there. Any |
| 5695 | number of these verbs may occur in a pattern. |
| 5696 | |
| 5697 | PCRE contains some optimizations that are used to speed up matching by |
| 5698 | running some checks at the start of each match attempt. For example, it |
| 5699 | may know the minimum length of matching subject, or that a particular |
| 5700 | character must be present. When one of these optimizations suppresses |
| 5701 | the running of a match, any included backtracking verbs will not, of |
| 5702 | course, be processed. You can suppress the start-of-match optimizations |
| 5703 | by setting the PCRE_NO_START_OPTIMIZE option when calling pcre_com- |
| 5704 | pile() or pcre_exec(), or by starting the pattern with (*NO_START_OPT). |
| 5705 | |
| 5706 | Experiments with Perl suggest that it too has similar optimizations, |
| 5707 | sometimes leading to anomalous results. |
| 5708 | |
| 5709 | Verbs that act immediately |
| 5710 | |
| 5711 | The following verbs act as soon as they are encountered. They may not |
| 5712 | be followed by a name. |
| 5713 | |
| 5714 | (*ACCEPT) |
| 5715 | |
| 5716 | This verb causes the match to end successfully, skipping the remainder |
| 5717 | of the pattern. However, when it is inside a subpattern that is called |
| 5718 | as a subroutine, only that subpattern is ended successfully. Matching |
| 5719 | then continues at the outer level. If (*ACCEPT) is inside capturing |
| 5720 | parentheses, the data so far is captured. For example: |
| 5721 | |
| 5722 | A((?:A|B(*ACCEPT)|C)D) |
| 5723 | |
| 5724 | This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is cap- |
| 5725 | tured by the outer parentheses. |
| 5726 | |
| 5727 | (*FAIL) or (*F) |
| 5728 | |
| 5729 | This verb causes a matching failure, forcing backtracking to occur. It |
| 5730 | is equivalent to (?!) but easier to read. The Perl documentation notes |
| 5731 | that it is probably useful only when combined with (?{}) or (??{}). |
| 5732 | Those are, of course, Perl features that are not present in PCRE. The |
| 5733 | nearest equivalent is the callout feature, as for example in this pat- |
| 5734 | tern: |
| 5735 | |
| 5736 | a+(?C)(*FAIL) |
| 5737 | |
| 5738 | A match with the string "aaaa" always fails, but the callout is taken |
| 5739 | before each backtrack happens (in this example, 10 times). |
| 5740 | |
| 5741 | Recording which path was taken |
| 5742 | |
| 5743 | There is one verb whose main purpose is to track how a match was |
| 5744 | arrived at, though it also has a secondary use in conjunction with |
| 5745 | advancing the match starting point (see (*SKIP) below). |
| 5746 | |
| 5747 | (*MARK:NAME) or (*:NAME) |
| 5748 | |
| 5749 | A name is always required with this verb. There may be as many |
| 5750 | instances of (*MARK) as you like in a pattern, and their names do not |
| 5751 | have to be unique. |
| 5752 | |
| 5753 | When a match succeeds, the name of the last-encountered (*MARK) on the |
| 5754 | matching path is passed back to the caller via the pcre_extra data |
| 5755 | structure, as described in the section on pcre_extra in the pcreapi |
| 5756 | documentation. Here is an example of pcretest output, where the /K mod- |
| 5757 | ifier requests the retrieval and outputting of (*MARK) data: |
| 5758 | |
| 5759 | re> /X(*MARK:A)Y|X(*MARK:B)Z/K |
| 5760 | data> XY |
| 5761 | 0: XY |
| 5762 | MK: A |
| 5763 | XZ |
| 5764 | 0: XZ |
| 5765 | MK: B |
| 5766 | |
| 5767 | The (*MARK) name is tagged with "MK:" in this output, and in this exam- |
| 5768 | ple it indicates which of the two alternatives matched. This is a more |
| 5769 | efficient way of obtaining this information than putting each alterna- |
| 5770 | tive in its own capturing parentheses. |
| 5771 | |
| 5772 | If (*MARK) is encountered in a positive assertion, its name is recorded |
| 5773 | and passed back if it is the last-encountered. This does not happen for |
| 5774 | negative assertions. |
| 5775 | |
| 5776 | After a partial match or a failed match, the name of the last encoun- |
| 5777 | tered (*MARK) in the entire match process is returned. For example: |
| 5778 | |
| 5779 | re> /X(*MARK:A)Y|X(*MARK:B)Z/K |
| 5780 | data> XP |
| 5781 | No match, mark = B |
| 5782 | |
| 5783 | Note that in this unanchored example the mark is retained from the |
| 5784 | match attempt that started at the letter "X". Subsequent match attempts |
| 5785 | starting at "P" and then with an empty string do not get as far as the |
| 5786 | (*MARK) item, but nevertheless do not reset it. |
| 5787 | |
| 5788 | Verbs that act after backtracking |
| 5789 | |
| 5790 | The following verbs do nothing when they are encountered. Matching con- |
| 5791 | tinues with what follows, but if there is no subsequent match, causing |
| 5792 | a backtrack to the verb, a failure is forced. That is, backtracking |
| 5793 | cannot pass to the left of the verb. However, when one of these verbs |
| 5794 | appears inside an atomic group, its effect is confined to that group, |
| 5795 | because once the group has been matched, there is never any backtrack- |
| 5796 | ing into it. In this situation, backtracking can "jump back" to the |
| 5797 | left of the entire atomic group. (Remember also, as stated above, that |
| 5798 | this localization also applies in subroutine calls and assertions.) |
| 5799 | |
| 5800 | These verbs differ in exactly what kind of failure occurs when back- |
| 5801 | tracking reaches them. |
| 5802 | |
| 5803 | (*COMMIT) |
| 5804 | |
| 5805 | This verb, which may not be followed by a name, causes the whole match |
| 5806 | to fail outright if the rest of the pattern does not match. Even if the |
| 5807 | pattern is unanchored, no further attempts to find a match by advancing |
| 5808 | the starting point take place. Once (*COMMIT) has been passed, |
| 5809 | pcre_exec() is committed to finding a match at the current starting |
| 5810 | point, or not at all. For example: |
| 5811 | |
| 5812 | a+(*COMMIT)b |
| 5813 | |
| 5814 | This matches "xxaab" but not "aacaab". It can be thought of as a kind |
| 5815 | of dynamic anchor, or "I've started, so I must finish." The name of the |
| 5816 | most recently passed (*MARK) in the path is passed back when (*COMMIT) |
| 5817 | forces a match failure. |
| 5818 | |
| 5819 | Note that (*COMMIT) at the start of a pattern is not the same as an |
| 5820 | anchor, unless PCRE's start-of-match optimizations are turned off, as |
| 5821 | shown in this pcretest example: |
| 5822 | |
| 5823 | re> /(*COMMIT)abc/ |
| 5824 | data> xyzabc |
| 5825 | 0: abc |
| 5826 | xyzabc\Y |
| 5827 | No match |
| 5828 | |
| 5829 | PCRE knows that any match must start with "a", so the optimization |
| 5830 | skips along the subject to "a" before running the first match attempt, |
| 5831 | which succeeds. When the optimization is disabled by the \Y escape in |
| 5832 | the second subject, the match starts at "x" and so the (*COMMIT) causes |
| 5833 | it to fail without trying any other starting points. |
| 5834 | |
| 5835 | (*PRUNE) or (*PRUNE:NAME) |
| 5836 | |
| 5837 | This verb causes the match to fail at the current starting position in |
| 5838 | the subject if the rest of the pattern does not match. If the pattern |
| 5839 | is unanchored, the normal "bumpalong" advance to the next starting |
| 5840 | character then happens. Backtracking can occur as usual to the left of |
| 5841 | (*PRUNE), before it is reached, or when matching to the right of |
| 5842 | (*PRUNE), but if there is no match to the right, backtracking cannot |
| 5843 | cross (*PRUNE). In simple cases, the use of (*PRUNE) is just an alter- |
| 5844 | native to an atomic group or possessive quantifier, but there are some |
| 5845 | uses of (*PRUNE) that cannot be expressed in any other way. The behav- |
| 5846 | iour of (*PRUNE:NAME) is the same as (*MARK:NAME)(*PRUNE). In an |
| 5847 | anchored pattern (*PRUNE) has the same effect as (*COMMIT). |
| 5848 | |
| 5849 | (*SKIP) |
| 5850 | |
| 5851 | This verb, when given without a name, is like (*PRUNE), except that if |
| 5852 | the pattern is unanchored, the "bumpalong" advance is not to the next |
| 5853 | character, but to the position in the subject where (*SKIP) was encoun- |
| 5854 | tered. (*SKIP) signifies that whatever text was matched leading up to |
| 5855 | it cannot be part of a successful match. Consider: |
| 5856 | |
| 5857 | a+(*SKIP)b |
| 5858 | |
| 5859 | If the subject is "aaaac...", after the first match attempt fails |
| 5860 | (starting at the first character in the string), the starting point |
| 5861 | skips on to start the next attempt at "c". Note that a possessive quan- |
| 5862 | tifer does not have the same effect as this example; although it would |
| 5863 | suppress backtracking during the first match attempt, the second |
| 5864 | attempt would start at the second character instead of skipping on to |
| 5865 | "c". |
| 5866 | |
| 5867 | (*SKIP:NAME) |
| 5868 | |
| 5869 | When (*SKIP) has an associated name, its behaviour is modified. If the |
| 5870 | following pattern fails to match, the previous path through the pattern |
| 5871 | is searched for the most recent (*MARK) that has the same name. If one |
| 5872 | is found, the "bumpalong" advance is to the subject position that cor- |
| 5873 | responds to that (*MARK) instead of to where (*SKIP) was encountered. |
| 5874 | If no (*MARK) with a matching name is found, the (*SKIP) is ignored. |
| 5875 | |
| 5876 | (*THEN) or (*THEN:NAME) |
| 5877 | |
| 5878 | This verb causes a skip to the next innermost alternative if the rest |
| 5879 | of the pattern does not match. That is, it cancels pending backtrack- |
| 5880 | ing, but only within the current alternative. Its name comes from the |
| 5881 | observation that it can be used for a pattern-based if-then-else block: |
| 5882 | |
| 5883 | ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ... |
| 5884 | |
| 5885 | If the COND1 pattern matches, FOO is tried (and possibly further items |
| 5886 | after the end of the group if FOO succeeds); on failure, the matcher |
| 5887 | skips to the second alternative and tries COND2, without backtracking |
| 5888 | into COND1. The behaviour of (*THEN:NAME) is exactly the same as |
| 5889 | (*MARK:NAME)(*THEN). If (*THEN) is not inside an alternation, it acts |
| 5890 | like (*PRUNE). |
| 5891 | |
| 5892 | Note that a subpattern that does not contain a | character is just a |
| 5893 | part of the enclosing alternative; it is not a nested alternation with |
| 5894 | only one alternative. The effect of (*THEN) extends beyond such a sub- |
| 5895 | pattern to the enclosing alternative. Consider this pattern, where A, |
| 5896 | B, etc. are complex pattern fragments that do not contain any | charac- |
| 5897 | ters at this level: |
| 5898 | |
| 5899 | A (B(*THEN)C) | D |
| 5900 | |
| 5901 | If A and B are matched, but there is a failure in C, matching does not |
| 5902 | backtrack into A; instead it moves to the next alternative, that is, D. |
| 5903 | However, if the subpattern containing (*THEN) is given an alternative, |
| 5904 | it behaves differently: |
| 5905 | |
| 5906 | A (B(*THEN)C | (*FAIL)) | D |
| 5907 | |
| 5908 | The effect of (*THEN) is now confined to the inner subpattern. After a |
| 5909 | failure in C, matching moves to (*FAIL), which causes the whole subpat- |
| 5910 | tern to fail because there are no more alternatives to try. In this |
| 5911 | case, matching does now backtrack into A. |
| 5912 | |
| 5913 | Note also that a conditional subpattern is not considered as having two |
| 5914 | alternatives, because only one is ever used. In other words, the | |
| 5915 | character in a conditional subpattern has a different meaning. Ignoring |
| 5916 | white space, consider: |
| 5917 | |
| 5918 | ^.*? (?(?=a) a | b(*THEN)c ) |
| 5919 | |
| 5920 | If the subject is "ba", this pattern does not match. Because .*? is |
| 5921 | ungreedy, it initially matches zero characters. The condition (?=a) |
| 5922 | then fails, the character "b" is matched, but "c" is not. At this |
| 5923 | point, matching does not backtrack to .*? as might perhaps be expected |
| 5924 | from the presence of the | character. The conditional subpattern is |
| 5925 | part of the single alternative that comprises the whole pattern, and so |
| 5926 | the match fails. (If there was a backtrack into .*?, allowing it to |
| 5927 | match "b", the match would succeed.) |
| 5928 | |
| 5929 | The verbs just described provide four different "strengths" of control |
| 5930 | when subsequent matching fails. (*THEN) is the weakest, carrying on the |
| 5931 | match at the next alternative. (*PRUNE) comes next, failing the match |
| 5932 | at the current starting position, but allowing an advance to the next |
| 5933 | character (for an unanchored pattern). (*SKIP) is similar, except that |
| 5934 | the advance may be more than one character. (*COMMIT) is the strongest, |
| 5935 | causing the entire match to fail. |
| 5936 | |
| 5937 | If more than one such verb is present in a pattern, the "strongest" one |
| 5938 | wins. For example, consider this pattern, where A, B, etc. are complex |
| 5939 | pattern fragments: |
| 5940 | |
| 5941 | (A(*COMMIT)B(*THEN)C|D) |
| 5942 | |
| 5943 | Once A has matched, PCRE is committed to this match, at the current |
| 5944 | starting position. If subsequently B matches, but C does not, the nor- |
| 5945 | mal (*THEN) action of trying the next alternative (that is, D) does not |
| 5946 | happen because (*COMMIT) overrides. |
| 5947 | |
| 5948 | |
| 5949 | SEE ALSO |
| 5950 | |
| 5951 | pcreapi(3), pcrecallout(3), pcrematching(3), pcresyntax(3), pcre(3). |
| 5952 | |
| 5953 | |
| 5954 | AUTHOR |
| 5955 | |
| 5956 | Philip Hazel |
| 5957 | University Computing Service |
| 5958 | Cambridge CB2 3QH, England. |
| 5959 | |
| 5960 | |
| 5961 | REVISION |
| 5962 | |
| 5963 | Last updated: 29 November 2011 |
| 5964 | Copyright (c) 1997-2011 University of Cambridge. |
| 5965 | ------------------------------------------------------------------------------ |
| 5966 | |
| 5967 | |
| 5968 | PCRESYNTAX(3) PCRESYNTAX(3) |
| 5969 | |
| 5970 | |
| 5971 | NAME |
| 5972 | PCRE - Perl-compatible regular expressions |
| 5973 | |
| 5974 | |
| 5975 | PCRE REGULAR EXPRESSION SYNTAX SUMMARY |
| 5976 | |
| 5977 | The full syntax and semantics of the regular expressions that are sup- |
| 5978 | ported by PCRE are described in the pcrepattern documentation. This |
| 5979 | document contains just a quick-reference summary of the syntax. |
| 5980 | |
| 5981 | |
| 5982 | QUOTING |
| 5983 | |
| 5984 | \x where x is non-alphanumeric is a literal x |
| 5985 | \Q...\E treat enclosed characters as literal |
| 5986 | |
| 5987 | |
| 5988 | CHARACTERS |
| 5989 | |
| 5990 | \a alarm, that is, the BEL character (hex 07) |
| 5991 | \cx "control-x", where x is any ASCII character |
| 5992 | \e escape (hex 1B) |
| 5993 | \f formfeed (hex 0C) |
| 5994 | \n newline (hex 0A) |
| 5995 | \r carriage return (hex 0D) |
| 5996 | \t tab (hex 09) |
| 5997 | \ddd character with octal code ddd, or backreference |
| 5998 | \xhh character with hex code hh |
| 5999 | \x{hhh..} character with hex code hhh.. |
| 6000 | |
| 6001 | |
| 6002 | CHARACTER TYPES |
| 6003 | |
| 6004 | . any character except newline; |
| 6005 | in dotall mode, any character whatsoever |
| 6006 | \C one byte, even in UTF-8 mode (best avoided) |
| 6007 | \d a decimal digit |
| 6008 | \D a character that is not a decimal digit |
| 6009 | \h a horizontal whitespace character |
| 6010 | \H a character that is not a horizontal whitespace character |
| 6011 | \N a character that is not a newline |
| 6012 | \p{xx} a character with the xx property |
| 6013 | \P{xx} a character without the xx property |
| 6014 | \R a newline sequence |
| 6015 | \s a whitespace character |
| 6016 | \S a character that is not a whitespace character |
| 6017 | \v a vertical whitespace character |
| 6018 | \V a character that is not a vertical whitespace character |
| 6019 | \w a "word" character |
| 6020 | \W a "non-word" character |
| 6021 | \X an extended Unicode sequence |
| 6022 | |
| 6023 | In PCRE, by default, \d, \D, \s, \S, \w, and \W recognize only ASCII |
| 6024 | characters, even in UTF-8 mode. However, this can be changed by setting |
| 6025 | the PCRE_UCP option. |
| 6026 | |
| 6027 | |
| 6028 | GENERAL CATEGORY PROPERTIES FOR \p and \P |
| 6029 | |
| 6030 | C Other |
| 6031 | Cc Control |
| 6032 | Cf Format |
| 6033 | Cn Unassigned |
| 6034 | Co Private use |
| 6035 | Cs Surrogate |
| 6036 | |
| 6037 | L Letter |
| 6038 | Ll Lower case letter |
| 6039 | Lm Modifier letter |
| 6040 | Lo Other letter |
| 6041 | Lt Title case letter |
| 6042 | Lu Upper case letter |
| 6043 | L& Ll, Lu, or Lt |
| 6044 | |
| 6045 | M Mark |
| 6046 | Mc Spacing mark |
| 6047 | Me Enclosing mark |
