path: root/lib/stdlib/doc/src/re.xml

<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE erlref SYSTEM "erlref.dtd">

<erlref>
  <header>
    <copyright>
      <year>2007</year>
      <year>2016</year>
      <holder>Ericsson AB, All Rights Reserved</holder>
    </copyright>
    <legalnotice>
  Licensed under the Apache License, Version 2.0 (the "License");
  you may not use this file except in compliance with the License.
  You may obtain a copy of the License at
 
      http://www.apache.org/licenses/LICENSE-2.0

  Unless required by applicable law or agreed to in writing, software
  distributed under the License is distributed on an "AS IS" BASIS,
  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  See the License for the specific language governing permissions and
  limitations under the License.

  The Initial Developer of the Original Code is Ericsson AB.
    </legalnotice>

    <title>re</title>
    <prepared>Patrik Nyblom</prepared>
    <responsible>Kenneth Lundin</responsible>
    <docno>1</docno>
    <approved></approved>
    <checked></checked>
    <date>2008-05-27</date>
    <rev>A</rev>
    <file>re.xml</file>
  </header>
  <module>re</module>
  <modulesummary>Perl-like regular expressions for Erlang.</modulesummary>
  <description>
    <p>This module contains regular expression matching functions for
      strings and binaries.</p>

    <p>The <seealso marker="#regexp_syntax">regular expression</seealso>
      syntax and semantics resemble that of Perl.</p>

    <p>The matching algorithms of the library are based on the
      PCRE library, but not all of the PCRE library is interfaced and
      some parts of the library go beyond what PCRE offers. The sections of
      the PCRE documentation that are relevant to this module are included
      here.</p>

    <note>
      <p>The Erlang literal syntax for strings uses the &quot;\&quot;
       (backslash) character as an escape code. You need to escape
       backslashes in literal strings, both in your code and in the shell,
       with an extra backslash, that is, &quot;\\&quot;.</p>
    </note>
  </description>

  <datatypes>
    <datatype>
      <name name="mp"/>
      <desc>
        <p>Opaque data type containing a compiled regular expression.
          <c>mp()</c> is guaranteed to be a tuple() having the atom
          <c>re_pattern</c> as its first element, to allow for matching in
          guards. The arity of the tuple or the content of the other fields
          can change in future Erlang/OTP releases.</p>
      </desc>
    </datatype>
    <datatype>
      <name name="nl_spec"/>
    </datatype>
    <datatype>
      <name name="compile_option"/>
    </datatype>
  </datatypes>

  <funcs>
    <func>
      <name name="compile" arity="1"/>
      <fsummary>Compile a regular expression into a match program</fsummary>
      <desc>
      <p>The same as <c>compile(<anno>Regexp</anno>,[])</c></p>
      </desc>
    </func> 

    <func>
      <name name="compile" arity="2"/>
      <fsummary>Compile a regular expression into a match program.</fsummary>
      <desc>
        <p>Compiles a regular expression, with the syntax
          described below, into an internal format to be used later as a
          parameter to
          <seealso marker="#run/2"><c>run/2</c></seealso> and
          <seealso marker="#run/3"><c>run/3</c></seealso>.</p>
        <p>Compiling the regular expression before matching is useful if
          the same expression is to be used in matching against multiple
          subjects during the lifetime of the program. Compiling once and
          executing many times is far more efficient than compiling each
          time one wants to match.</p>
        <p>When option <c>unicode</c> is specified, the regular expression
          is to be specified as a valid Unicode <c>charlist()</c>, otherwise as
          any valid <c>iodata()</c>.</p>
          <marker id="compile_options"/>
        <p>Options:</p>
        <taglist>
          <tag><c>unicode</c></tag>
          <item>
            <p>The regular expression is specified as a Unicode
              <c>charlist()</c> and the resulting regular expression code is to
              be run against a valid Unicode <c>charlist()</c> subject. Also
              consider option <c>ucp</c> when using Unicode characters.</p>
          </item>
          <tag><c>anchored</c></tag>
          <item>
            <p>The pattern is forced to be "anchored", that is, it is
              constrained to match only at the first matching point in the
              string that is searched (the "subject string"). This effect can
              also be achieved by appropriate constructs in the pattern
              itself.</p>
          </item>
          <tag><c>caseless</c></tag>
          <item>
            <p>Letters in the pattern match both uppercase and lowercase
              letters. It is equivalent to Perl option <c>/i</c> and can be
              changed within a pattern by a <c>(?i)</c> option setting.
              Uppercase and lowercase letters are defined as in the ISO 8859-1
              character set.</p>
          </item>
          <tag><c>dollar_endonly</c></tag>
          <item>
            <p>A dollar metacharacter in the pattern matches only at the end of
              the subject string. Without this option, a dollar also matches
              immediately before a newline at the end of the string (but not
              before any other newlines). This option is ignored if option
              <c>multiline</c> is specified. There is no equivalent option in
              Perl, and it cannot be set within a pattern.</p>
          </item>
          <tag><c>dotall</c></tag>
          <item>
            <p>A dot in the pattern matches all characters, including those
              indicating newline. Without it, a dot does not match when the
              current position is at a newline. This option is equivalent to
              Perl option <c>/s</c> and it can be changed within a pattern by a
              <c>(?s)</c> option setting. A negative class, such as <c>[^a]</c>,
              always matches newline characters, independent of the setting of
              this option.</p>
          </item>
          <tag><c>extended</c></tag>
          <item>
            <p>Whitespace data characters in the pattern are ignored except
              when escaped or inside a character class. Whitespace does not
              include character 'vt' (ASCII 11). Characters between an
              unescaped <c>#</c> outside a character class and the next newline,
              inclusive, are also ignored. This is equivalent to Perl option
              <c>/x</c> and can be changed within a pattern by a <c>(?x)</c>
              option setting.</p>
            <p>With this option, comments inside complicated patterns can be
              included. However, notice that this applies only to data
              characters. Whitespace characters can never appear within special
              character sequences in a pattern, for example within sequence
              <c>(?(</c> that introduces a conditional subpattern.</p>
          </item>
          <tag><c>firstline</c></tag>
          <item>
            <p>An unanchored pattern is required to match before or at the first
              newline in the subject string, although the matched text can
              continue over the newline.</p>
          </item>
          <tag><c>multiline</c></tag>
          <item>
            <p>By default, PCRE treats the subject string as consisting of a
              single line of characters (even if it contains newlines). The
              "start of line" metacharacter (<c>^</c>) matches only at the
              start of the string, while the "end of line" metacharacter
              (<c>$</c>) matches only at the end of the string, or before a
              terminating newline (unless option <c>dollar_endonly</c> is
              specified). This is the same as in Perl.</p>
            <p>When this option is specified, the "start of line" and "end of
              line" constructs match immediately following or immediately
              before internal newlines in the subject string, respectively, as
              well as at the very start and end. This is equivalent to Perl
              option <c>/m</c> and can be changed within a pattern by a
              <c>(?m)</c> option setting. If there are no newlines in a subject
              string, or no occurrences of <c>^</c> or <c>$</c> in a pattern,
              setting <c>multiline</c> has no effect.</p> </item>
          <tag><c>no_auto_capture</c></tag>
          <item>
            <p>Disables the use of numbered capturing parentheses in the
              pattern. Any opening parenthesis that is not followed by <c>?</c>
              behaves as if it is followed by <c>?:</c>. Named parentheses can
              still be used for capturing (and they acquire numbers in the
              usual way). There is no equivalent option in Perl.</p>
          </item>
          <tag><c>dupnames</c></tag>
          <item>
            <p>Names used to identify capturing subpatterns need not be unique.
              This can be helpful for certain types of pattern when it is known
              that only one instance of the named subpattern can ever be
              matched. More details of named subpatterns are provided below.</p>
          </item>
          <tag><c>ungreedy</c></tag>
          <item>
            <p>Inverts the "greediness" of the quantifiers so that they are not
              greedy by default, but become greedy if followed by "?". It is
              not compatible with Perl. It can also be set by a <c>(?U)</c>
              option setting within the pattern.</p>
          </item>
          <tag><c>{newline, NLSpec}</c></tag>
          <item>
            <p>Overrides the default definition of a newline in the subject
              string, which is LF (ASCII 10) in Erlang.</p>
            <taglist>
              <tag><c>cr</c></tag>
              <item>
                <p>Newline is indicated by a single character <c>cr</c>
                  (ASCII 13).</p>
              </item>
              <tag><c>lf</c></tag>
              <item>
                <p>Newline is indicated by a single character LF (ASCII 10), the
                  default.</p>
              </item>
              <tag><c>crlf</c></tag>
              <item>
                <p>Newline is indicated by the two-character CRLF (ASCII 13
                  followed by ASCII 10) sequence.</p>
              </item>
              <tag><c>anycrlf</c></tag>
              <item>
                <p>Any of the three preceding sequences is to be recognized.</p>
              </item>
              <tag><c>any</c></tag>
              <item>
                <p>Any of the newline sequences above, and the Unicode sequences
                  VT (vertical tab, U+000B), FF (formfeed, U+000C), NEL (next
                  line, U+0085), LS (line separator, U+2028), and PS (paragraph
                  separator, U+2029).</p>
              </item>
            </taglist>
          </item>
          <tag><c>bsr_anycrlf</c></tag>
          <item>
            <p>Specifies specifically that \R is to match only the CR,
              LF, or CRLF sequences, not the Unicode-specific newline
              characters.</p>
          </item>
          <tag><c>bsr_unicode</c></tag>
          <item>
            <p>Specifies specifically that \R is to match all the Unicode
              newline characters (including CRLF, and so on, the default).</p>
          </item>
          <tag><c>no_start_optimize</c></tag>
          <item>
            <p>Disables optimization that can malfunction if "Special
              start-of-pattern items" are present in the regular expression. A
              typical example would be when matching "DEFABC" against
              "(*COMMIT)ABC", where the start optimization of PCRE would skip
              the subject up to "A" and never realize that the (*COMMIT)
              instruction is to have made the matching fail. This option is only
              relevant if you use "start-of-pattern items", as discussed in
              section <seealso marker="#regexp_syntax_details">PCRE Regular Expression
              Details</seealso>.</p>
          </item>
          <tag><c>ucp</c></tag>
          <item>
            <p>Specifies that Unicode character properties are to be used when
              resolving \B, \b, \D, \d, \S, \s, \W and \w. Without this flag,
              only ISO Latin-1 properties are used. Using Unicode properties
              hurts performance, but is semantically correct when working with
              Unicode characters beyond the ISO Latin-1 range.</p>
          </item>
          <tag><c>never_utf</c></tag>
          <item>
            <p>Specifies that the (*UTF) and/or (*UTF8) "start-of-pattern
              items" are forbidden. This flag cannot be combined with option
              <c>unicode</c>. Useful if ISO Latin-1 patterns from an external
              source are to be compiled.</p>
          </item>
        </taglist>
      </desc>
    </func> 

    <func>
      <name name="inspect" arity="2"/>
      <fsummary>Inspects a compiled regular expression.</fsummary>
      <desc>
        <p>Takes a compiled regular expression and an item, and returns the
          relevant data from the regular expression. The only
          supported item is <c>namelist</c>, which returns the tuple
          <c>{namelist, [binary()]}</c>, containing the names of all (unique)
          named subpatterns in the regular expression. For example:</p>
        <code type="none">
1&gt; {ok,MP} = re:compile("(?&lt;A&gt;A)|(?&lt;B&gt;B)|(?&lt;C&gt;C)").
{ok,{re_pattern,3,0,0,
                &lt;&lt;69,82,67,80,119,0,0,0,0,0,0,0,1,0,0,0,255,255,255,255,
                  255,255,...&gt;&gt;}}
2&gt; re:inspect(MP,namelist).
{namelist,[&lt;&lt;"A"&gt;&gt;,&lt;&lt;"B"&gt;&gt;,&lt;&lt;"C"&gt;&gt;]}
3&gt; {ok,MPD} = re:compile("(?&lt;C&gt;A)|(?&lt;B&gt;B)|(?&lt;C&gt;C)",[dupnames]).
{ok,{re_pattern,3,0,0,
                &lt;&lt;69,82,67,80,119,0,0,0,0,0,8,0,1,0,0,0,255,255,255,255,
                  255,255,...&gt;&gt;}}
4&gt; re:inspect(MPD,namelist).                                   
{namelist,[&lt;&lt;"B"&gt;&gt;,&lt;&lt;"C"&gt;&gt;]}</code>
        <p>Notice in the second example that the duplicate name only occurs
          once in the returned list, and that the list is in alphabetical order
          regardless of where the names are positioned in the regular
          expression. The order of the names is the same as the order of
          captured subexpressions if <c>{capture, all_names}</c> is specified as
          an option to <seealso marker="#run/3"><c>run/3</c></seealso>.
          You can therefore create a name-to-value mapping from the result of
          <c>run/3</c> like this:</p>
        <code>
1&gt; {ok,MP} = re:compile("(?&lt;A&gt;A)|(?&lt;B&gt;B)|(?&lt;C&gt;C)").
{ok,{re_pattern,3,0,0,
                &lt;&lt;69,82,67,80,119,0,0,0,0,0,0,0,1,0,0,0,255,255,255,255,
                  255,255,...&gt;&gt;}}
2&gt; {namelist, N} = re:inspect(MP,namelist).
{namelist,[&lt;&lt;"A"&gt;&gt;,&lt;&lt;"B"&gt;&gt;,&lt;&lt;"C"&gt;&gt;]}
3&gt; {match,L} = re:run("AA",MP,[{capture,all_names,binary}]).
{match,[&lt;&lt;"A"&gt;&gt;,&lt;&lt;&gt;&gt;,&lt;&lt;&gt;&gt;]}
4&gt; NameMap = lists:zip(N,L).
[{&lt;&lt;"A"&gt;&gt;,&lt;&lt;"A"&gt;&gt;},{&lt;&lt;"B"&gt;&gt;,&lt;&lt;&gt;&gt;},{&lt;&lt;"C"&gt;&gt;,&lt;&lt;&gt;&gt;}]</code>
      </desc>
    </func>

    <func>
      <name name="replace" arity="3"/>
      <fsummary>Match a subject against regular expression and replace matching
        elements with Replacement.</fsummary>
      <desc>
        <p>Same as <c>replace(<anno>Subject</anno>, <anno>RE</anno>,
          <anno>Replacement</anno>, [])</c>.</p>
      </desc>
    </func> 

    <func>
      <name name="replace" arity="4"/>
      <fsummary>Match a subject against regular expression and replace matching
        elements with Replacement.</fsummary>
      <desc>
        <p>Replaces the matched part of the <c><anno>Subject</anno></c> string
          with the contents of <c><anno>Replacement</anno></c>.</p>
        <p>The permissible options are the same as for
          <seealso marker="#run/3"><c>run/3</c></seealso>, except that option<c>
          capture</c> is not allowed. Instead a <c>{return,
          <anno>ReturnType</anno>}</c> is present. The default return type is
          <c>iodata</c>, constructed in a way to minimize copying. The
          <c>iodata</c> result can be used directly in many I/O operations. If a
          flat <c>list()</c> is desired, specify <c>{return, list}</c>. If a
          binary is desired, specify <c>{return, binary}</c>.</p>
        <p>As in function <c>run/3</c>, an <c>mp()</c> compiled with option
          <c>unicode</c> requires <c><anno>Subject</anno></c> to be a Unicode
          <c>charlist()</c>. If compilation is done implicitly and the
          <c>unicode</c> compilation option is specified to this function, both
          the regular expression and <c><anno>Subject</anno></c> are to
          specified as valid Unicode <c>charlist()</c>s.</p>
        <p>The replacement string can contain the special character
          <c>&amp;</c>, which inserts the whole matching expression in the
          result, and the special sequence <c>\</c>N (where N is an integer &gt;
          0), <c>\g</c>N, or <c>\g{</c>N<c>}</c>, resulting in the subexpression
          number N, is inserted in the result. If no subexpression with that
          number is generated by the regular expression, nothing is
          inserted.</p>
        <p>To insert an &amp; or a \ in the result, precede it
          with a \. Notice that Erlang already gives a special meaning to
          \ in literal strings, so a single \ must be written as
          <c>"\\"</c> and therefore a double \ as <c>"\\\\"</c>.</p>
        <p><em>Example:</em></p>
        <code>
re:replace("abcd","c","[&amp;]",[{return,list}]).</code>
        <p>gives</p>
        <code>
"ab[c]d"</code>
        <p>while</p>
        <code>
re:replace("abcd","c","[\\&amp;]",[{return,list}]).</code>
        <p>gives</p>
        <code>
"ab[&amp;]d"</code>
        <p>As with <c>run/3</c>, compilation errors raise the <c>badarg</c>
          exception. <seealso marker="#compile/2"><c>compile/2</c></seealso>
          can be used to get more information about the error.</p>
      </desc>
    </func>

    <func>
      <name name="run" arity="2"/>
      <fsummary>Match a subject against regular expression and capture
        subpatterns.</fsummary>
      <desc>
        <p>Same as <c>run(<anno>Subject</anno>,<anno>RE</anno>,[])</c>.</p>
      </desc>
    </func> 

