<HTML ><HEAD ><TITLE >Pattern Syntax</TITLE ><META NAME="GENERATOR" CONTENT="Modular DocBook HTML Stylesheet Version 1.7"><LINK REL="HOME" TITLE="PHP Manual" HREF="index.html"><LINK REL="UP" TITLE="Regular Expression Functions (Perl-Compatible)" HREF="ref.pcre.html"><LINK REL="PREVIOUS" TITLE="Pattern Modifiers" HREF="pcre.pattern.modifiers.html"><LINK REL="NEXT" TITLE="preg_grep" HREF="function.preg-grep.html"><META HTTP-EQUIV="Content-type" CONTENT="text/html; charset=ISO-8859-1"></HEAD ><BODY CLASS="refentry" BGCOLOR="#FFFFFF" TEXT="#000000" LINK="#0000FF" VLINK="#840084" ALINK="#0000FF" ><DIV CLASS="NAVHEADER" ><TABLE SUMMARY="Header navigation table" WIDTH="100%" BORDER="0" CELLPADDING="0" CELLSPACING="0" ><TR ><TH COLSPAN="3" ALIGN="center" >PHP Manual</TH ></TR ><TR ><TD WIDTH="10%" ALIGN="left" VALIGN="bottom" ><A HREF="pcre.pattern.modifiers.html" ACCESSKEY="P" >Prev</A ></TD ><TD WIDTH="80%" ALIGN="center" VALIGN="bottom" ></TD ><TD WIDTH="10%" ALIGN="right" VALIGN="bottom" ><A HREF="function.preg-grep.html" ACCESSKEY="N" >Next</A ></TD ></TR ></TABLE ><HR ALIGN="LEFT" WIDTH="100%"></DIV ><H1 ><A NAME="pcre.pattern.syntax" ></A >Pattern Syntax</H1 ><DIV CLASS="refnamediv" ><A NAME="AEN81011" ></A >Pattern Syntax -- Describes PCRE regex syntax</DIV ><DIV CLASS="refsect1" ><A NAME="AEN81014" ></A ><H2 >Description</H2 ><P > The PCRE library is a set of functions that implement regular expression pattern matching using the same syntax and semantics as Perl 5, with just a few differences (see below). The current implementation corresponds to Perl 5.005. </P ></DIV ><DIV CLASS="refsect1" ><A NAME="AEN81017" ></A ><H2 >Differences From Perl</H2 ><P > The differences described here are with respect to Perl 5.005. <P ></P ><OL TYPE="1" ><LI ><P > By default, a whitespace character is any character that the C library function isspace() recognizes, though it is possible to compile PCRE with alternative character type tables. Normally isspace() matches space, formfeed, newline, carriage return, horizontal tab, and vertical tab. Perl 5 no longer includes vertical tab in its set of whitespace characters. The \v escape that was in the Perl documentation for a long time was never in fact recognized. However, the character itself was treated as whitespace at least up to 5.002. In 5.004 and 5.005 it does not match \s. </P ></LI ><LI ><P > PCRE does not allow repeat quantifiers on lookahead assertions. Perl permits them, but they do not mean what you might think. For example, (?!a){3} does not assert that the next three characters are not "a". It just asserts that the next character is not "a" three times. </P ></LI ><LI ><P > Capturing subpatterns that occur inside negative looka- head assertions are counted, but their entries in the offsets vector are never set. Perl sets its numerical vari- ables from any such patterns that are matched before the assertion fails to match something (thereby succeeding), but only if the negative lookahead assertion contains just one branch. </P ></LI ><LI ><P > Though binary zero characters are supported in the subject string, they are not allowed in a pattern string because it is passed as a normal C string, terminated by zero. The escape sequence "\\x00" can be used in the pattern to represent a binary zero. </P ></LI ><LI ><P > The following Perl escape sequences are not supported: \l, \u, \L, \U, \E, \Q. In fact these are implemented by Perl's general string-handling and are not part of its pat- tern matching engine. </P ></LI ><LI ><P > The Perl \G assertion is not supported as it is not relevant to single pattern matches. </P ></LI ><LI ><P > Fairly obviously, PCRE does not support the (?{code}) construction. </P ></LI ><LI ><P > There are at the time of writing some oddities in Perl 5.005_02 concerned with the settings of captured strings when part of a pattern is repeated. For example, matching "aba" against the pattern /^(a(b)?)+$/ sets $2 to the value "b", but matching "aabbaa" against /^(aa(bb)?)+$/ leaves $2 unset. However, if the pattern is changed to /^(aa(b(b))?)+$/ then $2 (and $3) get set. In Perl 5.004 $2 is set in both cases, and that is also <TT CLASS="constant" ><B >TRUE</B ></TT > of PCRE. If in the future Perl changes to a consistent state that is different, PCRE may change to follow. </P ></LI ><LI ><P > Another as yet unresolved discrepancy is that in Perl 5.005_02 the pattern /^(a)?(?(1)a|b)+$/ matches the string "a", whereas in PCRE it does not. However, in both Perl and PCRE /^(a)?a/ matched against "a" leaves $1 unset. </P ></LI ><LI ><P > PCRE provides some extensions to the Perl regular expression facilities: <P ></P ><OL TYPE="a" ><LI ><P > Although lookbehind assertions must match fixed length strings, each alternative branch of a lookbehind assertion can match a different length of string. Perl 5.005 requires them all to have the same length. </P ></LI ><LI ><P > If <A HREF="pcre.pattern.modifiers.html" >PCRE_DOLLAR_ENDONLY</A > is set and <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > is not set, the $ meta- character matches only at the very end of the string. </P ></LI ><LI ><P > If <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTRA</A > is set, a backslash followed by a letter with no special meaning is faulted. </P ></LI ><LI ><P > If <A HREF="pcre.pattern.modifiers.html" >PCRE_UNGREEDY</A > is set, the greediness of the repeti- tion quantifiers is inverted, that is, by default they are not greedy, but if followed by a question mark they are. </P ></LI ></OL > </P ></LI ></OL > </P ></DIV ><DIV CLASS="refsect1" ><A NAME="regexp.reference" ></A ><H2 >Regular Expression Details</H2 ><DIV CLASS="refsect2" ><A NAME="regexp.introduction" ></A ><H3 >Introduction</H3 ><P > The syntax and semantics of the regular expressions sup- ported by PCRE are described below. Regular expressions are also described in the Perl documentation and in a number of other books, some of which have copious examples. Jeffrey Friedl's "Mastering Regular Expressions", published by O'Reilly (ISBN 1-56592-257-3), covers them in great detail. The description here is intended as reference documentation. 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 charac- ters in the subject. As a trivial example, the pattern <TT CLASS="literal" >The quick brown fox</TT > matches a portion of a subject string that is identical to itself. </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.meta" ></A ><H3 >Meta-characters</H3 ><P > The power of regular expressions comes from the ability to include alternatives and repetitions in the pat- tern. These are encoded in the pattern by the use of <SPAN CLASS="emphasis" ><I CLASS="emphasis" >meta</I ></SPAN >- <SPAN CLASS="emphasis" ><I CLASS="emphasis" >characters</I ></SPAN >, which do not stand for themselves but instead are interpreted in some special way. </P ><P > There are two different sets of meta-characters: those that are recognized anywhere in the pattern except within square brackets, and those that are recognized in square brackets. Outside square brackets, the meta-characters are as follows: <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\</I ></SPAN ></DT ><DD ><P > general escape character with several uses </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >^</I ></SPAN ></DT ><DD ><P > assert start of subject (or line, in multiline mode) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >$</I ></SPAN ></DT ><DD ><P > assert end of subject (or line, in multiline mode) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >.</I ></SPAN ></DT ><DD ><P > match any character except newline (by default) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >[</I ></SPAN ></DT ><DD ><P > start character class definition </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >]</I ></SPAN ></DT ><DD ><P > end character class definition </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >|</I ></SPAN ></DT ><DD ><P > start of alternative branch </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >(</I ></SPAN ></DT ><DD ><P > start subpattern </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >)</I ></SPAN ></DT ><DD ><P > end subpattern </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >?</I ></SPAN ></DT ><DD ><P > extends the meaning of (, also 0 or 1 quantifier, also quantifier minimizer </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >*</I ></SPAN ></DT ><DD ><P > 0 or more quantifier </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >+</I ></SPAN ></DT ><DD ><P > 1 or more quantifier </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >{</I ></SPAN ></DT ><DD ><P > start min/max quantifier </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >}</I ></SPAN ></DT ><DD ><P > end min/max quantifier </P ></DD ></DL ></DIV > Part of a pattern that is in square brackets is called a "character class". In a character class the only meta- characters are: <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\</I ></SPAN ></DT ><DD ><P > general escape character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >^</I ></SPAN ></DT ><DD ><P > negate the class, but only if the first character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >-</I ></SPAN ></DT ><DD ><P > indicates character range </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >]</I ></SPAN ></DT ><DD ><P > terminates the character class </P ></DD ></DL ></DIV > The following sections describe the use of each of the meta-characters. </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.backslash" ></A ><H3 >backslash</H3 ><P > The backslash character has several uses. Firstly, if it is followed by a non-alphanumeric character, it takes away any special meaning that character may 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 applies whether or not the follow- ing character would otherwise be interpreted as a meta- character, so it is always safe to precede a non-alphanumeric with "\" to specify that it stands for itself. In particu- lar, if you want to match a backslash, you write "\\". </P ><P > If a pattern is compiled with the <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > option, whi- tespace in the pattern (other than in a character class) and characters between a "#" outside a character class and the next newline character are ignored. An escaping backslash can be used to include a whitespace or "#" character as part of the pattern. </P ><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, but when a pattern is being prepared by text editing, it is usually easier to use one of the following escape sequences than the binary character it represents: </P ><P > <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\a</I ></SPAN ></DT ><DD ><P > alarm, that is, the BEL character (hex 07) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\cx</I ></SPAN ></DT ><DD ><P > "control-x", where x is any character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\e</I ></SPAN ></DT ><DD ><P > escape (hex 1B) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\f</I ></SPAN ></DT ><DD ><P > formfeed (hex 0C) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\n</I ></SPAN ></DT ><DD ><P > newline (hex 0A) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\r</I ></SPAN ></DT ><DD ><P > carriage return (hex 0D) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\t</I ></SPAN ></DT ><DD ><P > tab (hex 09) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\xhh</I ></SPAN ></DT ><DD ><P > character with hex code hh </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\ddd</I ></SPAN ></DT ><DD ><P > character with octal code ddd, or backreference </P ></DD ></DL ></DIV > </P ><P > The precise effect of "<TT CLASS="literal" >\cx</TT >" is as follows: if "<TT CLASS="literal" >x</TT >" is a lower case letter, it is converted to upper case. Then bit 6 of the character (hex 40) is inverted. Thus "<TT CLASS="literal" >\cz</TT >" becomes hex 1A, but "<TT CLASS="literal" >\c{</TT >" becomes hex 3B, while "<TT CLASS="literal" >\c;</TT >" becomes hex 7B. </P ><P > After "<TT CLASS="literal" >\x</TT >", up to two hexadecimal digits are read (letters can be in upper or lower case). </P ><P > After "<TT CLASS="literal" >\0</TT >" up to two further octal digits are read. In both cases, if there are fewer than two digits, just those that are present are used. Thus the sequence "<TT CLASS="literal" >\0\x\07</TT >" specifies two binary zeros followed by a BEL character. Make sure you supply two digits after the initial zero if the 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 less than 10, or if there have been at least that many previous capturing left parentheses in the expression, the entire sequence is taken as a <SPAN CLASS="emphasis" ><I CLASS="emphasis" >back</I ></SPAN > <SPAN CLASS="emphasis" ><I CLASS="emphasis" >reference</I ></SPAN >. A description of how this works is given later, following the discussion of parenthesized subpatterns. </P ><P > Inside a character class, or if the decimal number is greater than 9 and there have not been that many capturing subpatterns, PCRE re-reads up to three octal digits following the backslash, and generates a single byte from the least significant 8 bits of the value. Any subsequent digits stand for themselves. For example: </P ><P > <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\040</I ></SPAN ></DT ><DD ><P > is another way of writing a space </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\40</I ></SPAN ></DT ><DD ><P > is the same, provided there are fewer than 40 previous capturing subpatterns </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\7</I ></SPAN ></DT ><DD ><P > is always a back reference </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\11</I ></SPAN ></DT ><DD ><P > might be a back reference, or another way of writing a tab </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\011</I ></SPAN ></DT ><DD ><P > is always a tab </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\0113</I ></SPAN ></DT ><DD ><P > is a tab followed by the character "3" </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\113</I ></SPAN ></DT ><DD ><P > is the character with octal code 113 (since there can be no more than 99 back references) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\377</I ></SPAN ></DT ><DD ><P > is a byte consisting entirely of 1 bits </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\81</I ></SPAN ></DT ><DD ><P > is either a back reference, or a binary zero followed by the two characters "8" and "1" </P ></DD ></DL ></DIV > </P ><P > Note that octal values of 100 or greater must not be intro- duced by a leading zero, because no more than three octal digits are ever read. </P ><P > All the sequences that define a single byte value can be used both inside and outside character classes. In addition, inside a character class, the sequence "<TT CLASS="literal" >\b</TT >" is interpreted as the backspace character (hex 08). Outside a character class it has a different meaning (see below). </P ><P > The third use of backslash is for specifying generic charac- ter types: </P ><P > <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\d</I ></SPAN ></DT ><DD ><P > any decimal digit </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\D</I ></SPAN ></DT ><DD ><P > any character that is not a decimal digit </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\s</I ></SPAN ></DT ><DD ><P > any whitespace character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\S</I ></SPAN ></DT ><DD ><P > any character that is not a whitespace character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\w</I ></SPAN ></DT ><DD ><P > any "word" character </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\W</I ></SPAN ></DT ><DD ><P > any "non-word" character </P ></DD ></DL ></DIV > </P ><P > Each pair of escape sequences partitions the complete set of characters into two disjoint sets. Any given character matches one, and only one, of each pair. </P ><P > A "word" character is any letter or digit or the underscore character, that is, any character which can be part of a Perl "<TT CLASS="literal" >word</TT >". The definition of letters and digits is controlled by PCRE's character tables, and may vary if locale-specific matching is taking place (see "Locale support" above). For example, in the "fr" (French) locale, some char- acter codes greater than 128 are used for accented letters, and these are matched by <TT CLASS="literal" >\w</TT >. </P ><P > These character type 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 of them fail, since there is no character to match. </P ><P > The fourth use of backslash is for certain simple asser- tions. An assertion specifies a condition that has to 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 backslashed assertions are </P ><P > <P ></P ><DIV CLASS="variablelist" ><DL ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\b</I ></SPAN ></DT ><DD ><P > word boundary </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\B</I ></SPAN ></DT ><DD ><P > not a word boundary </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\A</I ></SPAN ></DT ><DD ><P > start of subject (independent of multiline mode) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\Z</I ></SPAN ></DT ><DD ><P > end of subject or newline at end (independent of multiline mode) </P ></DD ><DT ><SPAN CLASS="emphasis" ><I CLASS="emphasis" >\z</I ></SPAN ></DT ><DD ><P > end of subject(independent of multiline mode) </P ></DD ></DL ></DIV > </P ><P > These assertions may not appear in character classes (but note that "<TT CLASS="literal" >\b</TT >" has a different meaning, namely the backspace character, inside a character class). </P ><P > A word boundary is a position in the subject string where the current character and the previous character do not both match <TT CLASS="literal" >\w</TT > or <TT CLASS="literal" >\W</TT > (i.e. one matches <TT CLASS="literal" >\w</TT > and the other matches <TT CLASS="literal" >\W</TT >), or the start or end of the string if the first or last character matches \w, respectively. </P ><P > The <TT CLASS="literal" >\A</TT >, <TT CLASS="literal" >\Z</TT >, and <TT CLASS="literal" >\z</TT > assertions differ from the traditional circumflex and dollar (described below) in that they only ever match at the very start and end of the subject string, whatever options are set. They are not affected by the <A HREF="pcre.pattern.modifiers.html" >PCRE_NOTBOL</A > or <A HREF="pcre.pattern.modifiers.html" >PCRE_NOTEOL</A > options. The difference between <TT CLASS="literal" >\Z</TT > and <TT CLASS="literal" >\z</TT > is that <TT CLASS="literal" >\Z</TT > matches before a newline that is the last character of the string as well as at the end of the string, whereas <TT CLASS="literal" >\z</TT > matches only at the end. </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.circudollar" ></A ><H3 >Circumflex and dollar</H3 ><P CLASS="literallayout" ><br> Outside a character class, in the default matching mode, the<br> circumflex character is an assertion which is true only if<br> the current matching point is at the start of the subject<br> string. Inside a character class, circumflex has an entirely<br> different meaning (see below).<br> <br> Circumflex need not be the first character of the pattern if<br> a number of alternatives are involved, but it should be the<br> first thing in each alternative in which it appears if the<br> pattern is ever to match that branch. If all possible alter-<br> natives start with a circumflex, that is, if the pattern is<br> constrained to match only at the start of the subject, it is<br> said to be an "anchored" pattern. (There are also other con-<br> structs that can cause a pattern to be anchored.)<br> <br> A dollar character is an assertion which is <TT CLASS="constant" ><B >TRUE</B ></TT > only if the<br> current matching point is at the end of the subject string,<br> or immediately before a newline character that is the last<br> character in the string (by default). Dollar need not be the<br> last character of the pattern if a number of alternatives<br> are involved, but it should be the last item in any branch<br> in which it appears. Dollar has no special meaning in a<br> character class.<br> <br> The meaning of dollar can be changed so that it matches only<br> at the very end of the string, by setting the<br> <A HREF="pcre.pattern.modifiers.html" >PCRE_DOLLAR_ENDONLY</A ><br> option at compile or matching time. This<br> does not affect the \Z assertion.<br> <br> The meanings of the circumflex and dollar characters are<br> changed if the <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > option is set. When this is<br> the case, they match immediately after and immediately<br> before an internal "\n" character, respectively, in addition<br> to matching at the start and end of the subject string. For<br> example, the pattern /^abc$/ matches the subject string<br> "def\nabc" in multiline mode, but not otherwise. Conse-<br> quently, patterns that are anchored in single line mode<br> because all branches start with "^" are not anchored in mul-<br> tiline mode. The <A HREF="pcre.pattern.modifiers.html" >PCRE_DOLLAR_ENDONLY</A > option is ignored if<br> <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > is set.<br> <br> Note that the sequences \A, \Z, and \z can be used to match<br> the start and end of the subject in both modes, and if all<br> branches of a pattern start with \A is it always anchored,<br> whether <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > is set or not.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.dot" ></A ><H3 >FULL STOP</H3 ><P CLASS="literallayout" ><br> Outside a character class, a dot in the pattern matches any<br> one character in the subject, including a non-printing<br> character, but not (by default) newline. If the <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > <br> option is set, then dots match newlines as well. The han-<br> dling of dot is entirely independent of the handling of cir-<br> cumflex and dollar, the only relationship being that they<br> both involve newline characters. Dot has no special meaning<br> in a character class.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.squarebrackets" ></A ><H3 >Square brackets</H3 ><P CLASS="literallayout" ><br> An opening square bracket introduces a character class, ter-<br> minated by a closing square bracket. A closing square<br> bracket on its own is not special. If a closing square<br> bracket is required as a member of the class, it should be<br> the first data character in the class (after an initial cir-<br> cumflex, if present) or escaped with a backslash.