| 6048 | Mn Non-spacing mark |
| 6049 | |
| 6050 | N Number |
| 6051 | Nd Decimal number |
| 6052 | Nl Letter number |
| 6053 | No Other number |
| 6054 | |
| 6055 | P Punctuation |
| 6056 | Pc Connector punctuation |
| 6057 | Pd Dash punctuation |
| 6058 | Pe Close punctuation |
| 6059 | Pf Final punctuation |
| 6060 | Pi Initial punctuation |
| 6061 | Po Other punctuation |
| 6062 | Ps Open punctuation |
| 6063 | |
| 6064 | S Symbol |
| 6065 | Sc Currency symbol |
| 6066 | Sk Modifier symbol |
| 6067 | Sm Mathematical symbol |
| 6068 | So Other symbol |
| 6069 | |
| 6070 | Z Separator |
| 6071 | Zl Line separator |
| 6072 | Zp Paragraph separator |
| 6073 | Zs Space separator |
| 6074 | |
| 6075 | |
| 6076 | PCRE SPECIAL CATEGORY PROPERTIES FOR \p and \P |
| 6077 | |
| 6078 | Xan Alphanumeric: union of properties L and N |
| 6079 | Xps POSIX space: property Z or tab, NL, VT, FF, CR |
| 6080 | Xsp Perl space: property Z or tab, NL, FF, CR |
| 6081 | Xwd Perl word: property Xan or underscore |
| 6082 | |
| 6083 | |
| 6084 | SCRIPT NAMES FOR \p AND \P |
| 6085 | |
| 6086 | Arabic, Armenian, Avestan, Balinese, Bamum, Bengali, Bopomofo, Braille, |
| 6087 | Buginese, Buhid, Canadian_Aboriginal, Carian, Cham, Cherokee, Common, |
| 6088 | Coptic, Cuneiform, Cypriot, Cyrillic, Deseret, Devanagari, Egyp- |
| 6089 | tian_Hieroglyphs, Ethiopic, Georgian, Glagolitic, Gothic, Greek, |
| 6090 | Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hiragana, Impe- |
| 6091 | rial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscriptional_Parthian, |
| 6092 | Javanese, Kaithi, Kannada, Katakana, Kayah_Li, Kharoshthi, Khmer, Lao, |
| 6093 | Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian, Lydian, Malayalam, |
| 6094 | Meetei_Mayek, Mongolian, Myanmar, New_Tai_Lue, Nko, Ogham, Old_Italic, |
| 6095 | Old_Persian, Old_South_Arabian, Old_Turkic, Ol_Chiki, Oriya, Osmanya, |
| 6096 | Phags_Pa, Phoenician, Rejang, Runic, Samaritan, Saurashtra, Shavian, |
| 6097 | Sinhala, Sundanese, Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le, |
| 6098 | Tai_Tham, Tai_Viet, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, |
| 6099 | Ugaritic, Vai, Yi. |
| 6100 | |
| 6101 | |
| 6102 | CHARACTER CLASSES |
| 6103 | |
| 6104 | [...] positive character class |
| 6105 | [^...] negative character class |
| 6106 | [x-y] range (can be used for hex characters) |
| 6107 | [[:xxx:]] positive POSIX named set |
| 6108 | [[:^xxx:]] negative POSIX named set |
| 6109 | |
| 6110 | alnum alphanumeric |
| 6111 | alpha alphabetic |
| 6112 | ascii 0-127 |
| 6113 | blank space or tab |
| 6114 | cntrl control character |
| 6115 | digit decimal digit |
| 6116 | graph printing, excluding space |
| 6117 | lower lower case letter |
| 6118 | print printing, including space |
| 6119 | punct printing, excluding alphanumeric |
| 6120 | space whitespace |
| 6121 | upper upper case letter |
| 6122 | word same as \w |
| 6123 | xdigit hexadecimal digit |
| 6124 | |
| 6125 | In PCRE, POSIX character set names recognize only ASCII characters by |
| 6126 | default, but some of them use Unicode properties if PCRE_UCP is set. |
| 6127 | You can use \Q...\E inside a character class. |
| 6128 | |
| 6129 | |
| 6130 | QUANTIFIERS |
| 6131 | |
| 6132 | ? 0 or 1, greedy |
| 6133 | ?+ 0 or 1, possessive |
| 6134 | ?? 0 or 1, lazy |
| 6135 | * 0 or more, greedy |
| 6136 | *+ 0 or more, possessive |
| 6137 | *? 0 or more, lazy |
| 6138 | + 1 or more, greedy |
| 6139 | ++ 1 or more, possessive |
| 6140 | +? 1 or more, lazy |
| 6141 | {n} exactly n |
| 6142 | {n,m} at least n, no more than m, greedy |
| 6143 | {n,m}+ at least n, no more than m, possessive |
| 6144 | {n,m}? at least n, no more than m, lazy |
| 6145 | {n,} n or more, greedy |
| 6146 | {n,}+ n or more, possessive |
| 6147 | {n,}? n or more, lazy |
| 6148 | |
| 6149 | |
| 6150 | ANCHORS AND SIMPLE ASSERTIONS |
| 6151 | |
| 6152 | \b word boundary |
| 6153 | \B not a word boundary |
| 6154 | ^ start of subject |
| 6155 | also after internal newline in multiline mode |
| 6156 | \A start of subject |
| 6157 | $ end of subject |
| 6158 | also before newline at end of subject |
| 6159 | also before internal newline in multiline mode |
| 6160 | \Z end of subject |
| 6161 | also before newline at end of subject |
| 6162 | \z end of subject |
| 6163 | \G first matching position in subject |
| 6164 | |
| 6165 | |
| 6166 | MATCH POINT RESET |
| 6167 | |
| 6168 | \K reset start of match |
| 6169 | |
| 6170 | |
| 6171 | ALTERNATION |
| 6172 | |
| 6173 | expr|expr|expr... |
| 6174 | |
| 6175 | |
| 6176 | CAPTURING |
| 6177 | |
| 6178 | (...) capturing group |
| 6179 | (?<name>...) named capturing group (Perl) |
| 6180 | (?'name'...) named capturing group (Perl) |
| 6181 | (?P<name>...) named capturing group (Python) |
| 6182 | (?:...) non-capturing group |
| 6183 | (?|...) non-capturing group; reset group numbers for |
| 6184 | capturing groups in each alternative |
| 6185 | |
| 6186 | |
| 6187 | ATOMIC GROUPS |
| 6188 | |
| 6189 | (?>...) atomic, non-capturing group |
| 6190 | |
| 6191 | |
| 6192 | COMMENT |
| 6193 | |
| 6194 | (?#....) comment (not nestable) |
| 6195 | |
| 6196 | |
| 6197 | OPTION SETTING |
| 6198 | |
| 6199 | (?i) caseless |
| 6200 | (?J) allow duplicate names |
| 6201 | (?m) multiline |
| 6202 | (?s) single line (dotall) |
| 6203 | (?U) default ungreedy (lazy) |
| 6204 | (?x) extended (ignore white space) |
| 6205 | (?-...) unset option(s) |
| 6206 | |
| 6207 | The following are recognized only at the start of a pattern or after |
| 6208 | one of the newline-setting options with similar syntax: |
| 6209 | |
| 6210 | (*NO_START_OPT) no start-match optimization (PCRE_NO_START_OPTIMIZE) |
| 6211 | (*UTF8) set UTF-8 mode (PCRE_UTF8) |
| 6212 | (*UCP) set PCRE_UCP (use Unicode properties for \d etc) |
| 6213 | |
| 6214 | |
| 6215 | LOOKAHEAD AND LOOKBEHIND ASSERTIONS |
| 6216 | |
| 6217 | (?=...) positive look ahead |
| 6218 | (?!...) negative look ahead |
| 6219 | (?<=...) positive look behind |
| 6220 | (?<!...) negative look behind |
| 6221 | |
| 6222 | Each top-level branch of a look behind must be of a fixed length. |
| 6223 | |
| 6224 | |
| 6225 | BACKREFERENCES |
| 6226 | |
| 6227 | \n reference by number (can be ambiguous) |
| 6228 | \gn reference by number |
| 6229 | \g{n} reference by number |
| 6230 | \g{-n} relative reference by number |
| 6231 | \k<name> reference by name (Perl) |
| 6232 | \k'name' reference by name (Perl) |
| 6233 | \g{name} reference by name (Perl) |
| 6234 | \k{name} reference by name (.NET) |
| 6235 | (?P=name) reference by name (Python) |
| 6236 | |
| 6237 | |
| 6238 | SUBROUTINE REFERENCES (POSSIBLY RECURSIVE) |
| 6239 | |
| 6240 | (?R) recurse whole pattern |
| 6241 | (?n) call subpattern by absolute number |
| 6242 | (?+n) call subpattern by relative number |
| 6243 | (?-n) call subpattern by relative number |
| 6244 | (?&name) call subpattern by name (Perl) |
| 6245 | (?P>name) call subpattern by name (Python) |
| 6246 | \g<name> call subpattern by name (Oniguruma) |
| 6247 | \g'name' call subpattern by name (Oniguruma) |
| 6248 | \g<n> call subpattern by absolute number (Oniguruma) |
| 6249 | \g'n' call subpattern by absolute number (Oniguruma) |
| 6250 | \g<+n> call subpattern by relative number (PCRE extension) |
| 6251 | \g'+n' call subpattern by relative number (PCRE extension) |
| 6252 | \g<-n> call subpattern by relative number (PCRE extension) |
| 6253 | \g'-n' call subpattern by relative number (PCRE extension) |
| 6254 | |
| 6255 | |
| 6256 | CONDITIONAL PATTERNS |
| 6257 | |
| 6258 | (?(condition)yes-pattern) |
| 6259 | (?(condition)yes-pattern|no-pattern) |
| 6260 | |
| 6261 | (?(n)... absolute reference condition |
| 6262 | (?(+n)... relative reference condition |
| 6263 | (?(-n)... relative reference condition |
| 6264 | (?(<name>)... named reference condition (Perl) |
| 6265 | (?('name')... named reference condition (Perl) |
| 6266 | (?(name)... named reference condition (PCRE) |
| 6267 | (?(R)... overall recursion condition |
| 6268 | (?(Rn)... specific group recursion condition |
| 6269 | (?(R&name)... specific recursion condition |
| 6270 | (?(DEFINE)... define subpattern for reference |
| 6271 | (?(assert)... assertion condition |
| 6272 | |
| 6273 | |
| 6274 | BACKTRACKING CONTROL |
| 6275 | |
| 6276 | The following act immediately they are reached: |
| 6277 | |
| 6278 | (*ACCEPT) force successful match |
| 6279 | (*FAIL) force backtrack; synonym (*F) |
| 6280 | |
| 6281 | The following act only when a subsequent match failure causes a back- |
| 6282 | track to reach them. They all force a match failure, but they differ in |
| 6283 | what happens afterwards. Those that advance the start-of-match point do |
| 6284 | so only if the pattern is not anchored. |
| 6285 | |
| 6286 | (*COMMIT) overall failure, no advance of starting point |
| 6287 | (*PRUNE) advance to next starting character |
| 6288 | (*SKIP) advance start to current matching position |
| 6289 | (*THEN) local failure, backtrack to next alternation |
| 6290 | |
| 6291 | |
| 6292 | NEWLINE CONVENTIONS |
| 6293 | |
| 6294 | These are recognized only at the very start of the pattern or after a |
| 6295 | (*BSR_...) or (*UTF8) or (*UCP) option. |
| 6296 | |
| 6297 | (*CR) carriage return only |
| 6298 | (*LF) linefeed only |
| 6299 | (*CRLF) carriage return followed by linefeed |
| 6300 | (*ANYCRLF) all three of the above |
| 6301 | (*ANY) any Unicode newline sequence |
| 6302 | |
| 6303 | |
| 6304 | WHAT \R MATCHES |
| 6305 | |
| 6306 | These are recognized only at the very start of the pattern or after a |
| 6307 | (*...) option that sets the newline convention or UTF-8 or UCP mode. |
| 6308 | |
| 6309 | (*BSR_ANYCRLF) CR, LF, or CRLF |
| 6310 | (*BSR_UNICODE) any Unicode newline sequence |
| 6311 | |
| 6312 | |
| 6313 | CALLOUTS |
| 6314 | |
| 6315 | (?C) callout |
| 6316 | (?Cn) callout with data n |
| 6317 | |
| 6318 | |
| 6319 | SEE ALSO |
| 6320 | |
| 6321 | pcrepattern(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3). |
| 6322 | |
| 6323 | |
| 6324 | AUTHOR |
| 6325 | |
| 6326 | Philip Hazel |
| 6327 | University Computing Service |
| 6328 | Cambridge CB2 3QH, England. |
| 6329 | |
| 6330 | |
| 6331 | REVISION |
| 6332 | |
| 6333 | Last updated: 21 November 2010 |
| 6334 | Copyright (c) 1997-2010 University of Cambridge. |
| 6335 | ------------------------------------------------------------------------------ |
| 6336 | |
| 6337 | |
| 6338 | PCREUNICODE(3) PCREUNICODE(3) |
| 6339 | |
| 6340 | |
| 6341 | NAME |
| 6342 | PCRE - Perl-compatible regular expressions |
| 6343 | |
| 6344 | |
| 6345 | UTF-8 AND UNICODE PROPERTY SUPPORT |
| 6346 | |
| 6347 | In order process UTF-8 strings, you must build PCRE to include UTF-8 |
| 6348 | support in the code, and, in addition, you must call pcre_compile() |
| 6349 | with the PCRE_UTF8 option flag, or the pattern must start with the |
| 6350 | sequence (*UTF8). When either of these is the case, both the pattern |
| 6351 | and any subject strings that are matched against it are treated as |
| 6352 | UTF-8 strings instead of strings of 1-byte characters. PCRE does not |
| 6353 | support any other formats (in particular, it does not support UTF-16). |
| 6354 | |
| 6355 | If you compile PCRE with UTF-8 support, but do not use it at run time, |
| 6356 | the library will be a bit bigger, but the additional run time overhead |
| 6357 | is limited to testing the PCRE_UTF8 flag occasionally, so should not be |
| 6358 | very big. |
| 6359 | |
| 6360 | If PCRE is built with Unicode character property support (which implies |
| 6361 | UTF-8 support), the escape sequences \p{..}, \P{..}, and \X are sup- |
| 6362 | ported. The available properties that can be tested are limited to the |
| 6363 | general category properties such as Lu for an upper case letter or Nd |
| 6364 | for a decimal number, the Unicode script names such as Arabic or Han, |
| 6365 | and the derived properties Any and L&. A full list is given in the |
| 6366 | pcrepattern documentation. Only the short names for properties are sup- |
| 6367 | ported. For example, \p{L} matches a letter. Its Perl synonym, \p{Let- |
| 6368 | ter}, is not supported. Furthermore, in Perl, many properties may |
| 6369 | optionally be prefixed by "Is", for compatibility with Perl 5.6. PCRE |
| 6370 | does not support this. |
| 6371 | |
| 6372 | Validity of UTF-8 strings |
| 6373 | |
| 6374 | When you set the PCRE_UTF8 flag, the strings passed as patterns and |
| 6375 | subjects are (by default) checked for validity on entry to the relevant |
| 6376 | functions. From release 7.3 of PCRE, the check is according the rules |
| 6377 | of RFC 3629, which are themselves derived from the Unicode specifica- |
| 6378 | tion. Earlier releases of PCRE followed the rules of RFC 2279, which |
| 6379 | allows the full range of 31-bit values (0 to 0x7FFFFFFF). The current |
| 6380 | check allows only values in the range U+0 to U+10FFFF, excluding U+D800 |
| 6381 | to U+DFFF. |
| 6382 | |
| 6383 | The excluded code points are the "Low Surrogate Area" of Unicode, of |
| 6384 | which the Unicode Standard says this: "The Low Surrogate Area does not |
| 6385 | contain any character assignments, consequently no character code |
| 6386 | charts or namelists are provided for this area. Surrogates are reserved |
| 6387 | for use with UTF-16 and then must be used in pairs." The code points |
| 6388 | that are encoded by UTF-16 pairs are available as independent code |
| 6389 | points in the UTF-8 encoding. (In other words, the whole surrogate |
| 6390 | thing is a fudge for UTF-16 which unfortunately messes up UTF-8.) |
| 6391 | |
| 6392 | If an invalid UTF-8 string is passed to PCRE, an error return is given. |
| 6393 | At compile time, the only additional information is the offset to the |
| 6394 | first byte of the failing character. The runtime functions pcre_exec() |
| 6395 | and pcre_dfa_exec() also pass back this information, as well as a more |
| 6396 | detailed reason code if the caller has provided memory in which to do |
| 6397 | this. |
| 6398 | |
| 6399 | In some situations, you may already know that your strings are valid, |
| 6400 | and therefore want to skip these checks in order to improve perfor- |
| 6401 | mance. If you set the PCRE_NO_UTF8_CHECK flag at compile time or at run |
| 6402 | time, PCRE assumes that the pattern or subject it is given (respec- |
| 6403 | tively) contains only valid UTF-8 codes. In this case, it does not |
| 6404 | diagnose an invalid UTF-8 string. |
| 6405 | |
| 6406 | If you pass an invalid UTF-8 string when PCRE_NO_UTF8_CHECK is set, |
| 6407 | what happens depends on why the string is invalid. If the string con- |
| 6408 | forms to the "old" definition of UTF-8 (RFC 2279), it is processed as a |
| 6409 | string of characters in the range 0 to 0x7FFFFFFF by pcre_dfa_exec() |
| 6410 | and the interpreted version of pcre_exec(). In other words, apart from |
| 6411 | the initial validity test, these functions (when in UTF-8 mode) handle |
| 6412 | strings according to the more liberal rules of RFC 2279. However, the |
| 6413 | just-in-time (JIT) optimization for pcre_exec() supports only RFC 3629. |
| 6414 | If you are using JIT optimization, or if the string does not even con- |
| 6415 | form to RFC 2279, the result is undefined. Your program may crash. |
| 6416 | |
| 6417 | If you want to process strings of values in the full range 0 to |
| 6418 | 0x7FFFFFFF, encoded in a UTF-8-like manner as per the old RFC, you can |
| 6419 | set PCRE_NO_UTF8_CHECK to bypass the more restrictive test. However, in |
| 6420 | this situation, you will have to apply your own validity check, and |
| 6421 | avoid the use of JIT optimization. |
| 6422 | |
| 6423 | General comments about UTF-8 mode |
| 6424 | |
| 6425 | 1. An unbraced hexadecimal escape sequence (such as \xb3) matches a |
| 6426 | two-byte UTF-8 character if the value is greater than 127. |
| 6427 | |
| 6428 | 2. Octal numbers up to \777 are recognized, and match two-byte UTF-8 |
| 6429 | characters for values greater than \177. |
| 6430 | |
| 6431 | 3. Repeat quantifiers apply to complete UTF-8 characters, not to indi- |
| 6432 | vidual bytes, for example: \x{100}{3}. |
| 6433 | |
| 6434 | 4. The dot metacharacter matches one UTF-8 character instead of a sin- |
| 6435 | gle byte. |
| 6436 | |
| 6437 | 5. The escape sequence \C can be used to match a single byte in UTF-8 |
| 6438 | mode, but its use can lead to some strange effects because it breaks up |
| 6439 | multibyte characters (see the description of \C in the pcrepattern doc- |
| 6440 | umentation). The use of \C is not supported in the alternative matching |
| 6441 | function pcre_dfa_exec(), nor is it supported in UTF-8 mode by the JIT |
