mingw/html/pod/perlre.html

   1 <?xml version="1.0" ?>
   2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
   3 <html xmlns="http://www.w3.org/1999/xhtml">
   4 <head>
   5 <title>perlre - Perl regular expressions</title>
   6 <meta http-equiv="content-type" content="text/html; charset=utf-8" />
   7 <link rev="made" href="mailto:" />
   8 </head>
   9
  10 <body style="background-color: white">
  11 <table border="0" width="100%" cellspacing="0" cellpadding="3">
  12 <tr><td class="block" style="background-color: #cccccc" valign="middle">
  13 <big><strong><span class="block">&nbsp;perlre - Perl regular expressions</span></strong></big>
  14 </td></tr>
  15 </table>
  16
  17 <p><a name="__index__"></a></p>
  18 <!-- INDEX BEGIN -->
  19
  20 <ul>
  21
  22         <li><a href="#name">NAME</a></li>
  23         <li><a href="#description">DESCRIPTION</a></li>
  24         <ul>
  25
  26                 <li><a href="#regular_expressions">Regular Expressions</a></li>
  27                 <li><a href="#extended_patterns">Extended Patterns</a></li>
  28                 <li><a href="#backtracking">Backtracking</a></li>
  29                 <li><a href="#version_8_regular_expressions">Version 8 Regular Expressions</a></li>
  30                 <li><a href="#warning_on__1_vs__1">Warning on \1 vs $1</a></li>
  31                 <li><a href="#repeated_patterns_matching_zerolength_substring">Repeated patterns matching zero-length substring</a></li>
  32                 <li><a href="#combining_pieces_together">Combining pieces together</a></li>
  33                 <li><a href="#creating_custom_re_engines">Creating custom RE engines</a></li>
  34         </ul>
  35
  36         <li><a href="#bugs">BUGS</a></li>
  37         <li><a href="#see_also">SEE ALSO</a></li>
  38 </ul>
  39 <!-- INDEX END -->
  40
  41 <hr />
  42 <p>
  43 </p>
  44 <h1><a name="name_x_regular_expression__x_regex__x_regexp_">NAME
  45   </a></h1>
  46 <p>perlre - Perl regular expressions</p>
  47 <p>
  48 </p>
  49 <hr />
  50 <h1><a name="description">DESCRIPTION</a></h1>
  51 <p>This page describes the syntax of regular expressions in Perl.</p>
  52 <p>If you haven't used regular expressions before, a quick-start
  53 introduction is available in <a href="file://C|\msysgit\mingw\html/pod/perlrequick.html">the perlrequick manpage</a>, and a longer tutorial
  54 introduction is available in <a href="file://C|\msysgit\mingw\html/pod/perlretut.html">the perlretut manpage</a>.</p>
  55 <p>For reference on how regular expressions are used in matching
  56 operations, plus various examples of the same, see discussions of
  57 <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_m_"><code>m//</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_s_"><code>s///</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_qr_"><code>qr//</code></a> and <code>??</code> in <a href="file://C|\msysgit\mingw\html/pod/perlop.html#regexp_quotelike_operators">Regexp Quote-Like Operators in the perlop manpage</a>.</p>
  58 <p>Matching operations can have various modifiers.  Modifiers
  59 that relate to the interpretation of the regular expression inside
  60 are listed below.  Modifiers that alter the way a regular expression
  61 is used by Perl are detailed in <a href="file://C|\msysgit\mingw\html/pod/perlop.html#regexp_quotelike_operators">Regexp Quote-Like Operators in the perlop manpage</a> and
  62 <a href="file://C|\msysgit\mingw\html/pod/perlop.html#gory_details_of_parsing_quoted_constructs">Gory details of parsing quoted constructs in the perlop manpage</a>.</p>
  63 <dl>
  64 <dt><strong><a name="item_i_x_3c_2fi_3e_x_3cregex_2c_case_2dinsensitive_3e_x">i
  65
  66 </a></strong>
  67
  68 <dd>
  69 <p>Do case-insensitive pattern matching.</p>
  70 </dd>
  71 <dd>
  72 <p>If <code>use locale</code> is in effect, the case map is taken from the current
  73 locale.  See <a href="file://C|\msysgit\mingw\html/pod/perllocale.html">the perllocale manpage</a>.</p>
  74 </dd>
  75 </li>
  76 <dt><strong><a name="item_m_x_3c_2fm_3e_x_3cregex_2c_multiline_3e_x_3cregexp">m
  77    </a></strong>
  78
  79 <dd>
  80 <p>Treat string as multiple lines.  That is, change ``^'' and ``$'' from matching
  81 the start or end of the string to matching the start or end of any
  82 line anywhere within the string.</p>
  83 </dd>
  84 </li>
  85 <dt><strong><a name="item_s_x_3c_2fs_3e_x_3cregex_2c_single_2dline_3e_x_3cre">s
  86
  87 </a></strong>
  88
  89 <dd>
  90 <p>Treat string as single line.  That is, change ``.'' to match any character
  91 whatsoever, even a newline, which normally it would not match.</p>
  92 </dd>
  93 <dd>
  94 <p>The <code>/s</code> and <code>/m</code> modifiers both override the <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$*</code></a> setting.  That
  95 is, no matter what <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$*</code></a> contains, <code>/s</code> without <code>/m</code> will force
  96 ``^'' to match only at the beginning of the string and ``$'' to match
  97 only at the end (or just before a newline at the end) of the string.
  98 Together, as /ms, they let the ``.'' match any character whatsoever,
  99 while still allowing ``^'' and ``$'' to match, respectively, just after
 100 and just before newlines within the string.</p>
 101 </dd>
 102 </li>
 103 <dt><strong><a name="item_x_x_3c_2fx_3e">x
 104 </a></strong>
 105
 106 <dd>
 107 <p>Extend your pattern's legibility by permitting whitespace and comments.</p>
 108 </dd>
 109 </li>
 110 </dl>
 111 <p>These are usually written as ``the <code>/x</code> modifier'', even though the delimiter
 112 in question might not really be a slash.  Any of these
 113 modifiers may also be embedded within the regular expression itself using
 114 the <code>(?...)</code> construct.  See below.</p>
 115 <p>The <code>/x</code> modifier itself needs a little more explanation.  It tells
 116 the regular expression parser to ignore whitespace that is neither
 117 backslashed nor within a character class.  You can use this to break up
 118 your regular expression into (slightly) more readable parts.  The <code>#</code>
 119 character is also treated as a metacharacter introducing a comment,
 120 just as in ordinary Perl code.  This also means that if you want real
 121 whitespace or <code>#</code> characters in the pattern (outside a character
 122 class, where they are unaffected by <code>/x</code>), that you'll either have to
 123 escape them or encode them using octal or hex escapes.  Taken together,
 124 these features go a long way towards making Perl's regular expressions
 125 more readable.  Note that you have to be careful not to include the
 126 pattern delimiter in the comment--perl has no way of knowing you did
 127 not intend to close the pattern early.  See the C-comment deletion code
 128 in <a href="file://C|\msysgit\mingw\html/pod/perlop.html">the perlop manpage</a>.
 129 </p>
 130 <p>
 131 </p>
 132 <h2><a name="regular_expressions">Regular Expressions</a></h2>
 133 <p>The patterns used in Perl pattern matching derive from supplied in
 134 the Version 8 regex routines.  (The routines are derived
 135 (distantly) from Henry Spencer's freely redistributable reimplementation
 136 of the V8 routines.)  See <a href="#version_8_regular_expressions">Version 8 Regular Expressions</a> for
 137 details.</p>
 138 <p>In particular the following metacharacters have their standard <em>egrep</em>-ish
 139 meanings:
 140
 141         </p>
 142 <pre>
 143     \   Quote the next metacharacter
 144     ^   Match the beginning of the line
 145     .   Match any character (except newline)
 146     $   Match the end of the line (or before newline at the end)
 147     |   Alternation
 148     ()  Grouping
 149     []  Character class</pre>
 150 <p>By default, the ``^'' character is guaranteed to match only the
 151 beginning of the string, the ``$'' character only the end (or before the
 152 newline at the end), and Perl does certain optimizations with the
 153 assumption that the string contains only one line.  Embedded newlines
 154 will not be matched by ``^'' or ``$''.  You may, however, wish to treat a
 155 string as a multi-line buffer, such that the ``^'' will match after any
 156 newline within the string, and ``$'' will match before any newline.  At the
 157 cost of a little more overhead, you can do this by using the /m modifier
 158 on the pattern match operator.  (Older programs did this by setting <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$*</code></a>,
 159 but this practice is now deprecated.)
 160   </p>
 161 <p>To simplify multi-line substitutions, the ``.'' character never matches a
 162 newline unless you use the <code>/s</code> modifier, which in effect tells Perl to pretend
 163 the string is a single line--even if it isn't.  The <code>/s</code> modifier also
 164 overrides the setting of <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$*</code></a>, in case you have some (badly behaved) older
 165 code that sets it in another module.
 166  </p>
 167 <p>The following standard quantifiers are recognized:
 168        </p>
 169 <pre>
 170     *      Match 0 or more times
 171     +      Match 1 or more times
 172     ?      Match 1 or 0 times
 173     {n}    Match exactly n times
 174     {n,}   Match at least n times
 175     {n,m}  Match at least n but not more than m times</pre>
 176 <p>(If a curly bracket occurs in any other context, it is treated
 177 as a regular character.  In particular, the lower bound
 178 is not optional.)  The ``*'' modifier is equivalent to <code>{0,}</code>, the ``+''
 179 modifier to <code>{1,}</code>, and the ``?'' modifier to <code>{0,1}</code>.  n and m are limited
 180 to integral values less than a preset limit defined when perl is built.