<br> <br> A character class matches a single character in the subject;<br> the character must be in the set of characters defined by<br> the class, unless the first character in the class is a cir-<br> cumflex, in which case the subject character must not be in<br> the set defined by the class. If a circumflex is actually<br> required as a member of the class, ensure it is not the<br> first character, or escape it with a backslash.<br> <br> For example, the character class [aeiou] matches any lower<br> case vowel, while [^aeiou] matches any character that is not<br> a lower case vowel. Note that a circumflex is just a con-<br> venient notation for specifying the characters which are in<br> the class by enumerating those that are not. It is not an<br> assertion: it still consumes a character from the subject<br> string, and fails if the current pointer is at the end of<br> the string.<br> <br> When caseless matching is set, any letters in a class<br> represent both their upper case and lower case versions, so<br> for example, a caseless [aeiou] matches "A" as well as "a",<br> and a caseless [^aeiou] does not match "A", whereas a case-<br> ful version would.<br> <br> The newline character is never treated in any special way in<br> character classes, whatever the setting of the <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > <br> or <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > options is. A class such as [^a] will<br> always match a newline.<br> <br> The minus (hyphen) character can be used to specify a range<br> of characters in a character class. For example, [d-m]<br> matches any letter between d and m, inclusive. If a minus<br> character is required in a class, it must be escaped with a<br> backslash or appear in a position where it cannot be inter-<br> preted as indicating a range, typically as the first or last<br> character in the class.<br> <br> It is not possible to have the literal character "]" as the<br> end character of a range. A pattern such as [W-]46] is<br> interpreted as a class of two characters ("W" and "-") fol-<br> lowed by a literal string "46]", so it would match "W46]" or<br> "-46]". However, if the "]" is escaped with a backslash it<br> is interpreted as the end of range, so [W-\]46] is inter-<br> preted as a single class containing a range followed by two<br> separate characters. The octal or hexadecimal representation<br> of "]" can also be used to end a range.<br> <br> Ranges operate in ASCII collating sequence. They can also be<br> used for characters specified numerically, for example<br> [\000-\037]. If a range that includes letters is used when<br> caseless matching is set, it matches the letters in either<br> case. For example, [W-c] is equivalent to [][\^_`wxyzabc],<br> matched caselessly, and if character tables for the "fr"<br> locale are in use, [\xc8-\xcb] matches accented E characters<br> in both cases.<br> <br> The character types \d, \D, \s, \S, \w, and \W may also<br> appear in a character class, and add the characters that<br> they match to the class. For example, [\dABCDEF] matches any<br> hexadecimal digit. A circumflex can conveniently be used<br> with the upper case character types to specify a more res-<br> tricted set of characters than the matching lower case type.<br> For example, the class [^\W_] matches any letter or digit,<br> but not underscore.<br> <br> All non-alphanumeric characters other than \, -, ^ (at the<br> start) and the terminating ] are non-special in character<br> classes, but it does no harm if they are escaped.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.verticalbar" ></A ><H3 >Vertical bar</H3 ><P CLASS="literallayout" ><br> Vertical bar characters are used to separate alternative<br> patterns. For example, the pattern<br> <br> gilbert|sullivan<br> <br> matches either "gilbert" or "sullivan". Any number of alternatives<br> may appear, and an empty alternative is permitted<br> (matching the empty string). The matching process tries<br> each alternative in turn, from left to right, and the first<br> one that succeeds is used. If the alternatives are within a<br> subpattern (defined below), "succeeds" means matching the<br> rest of the main pattern as well as the alternative in the<br> subpattern.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.internal-options" ></A ><H3 >Internal option setting</H3 ><P CLASS="literallayout" ><br> The settings of <A HREF="pcre.pattern.modifiers.html" >PCRE_CASELESS</A > , <br> <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > , <br> <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > ,<br> and <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > can be changed from within the pattern by<br> a sequence of Perl option letters enclosed between "(?" and<br> ")". The option letters are<br> <br> i for <A HREF="pcre.pattern.modifiers.html" >PCRE_CASELESS</A > <br> m for <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > <br> s for <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > <br> x for <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > <br> <br> For example, (?im) sets caseless, multiline matching. It is<br> also possible to unset these options by preceding the letter<br> with a hyphen, and a combined setting and unsetting such as<br> (?im-sx), which sets <A HREF="pcre.pattern.modifiers.html" >PCRE_CASELESS</A > and <A HREF="pcre.pattern.modifiers.html" >PCRE_MULTILINE</A > while<br> unsetting <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > and <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > , is also permitted.<br> If a letter appears both before and after the hyphen, the<br> option is unset.<br> <br> The scope of these option changes depends on where in the<br> pattern the setting occurs. For settings that are outside<br> any subpattern (defined below), the effect is the same as if<br> the options were set or unset at the start of matching. The<br> following patterns all behave in exactly the same way:<br> <br> (?i)abc<br> a(?i)bc<br> ab(?i)c<br> abc(?i)<br> <br> which in turn is the same as compiling the pattern abc with<br> <A HREF="pcre.pattern.modifiers.html" >PCRE_CASELESS</A > set.<br> In other words, such "top level" settings apply to the whole<br> pattern (unless there are other changes inside subpatterns).<br> If there is more than one setting of the same option at top level,<br> the rightmost setting is used.<br> <br> If an option change occurs inside a subpattern, the effect<br> is different. This is a change of behaviour in Perl 5.005.<br> An option change inside a subpattern affects only that part<br> of the subpattern that follows it, so<br> <br> (a(?i)b)c<br> <br> matches abc and aBc and no other strings (assuming<br> <A HREF="pcre.pattern.modifiers.html" >PCRE_CASELESS</A > is not used). By this means, options can be<br> made to have different settings in different parts of the<br> pattern. Any changes made in one alternative do carry on<br> into subsequent branches within the same subpattern. For<br> example,<br> <br> (a(?i)b|c)<br> <br> matches "ab", "aB", "c", and "C", even though when matching<br> "C" the first branch is abandoned before the option setting.<br> This is because the effects of option settings happen at<br> compile time. There would be some very weird behaviour otherwise.<br> <br> The PCRE-specific options <A HREF="pcre.pattern.modifiers.html" >PCRE_UNGREEDY</A > and <br> <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTRA</A > can<br> be changed in the same way as the Perl-compatible options by<br> using the characters U and X respectively. The (?X) flag<br> setting is special in that it must always occur earlier in<br> the pattern than any of the additional features it turns on,<br> even when it is at top level. It is best put at the start.