PCRE2PATTERN(3)
🌸 PCRE2PATTERN(3) 🌸 — PCRE2 regex details, bestie 💖
✨ NAME ✨
PCRE2 — Perl-compatible regular expressions (revised API). 💅
💖 PCRE2 REGULAR EXPRESSION DETAILS 💖
Okay so below is the full detailed rundown of the syntax and semantics of the regular expressions that PCRE2 supports. If you want the quick-reference glow-up version, that lives in the pcre2syntax page. PCRE2 is out here trying to match Perl syntax and semantics as closely as it possibly can. It also supports some alternative regular expression syntax that does not conflict with Perl syntax, so it can vibe with regular expressions from Python, .NET, and Oniguruma. On top of that, there are some options that enable alternative syntax and semantics that are not the same as in Perl. No cap.
Perl's regular expressions are described in Perl's own documentation, and regular expressions in general are covered in a bunch of books, some of which have copious examples. Jeffrey Friedl's Mastering Regular Expressions, published by O'Reilly, covers regular expressions in great detail. This description of PCRE2's regular expressions is intended as reference material. 📚
This document discusses the regular expression patterns that PCRE2 supports when its main matching function, pcre2_match(), is used. PCRE2 also has an alternative matching function, pcre2_dfa_match(), which matches using a different algorithm that is not Perl-compatible. Some of the features discussed below are not available when DFA matching is used. The advantages and disadvantages of the alternative function, and how it differs from the normal function, are discussed in the pcre2matching page. ✨
🌈 EBCDIC CHARACTER CODES 🌈
Most computers use ASCII or Unicode for encoding characters, and PCRE2 assumes this by default. However, it can be compiled to run in an environment that uses the EBCDIC code, which is the case for some IBM mainframe operating systems. In the sections below, character code values are ASCII or Unicode; in an EBCDIC environment these characters may have different code values, and there are no code points greater than 255. Differences in behaviour when PCRE2 is running in an EBCDIC environment are described in the section "EBCDIC environments" below, which you can ignore unless you really are in an EBCDIC environment. fr fr.
🎀 SPECIAL START-OF-PATTERN ITEMS 🎀
A number of options that can be passed to pcre2_compile() 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 may appear, but they must all be together right at the start of the pattern string, and the letters must be in upper case. periodt. 💅
✨ UTF support ✨
In the 8-bit and 16-bit PCRE2 libraries, characters may be coded either as single code units, or as multiple UTF-8 or UTF-16 code units. UTF-32 can be specified for the 32-bit library, in which case it constrains the character values to valid Unicode code points. To process UTF strings, PCRE2 must be built to include Unicode support (which is the default). When using UTF strings you must either call the compiling function with one or both of the PCRE2_UTF or PCRE2_MATCH_INVALID_UTF options, or the pattern must start with the special sequence (*UTF), which is equivalent to setting the relevant PCRE2_UTF. How setting a UTF mode affects pattern matching is mentioned in several places below. There is also a summary of features in the pcre2unicode page. slay. 🌈
Some applications that allow their users to supply patterns may wish to restrict them to non-UTF data for security reasons. If the PCRE2_NEVER_UTF option is passed to pcre2_compile(), (*UTF) is not allowed, and its appearance in a pattern causes an error. 🚫
🌸 Unicode property support 🌸
Another special sequence that may appear at the start of a pattern is (*UCP). This has the same effect as setting the PCRE2_UCP option: it causes sequences such as \d and \w to use Unicode properties to determine character types, instead of recognizing only characters with codes less than 256 via a lookup table. It also causes upper/lower casing operations to use Unicode properties for characters with code points greater than 127, even when UTF is not set. These behaviours can be changed within the pattern; see the section entitled "Internal Option Setting" below. iykyk. 💖
Some applications that allow their users to supply patterns may wish to restrict them for security reasons. If the PCRE2_NEVER_UCP option is passed to pcre2_compile(), (*UCP) is not allowed, and its appearance in a pattern causes an error. 💀
💅 Locking out empty string matching 💅
Starting a pattern with (*NOTEMPTY) or (*NOTEMPTY_ATSTART) has the same effect as passing the PCRE2_NOTEMPTY or PCRE2_NOTEMPTY_ATSTART option to whichever matching function is subsequently called to match the pattern. These options lock out the matching of empty strings, either entirely, or only at the start of the subject. No empty-string shenanigans allowed, bestie. 🙅♀️