 181 This is usually 32766 on the most common platforms.  The actual limit can
 182 be seen in the error message generated by code such as this:</p>
 183 <pre>
 184     $_ **= $_ , / {$_} / for 2 .. 42;</pre>
 185 <p>By default, a quantified subpattern is ``greedy'', that is, it will match as
 186 many times as possible (given a particular starting location) while still
 187 allowing the rest of the pattern to match.  If you want it to match the
 188 minimum number of times possible, follow the quantifier with a ``?''.  Note
 189 that the meanings don't change, just the ``greediness'':
 190
 191       </p>
 192 <pre>
 193     *?     Match 0 or more times
 194     +?     Match 1 or more times
 195     ??     Match 0 or 1 time
 196     {n}?   Match exactly n times
 197     {n,}?  Match at least n times
 198     {n,m}? Match at least n but not more than m times</pre>
 199 <p>Because patterns are processed as double quoted strings, the following
 200 also work:
 201
 202    </p>
 203 <pre>
 204     \t          tab                   (HT, TAB)
 205     \n          newline               (LF, NL)
 206     \r          return                (CR)
 207     \f          form feed             (FF)
 208     \a          alarm (bell)          (BEL)
 209     \e          escape (think troff)  (ESC)
 210     \033        octal char (think of a PDP-11)
 211     \x1B        hex char
 212     \x{263a}    wide hex char         (Unicode SMILEY)
 213     \c[         control char
 214     \N{name}    named char
 215     \l          lowercase next char (think vi)
 216     \u          uppercase next char (think vi)
 217     \L          lowercase till \E (think vi)
 218     \U          uppercase till \E (think vi)
 219     \E          end case modification (think vi)
 220     \Q          quote (disable) pattern metacharacters till \E</pre>
 221 <p>If <code>use locale</code> is in effect, the case map used by <code>\l</code>, <code>\L</code>, <code>\u</code>
 222 and <code>\U</code> is taken from the current locale.  See <a href="file://C|\msysgit\mingw\html/pod/perllocale.html">the perllocale manpage</a>.  For
 223 documentation of <code>\N{name}</code>, see <a href="file://C|\msysgit\mingw\html/lib/charnames.html">the charnames manpage</a>.</p>
 224 <p>You cannot include a literal <code>$</code> or <code>@</code> within a <code>\Q</code> sequence.
 225 An unescaped <code>$</code> or <code>@</code> interpolates the corresponding variable,
 226 while escaping will cause the literal string <code>\$</code> to be matched.
 227 You'll need to write something like <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_m_"><code>m/\Quser\E\@\Qhost/</code></a>.</p>
 228 <p>In addition, Perl defines the following:
 229
 230
 231  </p>
 232 <pre>
 233     \w  Match a &quot;word&quot; character (alphanumeric plus &quot;_&quot;)
 234     \W  Match a non-&quot;word&quot; character
 235     \s  Match a whitespace character
 236     \S  Match a non-whitespace character
 237     \d  Match a digit character
 238     \D  Match a non-digit character
 239     \pP Match P, named property.  Use \p{Prop} for longer names.
 240     \PP Match non-P
 241     \X  Match eXtended Unicode &quot;combining character sequence&quot;,
 242         equivalent to (?:\PM\pM*)
 243     \C  Match a single C char (octet) even under Unicode.
 244         NOTE: breaks up characters into their UTF-8 bytes,
 245         so you may end up with malformed pieces of UTF-8.
 246         Unsupported in lookbehind.</pre>
 247 <p>A <code>\w</code> matches a single alphanumeric character (an alphabetic
 248 character, or a decimal digit) or <code>_</code>, not a whole word.  Use <code>\w+</code>
 249 to match a string of Perl-identifier characters (which isn't the same
 250 as matching an English word).  If <code>use locale</code> is in effect, the list
 251 of alphabetic characters generated by <code>\w</code> is taken from the current
 252 locale.  See <a href="file://C|\msysgit\mingw\html/pod/perllocale.html">the perllocale manpage</a>.  You may use <code>\w</code>, <code>\W</code>, <code>\s</code>, <code>\S</code>,
 253 <code>\d</code>, and <code>\D</code> within character classes, but if you try to use them
 254 as endpoints of a range, that's not a range, the ``-'' is understood
 255 literally.  If Unicode is in effect, <code>\s</code> matches also ``\x{85}'',
 256 ``\x{2028}, and ''\x{2029}``, see <a href="file://C|\msysgit\mingw\html/pod/perlunicode.html">the perlunicode manpage</a> for more details about
 257 <code>\pP</code>, <code>\PP</code>, and <code>\X</code>, and <a href="file://C|\msysgit\mingw\html/pod/perluniintro.html">the perluniintro manpage</a> about Unicode in general.
 258 You can define your own <code>\p</code> and <code>\P</code> properties, see <a href="file://C|\msysgit\mingw\html/pod/perlunicode.html">the perlunicode manpage</a>.
 259   </p>
 260 <p>The POSIX character class syntax
 261 </p>
 262 <pre>
 263     [:class:]</pre>
 264 <p>is also available.  The available classes and their backslash
 265 equivalents (if available) are as follows:
 266
 267
 268       </p>
 269 <pre>
 270     alpha
 271     alnum
 272     ascii
 273     blank               [1]
 274     cntrl
 275     digit       \d
 276     graph
 277     lower
 278     print
 279     punct
 280     space       \s      [2]
 281     upper
 282     word        \w      [3]
 283     xdigit</pre>
 284 <dl>
 285 <dt><strong><a name="item__5b1_5d">[1]</a></strong>
 286
 287 <dd>
 288 <p>A GNU extension equivalent to <code>[ \t]</code>, ``all horizontal whitespace''.</p>
 289 </dd>
 290 </li>
 291 <dt><strong><a name="item__5b2_5d">[2]</a></strong>
 292
 293 <dd>
 294 <p>Not exactly equivalent to <code>\s</code> since the <code>[[:space:]]</code> includes
 295 also the (very rare) ``vertical tabulator'', ``\ck'', chr(11).</p>
 296 </dd>
 297 </li>
 298 <dt><strong><a name="item__5b3_5d">[3]</a></strong>
 299
 300 <dd>
 301 <p>A Perl extension, see above.</p>
 302 </dd>
 303 </li>
 304 </dl>
 305 <p>For example use <code>[:upper:]</code> to match all the uppercase characters.
 306 Note that the <code>[]</code> are part of the <code>[::]</code> construct, not part of the
 307 whole character class.  For example:</p>
 308 <pre>
 309     [01[:alpha:]%]</pre>
 310 <p>matches zero, one, any alphabetic character, and the percentage sign.</p>
 311 <p>The following equivalences to Unicode \p{} constructs and equivalent
 312 backslash character classes (if available), will hold:
 313   </p>
 314 <pre>
 315     [:...:]     \p{...}         backslash</pre>
 316 <pre>
 317     alpha       IsAlpha
 318     alnum       IsAlnum
 319     ascii       IsASCII
 320     blank       IsSpace
 321     cntrl       IsCntrl
 322     digit       IsDigit        \d
 323     graph       IsGraph
 324     lower       IsLower
 325     print       IsPrint
 326     punct       IsPunct
 327     space       IsSpace
 328                 IsSpacePerl    \s
 329     upper       IsUpper
 330     word        IsWord
 331     xdigit      IsXDigit</pre>
 332 <p>For example <code>[:lower:]</code> and <code>\p{IsLower}</code> are equivalent.</p>
 333 <p>If the <code>utf8</code> pragma is not used but the <code>locale</code> pragma is, the
 334 classes correlate with the usual <code>isalpha(3)</code> interface (except for
 335 ``word'' and ``blank'').</p>
 336 <p>The assumedly non-obviously named classes are:</p>
 337 <dl>
 338 <dt><strong><a name="item_cntrl_x_3ccntrl_3e">cntrl
 339 </a></strong>
 340
 341 <dd>
 342 <p>Any control character.  Usually characters that don't produce output as
 343 such but instead control the terminal somehow: for example newline and
 344 backspace are control characters.  All characters with <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ord"><code>ord()</code></a> less than
 345 32 are most often classified as control characters (assuming ASCII,
 346 the ISO Latin character sets, and Unicode), as is the character with
 347 the <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_ord"><code>ord()</code></a> value of 127 (<code>DEL</code>).</p>
 348 </dd>
 349 </li>
 350 <dt><strong><a name="item_graph_x_3cgraph_3e">graph
 351 </a></strong>
 352
 353 <dd>
 354 <p>Any alphanumeric or punctuation (special) character.</p>
 355 </dd>
 356 </li>
 357 <dt><strong><a name="item_print_x_3cprint_3e">print
 358 </a></strong>
 359
 360 <dd>
 361 <p>Any alphanumeric or punctuation (special) character or the space character.</p>
 362 </dd>
 363 </li>
 364 <dt><strong><a name="item_punct_x_3cpunct_3e">punct
 365 </a></strong>
 366
 367 <dd>
 368 <p>Any punctuation (special) character.</p>
 369 </dd>
 370 </li>
 371 <dt><strong><a name="item_xdigit_x_3cxdigit_3e">xdigit
 372 </a></strong>
 373
 374 <dd>
 375 <p>Any hexadecimal digit.  Though this may feel silly ([0-9A-Fa-f] would
 376 work just fine) it is included for completeness.</p>
 377 </dd>
 378 </li>
 379 </dl>
 380 <p>You can negate the [::] character classes by prefixing the class name
 381 with a '^'. This is a Perl extension.  For example:
 382 </p>
 383 <pre>
 384     POSIX       traditional Unicode</pre>
 385 <pre>
 386     [:^digit:]      \D      \P{IsDigit}
 387     [:^space:]      \S      \P{IsSpace}
 388     [:^word:]       \W      \P{IsWord}</pre>
 389 <p>Perl respects the POSIX standard in that POSIX character classes are
 390 only supported within a character class.  The POSIX character classes
 391 [.cc.] and [=cc=] are recognized but <strong>not</strong> supported and trying to
 392 use them will cause an error.</p>
 393 <p>Perl defines the following zero-width assertions:
 394
 395
 396
 397      </p>
 398 <pre>
 399     \b  Match a word boundary
 400     \B  Match a non-(word boundary)
 401     \A  Match only at beginning of string
 402     \Z  Match only at end of string, or before newline at the end
 403     \z  Match only at end of string
 404     \G  Match only at pos() (e.g. at the end-of-match position
 405         of prior m//g)</pre>
 406 <p>A word boundary (<code>\b</code>) is a spot between two characters
 407 that has a <code>\w</code> on one side of it and a <code>\W</code> on the other side
 408 of it (in either order), counting the imaginary characters off the
 409 beginning and end of the string as matching a <code>\W</code>.  (Within
 410 character classes <code>\b</code> represents backspace rather than a word
 411 boundary, just as it normally does in any double-quoted string.)