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.subpatterns" ></A ><H3 >subpatterns</H3 ><P CLASS="literallayout" ><br> Subpatterns are delimited by parentheses (round brackets),<br> which can be nested. Marking part of a pattern as a subpattern<br> does two things:<br> <br> 1. It localizes a set of alternatives. For example, the pat-<br> tern<br> <br> cat(aract|erpillar|)<br> <br> matches one of the words "cat", "cataract", or "caterpillar".<br> Without the parentheses, it would match "cataract",<br> "erpillar" or the empty string.<br> <br> 2. It sets up the subpattern as a capturing subpattern (as<br> defined above). When the whole pattern matches, that portion<br> of the subject string that matched the subpattern is<br> passed back to the caller via the <SPAN CLASS="emphasis" ><I CLASS="emphasis" >ovector</I ></SPAN ><br> argument of<br> <B CLASS="function" >pcre_exec()</B >. Opening parentheses are counted<br> from left to right (starting from 1) to obtain the numbers of the<br> capturing subpatterns.<br> <br> For example, if the string "the red king" is matched against<br> the pattern<br> <br> the ((red|white) (king|queen))<br> <br> the captured substrings are "red king", "red", and "king",<br> and are numbered 1, 2, and 3.<br> <br> The fact that plain parentheses fulfil two functions is not<br> always helpful. There are often times when a grouping subpattern<br> is required without a capturing requirement. If an<br> opening parenthesis is followed by "?:", the subpattern does<br> not do any capturing, and is not counted when computing the<br> number of any subsequent capturing subpatterns. For example,<br> if the string "the white queen" is matched against the<br> pattern<br> <br> the ((?:red|white) (king|queen))<br> <br> the captured substrings are "white queen" and "queen", and<br> are numbered 1 and 2. The maximum number of captured substrings<br> is 99, and the maximum number of all subpatterns,<br> both capturing and non-capturing, is 200.<br> <br> As a convenient shorthand, if any option settings are<br> required at the start of a non-capturing subpattern, the<br> option letters may appear between the "?" and the ":". Thus<br> the two patterns<br> <br> (?i:saturday|sunday)<br> (?:(?i)saturday|sunday)<br> <br> match exactly the same set of strings. Because alternative<br> branches are tried from left to right, and options are not<br> reset until the end of the subpattern is reached, an option<br> setting in one branch does affect subsequent branches, so<br> the above patterns match "SUNDAY" as well as "Saturday".<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.repetition" ></A ><H3 >Repetition</H3 ><P CLASS="literallayout" ><br> Repetition is specified by quantifiers, which can follow any<br> of the following items:<br> <br> a single character, possibly escaped<br> the . metacharacter<br> a character class<br> a back reference (see next section)<br> a parenthesized subpattern (unless it is an assertion -<br> see below)<br> <br> The general repetition quantifier specifies a minimum and<br> maximum number of permitted matches, by giving the two<br> numbers in curly brackets (braces), separated by a comma.<br> The numbers must be less than 65536, and the first must be<br> less than or equal to the second. For example:<br> <br> z{2,4}<br> <br> matches "zz", "zzz", or "zzzz". A closing brace on its own<br> is not a special character. If the second number is omitted,<br> but the comma is present, there is no upper limit; if the<br> second number and the comma are both omitted, the quantifier<br> specifies an exact number of required matches. Thus<br> <br> [aeiou]{3,}<br> <br> matches at least 3 successive vowels, but may match many<br> more, while<br> <br> \d{8}<br> <br> matches exactly 8 digits. An opening curly bracket that<br> appears in a position where a quantifier is not allowed, or<br> one that does not match the syntax of a quantifier, is taken<br> as a literal character. For example, {,6} is not a quantifier,<br> but a literal string of four characters.<br> <br> The quantifier {0} is permitted, causing the expression to<br> behave as if the previous item and the quantifier were not<br> present.<br> <br> For convenience (and historical compatibility) the three<br> most common quantifiers have single-character abbreviations:<br> <br> * is equivalent to {0,}<br> + is equivalent to {1,}<br> ? is equivalent to {0,1}<br> <br> It is possible to construct infinite loops by following a<br> subpattern that can match no characters with a quantifier<br> that has no upper limit, for example:<br> <br> (a?)*<br> <br> Earlier versions of Perl and PCRE used to give an error at<br> compile time for such patterns. However, because there are<br> cases where this can be useful, such patterns are now<br> accepted, but if any repetition of the subpattern does in<br> fact match no characters, the loop is forcibly broken.<br> <br> By default, the quantifiers are "greedy", that is, they<br> match as much as possible (up to the maximum number of permitted<br> times), without causing the rest of the pattern to<br> fail. The classic example of where this gives problems is in<br> trying to match comments in C programs. These appear between<br> the sequences /* and */ and within the sequence, individual<br> * and / characters may appear. An attempt to match C comments<br> by applying the pattern<br> <br> /\*.*\*/<br> <br> to the string<br> <br> /* first command */ not comment /* second comment */<br> <br> fails, because it matches the entire string due to the<br> greediness of the .* item.<br> <br> However, if a quantifier is followed by a question mark,<br> then it ceases to be greedy, and instead matches the minimum<br> number of times possible, so the pattern<br> <br> /\*.*?\*/<br> <br> does the right thing with the C comments. The meaning of the<br> various quantifiers is not otherwise changed, just the preferred<br> number of matches. Do not confuse this use of ques-<br> tion mark with its use as a quantifier in its own right.<br> Because it has two uses, it can sometimes appear doubled, as<br> in<br> <br> \d??\d<br> <br> which matches one digit by preference, but can match two if<br> that is the only way the rest of the pattern matches.<br> <br> If the <A HREF="pcre.pattern.modifiers.html" >PCRE_UNGREEDY</A > option is set (an option which is not<br> available in Perl) then the quantifiers are not greedy by<br> default, but individual ones can be made greedy by following<br> them with a question mark. In other words, it inverts the<br> default behaviour.<br> <br> When a parenthesized subpattern is quantified with a minimum<br> repeat count that is greater than 1 or with a limited maximum,<br> more store is required for the compiled pattern, in<br> proportion to the size of the minimum or maximum.<br> <br> If a pattern starts with .* or .{0,} and the <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > <br> option (equivalent to Perl's /s) is set, thus allowing the .<br> to match newlines, then the pattern is implicitly anchored,<br> because whatever follows will be tried against every character<br> position in the subject string, so there is no point in<br> retrying the overall match at any position after the first.<br> PCRE treats such a pattern as though it were preceded by \A.<br> In cases where it is known that the subject string contains<br> no newlines, it is worth setting <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > when the pattern begins with .* in order to<br> obtain this optimization, or<br> alternatively using ^ to indicate anchoring explicitly.