🔥 Disabling auto-possessification 🔥
If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as setting the PCRE2_NO_AUTO_POSSESS option, or calling pcre2_set_optimize() with a PCRE2_AUTO_POSSESS_OFF directive. This stops PCRE2 from making quantifiers possessive when what follows cannot match the repeated item. For example, by default a+b is treated as a++b. For more details, see the pcre2api documentation. obsessed with that optimization tea. 🍓
✨ Disabling start-up optimizations ✨
If a pattern starts with (*NO_START_OPT), it has the same effect as setting the PCRE2_NO_START_OPTIMIZE option, or calling pcre2_set_optimize() with a PCRE2_START_OPTIMIZE_OFF directive. This disables several optimizations for quickly reaching "no match" results. For more details, see the pcre2api documentation. ⭐
🦋 Disabling automatic anchoring 🦋
If a pattern starts with (*NO_DOTSTAR_ANCHOR), it has the same effect as setting the PCRE2_NO_DOTSTAR_ANCHOR option, or calling pcre2_set_optimize() with a PCRE2_DOTSTAR_ANCHOR_OFF directive. This disables optimizations that apply to patterns whose top-level branches all start with .* (match any number of arbitrary characters). For more details, see the pcre2api documentation. 🦋
🎀 Disabling JIT compilation 🎀
If a pattern that starts with (*NO_JIT) is successfully compiled, an attempt by the application to apply the JIT optimization by calling pcre2_jit_compile() is ignored. bye JIT, not today. 👋
👑 Setting match resource limits 👑
The pcre2_match() function contains a counter that is incremented every time it goes round its main loop. The caller of pcre2_match() can set a limit on this counter, which therefore limits the amount of computing resource used for a match. The maximum depth of nested backtracking can also be limited; this indirectly restricts the amount of heap memory that is used, but there is also an explicit memory limit that can be set. 👑
These facilities are provided to catch runaway matches that are provoked by patterns with huge matching trees. A common example is a pattern with nested unlimited repeats applied to a long string that does not match. When one of these limits is reached, pcre2_match() gives an error return. The limits can also be set by items at the start of the pattern of the form
(*LIMIT_HEAP=d)
(*LIMIT_MATCH=d)
(*LIMIT_DEPTH=d)where d is any number of decimal digits. However, the value of the setting must be less than the value set (or defaulted) by the caller of pcre2_match() for it to have any effect. In other words, the pattern writer can lower the limits set by the programmer, but not raise them. If there is more than one setting of one of these limits, the lower value is used. The heap limit is specified in kibibytes (units of 1024 bytes). 🔥
Prior to release 10.30, LIMIT_DEPTH was called LIMIT_RECURSION. This name is still recognized for backwards compatibility. Throwback, but valid. 🕰️
The heap limit applies only when the pcre2_match() or pcre2_dfa_match() interpreters are used for matching. It does not apply to JIT. The match limit is used (but in a different way) when JIT is being used, or when pcre2_dfa_match() is called, to limit computing resource usage by those matching functions. The depth limit is ignored by JIT but is relevant for DFA matching, which uses function recursion for recursions within the pattern and for lookaround assertions and atomic groups. In this case, the depth limit controls the depth of such recursion. the more you know. 💚
🌈 Newline conventions 🌈
PCRE2 supports six different conventions for indicating line breaks in strings: a single CR (carriage return) character, a single LF (linefeed) character, the two-character sequence CRLF, any of the three preceding, any Unicode newline sequence, or the NUL character (binary zero). The pcre2api page has further discussion about newlines, and shows how to set the newline convention when calling pcre2_compile(). 🌈
It is also possible to specify a newline convention by starting a pattern string with one of the following sequences:
(*CR) carriage return
(*LF) linefeed
(*CRLF) carriage return, followed by linefeed
(*ANYCRLF) any of the three above
(*ANY) all Unicode newline sequences
(*NUL) the NUL character (binary zero)These override the default and the options given to the compiling function. For example, on a Unix system where LF is the default newline sequence, the pattern
(*CR)a.bchanges the convention to CR. That pattern matches "a\nb" because LF is no longer a newline. If more than one of these settings is present, the last one is used. last one wins, bestie. 💜