| 6442 | optimization of pcre_exec(). If JIT optimization is requested for a |
| 6443 | UTF-8 pattern that contains \C, it will not succeed, and so the match- |
| 6444 | ing will be carried out by the normal interpretive function. |
| 6445 | |
| 6446 | 6. The character escapes \b, \B, \d, \D, \s, \S, \w, and \W correctly |
| 6447 | test characters of any code value, but, by default, the characters that |
| 6448 | PCRE recognizes as digits, spaces, or word characters remain the same |
| 6449 | set as before, all with values less than 256. This remains true even |
| 6450 | when PCRE is built to include Unicode property support, because to do |
| 6451 | otherwise would slow down PCRE in many common cases. Note in particular |
| 6452 | that this applies to \b and \B, because they are defined in terms of \w |
| 6453 | and \W. If you really want to test for a wider sense of, say, "digit", |
| 6454 | you can use explicit Unicode property tests such as \p{Nd}. Alterna- |
| 6455 | tively, if you set the PCRE_UCP option, the way that the character |
| 6456 | escapes work is changed so that Unicode properties are used to deter- |
| 6457 | mine which characters match. There are more details in the section on |
| 6458 | generic character types in the pcrepattern documentation. |
| 6459 | |
| 6460 | 7. Similarly, characters that match the POSIX named character classes |
| 6461 | are all low-valued characters, unless the PCRE_UCP option is set. |
| 6462 | |
| 6463 | 8. However, the horizontal and vertical whitespace matching escapes |
| 6464 | (\h, \H, \v, and \V) do match all the appropriate Unicode characters, |
| 6465 | whether or not PCRE_UCP is set. |
| 6466 | |
| 6467 | 9. Case-insensitive matching applies only to characters whose values |
| 6468 | are less than 128, unless PCRE is built with Unicode property support. |
| 6469 | Even when Unicode property support is available, PCRE still uses its |
| 6470 | own character tables when checking the case of low-valued characters, |
| 6471 | so as not to degrade performance. The Unicode property information is |
| 6472 | used only for characters with higher values. Furthermore, PCRE supports |
| 6473 | case-insensitive matching only when there is a one-to-one mapping |
| 6474 | between a letter's cases. There are a small number of many-to-one map- |
| 6475 | pings in Unicode; these are not supported by PCRE. |
| 6476 | |
| 6477 | |
| 6478 | AUTHOR |
| 6479 | |
| 6480 | Philip Hazel |
| 6481 | University Computing Service |
| 6482 | Cambridge CB2 3QH, England. |
| 6483 | |
| 6484 | |
| 6485 | REVISION |
| 6486 | |
| 6487 | Last updated: 19 October 2011 |
| 6488 | Copyright (c) 1997-2011 University of Cambridge. |
| 6489 | ------------------------------------------------------------------------------ |
| 6490 | |
| 6491 | |
| 6492 | PCREJIT(3) PCREJIT(3) |
| 6493 | |
| 6494 | |
| 6495 | NAME |
| 6496 | PCRE - Perl-compatible regular expressions |
| 6497 | |
| 6498 | |
| 6499 | PCRE JUST-IN-TIME COMPILER SUPPORT |
| 6500 | |
| 6501 | Just-in-time compiling is a heavyweight optimization that can greatly |
| 6502 | speed up pattern matching. However, it comes at the cost of extra pro- |
| 6503 | cessing before the match is performed. Therefore, it is of most benefit |
| 6504 | when the same pattern is going to be matched many times. This does not |
| 6505 | necessarily mean many calls of pcre_exec(); if the pattern is not |
| 6506 | anchored, matching attempts may take place many times at various posi- |
| 6507 | tions in the subject, even for a single call to pcre_exec(). If the |
| 6508 | subject string is very long, it may still pay to use JIT for one-off |
| 6509 | matches. |
| 6510 | |
| 6511 | JIT support applies only to the traditional matching function, |
| 6512 | pcre_exec(). It does not apply when pcre_dfa_exec() is being used. The |
| 6513 | code for this support was written by Zoltan Herczeg. |
| 6514 | |
| 6515 | |
| 6516 | AVAILABILITY OF JIT SUPPORT |
| 6517 | |
| 6518 | JIT support is an optional feature of PCRE. The "configure" option |
| 6519 | --enable-jit (or equivalent CMake option) must be set when PCRE is |
| 6520 | built if you want to use JIT. The support is limited to the following |
| 6521 | hardware platforms: |
| 6522 | |
| 6523 | ARM v5, v7, and Thumb2 |
| 6524 | Intel x86 32-bit and 64-bit |
| 6525 | MIPS 32-bit |
| 6526 | Power PC 32-bit and 64-bit (experimental) |
| 6527 | |
| 6528 | The Power PC support is designated as experimental because it has not |
| 6529 | been fully tested. If --enable-jit is set on an unsupported platform, |
| 6530 | compilation fails. |
| 6531 | |
| 6532 | A program that is linked with PCRE 8.20 or later can tell if JIT sup- |
| 6533 | port is available by calling pcre_config() with the PCRE_CONFIG_JIT |
| 6534 | option. The result is 1 when JIT is available, and 0 otherwise. How- |
| 6535 | ever, a simple program does not need to check this in order to use JIT. |
| 6536 | The API is implemented in a way that falls back to the ordinary PCRE |
| 6537 | code if JIT is not available. |
| 6538 | |
| 6539 | If your program may sometimes be linked with versions of PCRE that are |
| 6540 | older than 8.20, but you want to use JIT when it is available, you can |
| 6541 | test the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT |
| 6542 | macro such as PCRE_CONFIG_JIT, for compile-time control of your code. |
| 6543 | |
| 6544 | |
| 6545 | SIMPLE USE OF JIT |
| 6546 | |
| 6547 | You have to do two things to make use of the JIT support in the sim- |
| 6548 | plest way: |
| 6549 | |
| 6550 | (1) Call pcre_study() with the PCRE_STUDY_JIT_COMPILE option for |
| 6551 | each compiled pattern, and pass the resulting pcre_extra block to |
| 6552 | pcre_exec(). |
| 6553 | |
| 6554 | (2) Use pcre_free_study() to free the pcre_extra block when it is |
| 6555 | no longer needed instead of just freeing it yourself. This |
| 6556 | ensures that any JIT data is also freed. |
| 6557 | |
| 6558 | For a program that may be linked with pre-8.20 versions of PCRE, you |
| 6559 | can insert |
| 6560 | |
| 6561 | #ifndef PCRE_STUDY_JIT_COMPILE |
| 6562 | #define PCRE_STUDY_JIT_COMPILE 0 |
| 6563 | #endif |
| 6564 | |
| 6565 | so that no option is passed to pcre_study(), and then use something |
| 6566 | like this to free the study data: |
| 6567 | |
| 6568 | #ifdef PCRE_CONFIG_JIT |
| 6569 | pcre_free_study(study_ptr); |
| 6570 | #else |
| 6571 | pcre_free(study_ptr); |
| 6572 | #endif |
| 6573 | |
| 6574 | In some circumstances you may need to call additional functions. These |
| 6575 | are described in the section entitled "Controlling the JIT stack" |
| 6576 | below. |
| 6577 | |
| 6578 | If JIT support is not available, PCRE_STUDY_JIT_COMPILE is ignored, and |
| 6579 | no JIT data is set up. Otherwise, the compiled pattern is passed to the |
| 6580 | JIT compiler, which turns it into machine code that executes much |
| 6581 | faster than the normal interpretive code. When pcre_exec() is passed a |
| 6582 | pcre_extra block containing a pointer to JIT code, it obeys that |
| 6583 | instead of the normal code. The result is identical, but the code runs |
| 6584 | much faster. |
| 6585 | |
| 6586 | There are some pcre_exec() options that are not supported for JIT exe- |
| 6587 | cution. There are also some pattern items that JIT cannot handle. |
| 6588 | Details are given below. In both cases, execution automatically falls |
| 6589 | back to the interpretive code. |
| 6590 | |
| 6591 | If the JIT compiler finds an unsupported item, no JIT data is gener- |
| 6592 | ated. You can find out if JIT execution is available after studying a |
| 6593 | pattern by calling pcre_fullinfo() with the PCRE_INFO_JIT option. A |
| 6594 | result of 1 means that JIT compilation was successful. A result of 0 |
| 6595 | means that JIT support is not available, or the pattern was not studied |
| 6596 | with PCRE_STUDY_JIT_COMPILE, or the JIT compiler was not able to handle |
| 6597 | the pattern. |
| 6598 | |
| 6599 | Once a pattern has been studied, with or without JIT, it can be used as |
| 6600 | many times as you like for matching different subject strings. |
| 6601 | |
| 6602 | |
| 6603 | UNSUPPORTED OPTIONS AND PATTERN ITEMS |
| 6604 | |
| 6605 | The only pcre_exec() options that are supported for JIT execution are |
| 6606 | PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, and |
| 6607 | PCRE_NOTEMPTY_ATSTART. Note in particular that partial matching is not |
| 6608 | supported. |
| 6609 | |
| 6610 | The unsupported pattern items are: |
| 6611 | |
| 6612 | \C match a single byte; not supported in UTF-8 mode |
| 6613 | (?Cn) callouts |
| 6614 | (*COMMIT) ) |
| 6615 | (*MARK) ) |
| 6616 | (*PRUNE) ) the backtracking control verbs |
| 6617 | (*SKIP) ) |
| 6618 | (*THEN) ) |
| 6619 | |
| 6620 | Support for some of these may be added in future. |
| 6621 | |
| 6622 | |
| 6623 | RETURN VALUES FROM JIT EXECUTION |
| 6624 | |
| 6625 | When a pattern is matched using JIT execution, the return values are |
| 6626 | the same as those given by the interpretive pcre_exec() code, with the |
| 6627 | addition of one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means |
| 6628 | that the memory used for the JIT stack was insufficient. See "Control- |
| 6629 | ling the JIT stack" below for a discussion of JIT stack usage. For com- |
| 6630 | patibility with the interpretive pcre_exec() code, no more than two- |
| 6631 | thirds of the ovector argument is used for passing back captured sub- |
| 6632 | strings. |
| 6633 | |
| 6634 | The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if |
| 6635 | searching a very large pattern tree goes on for too long, as it is in |
| 6636 | the same circumstance when JIT is not used, but the details of exactly |
| 6637 | what is counted are not the same. The PCRE_ERROR_RECURSIONLIMIT error |
| 6638 | code is never returned by JIT execution. |
| 6639 | |
| 6640 | |
| 6641 | SAVING AND RESTORING COMPILED PATTERNS |
| 6642 | |
| 6643 | The code that is generated by the JIT compiler is architecture-spe- |
| 6644 | cific, and is also position dependent. For those reasons it cannot be |
| 6645 | saved (in a file or database) and restored later like the bytecode and |
| 6646 | other data of a compiled pattern. Saving and restoring compiled pat- |
| 6647 | terns is not something many people do. More detail about this facility |
| 6648 | is given in the pcreprecompile documentation. It should be possible to |
| 6649 | run pcre_study() on a saved and restored pattern, and thereby recreate |
| 6650 | the JIT data, but because JIT compilation uses significant resources, |
| 6651 | it is probably not worth doing this; you might as well recompile the |
| 6652 | original pattern. |
| 6653 | |
| 6654 | |
| 6655 | CONTROLLING THE JIT STACK |
| 6656 | |
| 6657 | When the compiled JIT code runs, it needs a block of memory to use as a |
| 6658 | stack. By default, it uses 32K on the machine stack. However, some |
| 6659 | large or complicated patterns need more than this. The error |
| 6660 | PCRE_ERROR_JIT_STACKLIMIT is given when there is not enough stack. |
| 6661 | Three functions are provided for managing blocks of memory for use as |
| 6662 | JIT stacks. There is further discussion about the use of JIT stacks in |
| 6663 | the section entitled "JIT stack FAQ" below. |
| 6664 | |
| 6665 | The pcre_jit_stack_alloc() function creates a JIT stack. Its arguments |
| 6666 | are a starting size and a maximum size, and it returns a pointer to an |
| 6667 | opaque structure of type pcre_jit_stack, or NULL if there is an error. |
| 6668 | The pcre_jit_stack_free() function can be used to free a stack that is |
| 6669 | no longer needed. (For the technically minded: the address space is |
| 6670 | allocated by mmap or VirtualAlloc.) |
| 6671 | |
| 6672 | JIT uses far less memory for recursion than the interpretive code, and |
| 6673 | a maximum stack size of 512K to 1M should be more than enough for any |
| 6674 | pattern. |
| 6675 | |
| 6676 | The pcre_assign_jit_stack() function specifies which stack JIT code |
| 6677 | should use. Its arguments are as follows: |
| 6678 | |
| 6679 | pcre_extra *extra |
| 6680 | pcre_jit_callback callback |
| 6681 | void *data |
| 6682 | |
| 6683 | The extra argument must be the result of studying a pattern with |
| 6684 | PCRE_STUDY_JIT_COMPILE. There are three cases for the values of the |
| 6685 | other two options: |
| 6686 | |
| 6687 | (1) If callback is NULL and data is NULL, an internal 32K block |
| 6688 | on the machine stack is used. |
| 6689 | |
| 6690 | (2) If callback is NULL and data is not NULL, data must be |
| 6691 | a valid JIT stack, the result of calling pcre_jit_stack_alloc(). |
| 6692 | |
| 6693 | (3) If callback not NULL, it must point to a function that is called |
| 6694 | with data as an argument at the start of matching, in order to |
| 6695 | set up a JIT stack. If the result is NULL, the internal 32K stack |
| 6696 | is used; otherwise the return value must be a valid JIT stack, |
| 6697 | the result of calling pcre_jit_stack_alloc(). |
| 6698 | |
| 6699 | You may safely assign the same JIT stack to more than one pattern, as |
| 6700 | long as they are all matched sequentially in the same thread. In a mul- |
| 6701 | tithread application, each thread must use its own JIT stack. |
| 6702 | |
| 6703 | Strictly speaking, even more is allowed. You can assign the same stack |
| 6704 | to any number of patterns as long as they are not used for matching by |
| 6705 | multiple threads at the same time. For example, you can assign the same |
| 6706 | stack to all compiled patterns, and use a global mutex in the callback |
| 6707 | to wait until the stack is available for use. However, this is an inef- |
| 6708 | ficient solution, and not recommended. |
| 6709 | |
| 6710 | This is a suggestion for how a typical multithreaded program might |
| 6711 | operate: |
| 6712 | |
| 6713 | During thread initalization |
| 6714 | thread_local_var = pcre_jit_stack_alloc(...) |
| 6715 | |
| 6716 | During thread exit |
| 6717 | pcre_jit_stack_free(thread_local_var) |
| 6718 | |
| 6719 | Use a one-line callback function |
| 6720 | return thread_local_var |
| 6721 | |
| 6722 | All the functions described in this section do nothing if JIT is not |
| 6723 | available, and pcre_assign_jit_stack() does nothing unless the extra |
| 6724 | argument is non-NULL and points to a pcre_extra block that is the |
| 6725 | result of a successful study with PCRE_STUDY_JIT_COMPILE. |
| 6726 | |
| 6727 | |
| 6728 | JIT STACK FAQ |
| 6729 | |
| 6730 | (1) Why do we need JIT stacks? |
| 6731 | |
| 6732 | PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack |
| 6733 | where the local data of the current node is pushed before checking its |
| 6734 | child nodes. Allocating real machine stack on some platforms is diffi- |
| 6735 | cult. For example, the stack chain needs to be updated every time if we |
| 6736 | extend the stack on PowerPC. Although it is possible, its updating |
| 6737 | time overhead decreases performance. So we do the recursion in memory. |
| 6738 | |
| 6739 | (2) Why don't we simply allocate blocks of memory with malloc()? |
| 6740 | |
| 6741 | Modern operating systems have a nice feature: they can reserve an |
| 6742 | address space instead of allocating memory. We can safely allocate mem- |
| 6743 | ory pages inside this address space, so the stack could grow without |
| 6744 | moving memory data (this is important because of pointers). Thus we can |
| 6745 | allocate 1M address space, and use only a single memory page (usually |
| 6746 | 4K) if that is enough. However, we can still grow up to 1M anytime if |
| 6747 | needed. |
| 6748 | |
| 6749 | (3) Who "owns" a JIT stack? |
| 6750 | |
| 6751 | The owner of the stack is the user program, not the JIT studied pattern |
| 6752 | or anything else. The user program must ensure that if a stack is used |
| 6753 | by pcre_exec(), (that is, it is assigned to the pattern currently run- |
| 6754 | ning), that stack must not be used by any other threads (to avoid over- |
| 6755 | writing the same memory area). The best practice for multithreaded pro- |
| 6756 | grams is to allocate a stack for each thread, and return this stack |
| 6757 | through the JIT callback function. |
| 6758 | |
| 6759 | (4) When should a JIT stack be freed? |
| 6760 | |
| 6761 | You can free a JIT stack at any time, as long as it will not be used by |
| 6762 | pcre_exec() again. When you assign the stack to a pattern, only a |
| 6763 | pointer is set. There is no reference counting or any other magic. You |
| 6764 | can free the patterns and stacks in any order, anytime. Just do not |
| 6765 | call pcre_exec() with a pattern pointing to an already freed stack, as |
| 6766 | that will cause SEGFAULT. (Also, do not free a stack currently used by |