 412 The <code>\A</code> and <code>\Z</code> are just like ``^'' and ``$'', except that they
 413 won't match multiple times when the <code>/m</code> modifier is used, while
 414 ``^'' and ``$'' will match at every internal line boundary.  To match
 415 the actual end of the string and not ignore an optional trailing
 416 newline, use <code>\z</code>.
 417     </p>
 418 <p>The <code>\G</code> assertion can be used to chain global matches (using
 419 <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_m_"><code>m//g</code></a>), as described in <a href="file://C|\msysgit\mingw\html/pod/perlop.html#regexp_quotelike_operators">Regexp Quote-Like Operators in the perlop manpage</a>.
 420 It is also useful when writing <code>lex</code>-like scanners, when you have
 421 several patterns that you want to match against consequent substrings
 422 of your string, see the previous reference.  The actual location
 423 where <code>\G</code> will match can also be influenced by using <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_pos"><code>pos()</code></a> as
 424 an lvalue: see <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_pos">pos in the perlfunc manpage</a>. Currently <code>\G</code> is only fully
 425 supported when anchored to the start of the pattern; while it
 426 is permitted to use it elsewhere, as in <code>/(?&lt;=\G..)./g</code>, some
 427 such uses (<code>/.\G/g</code>, for example) currently cause problems, and
 428 it is recommended that you avoid such usage for now.
 429 </p>
 430 <p>The bracketing construct <code>( ... )</code> creates capture buffers.  To
 431 refer to the digit'th buffer use \&lt;digit&gt; within the
 432 match.  Outside the match use ``$'' instead of ``\''.  (The
 433 \&lt;digit&gt; notation works in certain circumstances outside
 434 the match.  See the warning below about \1 vs $1 for details.)
 435 Referring back to another part of the match is called a
 436 <em>backreference</em>.
 437
 438  </p>
 439 <p>There is no limit to the number of captured substrings that you may
 440 use.  However Perl also uses \10, \11, etc. as aliases for \010,
 441 \011, etc.  (Recall that 0 means octal, so \011 is the character at
 442 number 9 in your coded character set; which would be the 10th character,
 443 a horizontal tab under ASCII.)  Perl resolves this
 444 ambiguity by interpreting \10 as a backreference only if at least 10
 445 left parentheses have opened before it.  Likewise \11 is a
 446 backreference only if at least 11 left parentheses have opened
 447 before it.  And so on.  \1 through \9 are always interpreted as
 448 backreferences.</p>
 449 <p>Examples:</p>
 450 <pre>
 451     s/^([^ ]*) *([^ ]*)/$2 $1/;     # swap first two words</pre>
 452 <pre>
 453      if (/(.)\1/) {                 # find first doubled char
 454          print &quot;'$1' is the first doubled character\n&quot;;
 455      }</pre>
 456 <pre>
 457     if (/Time: (..):(..):(..)/) {   # parse out values
 458         $hours = $1;
 459         $minutes = $2;
 460         $seconds = $3;
 461     }</pre>
 462 <p>Several special variables also refer back to portions of the previous
 463 match.  <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$+</code></a> returns whatever the last bracket match matched.
 464 <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a> returns the entire matched string.  (At one point <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item__0"><code>$0</code></a> did
 465 also, but now it returns the name of the program.)  <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$`</code></a> returns
 466 everything before the matched string.  <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$'</code></a> returns everything
 467 after the matched string. And <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___n"><code>$^N</code></a> contains whatever was matched by
 468 the most-recently closed group (submatch). <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___n"><code>$^N</code></a> can be used in
 469 extended patterns (see below), for example to assign a submatch to a
 470 variable.
 471     </p>
 472 <p>The numbered match variables ($1, $2, $3, etc.) and the related punctuation
 473 set (<a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$+</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$`</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$'</code></a>, and <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___n"><code>$^N</code></a>) are all dynamically scoped
 474 until the end of the enclosing block or until the next successful
 475 match, whichever comes first.  (See <a href="file://C|\msysgit\mingw\html/pod/perlsyn.html#compound_statements">Compound Statements in the perlsyn manpage</a>.)
 476
 477         </p>
 478 <p><strong>NOTE</strong>: failed matches in Perl do not reset the match variables,
 479 which makes it easier to write code that tests for a series of more
 480 specific cases and remembers the best match.</p>
 481 <p><strong>WARNING</strong>: Once Perl sees that you need one of <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$`</code></a>, or
 482 <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$'</code></a> anywhere in the program, it has to provide them for every
 483 pattern match.  This may substantially slow your program.  Perl
 484 uses the same mechanism to produce $1, $2, etc, so you also pay a
 485 price for each pattern that contains capturing parentheses.  (To
 486 avoid this cost while retaining the grouping behaviour, use the
 487 extended regular expression <code>(?: ... )</code> instead.)  But if you never
 488 use <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$`</code></a> or <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$'</code></a>, then patterns <em>without</em> capturing
 489 parentheses will not be penalized.  So avoid <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$'</code></a>, and <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$`</code></a>
 490 if you can, but if you can't (and some algorithms really appreciate
 491 them), once you've used them once, use them at will, because you've
 492 already paid the price.  As of 5.005, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a> is not so costly as the
 493 other two.
 494   </p>
 495 <p>Backslashed metacharacters in Perl are alphanumeric, such as <code>\b</code>,
 496 <code>\w</code>, <code>\n</code>.  Unlike some other regular expression languages, there
 497 are no backslashed symbols that aren't alphanumeric.  So anything
 498 that looks like \\, \(, \), \&lt;, \&gt;, \{, or \} is always
 499 interpreted as a literal character, not a metacharacter.  This was
 500 once used in a common idiom to disable or quote the special meanings
 501 of regular expression metacharacters in a string that you want to
 502 use for a pattern. Simply quote all non-``word'' characters:</p>
 503 <pre>
 504     $pattern =~ s/(\W)/\\$1/g;</pre>
 505 <p>(If <code>use locale</code> is set, then this depends on the current locale.)
 506 Today it is more common to use the <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_quotemeta"><code>quotemeta()</code></a> function or the <code>\Q</code>
 507 metaquoting escape sequence to disable all metacharacters' special
 508 meanings like this:</p>
 509 <pre>
 510     /$unquoted\Q$quoted\E$unquoted/</pre>
 511 <p>Beware that if you put literal backslashes (those not inside
 512 interpolated variables) between <code>\Q</code> and <code>\E</code>, double-quotish
 513 backslash interpolation may lead to confusing results.  If you
 514 <em>need</em> to use literal backslashes within <code>\Q...\E</code>,
 515 consult <a href="file://C|\msysgit\mingw\html/pod/perlop.html#gory_details_of_parsing_quoted_constructs">Gory details of parsing quoted constructs in the perlop manpage</a>.</p>
 516 <p>
 517 </p>
 518 <h2><a name="extended_patterns">Extended Patterns</a></h2>
 519 <p>Perl also defines a consistent extension syntax for features not
 520 found in standard tools like <strong>awk</strong> and <strong>lex</strong>.  The syntax is a
 521 pair of parentheses with a question mark as the first thing within
 522 the parentheses.  The character after the question mark indicates
 523 the extension.</p>
 524 <p>The stability of these extensions varies widely.  Some have been
 525 part of the core language for many years.  Others are experimental
 526 and may change without warning or be completely removed.  Check
 527 the documentation on an individual feature to verify its current
 528 status.</p>
 529 <p>A question mark was chosen for this and for the minimal-matching
 530 construct because 1) question marks are rare in older regular
 531 expressions, and 2) whenever you see one, you should stop and
 532 ``question'' exactly what is going on.  That's psychology...</p>
 533 <dl>
 534 <dt><strong><a name="item__28_3f_23text_29_x_3c_28_3f_23_29_3e"><code>(?#text)</code>
 535 </a></strong>
 536
 537 <dd>
 538 <p>A comment.  The text is ignored.  If the <code>/x</code> modifier enables
 539 whitespace formatting, a simple <code>#</code> will suffice.  Note that Perl closes
 540 the comment as soon as it sees a <code>)</code>, so there is no way to put a literal
 541 <code>)</code> in the comment.</p>
 542 </dd>
 543 </li>
 544 <dt><strong><a name="item__28_3fimsx_2dimsx_29_x_3c_28_3f_29_3e"><code>(?imsx-imsx)</code>
 545 </a></strong>
 546
 547 <dd>
 548 <p>One or more embedded pattern-match modifiers, to be turned on (or
 549 turned off, if preceded by <code>-</code>) for the remainder of the pattern or
 550 the remainder of the enclosing pattern group (if any). This is
 551 particularly useful for dynamic patterns, such as those read in from a
 552 configuration file, read in as an argument, are specified in a table