<br> <br> When a capturing subpattern is repeated, the value captured<br> is the substring that matched the final iteration. For example, after<br> <br> (tweedle[dume]{3}\s*)+<br> <br> has matched "tweedledum tweedledee" the value of the captured<br> substring is "tweedledee". However, if there are<br> nested capturing subpatterns, the corresponding captured<br> values may have been set in previous iterations. For example,<br> after<br> <br> /(a|(b))+/<br> <br> matches "aba" the value of the second captured substring is<br> "b".<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.back-references" ></A ><H3 >BACK REFERENCES</H3 ><P CLASS="literallayout" ><br> Outside a character class, a backslash followed by a digit<br> greater than 0 (and possibly further digits) is a back<br> reference to a capturing subpattern earlier (i.e. to its<br> left) in the pattern, provided there have been that many<br> previous capturing left parentheses.<br> <br> However, if the decimal number following the backslash is<br> less than 10, it is always taken as a back reference, and<br> causes an error only if there are not that many capturing<br> left parentheses in the entire pattern. In other words, the<br> parentheses that are referenced need not be to the left of<br> the reference for numbers less than 10. See the section<br> entitled "Backslash" above for further details of the handling<br> of digits following a backslash.<br> <br> A back reference matches whatever actually matched the capturing<br> subpattern in the current subject string, rather than<br> anything matching the subpattern itself. So the pattern<br> <br> (sens|respons)e and \1ibility<br> <br> matches "sense and sensibility" and "response and responsibility",<br> but not "sense and responsibility". If caseful<br> matching is in force at the time of the back reference, then<br> the case of letters is relevant. For example,<br> <br> ((?i)rah)\s+\1<br> <br> matches "rah rah" and "RAH RAH", but not "RAH rah", even<br> though the original capturing subpattern is matched caselessly.<br> <br> There may be more than one back reference to the same subpattern.<br> If a subpattern has not actually been used in a<br> particular match, then any back references to it always<br> fail. For example, the pattern<br> <br> (a|(bc))\2<br> <br> always fails if it starts to match "a" rather than "bc".<br> Because there may be up to 99 back references, all digits<br> following the backslash are taken as part of a potential<br> back reference number. If the pattern continues with a digit<br> character, then some delimiter must be used to terminate the<br> back reference. If the <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > option is set, this can<br> be whitespace. Otherwise an empty comment can be used.<br> <br> A back reference that occurs inside the parentheses to which<br> it refers fails when the subpattern is first used, so, for<br> example, (a\1) never matches. However, such references can<br> be useful inside repeated subpatterns. For example, the pattern<br> <br> (a|b\1)+<br> <br> matches any number of "a"s and also "aba", "ababaa" etc. At<br> each iteration of the subpattern, the back reference matches<br> the character string corresponding to the previous iteration.<br> In order for this to work, the pattern must be such<br> that the first iteration does not need to match the back<br> reference. This can be done using alternation, as in the<br> example above, or by a quantifier with a minimum of zero.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.assertions" ></A ><H3 >Assertions</H3 ><P CLASS="literallayout" ><br> An assertion is a test on the characters following or<br> preceding the current matching point that does not actually<br> consume any characters. The simple assertions coded as \b,<br> \B, \A, \Z, \z, ^ and $ are described above. More complicated<br> assertions are coded as subpatterns. There are two<br> kinds: those that look ahead of the current position in the<br> subject string, and those that look behind it.<br> <br> An assertion subpattern is matched in the normal way, except<br> that it does not cause the current matching position to be<br> changed. Lookahead assertions start with (?= for positive<br> assertions and (?! for negative assertions. For example,<br> <br> \w+(?=;)<br> <br> matches a word followed by a semicolon, but does not include<br> the semicolon in the match, and<br> <br> foo(?!bar)<br> <br> matches any occurrence of "foo" that is not followed by<br> "bar". Note that the apparently similar pattern<br> <br> (?!foo)bar<br> <br> does not find an occurrence of "bar" that is preceded by<br> something other than "foo"; it finds any occurrence of "bar"<br> whatsoever, because the assertion (?!foo) is always <TT CLASS="constant" ><B >TRUE</B ></TT ><br> when the next three characters are "bar". A lookbehind<br> assertion is needed to achieve this effect.<br> <br> Lookbehind assertions start with (?<= for positive assertions<br> and (?<! for negative assertions. For example,<br> <br> (?<!foo)bar<br> <br> does find an occurrence of "bar" that is not preceded by<br> "foo". The contents of a lookbehind assertion are restricted<br> such that all the strings it matches must have a fixed<br> length. However, if there are several alternatives, they do<br> not all have to have the same fixed length. Thus<br> <br> (?<=bullock|donkey)<br> <br> is permitted, but<br> <br> (?<!dogs?|cats?)<br> <br> causes an error at compile time. Branches that match different<br> length strings are permitted only at the top level of<br> a lookbehind assertion. This is an extension compared with<br> Perl 5.005, which requires all branches to match the same<br> length of string. An assertion such as<br> <br> (?<=ab(c|de))<br> <br> is not permitted, because its single top-level branch can<br> match two different lengths, but it is acceptable if rewritten<br> to use two top-level branches:<br> <br> (?<=abc|abde)<br> <br> The implementation of lookbehind assertions is, for each<br> alternative, to temporarily move the current position back<br> by the fixed width and then try to match. If there are<br> insufficient characters before the current position, the<br> match is deemed to fail. Lookbehinds in conjunction with<br> once-only subpatterns can be particularly useful for matching<br> at the ends of strings; an example is given at the end<br> of the section on once-only subpatterns.<br> <br> Several assertions (of any sort) may occur in succession.<br> For example,<br> <br> (?<=\d{3})(?<!999)foo<br> <br> matches "foo" preceded by three digits that are not "999".<br> Notice that each of the assertions is applied independently<br> at the same point in the subject string. First there is a<br> check that the previous three characters are all digits,<br> then there is a check that the same three characters are not<br> "999". This pattern does not match "foo" preceded by six<br> characters, the first of which are digits and the last three<br> of which are not "999". For example, it doesn't match<br> "123abcfoo". A pattern to do that is<br> <br> (?<=\d{3}...)(?<!999)foo<br> <br> This time the first assertion looks at the preceding six<br> characters, checking that the first three are digits, and<br> then the second assertion checks that the preceding three<br> characters are not "999".<br> <br> Assertions can be nested in any combination. For example,<br> <br> (?<=(?<!foo)bar)baz<br> <br> matches an occurrence of "baz" that is preceded by "bar"<br> which in turn is not preceded by "foo", while<br> <br> (?<=\d{3}(?!999)...)foo<br> <br> is another pattern which matches "foo" preceded by three<br> digits and any three characters that are not "999".