| 6767 | pcre_exec() in another thread). You can also replace the stack for a |
| 6768 | pattern at any time. You can even free the previous stack before |
| 6769 | assigning a replacement. |
| 6770 | |
| 6771 | (5) Should I allocate/free a stack every time before/after calling |
| 6772 | pcre_exec()? |
| 6773 | |
| 6774 | No, because this is too costly in terms of resources. However, you |
| 6775 | could implement some clever idea which release the stack if it is not |
| 6776 | used in let's say two minutes. The JIT callback can help to achive this |
| 6777 | without keeping a list of the currently JIT studied patterns. |
| 6778 | |
| 6779 | (6) OK, the stack is for long term memory allocation. But what happens |
| 6780 | if a pattern causes stack overflow with a stack of 1M? Is that 1M kept |
| 6781 | until the stack is freed? |
| 6782 | |
| 6783 | Especially on embedded sytems, it might be a good idea to release mem- |
| 6784 | ory sometimes without freeing the stack. There is no API for this at |
| 6785 | the moment. Probably a function call which returns with the currently |
| 6786 | allocated memory for any stack and another which allows releasing mem- |
| 6787 | ory (shrinking the stack) would be a good idea if someone needs this. |
| 6788 | |
| 6789 | (7) This is too much of a headache. Isn't there any better solution for |
| 6790 | JIT stack handling? |
| 6791 | |
| 6792 | No, thanks to Windows. If POSIX threads were used everywhere, we could |
| 6793 | throw out this complicated API. |
| 6794 | |
| 6795 | |
| 6796 | EXAMPLE CODE |
| 6797 | |
| 6798 | This is a single-threaded example that specifies a JIT stack without |
| 6799 | using a callback. |
| 6800 | |
| 6801 | int rc; |
| 6802 | int ovector[30]; |
| 6803 | pcre *re; |
| 6804 | pcre_extra *extra; |
| 6805 | pcre_jit_stack *jit_stack; |
| 6806 | |
| 6807 | re = pcre_compile(pattern, 0, &error, &erroffset, NULL); |
| 6808 | /* Check for errors */ |
| 6809 | extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error); |
| 6810 | jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024); |
| 6811 | /* Check for error (NULL) */ |
| 6812 | pcre_assign_jit_stack(extra, NULL, jit_stack); |
| 6813 | rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30); |
| 6814 | /* Check results */ |
| 6815 | pcre_free(re); |
| 6816 | pcre_free_study(extra); |
| 6817 | pcre_jit_stack_free(jit_stack); |
| 6818 | |
| 6819 | |
| 6820 | SEE ALSO |
| 6821 | |
| 6822 | pcreapi(3) |
| 6823 | |
| 6824 | |
| 6825 | AUTHOR |
| 6826 | |
| 6827 | Philip Hazel (FAQ by Zoltan Herczeg) |
| 6828 | University Computing Service |
| 6829 | Cambridge CB2 3QH, England. |
| 6830 | |
| 6831 | |
| 6832 | REVISION |
| 6833 | |
| 6834 | Last updated: 26 November 2011 |
| 6835 | Copyright (c) 1997-2011 University of Cambridge. |
| 6836 | ------------------------------------------------------------------------------ |
| 6837 | |
| 6838 | |
| 6839 | PCREPARTIAL(3) PCREPARTIAL(3) |
| 6840 | |
| 6841 | |
| 6842 | NAME |
| 6843 | PCRE - Perl-compatible regular expressions |
| 6844 | |
| 6845 | |
| 6846 | PARTIAL MATCHING IN PCRE |
| 6847 | |
| 6848 | In normal use of PCRE, if the subject string that is passed to |
| 6849 | pcre_exec() or pcre_dfa_exec() matches as far as it goes, but is too |
| 6850 | short to match the entire pattern, PCRE_ERROR_NOMATCH is returned. |
| 6851 | There are circumstances where it might be helpful to distinguish this |
| 6852 | case from other cases in which there is no match. |
| 6853 | |
| 6854 | Consider, for example, an application where a human is required to type |
| 6855 | in data for a field with specific formatting requirements. An example |
| 6856 | might be a date in the form ddmmmyy, defined by this pattern: |
| 6857 | |
| 6858 | ^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$ |
| 6859 | |
| 6860 | If the application sees the user's keystrokes one by one, and can check |
| 6861 | that what has been typed so far is potentially valid, it is able to |
| 6862 | raise an error as soon as a mistake is made, by beeping and not |
| 6863 | reflecting the character that has been typed, for example. This immedi- |
| 6864 | ate feedback is likely to be a better user interface than a check that |
| 6865 | is delayed until the entire string has been entered. Partial matching |
| 6866 | can also be useful when the subject string is very long and is not all |
| 6867 | available at once. |
| 6868 | |
| 6869 | PCRE supports partial matching by means of the PCRE_PARTIAL_SOFT and |
| 6870 | PCRE_PARTIAL_HARD options, which can be set when calling pcre_exec() or |
| 6871 | pcre_dfa_exec(). For backwards compatibility, PCRE_PARTIAL is a synonym |
| 6872 | for PCRE_PARTIAL_SOFT. The essential difference between the two options |
| 6873 | is whether or not a partial match is preferred to an alternative com- |
| 6874 | plete match, though the details differ between the two matching func- |
| 6875 | tions. If both options are set, PCRE_PARTIAL_HARD takes precedence. |
| 6876 | |
| 6877 | Setting a partial matching option for pcre_exec() disables the use of |
| 6878 | any just-in-time code that was set up by calling pcre_study() with the |
| 6879 | PCRE_STUDY_JIT_COMPILE option. It also disables two of PCRE's standard |
| 6880 | optimizations. PCRE remembers the last literal byte in a pattern, and |
| 6881 | abandons matching immediately if such a byte is not present in the sub- |
| 6882 | ject string. This optimization cannot be used for a subject string that |
| 6883 | might match only partially. If the pattern was studied, PCRE knows the |
| 6884 | minimum length of a matching string, and does not bother to run the |
| 6885 | matching function on shorter strings. This optimization is also dis- |
| 6886 | abled for partial matching. |
| 6887 | |
| 6888 | |
| 6889 | PARTIAL MATCHING USING pcre_exec() |
| 6890 | |
| 6891 | A partial match occurs during a call to pcre_exec() when the end of the |
| 6892 | subject string is reached successfully, but matching cannot continue |
| 6893 | because more characters are needed. However, at least one character in |
| 6894 | the subject must have been inspected. This character need not form part |
| 6895 | of the final matched string; lookbehind assertions and the \K escape |
| 6896 | sequence provide ways of inspecting characters before the start of a |
| 6897 | matched substring. The requirement for inspecting at least one charac- |
| 6898 | ter exists because an empty string can always be matched; without such |
| 6899 | a restriction there would always be a partial match of an empty string |
| 6900 | at the end of the subject. |
| 6901 | |
| 6902 | If there are at least two slots in the offsets vector when pcre_exec() |
| 6903 | returns with a partial match, the first slot is set to the offset of |
| 6904 | the earliest character that was inspected when the partial match was |
| 6905 | found. For convenience, the second offset points to the end of the sub- |
| 6906 | ject so that a substring can easily be identified. |
| 6907 | |
| 6908 | For the majority of patterns, the first offset identifies the start of |
| 6909 | the partially matched string. However, for patterns that contain look- |
| 6910 | behind assertions, or \K, or begin with \b or \B, earlier characters |
| 6911 | have been inspected while carrying out the match. For example: |
| 6912 | |
| 6913 | /(?<=abc)123/ |
| 6914 | |
| 6915 | This pattern matches "123", but only if it is preceded by "abc". If the |
| 6916 | subject string is "xyzabc12", the offsets after a partial match are for |
| 6917 | the substring "abc12", because all these characters are needed if |
| 6918 | another match is tried with extra characters added to the subject. |
| 6919 | |
| 6920 | What happens when a partial match is identified depends on which of the |
| 6921 | two partial matching options are set. |
| 6922 | |
| 6923 | PCRE_PARTIAL_SOFT with pcre_exec() |
| 6924 | |
| 6925 | If PCRE_PARTIAL_SOFT is set when pcre_exec() identifies a partial |
| 6926 | match, the partial match is remembered, but matching continues as nor- |
| 6927 | mal, and other alternatives in the pattern are tried. If no complete |
| 6928 | match can be found, pcre_exec() returns PCRE_ERROR_PARTIAL instead of |
| 6929 | PCRE_ERROR_NOMATCH. |
| 6930 | |
| 6931 | This option is "soft" because it prefers a complete match over a par- |
| 6932 | tial match. All the various matching items in a pattern behave as if |
| 6933 | the subject string is potentially complete. For example, \z, \Z, and $ |
| 6934 | match at the end of the subject, as normal, and for \b and \B the end |
| 6935 | of the subject is treated as a non-alphanumeric. |
| 6936 | |
| 6937 | If there is more than one partial match, the first one that was found |
| 6938 | provides the data that is returned. Consider this pattern: |
| 6939 | |
| 6940 | /123\w+X|dogY/ |
| 6941 | |
| 6942 | If this is matched against the subject string "abc123dog", both alter- |
| 6943 | natives fail to match, but the end of the subject is reached during |
| 6944 | matching, so PCRE_ERROR_PARTIAL is returned. The offsets are set to 3 |
| 6945 | and 9, identifying "123dog" as the first partial match that was found. |
| 6946 | (In this example, there are two partial matches, because "dog" on its |
| 6947 | own partially matches the second alternative.) |
| 6948 | |
| 6949 | PCRE_PARTIAL_HARD with pcre_exec() |
| 6950 | |
| 6951 | If PCRE_PARTIAL_HARD is set for pcre_exec(), it returns PCRE_ERROR_PAR- |
| 6952 | TIAL as soon as a partial match is found, without continuing to search |
| 6953 | for possible complete matches. This option is "hard" because it prefers |
| 6954 | an earlier partial match over a later complete match. For this reason, |
| 6955 | the assumption is made that the end of the supplied subject string may |
| 6956 | not be the true end of the available data, and so, if \z, \Z, \b, \B, |
| 6957 | or $ are encountered at the end of the subject, the result is |
| 6958 | PCRE_ERROR_PARTIAL. |
| 6959 | |
| 6960 | Setting PCRE_PARTIAL_HARD also affects the way pcre_exec() checks UTF-8 |
| 6961 | subject strings for validity. Normally, an invalid UTF-8 sequence |
| 6962 | causes the error PCRE_ERROR_BADUTF8. However, in the special case of a |
| 6963 | truncated UTF-8 character at the end of the subject, PCRE_ERROR_SHORT- |
| 6964 | UTF8 is returned when PCRE_PARTIAL_HARD is set. |
| 6965 | |
| 6966 | Comparing hard and soft partial matching |
| 6967 | |
| 6968 | The difference between the two partial matching options can be illus- |
| 6969 | trated by a pattern such as: |
| 6970 | |
| 6971 | /dog(sbody)?/ |
| 6972 | |
| 6973 | This matches either "dog" or "dogsbody", greedily (that is, it prefers |
| 6974 | the longer string if possible). If it is matched against the string |
| 6975 | "dog" with PCRE_PARTIAL_SOFT, it yields a complete match for "dog". |
| 6976 | However, if PCRE_PARTIAL_HARD is set, the result is PCRE_ERROR_PARTIAL. |
| 6977 | On the other hand, if the pattern is made ungreedy the result is dif- |
| 6978 | ferent: |
| 6979 | |
| 6980 | /dog(sbody)??/ |
| 6981 | |
| 6982 | In this case the result is always a complete match because pcre_exec() |
| 6983 | finds that first, and it never continues after finding a match. It |
| 6984 | might be easier to follow this explanation by thinking of the two pat- |
| 6985 | terns like this: |
| 6986 | |
| 6987 | /dog(sbody)?/ is the same as /dogsbody|dog/ |
| 6988 | /dog(sbody)??/ is the same as /dog|dogsbody/ |
| 6989 | |
| 6990 | The second pattern will never match "dogsbody" when pcre_exec() is |
| 6991 | used, because it will always find the shorter match first. |
| 6992 | |
| 6993 | |
| 6994 | PARTIAL MATCHING USING pcre_dfa_exec() |
| 6995 | |
| 6996 | The pcre_dfa_exec() function moves along the subject string character |
| 6997 | by character, without backtracking, searching for all possible matches |
| 6998 | simultaneously. If the end of the subject is reached before the end of |
| 6999 | the pattern, there is the possibility of a partial match, again pro- |
| 7000 | vided that at least one character has been inspected. |
| 7001 | |
| 7002 | When PCRE_PARTIAL_SOFT is set, PCRE_ERROR_PARTIAL is returned only if |
| 7003 | there have been no complete matches. Otherwise, the complete matches |
| 7004 | are returned. However, if PCRE_PARTIAL_HARD is set, a partial match |
| 7005 | takes precedence over any complete matches. The portion of the string |
| 7006 | that was inspected when the longest partial match was found is set as |
| 7007 | the first matching string, provided there are at least two slots in the |
| 7008 | offsets vector. |
| 7009 | |
| 7010 | Because pcre_dfa_exec() always searches for all possible matches, and |
| 7011 | there is no difference between greedy and ungreedy repetition, its be- |
| 7012 | haviour is different from pcre_exec when PCRE_PARTIAL_HARD is set. Con- |
| 7013 | sider the string "dog" matched against the ungreedy pattern shown |
| 7014 | above: |
| 7015 | |
| 7016 | /dog(sbody)??/ |
| 7017 | |
| 7018 | Whereas pcre_exec() stops as soon as it finds the complete match for |
| 7019 | "dog", pcre_dfa_exec() also finds the partial match for "dogsbody", and |
| 7020 | so returns that when PCRE_PARTIAL_HARD is set. |
| 7021 | |
| 7022 | |
| 7023 | PARTIAL MATCHING AND WORD BOUNDARIES |
| 7024 | |
| 7025 | If a pattern ends with one of sequences \b or \B, which test for word |
| 7026 | boundaries, partial matching with PCRE_PARTIAL_SOFT can give counter- |
| 7027 | intuitive results. Consider this pattern: |
| 7028 | |
| 7029 | /\bcat\b/ |
| 7030 | |
| 7031 | This matches "cat", provided there is a word boundary at either end. If |
| 7032 | the subject string is "the cat", the comparison of the final "t" with a |
| 7033 | following character cannot take place, so a partial match is found. |
| 7034 | However, pcre_exec() carries on with normal matching, which matches \b |
| 7035 | at the end of the subject when the last character is a letter, thus |
| 7036 | finding a complete match. The result, therefore, is not PCRE_ERROR_PAR- |
| 7037 | TIAL. The same thing happens with pcre_dfa_exec(), because it also |
| 7038 | finds the complete match. |
| 7039 | |
| 7040 | Using PCRE_PARTIAL_HARD in this case does yield PCRE_ERROR_PARTIAL, |
| 7041 | because then the partial match takes precedence. |
| 7042 | |
| 7043 | |
| 7044 | FORMERLY RESTRICTED PATTERNS |
| 7045 | |
| 7046 | For releases of PCRE prior to 8.00, because of the way certain internal |
| 7047 | optimizations were implemented in the pcre_exec() function, the |
| 7048 | PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be |
| 7049 | used with all patterns. From release 8.00 onwards, the restrictions no |
| 7050 | longer apply, and partial matching with pcre_exec() can be requested |
| 7051 | for any pattern. |
| 7052 | |
| 7053 | Items that were formerly restricted were repeated single characters and |
| 7054 | repeated metasequences. If PCRE_PARTIAL was set for a pattern that did |
| 7055 | not conform to the restrictions, pcre_exec() returned the error code |
| 7056 | PCRE_ERROR_BADPARTIAL (-13). This error code is no longer in use. The |
| 7057 | PCRE_INFO_OKPARTIAL call to pcre_fullinfo() to find out if a compiled |
| 7058 | pattern can be used for partial matching now always returns 1. |
| 7059 | |
| 7060 | |
| 7061 | EXAMPLE OF PARTIAL MATCHING USING PCRETEST |
| 7062 | |
| 7063 | If the escape sequence \P is present in a pcretest data line, the |
| 7064 | PCRE_PARTIAL_SOFT option is used for the match. Here is a run of |
| 7065 | pcretest that uses the date example quoted above: |
| 7066 | |
| 7067 | re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/ |
| 7068 | data> 25jun04\P |
| 7069 | 0: 25jun04 |
| 7070 | 1: jun |
| 7071 | data> 25dec3\P |
| 7072 | Partial match: 23dec3 |
| 7073 | data> 3ju\P |
| 7074 | Partial match: 3ju |
| 7075 | data> 3juj\P |
| 7076 | No match |
| 7077 | data> j\P |
| 7078 | No match |
| 7079 | |
| 7080 | The first data string is matched completely, so pcretest shows the |
| 7081 | matched substrings. The remaining four strings do not match the com- |
| 7082 | plete pattern, but the first two are partial matches. Similar output is |
| 7083 | obtained when pcre_dfa_exec() is used. |
| 7084 | |
| 7085 | If the escape sequence \P is present more than once in a pcretest data |
| 7086 | line, the PCRE_PARTIAL_HARD option is set for the match. |
| 7087 | |
| 7088 | |
| 7089 | MULTI-SEGMENT MATCHING WITH pcre_dfa_exec() |
| 7090 | |
| 7091 | When a partial match has been found using pcre_dfa_exec(), it is possi- |
| 7092 | ble to continue the match by providing additional subject data and |
| 7093 | calling pcre_dfa_exec() again with the same compiled regular expres- |