 553 somewhere, etc.  Consider the case that some of which want to be case
 554 sensitive and some do not.  The case insensitive ones need to include
 555 merely <code>(?i)</code> at the front of the pattern.  For example:</p>
 556 </dd>
 557 <dd>
 558 <pre>
 559     $pattern = &quot;foobar&quot;;
 560     if ( /$pattern/i ) { }</pre>
 561 </dd>
 562 <dd>
 563 <pre>
 564     # more flexible:</pre>
 565 </dd>
 566 <dd>
 567 <pre>
 568     $pattern = &quot;(?i)foobar&quot;;
 569     if ( /$pattern/ ) { }</pre>
 570 </dd>
 571 <dd>
 572 <p>These modifiers are restored at the end of the enclosing group. For example,</p>
 573 </dd>
 574 <dd>
 575 <pre>
 576     ( (?i) blah ) \s+ \1</pre>
 577 </dd>
 578 <dd>
 579 <p>will match a repeated (<em>including the case</em>!) word <code>blah</code> in any
 580 case, assuming <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_x"><code>x</code></a> modifier, and no <code>i</code> modifier outside this
 581 group.</p>
 582 </dd>
 583 </li>
 584 <dt><strong><a name="item__28_3f_3apattern_29_x_3c_28_3f_3a_29_3e"><code>(?:pattern)</code>
 585 </a></strong>
 586
 587 <dt><strong><a name="item__28_3fimsx_2dimsx_3apattern_29"><code>(?imsx-imsx:pattern)</code></a></strong>
 588
 589 <dd>
 590 <p>This is for clustering, not capturing; it groups subexpressions like
 591 ``()'', but doesn't make backreferences as ``()'' does.  So</p>
 592 </dd>
 593 <dd>
 594 <pre>
 595     @fields = split(/\b(?:a|b|c)\b/)</pre>
 596 </dd>
 597 <dd>
 598 <p>is like</p>
 599 </dd>
 600 <dd>
 601 <pre>
 602     @fields = split(/\b(a|b|c)\b/)</pre>
 603 </dd>
 604 <dd>
 605 <p>but doesn't spit out extra fields.  It's also cheaper not to capture
 606 characters if you don't need to.</p>
 607 </dd>
 608 <dd>
 609 <p>Any letters between <code>?</code> and <code>:</code> act as flags modifiers as with
 610 <code>(?imsx-imsx)</code>.  For example,</p>
 611 </dd>
 612 <dd>
 613 <pre>
 614     /(?s-i:more.*than).*million/i</pre>
 615 </dd>
 616 <dd>
 617 <p>is equivalent to the more verbose</p>
 618 </dd>
 619 <dd>
 620 <pre>
 621     /(?:(?s-i)more.*than).*million/i</pre>
 622 </dd>
 623 </li>
 624 <dt><strong><a name="item__28_3f_3dpattern_29_x_3c_28_3f_3d_29_3e_x_3clook_2"><code>(?=pattern)</code>
 625   </a></strong>
 626
 627 <dd>
 628 <p>A zero-width positive look-ahead assertion.  For example, <code>/\w+(?=\t)/</code>
 629 matches a word followed by a tab, without including the tab in <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>.</p>
 630 </dd>
 631 </li>
 632 <dt><strong><a name="item__28_3f_21pattern_29_x_3c_28_3f_21_29_3e_x_3clook_2"><code>(?!pattern)</code>
 633   </a></strong>
 634
 635 <dd>
 636 <p>A zero-width negative look-ahead assertion.  For example <code>/foo(?!bar)/</code>
 637 matches any occurrence of ``foo'' that isn't followed by ``bar''.  Note
 638 however that look-ahead and look-behind are NOT the same thing.  You cannot
 639 use this for look-behind.</p>
 640 </dd>
 641 <dd>
 642 <p>If you are looking for a ``bar'' that isn't preceded by a ``foo'', <code>/(?!foo)bar/</code>
 643 will not do what you want.  That's because the <code>(?!foo)</code> is just saying that
 644 the next thing cannot be ``foo''--and it's not, it's a ``bar'', so ``foobar'' will
 645 match.  You would have to do something like <code>/(?!foo)...bar/</code> for that.   We
 646 say ``like'' because there's the case of your ``bar'' not having three characters
 647 before it.  You could cover that this way: <code>/(?:(?!foo)...|^.{0,2})bar/</code>.
 648 Sometimes it's still easier just to say:</p>
 649 </dd>
 650 <dd>
 651 <pre>
 652     if (/bar/ &amp;&amp; $` !~ /foo$/)</pre>
 653 </dd>
 654 <dd>
 655 <p>For look-behind see below.</p>
 656 </dd>
 657 </li>
 658 <dt><strong><a name="item__28_3f_3c_3dpattern_29_x_3c_28_3f_3c_3d_29_3e_x_3c"><code>(?&lt;=pattern)</code>
 659   </a></strong>
 660
 661 <dd>
 662 <p>A zero-width positive look-behind assertion.  For example, <code>/(?&lt;=\t)\w+/</code>
 663 matches a word that follows a tab, without including the tab in <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$&amp;</code></a>.
 664 Works only for fixed-width look-behind.</p>
 665 </dd>
 666 </li>
 667 <dt><strong><a name="item__28_3f_3c_21pattern_29_x_3c_28_3f_3c_21_29_3e_x_3c"><code>(?&lt;!pattern)</code>
 668   </a></strong>
 669
 670 <dd>
 671 <p>A zero-width negative look-behind assertion.  For example <code>/(?&lt;!bar)foo/</code>
 672 matches any occurrence of ``foo'' that does not follow ``bar''.  Works
 673 only for fixed-width look-behind.</p>
 674 </dd>
 675 </li>
 676 <dt><strong><a name="item__28_3f_7b_code__7d_29_x_3c_28_3f_7b_7d_29_3e_x_3cr"><code>(?{ code })</code>
 677    </a></strong>
 678
 679 <dd>
 680 <p><strong>WARNING</strong>: This extended regular expression feature is considered
 681 highly experimental, and may be changed or deleted without notice.</p>
 682 </dd>
 683 <dd>
 684 <p>This zero-width assertion evaluates any embedded Perl code.  It
 685 always succeeds, and its <code>code</code> is not interpolated.  Currently,
 686 the rules to determine where the <code>code</code> ends are somewhat convoluted.</p>
 687 </dd>
 688 <dd>
 689 <p>This feature can be used together with the special variable <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___n"><code>$^N</code></a> to
 690 capture the results of submatches in variables without having to keep
 691 track of the number of nested parentheses. For example:</p>
 692 </dd>
 693 <dd>
 694 <pre>
 695   $_ = &quot;The brown fox jumps over the lazy dog&quot;;
 696   /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
 697   print &quot;color = $color, animal = $animal\n&quot;;</pre>
 698 </dd>
 699 <dd>
 700 <p>Inside the <code>(?{...})</code> block, <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___"><code>$_</code></a> refers to the string the regular
 701 expression is matching against. You can also use <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_pos"><code>pos()</code></a> to know what is
 702 the current position of matching within this string.</p>
 703 </dd>
 704 <dd>
 705 <p>The <code>code</code> is properly scoped in the following sense: If the assertion
 706 is backtracked (compare <a href="#backtracking">Backtracking</a>), all changes introduced after
 707 <code>local</code>ization are undone, so that</p>
 708 </dd>
 709 <dd>
 710 <pre>
 711   $_ = 'a' x 8;
 712   m&lt;
 713      (?{ $cnt = 0 })                    # Initialize $cnt.
 714      (
 715        a
 716        (?{
 717            local $cnt = $cnt + 1;       # Update $cnt, backtracking-safe.
 718        })
 719      )*
 720      aaaa
 721      (?{ $res = $cnt })                 # On success copy to non-localized
 722                                         # location.
 723    &gt;x;</pre>
 724 </dd>
 725 <dd>
 726 <p>will set <code>$res = 4</code>.  Note that after the match, $cnt returns to the globally
 727 introduced value, because the scopes that restrict <code>local</code> operators
 728 are unwound.</p>
 729 </dd>
 730 <dd>
 731 <p>This assertion may be used as a <a href="#item__28_3f_28condition_29yes_2dpattern_7cno_2dpattern_"><code>(?(condition)yes-pattern|no-pattern)</code></a>
 732 switch.  If <em>not</em> used in this way, the result of evaluation of
 733 <code>code</code> is put into the special variable <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___r"><code>$^R</code></a>.  This happens
 734 immediately, so <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___r"><code>$^R</code></a> can be used from other <code>(?{ code })</code> assertions
 735 inside the same regular expression.</p>
 736 </dd>
 737 <dd>
 738 <p>The assignment to <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___r"><code>$^R</code></a> above is properly localized, so the old
 739 value of <a href="file://C|\msysgit\mingw\html/pod/perlvar.html#item___r"><code>$^R</code></a> is restored if the assertion is backtracked; compare
 740 <a href="#backtracking">Backtracking</a>.</p>
 741 </dd>
 742 <dd>
 743 <p>For reasons of security, this construct is forbidden if the regular
 744 expression involves run-time interpolation of variables, unless the
 745 perilous <code>use re 'eval'</code> pragma has been used (see <a href="file://C|\msysgit\mingw\html/lib/re.html">the re manpage</a>), or the
 746 variables contain results of <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_qr_"><code>qr//</code></a> operator (see
 747 <a href="file://C|\msysgit\mingw\html/pod/perlop.html#qr_string_imosx">qr/STRING/imosx in the perlop manpage</a>).</p>
 748 </dd>
 749 <dd>
 750 <p>This restriction is because of the wide-spread and remarkably convenient
 751 custom of using run-time determined strings as patterns.  For example:</p>
 752 </dd>
 753 <dd>
 754 <pre>
 755     $re = &lt;&gt;;
 756     chomp $re;
 757     $string =~ /$re/;</pre>
 758 </dd>
 759 <dd>
 760 <p>Before Perl knew how to execute interpolated code within a pattern,
 761 this operation was completely safe from a security point of view,
 762 although it could raise an exception from an illegal pattern.  If
 763 you turn on the <code>use re 'eval'</code>, though, it is no longer secure,
 764 so you should only do so if you are also using taint checking.
 765 Better yet, use the carefully constrained evaluation within a Safe
 766 compartment.  See <a href="file://C|\msysgit\mingw\html/pod/perlsec.html">the perlsec manpage</a> for details about both these mechanisms.</p>
 767 </dd>
 768 </li>
 769 <dt><strong><a name="item__28_3f_3f_7b_code__7d_29_x_3c_28_3f_3f_7b_7d_29_3e"><code>(??{ code })</code>
 770
 771
 772   </a></strong>
 773
 774 <dd>
 775 <p><strong>WARNING</strong>: This extended regular expression feature is considered
 776 highly experimental, and may be changed or deleted without notice.