<br> <br> Assertion subpatterns are not capturing subpatterns, and may<br> not be repeated, because it makes no sense to assert the<br> same thing several times. If any kind of assertion contains<br> capturing subpatterns within it, these are counted for the<br> purposes of numbering the capturing subpatterns in the whole<br> pattern. However, substring capturing is carried out only<br> for positive assertions, because it does not make sense for<br> negative assertions.<br> <br> Assertions count towards the maximum of 200 parenthesized<br> subpatterns.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.onlyonce" ></A ><H3 >Once-only subpatterns</H3 ><P CLASS="literallayout" ><br> With both maximizing and minimizing repetition, failure of<br> what follows normally causes the repeated item to be re-<br> evaluated to see if a different number of repeats allows the<br> rest of the pattern to match. Sometimes it is useful to<br> prevent this, either to change the nature of the match, or<br> to cause it fail earlier than it otherwise might, when the<br> author of the pattern knows there is no point in carrying<br> on.<br> <br> Consider, for example, the pattern \d+foo when applied to<br> the subject line<br> <br> 123456bar<br> <br> After matching all 6 digits and then failing to match "foo",<br> the normal action of the matcher is to try again with only 5<br> digits matching the \d+ item, and then with 4, and so on,<br> before ultimately failing. Once-only subpatterns provide the<br> means for specifying that once a portion of the pattern has<br> matched, it is not to be re-evaluated in this way, so the<br> matcher would give up immediately on failing to match "foo"<br> the first time. The notation is another kind of special<br> parenthesis, starting with (?> as in this example:<br> <br> (?>\d+)bar<br> <br> This kind of parenthesis "locks up" the part of the pattern<br> it contains once it has matched, and a failure further into<br> the pattern is prevented from backtracking into it. Back-<br> tracking past it to previous items, however, works as normal.<br> <br> An alternative description is that a subpattern of this type<br> matches the string of characters that an identical standalone<br> pattern would match, if anchored at the current point<br> in the subject string.<br> <br> Once-only subpatterns are not capturing subpatterns. Simple<br> cases such as the above example can be thought of as a maximizing<br> repeat that must swallow everything it can. So,<br> while both \d+ and \d+? are prepared to adjust the number of<br> digits they match in order to make the rest of the pattern<br> match, (?>\d+) can only match an entire sequence of digits.<br> <br> This construction can of course contain arbitrarily complicated<br> subpatterns, and it can be nested.<br> <br> Once-only subpatterns can be used in conjunction with look-<br> behind assertions to specify efficient matching at the end<br> of the subject string. Consider a simple pattern such as<br> <br> abcd$<br> <br> when applied to a long string which does not match. Because<br> matching proceeds from left to right, PCRE will look for<br> each "a" in the subject and then see if what follows matches<br> the rest of the pattern. If the pattern is specified as<br> <br> ^.*abcd$<br> <br> then the initial .* matches the entire string at first, but<br> when this fails (because there is no following "a"), it<br> backtracks to match all but the last character, then all but<br> the last two characters, and so on. Once again the search<br> for "a" covers the entire string, from right to left, so we<br> are no better off. However, if the pattern is written as<br> <br> ^(?>.*)(?<=abcd)<br> <br> then there can be no backtracking for the .* item; it can<br> match only the entire string. The subsequent lookbehind<br> assertion does a single test on the last four characters. If<br> it fails, the match fails immediately. For long strings,<br> this approach makes a significant difference to the processing time.<br> <br> When a pattern contains an unlimited repeat inside a subpattern<br> that can itself be repeated an unlimited number of<br> times, the use of a once-only subpattern is the only way to<br> avoid some failing matches taking a very long time indeed.<br> The pattern<br> <br> (\D+|<\d+>)*[!?]<br> <br> matches an unlimited number of substrings that either consist<br> of non-digits, or digits enclosed in <>, followed by<br> either ! or ?. When it matches, it runs quickly. However, if<br> it is applied to<br> <br> aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa<br> <br> it takes a long time before reporting failure. This is<br> because the string can be divided between the two repeats in<br> a large number of ways, and all have to be tried. (The example<br> used [!?] rather than a single character at the end,<br> because both PCRE and Perl have an optimization that allows<br> for fast failure when a single character is used. They<br> remember the last single character that is required for a<br> match, and fail early if it is not present in the string.)<br> If the pattern is changed to<br> <br> ((?>\D+)|<\d+>)*[!?]<br> <br> sequences of non-digits cannot be broken, and failure happens quickly.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.conditional" ></A ><H3 >Conditional subpatterns</H3 ><P CLASS="literallayout" ><br> It is possible to cause the matching process to obey a subpattern <br> conditionally or to choose between two alternative<br> subpatterns, depending on the result of an assertion, or<br> whether a previous capturing subpattern matched or not. The<br> two possible forms of conditional subpattern are<br> <br> (?(condition)yes-pattern)<br> (?(condition)yes-pattern|no-pattern)<br> <br> If the condition is satisfied, the yes-pattern is used; otherwise<br> the no-pattern (if present) is used. If there are<br> more than two alternatives in the subpattern, a compile-time<br> error occurs.<br> <br> There are two kinds of condition. If the text between the<br> parentheses consists of a sequence of digits, then the<br> condition is satisfied if the capturing subpattern of that<br> number has previously matched. Consider the following pattern,<br> which contains non-significant white space to make it<br> more readable (assume the <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > option) and to<br> divide it into three parts for ease of discussion:<br> <br> ( \( )? [^()]+ (?(1) \) )<br> <br> The first part matches an optional opening parenthesis, and<br> if that character is present, sets it as the first captured<br> substring. The second part matches one or more characters<br> that are not parentheses. The third part is a conditional<br> subpattern that tests whether the first set of parentheses<br> matched or not. If they did, that is, if subject started<br> with an opening parenthesis, the condition is <TT CLASS="constant" ><B >TRUE</B ></TT >, and so<br> the yes-pattern is executed and a closing parenthesis is<br> required. Otherwise, since no-pattern is not present, the<br> subpattern matches nothing. In other words, this pattern<br> matches a sequence of non-parentheses, optionally enclosed<br> in parentheses.<br> <br> If the condition is not a sequence of digits, it must be an<br> assertion. This may be a positive or negative lookahead or<br> lookbehind assertion. Consider this pattern, again containing<br> non-significant white space, and with the two alternatives on<br> the second line:<br> <br> (?(?