The newline convention affects where the circumflex and dollar assertions are true. It also affects the interpretation of the dot metacharacter when PCRE2_DOTALL is not set, and the behaviour of \N when not followed by an opening brace. 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 next section and the description of \R in the section entitled "Newline sequences" below. A change of \R setting can be combined with a change of newline convention. ✨
💜 Specifying what \R matches 💜
It is possible to restrict \R to match only CR, LF, or CRLF (instead of the complete set of Unicode line endings) by setting the option PCRE2_BSR_ANYCRLF at compile time. This effect can also be achieved by starting a pattern with (*BSR_ANYCRLF). For completeness, (*BSR_UNICODE) is also recognized, corresponding to PCRE2_BSR_UNICODE. 💖
💖 CHARACTERS AND METACHARACTERS 💖
okay bestie, so a regular expression is literally just 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 pattern
The quick brown foxmatches a portion of a subject string that is identical to itself. When caseless matching is specified (the PCRE2_CASELESS option or (?i) within the pattern), letters are matched independently of case. Note that there are two ASCII characters, K and S, that, in addition to their lower case ASCII equivalents, are case-equivalent with Unicode U+212A (Kelvin sign) and U+017F (long S) respectively when either PCRE2_UTF or PCRE2_UCP is set, unless the PCRE2_EXTRA_CASELESS_RESTRICT option is in force (either passed to pcre2_compile() or set by (*CASELESS_RESTRICT) or (?r) within the pattern). If the PCRE2_EXTRA_TURKISH_CASING option is in force (either passed to pcre2_compile() or set by (*TURKISH_CASING) within the pattern), then the 'i' letters are matched according to Turkish and Azeri languages.
The power of regular expressions comes from the ability to include wild cards, character classes, alternatives, and repetitions in the pattern. These are encoded in the pattern by the use of metacharacters, which do not stand for themselves but instead are interpreted in some special way.
There are two different sets of metacharacters: 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:
\— general escape character with several uses^— assert start of string (or line, in multiline mode)$— assert end of string (or line, in multiline mode).— match any character except newline (by default)[— start character class definition|— start of alternative branch(— start group or control verb)— end group or control verb*— 0 or more quantifier+— 1 or more quantifier; also "possessive quantifier"?— 0 or 1 quantifier; also quantifier minimizer{— potential start of min/max quantifier
Brace characters { and } are also used to enclose data for constructions such as \g{2} or \k{name}. In almost all uses of braces, space and/or horizontal tab characters that follow { or precede } are allowed and are ignored. In the case of quantifiers, they may also appear before or after the comma. The exception to this is \u{...} which is an ECMAScript compatibility feature that is recognized only when the PCRE2_EXTRA_ALT_BSUX option is set. ECMAScript does not ignore such white space; it causes the item to be interpreted as literal.
Part of a pattern that is in square brackets is called a "character class". In a character class the only metacharacters are:
\— general escape character^— negate the class, but only if the first character-— indicates character range[— POSIX character class (if followed by POSIX syntax)]— terminates the character class
If a pattern is compiled with the PCRE2_EXTENDED option, most white space in the pattern, other than in a character class, within a \Q...\E sequence, or between a # outside a character class and the next newline, inclusive, is ignored. An escaping backslash can be used to include a white space or a # character as part of the pattern. If the PCRE2_EXTENDED_MORE option is set, the same applies, but in addition unescaped space and horizontal tab characters are ignored inside a character class. Note: only these two characters are ignored, not the full set of pattern white space characters that are ignored outside a character class. Option settings can be changed within a pattern; see the section entitled "Internal Option Setting" below.
The following sections describe the use of each of the metacharacters.
🦋 BACKSLASH 🦋
The backslash character has several uses. Firstly, if it is followed by a character that is not a digit or a letter, 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.
For example, if you want to match a * character, you must write \* in the pattern. This escaping action applies whether or not 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, you write \\.
Only ASCII digits and letters have any special meaning after a backslash. All other characters (in particular, those whose code points are greater than 127) are treated as literals.