 777 A simplified version of the syntax may be introduced for commonly
 778 used idioms.</p>
 779 </dd>
 780 <dd>
 781 <p>This is a ``postponed'' regular subexpression.  The <code>code</code> is evaluated
 782 at run time, at the moment this subexpression may match.  The result
 783 of evaluation is considered as a regular expression and matched as
 784 if it were inserted instead of this construct.</p>
 785 </dd>
 786 <dd>
 787 <p>The <code>code</code> is not interpolated.  As before, the rules to determine
 788 where the <code>code</code> ends are currently somewhat convoluted.</p>
 789 </dd>
 790 <dd>
 791 <p>The following pattern matches a parenthesized group:</p>
 792 </dd>
 793 <dd>
 794 <pre>
 795   $re = qr{
 796              \(
 797              (?:
 798                 (?&gt; [^()]+ )    # Non-parens without backtracking
 799               |
 800                 (??{ $re })     # Group with matching parens
 801              )*
 802              \)
 803           }x;</pre>
 804 </dd>
 805 </li>
 806 <dt><strong><a name="item__28_3f_3epattern_29_x_3cbacktrack_3e_x_3cbacktrack"><code>(?&gt;pattern)</code>
 807  </a></strong>
 808
 809 <dd>
 810 <p><strong>WARNING</strong>: This extended regular expression feature is considered
 811 highly experimental, and may be changed or deleted without notice.</p>
 812 </dd>
 813 <dd>
 814 <p>An ``independent'' subexpression, one which matches the substring
 815 that a <em>standalone</em> <code>pattern</code> would match if anchored at the given
 816 position, and it matches <em>nothing other than this substring</em>.  This
 817 construct is useful for optimizations of what would otherwise be
 818 ``eternal'' matches, because it will not backtrack (see <a href="#backtracking">Backtracking</a>).
 819 It may also be useful in places where the ``grab all you can, and do not
 820 give anything back'' semantic is desirable.</p>
 821 </dd>
 822 <dd>
 823 <p>For example: <code>^(?&gt;a*)ab</code> will never match, since <code>(?&gt;a*)</code>
 824 (anchored at the beginning of string, as above) will match <em>all</em>
 825 characters <code>a</code> at the beginning of string, leaving no <code>a</code> for
 826 <code>ab</code> to match.  In contrast, <code>a*ab</code> will match the same as <code>a+b</code>,
 827 since the match of the subgroup <code>a*</code> is influenced by the following
 828 group <code>ab</code> (see <a href="#backtracking">Backtracking</a>).  In particular, <code>a*</code> inside
 829 <code>a*ab</code> will match fewer characters than a standalone <code>a*</code>, since
 830 this makes the tail match.</p>
 831 </dd>
 832 <dd>
 833 <p>An effect similar to <code>(?&gt;pattern)</code> may be achieved by writing
 834 <code>(?=(pattern))\1</code>.  This matches the same substring as a standalone
 835 <code>a+</code>, and the following <code>\1</code> eats the matched string; it therefore
 836 makes a zero-length assertion into an analogue of <code>(?&gt;...)</code>.
 837 (The difference between these two constructs is that the second one
 838 uses a capturing group, thus shifting ordinals of backreferences
 839 in the rest of a regular expression.)</p>
 840 </dd>
 841 <dd>
 842 <p>Consider this pattern:</p>
 843 </dd>
 844 <dd>
 845 <pre>
 846     m{ \(
 847           (
 848             [^()]+              # x+
 849           |
 850             \( [^()]* \)
 851           )+
 852        \)
 853      }x</pre>
 854 </dd>
 855 <dd>
 856 <p>That will efficiently match a nonempty group with matching parentheses
 857 two levels deep or less.  However, if there is no such group, it
 858 will take virtually forever on a long string.  That's because there
 859 are so many different ways to split a long string into several
 860 substrings.  This is what <code>(.+)+</code> is doing, and <code>(.+)+</code> is similar
 861 to a subpattern of the above pattern.  Consider how the pattern
 862 above detects no-match on <code>((()aaaaaaaaaaaaaaaaaa</code> in several
 863 seconds, but that each extra letter doubles this time.  This
 864 exponential performance will make it appear that your program has
 865 hung.  However, a tiny change to this pattern</p>
 866 </dd>
 867 <dd>
 868 <pre>
 869     m{ \(
 870           (
 871             (?&gt; [^()]+ )        # change x+ above to (?&gt; x+ )
 872           |
 873             \( [^()]* \)
 874           )+
 875        \)
 876      }x</pre>
 877 </dd>
 878 <dd>
 879 <p>which uses <code>(?&gt;...)</code> matches exactly when the one above does (verifying
 880 this yourself would be a productive exercise), but finishes in a fourth
 881 the time when used on a similar string with 1000000 <code>a</code>s.  Be aware,
 882 however, that this pattern currently triggers a warning message under
 883 the <code>use warnings</code> pragma or <strong>-w</strong> switch saying it
 884 <code>&quot;matches null string many times in regex&quot;</code>.</p>
 885 </dd>
 886 <dd>
 887 <p>On simple groups, such as the pattern <code>(?&gt; [^()]+ )</code>, a comparable
 888 effect may be achieved by negative look-ahead, as in <code>[^()]+ (?! [^()] )</code>.
 889 This was only 4 times slower on a string with 1000000 <code>a</code>s.</p>
 890 </dd>
 891 <dd>
 892 <p>The ``grab all you can, and do not give anything back'' semantic is desirable
 893 in many situations where on the first sight a simple <code>()*</code> looks like
 894 the correct solution.  Suppose we parse text with comments being delimited
 895 by <code>#</code> followed by some optional (horizontal) whitespace.  Contrary to
 896 its appearance, <code>#[ \t]*</code> <em>is not</em> the correct subexpression to match
 897 the comment delimiter, because it may ``give up'' some whitespace if
 898 the remainder of the pattern can be made to match that way.  The correct
 899 answer is either one of these:</p>
 900 </dd>
 901 <dd>
 902 <pre>
 903     (?&gt;#[ \t]*)
 904     #[ \t]*(?![ \t])</pre>
 905 </dd>
 906 <dd>
 907 <p>For example, to grab non-empty comments into $1, one should use either
 908 one of these:</p>
 909 </dd>
 910 <dd>
 911 <pre>
 912     / (?&gt; \# [ \t]* ) (        .+ ) /x;
 913     /     \# [ \t]*   ( [^ \t] .* ) /x;</pre>
 914 </dd>
 915 <dd>
 916 <p>Which one you pick depends on which of these expressions better reflects
 917 the above specification of comments.</p>
 918 </dd>
 919 </li>
 920 <dt><strong><a name="item__28_3f_28condition_29yes_2dpattern_7cno_2dpattern_"><code>(?(condition)yes-pattern|no-pattern)</code>
 921 </a></strong>
 922
 923 <dt><strong><a name="item__28_3f_28condition_29yes_2dpattern_29"><code>(?(condition)yes-pattern)</code></a></strong>
 924
 925 <dd>
 926 <p><strong>WARNING</strong>: This extended regular expression feature is considered
 927 highly experimental, and may be changed or deleted without notice.</p>
 928 </dd>
 929 <dd>
 930 <p>Conditional expression.  <code>(condition)</code> should be either an integer in
 931 parentheses (which is valid if the corresponding pair of parentheses
 932 matched), or look-ahead/look-behind/evaluate zero-width assertion.</p>
 933 </dd>
 934 <dd>
 935 <p>For example:</p>
 936 </dd>
 937 <dd>
 938 <pre>
 939     m{ ( \( )?
 940        [^()]+
 941        (?(1) \) )
 942      }x</pre>
 943 </dd>
 944 <dd>
 945 <p>matches a chunk of non-parentheses, possibly included in parentheses
 946 themselves.</p>
 947 </dd>
 948 </li>
 949 </dl>
 950 <p>
 951 </p>
 952 <h2><a name="backtracking_x_backtrack__x_backtracking_">Backtracking
 953  </a></h2>
 954 <p>NOTE: This section presents an abstract approximation of regular
 955 expression behavior.  For a more rigorous (and complicated) view of
 956 the rules involved in selecting a match among possible alternatives,
 957 see <a href="#combining_pieces_together">Combining pieces together</a>.</p>
 958 <p>A fundamental feature of regular expression matching involves the
 959 notion called <em>backtracking</em>, which is currently used (when needed)
 960 by all regular expression quantifiers, namely <code>*</code>, <code>*?</code>, <code>+</code>,
 961 <code>+?</code>, <code>{n,m}</code>, and <code>{n,m}?</code>.  Backtracking is often optimized
 962 internally, but the general principle outlined here is valid.</p>
 963 <p>For a regular expression to match, the <em>entire</em> regular expression must
 964 match, not just part of it.  So if the beginning of a pattern containing a
 965 quantifier succeeds in a way that causes later parts in the pattern to
 966 fail, the matching engine backs up and recalculates the beginning
 967 part--that's why it's called backtracking.</p>
 968 <p>Here is an example of backtracking:  Let's say you want to find the
 969 word following ``foo'' in the string ``Food is on the foo table.'':</p>
 970 <pre>
 971     $_ = &quot;Food is on the foo table.&quot;;
 972     if ( /\b(foo)\s+(\w+)/i ) {
 973         print &quot;$2 follows $1.\n&quot;;
 974     }</pre>
 975 <p>When the match runs, the first part of the regular expression (<a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_b"><code>\b(foo)</code></a>)
 976 finds a possible match right at the beginning of the string, and loads up
 977 $1 with ``Foo''.  However, as soon as the matching engine sees that there's
 978 no whitespace following the ``Foo'' that it had saved in $1, it realizes its
 979 mistake and starts over again one character after where it had the
 980 tentative match.  This time it goes all the way until the next occurrence
 981 of ``foo''. The complete regular expression matches this time, and you get
 982 the expected output of ``table follows foo.''</p>
 983 <p>Sometimes minimal matching can help a lot.  Imagine you'd like to match
 984 everything between ``foo'' and ``bar''.  Initially, you write something
 985 like this:</p>
 986 <pre>
 987     $_ =  &quot;The food is under the bar in the barn.&quot;;
 988     if ( /foo(.*)bar/ ) {
 989         print &quot;got &lt;$1&gt;\n&quot;;
 990     }</pre>
 991 <p>Which perhaps unexpectedly yields:</p>
 992 <pre>
 993   got &lt;d is under the bar in the &gt;</pre>
 994 <p>That's because <code>.*</code> was greedy, so you get everything between the
 995 <em>first</em> ``foo'' and the <em>last</em> ``bar''.  Here it's more effective
 996 to use minimal matching to make sure you get the text between a ``foo''
 997 and the first ``bar'' thereafter.</p>
 998 <pre>
 999     if ( /foo(.*?)bar/ ) { print &quot;got &lt;$1&gt;\n&quot; }
1000   got &lt;d is under the &gt;</pre>
1001 <p>Here's another example: let's say you'd like to match a number at the end
1002 of a string, and you also want to keep the preceding part of the match.