    <func>
      <name name="run" arity="3"/>
      <fsummary>Match a subject against regular expression and capture
        subpatterns.</fsummary>
      <type_desc variable="CompileOpt">See <seealso marker="#compile_options">
        <c>compile/2</c></seealso>.</type_desc>
      <desc>
        <p>Executes a regular expression matching, and returns
          <c>match/{match, <anno>Captured</anno>}</c> or <c>nomatch</c>. The
          regular expression can be specified either as <c>iodata()</c> in
          which case it is automatically compiled (as by <c>compile/2</c>) and
          executed, or as a precompiled <c>mp()</c> in which case it is executed
          against the subject directly.</p>
        <p>When compilation is involved, exception <c>badarg</c> is thrown if a
          compilation error occurs. Call <c>compile/2</c> to get information
          about the location of the error in the regular expression.</p>
        <p>If the regular expression is previously compiled, the option list can
          only contain the following options:</p>
        <list type="bulleted">
          <item><c>anchored</c></item>
          <item><c>{capture, <anno>ValueSpec</anno>}/{capture,
            <anno>ValueSpec</anno>, <anno>Type</anno>}</c></item>
          <item><c>global</c></item>
          <item><c>{match_limit, integer() >= 0}</c></item>
          <item><c>{match_limit_recursion, integer() >= 0}</c></item>
          <item><c>{newline, <anno>NLSpec</anno>}</c></item>
          <item><c>notbol</c></item>
          <item><c>notempty</c></item>
          <item><c>notempty_atstart</c></item>
          <item><c>noteol</c></item>
          <item><c>{offset, integer() >= 0}</c></item>
          <item><c>report_errors</c></item>
        </list>
        <p>Otherwise all options valid for function <c>compile/2</c> are also
          allowed. Options allowed both for compilation and execution of a
          match, namely <c>anchored</c> and <c>{newline,
          <anno>NLSpec</anno>}</c>, affect both the compilation and execution if
          present together with a non-precompiled regular expression.</p>
        <p>If the regular expression was previously compiled with option
          <c>unicode</c>, <c><anno>Subject</anno></c> is to be provided as a
          valid Unicode <c>charlist()</c>, otherwise any <c>iodata()</c> will
          do. If compilation is involved and option <c>unicode</c> is specified,
          both <c><anno>Subject</anno></c> and the regular expression are to be
          specified as valid Unicode <c>charlists()</c>.</p>
        <p><c>{capture, <anno>ValueSpec</anno>}/{capture,
          <anno>ValueSpec</anno>, <anno>Type</anno>}</c> defines what to return
          from the function upon successful matching. The <c>capture</c> tuple
          can contain both a value specification, telling which of the captured
          substrings are to be returned, and a type specification, telling how
          captured substrings are to be returned (as index tuples, lists, or
          binaries). The options are described in detail below.</p>
        <p>If the capture options describe that no substring capturing is to be
          done (<c>{capture, none}</c>), the function returns the single atom
          <c>match</c> upon successful matching, otherwise the tuple
          <c>{match, <anno>ValueList</anno>}</c>. Disabling capturing can be
          done either by specifying <c>none</c> or an empty list as
          <c><anno>ValueSpec</anno></c>.</p>
        <p>Option <c>report_errors</c> adds the possibility that an error tuple
          is returned. The tuple either indicates a matching error
          (<c>match_limit</c> or <c>match_limit_recursion</c>), or a compilation
          error, where the error tuple has the format <c>{error, {compile,
          <anno>CompileErr</anno>}}</c>. Notice that if option
          <c>report_errors</c> is not specified, the function never returns
          error tuples, but reports compilation errors as a <c>badarg</c>
          exception and failed matches because of exceeded match limits simply
          as <c>nomatch</c>.</p>
        <p>The following options are relevant for execution:</p>
        <taglist>
          <tag><c>anchored</c></tag>
          <item>
            <p>Limits <c>run/3</c> to matching at the first matching
              position. If a pattern was compiled with <c>anchored</c>, or
              turned out to be anchored by virtue of its contents, it cannot
              be made unanchored at matching time, hence there is no
              <c>unanchored</c> option.</p></item>
          <tag><c>global</c></tag>
          <item>
            <p>Implements global (repetitive) search (flag <c>g</c> in Perl).
              Each match is returned as a separate <c>list()</c> containing the
              specific match and any matching subexpressions (or as specified
              by option <c>capture</c>. The <c><anno>Captured</anno></c> part
              of the return value is hence a <c>list()</c> of <c>list()</c>s
              when this option is specified.</p>
            <p>The interaction of option <c>global</c> with a regular
              expression that matches an empty string surprises some users.
              When option <c>global</c> is specified, <c>run/3</c> handles
              empty matches in the same way as Perl: a zero-length match at any
              point is also retried with options <c>[anchored,
              notempty_atstart]</c>. If that search gives a result of length
              &gt; 0, the result is included. Example:</p>
            <code>
re:run("cat","(|at)",[global]).</code>
            <p>The following matchings are performed:</p>
            <taglist>
              <tag>At offset <c>0</c></tag>
              <item>
                <p>The regular expression <c>(|at)</c> first match at the
                  initial position of string <c>cat</c>, giving the result set
                  <c>[{0,0},{0,0}]</c> (the second <c>{0,0}</c> is because of
                  the subexpression marked by the parentheses). As the length
                  of the match is 0, we do not advance to the next position
                  yet.</p>
              </item>
              <tag>At offset <c>0</c> with <c>[anchored,
                notempty_atstart]</c></tag>
              <item>
                <p>The search is retried with options <c>[anchored,
                  notempty_atstart]</c> at the same position, which does not
                  give any interesting result of longer length, so the search
                  position is advanced to the next character (<c>a</c>).</p>
              </item>
              <tag>At offset <c>1</c></tag>
              <item>
                <p>The search results in <c>[{1,0},{1,0}]</c>, so this search is
                  also repeated with the extra options.</p>
              </item>
              <tag>At offset <c>1</c> with <c>[anchored,
                notempty_atstart]</c></tag>
              <item>
                <p>Alternative <c>ab</c> is found and the result is
                  [{1,2},{1,2}]. The result is added to the list of results and
                  the position in the search string is advanced two steps.</p>
              </item>
              <tag>At offset <c>3</c></tag>
              <item>
                <p>The search once again matches the empty string, giving
                  <c>[{3,0},{3,0}]</c>.</p>
              </item>
              <tag>At offset <c>1</c> with <c>[anchored,
                notempty_atstart]</c></tag>
              <item>
                <p>This gives no result of length &gt; 0 and we are at the last
                  position, so the global search is complete.</p>
              </item>
            </taglist>
            <p>The result of the call is:</p>
            <code>
{match,[[{0,0},{0,0}],[{1,0},{1,0}],[{1,2},{1,2}],[{3,0},{3,0}]]}</code>
          </item>
          <tag><c>notempty</c></tag>
          <item>
            <p>An empty string is not considered to be a valid match if this
              option is specified. If alternatives in the pattern exist, they
              are tried. If all the alternatives match the empty string, the
              entire match fails.</p>
            <p><em>Example:</em></p>
            <p>If the following pattern is applied to a string not beginning
              with "a" or "b", it would normally match the empty string at the
              start of the subject:</p>
            <code>
a?b?</code>
            <p>With option <c>notempty</c>, this match is invalid, so
              <c>run/3</c> searches further into the string for occurrences of
              "a" or "b".</p>
          </item>
          <tag><c>notempty_atstart</c></tag>
          <item>
            <p>Like <c>notempty</c>, except that an empty string match that is
              not at the start of the subject is permitted. If the pattern is
              anchored, such a match can occur only if the pattern contains
              \K.</p>
            <p>Perl has no direct equivalent of <c>notempty</c> or
              <c>notempty_atstart</c>, but it does make a special case of a
              pattern match of the empty string within its split() function,
              and when using modifier <c>/g</c>. The Perl behavior can be
              emulated after matching a null string by first trying the
              match again at the same offset with <c>notempty_atstart</c> and
              <c>anchored</c>, and then, if that fails, by advancing the
              starting offset (see below) and trying an ordinary match
              again.</p>
          </item>
          <tag><c>notbol</c></tag>
          <item>
            <p>Specifies that the first character of the subject string is not
              the beginning of a line, so the circumflex metacharacter is not
              to match before it. Setting this without <c>multiline</c> (at
              compile time) causes circumflex never to match. This option only
              affects the behavior of the circumflex metacharacter. It does not
              affect \A.</p>
          </item>
          <tag><c>noteol</c></tag>
          <item>
            <p>Specifies that the end of the subject string is not the end of a
              line, so the dollar metacharacter is not to match it nor (except
              in multiline mode) a newline immediately before it. Setting this
              without <c>multiline</c> (at compile time) causes dollar never to
              match. This option affects only the behavior of the dollar
              metacharacter. It does not affect \Z or \z.</p>
          </item>
          <tag><c>report_errors</c></tag>
          <item>
            <p>Gives better control of the error handling in <c>run/3</c>. When
              specified, compilation errors (if the regular expression is not
              already compiled) and runtime errors are explicitly returned as
              an error tuple.</p> 
            <p>The following are the possible runtime errors:</p>
            <taglist>
              <tag><c>match_limit</c></tag>
              <item>
                <p>The PCRE library sets a limit on how many times the internal
                  match function can be called. Defaults to 10,000,000 in the
                  library compiled for Erlang. If <c>{error, match_limit}</c>
                  is returned, the execution of the regular expression has
                  reached this limit. This is normally to be regarded as a
                  <c>nomatch</c>, which is the default return value when this
                  occurs, but by specifying <c>report_errors</c>, you are
                  informed when the match fails because of too many internal
                  calls.</p>
              </item>
              <tag><c>match_limit_recursion</c></tag>
              <item>
                <p>This error is very similar to <c>match_limit</c>, but occurs
                  when the internal match function of PCRE is "recursively"
                  called more times than the <c>match_limit_recursion</c> limit,
                  which defaults to 10,000,000 as well. Notice that as long as
                  the <c>match_limit</c>
                  and <c>match_limit_default</c> values are
                  kept at the default values, the <c>match_limit_recursion</c>
                  error cannot occur, as the <c>match_limit</c> error occurs
                  before that (each recursive call is also a call, but not
                  conversely). Both limits can however be changed, either by
                  setting limits directly in the regular expression string (see
                  section <seealso marker="#regexp_syntax_details">PCRE Regular
                  Eexpression Details</seealso>) or by specifying options to
                  <c>run/3</c>.</p>
              </item>
            </taglist>
            <p>It is important to understand that what is referred to as
              "recursion" when limiting matches is not recursion on the C stack
              of the Erlang machine or on the Erlang process stack. The PCRE
              version compiled into the Erlang VM uses machine "heap" memory to
              store values that must be kept over recursion in regular
              expression matches.</p>
          </item>
          <tag><c>{match_limit, integer() >= 0}</c></tag>
          <item>
            <p>Limits the execution time of a match in an
              implementation-specific way. It is described as follows by the
              PCRE documentation:</p>
            <code>
The match_limit field provides a means of preventing PCRE from using
up a vast amount of resources when running patterns that are not going
to match, but which have a very large number of possibilities in their
search trees. The classic example is a pattern that uses nested
unlimited repeats.

Internally, pcre_exec() uses a function called match(), which it calls
repeatedly (sometimes recursively). The limit set by match_limit is
imposed on the number of times this function is called during a match,
which has the effect of limiting the amount of backtracking that can
take place. For patterns that are not anchored, the count restarts
from zero for each position in the subject string.</code>
            <p>This means that runaway regular expression matches can fail
              faster if the limit is lowered using this option. The default
              value 10,000,000 is compiled into the Erlang VM.</p>
            <note>
              <p>This option does in no way affect the execution of the Erlang
                VM in terms of "long running BIFs". <c>run/3</c> always gives
                control back to the scheduler of Erlang processes at intervals
                that ensures the real-time properties of the Erlang system.</p>
            </note>
          </item>
          <tag><c>{match_limit_recursion, integer() >= 0}</c></tag>
          <item>
            <p>Limits the execution time and memory consumption of a match in an
              implementation-specific way, very similar to <c>match_limit</c>.
              It is described as follows by the PCRE documentation:</p>
            <code>
The match_limit_recursion field is similar to match_limit, but instead
of limiting the total number of times that match() is called, it
limits the depth of recursion. The recursion depth is a smaller number
than the total number of calls, because not all calls to match() are
recursive. This limit is of use only if it is set smaller than
match_limit.

Limiting the recursion depth limits the amount of machine stack that
can be used, or, when PCRE has been compiled to use memory on the heap
instead of the stack, the amount of heap memory that can be used.</code>
            <p>The Erlang VM uses a PCRE library where heap memory is used when
              regular expression match recursion occurs. This therefore limits
              the use of machine heap, not C stack.</p>
            <p>Specifying a lower value can result in matches with deep
              recursion failing, when they should have matched:</p>
            <code type="none">
1&gt; re:run("aaaaaaaaaaaaaz","(a+)*z").
{match,[{0,14},{0,13}]}
2&gt; re:run("aaaaaaaaaaaaaz","(a+)*z",[{match_limit_recursion,5}]).
nomatch
3&gt; re:run("aaaaaaaaaaaaaz","(a+)*z",[{match_limit_recursion,5},report_errors]).
{error,match_limit_recursion}</code>
            <p>This option and option <c>match_limit</c> are only to be used in
              rare cases. Understanding of the PCRE library internals is
              recommended before tampering with these limits.</p>
          </item>
          <tag><c>{offset, integer() >= 0}</c></tag>
          <item>
            <p>Start matching at the offset (position) specified in the
              subject string. The offset is zero-based, so that the default is
              <c>{offset,0}</c> (all of the subject string).</p>
          </item>
          <tag><c>{newline, <anno>NLSpec</anno>}</c></tag>
          <item>
            <p>Overrides the default definition of a newline in the subject
              string, which is LF (ASCII 10) in Erlang.</p>
            <taglist>
              <tag><c>cr</c></tag>
              <item>
                <p>Newline is indicated by a single character CR (ASCII 13).</p>
              </item>
              <tag><c>lf</c></tag>
              <item>
                <p>Newline is indicated by a single character LF (ASCII 10),
                  the default.</p>
              </item>
              <tag><c>crlf</c></tag>
              <item>
                <p>Newline is indicated by the two-character CRLF (ASCII 13
                  followed by ASCII 10) sequence.</p>
              </item>
              <tag><c>anycrlf</c></tag>
              <item>
                <p>Any of the three preceding sequences is be recognized.</p>
              </item>
              <tag><c>any</c></tag>
              <item>
                <p>Any of the newline sequences above, and the Unicode
                sequences VT (vertical tab, U+000B), FF (formfeed, U+000C), NEL
                (next line, U+0085), LS (line separator, U+2028), and PS
                (paragraph separator, U+2029).</p>
              </item>
            </taglist>
          </item>
          <tag><c>bsr_anycrlf</c></tag>
          <item>
            <p>Specifies specifically that \R is to match only the CR
              LF, or CRLF sequences, not the Unicode-specific newline
              characters. (Overrides the compilation option.)</p>
          </item>
          <tag><c>bsr_unicode</c></tag>
          <item>
            <p>Specifies specifically that \R is to match all the Unicode
              newline characters (including CRLF, and so on, the default).
              (Overrides the compilation option.)</p>
          </item>
          <tag><c>{capture, <anno>ValueSpec</anno>}</c>/<c>{capture,
            <anno>ValueSpec</anno>, <anno>Type</anno>}</c></tag>
          <item>
            <p>Specifies which captured substrings are returned and in what
              format. By default, <c>run/3</c> captures all of the matching
              part of the substring and all capturing subpatterns (all of the
              pattern is automatically captured). The default return type is
              (zero-based) indexes of the captured parts of the string,
              specified as <c>{Offset,Length}</c> pairs (the <c>index</c>
              <c><anno>Type</anno></c> of capturing).</p>
            <p>As an example of the default behavior, the following call
              returns, as first and only captured string, the matching part of
              the subject ("abcd" in the middle) as an index pair <c>{3,4}</c>,
              where character positions are zero-based, just as in offsets:</p>
            <code>
re:run("ABCabcdABC","abcd",[]).</code>      
            <p>The return value of this call is:</p>
            <code>
{match,[{3,4}]}</code>
            <p>Another (and quite common) case is where the regular expression
              matches all of the subject:</p>
            <code>
re:run("ABCabcdABC",".*abcd.*",[]).</code>
            <p>Here the return value correspondingly points out all of the
              string, beginning at index 0, and it is 10 characters long:</p>
            <code>
{match,[{0,10}]}</code>
            <p>If the regular expression contains capturing subpatterns, like
              in:</p>
            <code>
re:run("ABCabcdABC",".*(abcd).*",[]).</code>      
            <p>all of the matched subject is captured, as well as the captured
              substrings:</p>
            <code>
{match,[{0,10},{3,4}]}</code>
            <p>The complete matching pattern always gives the first return
              value in the list and the remaining subpatterns are added in the
              order they occurred in the regular expression.</p>
            <p>The capture tuple is built up as follows:</p>
            <taglist>
              <tag><c><anno>ValueSpec</anno></c></tag>
              <item>
                <p>Specifies which captured (sub)patterns are to be returned.
                  <c><anno>ValueSpec</anno></c> can either be an atom describing
                  a predefined set of return values, or a list containing the
                  indexes or the names of specific subpatterns to return.</p>
                <p>The following are the predefined sets of subpatterns:</p>
                <taglist>
                  <tag><c>all</c></tag>
                  <item>
                    <p>All captured subpatterns including the complete matching
                      string. This is the default.</p>
                  </item>
                  <tag><c>all_names</c></tag>
                  <item>
                    <p>All <em>named</em> subpatterns in the regular expression,
                      as if a <c>list()</c> of all the names <em>in
                      alphabetical order</em> was specified. The list of all
                      names can also be retrieved with
                      <seealso marker="#inspect/2">
                      <c>inspect/2</c></seealso>.</p>
                  </item>
                  <tag><c>first</c></tag>
                  <item>
                    <p>Only the first captured subpattern, which is always the
                      complete matching part of the subject. All explicitly
                      captured subpatterns are discarded.</p>
                  </item>
                  <tag><c>all_but_first</c></tag>
                  <item>
                    <p>All but the first matching subpattern, that is, all
                      explicitly captured subpatterns, but not the complete
                      matching part of the subject string. This is useful if
                      the regular expression as a whole matches a large part of
                      the subject, but the part you are interested in is in an
                      explicitly captured subpattern. If the return type is
                      <c>list</c> or <c>binary</c>, not returning subpatterns
                      you are not interested in is a good way to optimize.</p>
                  </item>
                  <tag><c>none</c></tag>
                  <item>
                    <p>Returns no matching subpatterns, gives the single
                      atom <c>match</c> as the return value of the function
                      when matching successfully instead of the <c>{match,
                      list()}</c> return. Specifying an empty list gives the
                      same behavior.</p>
                  </item>
                </taglist>
                <p>The value list is a list of indexes for the subpatterns to
                  return, where index 0 is for all of the pattern, and 1 is for
                  the first explicit capturing subpattern in the regular
                  expression, and so on. When using named captured subpatterns
                  (see below) in the regular expression, one can use
                  <c>atom()</c>s or <c>string()</c>s to specify the subpatterns
                  to be returned. For example, consider the regular
                  expression:</p>
                <code>
".*(abcd).*"</code>
                 <p>matched against string "ABCabcdABC", capturing only the
                   "abcd" part (the first explicit subpattern):</p>
                <code>
re:run("ABCabcdABC",".*(abcd).*",[{capture,[1]}]).</code>
                <p>The call gives the following result, as the first explicitly
                  captured subpattern is "(abcd)", matching "abcd" in the
                  subject, at (zero-based) position 3, of length 4:</p>
                <code>
{match,[{3,4}]}</code>
                <p>Consider the same regular expression, but with the subpattern
                  explicitly named 'FOO':</p>
                <code>
".*(?&lt;FOO&gt;abcd).*"</code>
                <p>With this expression, we could still give the index of the
                  subpattern with the following call:</p>
                <code>
re:run("ABCabcdABC",".*(?&lt;FOO&gt;abcd).*",[{capture,[1]}]).</code>
                <p>giving the same result as before. But, as the subpattern is
                  named, we can also specify its name in the value list:</p>
                <code>
re:run("ABCabcdABC",".*(?&lt;FOO&gt;abcd).*",[{capture,['FOO']}]).</code>
                <p>This would give the same result as the earlier examples,
                  namely:</p>
                <code>
{match,[{3,4}]}</code>
                <p>The values list can specify indexes or names not present in
                  the regular expression, in which case the return values vary
                  depending on the type. If the type is <c>index</c>, the tuple
                  <c>{-1,0}</c> is returned for values with no corresponding
                  subpattern in the regular expression, but for the other types
                  (<c>binary</c> and <c>list</c>), the values are the empty
                  binary or list, respectively.</p>
              </item> 
              <tag><c><anno>Type</anno></c></tag>
              <item>
                <p>Optionally specifies how captured substrings are to be
                  returned. If omitted, the default of <c>index</c> is used.</p>
                <p><c><anno>Type</anno></c> can be one of the following:</p>
                <taglist>
                  <tag><c>index</c></tag> 
                  <item>
                    <p>Returns captured substrings as pairs of byte indexes
                      into the subject string and length of the matching string
                      in the subject (as if the subject string was flattened
                      with <seealso marker="erts:erlang#iolist_to_binary/1">
                      <c>erlang:iolist_to_binary/1</c></seealso> or
                      <seealso marker="unicode#characters_to_binary/2">
                      <c>unicode:characters_to_binary/2</c></seealso> before
                      matching). Notice that option <c>unicode</c> results in
                      <em>byte-oriented</em> indexes in a (possibly virtual)
                      <em>UTF-8 encoded</em> binary. A byte index tuple
                      <c>{0,2}</c> can therefore represent one or two
                      characters when <c>unicode</c> is in effect. This can seem
                      counter-intuitive, but has been deemed the most effective
                      and useful way to do it. To return lists instead can
                      result in simpler code if that is desired. This return
                      type is the default.</p>
                  </item>
                  <tag><c>list</c></tag>
                  <item>
                    <p>Returns matching substrings as lists of characters
                      (Erlang <c>string()</c>s). It option <c>unicode</c> is
                      used in combination with the \C sequence in the
                      regular expression, a captured subpattern can contain
                      bytes that are not valid UTF-8 (\C matches bytes
                      regardless of character encoding). In that case the
                      <c>list</c> capturing can result in the same types of
                      tuples that
                      <seealso marker="unicode#characters_to_list/2">
                      <c>unicode:characters_to_list/2</c></seealso> can return,
                      namely three-tuples with tag <c>incomplete</c> or
                      <c>error</c>, the successfully converted characters and
                      the invalid UTF-8 tail of the conversion as a binary. The
                      best strategy is to avoid using the \C sequence
                      when capturing lists.</p>
                  </item>
                  <tag><c>binary</c></tag> 
                  <item>
                    <p>Returns matching substrings as binaries. If option
                      <c>unicode</c> is used, these binaries are in UTF-8. If
                      the \C sequence is used together with
                      <c>unicode</c>, the binaries can be invalid UTF-8.</p>
                  </item>
                </taglist>
              </item>
            </taglist>
            <p>In general, subpatterns that were not assigned a value in the
              match are returned as the tuple <c>{-1,0}</c> when <c>type</c> is
              <c>index</c>. Unassigned subpatterns are returned as the empty
              binary or list, respectively, for other return types. Consider
              the following regular expression:</p>
            <code>
".*((?&lt;FOO&gt;abdd)|a(..d)).*"</code>
            <p>There are three explicitly capturing subpatterns, where the
              opening parenthesis position determines the order in the result,
              hence <c>((?&lt;FOO&gt;abdd)|a(..d))</c> is subpattern index 1,
              <c>(?&lt;FOO&gt;abdd)</c> is subpattern index 2, and <c>(..d)</c>
              is subpattern index 3. When matched against the following
             string:</p>
            <code>
"ABCabcdABC"</code>
            <p>the subpattern at index 2 does not match, as "abdd" is not
              present in the string, but the complete pattern matches (because
              of the alternative <c>a(..d)</c>). The subpattern at index 2 is
              therefore unassigned and the default return value is:</p>
            <code>
{match,[{0,10},{3,4},{-1,0},{4,3}]}</code>
            <p>Setting the capture <c><anno>Type</anno></c> to <c>binary</c>
              gives:</p>
            <code>
{match,[&lt;&lt;"ABCabcdABC"&gt;&gt;,&lt;&lt;"abcd"&gt;&gt;,&lt;&lt;&gt;&gt;,&lt;&lt;"bcd"&gt;&gt;]}</code>
            <p>Here the empty binary (<c>&lt;&lt;&gt;&gt;</c>) represents the
              unassigned subpattern. In the <c>binary</c> case, some information
              about the matching is therefore lost, as
	      <c>&lt;&lt;&gt;&gt;</c> can
              also be an empty string captured.</p>
            <p>If differentiation between empty matches and non-existing
              subpatterns is necessary, use the <c>type</c> <c>index</c> and do
              the conversion to the final type in Erlang code.</p>
            <p>When option <c>global</c> is speciified, the <c>capture</c>
              specification affects each match separately, so that:</p>
            <code>
re:run("cacb","c(a|b)",[global,{capture,[1],list}]).</code>
            <p>gives</p>
            <code>
{match,[["a"],["b"]]}</code>
          </item>
        </taglist>
        <p>For a descriptions of options only affecting the compilation step,
          see <seealso marker="#compile/2"><c>compile/2</c></seealso>.</p>
      </desc>
    </func>

    <func>
      <name name="split" arity="2"/>
      <fsummary>Split a string by tokens specified as a regular expression.
      </fsummary>
      <desc>
        <p>Same as <c>split(<anno>Subject</anno>, <anno>RE</anno>, [])</c>.</p>
      </desc>
    </func>