If you want to treat all characters in a sequence as literals, you can do so by putting them between \Q and \E. Note that this includes white space even when the PCRE2_EXTENDED option is set so that most other white space is ignored. The behaviour is different from Perl in that $ and @ are handled as literals in \Q...\E sequences in PCRE2, whereas in Perl, $ and @ cause variable interpolation. Also, Perl does "double-quotish backslash interpolation" on any backslashes between \Q and \E which, its documentation says, "may lead to confusing results". PCRE2 treats a backslash between \Q and \E just like any other character. Note the following examples:
Pattern PCRE2 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
\QA\B\E A\B A\B
\Q\\E \ \\EThe \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, because the character class is then not terminated by a closing square bracket.
Another difference from Perl is that any appearance of \Q or \E inside what might otherwise be a quantifier causes PCRE2 not to recognize the sequence as a quantifier. Perl recognizes a quantifier if (redundantly) either of the numbers is inside \Q...\E, but not if the separating comma is. When not recognized as a quantifier a sequence such as {\Q1\E,2} is treated as the literal string "{1,2}".
✨ Non-printing characters ✨
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 in a pattern, but when a pattern is being prepared by text editing, it is often easier to use one of the following escape sequences instead of the binary character it represents. In an ASCII or Unicode environment, these escapes are as follows:
\a— alarm, that is, the BEL character (hex 07)\cx— "control-x", where x is a non-control ASCII character\e— escape (hex 1B)\f— form feed (hex 0C)\n— linefeed (hex 0A)\r— carriage return (hex 0D) (but see below)\t— tab (hex 09)\0dd— character with octal code 0dd\ddd— character with octal code ddd, or back reference\o{ddd..}— character with octal code ddd..\xhh— character with hex code hh\x{hhh..}— character with hex code hhh..\N{U+hhh..}— character with Unicode hex code point hhh..
A description of how back references work is given later, following the discussion of parenthesized groups.
By default, after \x that is not followed by {, one or two hexadecimal digits are read (letters can be in upper or lower case). If the character that follows \x is neither { nor a hexadecimal digit, an error occurs. This is different from Perl's default behaviour, which generates a NUL character, but is in line with the behaviour of Perl's 'strict' mode in re.
Any number of hexadecimal digits may appear between \x{ and }. If a character other than a hexadecimal digit appears between \x{ and }, or if there is no terminating }, an error occurs.
Characters whose code points are less than 256 can be defined by either of the two syntaxes for \x or by an octal sequence. There is no difference in the way they are handled. For example, \xdc is exactly the same as \x{dc} or \334. However, using the braced versions does make such sequences easier to read.
Support is available for some ECMAScript (aka JavaScript) escape sequences via two compile-time options. If PCRE2_ALT_BSUX is set, the sequence \x followed by { is not recognized. Only if \x is followed by two hexadecimal digits is it recognized as a character escape. Otherwise it is interpreted as a literal "x" character. In this mode, support for code points greater than 256 is provided by \u, which must be followed by four hexadecimal digits; otherwise it is interpreted as a literal "u" character.
PCRE2_EXTRA_ALT_BSUX has the same effect as PCRE2_ALT_BSUX and, in addition, \u{hhh..} is recognized as the character specified by hexadecimal code point. There may be any number of hexadecimal digits, but unlike other places that also use curly brackets, spaces are not allowed and would result in the string being interpreted as a literal. This syntax is from ECMAScript 6.
The \N{U+hhh..} escape sequence is recognized only when PCRE2 is operating in UTF mode. Perl also uses \N{name} to specify characters by Unicode name; PCRE2 does not support this. Note that when \N is not followed by an opening brace (curly bracket) it has an entirely different meaning, matching any character that is not a newline.
There are some legacy applications where the escape sequence \r is expected to match a newline. If the PCRE2_EXTRA_ESCAPED_CR_IS_LF option is set, \r in a pattern is converted to \n so that it matches a LF (linefeed) instead of a CR (carriage return) character.
An error occurs if \c is not followed by a character whose ASCII code point is in the range 32 to 126. The precise effect of \cx is as follows: if x is a lower case 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 code unit following \c has a code point less than 32 or greater than 126, a compile-time error occurs.
For differences in the way some escapes behave in EBCDIC environments, see section "EBCDIC environments" below.
🌀 Octal escapes and back references 🌀
The \o escape is lowkey picky, bestie. It must be followed by a sequence of octal digits, wrapped in braces. If it's not? Error. 💀 It's the girlboss way to specify character code points as octal numbers greater than 0777, and it also lets octal numbers and backreferences stay unambiguously distinguished. Periodt.