1003 So you write this:</p>
1004 <pre>
1005     $_ = &quot;I have 2 numbers: 53147&quot;;
1006     if ( /(.*)(\d*)/ ) {                                # Wrong!
1007         print &quot;Beginning is &lt;$1&gt;, number is &lt;$2&gt;.\n&quot;;
1008     }</pre>
1009 <p>That won't work at all, because <code>.*</code> was greedy and gobbled up the
1010 whole string. As <code>\d*</code> can match on an empty string the complete
1011 regular expression matched successfully.</p>
1012 <pre>
1013     Beginning is &lt;I have 2 numbers: 53147&gt;, number is &lt;&gt;.</pre>
1014 <p>Here are some variants, most of which don't work:</p>
1015 <pre>
1016     $_ = &quot;I have 2 numbers: 53147&quot;;
1017     @pats = qw{
1018         (.*)(\d*)
1019         (.*)(\d+)
1020         (.*?)(\d*)
1021         (.*?)(\d+)
1022         (.*)(\d+)$
1023         (.*?)(\d+)$
1024         (.*)\b(\d+)$
1025         (.*\D)(\d+)$
1026     };</pre>
1027 <pre>
1028     for $pat (@pats) {
1029         printf &quot;%-12s &quot;, $pat;
1030         if ( /$pat/ ) {
1031             print &quot;&lt;$1&gt; &lt;$2&gt;\n&quot;;
1032         } else {
1033             print &quot;FAIL\n&quot;;
1034         }
1035     }</pre>
1036 <p>That will print out:</p>
1037 <pre>
1038     (.*)(\d*)    &lt;I have 2 numbers: 53147&gt; &lt;&gt;
1039     (.*)(\d+)    &lt;I have 2 numbers: 5314&gt; &lt;7&gt;
1040     (.*?)(\d*)   &lt;&gt; &lt;&gt;
1041     (.*?)(\d+)   &lt;I have &gt; &lt;2&gt;
1042     (.*)(\d+)$   &lt;I have 2 numbers: 5314&gt; &lt;7&gt;
1043     (.*?)(\d+)$  &lt;I have 2 numbers: &gt; &lt;53147&gt;
1044     (.*)\b(\d+)$ &lt;I have 2 numbers: &gt; &lt;53147&gt;
1045     (.*\D)(\d+)$ &lt;I have 2 numbers: &gt; &lt;53147&gt;</pre>
1046 <p>As you see, this can be a bit tricky.  It's important to realize that a
1047 regular expression is merely a set of assertions that gives a definition
1048 of success.  There may be 0, 1, or several different ways that the
1049 definition might succeed against a particular string.  And if there are
1050 multiple ways it might succeed, you need to understand backtracking to
1051 know which variety of success you will achieve.</p>
1052 <p>When using look-ahead assertions and negations, this can all get even
1053 trickier.  Imagine you'd like to find a sequence of non-digits not
1054 followed by ``123''.  You might try to write that as</p>
1055 <pre>
1056     $_ = &quot;ABC123&quot;;
1057     if ( /^\D*(?!123)/ ) {              # Wrong!
1058         print &quot;Yup, no 123 in $_\n&quot;;
1059     }</pre>
1060 <p>But that isn't going to match; at least, not the way you're hoping.  It
1061 claims that there is no 123 in the string.  Here's a clearer picture of
1062 why that pattern matches, contrary to popular expectations:</p>
1063 <pre>
1064     $x = 'ABC123';
1065     $y = 'ABC445';</pre>
1066 <pre>
1067     print &quot;1: got $1\n&quot; if $x =~ /^(ABC)(?!123)/;
1068     print &quot;2: got $1\n&quot; if $y =~ /^(ABC)(?!123)/;</pre>
1069 <pre>
1070     print &quot;3: got $1\n&quot; if $x =~ /^(\D*)(?!123)/;
1071     print &quot;4: got $1\n&quot; if $y =~ /^(\D*)(?!123)/;</pre>
1072 <p>This prints</p>
1073 <pre>
1074     2: got ABC
1075     3: got AB
1076     4: got ABC</pre>
1077 <p>You might have expected test 3 to fail because it seems to a more
1078 general purpose version of test 1.  The important difference between
1079 them is that test 3 contains a quantifier (<code>\D*</code>) and so can use
1080 backtracking, whereas test 1 will not.  What's happening is
1081 that you've asked ``Is it true that at the start of $x, following 0 or more
1082 non-digits, you have something that's not 123?''  If the pattern matcher had
1083 let <code>\D*</code> expand to ``ABC'', this would have caused the whole pattern to
1084 fail.</p>
1085 <p>The search engine will initially match <code>\D*</code> with ``ABC''.  Then it will
1086 try to match <code>(?!123</code> with ``123'', which fails.  But because
1087 a quantifier (<code>\D*</code>) has been used in the regular expression, the
1088 search engine can backtrack and retry the match differently
1089 in the hope of matching the complete regular expression.</p>
1090 <p>The pattern really, <em>really</em> wants to succeed, so it uses the
1091 standard pattern back-off-and-retry and lets <code>\D*</code> expand to just ``AB'' this
1092 time.  Now there's indeed something following ``AB'' that is not
1093 ``123''.  It's ``C123'', which suffices.</p>
1094 <p>We can deal with this by using both an assertion and a negation.
1095 We'll say that the first part in $1 must be followed both by a digit
1096 and by something that's not ``123''.  Remember that the look-aheads
1097 are zero-width expressions--they only look, but don't consume any
1098 of the string in their match.  So rewriting this way produces what
1099 you'd expect; that is, case 5 will fail, but case 6 succeeds:</p>
1100 <pre>
1101     print &quot;5: got $1\n&quot; if $x =~ /^(\D*)(?=\d)(?!123)/;
1102     print &quot;6: got $1\n&quot; if $y =~ /^(\D*)(?=\d)(?!123)/;</pre>
1103 <pre>
1104     6: got ABC</pre>
1105 <p>In other words, the two zero-width assertions next to each other work as though
1106 they're ANDed together, just as you'd use any built-in assertions:  <code>/^$/</code>
1107 matches only if you're at the beginning of the line AND the end of the
1108 line simultaneously.  The deeper underlying truth is that juxtaposition in
1109 regular expressions always means AND, except when you write an explicit OR
1110 using the vertical bar.  <code>/ab/</code> means match ``a'' AND (then) match ``b'',
1111 although the attempted matches are made at different positions because ``a''
1112 is not a zero-width assertion, but a one-width assertion.</p>
1113 <p><strong>WARNING</strong>: particularly complicated regular expressions can take
1114 exponential time to solve because of the immense number of possible
1115 ways they can use backtracking to try match.  For example, without
1116 internal optimizations done by the regular expression engine, this will
1117 take a painfully long time to run:</p>
1118 <pre>
1119     'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/</pre>
1120 <p>And if you used <code>*</code>'s in the internal groups instead of limiting them
1121 to 0 through 5 matches, then it would take forever--or until you ran
1122 out of stack space.  Moreover, these internal optimizations are not
1123 always applicable.  For example, if you put <code>{0,5}</code> instead of <code>*</code>
1124 on the external group, no current optimization is applicable, and the
1125 match takes a long time to finish.</p>
1126 <p>A powerful tool for optimizing such beasts is what is known as an
1127 ``independent group'',
1128 which does not backtrack (see <code>&lt; (?&gt;pattern) &gt;</code>).  Note also that
1129 zero-length look-ahead/look-behind assertions will not backtrack to make
1130 the tail match, since they are in ``logical'' context: only
1131 whether they match is considered relevant.  For an example
1132 where side-effects of look-ahead <em>might</em> have influenced the
1133 following match, see <code>&lt; (?&gt;pattern) &gt;</code>.</p>
1134 <p>
1135 </p>
1136 <h2><a name="version_8_regular_expressions_x_regular_expression__version_8__x_regex__version_8__x_regexp__version_8_">Version 8 Regular Expressions
1137   </a></h2>
1138 <p>In case you're not familiar with the ``regular'' Version 8 regex
1139 routines, here are the pattern-matching rules not described above.</p>
1140 <p>Any single character matches itself, unless it is a <em>metacharacter</em>
1141 with a special meaning described here or above.  You can cause
1142 characters that normally function as metacharacters to be interpreted
1143 literally by prefixing them with a ``\'' (e.g., ``\.'' matches a ``.'', not any
1144 character; ``\\'' matches a ``\'').  A series of characters matches that
1145 series of characters in the target string, so the pattern <code>blurfl</code>
1146 would match ``blurfl'' in the target string.</p>
1147 <p>You can specify a character class, by enclosing a list of characters
1148 in <code>[]</code>, which will match any one character from the list.  If the
1149 first character after the ``['' is ``^'', the class matches any character not
1150 in the list.  Within a list, the ``-'' character specifies a
1151 range, so that <code>a-z</code> represents all characters between ``a'' and ``z'',
1152 inclusive.  If you want either ``-'' or ``]'' itself to be a member of a
1153 class, put it at the start of the list (possibly after a ``^''), or
1154 escape it with a backslash.  ``-'' is also taken literally when it is
1155 at the end of the list, just before the closing ``]''.  (The
1156 following all specify the same class of three characters: <code>[-az]</code>,
1157 <code>[az-]</code>, and <code>[a\-z]</code>.  All are different from <code>[a-z]</code>, which
1158 specifies a class containing twenty-six characters, even on EBCDIC
1159 based coded character sets.)  Also, if you try to use the character
1160 classes <code>\w</code>, <code>\W</code>, <code>\s</code>, <code>\S</code>, <code>\d</code>, or <code>\D</code> as endpoints of
1161 a range, that's not a range, the ``-'' is understood literally.</p>
1162 <p>Note also that the whole range idea is rather unportable between
1163 character sets--and even within character sets they may cause results
1164 you probably didn't expect.  A sound principle is to use only ranges
1165 that begin from and end at either alphabets of equal case ([a-e],
1166 [A-E]), or digits ([0-9]).  Anything else is unsafe.  If in doubt,