    <func>
      <name name="split" arity="3"/>
      <fsummary>Split a string by tokens specified as a regular expression</fsummary>
      <type_desc variable="CompileOpt">See <seealso marker="#compile_options">
        <c>compile/2</c></seealso>.</type_desc>
      <desc>
        <p>Splits the input into parts by finding tokens according to the
          regular expression supplied. The splitting is basically done by
          running a global regular expression match and dividing the initial
          string wherever a match occurs. The matching part of the string is
          removed from the output.</p>
        <p>As in <seealso marker="#run/3"><c>run/3</c></seealso>, an <c>mp()</c>
          compiled with option <c>unicode</c> requires
          <c><anno>Subject</anno></c> to be a Unicode <c>charlist()</c>. If
          compilation is done implicitly and the <c>unicode</c> compilation
          option is specified to this function, both the regular expression and
          <c><anno>Subject</anno></c> are to be specified as valid Unicode
          <c>charlist()</c>s.</p>
        <p>The result is given as a list of &quot;strings&quot;, the preferred
          data type specified in option <c>return</c> (default
          <c>iodata</c>).</p>
        <p>If subexpressions are specified in the regular expression, the
          matching subexpressions are returned in the resulting list as
          well. For example:</p>
        <code>
re:split("Erlang","[ln]",[{return,list}]).</code>
        <p>gives</p>
        <code>
["Er","a","g"]</code>
        <p>while</p>
        <code>
re:split("Erlang","([ln])",[{return,list}]).</code>
        <p>gives</p>
        <code>
["Er","l","a","n","g"]</code>
        <p>The text matching the subexpression (marked by the parentheses in the
          regular expression) is inserted in the result list where it was found.
          This means that concatenating the result of a split where the whole
          regular expression is a single subexpression (as in the last example)
          always results in the original string.</p>
        <p>As there is no matching subexpression for the last part in the
          example (the &quot;g&quot;), nothing is inserted after that. To make
          the group of strings and the parts matching the subexpressions more
          obvious, one can use option <c>group</c>, which groups together the
          part of the subject string with the parts matching the subexpressions
          when the string was split:</p>
        <code>
re:split("Erlang","([ln])",[{return,list},group]).</code>
        <p>gives</p>
        <code>
[["Er","l"],["a","n"],["g"]]</code>      
        <p>Here the regular expression first matched the &quot;l&quot;,
          causing &quot;Er&quot; to be the first part in the result. When
          the regular expression matched, the (only) subexpression was
          bound to the &quot;l&quot;, so the &quot;l&quot; is inserted
          in the group together with &quot;Er&quot;. The next match is of
          the &quot;n&quot;, making &quot;a&quot; the next part to be
          returned. As the subexpression is bound to substring
          &quot;n&quot; in this case, the &quot;n&quot; is inserted into
          this group. The last group consists of the remaining string,
          as no more matches are found.</p>
        <p>By default, all parts of the string, including the empty strings,
          are returned from the function, for example:</p>
        <code>
re:split("Erlang","[lg]",[{return,list}]).</code>
        <p>gives</p>
        <code>
["Er","an",[]]</code>
        <p>as the matching of the &quot;g&quot; in the end of the string
          leaves an empty rest, which is also returned. This behavior
          differs from the default behavior of the split function in
          Perl, where empty strings at the end are by default removed. To
          get the &quot;trimming&quot; default behavior of Perl, specify
          <c>trim</c> as an option:</p>
        <code>
re:split("Erlang","[lg]",[{return,list},trim]).</code>
        <p>gives</p>
        <code>
["Er","an"]</code> 
        <p>The &quot;trim&quot; option says; &quot;give me as many parts as
          possible except the empty ones&quot;, which sometimes can be
          useful. You can also specify how many parts you want, by specifying
          <c>{parts,</c>N<c>}</c>:</p>
        <code>
re:split("Erlang","[lg]",[{return,list},{parts,2}]).</code>    
        <p>gives</p>
        <code>
["Er","ang"]</code>
        <p>Notice that the last part is &quot;ang&quot;, not
          &quot;an&quot;, as splitting was specified into two parts,
          and the splitting stops when enough parts are given, which is
          why the result differs from that of <c>trim</c>.</p>
        <p>More than three parts are not possible with this indata, so</p>
        <code>
re:split("Erlang","[lg]",[{return,list},{parts,4}]).</code>
        <p>gives the same result as the default, which is to be
          viewed as &quot;an infinite number of parts&quot;.</p> 
        <p>Specifying <c>0</c> as the number of parts gives the same
          effect as option <c>trim</c>. If subexpressions are
          captured, empty subexpressions matched at the end are also
          stripped from the result if <c>trim</c> or <c>{parts,0}</c> is
          specified.</p>
        <p>The <c>trim</c> behavior corresponds exactly to the Perl default.
          <c>{parts,N}</c>, where N is a positive integer, corresponds
          exactly to the Perl behavior with a positive numerical third
          parameter. The default behavior of <c>split/3</c> corresponds
          to the Perl behavior when a negative integer is specified as
          the third parameter for the Perl routine.</p>
        <p>Summary of options not previously described for function
          <c>run/3</c>:</p>
        <taglist>
          <tag><c>{return,<anno>ReturnType</anno>}</c></tag>
          <item>
            <p>Specifies how the parts of the original string are presented in
              the result list. Valid types:</p>
            <taglist>
              <tag><c>iodata</c></tag>
              <item>
                <p>The variant of <c>iodata()</c> that gives the least copying
                  of data with the current implementation (often a binary, but
                  do not depend on it).</p></item>
              <tag><c>binary</c></tag>
              <item>
                <p>All parts returned as binaries.</p></item>
              <tag><c>list</c></tag>
              <item>
                <p>All parts returned as lists of characters
                  (&quot;strings&quot;).</p>
              </item>
            </taglist>
          </item>
          <tag><c>group</c></tag>
          <item>
            <p>Groups together the part of the string with
              the parts of the string matching the subexpressions of the
              regular expression.</p>
            <p>The return value from the function is in this case a
              <c>list()</c> of <c>list()</c>s. Each sublist begins with the
              string picked out of the subject string, followed by the parts
              matching each of the subexpressions in order of occurrence in the
              regular expression.</p>
          </item>
          <tag><c>{parts,N}</c></tag>
          <item>
            <p>Specifies the number of parts the subject string is to be
              split into.</p>
            <p>The number of parts is to be a positive integer for a specific
              maximum number of parts, and <c>infinity</c> for the
              maximum number of parts possible (the default). Specifying
              <c>{parts,0}</c> gives as many parts as possible disregarding
              empty parts at the end, the same as specifying <c>trim</c>.</p>
          </item>
          <tag><c>trim</c></tag>
          <item>
            <p>Specifies that empty parts at the end of the result list are
              to be disregarded. The same as specifying <c>{parts,0}</c>. This
              corresponds to the default behavior of the <c>split</c>
              built-in function in Perl.</p>
          </item>
        </taglist>
      </desc>
    </func>     
  </funcs>

  <section>
    <marker id="regexp_syntax"></marker>
    <title>Perl-Like Regular Expression Syntax</title>
    <p>The following sections contain reference material for the regular
      expressions used by this module. The information is based on the PCRE
      documentation, with changes where this module behaves differently to
      the PCRE library.</p>
  </section>

  <section>
    <marker id="regexp_syntax_details"></marker>
    <title>PCRE Regular Expression Details</title>
    <p>The syntax and semantics of the regular expressions supported by PCRE are
      described in detail in the following sections. Perl's regular expressions
      are described in its own documentation, and regular expressions in general
      are covered in many books, some with copious examples.
      Jeffrey Friedl's "Mastering Regular Expressions", published by O'Reilly,
      covers regular expressions in great detail. This description of the PCRE
      regular expressions is intended as reference material.</p>

    <p>The reference material is divided into the following sections:</p>

    <list type="bulleted">
      <item><seealso marker="#sect1">Special Start-of-Pattern Items</seealso>
      </item>
      <item><seealso marker="#sect2">Characters and Metacharacters</seealso>
      </item>
      <item><seealso marker="#sect3">Backslash</seealso></item>
      <item><seealso marker="#sect4">Circumflex and Dollar</seealso></item>
      <item><seealso marker="#sect5">Full Stop (Period, Dot) and \N</seealso>
      </item>
      <item><seealso marker="#sect6">Matching a Single Data Unit</seealso>
      </item>
      <item><seealso marker="#sect7">Square Brackets and Character
        Classes</seealso></item>
      <item><seealso marker="#sect8">Posix Character Classes</seealso></item>
      <item><seealso marker="#sect9">Vertical Bar</seealso></item>
      <item><seealso marker="#sect10">Internal Option Setting</seealso></item>
      <item><seealso marker="#sect11">Subpatterns</seealso></item>
      <item><seealso marker="#sect12">Duplicate Subpattern Numbers</seealso>
      </item>
      <item><seealso marker="#sect13">Named Subpatterns</seealso></item>
      <item><seealso marker="#sect14">Repetition</seealso></item>
      <item><seealso marker="#sect15">Atomic Grouping and Possessive
        Quantifiers</seealso></item>
      <item><seealso marker="#sect16">Back References</seealso></item>
      <item><seealso marker="#sect17">Assertions</seealso></item>
      <item><seealso marker="#sect18">Conditional Subpatterns</seealso></item>
      <item><seealso marker="#sect19">Comments</seealso></item>
      <item><seealso marker="#sect20">Recursive Patterns</seealso></item>
      <item><seealso marker="#sect21">Subpatterns as Subroutines</seealso>
      </item>
      <item><seealso marker="#sect22">Oniguruma Subroutine Syntax</seealso>
      </item>
      <item><seealso marker="#sect23">Backtracking Control</seealso></item>
    </list>
  </section>

  <section>
    <marker id="sect1"></marker>
    <title>Special Start-of-Pattern Items</title>
    <p>Some options that can be passed to <seealso marker="#compile/2">
      <c>compile/2</c></seealso> can also be set by special items at the start
      of a pattern. These are not Perl-compatible, but are provided to make
      these options accessible to pattern writers who are not able to change
      the program that processes the pattern. Any number of these items can
      appear, but they must all be together right at the start of the
      pattern string, and the letters must be in upper case.</p>

    <p><em>UTF Support</em></p>

    <p>Unicode support is basically UTF-8 based. To use Unicode characters, you
      either call <seealso marker="#compile/2"><c>compile/2</c></seealso> or
      <seealso marker="#run/3"><c>run/3</c></seealso> with option
      <c>unicode</c>, or the pattern must start with one of these special
      sequences:</p>

    <code>
(*UTF8)
(*UTF)</code>

    <p>Both options give the same effect, the input string is interpreted as
      UTF-8. Notice that with these instructions, the automatic conversion of
      lists to UTF-8 is not performed by the <c>re</c> functions. Therefore,
      using these sequences is not recommended.
      Add option <c>unicode</c> when running
      <seealso marker="#compile/2"><c>compile/2</c></seealso> instead.</p>

    <p>Some applications that allow their users to supply patterns can wish to
      restrict them to non-UTF data for security reasons. If option
      <c>never_utf</c> is set at compile time, (*UTF), and so on, are not
      allowed, and their appearance causes an error.</p>

    <p><em>Unicode Property Support</em></p>

    <p>The following is another special sequence that can appear at the start of
      a pattern:</p>

    <code>
(*UCP)</code>

    <p>This has the same effect as setting option <c>ucp</c>: it causes
      sequences such as \d and \w to use Unicode properties to
      determine character types, instead of recognizing only characters with
      codes &lt; 256 through a lookup table.</p>

    <p><em>Disabling Startup Optimizations</em></p>

    <p>If a pattern starts with <c>(*NO_START_OPT)</c>,
    it has the same effect as
    setting option <c>no_start_optimize</c> at compile time.</p>

    <p><em>Newline Conventions</em></p>
    <marker id="newline_conventions"></marker>

    <p>PCRE supports five conventions for indicating line breaks in strings: a
      single CR (carriage return) character, a single LF (line feed) character,
      the two-character sequence CRLF, any of the three preceding, and any
      Unicode newline sequence.</p>

    <p>A newline convention can also be specified by starting a pattern string
      with one of the following five sequences:</p>

    <taglist>
      <tag>(*CR)</tag><item>Carriage return</item>
      <tag>(*LF)</tag><item>Line feed</item>
      <tag>(*CRLF)</tag><item>>Carriage return followed by
        line feed</item>
      <tag>(*ANYCRLF)</tag><item>Any of the three above</item>
      <tag>(*ANY)</tag><item>All Unicode newline sequences</item>
    </taglist>

    <p>These override the default and the options specified to
      <seealso marker="#compile/2"><c>compile/2</c></seealso>. For example, the
      following pattern changes the convention to CR:</p>

    <code>
(*CR)a.b</code>

    <p>This pattern matches <c>a\nb</c>, as LF is no longer a newline.
    If more than one of them is present, the last one is used.</p>

    <p>The newline convention affects where the circumflex and dollar assertions
      are true. It also affects the interpretation of the dot metacharacter when
      <c>dotall</c> is not set, and the behavior of \N. However, it does not
      affect what the \R escape sequence matches. By default, this is any
      Unicode newline sequence, for Perl compatibility. However, this can be
      changed; see the description of \R in section
      <seealso marker="#newline_sequences">Newline Sequences</seealso>. A change
      of the \R setting can be combined with a change of the newline
      convention.</p>

    <p><em>Setting Match and Recursion Limits</em></p>

    <p>The caller of <seealso marker="#run/3"><c>run/3</c></seealso> can set a
      limit on the number of times the internal match() function is called and
      on the maximum depth of recursive calls. These facilities are provided to
      catch runaway matches that are provoked by patterns with huge matching
      trees (a typical example is a pattern with nested unlimited repeats) and
      to avoid running out of system stack by too much recursion. When one of
      these limits is reached, <c>pcre_exec()</c> gives an error return. The
      limits can also be set by items at the start of the pattern of the
      following forms:</p>

    <code>
(*LIMIT_MATCH=d)
(*LIMIT_RECURSION=d)</code>

    <p>Here d is any number of decimal digits. However, the value of the setting
      must be less than the value set by the caller of <c>run/3</c> for it to
      have any effect. That is, the pattern writer can lower the limit set by
      the programmer, but not raise it. If there is more than one setting of one
      of these limits, the lower value is used.</p>

    <p>The default value for both the limits is 10,000,000 in the Erlang
      VM. Notice that the recursion limit does not affect the stack depth of the
      VM, as PCRE for Erlang is compiled in such a way that the match function
      never does recursion on the C stack.</p>
  </section>

  <section>
    <marker id="sect2"></marker>
    <title>Characters and Metacharacters</title>
    <!-- .rs -->
    <p>A regular expression is a pattern that is matched against a subject
      string from left to right. Most characters stand for themselves in a
      pattern and match the corresponding characters in the subject. As a
      trivial example, the following pattern matches a portion of a subject
      string that is identical to itself:</p>

    <code>
The quick brown fox</code>

    <p>When caseless matching is specified (option <c>caseless</c>), letters
      are matched independently of case.</p>

    <p>The power of regular expressions comes from the ability to include
      alternatives and repetitions in the pattern. These are encoded in the
      pattern by the use of <em>metacharacters</em>, which do not stand for
      themselves but instead are interpreted in some special way.</p>

    <p>Two sets of metacharacters exist: those that are recognized anywhere in
      the pattern except within square brackets, and those that are recognized
      within square brackets. Outside square brackets, the metacharacters are
      as follows:</p>

    <taglist>
      <tag>\</tag><item>General escape character with many uses</item>
      <tag>^</tag><item>Assert start of string (or line, in multiline mode)
        </item>
      <tag>$</tag><item>Assert end of string (or line, in multiline mode)</item>
      <tag>.</tag><item>Match any character except newline (by default)</item>
      <tag>[</tag><item>Start character class definition</item>
      <tag>|</tag><item>Start of alternative branch</item>
      <tag>(</tag><item>Start subpattern</item>
      <tag>)</tag><item>End subpattern</item>
      <tag>?</tag><item>Extends the meaning of (, also 0 or 1 quantifier, also
        quantifier minimizer</item>
      <tag>*</tag><item>0 or more quantifiers</item>
      <tag>+</tag><item>1 or more quantifier, also "possessive quantifier"
        </item>
      <tag>{</tag><item>Start min/max quantifier</item>
</taglist>

    <p>Part of a pattern within square brackets is called a "character class".
      The following are the only metacharacters in a character class:</p>

    <taglist>
      <tag>\</tag><item>General escape character</item>
      <tag>^</tag><item>Negate the class, but only if the first character</item>
      <tag>-</tag><item>Indicates character range</item>
      <tag>[</tag><item>Posix character class (only if followed by Posix syntax)
        </item>
      <tag>]</tag><item>Terminates the character class</item>
    </taglist>

    <p>The following sections describe the use of each metacharacter.</p>
  </section>

  <section>
  <marker id="sect3"></marker>
  <title>Backslash</title>
    <p>The backslash character has many uses. First, if it is followed by a
      character that is not a number or a letter, it takes away any special
      meaning that a character can have. This use of backslash as an escape
      character applies both inside and outside character classes.</p>

    <p>For example, if you want to match a * character, you write \* in the
      pattern. This escaping action applies if the following character would
      otherwise be interpreted as a metacharacter, so it is always safe to
      precede a non-alphanumeric with backslash to specify that it stands for
      itself. In particular, if you want to match a backslash, write \\.</p>

    <p>In <c>unicode</c> mode, only ASCII numbers and letters have any special
      meaning after a backslash. All other characters (in particular, those
      whose code points are &gt; 127) are treated as literals.</p>

    <p>If a pattern is compiled with option <c>extended</c>, whitespace in the
      pattern (other than in a character class) and characters between a #
      outside a character class and the next newline are ignored. An escaping
      backslash can be used to include a whitespace or # character as part of
      the pattern.</p>

    <p>To remove the special meaning from a sequence of characters, put them
      between \Q and \E. This is different from Perl in that $ and @ are
      handled as literals in \Q...\E sequences in PCRE, while $ and @ cause
      variable interpolation in Perl. Notice the following examples:</p>

<code type="none">
Pattern            PCRE matches   Perl matches

\Qabc$xyz\E        abc$xyz        abc followed by the contents of $xyz
\Qabc\$xyz\E       abc\$xyz       abc\$xyz
\Qabc\E\$\Qxyz\E   abc$xyz        abc$xyz</code>


    <p>The \Q...\E sequence is recognized both inside and outside character
      classes. An isolated \E that is not preceded by \Q is ignored. If \Q is
      not followed by \E later in the pattern, the literal interpretation
      continues to the end of the pattern (that is, \E is assumed at the end).
      If the isolated \Q is inside a character class, this causes an error, as
      the character class is not terminated.</p>

    <p><em>Non-Printing Characters</em></p>
    <marker id="non_printing_characters"></marker>

    <p>A second use of backslash provides a way of encoding non-printing
      characters in patterns in a visible manner. There is no restriction on the
      appearance of non-printing characters, apart from the binary zero that
      terminates a pattern. When a pattern is prepared by text editing, it is
      often easier to use one of the following escape sequences than the binary
      character it represents:</p>

    <taglist>
      <tag>\a</tag><item>Alarm, that is, the BEL character (hex 07)</item>
      <tag>\cx</tag><item>"Control-x", where x is any ASCII character</item>
      <tag>\e</tag><item>Escape (hex 1B)</item>
      <tag>\f</tag><item>Form feed (hex 0C)</item>
      <tag>\n</tag><item>Line feed (hex 0A)</item>
      <tag>\r</tag><item>Carriage return (hex 0D)</item>
      <tag>\t</tag><item>Tab (hex 09)</item>
      <tag>\ddd</tag><item>Character with octal code ddd, or back reference
        </item>
      <tag>\xhh</tag><item>Character with hex code hh</item>
      <tag>\x{hhh..}</tag><item>Character with hex code hhh..</item>
    </taglist>

    <p>The precise effect of \cx on ASCII characters is as follows: if x is a
      lowercase letter, it is converted to upper case. Then bit 6 of the
      character (hex 40) is inverted. Thus \cA to \cZ become hex 01 to hex 1A
      (A is 41, Z is 5A), but \c{ becomes hex 3B ({ is 7B), and \c; becomes
      hex 7B (; is 3B). If the data item (byte or 16-bit value) following \c
      has a value &gt; 127, a compile-time error occurs. This locks out
      non-ASCII characters in all modes.</p>

    <p>The \c facility was designed for use with ASCII characters, but with the
      extension to Unicode it is even less useful than it once was.</p>

    <p>By default, after \x, from zero to two hexadecimal digits are read
      (letters can be in upper or lower case). Any number of hexadecimal digits
      can appear between \x{ and }, but the character code is constrained as
      follows:</p>

    <taglist>
      <tag>8-bit non-Unicode mode</tag>
      <item>&lt; 0x100</item>
      <tag>8-bit UTF-8 mode</tag>
      <item>&lt; 0x10ffff and a valid code point</item>
    </taglist>

    <p>Invalid Unicode code points are the range 0xd800 to 0xdfff (the so-called
      "surrogate" code points), and 0xffef.</p>

    <p>If characters other than hexadecimal digits appear between \x{ and },
      or if there is no terminating }, this form of escape is not recognized.
      Instead, the initial \x is interpreted as a basic hexadecimal escape,
      with no following digits, giving a character whose value is zero.</p>

    <p>Characters whose value is &lt; 256 can be defined by either of the two
      syntaxes for \x. There is no difference in the way they are handled. For
      example, \xdc is the same as \x{dc}.</p>

    <p>After \0 up to two further octal digits are read. If there are fewer than
      two digits, only those that are present are used. Thus the sequence
      \0\x\07 specifies two binary zeros followed by a BEL character (code value
      7). Ensure to supply two digits after the initial zero if the pattern
      character that follows is itself an octal digit.</p>

    <p>The handling of a backslash followed by a digit other than 0 is
      complicated. Outside a character class, PCRE reads it and any following
      digits as a decimal number. If the number is &lt; 10, or if there have
      been at least that many previous capturing left parentheses in the
      expression, the entire sequence is taken as a <em>back reference</em>. A
      description of how this works is provided later, following the discussion
      of parenthesized subpatterns.</p>