=[^a-z]*[a-z])<br> \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )<br> <br> The condition is a positive lookahead assertion that matches<br> an optional sequence of non-letters followed by a letter. In<br> other words, it tests for the presence of at least one<br> letter in the subject. If a letter is found, the subject is<br> matched against the first alternative; otherwise it is<br> matched against the second. This pattern matches strings in<br> one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are<br> letters and dd are digits.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.comments" ></A ><H3 >Comments</H3 ><P CLASS="literallayout" ><br> The sequence (?# marks the start of a comment which<br> continues up to the next closing parenthesis. Nested<br> parentheses are not permitted. The characters that make up a<br> comment play no part in the pattern matching at all.<br> <br> If the <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A > option is set, an unescaped # character<br> outside a character class introduces a comment that contin-<br> ues up to the next newline character in the pattern.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.recursive" ></A ><H3 >Recursive patterns</H3 ><P CLASS="literallayout" ><br> Consider the problem of matching a string in parentheses,<br> allowing for unlimited nested parentheses. Without the use<br> of recursion, the best that can be done is to use a pattern<br> that matches up to some fixed depth of nesting. It is not<br> possible to handle an arbitrary nesting depth. Perl 5.6 has<br> provided an experimental facility that allows regular<br> expressions to recurse (amongst other things). The special <br> item (?R) is provided for the specific case of recursion. <br> This PCRE pattern solves the parentheses problem (assume <br> the <A HREF="pcre.pattern.modifiers.html" >PCRE_EXTENDED</A ><br> option is set so that white space is <br> ignored):<br> <br> \( ( (?>[^()]+) | (?R) )* \)<br> <br> First it matches an opening parenthesis. Then it matches any<br> number of substrings which can either be a sequence of non-<br> parentheses, or a recursive match of the pattern itself<br> (i.e. a correctly parenthesized substring). Finally there is<br> a closing parenthesis.<br> <br> This particular example pattern contains nested unlimited<br> repeats, and so the use of a once-only subpattern for matching<br> strings of non-parentheses is important when applying<br> the pattern to strings that do not match. For example, when<br> it is applied to<br> <br> (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()<br> <br> it yields "no match" quickly. However, if a once-only subpattern<br> is not used, the match runs for a very long time<br> indeed because there are so many different ways the + and *<br> repeats can carve up the subject, and all have to be tested<br> before failure can be reported.<br> <br> The values set for any capturing subpatterns are those from<br> the outermost level of the recursion at which the subpattern<br> value is set. If the pattern above is matched against<br> <br> (ab(cd)ef)<br> <br> the value for the capturing parentheses is "ef", which is<br> the last value taken on at the top level. If additional<br> parentheses are added, giving<br> <br> \( ( ( (?>[^()]+) | (?R) )* ) \)<br> ^ ^<br> ^ ^ then the string they capture<br> is "ab(cd)ef", the contents of the top level parentheses. If<br> there are more than 15 capturing parentheses in a pattern,<br> PCRE has to obtain extra memory to store data during a<br> recursion, which it does by using pcre_malloc, freeing it<br> via pcre_free afterwards. If no memory can be obtained, it<br> saves data for the first 15 capturing parentheses only, as<br> there is no way to give an out-of-memory error from within a<br> recursion.<br> </P ></DIV ><DIV CLASS="refsect2" ><A NAME="regexp.reference.performances" ></A ><H3 >Performances</H3 ><P CLASS="literallayout" ><br> Certain items that may appear in patterns are more efficient<br> than others. It is more efficient to use a character class<br> like [aeiou] than a set of alternatives such as (a|e|i|o|u).<br> In general, the simplest construction that provides the<br> required behaviour is usually the most efficient. Jeffrey<br> Friedl's book contains a lot of discussion about optimizing<br> regular expressions for efficient performance.<br> <br> When a pattern begins with .* and the <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > option is<br> set, the pattern is implicitly anchored by PCRE, since it<br> can match only at the start of a subject string. However, if<br> <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > is not set, PCRE cannot make this optimization,<br> because the . metacharacter does not then match a newline,<br> and if the subject string contains newlines, the pattern may<br> match from the character immediately following one of them<br> instead of from the very start. For example, the pattern<br> <br> (.*) second<br> <br> matches the subject "first\nand second" (where \n stands for<br> a newline character) with the first captured substring being<br> "and". In order to do this, PCRE has to retry the match<br> starting after every newline in the subject.<br> <br> If you are using such a pattern with subject strings that do<br> not contain newlines, the best performance is obtained by<br> setting <A HREF="pcre.pattern.modifiers.html" >PCRE_DOTALL</A > , or starting the pattern with ^.* to<br> indicate explicit anchoring. That saves PCRE from having to<br> scan along the subject looking for a newline to restart at.<br> <br> Beware of patterns that contain nested indefinite repeats.<br> These can take a long time to run when applied to a string<br> that does not match. Consider the pattern fragment<br> <br> (a+)*<br> <br> This can match "aaaa" in 33 different ways, and this number<br> increases very rapidly as the string gets longer. (The *<br> repeat can match 0, 1, 2, 3, or 4 times, and for each of<br> those cases other than 0, the + repeats can match different<br> numbers of times.) When the remainder of the pattern is such<br> that the entire match is going to fail, PCRE has in principle<br> to try every possible variation, and this can take an<br> extremely long time.<br> <br> An optimization catches some of the more simple cases such<br> as<br> <br> (a+)*b<br> <br> where a literal character follows. Before embarking on the<br> standard matching procedure, PCRE checks that there is a "b"<br> later in the subject string, and if there is not, it fails<br> the match immediately. However, when there is no following<br> literal this optimization cannot be used. You can see the<br> difference by comparing the behaviour of<br> <br> (a+)*\d<br> <br> with the pattern above. The former gives a failure almost<br> instantly when applied to a whole line of "a" characters,<br> whereas the latter takes an appreciable time with strings<br> longer than about 20 characters.<br> </P ></DIV ></DIV ><DIV CLASS="NAVFOOTER" ><HR ALIGN="LEFT" WIDTH="100%"><TABLE SUMMARY="Footer navigation table" WIDTH="100%" BORDER="0" CELLPADDING="0" CELLSPACING="0" ><TR ><TD WIDTH="33%" ALIGN="left" VALIGN="top" ><A HREF="pcre.pattern.modifiers.html" ACCESSKEY="P" >Prev</A ></TD ><TD WIDTH="34%" ALIGN="center" VALIGN="top" ><A HREF="index.html" ACCESSKEY="H" >Home</A ></TD ><TD WIDTH="33%" ALIGN="right" VALIGN="top" ><A HREF="function.preg-grep.html" ACCESSKEY="N" >Next</A ></TD ></TR ><TR ><TD WIDTH="33%" ALIGN="left" VALIGN="top" >Pattern Modifiers</TD ><TD WIDTH="34%" ALIGN="center" VALIGN="top" ><A HREF="ref.pcre.html" ACCESSKEY="U" >Up</A ></TD ><TD WIDTH="33%" ALIGN="right" VALIGN="top" >preg_grep</TD ></TR ></TABLE ></DIV ></BODY ></HTML >