1167 spell out the character sets in full.</p>
1168 <p>Characters may be specified using a metacharacter syntax much like that
1169 used in C: ``\n'' matches a newline, ``\t'' a tab, ``\r'' a carriage return,
1170 ``\f'' a form feed, etc.  More generally, \<em>nnn</em>, where <em>nnn</em> is a string
1171 of octal digits, matches the character whose coded character set value
1172 is <em>nnn</em>.  Similarly, \x<em>nn</em>, where <em>nn</em> are hexadecimal digits,
1173 matches the character whose numeric value is <em>nn</em>. The expression \c<em>x</em>
1174 matches the character control-<em>x</em>.  Finally, the ``.'' metacharacter
1175 matches any character except ``\n'' (unless you use <code>/s</code>).</p>
1176 <p>You can specify a series of alternatives for a pattern using ``|'' to
1177 separate them, so that <code>fee|fie|foe</code> will match any of ``fee'', ``fie'',
1178 or ``foe'' in the target string (as would <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_f"><code>f(e|i|o)e</code></a>).  The
1179 first alternative includes everything from the last pattern delimiter
1180 (``('', ``['', or the beginning of the pattern) up to the first ``|'', and
1181 the last alternative contains everything from the last ``|'' to the next
1182 pattern delimiter.  That's why it's common practice to include
1183 alternatives in parentheses: to minimize confusion about where they
1184 start and end.</p>
1185 <p>Alternatives are tried from left to right, so the first
1186 alternative found for which the entire expression matches, is the one that
1187 is chosen. This means that alternatives are not necessarily greedy. For
1188 example: when matching <code>foo|foot</code> against ``barefoot'', only the ``foo''
1189 part will match, as that is the first alternative tried, and it successfully
1190 matches the target string. (This might not seem important, but it is
1191 important when you are capturing matched text using parentheses.)</p>
1192 <p>Also remember that ``|'' is interpreted as a literal within square brackets,
1193 so if you write <code>[fee|fie|foe]</code> you're really only matching <code>[feio|]</code>.</p>
1194 <p>Within a pattern, you may designate subpatterns for later reference
1195 by enclosing them in parentheses, and you may refer back to the
1196 <em>n</em>th subpattern later in the pattern using the metacharacter
1197 \<em>n</em>.  Subpatterns are numbered based on the left to right order
1198 of their opening parenthesis.  A backreference matches whatever
1199 actually matched the subpattern in the string being examined, not
1200 the rules for that subpattern.  Therefore, <code>(0|0x)\d*\s\1\d*</code> will
1201 match ``0x1234 0x4321'', but not ``0x1234 01234'', because subpattern
1202 1 matched ``0x'', even though the rule <code>0|0x</code> could potentially match
1203 the leading 0 in the second number.</p>
1204 <p>
1205 </p>
1206 <h2><a name="warning_on__1_vs__1">Warning on \1 vs $1</a></h2>
1207 <p>Some people get too used to writing things like:</p>
1208 <pre>
1209     $pattern =~ s/(\W)/\\\1/g;</pre>
1210 <p>This is grandfathered for the RHS of a substitute to avoid shocking the
1211 <strong>sed</strong> addicts, but it's a dirty habit to get into.  That's because in
1212 PerlThink, the righthand side of an <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_s_"><code>s///</code></a> is a double-quoted string.  <code>\1</code> in
1213 the usual double-quoted string means a control-A.  The customary Unix
1214 meaning of <code>\1</code> is kludged in for <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_s_"><code>s///</code></a>.  However, if you get into the habit
1215 of doing that, you get yourself into trouble if you then add an <code>/e</code>
1216 modifier.</p>
1217 <pre>
1218     s/(\d+)/ \1 + 1 /eg;        # causes warning under -w</pre>
1219 <p>Or if you try to do</p>
1220 <pre>
1221     s/(\d+)/\1000/;</pre>
1222 <p>You can't disambiguate that by saying <code>\{1}000</code>, whereas you can fix it with
1223 <code>${1}000</code>.  The operation of interpolation should not be confused
1224 with the operation of matching a backreference.  Certainly they mean two
1225 different things on the <em>left</em> side of the <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_s_"><code>s///</code></a>.</p>
1226 <p>
1227 </p>
1228 <h2><a name="repeated_patterns_matching_zerolength_substring">Repeated patterns matching zero-length substring</a></h2>
1229 <p><strong>WARNING</strong>: Difficult material (and prose) ahead.  This section needs a rewrite.</p>
1230 <p>Regular expressions provide a terse and powerful programming language.  As
1231 with most other power tools, power comes together with the ability
1232 to wreak havoc.</p>
1233 <p>A common abuse of this power stems from the ability to make infinite
1234 loops using regular expressions, with something as innocuous as:</p>
1235 <pre>
1236     'foo' =~ m{ ( o? )* }x;</pre>
1237 <p>The <code>o?</code> can match at the beginning of <code>'foo'</code>, and since the position
1238 in the string is not moved by the match, <code>o?</code> would match again and again
1239 because of the <code>*</code> modifier.  Another common way to create a similar cycle
1240 is with the looping modifier <code>//g</code>:</p>
1241 <pre>
1242     @matches = ( 'foo' =~ m{ o? }xg );</pre>
1243 <p>or</p>
1244 <pre>
1245     print &quot;match: &lt;$&amp;&gt;\n&quot; while 'foo' =~ m{ o? }xg;</pre>
1246 <p>or the loop implied by split().</p>
1247 <p>However, long experience has shown that many programming tasks may
1248 be significantly simplified by using repeated subexpressions that
1249 may match zero-length substrings.  Here's a simple example being:</p>
1250 <pre>
1251     @chars = split //, $string;           # // is not magic in split
1252     ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /</pre>
1253 <p>Thus Perl allows such constructs, by <em>forcefully breaking
1254 the infinite loop</em>.  The rules for this are different for lower-level
1255 loops given by the greedy modifiers <code>*+{}</code>, and for higher-level
1256 ones like the <code>/g</code> modifier or <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_split"><code>split()</code></a> operator.</p>
1257 <p>The lower-level loops are <em>interrupted</em> (that is, the loop is
1258 broken) when Perl detects that a repeated expression matched a
1259 zero-length substring.   Thus</p>
1260 <pre>
1261    m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x;</pre>
1262 <p>is made equivalent to</p>
1263 <pre>
1264    m{   (?: NON_ZERO_LENGTH )*
1265       |
1266         (?: ZERO_LENGTH )?
1267     }x;</pre>
1268 <p>The higher level-loops preserve an additional state between iterations:
1269 whether the last match was zero-length.  To break the loop, the following
1270 match after a zero-length match is prohibited to have a length of zero.
1271 This prohibition interacts with backtracking (see <a href="#backtracking">Backtracking</a>),
1272 and so the <em>second best</em> match is chosen if the <em>best</em> match is of
1273 zero length.</p>
1274 <p>For example:</p>
1275 <pre>
1276     $_ = 'bar';
1277     s/\w??/&lt;$&amp;&gt;/g;</pre>
1278 <p>results in <code>&lt;&gt;&lt;b&gt;&lt;&gt;&lt;a&gt;&lt;&gt;&lt;r&gt;&lt;&gt;</code>.  At each position of the string the best
1279 match given by non-greedy <code>??</code> is the zero-length match, and the <em>second
1280 best</em> match is what is matched by <code>\w</code>.  Thus zero-length matches
1281 alternate with one-character-long matches.</p>
1282 <p>Similarly, for repeated <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_m_"><code>m/()/g</code></a> the second-best match is the match at the
1283 position one notch further in the string.</p>
1284 <p>The additional state of being <em>matched with zero-length</em> is associated with
1285 the matched string, and is reset by each assignment to pos().
1286 Zero-length matches at the end of the previous match are ignored
1287 during <a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_split"><code>split</code></a>.</p>
1288 <p>
1289 </p>
1290 <h2><a name="combining_pieces_together">Combining pieces together</a></h2>
1291 <p>Each of the elementary pieces of regular expressions which were described
1292 before (such as <code>ab</code> or <code>\Z</code>) could match at most one substring
1293 at the given position of the input string.  However, in a typical regular
1294 expression these elementary pieces are combined into more complicated
1295 patterns using combining operators <a href="#item_st"><code>ST</code></a>, <a href="#item_s_7ct"><code>S|T</code></a>, <code>S*</code> etc
1296 (in these examples <a href="#item_s"><code>S</code></a> and <code>T</code> are regular subexpressions).</p>
1297 <p>Such combinations can include alternatives, leading to a problem of choice:
1298 if we match a regular expression <code>a|ab</code> against <code>&quot;abc&quot;</code>, will it match
1299 substring <code>&quot;a&quot;</code> or <code>&quot;ab&quot;</code>?  One way to describe which substring is
1300 actually matched is the concept of backtracking (see <a href="#backtracking">Backtracking</a>).
1301 However, this description is too low-level and makes you think
1302 in terms of a particular implementation.</p>
1303 <p>Another description starts with notions of ``better''/``worse''.  All the
1304 substrings which may be matched by the given regular expression can be
1305 sorted from the ``best'' match to the ``worst'' match, and it is the ``best''
1306 match which is chosen.  This substitutes the question of ``what is chosen?''