    <p>Inside a character class, or if the decimal number is &gt; 9 and there
      have not been that many capturing subpatterns, PCRE re-reads up to three
      octal digits following the backslash, and uses them to generate a data
      character. Any subsequent digits stand for themselves. The value of the
      character is constrained in the same way as characters specified in
      hexadecimal. For example:</p>

    <taglist>
      <tag>\040</tag>
        <item>Another way of writing an ASCII space</item>
      <tag>\40</tag>
        <item>The same, provided there are &lt; 40 previous capturing
          subpatterns</item>
      <tag>\7</tag>
        <item>Always a back reference</item>
      <tag>\11</tag>
        <item>Can be a back reference, or another way of writing a tab</item>
      <tag>\011</tag>
        <item>Always a tab</item>
      <tag>\0113</tag>
        <item>A tab followed by character "3"</item>
      <tag>\113</tag>
        <item>Can be a back reference, otherwise the character with octal code
          113 </item>
      <tag>\377</tag>
        <item>Can be a back reference, otherwise value 255 (decimal)</item>
      <tag>\81</tag>
        <item>Either a back reference, or a binary zero followed by the two
          characters "8" and "1"</item>
    </taglist>

    <p>Notice that octal values &gt;= 100 must not be introduced by a leading
      zero, as no more than three octal digits are ever read.</p>

    <p>All the sequences that define a single character value can be used both
      inside and outside character classes. Also, inside a character class, \b
      is interpreted as the backspace character (hex 08).</p>

    <p>\N is not allowed in a character class. \B, \R, and \X are not special
      inside a character class. Like other unrecognized escape sequences, they
      are treated as the literal characters "B", "R", and "X". Outside a
      character class, these sequences have different meanings.</p>

    <p><em>Unsupported Escape Sequences</em></p>

    <p>In Perl, the sequences \l, \L, \u, and \U are recognized by its string
      handler and used to modify the case of following characters. PCRE does not
      support these escape sequences.</p>

    <p><em>Absolute and Relative Back References</em></p>

    <p>The sequence \g followed by an unsigned or a negative number, optionally
      enclosed in braces, is an absolute or relative back reference. A named
      back reference can be coded as \g{name}. Back references are discussed
      later, following the discussion of parenthesized subpatterns.</p>

    <p><em>Absolute and Relative Subroutine Calls</em></p>

    <p>For compatibility with Oniguruma, the non-Perl syntax \g followed by a
      name or a number enclosed either in angle brackets or single quotes, is
      alternative syntax for referencing a subpattern as a "subroutine".
      Details are discussed later. Notice that \g{...} (Perl syntax) and
      \g&lt;...&gt; (Oniguruma syntax) are <em>not</em> synonymous. The former
      is a back reference and the latter is a subroutine call.</p>

    <p><em>Generic Character Types</em></p>
    <marker id="generic_character_types"></marker>

    <p>Another use of backslash is for specifying generic character types:</p>

    <taglist>  
      <tag>\d</tag><item>Any decimal digit</item>
      <tag>\D</tag><item>Any character that is not a decimal digit</item>
      <tag>\h</tag><item>Any horizontal whitespace character</item>
      <tag>\H</tag><item>Any character that is not a horizontal whitespace
        character</item>
      <tag>\s</tag><item>Any whitespace character</item>
      <tag>\S</tag><item>Any character that is not a whitespace character
        </item>
      <tag>\v</tag><item>Any vertical whitespace character</item>
      <tag>\V</tag><item>Any character that is not a vertical whitespace
        character</item>
      <tag>\w</tag><item>Any "word" character</item>
      <tag>\W</tag><item>Any "non-word" character</item>
    </taglist>

    <p>There is also the single sequence \N, which matches a non-newline
      character. This is the same as the "." metacharacter when <c>dotall</c>
      is not set. Perl also uses \N to match characters by name, but PCRE does
      not support this.</p>

    <p>Each pair of lowercase and uppercase escape sequences partitions the
      complete set of characters into two disjoint sets. Any given character
      matches one, and only one, of each pair. The sequences can appear both
      inside and outside character classes. They each match one character of the
      appropriate type. If the current matching point is at the end of the
      subject string, all fail, as there is no character to match.</p>

    <p>For compatibility with Perl, \s does not match the VT character
      (code 11). This makes it different from the Posix "space" class. The \s
      characters are HT (9), LF (10), FF (12), CR (13), and space (32). If "use
      locale;" is included in a Perl script, \s can match the VT character. In
      PCRE, it never does.</p>

    <p>A "word" character is an underscore or any character that is a letter or
      a digit. By default, the definition of letters and digits is controlled by
      the PCRE low-valued character tables, in Erlang's case (and without option
      <c>unicode</c>), the ISO Latin-1 character set.</p>

    <p>By default, in <c>unicode</c> mode, characters with values &gt; 255, that
      is, all characters outside the ISO Latin-1 character set, never match \d,
      \s, or \w, and always match \D, \S, and \W. These sequences retain their
      original meanings from before UTF support was available, mainly for
      efficiency reasons. However, if option <c>ucp</c> is set, the behavior is
      changed so that Unicode properties are used to determine character types,
      as follows:</p>

    <taglist>
      <tag>\d</tag><item>Any character that \p{Nd} matches (decimal digit)
        </item>
      <tag>\s</tag><item>Any character that \p{Z} matches, plus HT, LF, FF, CR
        </item>
      <tag>\w</tag><item>Any character that \p{L} or \p{N} matches, plus
        underscore</item>
    </taglist>

    <p>The uppercase escapes match the inverse sets of characters. Notice that
      \d matches only decimal digits, while \w matches any Unicode digit, any
      Unicode letter, and underscore. Notice also that <c>ucp</c> affects \b and
      \B, as they are defined in terms of \w and \W. Matching these sequences is
      noticeably slower when <c>ucp</c> is set.</p>

    <p>The sequences \h, \H, \v, and \V are features that were added to Perl in
      release 5.10. In contrast to the other sequences, which match only ASCII
      characters by default, these always match certain high-valued code points,
      regardless if <c>ucp</c> is set.</p>

    <p>The following are the horizontal space characters:</p>

    <taglist>
      <tag>U+0009</tag><item>Horizontal tab (HT)</item>
      <tag>U+0020</tag><item>Space</item>
      <tag>U+00A0</tag><item>Non-break space</item>
      <tag>U+1680</tag><item>Ogham space mark</item>
      <tag>U+180E</tag><item>Mongolian vowel separator</item>
      <tag>U+2000</tag><item>En quad</item>
      <tag>U+2001</tag><item>Em quad</item>
      <tag>U+2002</tag><item>En space</item>
      <tag>U+2003</tag><item>Em space</item>
      <tag>U+2004</tag><item>Three-per-em space</item>
      <tag>U+2005</tag><item>Four-per-em space</item>
      <tag>U+2006</tag><item>Six-per-em space</item>
      <tag>U+2007</tag><item>Figure space</item>
      <tag>U+2008</tag><item>Punctuation space</item>
      <tag>U+2009</tag><item>Thin space</item>
      <tag>U+200A</tag><item>Hair space</item>
      <tag>U+202F</tag><item>Narrow no-break space</item>
      <tag>U+205F</tag><item>Medium mathematical space</item>
      <tag>U+3000</tag><item>Ideographic space</item>
    </taglist>

    <p>The following are the vertical space characters:</p>

    <taglist>
      <tag>U+000A</tag><item>Line feed (LF)</item>
      <tag>U+000B</tag><item>Vertical tab (VT)</item>
      <tag>U+000C</tag><item>Form feed (FF)</item>
      <tag>U+000D</tag><item>Carriage return (CR)</item>
      <tag>U+0085</tag><item>Next line (NEL)</item>
      <tag>U+2028</tag><item>Line separator</item>
      <tag>U+2029</tag><item>Paragraph separator</item>
    </taglist>

    <p>In 8-bit, non-UTF-8 mode, only the characters with code points &lt; 256
      are relevant.</p>

    <p><em>Newline Sequences</em></p>
    <marker id="newline_sequences"></marker>

    <p>Outside a character class, by default, the escape sequence \R matches any
      Unicode newline sequence. In non-UTF-8 mode, \R is equivalent to the
      following:</p>

    <code>
(?&gt;\r\n|\n|\x0b|\f|\r|\x85)</code>

    <p>This is an example of an "atomic group", details are provided below.</p>

    <p>This particular group matches either the two-character sequence CR
      followed by LF, or one of the single characters LF (line feed, U+000A),
      VT (vertical tab, U+000B), FF (form feed, U+000C), CR (carriage return,
      U+000D), or NEL (next line, U+0085). The two-character sequence is
      treated as a single unit that cannot be split.</p>

    <p>In Unicode mode, two more characters whose code points are &gt; 255 are
      added: LS (line separator, U+2028) and PS (paragraph separator, U+2029).
      Unicode character property support is not needed for these characters to
      be recognized.</p>

    <p>\R can be restricted to match only CR, LF, or CRLF (instead of the
      complete set of Unicode line endings) by setting option <c>bsr_anycrlf</c>
      either at compile time or when the pattern is matched. (BSR is an acronym
      for "backslash R".) This can be made the default when PCRE is built; if
      so, the other behavior can be requested through option
      <c>bsr_unicode</c>. These settings can also be specified by starting a
      pattern string with one of the following sequences:</p>

    <taglist>
      <tag>(*BSR_ANYCRLF)</tag>
      <item>CR, LF, or CRLF only</item>
      <tag>(*BSR_UNICODE)</tag>
      <item>Any Unicode newline sequence</item>
    </taglist>

    <p>These override the default and the options specified to the compiling
      function, but they can themselves be overridden by options specified to a
      matching function. Notice that these special settings, which are not
      Perl-compatible, are recognized only at the very start of a pattern, and
      that they must be in upper case. If more than one of them is present, the
      last one is used. They can be combined with a change of newline
      convention; for example, a pattern can start with:</p>

    <code>
(*ANY)(*BSR_ANYCRLF)</code>

    <p>They can also be combined with the (*UTF8), (*UTF), or (*UCP) special
      sequences. Inside a character class, \R is treated as an unrecognized
      escape sequence, and so matches the letter "R" by default.</p>

    <p><em>Unicode Character Properties</em></p>

    <p>Three more escape sequences that match characters with specific
      properties are available. When in 8-bit non-UTF-8 mode, these sequences
      are limited to testing characters whose code points are &lt;
      256, but they do work in this mode. The following are the extra escape
      sequences:</p>

    <taglist>
      <tag>\p{<em>xx</em>}</tag>
      <item>A character with property <em>xx</em></item>
      <tag>\P{<em>xx</em>}</tag>
      <item>A character without property <em>xx</em></item>
      <tag>\X</tag>
      <item>A Unicode extended grapheme cluster</item>
    </taglist>

    <p>The property names represented by <em>xx</em> above are limited to the
      Unicode script names, the general category properties, "Any", which
      matches any character (including newline), and some special PCRE
      properties (described in the next section). Other Perl properties, such as
      "InMusicalSymbols", are currently not supported by PCRE. Notice that
      \P{Any} does not match any characters and always causes a match
      failure.</p>

    <p>Sets of Unicode characters are defined as belonging to certain scripts.
      A character from one of these sets can be matched using a script name, for
      example:</p>

    <code>
\p{Greek} \P{Han}</code>

    <p>Those that are not part of an identified script are lumped together as
      "Common". The following is the current list of scripts:</p>

    <list type="bulleted">
      <item>Arabic</item>
      <item>Armenian</item>
      <item>Avestan</item>
      <item>Balinese</item>
      <item>Bamum</item>
      <item>Batak</item>
      <item>Bengali</item>
      <item>Bopomofo</item>
      <item>Braille</item>
      <item>Buginese</item>
      <item>Buhid</item>
      <item>Canadian_Aboriginal</item>
      <item>Carian</item>
      <item>Chakma</item>
      <item>Cham</item>
      <item>Cherokee</item>
      <item>Common</item>
      <item>Coptic</item>
      <item>Cuneiform</item>
      <item>Cypriot</item>
      <item>Cyrillic</item>
      <item>Deseret</item>
      <item>Devanagari</item>
      <item>Egyptian_Hieroglyphs</item>
      <item>Ethiopic</item>
      <item>Georgian</item>
      <item>Glagolitic</item>
      <item>Gothic</item>
      <item>Greek</item>
      <item>Gujarati</item>
      <item>Gurmukhi</item>
      <item>Han</item>
      <item>Hangul</item>
      <item>Hanunoo</item>
      <item>Hebrew</item>
      <item>Hiragana</item>
      <item>Imperial_Aramaic</item>
      <item>Inherited</item>
      <item>Inscriptional_Pahlavi</item>
      <item>Inscriptional_Parthian</item>
      <item>Javanese</item>
      <item>Kaithi</item>
      <item>Kannada</item>
      <item>Katakana</item>
      <item>Kayah_Li</item>
      <item>Kharoshthi</item>
      <item>Khmer</item>
      <item>Lao</item>
      <item>Latin</item>
      <item>Lepcha</item>
      <item>Limbu</item>
      <item>Linear_B</item>
      <item>Lisu</item>
      <item>Lycian</item>
      <item>Lydian</item>
      <item>Malayalam</item>
      <item>Mandaic</item>
      <item>Meetei_Mayek</item>
      <item>Meroitic_Cursive</item>
      <item>Meroitic_Hieroglyphs</item>
      <item>Miao</item>
      <item>Mongolian</item>
      <item>Myanmar</item>
      <item>New_Tai_Lue</item>
      <item>Nko</item>
      <item>Ogham</item>
      <item>Old_Italic</item>
      <item>Old_Persian</item>
      <item>Oriya</item>
      <item>Old_South_Arabian</item>
      <item>Old_Turkic</item>
      <item>Ol_Chiki</item>
      <item>Osmanya</item>
      <item>Phags_Pa</item>
      <item>Phoenician</item>
      <item>Rejang</item>
      <item>Runic</item>
      <item>Samaritan</item>
      <item>Saurashtra</item>
      <item>Sharada</item>
      <item>Shavian</item>
      <item>Sinhala</item>
      <item>Sora_Sompeng</item>
      <item>Sundanese</item>
      <item>Syloti_Nagri</item>
      <item>Syriac</item>
      <item>Tagalog</item>
      <item>Tagbanwa</item>
      <item>Tai_Le</item>
      <item>Tai_Tham</item>
      <item>Tai_Viet</item>
      <item>Takri</item>
      <item>Tamil</item>
      <item>Telugu</item>
      <item>Thaana</item>
      <item>Thai</item>
      <item>Tibetan</item>
      <item>Tifinagh</item>
      <item>Ugaritic</item>
      <item>Vai</item>
      <item>Yi</item>
    </list>

    <p>Each character has exactly one Unicode general category property,
      specified by a two-letter acronym. For compatibility with Perl, negation
      can be specified by including a circumflex between the opening brace and
      the property name. For example, \p{^Lu} is the same as \P{Lu}.</p>

    <p>If only one letter is specified with \p or \P, it includes all the
      general category properties that start with that letter. In this case, in
      the absence of negation, the curly brackets in the escape sequence are
      optional. The following two examples have the same effect:</p>

    <code>
\p{L}
\pL</code>

    <p>The following general category property codes are supported:</p>

    <taglist>  
      <tag>C</tag><item>Other</item>
      <tag>Cc</tag><item>Control</item>
      <tag>Cf</tag><item>Format</item>
      <tag>Cn</tag><item>Unassigned</item>
      <tag>Co</tag><item>Private use</item>
      <tag>Cs</tag><item>Surrogate</item> 
      <tag>L</tag><item>Letter</item>
      <tag>Ll</tag><item>Lowercase letter</item>
      <tag>Lm</tag><item>Modifier letter</item>
      <tag>Lo</tag><item>Other letter</item>
      <tag>Lt</tag><item>Title case letter</item>
      <tag>Lu</tag><item>Uppercase letter</item>  
      <tag>M</tag><item>Mark</item>
      <tag>Mc</tag><item>Spacing mark</item>
      <tag>Me</tag><item>Enclosing mark</item>
      <tag>Mn</tag><item>Non-spacing mark</item>  
      <tag>N</tag><item>Number</item>
      <tag>Nd</tag><item>Decimal number</item>
      <tag>Nl</tag><item>Letter number</item>
      <tag>No</tag><item>Other number</item>  
      <tag>P</tag><item>Punctuation</item>
      <tag>Pc</tag><item>Connector punctuation</item>
      <tag>Pd</tag><item>Dash punctuation</item>
      <tag>Pe</tag><item>Close punctuation</item>
      <tag>Pf</tag><item>Final punctuation</item>
      <tag>Pi</tag><item>Initial punctuation</item>
      <tag>Po</tag><item>Other punctuation</item>
      <tag>Ps</tag><item>Open punctuation</item>
      <tag>S</tag><item>Symbol</item>
      <tag>Sc</tag><item>Currency symbol</item>
      <tag>Sk</tag><item>Modifier symbol</item>
      <tag>Sm</tag><item>Mathematical symbol</item>
      <tag>So</tag><item>Other symbol</item> 
      <tag>Z</tag><item>Separator</item>
      <tag>Zl</tag><item>Line separator</item>
      <tag>Zp</tag><item>Paragraph separator</item>
      <tag>Zs</tag><item>Space separator</item>
    </taglist>

    <p>The special property L&amp; is also supported. It matches a character
      that has the Lu, Ll, or Lt property, that is, a letter that is not
      classified as a modifier or "other".</p>

    <p>The Cs (Surrogate) property applies only to characters in the range
      U+D800 to U+DFFF. Such characters are invalid in Unicode strings and so
      cannot be tested by PCRE. Perl does not support the Cs property.</p>

    <p>The long synonyms for property names supported by Perl (such as
      \p{Letter}) are not supported by PCRE. It is not permitted to prefix any
      of these properties with "Is".</p>

    <p>No character in the Unicode table has the Cn (unassigned) property.
      This property is instead assumed for any code point that is not in the
      Unicode table.</p>

    <p>Specifying caseless matching does not affect these escape sequences. For
      example, \p{Lu} always matches only uppercase letters. This is different
      from the behavior of current versions of Perl.</p>

    <p>Matching characters by Unicode property is not fast, as PCRE must do a
      multistage table lookup to find a character property. That is why the
      traditional escape sequences such as \d and \w do not use Unicode
      properties in PCRE by default. However, you can make them do so by setting
      option <c>ucp</c> or by starting the pattern with (*UCP).</p>

    <p><em>Extended Grapheme Clusters</em></p>

    <p>The \X escape matches any number of Unicode characters that form an
      "extended grapheme cluster", and treats the sequence as an atomic group
      (see below). Up to and including release 8.31, PCRE matched an earlier,
      simpler definition that was equivalent to <c>(?&gt;\PM\pM*)</c>. That is,
      it matched a character without the "mark" property, followed by zero or
      more characters with the "mark" property. Characters with the "mark"
      property are typically non-spacing accents that affect the preceding
      character.</p>

    <p>This simple definition was extended in Unicode to include more
      complicated kinds of composite character by giving each character a
      grapheme breaking property, and creating rules that use these properties
      to define the boundaries of extended grapheme clusters. In PCRE releases
      later than 8.31, \X matches one of these clusters.</p>

    <p>\X always matches at least one character. Then it decides whether to add
      more characters according to the following rules for ending a cluster:</p>

    <list type="ordered">
      <item>
        <p>End at the end of the subject string.</p>
      </item>
      <item>
        <p>Do not end between CR and LF; otherwise end after any control
          character.</p>
      </item>
      <item>
        <p>Do not break Hangul (a Korean script) syllable sequences. Hangul
          characters are of five types: L, V, T, LV, and LVT. An L character can
          be followed by an L, V, LV, or LVT character. An LV or V character can
          be followed by a V or T character. An LVT or T character can be
          followed only by a T character.</p>
      </item>
      <item>
        <p>Do not end before extending characters or spacing marks. Characters
          with the "mark" property always have the "extend" grapheme breaking
          property.</p>
      </item>
      <item>
        <p>Do not end after prepend characters.</p>
      </item>
      <item>
        <p>Otherwise, end the cluster.</p>
      </item>
    </list>

    <p><em>PCRE Additional Properties</em></p>

    <p>In addition to the standard Unicode properties described earlier, PCRE
      supports four more that make it possible to convert traditional escape
      sequences, such as \w and \s, and Posix character classes to use Unicode
      properties. PCRE uses these non-standard, non-Perl properties internally
      when <c>PCRE_UCP</c> is set. However, they can also be used explicitly.
      The properties are as follows:</p>

    <taglist>
      <tag>Xan</tag>
      <item>
        <p>Any alphanumeric character. Matches characters that have either the
          L (letter) or the N (number) property.</p>
      </item>
      <tag>Xps</tag>
      <item>
        <p>Any Posix space character. Matches the characters tab, line feed,
          vertical tab, form feed, carriage return, and any other character
          that has the Z (separator) property.</p>
      </item>
      <tag>Xsp</tag>
      <item>
        <p>Any Perl space character. Matches the same as Xps, except that
          vertical tab is excluded.</p>
      </item>
      <tag>Xwd</tag>
      <item>
        <p>Any Perl "word" character. Matches the same characters as Xan, plus
          underscore.</p>
      </item>
    </taglist>

    <p>There is another non-standard property, Xuc, which matches any character
      that can be represented by a Universal Character Name in C++ and other
      programming languages. These are the characters $, @, ` (grave accent),
      and all characters with Unicode code points &gt;= U+00A0, except for the
      surrogates U+D800 to U+DFFF. Notice that most base (ASCII) characters are
      excluded. (Universal Character Names are of the form \uHHHH or \UHHHHHHHH,
      where H is a hexadecimal digit. Notice that the Xuc property does not
      match these sequences but the characters that they represent.)</p>