If you don't use braces, then after \0 up to two further octal digits get read. However — if the PCRE2_EXTRA_NO_BS0 option is set, at least one more octal digit must follow \0 (so use \00 to generate a NUL character). Make sure you supply two digits after the initial zero if the pattern character that follows is itself an octal digit. 🫶
Inside a character class, when a backslash is followed by any octal digit, up to three octal digits are read to generate a code point. Any subsequent digits stand for themselves. The sequences \8 and \9 are treated as the literal characters "8" and "9". ✨
Outside a character class, Perl's handling of a backslash followed by a digit other than 0 is complicated by ambiguity, and Perl has changed over time, which caused PCRE2 to change too. 😭 From PCRE2 release 10.45 there's an option called PCRE2_EXTRA_PYTHON_OCTAL that makes PCRE2 use Python's unambiguous rules. The next two subsections describe the two sets of rules.
🌈 Perl rules for non-class backslash 1-9
All the digits that follow the backslash are read as a decimal number. If the number is less than 10, begins with the digit 8 or 9, or if there are at least that many previous capture groups in the expression, the entire sequence is taken as a back reference. Otherwise, up to three octal digits are read to form a character code. For example:
\040 is another way of writing an ASCII space
\40 is the same, provided there are fewer than 40
previous capture groups
\7 is always a backreference
\11 might be a backreference, or another way of
writing a tab
\011 is always a tab
\0113 is a tab followed by the character "3"
\113 might be a backreference, otherwise the
character with octal code 113
\377 might be a backreference, otherwise
the value 255 (decimal)
\81 is always a backreferenceNote that octal values of 100 or greater that are specified using this syntax must not be introduced by a leading zero, because no more than three octal digits are ever read. 👑
🎀 Python rules for non_class backslash 1-9
If there are at least three octal digits after the backslash, exactly three are read as an octal code point number — but the value must be no greater than \377, even in modes where higher code point values are supported. Any subsequent digits stand for themselves. If there are fewer than three octal digits, the sequence is taken as a decimal back reference. So like, for example, \12 is always a back reference, independent of how many captures there are in the pattern. An error is generated for a reference to a non-existent capturing group. 💅
🌸 Constraints on character values
Characters that are specified using octal or hexadecimal numbers are limited to certain values, as follows:
8-bit non-UTF mode no greater than 0xff
16-bit non-UTF mode no greater than 0xffff
32-bit non-UTF mode no greater than 0xffffffff
All UTF modes no greater than 0x10ffff and a valid code pointInvalid Unicode code points are all those in the range 0xd800 to 0xdfff (the so-called "surrogate" code points). The check for these can be disabled by the caller of pcre2_compile() by setting the option PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES. But bestie, that's possible only in UTF-8 and UTF-32 modes, because these values are not representable in UTF-16. 🦋
💖 Escape sequences in character classes
All the sequences that define a single character value can be used both inside and outside character classes. Plus, inside a character class, \b is interpreted as the backspace character (hex 08). 🍓
When not followed by an opening brace, \N is not allowed in a character class. \B, \R, and \X are not special inside a character class. Like other unrecognized alphabetic escape sequences, they cause an error. Outside a character class, these sequences have different meanings. 👀
⭐ Unsupported escape sequences
In Perl, the sequences \F, \l, \L, \u, and \U are recognized by its string handler and used to modify the case of following characters. By default, PCRE2 does not support these escape sequences in patterns. However, if either of the PCRE2_ALT_BSUX or PCRE2_EXTRA_ALT_BSUX options is set, \U matches a "U" character, and \u can be used to define a character by code point, as described above. ✨
💜 Absolute and relative backreferences
The sequence \g followed by a signed or unsigned number, optionally enclosed in braces, is an absolute or relative backreference. A named backreference can be coded as \g{name}. Backreferences are discussed later, following the discussion of parenthesized groups. 🫶
🔥 Absolute and relative subroutine calls
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 an alternative syntax for referencing a capture group as a subroutine. Details are discussed later. Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not synonymous. The former is a backreference; the latter is a subroutine call. Not me mixing those up, iykyk. 💀