1307 by the question of ``which matches are better, and which are worse?''.</p>
1308 <p>Again, for elementary pieces there is no such question, since at most
1309 one match at a given position is possible.  This section describes the
1310 notion of better/worse for combining operators.  In the description
1311 below <a href="#item_s"><code>S</code></a> and <code>T</code> are regular subexpressions.</p>
1312 <dl>
1313 <dt><strong><a name="item_st"><code>ST</code></a></strong>
1314
1315 <dd>
1316 <p>Consider two possible matches, <code>AB</code> and <code>A'B'</code>, <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_a"><code>A</code></a> and <code>A'</code> are
1317 substrings which can be matched by <a href="#item_s"><code>S</code></a>, <code>B</code> and <code>B'</code> are substrings
1318 which can be matched by <code>T</code>.</p>
1319 </dd>
1320 <dd>
1321 <p>If <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_a"><code>A</code></a> is better match for <a href="#item_s"><code>S</code></a> than <code>A'</code>, <code>AB</code> is a better
1322 match than <code>A'B'</code>.</p>
1323 </dd>
1324 <dd>
1325 <p>If <a href="file://C|\msysgit\mingw\html/pod/perlguts.html#item_a"><code>A</code></a> and <code>A'</code> coincide: <code>AB</code> is a better match than <code>AB'</code> if
1326 <code>B</code> is better match for <code>T</code> than <code>B'</code>.</p>
1327 </dd>
1328 </li>
1329 <dt><strong><a name="item_s_7ct"><code>S|T</code></a></strong>
1330
1331 <dd>
1332 <p>When <a href="#item_s"><code>S</code></a> can match, it is a better match than when only <code>T</code> can match.</p>
1333 </dd>
1334 <dd>
1335 <p>Ordering of two matches for <a href="#item_s"><code>S</code></a> is the same as for <a href="#item_s"><code>S</code></a>.  Similar for
1336 two matches for <code>T</code>.</p>
1337 </dd>
1338 </li>
1339 <dt><strong><a name="item_s"><code>S{REPEAT_COUNT}</code></a></strong>
1340
1341 <dd>
1342 <p>Matches as <code>SSS...S</code> (repeated as many times as necessary).</p>
1343 </dd>
1344 </li>
1345 <dt><strong><code>S{min,max}</code></strong>
1346
1347 <dd>
1348 <p>Matches as <a href="#item_s"><code>S{max}|S{max-1}|...|S{min+1}|S{min}</code></a>.</p>
1349 </dd>
1350 </li>
1351 <dt><strong><a name="item_s_7bmin_2cmax_7d_3f"><code>S{min,max}?</code></a></strong>
1352
1353 <dd>
1354 <p>Matches as <a href="#item_s"><code>S{min}|S{min+1}|...|S{max-1}|S{max}</code></a>.</p>
1355 </dd>
1356 </li>
1357 <dt><strong><a name="item_s_3f_2c_s_2a_2c_s_2b"><code>S?</code>, <code>S*</code>, <code>S+</code></a></strong>
1358
1359 <dd>
1360 <p>Same as <a href="#item_s"><code>S{0,1}</code></a>, <a href="#item_s"><code>S{0,BIG_NUMBER}</code></a>, <a href="#item_s"><code>S{1,BIG_NUMBER}</code></a> respectively.</p>
1361 </dd>
1362 </li>
1363 <dt><strong><a name="item_s_3f_3f_2c_s_2a_3f_2c_s_2b_3f"><code>S??</code>, <code>S*?</code>, <code>S+?</code></a></strong>
1364
1365 <dd>
1366 <p>Same as <code>S{0,1}?</code>, <code>S{0,BIG_NUMBER}?</code>, <code>S{1,BIG_NUMBER}?</code> respectively.</p>
1367 </dd>
1368 </li>
1369 <dt><strong><a name="item__28_3f_3es_29"><code>(?&gt;S)</code></a></strong>
1370
1371 <dd>
1372 <p>Matches the best match for <a href="#item_s"><code>S</code></a> and only that.</p>
1373 </dd>
1374 </li>
1375 <dt><strong><a name="item__28_3f_3ds_29_2c__28_3f_3c_3ds_29"><code>(?=S)</code>, <code>(?&lt;=S)</code></a></strong>
1376
1377 <dd>
1378 <p>Only the best match for <a href="#item_s"><code>S</code></a> is considered.  (This is important only if
1379 <a href="#item_s"><code>S</code></a> has capturing parentheses, and backreferences are used somewhere
1380 else in the whole regular expression.)</p>
1381 </dd>
1382 </li>
1383 <dt><strong><a name="item__28_3f_21s_29_2c__28_3f_3c_21s_29"><code>(?!S)</code>, <code>(?&lt;!S)</code></a></strong>
1384
1385 <dd>
1386 <p>For this grouping operator there is no need to describe the ordering, since
1387 only whether or not <a href="#item_s"><code>S</code></a> can match is important.</p>
1388 </dd>
1389 </li>
1390 <dt><strong><a name="item__28_3f_3f_7b_expr__7d_29"><code>(??{ EXPR })</code></a></strong>
1391
1392 <dd>
1393 <p>The ordering is the same as for the regular expression which is
1394 the result of EXPR.</p>
1395 </dd>
1396 </li>
1397 <dt><strong><code>(?(condition)yes-pattern|no-pattern)</code></strong>
1398
1399 <dd>
1400 <p>Recall that which of <code>yes-pattern</code> or <code>no-pattern</code> actually matches is
1401 already determined.  The ordering of the matches is the same as for the
1402 chosen subexpression.</p>
1403 </dd>
1404 </li>
1405 </dl>
1406 <p>The above recipes describe the ordering of matches <em>at a given position</em>.
1407 One more rule is needed to understand how a match is determined for the
1408 whole regular expression: a match at an earlier position is always better
1409 than a match at a later position.</p>
1410 <p>
1411 </p>
1412 <h2><a name="creating_custom_re_engines">Creating custom RE engines</a></h2>
1413 <p>Overloaded constants (see <a href="file://C|\msysgit\mingw\html/lib/overload.html">the overload manpage</a>) provide a simple way to extend
1414 the functionality of the RE engine.</p>
1415 <p>Suppose that we want to enable a new RE escape-sequence <code>\Y|</code> which
1416 matches at boundary between whitespace characters and non-whitespace
1417 characters.  Note that <code>(?=\S)(?&lt;!\S)|(?!\S)(?&lt;=\S)</code> matches exactly
1418 at these positions, so we want to have each <code>\Y|</code> in the place of the
1419 more complicated version.  We can create a module <code>customre</code> to do
1420 this:</p>
1421 <pre>
1422     package customre;
1423     use overload;</pre>
1424 <pre>
1425     sub import {
1426       shift;
1427       die &quot;No argument to customre::import allowed&quot; if @_;
1428       overload::constant 'qr' =&gt; \&amp;convert;
1429     }</pre>
1430 <pre>
1431     sub invalid { die &quot;/$_[0]/: invalid escape '\\$_[1]'&quot;}</pre>
1432 <pre>
1433     # We must also take care of not escaping the legitimate \\Y|
1434     # sequence, hence the presence of '\\' in the conversion rules.
1435     my %rules = ( '\\' =&gt; '\\\\',
1436                   'Y|' =&gt; qr/(?=\S)(?&lt;!\S)|(?!\S)(?&lt;=\S)/ );
1437     sub convert {
1438       my $re = shift;
1439       $re =~ s{
1440                 \\ ( \\ | Y . )
1441               }
1442               { $rules{$1} or invalid($re,$1) }sgex;
1443       return $re;
1444     }</pre>
1445 <p>Now <code>use customre</code> enables the new escape in constant regular
1446 expressions, i.e., those without any runtime variable interpolations.
1447 As documented in <a href="file://C|\msysgit\mingw\html/lib/overload.html">the overload manpage</a>, this conversion will work only over
1448 literal parts of regular expressions.  For <code>\Y|$re\Y|</code> the variable
1449 part of this regular expression needs to be converted explicitly
1450 (but only if the special meaning of <code>\Y|</code> should be enabled inside $re):</p>
1451 <pre>
1452     use customre;
1453     $re = &lt;&gt;;
1454     chomp $re;
1455     $re = customre::convert $re;
1456     /\Y|$re\Y|/;</pre>
1457 <p>
1458 </p>
1459 <hr />
1460 <h1><a name="bugs">BUGS</a></h1>
1461 <p>This document varies from difficult to understand to completely
1462 and utterly opaque.  The wandering prose riddled with jargon is
1463 hard to fathom in several places.</p>
1464 <p>This document needs a rewrite that separates the tutorial content
1465 from the reference content.</p>
1466 <p>
1467 </p>
1468 <hr />
1469 <h1><a name="see_also">SEE ALSO</a></h1>
1470 <p><a href="file://C|\msysgit\mingw\html/pod/perlrequick.html">the perlrequick manpage</a>.</p>
1471 <p><a href="file://C|\msysgit\mingw\html/pod/perlretut.html">the perlretut manpage</a>.</p>
1472 <p><a href="file://C|\msysgit\mingw\html/pod/perlop.html#regexp_quotelike_operators">Regexp Quote-Like Operators in the perlop manpage</a>.</p>
1473 <p><a href="file://C|\msysgit\mingw\html/pod/perlop.html#gory_details_of_parsing_quoted_constructs">Gory details of parsing quoted constructs in the perlop manpage</a>.</p>
1474 <p><a href="file://C|\msysgit\mingw\html/pod/perlfaq6.html">the perlfaq6 manpage</a>.</p>
1475 <p><a href="file://C|\msysgit\mingw\html/pod/perlfunc.html#item_pos">pos in the perlfunc manpage</a>.</p>
1476 <p><a href="file://C|\msysgit\mingw\html/pod/perllocale.html">the perllocale manpage</a>.</p>
1477 <p><a href="file://C|\msysgit\mingw\html/pod/perlebcdic.html">the perlebcdic manpage</a>.</p>
1478 <p><em>Mastering Regular Expressions</em> by Jeffrey Friedl, published
1479 by O'Reilly and Associates.</p>
1480 <table border="0" width="100%" cellspacing="0" cellpadding="3">
1481 <tr><td class="block" style="background-color: #cccccc" valign="middle">
1482 <big><strong><span class="block">&nbsp;perlre - Perl regular expressions</span></strong></big>
1483 </td></tr>
1484 </table>
1485
1486 </body>
1487
1488 </html>