    <p><em>Resetting the Match Start</em></p>

    <p>The escape sequence \K causes any previously matched characters not to
      be included in the final matched sequence. For example, the following
      pattern matches "foobar", but reports that it has matched "bar":</p> 

    <code>
foo\Kbar</code>

    <p>This feature is similar to a lookbehind assertion
      <!--  HTML &lt;a href="#lookbehind"&gt; -->
      <!--  &lt;/a&gt; -->
      (described below). However, in this case, the part of the subject before
      the real match does not have to be of fixed length, as lookbehind
      assertions do. The use of \K does not interfere with the setting of
      captured substrings. For example, when the following pattern matches
      "foobar", the first substring is still set to "foo":</p>

<code>
(foo)\Kbar</code>

    <p>Perl documents that the use of \K within assertions is "not well
      defined". In PCRE, \K is acted upon when it occurs inside positive
      assertions, but is ignored in negative assertions.</p>

    <p><em>Simple Assertions</em></p>

    <p>The final use of backslash is for certain simple assertions. An
      assertion specifies a condition that must be met at a particular point in
      a match, without consuming any characters from the subject string. The
      use of subpatterns for more complicated assertions is described below. The
      following are the backslashed assertions:</p>

    <taglist>
      <tag>\b</tag><item>Matches at a word boundary.</item>
      <tag>\B</tag><item>Matches when not at a word boundary.</item>
      <tag>\A</tag><item>Matches at the start of the subject.</item>
      <tag>\Z</tag><item>Matches at the end of the subject, and before a newline
        at the end of the subject.</item>
      <tag>\z</tag><item>Matches only at the end of the subject.</item>
      <tag>\G</tag><item>Matches at the first matching position in the subject.
        </item>
    </taglist>

    <p>Inside a character class, \b has a different meaning; it matches the
      backspace character. If any other of these assertions appears in a
      character class, by default it matches the corresponding literal character
      (for example, \B matches the letter B).</p>

    <p>A word boundary is a position in the subject string where the current
      character and the previous character do not both match \w or \W (that is,
      one matches \w and the other matches \W), or the start or end of the
      string if the first or last character matches \w, respectively. In UTF
      mode, the meanings of \w and \W can be changed by setting option
      <c>ucp</c>. When this is done, it also affects \b and \B. PCRE and Perl do
      not have a separate "start of word" or "end of word" metasequence.
      However, whatever follows \b normally determines which it is. For example,
      the fragment \ba matches "a" at the start of a word.</p>

    <p>The \A, \Z, and \z assertions differ from the traditional circumflex and
      dollar (described in the next section) in that they only ever match at the
      very start and end of the subject string, whatever options are set. Thus,
      they are independent of multiline mode. These three assertions are not
      affected by options <c>notbol</c> or <c>noteol</c>, which affect only the
      behavior of the circumflex and dollar metacharacters. However, if argument
      <c>startoffset</c> of <seealso marker="#run/3"><c>run/3</c></seealso> is
      non-zero, indicating that matching is to start at a point other than the
      beginning of the subject, \A can never match. The difference between \Z
      and \z is that \Z matches before a newline at the end of the string and
      at the very end, while \z matches only at the end.</p>

    <p>The \G assertion is true only when the current matching position is at
      the start point of the match, as specified by argument <c>startoffset</c>
      of <c>run/3</c>. It differs from \A when the value of <c>startoffset</c>
      is non-zero. By calling <c>run/3</c> multiple times with appropriate
      arguments, you can mimic the Perl option <c>/g</c>, and it is in this
      kind of implementation where \G can be useful.</p>

    <p>Notice, however, that the PCRE interpretation of \G, as the start of the
      current match, is subtly different from Perl, which defines it as the end
      of the previous match. In Perl, these can be different when the previously
      matched string was empty. As PCRE does only one match at a time, it cannot
      reproduce this behavior.</p>

    <p>If all the alternatives of a pattern begin with \G, the expression is
      anchored to the starting match position, and the "anchored" flag is set in
      the compiled regular expression.</p>
  </section>

  <section>
    <marker id="sect4"></marker>
    <title>Circumflex and Dollar</title>
    <p>The circumflex and dollar metacharacters are zero-width assertions. That
      is, they test for a particular condition to be true without consuming any
      characters from the subject string.</p>

    <p>Outside a character class, in the default matching mode, the circumflex
      character is an assertion that is true only if the current matching point
      is at the start of the subject string. If argument <c>startoffset</c> of
      <seealso marker="#run/3"><c>run/3</c></seealso> is non-zero, circumflex
      can never match if option <c>multiline</c> is unset. Inside a character
      class, circumflex has an entirely different meaning (see below).</p>

    <p>Circumflex needs not to be the first character of the pattern if
      some alternatives are involved, but it is to be the first thing in
      each alternative in which it appears if the pattern is ever to match that
      branch. If all possible alternatives start with a circumflex, that is, if
      the pattern is constrained to match only at the start of the subject, it
      is said to be an "anchored" pattern. (There are also other constructs that
      can cause a pattern to be anchored.)</p>

    <p>The dollar character is an assertion that is true only if the current
      matching point is at the end of the subject string, or immediately before
      a newline at the end of the string (by default). Notice however that it
      does not match the newline. Dollar needs not to be the last character of
      the pattern if some alternatives are involved, but it is to be the
      last item in any branch in which it appears. Dollar has no special meaning
      in a character class.</p>

    <p>The meaning of dollar can be changed so that it matches only at the very
      end of the string, by setting option <c>dollar_endonly</c> at compile
      time. This does not affect the \Z assertion.</p>

    <p>The meanings of the circumflex and dollar characters are changed if
      option <c>multiline</c> is set. When this is the case, a circumflex
      matches immediately after internal newlines and at the start of the
      subject string. It does not match after a newline that ends the string. A
      dollar matches before any newlines in the string, and at the very end,
      when <c>multiline</c> is set. When newline is specified as the
      two-character sequence CRLF, isolated CR and LF characters do not
      indicate newlines.</p>

    <p>For example, the pattern /^abc$/ matches the subject string "def\nabc"
      (where \n represents a newline) in multiline mode, but not otherwise.
      So, patterns that are anchored in single-line mode because all
      branches start with ^ are not anchored in multiline mode, and a match for
      circumflex is possible when argument <em>startoffset</em> of <c>run/3</c>
      is non-zero. Option <c>dollar_endonly</c> is ignored if <c>multiline</c>
      is set.</p>

    <p>Notice that the sequences \A, \Z, and \z can be used to match the start
      and end of the subject in both modes. If all branches of a pattern start
      with \A, it is always anchored, regardless if <c>multiline</c> is set.</p>
  </section>

  <section>
    <marker id="sect5"></marker>
    <title>Full Stop (Period, Dot) and \N</title>
    <p>Outside a character class, a dot in the pattern matches any character in
      the subject string except (by default) a character that signifies the end
      of a line.</p>

    <p>When a line ending is defined as a single character, dot never matches
      that character. When the two-character sequence CRLF is used, dot does not
      match CR if it is immediately followed by LF, otherwise it matches all
      characters (including isolated CRs and LFs). When any Unicode line endings
      are recognized, dot does not match CR, LF, or any of the other
      line-ending characters.</p>

    <p>The behavior of dot regarding newlines can be changed.  If option
      <c>dotall</c> is set, a dot matches any character, without exception. If
      the two-character sequence CRLF is present in the subject string, it takes
      two dots to match it.</p>

    <p>The handling of dot is entirely independent of the handling of circumflex
      and dollar, the only relationship is that both involve newlines. Dot has
      no special meaning in a character class.</p>

    <p>The escape sequence \N behaves like a dot, except that it is not affected
      by option <c>PCRE_DOTALL</c>. That is, it matches any character except one
      that signifies the end of a line. Perl also uses \N to match characters by
      name but PCRE does not support this.</p>
  </section>

  <section>
    <marker id="sect6"></marker>
    <title>Matching a Single Data Unit</title>
    <p>Outside a character class, the escape sequence \C matches any data unit,
      regardless if a UTF mode is set. One data unit is one byte. Unlike a dot,
      \C always matches line-ending characters. The feature is provided in Perl
      to match individual bytes in UTF-8 mode, but it is unclear how it can
      usefully be used. As \C breaks up characters into individual data units,
      matching one unit with \C in a UTF mode means that the remaining string
      can start with a malformed UTF character. This has undefined results, as
      PCRE assumes that it deals with valid UTF strings.</p>

    <p>PCRE does not allow \C to appear in lookbehind assertions (described
      below) in a UTF mode, as this would make it impossible to calculate the
     length of the lookbehind.</p>

    <p>The \C escape sequence is best avoided. However, one way of using it that
      avoids the problem of malformed UTF characters is to use a lookahead to
      check the length of the next character, as in the following pattern, which
      can be used with a UTF-8 string (ignore whitespace and line breaks):</p>

<code type="none">
(?| (?=[\x00-\x7f])(\C) |
    (?=[\x80-\x{7ff}])(\C)(\C) |
    (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
    (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))</code>

    <p>A group that starts with (?| resets the capturing parentheses numbers in
      each alternative (see section <seealso marker="#sect12">Duplicate
      Subpattern Numbers</seealso>). The assertions at the start of each branch
      check the next UTF-8 character for values whose encoding uses 1, 2, 3, or
      4 bytes, respectively. The individual bytes of the character are then
      captured by the appropriate number of groups.</p>
  </section>

  <section>
    <marker id="sect7"></marker>
    <title>Square Brackets and Character Classes</title>
    <p>An opening square bracket introduces a character class, terminated by a
      closing square bracket. A closing square bracket on its own is not special
      by default. However, if option <c>PCRE_JAVASCRIPT_COMPAT</c> is set, a
      lone closing square bracket causes a compile-time error. If a closing
      square bracket is required as a member of the class, it is to be the first
      data character in the class (after an initial circumflex, if present) or
      escaped with a backslash.</p>

    <p>A character class matches a single character in the subject. In a UTF
      mode, the character can be more than one data unit long. A matched
      character must be in the set of characters defined by the class, unless
      the first character in the class definition is a circumflex, in which case
      the subject character must not be in the set defined by the class. If a
      circumflex is required as a member of the class, ensure that it is not the
      first character, or escape it with a backslash.</p>

    <p>For example, the character class <c>[aeiou]</c> matches any lowercase
      vowel, while <c>[^aeiou]</c> matches any character that is not a lowercase
      vowel. Notice that a circumflex is just a convenient notation for
      specifying the characters that are in the class by enumerating those that
      are not. A class that starts with a circumflex is not an assertion; it
      still consumes a character from the subject string, and therefore it fails
      if the current pointer is at the end of the string.</p>

    <p>In UTF-8 mode, characters with values &gt; 255 (0xffff) can be included
      in a class as a literal string of data units, or by using the \x{ escaping
      mechanism.</p>

    <p>When caseless matching is set, any letters in a class represent both
      their uppercase and lowercase versions. For example, a caseless
      <c>[aeiou]</c> matches "A" and "a", and a caseless <c>[^aeiou]</c> does
      not match "A", but a caseful version would. In a UTF mode, PCRE always
      understands the concept of case for characters whose values are &lt; 256,
      so caseless matching is always possible. For characters with higher
      values, the concept of case is supported only if PCRE is compiled with
      Unicode property support. If you want to use caseless matching in a UTF
      mode for characters &gt;=, ensure that PCRE is compiled with Unicode
      property support and with UTF support.</p>

    <p>Characters that can indicate line breaks are never treated in any special
      way when matching character classes, whatever line-ending sequence is in
      use, and whatever setting of options <c>PCRE_DOTALL</c> and
      <c>PCRE_MULTILINE</c> is used. A class such as [^a] always matches one of
      these characters.</p>

    <p>The minus (hyphen) character can be used to specify a range of characters
      in a character class. For example, [d-m] matches any letter between d and
      m, inclusive. If a minus character is required in a class, it must be
      escaped with a backslash or appear in a position where it cannot be
      interpreted as indicating a range, typically as the first or last
      character in the class.</p>

    <p>The literal character "]" cannot be the end character of a range. A
      pattern such as [W-]46] is interpreted as a class of two characters ("W"
      and "-") followed by a literal string "46]", so it would match "W46]" or
      "-46]". However, if "]" is escaped with a backslash, it is interpreted as
      the end of range, so [W-\]46] is interpreted as a class containing a range
      followed by two other characters. The octal or hexadecimal representation
      of "]" can also be used to end a range.</p>

    <p>Ranges operate in the collating sequence of character values. They can
      also be used for characters specified numerically, for example,
      [\000-\037]. Ranges can include any characters that are valid for the
      current mode.</p>

    <p>If a range that includes letters is used when caseless matching is set,
      it matches the letters in either case. For example, [W-c] is equivalent to
      [][\\^_`wxyzabc], matched caselessly. In a non-UTF mode, if character
      tables for a French locale are in use, [\xc8-\xcb] matches accented E
      characters in both cases. In UTF modes, PCRE supports the concept of case
      for characters with values &gt; 255 only when it is compiled with Unicode
      property support.</p>

    <p>The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, \V,
      \w, and \W can appear in a character class, and add the characters that
      they match to the class. For example, [\dABCDEF] matches any hexadecimal
      digit. In UTF modes, option <c>ucp</c> affects the meanings of \d, \s, \w
      and their uppercase partners, just as it does when they appear outside a
      character class, as described in section
      <seealso marker="#generic_character_types">Generic Character
      Types</seealso> earlier. The escape sequence \b has a different meaning
      inside a character class; it matches the backspace character. The
      sequences \B, \N, \R, and \X are not special inside a character class.
      Like any other unrecognized escape sequences, they are treated as the
      literal characters "B", "N", "R", and "X".</p>

    <p>A circumflex can conveniently be used with the uppercase character types
      to specify a more restricted set of characters than the matching lowercase
      type. For example, class [^\W_] matches any letter or digit, but not
      underscore, while [\w] includes underscore. A positive character class
      is to be read as "something OR something OR ..." and a negative class as
      "NOT something AND NOT something AND NOT ...".</p>

    <p>Only the following metacharacters are recognized in character
      classes:</p>

    <list type="bulleted">
      <item>Backslash</item>
      <item>Hyphen (only where it can be interpreted as specifying a
        range)</item>
      <item>Circumflex (only at the start)</item>
      <item>Opening square bracket (only when it can be interpreted as
        introducing a Posix class name; see the next section)</item>
      <item>Terminating closing square bracket</item>
    </list>

    <p>However, escaping other non-alphanumeric characters does no harm.</p>
  </section>

  <section>
    <marker id="sect8"></marker>
    <title>Posix Character Classes</title>
    <p>Perl supports the Posix notation for character classes. This uses names
      enclosed by [: and :] within the enclosing square brackets. PCRE also
      supports this notation. For example, the following matches "0", "1", any
      alphabetic character, or "%":</p>

    <code>
[01[:alpha:]%]</code>

    <p>The following are the supported class names:</p>

    <taglist>  
      <tag>alnum</tag><item>Letters and digits</item>
      <tag>alpha</tag><item>Letters</item>
      <tag>ascii</tag><item>Character codes 0-127</item>
      <tag>blank</tag><item>Space or tab only</item>
      <tag>cntrl</tag><item>Control characters</item>
      <tag>digit</tag><item>Decimal digits (same as \d)</item>
      <tag>graph</tag><item>Printing characters, excluding space</item>
      <tag>lower</tag><item>Lowercase letters</item>
      <tag>print</tag><item>Printing characters, including space</item>
      <tag>punct</tag><item>Printing characters, excluding letters, digits, and
        space</item>
      <tag>space</tag><item>Whitespace (not quite the same as \s)</item>
      <tag>upper</tag><item>Uppercase letters</item>
      <tag>word</tag><item>"Word" characters (same as \w)</item>
      <tag>xdigit</tag><item>Hexadecimal digits</item>
    </taglist>

    <p>The "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13),
      and space (32). Notice that this list includes the VT character (code 11).
      This makes "space" different to \s, which does not include VT (for Perl
      compatibility).</p>

    <p>The name "word" is a Perl extension, and "blank" is a GNU extension from
      Perl 5.8. Another Perl extension is negation, which is indicated by a ^
      character after the colon. For example, the following matches "1", "2",
      or any non-digit:</p>

    <code>
[12[:^digit:]]</code>

    <p>PCRE (and Perl) also recognize the Posix syntax [.ch.] and [=ch=] where
      "ch" is a "collating element", but these are not supported, and an error
      is given if they are encountered.</p>

    <p>By default, in UTF modes, characters with values &gt; 255 do not match
      any of the Posix character classes. However, if option <c>PCRE_UCP</c> is
      passed to <c>pcre_compile()</c>, some of the classes are changed so that
      Unicode character properties are used. This is achieved by replacing the
      Posix classes by other sequences, as follows:</p>

    <taglist>
      <tag>[:alnum:]</tag><item>Becomes <em>\p{Xan}</em></item>
      <tag>[:alpha:]</tag><item>Becomes <em>\p{L}</em></item>
      <tag>[:blank:]</tag><item>Becomes <em>\h</em></item>
      <tag>[:digit:]</tag><item>Becomes <em>\p{Nd}</em></item>
      <tag>[:lower:]</tag><item>Becomes <em>\p{Ll}</em></item>
      <tag>[:space:]</tag><item>Becomes <em>\p{Xps}</em></item>
      <tag>[:upper:]</tag><item>Becomes <em>\p{Lu}</em></item>
      <tag>[:word:]</tag><item>Becomes <em>\p{Xwd}</em></item>
    </taglist>

    <p>Negated versions, such as [:^alpha:], use \P instead of \p. The other
      Posix classes are unchanged, and match only characters with code points
      &lt; 256.</p>
  </section>

  <section>
    <marker id="sect9"></marker>
    <title>Vertical Bar</title>
    <p>Vertical bar characters are used to separate alternative patterns. For
      example, the following pattern matches either "gilbert" or "sullivan":</p>

    <code>
gilbert|sullivan</code>

    <p>Any number of alternatives can appear, and an empty alternative is
      permitted (matching the empty string). The matching process tries each
      alternative in turn, from left to right, and the first that succeeds is
      used. If the alternatives are within a subpattern (defined in section
      <seealso marker="#sect11">Subpatterns</seealso>), "succeeds" means
      matching the remaining main pattern and the alternative in the
      subpattern.</p>
  </section>

  <section>
    <marker id="sect10"></marker>
    <title>Internal Option Setting</title>
    <p>The settings of the Perl-compatible options <c>caseless</c>,
      <c>multiline</c>, <c>dotall</c>, and <c>extended</c> can be changed from
      within the pattern by a sequence of Perl option letters enclosed between
      "(?" and ")". The option letters are as follows:</p>

    <taglist>
      <tag>i</tag><item>For <c>caseless</c></item>
      <tag>m</tag><item>For <c>multiline</c></item>
      <tag>s</tag><item>For <c>dotall</c></item>
      <tag>x</tag><item>For <c>extended</c></item>
    </taglist>

    <p>For example, <c>(?im)</c> sets caseless, multiline matching. These
      options can also be unset by preceding the letter with a hyphen. A
      combined setting and unsetting such as <c>(?im-sx)</c>, which sets
      <c>caseless</c> and <c>multiline</c>, while unsetting <c>dotall</c> and
      <c>extended</c>, is also permitted. If a letter appears both before and
      after the hyphen, the option is unset.</p>

    <p>The PCRE-specific options <c>dupnames</c>, <c>ungreedy</c>, and
      <c>extra</c> can be changed in the same way as the Perl-compatible
      options by using the characters J, U, and X respectively.</p>

    <p>When one of these option changes occurs at top-level (that is, not inside
      subpattern parentheses), the change applies to the remainder of the
      pattern that follows. If the change is placed right at the start of a
      pattern, PCRE extracts it into the global options.</p>
    <p>An option change within a subpattern (see section
      <seealso marker="#sect11">Subpatterns</seealso>) affects only that part of
      the subpattern that follows it. So, the following matches abc and aBc and
      no other strings (assuming <c>caseless</c> is not used):</p>

    <code>
(a(?i)b)c</code>

    <p>By this means, options can be made to have different settings in
      different parts of the pattern. Any changes made in one alternative do
      carry on into subsequent branches within the same subpattern. For
      example:</p>

    <code>
(a(?i)b|c)</code>

    <p>matches "ab", "aB", "c", and "C", although when matching "C" the first
      branch is abandoned before the option setting. This is because the effects
      of option settings occur at compile time. There would be some weird
      behavior otherwise.</p>

    <note>
      <p>Other PCRE-specific options can be set by the application when the
        compiling or matching functions are called. Sometimes the pattern can
        contain special leading sequences, such as (*CRLF), to override what
        the application has set or what has been defaulted. Details are provided
        in section <seealso marker="#newline_sequences">
        Newline Sequences</seealso> earlier.</p>
      <p>The (*UTF8) and (*UCP) leading sequences can be used to set UTF and
        Unicode property modes. They are equivalent to setting options
        <c>unicode</c> and <c>ucp</c>, respectively. The (*UTF) sequence is a
        generic version that can be used with any of the libraries. However,
        the application can set option <c>never_utf</c>, which locks out the
        use of the (*UTF) sequences.</p>
    </note>
  </section>

  <section>
    <marker id="sect11"></marker>
    <title>Subpatterns</title>
    <p>Subpatterns are delimited by parentheses (round brackets), which can be
      nested. Turning part of a pattern into a subpattern does two things:</p>