| 7094 | sion, this time setting the PCRE_DFA_RESTART option. You must pass the |
| 7095 | same working space as before, because this is where details of the pre- |
| 7096 | vious partial match are stored. Here is an example using pcretest, |
| 7097 | using the \R escape sequence to set the PCRE_DFA_RESTART option (\D |
| 7098 | specifies the use of pcre_dfa_exec()): |
| 7099 | |
| 7100 | re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/ |
| 7101 | data> 23ja\P\D |
| 7102 | Partial match: 23ja |
| 7103 | data> n05\R\D |
| 7104 | 0: n05 |
| 7105 | |
| 7106 | The first call has "23ja" as the subject, and requests partial match- |
| 7107 | ing; the second call has "n05" as the subject for the continued |
| 7108 | (restarted) match. Notice that when the match is complete, only the |
| 7109 | last part is shown; PCRE does not retain the previously partially- |
| 7110 | matched string. It is up to the calling program to do that if it needs |
| 7111 | to. |
| 7112 | |
| 7113 | You can set the PCRE_PARTIAL_SOFT or PCRE_PARTIAL_HARD options with |
| 7114 | PCRE_DFA_RESTART to continue partial matching over multiple segments. |
| 7115 | This facility can be used to pass very long subject strings to |
| 7116 | pcre_dfa_exec(). |
| 7117 | |
| 7118 | |
| 7119 | MULTI-SEGMENT MATCHING WITH pcre_exec() |
| 7120 | |
| 7121 | From release 8.00, pcre_exec() can also be used to do multi-segment |
| 7122 | matching. Unlike pcre_dfa_exec(), it is not possible to restart the |
| 7123 | previous match with a new segment of data. Instead, new data must be |
| 7124 | added to the previous subject string, and the entire match re-run, |
| 7125 | starting from the point where the partial match occurred. Earlier data |
| 7126 | can be discarded. It is best to use PCRE_PARTIAL_HARD in this situa- |
| 7127 | tion, because it does not treat the end of a segment as the end of the |
| 7128 | subject when matching \z, \Z, \b, \B, and $. Consider an unanchored |
| 7129 | pattern that matches dates: |
| 7130 | |
| 7131 | re> /\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d/ |
| 7132 | data> The date is 23ja\P\P |
| 7133 | Partial match: 23ja |
| 7134 | |
| 7135 | At this stage, an application could discard the text preceding "23ja", |
| 7136 | add on text from the next segment, and call pcre_exec() again. Unlike |
| 7137 | pcre_dfa_exec(), the entire matching string must always be available, |
| 7138 | and the complete matching process occurs for each call, so more memory |
| 7139 | and more processing time is needed. |
| 7140 | |
| 7141 | Note: If the pattern contains lookbehind assertions, or \K, or starts |
| 7142 | with \b or \B, the string that is returned for a partial match will |
| 7143 | include characters that precede the partially matched string itself, |
| 7144 | because these must be retained when adding on more characters for a |
| 7145 | subsequent matching attempt. |
| 7146 | |
| 7147 | |
| 7148 | ISSUES WITH MULTI-SEGMENT MATCHING |
| 7149 | |
| 7150 | Certain types of pattern may give problems with multi-segment matching, |
| 7151 | whichever matching function is used. |
| 7152 | |
| 7153 | 1. If the pattern contains a test for the beginning of a line, you need |
| 7154 | to pass the PCRE_NOTBOL option when the subject string for any call |
| 7155 | does start at the beginning of a line. There is also a PCRE_NOTEOL |
| 7156 | option, but in practice when doing multi-segment matching you should be |
| 7157 | using PCRE_PARTIAL_HARD, which includes the effect of PCRE_NOTEOL. |
| 7158 | |
| 7159 | 2. Lookbehind assertions at the start of a pattern are catered for in |
| 7160 | the offsets that are returned for a partial match. However, in theory, |
| 7161 | a lookbehind assertion later in the pattern could require even earlier |
| 7162 | characters to be inspected, and it might not have been reached when a |
| 7163 | partial match occurs. This is probably an extremely unlikely case; you |
| 7164 | could guard against it to a certain extent by always including extra |
| 7165 | characters at the start. |
| 7166 | |
| 7167 | 3. Matching a subject string that is split into multiple segments may |
| 7168 | not always produce exactly the same result as matching over one single |
| 7169 | long string, especially when PCRE_PARTIAL_SOFT is used. The section |
| 7170 | "Partial Matching and Word Boundaries" above describes an issue that |
| 7171 | arises if the pattern ends with \b or \B. Another kind of difference |
| 7172 | may occur when there are multiple matching possibilities, because (for |
| 7173 | PCRE_PARTIAL_SOFT) a partial match result is given only when there are |
| 7174 | no completed matches. This means that as soon as the shortest match has |
| 7175 | been found, continuation to a new subject segment is no longer possi- |
| 7176 | ble. Consider again this pcretest example: |
| 7177 | |
| 7178 | re> /dog(sbody)?/ |
| 7179 | data> dogsb\P |
| 7180 | 0: dog |
| 7181 | data> do\P\D |
| 7182 | Partial match: do |
| 7183 | data> gsb\R\P\D |
| 7184 | 0: g |
| 7185 | data> dogsbody\D |
| 7186 | 0: dogsbody |
| 7187 | 1: dog |
| 7188 | |
| 7189 | The first data line passes the string "dogsb" to pcre_exec(), setting |
| 7190 | the PCRE_PARTIAL_SOFT option. Although the string is a partial match |
| 7191 | for "dogsbody", the result is not PCRE_ERROR_PARTIAL, because the |
| 7192 | shorter string "dog" is a complete match. Similarly, when the subject |
| 7193 | is presented to pcre_dfa_exec() in several parts ("do" and "gsb" being |
| 7194 | the first two) the match stops when "dog" has been found, and it is not |
| 7195 | possible to continue. On the other hand, if "dogsbody" is presented as |
| 7196 | a single string, pcre_dfa_exec() finds both matches. |
| 7197 | |
| 7198 | Because of these problems, it is best to use PCRE_PARTIAL_HARD when |
| 7199 | matching multi-segment data. The example above then behaves differ- |
| 7200 | ently: |
| 7201 | |
| 7202 | re> /dog(sbody)?/ |
| 7203 | data> dogsb\P\P |
| 7204 | Partial match: dogsb |
| 7205 | data> do\P\D |
| 7206 | Partial match: do |
| 7207 | data> gsb\R\P\P\D |
| 7208 | Partial match: gsb |
| 7209 | |
| 7210 | 4. Patterns that contain alternatives at the top level which do not all |
| 7211 | start with the same pattern item may not work as expected when |
| 7212 | PCRE_DFA_RESTART is used with pcre_dfa_exec(). For example, consider |
| 7213 | this pattern: |
| 7214 | |
| 7215 | 1234|3789 |
| 7216 | |
| 7217 | If the first part of the subject is "ABC123", a partial match of the |
| 7218 | first alternative is found at offset 3. There is no partial match for |
| 7219 | the second alternative, because such a match does not start at the same |
| 7220 | point in the subject string. Attempting to continue with the string |
| 7221 | "7890" does not yield a match because only those alternatives that |
| 7222 | match at one point in the subject are remembered. The problem arises |
| 7223 | because the start of the second alternative matches within the first |
| 7224 | alternative. There is no problem with anchored patterns or patterns |
| 7225 | such as: |
| 7226 | |
| 7227 | 1234|ABCD |
| 7228 | |
| 7229 | where no string can be a partial match for both alternatives. This is |
| 7230 | not a problem if pcre_exec() is used, because the entire match has to |
| 7231 | be rerun each time: |
| 7232 | |
| 7233 | re> /1234|3789/ |
| 7234 | data> ABC123\P\P |
| 7235 | Partial match: 123 |
| 7236 | data> 1237890 |
| 7237 | 0: 3789 |
| 7238 | |
| 7239 | Of course, instead of using PCRE_DFA_RESTART, the same technique of re- |
| 7240 | running the entire match can also be used with pcre_dfa_exec(). Another |
| 7241 | possibility is to work with two buffers. If a partial match at offset n |
| 7242 | in the first buffer is followed by "no match" when PCRE_DFA_RESTART is |
| 7243 | used on the second buffer, you can then try a new match starting at |
| 7244 | offset n+1 in the first buffer. |
| 7245 | |
| 7246 | |
| 7247 | AUTHOR |
| 7248 | |
| 7249 | Philip Hazel |
| 7250 | University Computing Service |
| 7251 | Cambridge CB2 3QH, England. |
| 7252 | |
| 7253 | |
| 7254 | REVISION |
| 7255 | |
| 7256 | Last updated: 26 August 2011 |
| 7257 | Copyright (c) 1997-2011 University of Cambridge. |
| 7258 | ------------------------------------------------------------------------------ |
| 7259 | |
| 7260 | |
| 7261 | PCREPRECOMPILE(3) PCREPRECOMPILE(3) |
| 7262 | |
| 7263 | |
| 7264 | NAME |
| 7265 | PCRE - Perl-compatible regular expressions |
| 7266 | |
| 7267 | |
| 7268 | SAVING AND RE-USING PRECOMPILED PCRE PATTERNS |
| 7269 | |
| 7270 | If you are running an application that uses a large number of regular |
| 7271 | expression patterns, it may be useful to store them in a precompiled |
| 7272 | form instead of having to compile them every time the application is |
| 7273 | run. If you are not using any private character tables (see the |
| 7274 | pcre_maketables() documentation), this is relatively straightforward. |
| 7275 | If you are using private tables, it is a little bit more complicated. |
| 7276 | However, if you are using the just-in-time optimization feature of |
| 7277 | pcre_study(), it is not possible to save and reload the JIT data. |
| 7278 | |
| 7279 | If you save compiled patterns to a file, you can copy them to a differ- |
| 7280 | ent host and run them there. This works even if the new host has the |
| 7281 | opposite endianness to the one on which the patterns were compiled. |
| 7282 | There may be a small performance penalty, but it should be insignifi- |
| 7283 | cant. However, compiling regular expressions with one version of PCRE |
| 7284 | for use with a different version is not guaranteed to work and may |
| 7285 | cause crashes, and saving and restoring a compiled pattern loses any |
| 7286 | JIT optimization data. |
| 7287 | |
| 7288 | |
| 7289 | SAVING A COMPILED PATTERN |
| 7290 | |
| 7291 | The value returned by pcre_compile() points to a single block of memory |
| 7292 | that holds the compiled pattern and associated data. You can find the |
| 7293 | length of this block in bytes by calling pcre_fullinfo() with an argu- |
| 7294 | ment of PCRE_INFO_SIZE. You can then save the data in any appropriate |
| 7295 | manner. Here is sample code that compiles a pattern and writes it to a |
| 7296 | file. It assumes that the variable fd refers to a file that is open for |
| 7297 | output: |
| 7298 | |
| 7299 | int erroroffset, rc, size; |
| 7300 | char *error; |
| 7301 | pcre *re; |
| 7302 | |
| 7303 | re = pcre_compile("my pattern", 0, &error, &erroroffset, NULL); |
| 7304 | if (re == NULL) { ... handle errors ... } |
| 7305 | rc = pcre_fullinfo(re, NULL, PCRE_INFO_SIZE, &size); |
| 7306 | if (rc < 0) { ... handle errors ... } |
| 7307 | rc = fwrite(re, 1, size, fd); |
| 7308 | if (rc != size) { ... handle errors ... } |
| 7309 | |
| 7310 | In this example, the bytes that comprise the compiled pattern are |
| 7311 | copied exactly. Note that this is binary data that may contain any of |
| 7312 | the 256 possible byte values. On systems that make a distinction |
| 7313 | between binary and non-binary data, be sure that the file is opened for |
| 7314 | binary output. |
| 7315 | |
| 7316 | If you want to write more than one pattern to a file, you will have to |
| 7317 | devise a way of separating them. For binary data, preceding each pat- |
| 7318 | tern with its length is probably the most straightforward approach. |
| 7319 | Another possibility is to write out the data in hexadecimal instead of |
| 7320 | binary, one pattern to a line. |
| 7321 | |
| 7322 | Saving compiled patterns in a file is only one possible way of storing |
| 7323 | them for later use. They could equally well be saved in a database, or |
| 7324 | in the memory of some daemon process that passes them via sockets to |
| 7325 | the processes that want them. |
| 7326 | |
| 7327 | If the pattern has been studied, it is also possible to save the normal |
| 7328 | study data in a similar way to the compiled pattern itself. However, if |
| 7329 | the PCRE_STUDY_JIT_COMPILE was used, the just-in-time data that is cre- |
| 7330 | ated cannot be saved because it is too dependent on the current envi- |
| 7331 | ronment. When studying generates additional information, pcre_study() |
| 7332 | returns a pointer to a pcre_extra data block. Its format is defined in |
| 7333 | the section on matching a pattern in the pcreapi documentation. The |
| 7334 | study_data field points to the binary study data, and this is what you |
| 7335 | must save (not the pcre_extra block itself). The length of the study |
| 7336 | data can be obtained by calling pcre_fullinfo() with an argument of |
| 7337 | PCRE_INFO_STUDYSIZE. Remember to check that pcre_study() did return a |
| 7338 | non-NULL value before trying to save the study data. |
| 7339 | |
| 7340 | |
| 7341 | RE-USING A PRECOMPILED PATTERN |
| 7342 | |
| 7343 | Re-using a precompiled pattern is straightforward. Having reloaded it |
| 7344 | into main memory, you pass its pointer to pcre_exec() or |
| 7345 | pcre_dfa_exec() in the usual way. This should work even on another |
| 7346 | host, and even if that host has the opposite endianness to the one |
| 7347 | where the pattern was compiled. |
| 7348 | |
| 7349 | However, if you passed a pointer to custom character tables when the |
| 7350 | pattern was compiled (the tableptr argument of pcre_compile()), you |
| 7351 | must now pass a similar pointer to pcre_exec() or pcre_dfa_exec(), |
| 7352 | because the value saved with the compiled pattern will obviously be |
| 7353 | nonsense. A field in a pcre_extra() block is used to pass this data, as |
| 7354 | described in the section on matching a pattern in the pcreapi documen- |
| 7355 | tation. |
| 7356 | |
| 7357 | If you did not provide custom character tables when the pattern was |
| 7358 | compiled, the pointer in the compiled pattern is NULL, which causes |
| 7359 | pcre_exec() to use PCRE's internal tables. Thus, you do not need to |
| 7360 | take any special action at run time in this case. |
| 7361 | |
| 7362 | If you saved study data with the compiled pattern, you need to create |
| 7363 | your own pcre_extra data block and set the study_data field to point to |
| 7364 | the reloaded study data. You must also set the PCRE_EXTRA_STUDY_DATA |
| 7365 | bit in the flags field to indicate that study data is present. Then |
| 7366 | pass the pcre_extra block to pcre_exec() or pcre_dfa_exec() in the |
| 7367 | usual way. If the pattern was studied for just-in-time optimization, |
| 7368 | that data cannot be saved, and so is lost by a save/restore cycle. |
| 7369 | |
| 7370 | |
| 7371 | COMPATIBILITY WITH DIFFERENT PCRE RELEASES |
| 7372 | |
| 7373 | In general, it is safest to recompile all saved patterns when you |
| 7374 | update to a new PCRE release, though not all updates actually require |
| 7375 | this. |
| 7376 | |
| 7377 | |
| 7378 | AUTHOR |
| 7379 | |
| 7380 | Philip Hazel |
| 7381 | University Computing Service |
| 7382 | Cambridge CB2 3QH, England. |
| 7383 | |
| 7384 | |
| 7385 | REVISION |
| 7386 | |
| 7387 | Last updated: 26 August 2011 |
| 7388 | Copyright (c) 1997-2011 University of Cambridge. |
| 7389 | ------------------------------------------------------------------------------ |
| 7390 | |
| 7391 | |
| 7392 | PCREPERFORM(3) PCREPERFORM(3) |
| 7393 | |
| 7394 | |
| 7395 | NAME |
| 7396 | PCRE - Perl-compatible regular expressions |
| 7397 | |
| 7398 | |
| 7399 | PCRE PERFORMANCE |
| 7400 | |
| 7401 | Two aspects of performance are discussed below: memory usage and pro- |
| 7402 | cessing time. The way you express your pattern as a regular expression |
| 7403 | can affect both of them. |
| 7404 | |
| 7405 | |
| 7406 | COMPILED PATTERN MEMORY USAGE |
| 7407 | |
| 7408 | Patterns are compiled by PCRE into a reasonably efficient byte code, so |
| 7409 | that most simple patterns do not use much memory. However, there is one |
| 7410 | case where the memory usage of a compiled pattern can be unexpectedly |
| 7411 | large. If a parenthesized subpattern has a quantifier with a minimum |
| 7412 | greater than 1 and/or a limited maximum, the whole subpattern is |
| 7413 | repeated in the compiled code. For example, the pattern |
| 7414 | |
| 7415 | (abc|def){2,4} |
| 7416 | |
| 7417 | is compiled as if it were |
| 7418 | |
| 7419 | (abc|def)(abc|def)((abc|def)(abc|def)?)? |
| 7420 | |
| 7421 | (Technical aside: It is done this way so that backtrack points within |
| 7422 | each of the repetitions can be independently maintained.) |
| 7423 | |
| 7424 | For regular expressions whose quantifiers use only small numbers, this |
| 7425 | is not usually a problem. However, if the numbers are large, and par- |