    <taglist>
      <tag>1.</tag>
      <item>
        <p>It localizes a set of alternatives. For example, the following
          pattern matches "cataract", "caterpillar", or "cat":</p>
        <code>
cat(aract|erpillar|)</code>
        <p>Without the parentheses, it would match "cataract", "erpillar", or an
          empty string.</p>
      </item>
      <tag>2.</tag>
      <item>
        <p>It sets up the subpattern as a capturing subpattern. That is, when
          the complete pattern matches, that portion of the subject string that
          matched the subpattern is passed back to the caller through the
          return value of <seealso marker="#run/3"><c>run/3</c></seealso>.</p>
      </item>
    </taglist>

    <p>Opening parentheses are counted from left to right (starting from 1) to
      obtain numbers for the capturing subpatterns. For example, if the string
      "the red king" is matched against the following pattern, the captured
      substrings are "red king", "red", and "king", and are numbered 1, 2, and
      3, respectively:</p>

    <code>
the ((red|white) (king|queen))</code>

    <p>It is not always helpful that plain parentheses fulfill two functions.
      Often a grouping subpattern is required without a capturing requirement.
      If an opening parenthesis is followed by a question mark and a colon, the
      subpattern does not do any capturing, and is not counted when computing
      the number of any subsequent capturing subpatterns. For example, if the
      string "the white queen" is matched against the following pattern, the
      captured substrings are "white queen" and "queen", and are numbered 1 and
      2:</p>

    <code>
the ((?:red|white) (king|queen))</code>

    <p>The maximum number of capturing subpatterns is 65535.</p>

    <p>As a convenient shorthand, if any option settings are required at the
      start of a non-capturing subpattern, the option letters can appear between
      "?" and ":". Thus, the following two patterns match the same set of
      strings:</p>

    <code>
(?i:saturday|sunday)
(?:(?i)saturday|sunday)</code>

    <p>As alternative branches are tried from left to right, and options are not
      reset until the end of the subpattern is reached, an option setting in one
      branch does affect subsequent branches, so the above patterns match both
      "SUNDAY" and "Saturday".</p>
  </section>

  <section>
    <marker id="sect12"></marker>
    <title>Duplicate Subpattern Numbers</title>
    <p>Perl 5.10 introduced a feature where each alternative in a subpattern
      uses the same numbers for its capturing parentheses. Such a subpattern
      starts with <c>(?|</c> and is itself a non-capturing subpattern. For
      example, consider the following pattern:</p>

    <code>
(?|(Sat)ur|(Sun))day</code>

    <p>As the two alternatives are inside a <c>(?|</c> group, both sets of
      capturing parentheses are numbered one. Thus, when the pattern matches,
      you can look at captured substring number one, whichever alternative
      matched. This construct is useful when you want to capture a part, but
      not all, of one of many alternatives. Inside a <c>(?|</c> group,
      parentheses are numbered as usual, but the number is reset at the start
      of each branch. The numbers of any capturing parentheses that follow the
      subpattern start after the highest number used in any branch.
      The following example is from the Perl documentation; the numbers
      underneath show in which buffer the captured content is stored:</p>

<code type="none">
# before  ---------------branch-reset----------- after
/ ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
# 1            2         2  3        2     3     4</code>

    <p>A back reference to a numbered subpattern uses the most recent value that
      is set for that number by any subpattern. The following pattern matches
      "abcabc" or "defdef":</p>

    <code>
/(?|(abc)|(def))\1/</code>

    <p>In contrast, a subroutine call to a numbered subpattern always refers to
      the first one in the pattern with the given number. The following pattern
      matches "abcabc" or "defabc":</p>

    <code>
/(?|(abc)|(def))(?1)/</code>

    <p>If a condition test for a subpattern having matched refers to a
      non-unique number, the test is true if any of the subpatterns of that
      number have matched.</p>

    <p>An alternative approach using this "branch reset" feature is to use
      duplicate named subpatterns, as described in the next section.</p>
  </section>

  <section>
    <marker id="sect13"></marker>
    <title>Named Subpatterns</title>
    <p>Identifying capturing parentheses by number is simple, but it can be
      hard to keep track of the numbers in complicated regular expressions.
      Also, if an expression is modified, the numbers can change. To help with
      this difficulty, PCRE supports the naming of subpatterns. This feature was
      not added to Perl until release 5.10. Python had the feature earlier, and
      PCRE introduced it at release 4.0, using the Python syntax. PCRE now
      supports both the Perl and the Python syntax. Perl allows identically
      numbered subpatterns to have different names, but PCRE does not.</p>

    <p>In PCRE, a subpattern can be named in one of three ways:
      <c>(?&lt;name&gt;...)</c> or <c>(?'name'...)</c> as in Perl, or
      <c>(?P&lt;name&gt;...)</c> as in Python. References to capturing
      parentheses from other parts of the pattern, such as back references,
      recursion, and conditions, can be made by name and by number.</p>

    <p>Names consist of up to 32 alphanumeric characters and underscores. Named
      capturing parentheses are still allocated numbers as well as names,
      exactly as if the names were not present.
      The <c>capture</c> specification to <seealso marker="#run/3">
      <c>run/3</c></seealso> can use named values if they are present in the
      regular expression.</p>

    <p>By default, a name must be unique within a pattern, but this constraint
      can be relaxed by setting option <c>dupnames</c> at compile time.
      (Duplicate names are also always permitted for subpatterns with the same
      number, set up as described in the previous section.) Duplicate names can
      be useful for patterns where only one instance of the named parentheses
      can match. Suppose that you want to match the name of a weekday, either as
      a 3-letter abbreviation or as the full name, and in both cases you want to
      extract the abbreviation. The following pattern (ignoring the line
      breaks) does the job:</p>

    <code type="none">
(?&lt;DN&gt;Mon|Fri|Sun)(?:day)?|
(?&lt;DN&gt;Tue)(?:sday)?|
(?&lt;DN&gt;Wed)(?:nesday)?|
(?&lt;DN&gt;Thu)(?:rsday)?|
(?&lt;DN&gt;Sat)(?:urday)?</code>

    <p>There are five capturing substrings, but only one is ever set after a
      match. (An alternative way of solving this problem is to use a "branch
      reset" subpattern, as described in the previous section.)</p>

    <p>For capturing named subpatterns which names are not unique, the first
      matching occurrence (counted from left to right in the subject) is
      returned from <seealso marker="#run/3"><c>run/3</c></seealso>, if the name
      is specified in the <c>values</c> part of the <c>capture</c> statement.
      The <c>all_names</c> capturing value matches all the names in the same
      way.</p>

    <note>
      <p>You cannot use different names to distinguish between two subpatterns
        with the same number, as PCRE uses only the numbers when matching. For
        this reason, an error is given at compile time if different names are
        specified to subpatterns with the same number. However, you can specify
        the same name to subpatterns with the same number, even when
        <c>dupnames</c> is not set.</p>
    </note>
  </section>

  <section>
    <marker id="sect14"></marker>
    <title>Repetition</title>
    <p>Repetition is specified by quantifiers, which can follow any of the
      following items:</p>

    <list type="bulleted">
      <item>A literal data character</item>
      <item>The dot metacharacter</item>
      <item>The \C escape sequence</item>
      <item>The \X escape sequence</item>
      <item>The \R escape sequence</item>
      <item>An escape such as \d or \pL that matches a single character</item>
      <item>A character class</item>
      <item>A back reference (see the next section)</item>
      <item>A parenthesized subpattern (including assertions)</item>
      <item>A subroutine call to a subpattern (recursive or otherwise)</item>
    </list>

    <p>The general repetition quantifier specifies a minimum and maximum number
      of permitted matches, by giving the two numbers in curly brackets
      (braces), separated by a comma. The numbers must be &lt; 65536, and the
      first must be less than or equal to the second. For example, the following
      matches "zz", "zzz", or "zzzz":</p>

    <code>
z{2,4}</code>

    <p>A closing brace on its own is not a special character. If the second
      number is omitted, but the comma is present, there is no upper limit. If
      the second number and the comma are both omitted, the quantifier specifies
      an exact number of required matches. Thus, the following matches at least
      three successive vowels, but can match many more:</p>

    <code>
[aeiou]{3,}</code>

    <p>The following matches exactly eight digits:</p>

    <code>
\d{8}</code>

    <p>An opening curly bracket that appears in a position where a quantifier is
      not allowed, or one that does not match the syntax of a quantifier, is
      taken as a literal character. For example, {,6} is not a quantifier, but a
      literal string of four characters.</p>

    <p>In Unicode mode, quantifiers apply to characters rather than to
      individual data units. Thus, for example, \x{100}{2} matches two
      characters, each of which is represented by a 2-byte sequence in a
      UTF-8 string. Similarly, \X{3} matches three Unicode extended grapheme
      clusters, each of which can be many data units long (and they can be of
      different lengths).</p>

    <p>The quantifier {0} is permitted, causing the expression to behave as if
      the previous item and the quantifier were not present. This can be useful
      for subpatterns that are referenced as subroutines from elsewhere in the
      pattern (but see also section <seealso marker="#defining_subpatterns">
      Defining Subpatterns for Use by Reference Only</seealso>). Items other
      than subpatterns that have a {0} quantifier are omitted from the compiled
      pattern.</p>

    <p>For convenience, the three most common quantifiers have single-character
      abbreviations:</p>

    <taglist>  
      <tag>*</tag><item>Equivalent to {0,}</item>
      <tag>+</tag><item>Equivalent to {1,}</item>
      <tag>?</tag><item>Equivalent to {0,1}</item>
    </taglist>

    <p>Infinite loops can be constructed by following a subpattern that can
      match no characters with a quantifier that has no upper limit, for
      example:</p>

    <code>
(a?)*</code>

    <p>Earlier versions of Perl and PCRE used to give an error at compile time
      for such patterns. However, as there are cases where this can be useful,
      such patterns are now accepted. However, if any repetition of the
      subpattern matches no characters, the loop is forcibly broken.</p>

    <p>By default, the quantifiers are "greedy", that is, they match as much as
      possible (up to the maximum number of permitted times), without causing
      the remaining pattern to fail. The classic example of where this gives
      problems is in trying to match comments in C programs. These appear
      between /* and */. Within the comment, individual * and / characters can
      appear. An attempt to match C comments by applying the pattern</p>

    <code>
/\*.*\*/</code>

    <p>to the string</p>

    <code>
/* first comment */  not comment  /* second comment */</code>

    <p>fails, as it matches the entire string owing to the greediness of the .* 
      item.</p>

    <p>However, if a quantifier is followed by a question mark, it ceases to be
      greedy, and instead matches the minimum number of times possible, so the
      following pattern does the right thing with the C comments:</p>

    <code>
/\*.*?\*/</code>

    <p>The meaning of the various quantifiers is not otherwise changed, only
      the preferred number of matches. Do not confuse this use of question mark
      with its use as a quantifier in its own right. As it has two uses, it can
      sometimes appear doubled, as in</p>

    <code>
\d??\d</code>

    <p>which matches one digit by preference, but can match two if that is the
      only way the remaining pattern matches.</p>

    <p>If option <c>ungreedy</c> is set (an option that is not available in
      Perl), the quantifiers are not greedy by default, but individual ones can
      be made greedy by following them with a question mark. That is, it inverts
      the default behavior.</p>

    <p>When a parenthesized subpattern is quantified with a minimum repeat count
      that is &gt; 1 or with a limited maximum, more memory is required for the
      compiled pattern, in proportion to the size of the minimum or maximum.</p>

    <p>If a pattern starts with .* or .{0,} and option <c>dotall</c> (equivalent
      to Perl option <c>/s</c>) is set, thus allowing the dot to match newlines,
      the pattern is implicitly anchored, because whatever follows is tried
      against every character position in the subject string. So, there is no
      point in retrying the overall match at any position after the first. PCRE
      normally treats such a pattern as if it was preceded by \A.</p>

    <p>In cases where it is known that the subject string contains no newlines,
      it is worth setting <c>dotall</c> to obtain this optimization, or
      alternatively using ^ to indicate anchoring explicitly.</p>

    <p>However, there are some cases where the optimization cannot be used. When
      .* is inside capturing parentheses that are the subject of a back
      reference elsewhere in the pattern, a match at the start can fail where a
      later one succeeds. Consider, for example:</p>

    <code>
(.*)abc\1</code>

    <p>If the subject is "xyz123abc123", the match point is the fourth
      character. Therefore, such a pattern is not implicitly anchored.</p>

    <p>Another case where implicit anchoring is not applied is when the leading
      .* is inside an atomic group. Once again, a match at the start can fail
      where a later one succeeds. Consider the following pattern:</p>

    <code>
(?&gt;.*?a)b</code>

    <p>It matches "ab" in the subject "aab". The use of the backtracking control
      verbs (*PRUNE) and (*SKIP) also disable this optimization.</p>

    <p>When a capturing subpattern is repeated, the value captured is the
      substring that matched the final iteration. For example, after</p>

    <code>
(tweedle[dume]{3}\s*)+</code>

    <p>has matched "tweedledum tweedledee", the value of the captured substring
      is "tweedledee". However, if there are nested capturing subpatterns, the
      corresponding captured values can have been set in previous iterations.
      For example, after</p>

    <code>
/(a|(b))+/</code>

    <p>matches "aba", the value of the second captured substring is "b".</p>
  </section>

  <section>
    <marker id="sect15"></marker>
    <title>Atomic Grouping and Possessive Quantifiers</title>
    <p>With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
      repetition, failure of what follows normally causes the repeated item to
      be re-evaluated to see if a different number of repeats allows the
      remaining pattern to match. Sometimes it is useful to prevent this, either
      to change the nature of the match, or to cause it to fail earlier than it
      otherwise might, when the author of the pattern knows that there is no
      point in carrying on.</p>

    <p>Consider, for example, the pattern \d+foo when applied to the following
      subject line:</p>

    <code>
123456bar</code>

    <p>After matching all six digits and then failing to match "foo", the normal
      action of the matcher is to try again with only five digits matching item
      \d+, and then with four, and so on, before ultimately failing. "Atomic
      grouping" (a term taken from Jeffrey Friedl's book) provides the means for
      specifying that once a subpattern has matched, it is not to be
      re-evaluated in this way.</p>

    <p>If atomic grouping is used for the previous example, the matcher gives up
      immediately on failing to match "foo" the first time. The notation is a
      kind of special parenthesis, starting with <c>(?&gt;</c> as in the
      following example:</p>

    <code>
(?&gt;\d+)foo</code>

    <p>This kind of parenthesis "locks up" the part of the pattern it contains
      once it has matched, and a failure further into the pattern is prevented
      from backtracking into it. Backtracking past it to previous items,
      however, works as normal.</p>

    <p>An alternative description is that a subpattern of this type matches the
      string of characters that an identical standalone pattern would match, if
      anchored at the current point in the subject string.</p>

    <p>Atomic grouping subpatterns are not capturing subpatterns. Simple cases
      such as the above example can be thought of as a maximizing repeat that
      must swallow everything it can. So, while both \d+ and \d+? are prepared
      to adjust the number of digits they match to make the remaining pattern
      match, <c>(?&gt;\d+)</c> can only match an entire sequence of digits.</p>

    <p>Atomic groups in general can contain any complicated
      subpatterns, and can be nested. However, when the subpattern for an atomic
      group is just a single repeated item, as in the example above, a simpler
      notation, called a "possessive quantifier" can be used. This consists of
      an extra + character following a quantifier. Using this notation, the
      previous example can be rewritten as</p>

    <code>
\d++foo</code>

    <p>Notice that a possessive quantifier can be used with an entire group,
      for example:</p>

    <code>
(abc|xyz){2,3}+</code>

    <p>Possessive quantifiers are always greedy; the setting of option
      <c>ungreedy</c> is ignored. They are a convenient notation for the simpler
      forms of an atomic group. However, there is no difference in the meaning
      of a possessive quantifier and the equivalent atomic group, but there can
      be a performance difference; possessive quantifiers are probably slightly
      faster.</p>

    <p>The possessive quantifier syntax is an extension to the Perl 5.8 syntax.
      Jeffrey Friedl originated the idea (and the name) in the first edition of
      his book. Mike McCloskey liked it, so implemented it when he built the
      Sun Java package, and PCRE copied it from there. It ultimately found its
      way into Perl at release 5.10.</p>

    <p>PCRE has an optimization that automatically "possessifies" certain simple
      pattern constructs. For example, the sequence A+B is treated as A++B, as
      there is no point in backtracking into a sequence of A:s when B must
      follow.</p>

    <p>When a pattern contains an unlimited repeat inside a subpattern that can
      itself be repeated an unlimited number of times, the use of an atomic
      group is the only way to avoid some failing matches taking a long time.
      The pattern</p>

    <code>
(\D+|&lt;\d+&gt;)*[!?]</code>

    <p>matches an unlimited number of substrings that either consist of
      non-digits, or digits enclosed in &lt;&gt;, followed by ! or ?. When it
      matches, it runs quickly. However, if it is applied to</p>

    <code>
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa</code>

    <p>it takes a long time before reporting failure. This is because the string
      can be divided between the internal \D+ repeat and the external * repeat
      in many ways, and all must be tried. (The example uses [!?] rather than a
      single character at the end, as both PCRE and Perl have an optimization
      that allows for fast failure when a single character is used. They
      remember the last single character that is required for a match, and fail
      early if it is not present in the string.) If the pattern is changed so
      that it uses an atomic group, like the following, sequences of non-digits
      cannot be broken, and failure happens quickly:</p>

    <code>
((?&gt;\D+)|&lt;\d+&gt;)*[!?]</code>
  </section>

  <section>
    <marker id="sect16"></marker>
    <title>Back References</title>
    <p>Outside a character class, a backslash followed by a digit &gt; 0 (and
      possibly further digits) is a back reference to a capturing subpattern
      earlier (that is, to its left) in the pattern, provided there have been
      that many previous capturing left parentheses.</p>

    <p>However, if the decimal number following the backslash is &lt; 10, it is
      always taken as a back reference, and causes an error only if there are
      not that many capturing left parentheses in the entire pattern. That is,
      the parentheses that are referenced do need not be to the left of the
      reference for numbers &lt; 10. A "forward back reference" of this type can
      make sense when a repetition is involved and the subpattern to the right
      has participated in an earlier iteration.</p>

    <p>It is not possible to have a numerical "forward back reference" to a
      subpattern whose number is 10 or more using this syntax, as a sequence
      such as \50 is interpreted as a character defined in octal. For more
      details of the handling of digits following a backslash, see section
      <seealso marker="#non_printing_characters">Non-Printing
      Characters</seealso> earlier. There is no such problem when named
      parentheses are used. A back reference to any subpattern is possible
      using named parentheses (see below).</p>

    <p>Another way to avoid the ambiguity inherent in the use of digits
      following a backslash is to use the \g escape sequence. This escape must
      be followed by an unsigned number or a negative number, optionally
      enclosed in braces. The following examples are identical:</p>

    <code>
(ring), \1
(ring), \g1
(ring), \g{1}</code>

    <p>An unsigned number specifies an absolute reference without the ambiguity
      that is present in the older syntax. It is also useful when literal digits
      follow the reference. A negative number is a relative reference. Consider
      the following example:</p>

    <code>
(abc(def)ghi)\g{-1}</code>

    <p>The sequence \g{-1} is a reference to the most recently started capturing
      subpattern before \g, that is, it is equivalent to \2 in this example.
      Similarly, \g{-2} would be equivalent to \1. The use of relative
      references can be helpful in long patterns, and also in patterns that are
      created by joining fragments containing references within themselves.</p>

    <p>A back reference matches whatever matched the capturing subpattern in the
      current subject string, rather than anything matching the subpattern
      itself (section <seealso marker="#sect21">Subpattern as
      Subroutines</seealso> describes a way of doing that). So, the
      following pattern matches "sense and sensibility" and "response and
      responsibility", but not "sense and responsibility":</p>

    <code>
(sens|respons)e and \1ibility</code>

    <p>If caseful matching is in force at the time of the back reference, the
      case of letters is relevant. For example, the following matches "rah rah"
      and "RAH RAH", but not "RAH rah", although the original capturing
      subpattern is matched caselessly:</p>

    <code>
((?i)rah)\s+\1</code>

    <p>There are many different ways of writing back references to named
      subpatterns. The .NET syntax <c>\k{name}</c> and the Perl syntax
      <c>\k&lt;name&gt;</c> or <c>\k'name'</c> are supported, as is the Python
      syntax <c>(?P=name)</c>. The unified back reference syntax in Perl 5.10,
      in which \g can be used for both numeric and named references, is also
      supported. The previous example can be rewritten in the following
      ways:</p>

    <code>
(?&lt;p1&gt;(?i)rah)\s+\k&lt;p1&gt;
(?'p1'(?i)rah)\s+\k{p1}
(?P&lt;p1&gt;(?i)rah)\s+(?P=p1)
(?&lt;p1&gt;(?i)rah)\s+\g{p1}</code>

    <p>A subpattern that is referenced by name can appear in the pattern before
      or after the reference.</p>

    <p>There can be more than one back reference to the same subpattern. If a
      subpattern has not been used in a particular match, any back references to
      it always fails. For example, the following pattern always fails if it
      starts to match "a" rather than "bc":</p>

    <code>
(a|(bc))\2</code>

    <p>As there can be many capturing parentheses in a pattern, all digits
      following the backslash are taken as part of a potential back reference
      number. If the pattern continues with a digit character, some delimiter
      must be used to terminate the back reference. If option <c>extended</c> is
      set, this can be whitespace. Otherwise an empty comment (see section
      <seealso marker="#sect19">Comments</seealso>) can be used.</p>

    <p><em>Recursive Back References</em></p>

    <p>A back reference that occurs inside the parentheses to which it refers
      fails when the subpattern is first used, so, for example, (a\1) never
      matches. However, such references can be useful inside repeated
      subpatterns. For example, the following pattern matches any number of
      "a"s and also "aba", "ababbaa", and so on:</p>

    <code>
(a|b\1)+</code>

    <p>At each iteration of the subpattern, the back reference matches the
      character string corresponding to the previous iteration. In order for
      this to work, the pattern must be such that the first iteration does not
      need to match the back reference. This can be done using alternation, as
      in the example above, or by a quantifier with a minimum of zero.</p>

    <p>Back references of this type cause the group that they reference to be
      treated as an atomic group. Once the whole group has been matched, a
      subsequent matching failure cannot cause backtracking into the middle of
      the group.</p>
  </section>

  <section>
    <marker id="sect17"></marker>
    <title>Assertions</title>
    <p>An assertion is a test on the characters following or preceding the
      current matching point that does not consume any characters. The simple
      assertions coded as \b, \B, \A, \G, \Z, \z, ^, and $ are described in
      the previous sections.</p>

    <p>More complicated assertions are coded as subpatterns. There are two
      kinds: those that look ahead of the current position in the subject
      string, and those that look behind it. An assertion subpattern is matched
      in the normal way, except that it does not cause the current matching
      position to be changed.</p>

    <p>Assertion subpatterns are not capturing subpatterns. If such an assertion
      contains capturing subpatterns within it, these are counted for the
      purposes of numbering the capturing subpatterns in the whole pattern.
      However, substring capturing is done only for positive assertions. (Perl
      sometimes, but not always, performs capturing in negative assertions.)</p>

    <p>For compatibility with Perl, assertion subpatterns can be repeated.
      However, it makes no sense to assert the same thing many times, the side
      effect of capturing parentheses can occasionally be useful. In practice,
      there are only three cases:</p>

    <list type="bulleted">
      <item>
        <p>If the quantifier is {0}, the assertion is never obeyed during
          matching. However, it can contain internal capturing parenthesized
          groups that are called from elsewhere through the subroutine
          mechanism.</p>
      </item>
      <item>
        <p>If quantifier is {0,n}, where n &gt; 0, it is treated as if it was
          {0,1}. At runtime, the remaining pattern match is tried with and
          without the assertion, the order depends on the greediness of the
          quantifier.</p>
      </item>
      <item>
        <p>If the minimum repetition is &gt; 0, the quantifier is ignored. The
          assertion is obeyed only once when encountered during matching.</p>
      </item>
    </list>

    <p><em>Lookahead Assertions</em></p>

    <p>Lookahead assertions start with (?= for positive assertions and (?! for
      negative assertions. For example, the following matches a word followed by
      a semicolon, but does not include the semicolon in the match:</p>

    <code>
\w+(?=;)</code>

    <p>The following matches any occurrence of "foo" that is not followed by
      "bar":</p>

    <code>
foo(?!bar)</code>

    <p>Notice that the apparently similar pattern</p>

    <code>
(?!foo)bar</code>

    <p>does not find an occurrence of "bar" that is preceded by something other
      than "foo". It finds any occurrence of "bar" whatsoever, as the assertion
      (?!foo) is always true when the next three characters are "bar". A
      lookbehind assertion is needed to achieve the other effect.</p>

    <p>If you want to force a matching failure at some point in a pattern, the
      most convenient way to do it is with (?!), as an empty string always
      matches. So, an assertion that requires there is not to be an empty
      string must always fail. The backtracking control verb (*FAIL) or (*F) is
      a synonym for (?!).</p>

    <p><em>Lookbehind Assertions</em></p>

    <p>Lookbehind assertions start with (?&lt;= for positive assertions and
      (?&lt;! for negative assertions. For example, the following finds an
      occurrence of "bar" that is not preceded by "foo":</p>

    <code>
(?&lt;!foo)bar</code>

    <p>The contents of a lookbehind assertion are restricted such that all the
      strings it matches must have a fixed length. However, if there are many
      top-level alternatives, they do not all have to have the same fixed
      length. Thus, the following is permitted:</p>

    <code>
(?&lt;=bullock|donkey)</code>

    <p>The following causes an error at compile time:</p>

    <code>
(?&lt;!dogs?|cats?)</code>

    <p>Branches that match different length strings are permitted only at the
      top-level of a lookbehind assertion. This is an extension compared with
      Perl, which requires all branches to match the same length of string. An
      assertion such as the following is not permitted, as its single top-level
      branch can match two different lengths:</p>

    <code>
(?&lt;=ab(c|de))</code>

    <p>However, it is acceptable to PCRE if rewritten to use two top-level
      branches:</p>

    <code>
(?&lt;=abc|abde)</code>

    <p>Sometimes the escape sequence \K (see above) can be used instead of
      a lookbehind assertion to get round the fixed-length restriction.</p>

    <p>The implementation of lookbehind assertions is, for each alternative, to
      move the current position back temporarily by the fixed length and then
      try to match. If there are insufficient characters before the current
      position, the assertion fails.</p>

    <p>In a UTF mode, PCRE does not allow the \C escape (which matches a single
      data unit even in a UTF mode) to appear in lookbehind assertions, as it
      makes it impossible to calculate the length of the lookbehind. The \X and
      \R escapes, which can match different numbers of data units, are not
      permitted either.</p>

    <p>"Subroutine" calls (see below), such as (?2) or (?&amp;X), are permitted
      in lookbehinds, as long as the subpattern matches a fixed-length string.
       Recursion, however, is not supported.</p>

    <p>Possessive quantifiers can be used with lookbehind
      assertions to specify efficient matching of fixed-length strings at the
      end of subject strings. Consider the following simple pattern when applied
      to a long string that does not match:</p>

    <code>
abcd$</code>

    <p>As matching proceeds from left to right, PCRE looks for each "a" in the
      subject and then sees if what follows matches the remaining pattern. If
      the pattern is specified as</p>

    <code>
^.*abcd$</code>

    <p>the initial .* matches the entire string at first. However, when this
      fails (as there is no following "a"), it backtracks to match all but the
      last character, then all but the last two characters, and so on. Once
      again the search for "a" covers the entire string, from right to left, so
      we are no better off. However, if the pattern is written as</p>

    <code>
^.*+(?&lt;=abcd)</code>

    <p>there can be no backtracking for the .*+ item; it can match only the
      entire string. The subsequent lookbehind assertion does a single test on
      the last four characters. If it fails, the match fails immediately. For
      long strings, this approach makes a significant difference to the
      processing time.</p>

    <p><em>Using Multiple Assertions</em></p>

    <p>Many assertions (of any sort) can occur in succession. For example, the
      following matches "foo" preceded by three digits that are not "999":</p>

    <code>
(?&lt;=\d{3})(?&lt;!999)foo</code>

    <p>Notice that each of the assertions is applied independently at the same
      point in the subject string. First there is a check that the previous
      three characters are all digits, and then there is a check that the same
      three characters are not "999".  This pattern does <em>not</em> match
      "foo" preceded by six characters, the first of which are digits and the
      last three of which are not "999". For example, it does not match
      "123abcfoo". A pattern to do that is the following:</p>

    <code>
(?&lt;=\d{3}...)(?&lt;!999)foo</code>

    <p>This time the first assertion looks at the preceding six characters,
      checks that the first three are digits, and then the second assertion
      checks that the preceding three characters are not "999".</p>

    <p>Assertions can be nested in any combination. For example, the following
      matches an occurrence of "baz" that is preceded by "bar", which in turn is
      not preceded by "foo":</p>

    <code>
(?&lt;=(?&lt;!foo)bar)baz</code>

    <p>The following pattern matches "foo" preceded by three digits and any
      three characters that are not "999":</p>

    <code>
(?&lt;=\d{3}(?!999)...)foo</code>
  </section>

  <section>
    <marker id="sect18"></marker>
    <title>Conditional Subpatterns</title>
    <p>It is possible to cause the matching process to obey a subpattern
      conditionally or to choose between two alternative subpatterns, depending
      on the result of an assertion, or whether a specific capturing subpattern
      has already been matched. The following are the two possible forms of
      conditional subpattern:</p>

    <code>
(?(condition)yes-pattern)
(?(condition)yes-pattern|no-pattern)</code>

    <p>If the condition is satisfied, the yes-pattern is used, otherwise the
      no-pattern (if present). If more than two alternatives exist in the
      subpattern, a compile-time error occurs. Each of the two alternatives can
      itself contain nested subpatterns of any form, including conditional
      subpatterns; the restriction to two alternatives applies only at the level
      of the condition. The following pattern fragment is an example where the
      alternatives are complex:</p>

    <code>
(?(1) (A|B|C) | (D | (?(2)E|F) | E) )</code>

    <p>There are four kinds of condition: references to subpatterns, references
      to recursion, a pseudo-condition called DEFINE, and assertions.</p>

    <p><em>Checking for a Used Subpattern By Number</em></p>

    <p>If the text between the parentheses consists of a sequence of digits,
      the condition is true if a capturing subpattern of that number has
      previously matched. If more than one capturing subpattern with the same
      number exists (see section <seealso marker="#sect12">
      Duplicate Subpattern Numbers</seealso> earlier), the condition is true if
      any of them have matched. An alternative notation is to precede the
      digits with a plus or minus sign. In this case, the subpattern number is
      relative rather than absolute. The most recently opened parentheses can be
      referenced by (?(-1), the next most recent by (?(-2), and so on. Inside
      loops, it can also make sense to refer to subsequent groups. The next
      parentheses to be opened can be referenced as (?(+1), and so on. (The
      value zero in any of these forms is not used; it provokes a compile-time
      error.)</p>

    <p>Consider the following pattern, which contains non-significant whitespace
      to make it more readable (assume option <c>extended</c>) and to divide it
      into three parts for ease of discussion:</p>

    <code>
( \( )?    [^()]+    (?(1) \) )</code>

    <p>The first part matches an optional opening parenthesis, and if that
      character is present, sets it as the first captured substring. The second
      part matches one or more characters that are not parentheses. The third
      part is a conditional subpattern that tests whether the first set of
      parentheses matched or not. If they did, that is, if subject started with
      an opening parenthesis, the condition is true, and so the yes-pattern is
      executed and a closing parenthesis is required. Otherwise, as no-pattern
      is not present, the subpattern matches nothing. That is, this pattern
      matches a sequence of non-parentheses, optionally enclosed in
      parentheses.</p>

    <p>If this pattern is embedded in a larger one, a relative reference can be
      used:</p>

    <code>
...other stuff... ( \( )?    [^()]+    (?(-1) \) ) ...</code>

    <p>This makes the fragment independent of the parentheses in the larger
      pattern.</p>

    <p><em>Checking for a Used Subpattern By Name</em></p>

    <p>Perl uses the syntax (?(&lt;name&gt;)...) or (?('name')...) to test for a
      used subpattern by name. For compatibility with earlier versions of PCRE,
      which had this facility before Perl, the syntax (?(name)...) is also
      recognized. However, there is a possible ambiguity with this syntax, as
      subpattern names can consist entirely of digits. PCRE looks first for a
      named subpattern; if it cannot find one and the name consists entirely of
      digits, PCRE looks for a subpattern of that number, which must be &gt; 0.
      Using subpattern names that consist entirely of digits is not
      recommended.</p>

    <p>Rewriting the previous example to use a named subpattern gives:</p>

    <code>
(?&lt;OPEN&gt; \( )?    [^()]+    (?(&lt;OPEN&gt;) \) )</code>

    <p>If the name used in a condition of this kind is a duplicate, the test is
      applied to all subpatterns of the same name, and is true if any one of
      them has matched.</p>

    <p><em>Checking for Pattern Recursion</em></p>

    <p>If the condition is the string (R), and there is no subpattern with the
      name R, the condition is true if a recursive call to the whole pattern or
      any subpattern has been made. If digits or a name preceded by ampersand
      follow the letter R, for example:</p>

    <code>
(?(R3)...) or (?(R&amp;name)...)</code>

    <p>the condition is true if the most recent recursion is into a subpattern
      whose number or name is given. This condition does not check the entire
      recursion stack. If the name used in a condition of this kind is a
      duplicate, the test is applied to all subpatterns of the same name, and is
      true if any one of them is the most recent recursion.</p>

    <p>At "top-level", all these recursion test conditions are false. The syntax
      for recursive patterns is described below.</p>

    <p><em>Defining Subpatterns for Use By Reference Only</em></p>
    <marker id="defining_subpatterns"/>

    <p>If the condition is the string (DEFINE), and there is no subpattern with
      the name DEFINE, the condition is always false. In this case, there can be
      only one alternative in the subpattern. It is always skipped if control
      reaches this point in the pattern. The idea of DEFINE is that it can be
      used to define "subroutines" that can be referenced from elsewhere. (The
      use of subroutines is described below.) For example, a pattern to match
      an IPv4 address, such as "192.168.23.245", can be written like this
      (ignore whitespace and line breaks):</p>

    <code>
(?(DEFINE) (?&lt;byte&gt; 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) ) \b (?&amp;byte) (\.(?&amp;byte)){3} \b</code>

    <p>The first part of the pattern is a DEFINE group inside which is a another
      group named "byte" is defined. This matches an individual component of an
      IPv4 address (a number &lt; 256). When matching takes place, this part of
      the pattern is skipped, as DEFINE acts like a false condition. The
      remaining pattern uses references to the named group to match the four
      dot-separated components of an IPv4 address, insisting on a word boundary
      at each end.</p>

    <p><em>Assertion Conditions</em></p>

    <p>If the condition is not in any of the above formats, it must be an
      assertion. This can be a positive or negative lookahead or lookbehind
      assertion. Consider the following pattern, containing non-significant
      whitespace, and with the two alternatives on the second line:</p>

    <code type="none">
(?(?=[^a-z]*[a-z])
\d{2}-[a-z]{3}-\d{2}  |  \d{2}-\d{2}-\d{2} )</code>

    <p>The condition is a positive lookahead assertion that matches an optional
      sequence of non-letters followed by a letter. That is, it tests for the
      presence of at least one letter in the subject. If a letter is found, the
      subject is matched against the first alternative, otherwise it is matched
      against the second. This pattern matches strings in one of the two forms
      dd-aaa-dd or dd-dd-dd, where aaa are letters and dd are digits.</p>
  </section>

  <section>
    <marker id="sect19"></marker>
    <title>Comments</title>
    <p>There are two ways to include comments in patterns that are processed by
      PCRE. In both cases, the start of the comment must not be in a character
      class, or in the middle of any other sequence of related characters such
      as (?: or a subpattern name or number. The characters that make up a
      comment play no part in the pattern matching.</p>

    <p>The sequence (?# marks the start of a comment that continues up to the
      next closing parenthesis. Nested parentheses are not permitted. If option
      PCRE_EXTENDED is set, an unescaped # character also introduces a comment,
      which in this case continues to immediately after the next newline
      character or character sequence in the pattern. Which characters are
      interpreted as newlines is controlled by the options passed to a
      compiling function or by a special sequence at the start of the pattern,
      as described in section <seealso marker="#newline_conventions">
      Newline Conventions</seealso> earlier.</p>

    <p>Notice that the end of this type of comment is a literal newline sequence
      in the pattern; escape sequences that happen to represent a newline do not
      count. For example, consider the following pattern when <c>extended</c> is
      set, and the default newline convention is in force:</p>

    <code>
abc #comment \n still comment</code>

    <p>On encountering character #, <c>pcre_compile()</c> skips along, looking
      for a newline in the pattern. The sequence \n is still literal at this
      stage, so it does not terminate the comment. Only a character with code
      value 0x0a (the default newline) does so.</p>
  </section>

  <section>
    <marker id="sect20"></marker>
    <title>Recursive Patterns</title>
    <p>Consider the problem of matching a string in parentheses, allowing for
      unlimited nested parentheses. Without the use of recursion, the best that
      can be done is to use a pattern that matches up to some fixed depth of
      nesting. It is not possible to handle an arbitrary nesting depth.</p>

    <p>For some time, Perl has provided a facility that allows regular
      expressions to recurse (among other things). It does this by
      interpolating Perl code in the expression at runtime, and the code can
      refer to the expression itself. A Perl pattern using code interpolation to
      solve the parentheses problem can be created like this:</p>

    <code>
$re = qr{\( (?: (?&gt;[^()]+) | (?p{$re}) )* \)}x;</code>

    <p>Item (?p{...}) interpolates Perl code at runtime, and in this case refers
      recursively to the pattern in which it appears.</p>

    <p>Obviously, PCRE cannot support the interpolation of Perl code. Instead,
      it supports special syntax for recursion of the entire pattern, and for
      individual subpattern recursion. After its introduction in PCRE and
      Python, this kind of recursion was later introduced into Perl at
      release 5.10.</p>

    <p>A special item that consists of (? followed by a number &gt; 0 and a
      closing parenthesis is a recursive subroutine call of the subpattern of
      the given number, if it occurs inside that subpattern. (If not,
      it is a non-recursive subroutine call, which is described in the next
      section.) The special item (?R) or (?0) is a recursive call of the entire
      regular expression.</p>

    <p>This PCRE pattern solves the nested parentheses problem (assume that
      option <c>extended</c> is set so that whitespace is ignored):</p>

    <code>
\( ( [^()]++ | (?R) )* \)</code>

    <p>First it matches an opening parenthesis. Then it matches any number of
      substrings, which can either be a sequence of non-parentheses or a
      recursive match of the pattern itself (that is, a correctly parenthesized
      substring). Finally there is a closing parenthesis. Notice the use of a
      possessive quantifier to avoid backtracking into sequences of
      non-parentheses.</p>

    <p>If this was part of a larger pattern, you would not want to recurse the
      entire pattern, so instead you can use:</p>

    <code>
( \( ( [^()]++ | (?1) )* \) )</code>

    <p>The pattern is here within parentheses so that the recursion refers to
      them instead of the whole pattern.</p>

    <p>In a larger pattern, keeping track of parenthesis numbers can be tricky.
      This is made easier by the use of relative references. Instead of (?1) in
      the pattern above, you can write (?-2) to refer to the second most
      recently opened parentheses preceding the recursion. That is, a negative
      number counts capturing parentheses leftwards from the point at which it
      is encountered.</p>

    <p>It is also possible to refer to later opened parentheses, by
      writing references such as (?+2). However, these cannot be recursive, as
      the reference is not inside the parentheses that are referenced. They are
      always non-recursive subroutine calls, as described in the next
      section.</p>

    <p>An alternative approach is to use named parentheses instead. The Perl
      syntax for this is (?&amp;name). The earlier PCRE syntax (?P&gt;name) is
      also supported. We can rewrite the above example as follows:</p>

    <code>
(?&lt;pn&gt; \( ( [^()]++ | (?&amp;pn) )* \) )</code>

    <p>If there is more than one subpattern with the same name, the earliest
      one is used.</p>

    <p>This particular example pattern that we have studied contains nested
      unlimited repeats, and so the use of a possessive quantifier for matching
      strings of non-parentheses is important when applying the pattern to
      strings that do not match. For example, when this pattern is applied
      to</p>

    <code>
(aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()</code>

    <p>it gives "no match" quickly. However, if a possessive quantifier is not
      used, the match runs for a long time, as there are so many different
      ways the + and * repeats can carve up the subject, and all must be tested
      before failure can be reported.</p>

    <p>At the end of a match, the values of capturing parentheses are those from
      the outermost level. If the pattern above is matched against</p>

    <code>
(ab(cd)ef)</code>

    <p>the value for the inner capturing parentheses (numbered 2) is "ef",
      which is the last value taken on at the top-level. If a capturing
      subpattern is not matched at the top level, its final captured value is
      unset, even if it was (temporarily) set at a deeper level during the
      matching process.</p>

    <p>Do not confuse item (?R) with condition (R), which tests for recursion.
      Consider the following pattern, which matches text in angle brackets,
      allowing for arbitrary nesting. Only digits are allowed in nested brackets
      (that is, when recursing), while any characters are permitted at the
      outer level.</p>

    <code>
&lt; (?: (?(R) \d++  | [^&lt;&gt;]*+) | (?R)) * &gt;</code>

    <p>Here (?(R) is the start of a conditional subpattern, with two different
      alternatives for the recursive and non-recursive cases. Item (?R) is the
      actual recursive call.</p>

    <p><em>Differences in Recursion Processing between PCRE and Perl</em></p>

    <p>Recursion processing in PCRE differs from Perl in two important ways. In
      PCRE (like Python, but unlike Perl), a recursive subpattern call is always
      treated as an atomic group. That is, once it has matched some of the
      subject string, it is never re-entered, even if it contains untried
      alternatives and there is a subsequent matching failure. This can be
      illustrated by the following pattern, which means to match a palindromic
      string containing an odd number of characters (for example, "a", "aba",
      "abcba", "abcdcba"):</p>

    <code>
^(.|(.)(?1)\2)$</code>

    <p>The idea is that it either matches a single character, or two identical
      characters surrounding a subpalindrome. In Perl, this pattern works; in
      PCRE it does not work if the pattern is longer than three characters.
      Consider the subject string "abcba".</p>

    <p>At the top level, the first character is matched, but as it is not at
      the end of the string, the first alternative fails, the second
      alternative is taken, and the recursion kicks in. The recursive call to
      subpattern 1 successfully matches the next character ("b"). (Notice that
      the beginning and end of line tests are not part of the recursion.)</p>

    <p>Back at the top level, the next character ("c") is compared with what
      subpattern 2 matched, which was "a". This fails. As the recursion is
      treated as an atomic group, there are now no backtracking points, and so
      the entire match fails. (Perl can now re-enter the recursion
      and try the second alternative.) However, if the pattern is written with
      the alternatives in the other order, things are different:</p>

    <code>
^((.)(?1)\2|.)$</code>

    <p>This time, the recursing alternative is tried first, and continues to
      recurse until it runs out of characters, at which point the recursion
      fails. But this time we have another alternative to try at the higher
      level. That is the significant difference: in the previous case the
      remaining alternative is at a deeper recursion level, which PCRE cannot
      use.</p>