| 7426 | ticularly if such repetitions are nested, the memory usage can become |
| 7427 | an embarrassment. For example, the very simple pattern |
| 7428 | |
| 7429 | ((ab){1,1000}c){1,3} |
| 7430 | |
| 7431 | uses 51K bytes when compiled. When PCRE is compiled with its default |
| 7432 | internal pointer size of two bytes, the size limit on a compiled pat- |
| 7433 | tern is 64K, and this is reached with the above pattern if the outer |
| 7434 | repetition is increased from 3 to 4. PCRE can be compiled to use larger |
| 7435 | internal pointers and thus handle larger compiled patterns, but it is |
| 7436 | better to try to rewrite your pattern to use less memory if you can. |
| 7437 | |
| 7438 | One way of reducing the memory usage for such patterns is to make use |
| 7439 | of PCRE's "subroutine" facility. Re-writing the above pattern as |
| 7440 | |
| 7441 | ((ab)(?2){0,999}c)(?1){0,2} |
| 7442 | |
| 7443 | reduces the memory requirements to 18K, and indeed it remains under 20K |
| 7444 | even with the outer repetition increased to 100. However, this pattern |
| 7445 | is not exactly equivalent, because the "subroutine" calls are treated |
| 7446 | as atomic groups into which there can be no backtracking if there is a |
| 7447 | subsequent matching failure. Therefore, PCRE cannot do this kind of |
| 7448 | rewriting automatically. Furthermore, there is a noticeable loss of |
| 7449 | speed when executing the modified pattern. Nevertheless, if the atomic |
| 7450 | grouping is not a problem and the loss of speed is acceptable, this |
| 7451 | kind of rewriting will allow you to process patterns that PCRE cannot |
| 7452 | otherwise handle. |
| 7453 | |
| 7454 | |
| 7455 | STACK USAGE AT RUN TIME |
| 7456 | |
| 7457 | When pcre_exec() is used for matching, certain kinds of pattern can |
| 7458 | cause it to use large amounts of the process stack. In some environ- |
| 7459 | ments the default process stack is quite small, and if it runs out the |
| 7460 | result is often SIGSEGV. This issue is probably the most frequently |
| 7461 | raised problem with PCRE. Rewriting your pattern can often help. The |
| 7462 | pcrestack documentation discusses this issue in detail. |
| 7463 | |
| 7464 | |
| 7465 | PROCESSING TIME |
| 7466 | |
| 7467 | Certain items in regular expression patterns are processed more effi- |
| 7468 | ciently than others. It is more efficient to use a character class like |
| 7469 | [aeiou] than a set of single-character alternatives such as |
| 7470 | (a|e|i|o|u). In general, the simplest construction that provides the |
| 7471 | required behaviour is usually the most efficient. Jeffrey Friedl's book |
| 7472 | contains a lot of useful general discussion about optimizing regular |
| 7473 | expressions for efficient performance. This document contains a few |
| 7474 | observations about PCRE. |
| 7475 | |
| 7476 | Using Unicode character properties (the \p, \P, and \X escapes) is |
| 7477 | slow, because PCRE has to scan a structure that contains data for over |
| 7478 | fifteen thousand characters whenever it needs a character's property. |
| 7479 | If you can find an alternative pattern that does not use character |
| 7480 | properties, it will probably be faster. |
| 7481 | |
| 7482 | By default, the escape sequences \b, \d, \s, and \w, and the POSIX |
| 7483 | character classes such as [:alpha:] do not use Unicode properties, |
| 7484 | partly for backwards compatibility, and partly for performance reasons. |
| 7485 | However, you can set PCRE_UCP if you want Unicode character properties |
| 7486 | to be used. This can double the matching time for items such as \d, |
| 7487 | when matched with pcre_exec(); the performance loss is less with |
| 7488 | pcre_dfa_exec(), and in both cases there is not much difference for \b. |
| 7489 | |
| 7490 | When a pattern begins with .* not in parentheses, or in parentheses |
| 7491 | that are not the subject of a backreference, and the PCRE_DOTALL option |
| 7492 | is set, the pattern is implicitly anchored by PCRE, since it can match |
| 7493 | only at the start of a subject string. However, if PCRE_DOTALL is not |
| 7494 | set, PCRE cannot make this optimization, because the . metacharacter |
| 7495 | does not then match a newline, and if the subject string contains new- |
| 7496 | lines, the pattern may match from the character immediately following |
| 7497 | one of them instead of from the very start. For example, the pattern |
| 7498 | |
| 7499 | .*second |
| 7500 | |
| 7501 | matches the subject "first\nand second" (where \n stands for a newline |
| 7502 | character), with the match starting at the seventh character. In order |
| 7503 | to do this, PCRE has to retry the match starting after every newline in |
| 7504 | the subject. |
| 7505 | |
| 7506 | If you are using such a pattern with subject strings that do not con- |
| 7507 | tain newlines, the best performance is obtained by setting PCRE_DOTALL, |
| 7508 | or starting the pattern with ^.* or ^.*? to indicate explicit anchor- |
| 7509 | ing. That saves PCRE from having to scan along the subject looking for |
| 7510 | a newline to restart at. |
| 7511 | |
| 7512 | Beware of patterns that contain nested indefinite repeats. These can |
| 7513 | take a long time to run when applied to a string that does not match. |
| 7514 | Consider the pattern fragment |
| 7515 | |
| 7516 | ^(a+)* |
| 7517 | |
| 7518 | This can match "aaaa" in 16 different ways, and this number increases |
| 7519 | very rapidly as the string gets longer. (The * repeat can match 0, 1, |
| 7520 | 2, 3, or 4 times, and for each of those cases other than 0 or 4, the + |
| 7521 | repeats can match different numbers of times.) When the remainder of |
| 7522 | the pattern is such that the entire match is going to fail, PCRE has in |
| 7523 | principle to try every possible variation, and this can take an |
| 7524 | extremely long time, even for relatively short strings. |
| 7525 | |
| 7526 | An optimization catches some of the more simple cases such as |
| 7527 | |
| 7528 | (a+)*b |
| 7529 | |
| 7530 | where a literal character follows. Before embarking on the standard |
| 7531 | matching procedure, PCRE checks that there is a "b" later in the sub- |
| 7532 | ject string, and if there is not, it fails the match immediately. How- |
| 7533 | ever, when there is no following literal this optimization cannot be |
| 7534 | used. You can see the difference by comparing the behaviour of |
| 7535 | |
| 7536 | (a+)*\d |
| 7537 | |
| 7538 | with the pattern above. The former gives a failure almost instantly |
| 7539 | when applied to a whole line of "a" characters, whereas the latter |
| 7540 | takes an appreciable time with strings longer than about 20 characters. |
| 7541 | |
| 7542 | In many cases, the solution to this kind of performance issue is to use |
| 7543 | an atomic group or a possessive quantifier. |
| 7544 | |
| 7545 | |
| 7546 | AUTHOR |
| 7547 | |
| 7548 | Philip Hazel |
| 7549 | University Computing Service |
| 7550 | Cambridge CB2 3QH, England. |
| 7551 | |
| 7552 | |
| 7553 | REVISION |
| 7554 | |
| 7555 | Last updated: 16 May 2010 |
| 7556 | Copyright (c) 1997-2010 University of Cambridge. |
| 7557 | ------------------------------------------------------------------------------ |
| 7558 | |
| 7559 | |
| 7560 | PCREPOSIX(3) PCREPOSIX(3) |
| 7561 | |
| 7562 | |
| 7563 | NAME |
| 7564 | PCRE - Perl-compatible regular expressions. |
| 7565 | |
| 7566 | |
| 7567 | SYNOPSIS OF POSIX API |
| 7568 | |
| 7569 | #include <pcreposix.h> |
| 7570 | |
| 7571 | int regcomp(regex_t *preg, const char *pattern, |
| 7572 | int cflags); |
| 7573 | |
| 7574 | int regexec(regex_t *preg, const char *string, |
| 7575 | size_t nmatch, regmatch_t pmatch[], int eflags); |
| 7576 | |
| 7577 | size_t regerror(int errcode, const regex_t *preg, |
| 7578 | char *errbuf, size_t errbuf_size); |
| 7579 | |
| 7580 | void regfree(regex_t *preg); |
| 7581 | |
| 7582 | |
| 7583 | DESCRIPTION |
| 7584 | |
| 7585 | This set of functions provides a POSIX-style API to the PCRE regular |
| 7586 | expression package. See the pcreapi documentation for a description of |
| 7587 | PCRE's native API, which contains much additional functionality. |
| 7588 | |
| 7589 | The functions described here are just wrapper functions that ultimately |
| 7590 | call the PCRE native API. Their prototypes are defined in the |
| 7591 | pcreposix.h header file, and on Unix systems the library itself is |
| 7592 | called pcreposix.a, so can be accessed by adding -lpcreposix to the |
| 7593 | command for linking an application that uses them. Because the POSIX |
| 7594 | functions call the native ones, it is also necessary to add -lpcre. |
| 7595 | |
| 7596 | I have implemented only those POSIX option bits that can be reasonably |
| 7597 | mapped to PCRE native options. In addition, the option REG_EXTENDED is |
| 7598 | defined with the value zero. This has no effect, but since programs |
| 7599 | that are written to the POSIX interface often use it, this makes it |
| 7600 | easier to slot in PCRE as a replacement library. Other POSIX options |
| 7601 | are not even defined. |
| 7602 | |
| 7603 | There are also some other options that are not defined by POSIX. These |
| 7604 | have been added at the request of users who want to make use of certain |
| 7605 | PCRE-specific features via the POSIX calling interface. |
| 7606 | |
| 7607 | When PCRE is called via these functions, it is only the API that is |
| 7608 | POSIX-like in style. The syntax and semantics of the regular expres- |
| 7609 | sions themselves are still those of Perl, subject to the setting of |
| 7610 | various PCRE options, as described below. "POSIX-like in style" means |
| 7611 | that the API approximates to the POSIX definition; it is not fully |
| 7612 | POSIX-compatible, and in multi-byte encoding domains it is probably |
| 7613 | even less compatible. |
| 7614 | |
| 7615 | The header for these functions is supplied as pcreposix.h to avoid any |
| 7616 | potential clash with other POSIX libraries. It can, of course, be |
| 7617 | renamed or aliased as regex.h, which is the "correct" name. It provides |
| 7618 | two structure types, regex_t for compiled internal forms, and reg- |
| 7619 | match_t for returning captured substrings. It also defines some con- |
| 7620 | stants whose names start with "REG_"; these are used for setting |
| 7621 | options and identifying error codes. |
| 7622 | |
| 7623 | |
| 7624 | COMPILING A PATTERN |
| 7625 | |
| 7626 | The function regcomp() is called to compile a pattern into an internal |
| 7627 | form. The pattern is a C string terminated by a binary zero, and is |
| 7628 | passed in the argument pattern. The preg argument is a pointer to a |
| 7629 | regex_t structure that is used as a base for storing information about |
| 7630 | the compiled regular expression. |
| 7631 | |
| 7632 | The argument cflags is either zero, or contains one or more of the bits |
| 7633 | defined by the following macros: |
| 7634 | |
| 7635 | REG_DOTALL |
| 7636 | |
| 7637 | The PCRE_DOTALL option is set when the regular expression is passed for |
| 7638 | compilation to the native function. Note that REG_DOTALL is not part of |
| 7639 | the POSIX standard. |
| 7640 | |
| 7641 | REG_ICASE |
| 7642 | |
| 7643 | The PCRE_CASELESS option is set when the regular expression is passed |
| 7644 | for compilation to the native function. |
| 7645 | |
| 7646 | REG_NEWLINE |
| 7647 | |
| 7648 | The PCRE_MULTILINE option is set when the regular expression is passed |
| 7649 | for compilation to the native function. Note that this does not mimic |
| 7650 | the defined POSIX behaviour for REG_NEWLINE (see the following sec- |
| 7651 | tion). |
| 7652 | |
| 7653 | REG_NOSUB |
| 7654 | |
| 7655 | The PCRE_NO_AUTO_CAPTURE option is set when the regular expression is |
| 7656 | passed for compilation to the native function. In addition, when a pat- |
| 7657 | tern that is compiled with this flag is passed to regexec() for match- |
| 7658 | ing, the nmatch and pmatch arguments are ignored, and no captured |
| 7659 | strings are returned. |
| 7660 | |
| 7661 | REG_UCP |
| 7662 | |
| 7663 | The PCRE_UCP option is set when the regular expression is passed for |
| 7664 | compilation to the native function. This causes PCRE to use Unicode |
| 7665 | properties when matchine \d, \w, etc., instead of just recognizing |
| 7666 | ASCII values. Note that REG_UTF8 is not part of the POSIX standard. |
| 7667 | |
| 7668 | REG_UNGREEDY |
| 7669 | |
| 7670 | The PCRE_UNGREEDY option is set when the regular expression is passed |
| 7671 | for compilation to the native function. Note that REG_UNGREEDY is not |
| 7672 | part of the POSIX standard. |
| 7673 | |
| 7674 | REG_UTF8 |
| 7675 | |
| 7676 | The PCRE_UTF8 option is set when the regular expression is passed for |
| 7677 | compilation to the native function. This causes the pattern itself and |
| 7678 | all data strings used for matching it to be treated as UTF-8 strings. |
| 7679 | Note that REG_UTF8 is not part of the POSIX standard. |
| 7680 | |
| 7681 | In the absence of these flags, no options are passed to the native |
| 7682 | function. This means the the regex is compiled with PCRE default |
| 7683 | semantics. In particular, the way it handles newline characters in the |
| 7684 | subject string is the Perl way, not the POSIX way. Note that setting |
| 7685 | PCRE_MULTILINE has only some of the effects specified for REG_NEWLINE. |
| 7686 | It does not affect the way newlines are matched by . (they are not) or |
| 7687 | by a negative class such as [^a] (they are). |
| 7688 | |
| 7689 | The yield of regcomp() is zero on success, and non-zero otherwise. The |
| 7690 | preg structure is filled in on success, and one member of the structure |
| 7691 | is public: re_nsub contains the number of capturing subpatterns in the |
| 7692 | regular expression. Various error codes are defined in the header file. |
| 7693 | |
| 7694 | NOTE: If the yield of regcomp() is non-zero, you must not attempt to |
| 7695 | use the contents of the preg structure. If, for example, you pass it to |
| 7696 | regexec(), the result is undefined and your program is likely to crash. |
| 7697 | |
| 7698 | |
| 7699 | MATCHING NEWLINE CHARACTERS |
| 7700 | |
| 7701 | This area is not simple, because POSIX and Perl take different views of |
| 7702 | things. It is not possible to get PCRE to obey POSIX semantics, but |
| 7703 | then PCRE was never intended to be a POSIX engine. The following table |
| 7704 | lists the different possibilities for matching newline characters in |
| 7705 | PCRE: |
| 7706 | |
| 7707 | Default Change with |
| 7708 | |
| 7709 | . matches newline no PCRE_DOTALL |
| 7710 | newline matches [^a] yes not changeable |
| 7711 | $ matches \n at end yes PCRE_DOLLARENDONLY |
| 7712 | $ matches \n in middle no PCRE_MULTILINE |
| 7713 | ^ matches \n in middle no PCRE_MULTILINE |
| 7714 | |
| 7715 | This is the equivalent table for POSIX: |
| 7716 | |
| 7717 | Default Change with |
| 7718 | |
| 7719 | . matches newline yes REG_NEWLINE |
| 7720 | newline matches [^a] yes REG_NEWLINE |
| 7721 | $ matches \n at end no REG_NEWLINE |
| 7722 | $ matches \n in middle no REG_NEWLINE |
| 7723 | ^ matches \n in middle no REG_NEWLINE |
| 7724 | |
| 7725 | PCRE's behaviour is the same as Perl's, except that there is no equiva- |
| 7726 | lent for PCRE_DOLLAR_ENDONLY in Perl. In both PCRE and Perl, there is |
| 7727 | no way to stop newline from matching [^a]. |
| 7728 | |
| 7729 | The default POSIX newline handling can be obtained by setting |
| 7730 | PCRE_DOTALL and PCRE_DOLLAR_ENDONLY, but there is no way to make PCRE |
| 7731 | behave exactly as for the REG_NEWLINE action. |
| 7732 | |
| 7733 | |
| 7734 | MATCHING A PATTERN |
| 7735 | |
| 7736 | The function regexec() is called to match a compiled pattern preg |
| 7737 | against a given string, which is by default terminated by a zero byte |
| 7738 | (but see REG_STARTEND below), subject to the options in eflags. These |
| 7739 | can be: |
| 7740 | |
| 7741 | REG_NOTBOL |
| 7742 | |
| 7743 | The PCRE_NOTBOL option is set when calling the underlying PCRE matching |
| 7744 | function. |
| 7745 | |
| 7746 | REG_NOTEMPTY |
| 7747 | |
| 7748 | The PCRE_NOTEMPTY option is set when calling the underlying PCRE match- |
| 7749 | ing function. Note that REG_NOTEMPTY is not part of the POSIX standard. |
| 7750 | However, setting this option can give more POSIX-like behaviour in some |
| 7751 | situations. |
| 7752 | |
| 7753 | REG_NOTEOL |
| 7754 | |
| 7755 | The PCRE_NOTEOL option is set when calling the underlying PCRE matching |
| 7756 | function. |
| 7757 | |
| 7758 | REG_STARTEND |
| 7759 | |
| 7760 | The string is considered to start at string + pmatch[0].rm_so and to |
| 7761 | have a terminating NUL located at string + pmatch[0].rm_eo (there need |
| 7762 | not actually be a NUL at that location), regardless of the value of |
| 7763 | nmatch. This is a BSD extension, compatible with but not specified by |
| 7764 | IEEE Standard 1003.2 (POSIX.2), and should be used with caution in |
| 7765 | software intended to be portable to other systems. Note that a non-zero |
| 7766 | rm_so does not imply REG_NOTBOL; REG_STARTEND affects only the location |
| 7767 | of the string, not how it is matched. |
| 7768 | |
| 7769 | If the pattern was compiled with the REG_NOSUB flag, no data about any |
| 7770 | matched strings is returned. The nmatch and pmatch arguments of |
| 7771 | regexec() are ignored. |
| 7772 | |
| 7773 | If the value of nmatch is zero, or if the value pmatch is NULL, no data |
| 7774 | about any matched strings is returned. |
| 7775 | |
| 7776 | Otherwise,the portion of the string that was matched, and also any cap- |
| 7777 | tured substrings, are returned via the pmatch argument, which points to |
| 7778 | an array of nmatch structures of type regmatch_t, containing the mem- |
| 7779 | bers rm_so and rm_eo. These contain the offset to the first character |
| 7780 | of each substring and the offset to the first character after the end |
| 7781 | of each substring, respectively. The 0th element of the vector relates |
| 7782 | to the entire portion of string that was matched; subsequent elements |
| 7783 | relate to the capturing subpatterns of the regular expression. Unused |
| 7784 | entries in the array have both structure members set to -1. |
| 7785 | |
| 7786 | A successful match yields a zero return; various error codes are |
| 7787 | defined in the header file, of which REG_NOMATCH is the "expected" |
| 7788 | failure code. |
| 7789 | |
| 7790 | |
| 7791 | ERROR MESSAGES |
| 7792 | |
| 7793 | The regerror() function maps a non-zero errorcode from either regcomp() |
| 7794 | or regexec() to a printable message. If preg is not NULL, the error |
| 7795 | should have arisen from the use of that structure. A message terminated |
| 7796 | by a binary zero is placed in errbuf. The length of the message, |
| 7797 | including the zero, is limited to errbuf_size. The yield of the func- |
| 7798 | tion is the size of buffer needed to hold the whole message. |
| 7799 | |
| 7800 | |
| 7801 | MEMORY USAGE |
| 7802 | |
| 7803 | Compiling a regular expression causes memory to be allocated and asso- |
| 7804 | ciated with the preg structure. The function regfree() frees all such |
| 7805 | memory, after which preg may no longer be used as a compiled expres- |
| 7806 | sion. |
| 7807 | |
| 7808 | |
| 7809 | AUTHOR |
| 7810 | |
| 7811 | Philip Hazel |
| 7812 | University Computing Service |
| 7813 | Cambridge CB2 3QH, England. |
| 7814 | |
| 7815 | |
| 7816 | REVISION |
| 7817 | |
| 7818 | Last updated: 16 May 2010 |
| 7819 | Copyright (c) 1997-2010 University of Cambridge. |
| 7820 | ------------------------------------------------------------------------------ |
| 7821 | |
| 7822 | |
| 7823 | PCRECPP(3) PCRECPP(3) |
| 7824 | |
| 7825 | |
| 7826 | NAME |
| 7827 | PCRE - Perl-compatible regular expressions. |
| 7828 | |
| 7829 | |
| 7830 | SYNOPSIS OF C++ WRAPPER |
| 7831 | |
| 7832 | #include <pcrecpp.h> |
| 7833 | |
| 7834 | |
| 7835 | DESCRIPTION |
| 7836 | |
| 7837 | The C++ wrapper for PCRE was provided by Google Inc. Some additional |
| 7838 | functionality was added by Giuseppe Maxia. This brief man page was con- |
| 7839 | structed from the notes in the pcrecpp.h file, which should be con- |
| 7840 | sulted for further details. |
| 7841 | |
| 7842 | |
| 7843 | MATCHING INTERFACE |
| 7844 | |
| 7845 | The "FullMatch" operation checks that supplied text matches a supplied |
| 7846 | pattern exactly. If pointer arguments are supplied, it copies matched |
| 7847 | sub-strings that match sub-patterns into them. |
| 7848 | |
| 7849 | Example: successful match |
| 7850 | pcrecpp::RE re("h.*o"); |
| 7851 | re.FullMatch("hello"); |
| 7852 | |
| 7853 | Example: unsuccessful match (requires full match): |
| 7854 | pcrecpp::RE re("e"); |
| 7855 | !re.FullMatch("hello"); |
| 7856 | |
| 7857 | Example: creating a temporary RE object: |
| 7858 | pcrecpp::RE("h.*o").FullMatch("hello"); |
| 7859 | |
| 7860 | You can pass in a "const char*" or a "string" for "text". The examples |
| 7861 | below tend to use a const char*. You can, as in the different examples |
| 7862 | above, store the RE object explicitly in a variable or use a temporary |
| 7863 | RE object. The examples below use one mode or the other arbitrarily. |
| 7864 | Either could correctly be used for any of these examples. |
| 7865 | |
| 7866 | You must supply extra pointer arguments to extract matched subpieces. |
| 7867 | |
| 7868 | Example: extracts "ruby" into "s" and 1234 into "i" |
| 7869 | int i; |
| 7870 | string s; |
| 7871 | pcrecpp::RE re("(\\w+):(\\d+)"); |
| 7872 | re.FullMatch("ruby:1234", &s, &i); |
| 7873 | |
| 7874 | Example: does not try to extract any extra sub-patterns |
| 7875 | re.FullMatch("ruby:1234", &s); |
| 7876 | |
| 7877 | Example: does not try to extract into NULL |
| 7878 | re.FullMatch("ruby:1234", NULL, &i); |
| 7879 | |
| 7880 | Example: integer overflow causes failure |
| 7881 | !re.FullMatch("ruby:1234567891234", NULL, &i); |
| 7882 | |
| 7883 | Example: fails because there aren't enough sub-patterns: |
| 7884 | !pcrecpp::RE("\\w+:\\d+").FullMatch("ruby:1234", &s); |
| 7885 | |
| 7886 | Example: fails because string cannot be stored in integer |
| 7887 | !pcrecpp::RE("(.*)").FullMatch("ruby", &i); |
| 7888 | |
| 7889 | The provided pointer arguments can be pointers to any scalar numeric |
| 7890 | type, or one of: |
| 7891 | |
| 7892 | string (matched piece is copied to string) |
| 7893 | StringPiece (StringPiece is mutated to point to matched piece) |
| 7894 | T (where "bool T::ParseFrom(const char*, int)" exists) |
| 7895 | NULL (the corresponding matched sub-pattern is not copied) |
| 7896 | |
| 7897 | The function returns true iff all of the following conditions are sat- |
| 7898 | isfied: |
| 7899 | |
| 7900 | a. "text" matches "pattern" exactly; |
| 7901 | |
| 7902 | b. The number of matched sub-patterns is >= number of supplied |
| 7903 | pointers; |
| 7904 | |
| 7905 | c. The "i"th argument has a suitable type for holding the |
| 7906 | string captured as the "i"th sub-pattern. If you pass in |
| 7907 | void * NULL for the "i"th argument, or a non-void * NULL |
| 7908 | of the correct type, or pass fewer arguments than the |
| 7909 | number of sub-patterns, "i"th captured sub-pattern is |
| 7910 | ignored. |
| 7911 | |
| 7912 | CAVEAT: An optional sub-pattern that does not exist in the matched |
| 7913 | string is assigned the empty string. Therefore, the following will |
| 7914 | return false (because the empty string is not a valid number): |
| 7915 | |
| 7916 | int number; |
| 7917 | pcrecpp::RE::FullMatch("abc", "[a-z]+(\\d+)?", &number); |
| 7918 | |
| 7919 | The matching interface supports at most 16 arguments per call. If you |
| 7920 | need more, consider using the more general interface |
| 7921 | pcrecpp::RE::DoMatch. See pcrecpp.h for the signature for DoMatch. |
| 7922 | |
| 7923 | NOTE: Do not use no_arg, which is used internally to mark the end of a |
| 7924 | list of optional arguments, as a placeholder for missing arguments, as |
| 7925 | this can lead to segfaults. |
| 7926 | |
| 7927 | |
| 7928 | QUOTING METACHARACTERS |
| 7929 | |
| 7930 | You can use the "QuoteMeta" operation to insert backslashes before all |
| 7931 | potentially meaningful characters in a string. The returned string, |
| 7932 | used as a regular expression, will exactly match the original string. |
| 7933 | |
| 7934 | Example: |
| 7935 | string quoted = RE::QuoteMeta(unquoted); |
| 7936 | |
| 7937 | Note that it's legal to escape a character even if it has no special |
| 7938 | meaning in a regular expression -- so this function does that. (This |
| 7939 | also makes it identical to the perl function of the same name; see |
| 7940 | "perldoc -f quotemeta".) For example, "1.5-2.0?" becomes |
| 7941 | "1\.5\-2\.0\?". |
| 7942 | |
| 7943 | |
| 7944 | PARTIAL MATCHES |
| 7945 | |
| 7946 | You can use the "PartialMatch" operation when you want the pattern to |
| 7947 | match any substring of the text. |
| 7948 | |
| 7949 | Example: simple search for a string: |
| 7950 | pcrecpp::RE("ell").PartialMatch("hello"); |
| 7951 | |
| 7952 | Example: find first number in a string: |
| 7953 | int number; |
| 7954 | pcrecpp::RE re("(\\d+)"); |
| 7955 | re.PartialMatch("x*100 + 20", &number); |
| 7956 | assert(number == 100); |
| 7957 | |
| 7958 | |
| 7959 | UTF-8 AND THE MATCHING INTERFACE |
| 7960 | |
| 7961 | By default, pattern and text are plain text, one byte per character. |
| 7962 | The UTF8 flag, passed to the constructor, causes both pattern and |
| 7963 | string to be treated as UTF-8 text, still a byte stream but potentially |
| 7964 | multiple bytes per character. In practice, the text is likelier to be |
| 7965 | UTF-8 than the pattern, but the match returned may depend on the UTF8 |
| 7966 | flag, so always use it when matching UTF8 text. For example, "." will |
| 7967 | match one byte normally but with UTF8 set may match up to three bytes |
| 7968 | of a multi-byte character. |
| 7969 | |
| 7970 | Example: |
| 7971 | pcrecpp::RE_Options options; |
| 7972 | options.set_utf8(); |
| 7973 | pcrecpp::RE re(utf8_pattern, options); |
| 7974 | re.FullMatch(utf8_string); |
| 7975 | |
| 7976 | Example: using the convenience function UTF8(): |
| 7977 | pcrecpp::RE re(utf8_pattern, pcrecpp::UTF8()); |
| 7978 | re.FullMatch(utf8_string); |
| 7979 | |
| 7980 | NOTE: The UTF8 flag is ignored if pcre was not configured with the |
| 7981 | --enable-utf8 flag. |
| 7982 | |
| 7983 | |
| 7984 | PASSING MODIFIERS TO THE REGULAR EXPRESSION ENGINE |
| 7985 | |
| 7986 | PCRE defines some modifiers to change the behavior of the regular |
| 7987 | expression engine. The C++ wrapper defines an auxiliary class, |
| 7988 | RE_Options, as a vehicle to pass such modifiers to a RE class. Cur- |
| 7989 | rently, the following modifiers are supported: |
| 7990 | |
| 7991 | modifier description Perl corresponding |
| 7992 | |
| 7993 | PCRE_CASELESS case insensitive match /i |
| 7994 | PCRE_MULTILINE multiple lines match /m |
| 7995 | PCRE_DOTALL dot matches newlines /s |
| 7996 | PCRE_DOLLAR_ENDONLY $ matches only at end N/A |
| 7997 | PCRE_EXTRA strict escape parsing N/A |
| 7998 | PCRE_EXTENDED ignore whitespaces /x |
| 7999 | PCRE_UTF8 handles UTF8 chars built-in |
| 8000 | PCRE_UNGREEDY reverses * and *? N/A |
| 8001 | PCRE_NO_AUTO_CAPTURE disables capturing parens N/A (*) |
| 8002 | |
| 8003 | (*) Both Perl and PCRE allow non capturing parentheses by means of the |
| 8004 | "?:" modifier within the pattern itself. e.g. (?:ab|cd) does not cap- |
| 8005 | ture, while (ab|cd) does. |
| 8006 | |
| 8007 | For a full account on how each modifier works, please check the PCRE |
| 8008 | API reference page. |
| 8009 | |
| 8010 | For each modifier, there are two member functions whose name is made |
| 8011 | out of the modifier in lowercase, without the "PCRE_" prefix. For |
| 8012 | instance, PCRE_CASELESS is handled by |
| 8013 | |
| 8014 | bool caseless() |
| 8015 | |
| 8016 | which returns true if the modifier is set, and |
| 8017 | |
| 8018 | RE_Options & set_caseless(bool) |
| 8019 | |
| 8020 | which sets or unsets the modifier. Moreover, PCRE_EXTRA_MATCH_LIMIT can |
| 8021 | be accessed through the set_match_limit() and match_limit() member |
| 8022 | functions. Setting match_limit to a non-zero value will limit the exe- |
| 8023 | cution of pcre to keep it from doing bad things like blowing the stack |
| 8024 | or taking an eternity to return a result. A value of 5000 is good |
| 8025 | enough to stop stack blowup in a 2MB thread stack. Setting match_limit |
| 8026 | to zero disables match limiting. Alternatively, you can call |
| 8027 | match_limit_recursion() which uses PCRE_EXTRA_MATCH_LIMIT_RECURSION to |
| 8028 | limit how much PCRE recurses. match_limit() limits the number of |
| 8029 | matches PCRE does; match_limit_recursion() limits the depth of internal |
| 8030 | recursion, and therefore the amount of stack that is used. |
| 8031 | |
| 8032 | Normally, to pass one or more modifiers to a RE class, you declare a |
| 8033 | RE_Options object, set the appropriate options, and pass this object to |
| 8034 | a RE constructor. Example: |
| 8035 | |
| 8036 | RE_Options opt; |
| 8037 | opt.set_caseless(true); |
| 8038 | if (RE("HELLO", opt).PartialMatch("hello world")) ... |
| 8039 | |
| 8040 | RE_options has two constructors. The default constructor takes no argu- |
| 8041 | ments and creates a set of flags that are off by default. The optional |
| 8042 | parameter option_flags is to facilitate transfer of legacy code from C |
| 8043 | programs. This lets you do |
| 8044 | |
| 8045 | RE(pattern, |
| 8046 | RE_Options(PCRE_CASELESS|PCRE_MULTILINE)).PartialMatch(str); |
| 8047 | |
| 8048 | However, new code is better off doing |
| 8049 | |
| 8050 | RE(pattern, |
| 8051 | RE_Options().set_caseless(true).set_multiline(true)) |
| 8052 | .PartialMatch(str); |
| 8053 | |
| 8054 | If you are going to pass one of the most used modifiers, there are some |
| 8055 | convenience functions that return a RE_Options class with the appropri- |
| 8056 | ate modifier already set: CASELESS(), UTF8(), MULTILINE(), DOTALL(), |
| 8057 | and EXTENDED(). |
| 8058 | |
| 8059 | If you need to set several options at once, and you don't want to go |
| 8060 | through the pains of declaring a RE_Options object and setting several |
| 8061 | options, there is a parallel method that give you such ability on the |
| 8062 | fly. You can concatenate several set_xxxxx() member functions, since |
| 8063 | each of them returns a reference to its class object. For example, to |
| 8064 | pass PCRE_CASELESS, PCRE_EXTENDED, and PCRE_MULTILINE to a RE with one |
| 8065 | statement, you may write: |
| 8066 | |
| 8067 | RE(" ^ xyz \\s+ .* blah$", |
| 8068 | RE_Options() |
| 8069 | .set_caseless(true) |
| 8070 | .set_extended(true) |
| 8071 | .set_multiline(true)).PartialMatch(sometext); |
| 8072 | |
| 8073 | |
| 8074 | SCANNING TEXT INCREMENTALLY |
| 8075 | |
| 8076 | The "Consume" operation may be useful if you want to repeatedly match |
| 8077 | regular expressions at the front of a string and skip over them as they |
| 8078 | match. This requires use of the "StringPiece" type, which represents a |
| 8079 | sub-range of a real string. Like RE, StringPiece is defined in the |
| 8080 | pcrecpp namespace. |
| 8081 | |
| 8082 | Example: read lines of the form "var = value" from a string. |
| 8083 | string contents = ...; // Fill string somehow |
| 8084 | pcrecpp::StringPiece input(contents); // Wrap in a StringPiece |
| 8085 | |
| 8086 | string var; |
| 8087 | int value; |
| 8088 | pcrecpp::RE re("(\\w+) = (\\d+)\n"); |
| 8089 | while (re.Consume(&input, &var, &value)) { |
| 8090 | ...; |
| 8091 | } |
| 8092 | |
| 8093 | Each successful call to "Consume" will set "var/value", and also |
| 8094 | advance "input" so it points past the matched text. |
| 8095 | |
| 8096 | The "FindAndConsume" operation is similar to "Consume" but does not |
| 8097 | anchor your match at the beginning of the string. For example, you |
| 8098 | could extract all words from a string by repeatedly calling |
| 8099 | |
| 8100 | pcrecpp::RE("(\\w+)").FindAndConsume(&input, &word) |
| 8101 | |
| 8102 | |
| 8103 | PARSING HEX/OCTAL/C-RADIX NUMBERS |
| 8104 | |
| 8105 | By default, if you pass a pointer to a numeric value, the corresponding |
| 8106 | text is interpreted as a base-10 number. You can instead wrap the |
| 8107 | pointer with a call to one of the operators Hex(), Octal(), or CRadix() |
| 8108 | to interpret the text in another base. The CRadix operator interprets |
| 8109 | C-style "0" (base-8) and "0x" (base-16) prefixes, but defaults to |
| 8110 | base-10. |
| 8111 | |
| 8112 | Example: |
| 8113 | int a, b, c, d; |
| 8114 | pcrecpp::RE re("(.*) (.*) (.*) (.*)"); |
| 8115 | re.FullMatch("100 40 0100 0x40", |
| 8116 | pcrecpp::Octal(&a), pcrecpp::Hex(&b), |
| 8117 | pcrecpp::CRadix(&c), pcrecpp::CRadix(&d)); |
| 8118 | |
| 8119 | will leave 64 in a, b, c, and d. |
| 8120 | |
| 8121 | |
| 8122 | REPLACING PARTS OF STRINGS |
| 8123 | |
| 8124 | You can replace the first match of "pattern" in "str" with "rewrite". |
| 8125 | Within "rewrite", backslash-escaped digits (\1 to \9) can be used to |
| 8126 | insert text matching corresponding parenthesized group from the pat- |
| 8127 | tern. \0 in "rewrite" refers to the entire matching text. For example: |
| 8128 | |
| 8129 | string s = "yabba dabba doo"; |