    <p>To change the pattern so that it matches all palindromic strings, not
      only those with an odd number of characters, it is tempting to change the
      pattern to this:</p>

    <code>
^((.)(?1)\2|.?)$</code>

    <p>Again, this works in Perl, but not in PCRE, and for the same reason. When
      a deeper recursion has matched a single character, it cannot be entered
      again to match an empty string. The solution is to separate the two cases,
      and write out the odd and even cases as alternatives at the higher
      level:</p>

    <code>
^(?:((.)(?1)\2|)|((.)(?3)\4|.))</code>

    <p>If you want to match typical palindromic phrases, the pattern must ignore
      all non-word characters, which can be done as follows:</p>

    <code>
^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$</code>

    <p>If run with option <c>caseless</c>, this pattern matches phrases such as
      "A man, a plan, a canal: Panama!" and it works well in both PCRE and Perl.
      Notice the use of the possessive quantifier *+ to avoid backtracking into
      sequences of non-word characters. Without this, PCRE takes much longer
      (10 times or more) to match typical phrases, and Perl takes so long that
      you think it has gone into a loop.</p>

    <note>
      <p>The palindrome-matching patterns above work only if the subject string
        does not start with a palindrome that is shorter than the entire string.
        For example, although "abcba" is correctly matched, if the subject is
        "ababa", PCRE finds palindrome "aba" at the start, and then fails at top
        level, as the end of the string does not follow. Once again, it cannot
        jump back into the recursion to try other alternatives, so the entire
        match fails.</p>
    </note>

    <p>The second way in which PCRE and Perl differ in their recursion
      processing is in the handling of captured values. In Perl, when a
      subpattern is called recursively or as a subpattern (see the next
      section), it has no access to any values that were captured outside the
      recursion. In PCRE these values can be referenced. Consider the following
      pattern:</p>

    <code>
^(.)(\1|a(?2))</code>

    <p>In PCRE, it matches "bab". The first capturing parentheses match "b",
      then in the second group, when the back reference \1 fails to match "b",
      the second alternative matches "a", and then recurses. In the recursion,
      \1 does now match "b" and so the whole match succeeds. In Perl, the
      pattern fails to match because inside the recursive call \1 cannot access
      the externally set value.</p>
  </section>

  <section>
    <marker id="sect21"></marker>
    <title>Subpatterns as Subroutines</title>
    <p>If the syntax for a recursive subpattern call (either by number or by
      name) is used outside the parentheses to which it refers, it operates
      like a subroutine in a programming language. The called subpattern can be
      defined before or after the reference. A numbered reference can be
      absolute or relative, as in the following examples:</p>

    <code>
(...(absolute)...)...(?2)...
(...(relative)...)...(?-1)...
(...(?+1)...(relative)...</code>

    <p>An earlier example pointed out that the following pattern matches "sense
      and sensibility" and "response and responsibility", but not "sense and
      responsibility":</p>

    <code>
(sens|respons)e and \1ibility</code>

    <p>If instead the following pattern is used, it matches "sense and
      responsibility" and the other two strings:</p>

    <code>
(sens|respons)e and (?1)ibility</code>

    <p>Another example is provided in the discussion of DEFINE earlier.</p>

    <p>All subroutine calls, recursive or not, are always treated as atomic
      groups. That is, once a subroutine has matched some of the subject string,
      it is never re-entered, even if it contains untried alternatives and there
      is a subsequent matching failure. Any capturing parentheses that are set
      during the subroutine call revert to their previous values afterwards.</p>

    <p>Processing options such as case-independence are fixed when a subpattern
      is defined, so if it is used as a subroutine, such options cannot be
      changed for different calls. For example, the following pattern matches
      "abcabc" but not "abcABC", as the change of processing option does not
      affect the called subpattern:</p>

    <code>
(abc)(?i:(?-1))</code>
  </section>

  <section>
    <marker id="sect22"></marker>
    <title>Oniguruma Subroutine Syntax</title>
    <p>For compatibility with Oniguruma, the non-Perl syntax \g followed by a
      name or a number enclosed either in angle brackets or single quotes, is
      alternative syntax for referencing a subpattern as a subroutine, possibly
      recursively. Here follows two of the examples used above, rewritten using
      this syntax:</p>

    <code>
(?&lt;pn&gt; \( ( (?&gt;[^()]+) | \g&lt;pn&gt; )* \) )
(sens|respons)e and \g'1'ibility</code>

    <p>PCRE supports an extension to Oniguruma: if a number is preceded by a
      plus or minus sign, it is taken as a relative reference, for example:</p>

    <code>
(abc)(?i:\g&lt;-1&gt;)</code>

    <p>Notice that \g{...} (Perl syntax) and \g&lt;...&gt; (Oniguruma syntax)
      are <em>not</em> synonymous. The former is a back reference; the latter
      is a subroutine call.</p>
  </section>

  <section>
    <marker id="sect23"></marker>
    <title>Backtracking Control</title>
    <p>Perl 5.10 introduced some "Special Backtracking Control Verbs",
      which are still described in the Perl documentation as "experimental and
      subject to change or removal in a future version of Perl". It goes on to
      say: "Their usage in production code should be noted to avoid problems
      during upgrades." The same remarks apply to the PCRE features described
      in this section.</p>

    <p>The new verbs make use of what was previously invalid syntax: an opening
      parenthesis followed by an asterisk. They are generally of the form
      (*VERB) or (*VERB:NAME). Some can take either form, possibly behaving
      differently depending on whether a name is present. A name is any sequence
      of characters that does not include a closing parenthesis. The maximum
      name length is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit
      libraries. If the name is empty, that is, if the closing parenthesis
      immediately follows the colon, the effect is as if the colon was not
      there. Any number of these verbs can occur in a pattern.</p>

    <p>The behavior of these verbs in repeated groups, assertions, and in
      subpatterns called as subroutines (whether or not recursively) is
      described below.</p>

    <p><em>Optimizations That Affect Backtracking Verbs</em></p>

    <p>PCRE contains some optimizations that are used to speed up matching by
      running some checks at the start of each match attempt. For example, it
      can know the minimum length of matching subject, or that a particular
      character must be present. When one of these optimizations bypasses the
      running of a match, any included backtracking verbs are not processed.
      processed. You can suppress the start-of-match optimizations by setting
      option <c>no_start_optimize</c> when calling
      <seealso marker="#compile/2"><c>compile/2</c></seealso> or
      <seealso marker="#run/3"><c>run/3</c></seealso>, or by starting the
      pattern with (*NO_START_OPT).</p>

    <p>Experiments with Perl suggest that it too has similar optimizations,
      sometimes leading to anomalous results.</p>

    <p><em>Verbs That Act Immediately</em></p>

    <p>The following verbs act as soon as they are encountered. They must not
      be followed by a name.</p>

    <code>
(*ACCEPT)</code>

    <p>This verb causes the match to end successfully, skipping the remainder of
      the pattern. However, when it is inside a subpattern that is called as a
      subroutine, only that subpattern is ended successfully. Matching then
      continues at the outer level. If (*ACCEPT) is triggered in a positive
      assertion, the assertion succeeds; in a negative assertion, the assertion
      fails.</p>

    <p>If (*ACCEPT) is inside capturing parentheses, the data so far is
      captured. For example, the following matches "AB", "AAD", or "ACD". When
      it matches "AB", "B" is captured by the outer parentheses.</p>

    <code>
A((?:A|B(*ACCEPT)|C)D)</code>

    <p>The following verb causes a matching failure, forcing backtracking to
      occur. It is equivalent to (?!) but easier to read.</p>

    <code>
(*FAIL) or (*F)</code>

    <p>The Perl documentation states that it is probably useful only when
      combined with (?{}) or (??{}). Those are Perl features that
      are not present in PCRE.</p>

    <p>A match with the string "aaaa" always fails, but the callout is taken
      before each backtrack occurs (in this example, 10 times).</p>

    <p><em>Recording Which Path Was Taken</em></p>

    <p>The main purpose of this verb is to track how a match was arrived at,
      although it also has a secondary use in with advancing the match
      starting point (see (*SKIP) below).</p>

    <note>
      <p>In Erlang, there is no interface to retrieve a mark with
        <seealso marker="#run/2"><c>run/2,3</c></seealso>, so only the secondary
        purpose is relevant to the Erlang programmer.</p>

      <p>The rest of this section is therefore deliberately not adapted for
        reading by the Erlang programmer, but the examples can help in
        understanding NAMES as they can be used by (*SKIP).</p>
    </note>

    <code>
(*MARK:NAME) or (*:NAME)</code>

    <p>A name is always required with this verb. There can be as many instances
      of (*MARK) as you like in a pattern, and their names do not have to be
      unique.</p>

    <p>When a match succeeds, the name of the last encountered (*MARK:NAME),
      (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to the
      caller as described in section "Extra data for <c>pcre_exec()</c>" in the
      <c>pcreapi</c> documentation. In the following example of <c>pcretest</c>
      output, the /K modifier requests the retrieval and outputting of (*MARK)
      data:</p>

<code>
  re&gt; /X(*MARK:A)Y|X(*MARK:B)Z/K
data&gt; XY
 0: XY
MK: A
XZ
 0: XZ
MK: B</code>

    <p>The (*MARK) name is tagged with "MK:" in this output, and in this example
      it indicates which of the two alternatives matched. This is a more
      efficient way of obtaining this information than putting each alternative
      in its own capturing parentheses.</p>

    <p>If a verb with a name is encountered in a positive assertion that is
      true, the name is recorded and passed back if it is the last encountered.
      This does not occur for negative assertions or failing positive
      assertions.</p>

    <p>After a partial match or a failed match, the last encountered name in the
      entire match process is returned, for example:</p>

    <code>
  re&gt; /X(*MARK:A)Y|X(*MARK:B)Z/K
data&gt; XP
No match, mark = B</code>

    <p>Notice that in this unanchored example, the mark is retained from the
      match attempt that started at letter "X" in the subject. Subsequent match
      attempts starting at "P" and then with an empty string do not get as far
      as the (*MARK) item, nevertheless do not reset it.</p>

    <p><em>Verbs That Act after Backtracking</em></p>

    <p>The following verbs do nothing when they are encountered. Matching
      continues with what follows, but if there is no subsequent match, causing
      a backtrack to the verb, a failure is forced. That is, backtracking cannot
      pass to the left of the verb. However, when one of these verbs appears
      inside an atomic group or an assertion that is true, its effect is
      confined to that group, as once the group has been matched, there is never
      any backtracking into it. In this situation, backtracking can "jump back"
      to the left of the entire atomic group or assertion. (Remember also, as
      stated above, that this localization also applies in subroutine
      calls.)</p>

    <p>These verbs differ in exactly what kind of failure occurs when
      backtracking reaches them. The behavior described below is what occurs
      when the verb is not in a subroutine or an assertion. Subsequent sections
      cover these special cases.</p>

    <p>The following verb, which must not be followed by a name, causes the
      whole match to fail outright if there is a later matching failure that
      causes backtracking to reach it. Even if the pattern is unanchored, no
      further attempts to find a match by advancing the starting point take
      place.</p>

    <code>
(*COMMIT)</code>

    <p>If (*COMMIT) is the only backtracking verb that is encountered, once it
      has been passed, <seealso marker="#run/2"><c>run/2,3</c></seealso> is
      committed to find a match at the current starting point, or not at all,
      for example:</p>

    <code>
a+(*COMMIT)b</code>

    <p>This matches "xxaab" but not "aacaab". It can be thought of as a kind of
      dynamic anchor, or "I've started, so I must finish". The name of the most
      recently passed (*MARK) in the path is passed back when (*COMMIT) forces
      a match failure.</p>

    <p>If more than one backtracking verb exists in a pattern, a different one
      that follows (*COMMIT) can be triggered first, so merely passing (*COMMIT)
      during a match does not always guarantee that a match must be at this
      starting point.</p>

    <p>Notice that (*COMMIT) at the start of a pattern is not the same as an
      anchor, unless the PCRE start-of-match optimizations are turned off, as
      shown in the following example:</p>

<code type="none">
1&gt; re:run("xyzabc","(*COMMIT)abc",[{capture,all,list}]).
{match,["abc"]}
2&gt; re:run("xyzabc","(*COMMIT)abc",[{capture,all,list},no_start_optimize]).
nomatch</code>

    <p>PCRE knows that any match must start with "a", so the optimization skips
      along the subject to "a" before running the first match attempt, which
      succeeds. When the optimization is disabled by option
      <c>no_start_optimize</c>, the match starts at "x" and so the (*COMMIT)
      causes it to fail without trying any other starting points.</p>

    <p>The following verb causes the match to fail at the current starting
      position in the subject if there is a later matching failure that causes
      backtracking to reach it:</p>

    <code>
(*PRUNE) or (*PRUNE:NAME)</code>

    <p>If the pattern is unanchored, the normal "bumpalong" advance to the next
      starting character then occurs. Backtracking can occur as usual to the
      left of (*PRUNE), before it is reached, or when matching to the right of
      (*PRUNE), but if there is no match to the right, backtracking cannot
      cross (*PRUNE). In simple cases, the use of (*PRUNE) is just an
      alternative to an atomic group or possessive quantifier, but there are
      some uses of (*PRUNE) that cannot be expressed in any other way. In an
      anchored pattern, (*PRUNE) has the same effect as (*COMMIT).</p>

    <p>The behavior of (*PRUNE:NAME) is the not the same as
      (*MARK:NAME)(*PRUNE). It is like (*MARK:NAME) in that the name is
      remembered for passing back to the caller. However, (*SKIP:NAME) searches
      only for names set with (*MARK).</p>

    <note>
      <p>The fact that (*PRUNE:NAME) remembers the name is useless to the Erlang
        programmer, as names cannot be retrieved.</p>
    </note>

    <p>The following verb, when specified without a name, is like (*PRUNE),
      except that if the pattern is unanchored, the "bumpalong" advance is not
      to the next character, but to the position in the subject where (*SKIP)
      was encountered.</p>

    <code>
(*SKIP)</code>

    <p>(*SKIP) signifies that whatever text was matched leading up to it cannot
      be part of a successful match. Consider:</p>

    <code>
a+(*SKIP)b</code>

    <p>If the subject is "aaaac...", after the first match attempt fails
      (starting at the first character in the string), the starting point skips
      on to start the next attempt at "c". Notice that a possessive quantifier
      does not have the same effect as this example; although it would suppress
      backtracking during the first match attempt, the second attempt would
      start at the second character instead of skipping on to "c".</p>

    <p>When (*SKIP) has an associated name, its behavior is modified:</p>

    <code>
(*SKIP:NAME)</code>

    <p>When this is triggered, the previous path through the pattern is searched
      for the most recent (*MARK) that has the same name. If one is found, the
      "bumpalong" advance is to the subject position that corresponds to that
      (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with a
      matching name is found, (*SKIP) is ignored.</p>

    <p>Notice that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It
      ignores names that are set by (*PRUNE:NAME) or (*THEN:NAME).</p>

    <p>The following verb causes a skip to the next innermost alternative when
      backtracking reaches it. That is, it cancels any further backtracking
      within the current alternative.</p>

    <code>
(*THEN) or (*THEN:NAME)</code>

    <p>The verb name comes from the observation that it can be used for a
      pattern-based if-then-else block:</p>

    <code>
( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...</code>

    <p>If the COND1 pattern matches, FOO is tried (and possibly further items
      after the end of the group if FOO succeeds). On failure, the matcher skips
      to the second alternative and tries COND2, without backtracking into
      COND1. If that succeeds and BAR fails, COND3 is tried. If BAZ then fails,
      there are no more alternatives, so there is a backtrack to whatever
      came before the entire group. If (*THEN) is not inside an alternation, it
      acts like (*PRUNE).</p>

    <p>The behavior of (*THEN:NAME) is the not the same as (*MARK:NAME)(*THEN).
      It is like (*MARK:NAME) in that the name is remembered for passing back to
      the caller. However, (*SKIP:NAME) searches only for names set with
      (*MARK).</p>

    <note>
      <p>The fact that (*THEN:NAME) remembers the name is useless to the Erlang
        programmer, as names cannot be retrieved.</p>
    </note>

    <p>A subpattern that does not contain a | character is just a part of the
      enclosing alternative; it is not a nested alternation with only one
      alternative. The effect of (*THEN) extends beyond such a subpattern to the
      enclosing alternative. Consider the following pattern, where A, B, and so
      on, are complex pattern fragments that do not contain any | characters at
      this level:</p>

    <code>
A (B(*THEN)C) | D</code>

    <p>If A and B are matched, but there is a failure in C, matching does not
      backtrack into A; instead it moves to the next alternative, that is, D.
      However, if the subpattern containing (*THEN) is given an alternative, it
      behaves differently:</p>

    <code>
A (B(*THEN)C | (*FAIL)) | D</code>

    <p>The effect of (*THEN) is now confined to the inner subpattern. After a
      failure in C, matching moves to (*FAIL), which causes the whole subpattern
      to fail, as there are no more alternatives to try. In this case, matching
      does now backtrack into A.</p>

    <p>Notice that a conditional subpattern is not considered as having two
      alternatives, as only one is ever used. That is, the | character in a
      conditional subpattern has a different meaning. Ignoring whitespace,
      consider:</p>

    <code>
^.*? (?(?=a) a | b(*THEN)c )</code>

    <p>If the subject is "ba", this pattern does not match. As .*? is ungreedy,
      it initially matches zero characters. The condition (?=a) then fails, the
      character "b" is matched, but "c" is not. At this point, matching does not
      backtrack to .*? as can perhaps be expected from the presence of the |
      character. The conditional subpattern is part of the single alternative
      that comprises the whole pattern, and so the match fails. (If there was a
      backtrack into .*?, allowing it to match "b", the match would
      succeed.)</p>

    <p>The verbs described above provide four different "strengths" of control
      when subsequent matching fails:</p>

    <list type="bulleted">
      <item>
        <p>(*THEN) is the weakest, carrying on the match at the next
          alternative.</p>
      </item>
      <item>
        <p>(*PRUNE) comes next, fails the match at the current starting
          position, but allows an advance to the next character (for an
          unanchored pattern).</p>
      </item>
      <item>
        <p>(*SKIP) is similar, except that the advance can be more than one
          character.</p>
      </item>
      <item>
        <p>(*COMMIT) is the strongest, causing the entire match to fail.</p>
      </item>
    </list>

    <p><em>More than One Backtracking Verb</em></p>

    <p>If more than one backtracking verb is present in a pattern, the one that
      is backtracked onto first acts. For example, consider the following
      pattern, where A, B, and so on, are complex pattern fragments:</p>

    <code>
(A(*COMMIT)B(*THEN)C|ABD)</code>

    <p>If A matches but B fails, the backtrack to (*COMMIT) causes the entire
      match to fail. However, if A and B match, but C fails, the backtrack to
      (*THEN) causes the next alternative (ABD) to be tried. This behavior is
      consistent, but is not always the same as in Perl. It means that if two or
      more backtracking verbs appear in succession, the last of them has no
      effect. Consider the following example:</p>

    <code>
...(*COMMIT)(*PRUNE)...</code>

    <p>If there is a matching failure to the right, backtracking onto (*PRUNE)
      cases it to be triggered, and its action is taken. There can never be a
      backtrack onto (*COMMIT).</p>

    <p><em>Backtracking Verbs in Repeated Groups</em></p>

    <p>PCRE differs from Perl in its handling of backtracking verbs in repeated
      groups. For example, consider:</p>

    <code>
/(a(*COMMIT)b)+ac/</code>

    <p>If the subject is "abac", Perl matches, but PCRE fails because the
      (*COMMIT) in the second repeat of the group acts.</p>

    <p><em>Backtracking Verbs in Assertions</em></p>

    <p>(*FAIL) in an assertion has its normal effect: it forces an immediate
      backtrack.</p>

    <p>(*ACCEPT) in a positive assertion causes the assertion to succeed without
      any further processing. In a negative assertion, (*ACCEPT) causes the
      assertion to fail without any further processing.</p>

    <p>The other backtracking verbs are not treated specially if they appear in
      a positive assertion. In particular, (*THEN) skips to the next alternative
      in the innermost enclosing group that has alternations, regardless if this
      is within the assertion.</p>

    <p>Negative assertions are, however, different, to ensure that changing a
      positive assertion into a negative assertion changes its result.
      Backtracking into (*COMMIT), (*SKIP), or (*PRUNE) causes a negative
      assertion to be true, without considering any further alternative branches
      in the assertion. Backtracking into (*THEN) causes it to skip to the next
      enclosing alternative within the assertion (the normal behavior), but if
      the assertion does not have such an alternative, (*THEN) behaves like
      (*PRUNE).</p>

    <p><em>Backtracking Verbs in Subroutines</em></p>

    <p>These behaviors occur regardless if the subpattern is called recursively.
      The treatment of subroutines in Perl is different in some cases.</p>

    <list type="bulleted">
      <item>
        <p>(*FAIL) in a subpattern called as a subroutine has its normal effect:
          it forces an immediate backtrack.</p>
      </item>
      <item>
        <p>(*ACCEPT) in a subpattern called as a subroutine causes the
          subroutine match to succeed without any further processing. Matching
          then continues after the subroutine call.</p>
      </item>
      <item>
        <p>(*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a
          subroutine cause the subroutine match to fail.</p>
      </item>
      <item>
        <p>(*THEN) skips to the next alternative in the innermost enclosing
          group within the subpattern that has alternatives. If there is no such
          group within the subpattern, (*THEN) causes the subroutine match to
          fail.</p>
      </item>
    </list>
  </section>
</erlref>