| 8130 | pcrecpp::RE("b+").Replace("d", &s); |
| 8131 | |
| 8132 | will leave "s" containing "yada dabba doo". The result is true if the |
| 8133 | pattern matches and a replacement occurs, false otherwise. |
| 8134 | |
| 8135 | GlobalReplace is like Replace except that it replaces all occurrences |
| 8136 | of the pattern in the string with the rewrite. Replacements are not |
| 8137 | subject to re-matching. For example: |
| 8138 | |
| 8139 | string s = "yabba dabba doo"; |
| 8140 | pcrecpp::RE("b+").GlobalReplace("d", &s); |
| 8141 | |
| 8142 | will leave "s" containing "yada dada doo". It returns the number of |
| 8143 | replacements made. |
| 8144 | |
| 8145 | Extract is like Replace, except that if the pattern matches, "rewrite" |
| 8146 | is copied into "out" (an additional argument) with substitutions. The |
| 8147 | non-matching portions of "text" are ignored. Returns true iff a match |
| 8148 | occurred and the extraction happened successfully; if no match occurs, |
| 8149 | the string is left unaffected. |
| 8150 | |
| 8151 | |
| 8152 | AUTHOR |
| 8153 | |
| 8154 | The C++ wrapper was contributed by Google Inc. |
| 8155 | Copyright (c) 2007 Google Inc. |
| 8156 | |
| 8157 | |
| 8158 | REVISION |
| 8159 | |
| 8160 | Last updated: 17 March 2009 |
| 8161 | Minor typo fixed: 25 July 2011 |
| 8162 | ------------------------------------------------------------------------------ |
| 8163 | |
| 8164 | |
| 8165 | PCRESAMPLE(3) PCRESAMPLE(3) |
| 8166 | |
| 8167 | |
| 8168 | NAME |
| 8169 | PCRE - Perl-compatible regular expressions |
| 8170 | |
| 8171 | |
| 8172 | PCRE SAMPLE PROGRAM |
| 8173 | |
| 8174 | A simple, complete demonstration program, to get you started with using |
| 8175 | PCRE, is supplied in the file pcredemo.c in the PCRE distribution. A |
| 8176 | listing of this program is given in the pcredemo documentation. If you |
| 8177 | do not have a copy of the PCRE distribution, you can save this listing |
| 8178 | to re-create pcredemo.c. |
| 8179 | |
| 8180 | The program compiles the regular expression that is its first argument, |
| 8181 | and matches it against the subject string in its second argument. No |
| 8182 | PCRE options are set, and default character tables are used. If match- |
| 8183 | ing succeeds, the program outputs the portion of the subject that |
| 8184 | matched, together with the contents of any captured substrings. |
| 8185 | |
| 8186 | If the -g option is given on the command line, the program then goes on |
| 8187 | to check for further matches of the same regular expression in the same |
| 8188 | subject string. The logic is a little bit tricky because of the possi- |
| 8189 | bility of matching an empty string. Comments in the code explain what |
| 8190 | is going on. |
| 8191 | |
| 8192 | If PCRE is installed in the standard include and library directories |
| 8193 | for your operating system, you should be able to compile the demonstra- |
| 8194 | tion program using this command: |
| 8195 | |
| 8196 | gcc -o pcredemo pcredemo.c -lpcre |
| 8197 | |
| 8198 | If PCRE is installed elsewhere, you may need to add additional options |
| 8199 | to the command line. For example, on a Unix-like system that has PCRE |
| 8200 | installed in /usr/local, you can compile the demonstration program |
| 8201 | using a command like this: |
| 8202 | |
| 8203 | gcc -o pcredemo -I/usr/local/include pcredemo.c \ |
| 8204 | -L/usr/local/lib -lpcre |
| 8205 | |
| 8206 | In a Windows environment, if you want to statically link the program |
| 8207 | against a non-dll pcre.a file, you must uncomment the line that defines |
| 8208 | PCRE_STATIC before including pcre.h, because otherwise the pcre_mal- |
| 8209 | loc() and pcre_free() exported functions will be declared |
| 8210 | __declspec(dllimport), with unwanted results. |
| 8211 | |
| 8212 | Once you have compiled and linked the demonstration program, you can |
| 8213 | run simple tests like this: |
| 8214 | |
| 8215 | ./pcredemo 'cat|dog' 'the cat sat on the mat' |
| 8216 | ./pcredemo -g 'cat|dog' 'the dog sat on the cat' |
| 8217 | |
| 8218 | Note that there is a much more comprehensive test program, called |
| 8219 | pcretest, which supports many more facilities for testing regular |
| 8220 | expressions and the PCRE library. The pcredemo program is provided as a |
| 8221 | simple coding example. |
| 8222 | |
| 8223 | If you try to run pcredemo when PCRE is not installed in the standard |
| 8224 | library directory, you may get an error like this on some operating |
| 8225 | systems (e.g. Solaris): |
| 8226 | |
| 8227 | ld.so.1: a.out: fatal: libpcre.so.0: open failed: No such file or |
| 8228 | directory |
| 8229 | |
| 8230 | This is caused by the way shared library support works on those sys- |
| 8231 | tems. You need to add |
| 8232 | |
| 8233 | -R/usr/local/lib |
| 8234 | |
| 8235 | (for example) to the compile command to get round this problem. |
| 8236 | |
| 8237 | |
| 8238 | AUTHOR |
| 8239 | |
| 8240 | Philip Hazel |
| 8241 | University Computing Service |
| 8242 | Cambridge CB2 3QH, England. |
| 8243 | |
| 8244 | |
| 8245 | REVISION |
| 8246 | |
| 8247 | Last updated: 17 November 2010 |
| 8248 | Copyright (c) 1997-2010 University of Cambridge. |
| 8249 | ------------------------------------------------------------------------------ |
| 8250 | PCRELIMITS(3) PCRELIMITS(3) |
| 8251 | |
| 8252 | |
| 8253 | NAME |
| 8254 | PCRE - Perl-compatible regular expressions |
| 8255 | |
| 8256 | |
| 8257 | SIZE AND OTHER LIMITATIONS |
| 8258 | |
| 8259 | There are some size limitations in PCRE but it is hoped that they will |
| 8260 | never in practice be relevant. |
| 8261 | |
| 8262 | The maximum length of a compiled pattern is 65539 (sic) bytes if PCRE |
| 8263 | is compiled with the default internal linkage size of 2. If you want to |
| 8264 | process regular expressions that are truly enormous, you can compile |
| 8265 | PCRE with an internal linkage size of 3 or 4 (see the README file in |
| 8266 | the source distribution and the pcrebuild documentation for details). |
| 8267 | In these cases the limit is substantially larger. However, the speed |
| 8268 | of execution is slower. |
| 8269 | |
| 8270 | All values in repeating quantifiers must be less than 65536. |
| 8271 | |
| 8272 | There is no limit to the number of parenthesized subpatterns, but there |
| 8273 | can be no more than 65535 capturing subpatterns. |
| 8274 | |
| 8275 | There is a limit to the number of forward references to subsequent sub- |
| 8276 | patterns of around 200,000. Repeated forward references with fixed |
| 8277 | upper limits, for example, (?2){0,100} when subpattern number 2 is to |
| 8278 | the right, are included in the count. There is no limit to the number |
| 8279 | of backward references. |
| 8280 | |
| 8281 | The maximum length of name for a named subpattern is 32 characters, and |
| 8282 | the maximum number of named subpatterns is 10000. |
| 8283 | |
| 8284 | The maximum length of a subject string is the largest positive number |
| 8285 | that an integer variable can hold. However, when using the traditional |
| 8286 | matching function, PCRE uses recursion to handle subpatterns and indef- |
| 8287 | inite repetition. This means that the available stack space may limit |
| 8288 | the size of a subject string that can be processed by certain patterns. |
| 8289 | For a discussion of stack issues, see the pcrestack documentation. |
| 8290 | |
| 8291 | |
| 8292 | AUTHOR |
| 8293 | |
| 8294 | Philip Hazel |
| 8295 | University Computing Service |
| 8296 | Cambridge CB2 3QH, England. |
| 8297 | |
| 8298 | |
| 8299 | REVISION |
| 8300 | |
| 8301 | Last updated: 30 November 2011 |
| 8302 | Copyright (c) 1997-2011 University of Cambridge. |
| 8303 | ------------------------------------------------------------------------------ |
| 8304 | |
| 8305 | |
| 8306 | PCRESTACK(3) PCRESTACK(3) |
| 8307 | |
| 8308 | |
| 8309 | NAME |
| 8310 | PCRE - Perl-compatible regular expressions |
| 8311 | |
| 8312 | |
| 8313 | PCRE DISCUSSION OF STACK USAGE |
| 8314 | |
| 8315 | When you call pcre_exec(), it makes use of an internal function called |
| 8316 | match(). This calls itself recursively at branch points in the pattern, |
| 8317 | in order to remember the state of the match so that it can back up and |
| 8318 | try a different alternative if the first one fails. As matching pro- |
| 8319 | ceeds deeper and deeper into the tree of possibilities, the recursion |
| 8320 | depth increases. The match() function is also called in other circum- |
| 8321 | stances, for example, whenever a parenthesized sub-pattern is entered, |
| 8322 | and in certain cases of repetition. |
| 8323 | |
| 8324 | Not all calls of match() increase the recursion depth; for an item such |
| 8325 | as a* it may be called several times at the same level, after matching |
| 8326 | different numbers of a's. Furthermore, in a number of cases where the |
| 8327 | result of the recursive call would immediately be passed back as the |
| 8328 | result of the current call (a "tail recursion"), the function is just |
| 8329 | restarted instead. |
| 8330 | |
| 8331 | The above comments apply when pcre_exec() is run in its normal inter- |
| 8332 | pretive manner. If the pattern was studied with the PCRE_STUDY_JIT_COM- |
| 8333 | PILE option, and just-in-time compiling was successful, and the options |
| 8334 | passed to pcre_exec() were not incompatible, the matching process uses |
| 8335 | the JIT-compiled code instead of the match() function. In this case, |
| 8336 | the memory requirements are handled entirely differently. See the pcre- |
| 8337 | jit documentation for details. |
| 8338 | |
| 8339 | The pcre_dfa_exec() function operates in an entirely different way, and |
| 8340 | uses recursion only when there is a regular expression recursion or |
| 8341 | subroutine call in the pattern. This includes the processing of asser- |
| 8342 | tion and "once-only" subpatterns, which are handled like subroutine |
| 8343 | calls. Normally, these are never very deep, and the limit on the com- |
| 8344 | plexity of pcre_dfa_exec() is controlled by the amount of workspace it |
| 8345 | is given. However, it is possible to write patterns with runaway infi- |
| 8346 | nite recursions; such patterns will cause pcre_dfa_exec() to run out of |
| 8347 | stack. At present, there is no protection against this. |
| 8348 | |
| 8349 | The comments that follow do NOT apply to pcre_dfa_exec(); they are rel- |
| 8350 | evant only for pcre_exec() without the JIT optimization. |
| 8351 | |
| 8352 | Reducing pcre_exec()'s stack usage |
| 8353 | |
| 8354 | Each time that match() is actually called recursively, it uses memory |
| 8355 | from the process stack. For certain kinds of pattern and data, very |
| 8356 | large amounts of stack may be needed, despite the recognition of "tail |
| 8357 | recursion". You can often reduce the amount of recursion, and there- |
| 8358 | fore the amount of stack used, by modifying the pattern that is being |
| 8359 | matched. Consider, for example, this pattern: |
| 8360 | |
| 8361 | ([^<]|<(?!inet))+ |
| 8362 | |
| 8363 | It matches from wherever it starts until it encounters "<inet" or the |
| 8364 | end of the data, and is the kind of pattern that might be used when |
| 8365 | processing an XML file. Each iteration of the outer parentheses matches |
| 8366 | either one character that is not "<" or a "<" that is not followed by |
| 8367 | "inet". However, each time a parenthesis is processed, a recursion |
| 8368 | occurs, so this formulation uses a stack frame for each matched charac- |
| 8369 | ter. For a long string, a lot of stack is required. Consider now this |
| 8370 | rewritten pattern, which matches exactly the same strings: |
| 8371 | |
| 8372 | ([^<]++|<(?!inet))+ |
| 8373 | |
| 8374 | This uses very much less stack, because runs of characters that do not |
| 8375 | contain "<" are "swallowed" in one item inside the parentheses. Recur- |
| 8376 | sion happens only when a "<" character that is not followed by "inet" |
| 8377 | is encountered (and we assume this is relatively rare). A possessive |
| 8378 | quantifier is used to stop any backtracking into the runs of non-"<" |
| 8379 | characters, but that is not related to stack usage. |
| 8380 | |
| 8381 | This example shows that one way of avoiding stack problems when match- |
| 8382 | ing long subject strings is to write repeated parenthesized subpatterns |
| 8383 | to match more than one character whenever possible. |
| 8384 | |
| 8385 | Compiling PCRE to use heap instead of stack for pcre_exec() |
| 8386 | |
| 8387 | In environments where stack memory is constrained, you might want to |
| 8388 | compile PCRE to use heap memory instead of stack for remembering back- |
| 8389 | up points when pcre_exec() is running. This makes it run a lot more |
| 8390 | slowly, however. Details of how to do this are given in the pcrebuild |
| 8391 | documentation. When built in this way, instead of using the stack, PCRE |
| 8392 | obtains and frees memory by calling the functions that are pointed to |
| 8393 | by the pcre_stack_malloc and pcre_stack_free variables. By default, |
| 8394 | these point to malloc() and free(), but you can replace the pointers to |
| 8395 | cause PCRE to use your own functions. Since the block sizes are always |
| 8396 | the same, and are always freed in reverse order, it may be possible to |
| 8397 | implement customized memory handlers that are more efficient than the |
| 8398 | standard functions. |
| 8399 | |
| 8400 | Limiting pcre_exec()'s stack usage |
| 8401 | |
| 8402 | You can set limits on the number of times that match() is called, both |
| 8403 | in total and recursively. If a limit is exceeded, pcre_exec() returns |
| 8404 | an error code. Setting suitable limits should prevent it from running |
| 8405 | out of stack. The default values of the limits are very large, and |
| 8406 | unlikely ever to operate. They can be changed when PCRE is built, and |
| 8407 | they can also be set when pcre_exec() is called. For details of these |
| 8408 | interfaces, see the pcrebuild documentation and the section on extra |
| 8409 | data for pcre_exec() in the pcreapi documentation. |
| 8410 | |
| 8411 | As a very rough rule of thumb, you should reckon on about 500 bytes per |
| 8412 | recursion. Thus, if you want to limit your stack usage to 8Mb, you |
| 8413 | should set the limit at 16000 recursions. A 64Mb stack, on the other |
| 8414 | hand, can support around 128000 recursions. |
| 8415 | |
| 8416 | In Unix-like environments, the pcretest test program has a command line |
| 8417 | option (-S) that can be used to increase the size of its stack. As long |
| 8418 | as the stack is large enough, another option (-M) can be used to find |
| 8419 | the smallest limits that allow a particular pattern to match a given |
| 8420 | subject string. This is done by calling pcre_exec() repeatedly with |
| 8421 | different limits. |
| 8422 | |
| 8423 | Changing stack size in Unix-like systems |
| 8424 | |
| 8425 | In Unix-like environments, there is not often a problem with the stack |
| 8426 | unless very long strings are involved, though the default limit on |
| 8427 | stack size varies from system to system. Values from 8Mb to 64Mb are |
| 8428 | common. You can find your default limit by running the command: |
| 8429 | |
| 8430 | ulimit -s |
| 8431 | |
| 8432 | Unfortunately, the effect of running out of stack is often SIGSEGV, |
| 8433 | though sometimes a more explicit error message is given. You can nor- |
| 8434 | mally increase the limit on stack size by code such as this: |
| 8435 | |
| 8436 | struct rlimit rlim; |
| 8437 | getrlimit(RLIMIT_STACK, &rlim); |
| 8438 | rlim.rlim_cur = 100*1024*1024; |
| 8439 | setrlimit(RLIMIT_STACK, &rlim); |
| 8440 | |
| 8441 | This reads the current limits (soft and hard) using getrlimit(), then |
| 8442 | attempts to increase the soft limit to 100Mb using setrlimit(). You |
| 8443 | must do this before calling pcre_exec(). |
| 8444 | |
| 8445 | Changing stack size in Mac OS X |
| 8446 | |
| 8447 | Using setrlimit(), as described above, should also work on Mac OS X. It |
| 8448 | is also possible to set a stack size when linking a program. There is a |
| 8449 | discussion about stack sizes in Mac OS X at this web site: |
| 8450 | http://developer.apple.com/qa/qa2005/qa1419.html. |
| 8451 | |
| 8452 | |
| 8453 | AUTHOR |
| 8454 | |
| 8455 | Philip Hazel |
| 8456 | University Computing Service |
| 8457 | Cambridge CB2 3QH, England. |
| 8458 | |
| 8459 | |
| 8460 | REVISION |
| 8461 | |
| 8462 | Last updated: 26 August 2011 |
| 8463 | Copyright (c) 1997-2011 University of Cambridge. |
| 8464 | ------------------------------------------------------------------------------ |
| 8465 | |
| 8466 | |