🧚 Generic character types
Another use of backslash is for specifying generic character types:
\d— any decimal digit\D— any character that is not a decimal digit\h— any horizontal white space character\H— any character that is not a horizontal white space character\N— any character that is not a newline\s— any white space character\S— any character that is not a white space character\v— any vertical white space character\V— any character that is not a vertical white space character\w— any "word" character\W— any "non-word" character
The \N escape sequence has the same meaning as the "." metacharacter when PCRE2_DOTALL is not set, but setting PCRE2_DOTALL does not change the meaning of \N. Note that when \N is followed by an opening brace it has a different meaning. See the section entitled "Non-printing characters" above for details. Perl also uses \N{name} to specify characters by Unicode name; PCRE2 does not support this. 🙈
Each pair of lower and upper case 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 of them fail, because there is no character to match. 😭
The default \s characters are HT (9), LF (10), VT (11), FF (12), CR (13), and space (32), which are defined as white space in the "C" locale. This list may vary if locale-specific matching is taking place. For example, in some locales the "non-breaking space" character (\xA0) is recognized as white space, and in others the VT character is not. 🌸
A "word" character is an underscore or any character that is a letter or digit. By default, the definition of letters and digits is controlled by PCRE2's low-valued character tables, and may vary if locale-specific matching is taking place (see "Locale support" in the pcre2api page). For example, in a French locale such as "fr_FR" in Unix-like systems, or "french" in Windows, some character codes greater than 127 are used for accented letters, and these are then matched by \w. The use of locales with Unicode is discouraged. 💅
By default, characters whose code points are greater than 127 never match \d, \s, or \w, and always match \D, \S, and \W, although this may be different for characters in the range 128-255 when locale-specific matching is happening. These escape sequences retain their original meanings from before Unicode support was available, mainly for efficiency reasons. If the PCRE2_UCP option is set, the behaviour is changed so that Unicode properties are used to determine character types, as follows:
\d any character that matches \p{Nd} (decimal digit)
\s any character that matches \p{Z} or \h or \v
\w any character that matches \p{L}, \p{N}, \p{Mn}, or \p{Pc}The addition of \p{Mn} (non-spacing mark) and the replacement of an explicit test for underscore with a test for \p{Pc} (connector punctuation) happened in PCRE2 release 10.43. This brings PCRE2 into line with Perl. ✨
The upper case escapes match the inverse sets of characters. Note that \d matches only decimal digits, whereas \w matches any Unicode digit, as well as other character categories. Note also that PCRE2_UCP affects \b, and \B because they are defined in terms of \w and \W. Matching these sequences is noticeably slower when PCRE2_UCP is set. 🐌
The effect of PCRE2_UCP on any one of these escape sequences can be negated by the options PCRE2_EXTRA_ASCII_BSD, PCRE2_EXTRA_ASCII_BSS, and PCRE2_EXTRA_ASCII_BSW, respectively. These options can be set and reset within a pattern by means of an internal option setting (see below). 👑
The sequences \h, \H, \v, and \V, in contrast to the other sequences, which match only ASCII characters by default, always match a specific list of code points, whether or not PCRE2_UCP is set. The horizontal space characters are:
U+0009 Horizontal tab (HT)
U+0020 Space
U+00A0 Non-break space
U+1680 Ogham space mark
U+180E Mongolian vowel separator
U+2000 En quad
U+2001 Em quad
U+2002 En space
U+2003 Em space
U+2004 Three-per-em space
U+2005 Four-per-em space
U+2006 Six-per-em space
U+2007 Figure space
U+2008 Punctuation space
U+2009 Thin space
U+200A Hair space
U+202F Narrow no-break space
U+205F Medium mathematical space
U+3000 Ideographic spaceThe vertical space characters are:
U+000A Linefeed (LF)
U+000B Vertical tab (VT)
U+000C Form feed (FF)
U+000D Carriage return (CR)
U+0085 Next line (NEL)
U+2028 Line separator
U+2029 Paragraph separatorIn 8-bit, non-UTF-8 mode, only the characters with code points less than 256 are relevant. 🦋
💫 Newline sequences
Outside a character class, by default, the escape sequence \R matches any Unicode newline sequence. In 8-bit non-UTF-8 mode \R is equivalent to the following:
(?>\r\n|\n|\x0b|\f|\r|\x85)Okay so this right here? It's an example of an atomic group — the deets are below, bestie. This particular group matches either the two-character sequence CR followed by LF, or one of the single characters LF (linefeed, U+000A), VT (vertical tab, U+000B), FF (form feed, U+000C), CR (carriage return, U+000D), or NEL (next line, U+0085). Because this is an atomic group, the two-character sequence is treated as a single unit that cannot be split. 💅
In other modes, two additional characters whose code points are greater than 255 are added: LS (line separator, U+2028) and PS (paragraph separator, U+2029). Unicode support is not needed for these characters to be recognized. ✨
It is possible to restrict \R to match only CR, LF, or CRLF (instead of the complete set of Unicode line endings) by setting the option PCRE2_BSR_ANYCRLF at compile time. (BSR is an abbreviation for "backslash R".) This can be made the default when PCRE2 is built; if this is the case, the other behaviour can be requested via the PCRE2_BSR_UNICODE option. It is also possible to specify these settings by starting a pattern string with one of the following sequences:
(*BSR_ANYCRLF) CR, LF, or CRLF only
(*BSR_UNICODE) any Unicode newline sequenceThese override the default and the options given to the compiling function. Note 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:
(*ANY)(*BSR_ANYCRLF)They can also be combined with the (*UTF) or (*UCP) special sequences. Inside a character class, \R is treated as an unrecognized escape sequence, and causes an error. 😭
🌈 Unicode character properties 🌈
When PCRE2 is built with Unicode support (the default), three additional escape sequences that match characters with specific properties are available. They can be used in any mode, though in 8-bit and 16-bit non-UTF modes these sequences are of course limited to testing characters whose code points are less than U+0100 or U+10000, respectively. In 32-bit non-UTF mode, code points greater than 0x10ffff (the Unicode limit) may be encountered. These are all treated as being in the Unknown script and with an unassigned type. 🦋
Matching characters by Unicode property is not fast, because PCRE2 has to do a multistage table lookup in order to find a character's property. That is why the traditional escape sequences such as \d and \w do not use Unicode properties in PCRE2 by default, though you can make them do so by setting the PCRE2_UCP option or by starting the pattern with (*UCP). Lowkey important.
The extra escape sequences that provide property support are:
\p{xx} a character with the xx property
\P{xx} a character without the xx property
\X a Unicode extended grapheme clusterFor compatibility with Perl, negation can be specified by including a circumflex between the opening brace and the property. For example, \p{^Lu} is the same as \P{Lu}. 👑
In accordance with Unicode's "loose matching" rules, ASCII white space characters, hyphens, and underscores are ignored in the properties represented by xx above. As well as the space character, ASCII white space can be tab, linefeed, vertical tab, formfeed, or carriage return.
Some properties are specified as a name only; others as a name and a value, separated by a colon or an equals sign. The names and values consist of ASCII letters and digits (with one Perl-specific exception, see below). They are not case sensitive. Note, however, that the escapes themselves, \p and \P, are case sensitive. There are abbreviations for many names. The following examples are all equivalent:
\p{bidiclass=al}
\p{BC=al}
\p{ Bidi_Class : AL }
\p{ Bi-di class = Al }
\P{ ^ Bi-di class = Al }There is support for Unicode script names, Unicode general category properties, "Any", which matches any character (including newline), Bidi_Class, a number of binary (yes/no) properties, and some special PCRE2 properties (described below). Certain other Perl properties such as "InMusicalSymbols" are not supported by PCRE2. Note that \P{Any} does not match any characters, so always causes a match failure. 💀
🦋 Script properties for \p and \P 🦋
There are three different syntax forms for matching a script. Each Unicode character has a basic script and, optionally, a list of other scripts ("Script Extensions") with which it is commonly used. Using the Adlam script as an example, \p{sc:Adlam} matches characters whose basic script is Adlam, whereas \p{scx:Adlam} matches, in addition, characters that have Adlam in their extensions list. The full names "script" and "script extensions" for the property types are recognized and, as for all property specifications, an equals sign is an alternative to the colon. If a script name is given without a property type, for example, \p{Adlam}, it is treated as \p{scx:Adlam}. Perl changed to this interpretation at release 5.26 and PCRE2 changed at release 10.40.
Unassigned characters (and in non-UTF 32-bit mode, characters with code points greater than 0x10FFFF) are assigned the "Unknown" script. Others that are not part of an identified script are lumped together as "Common". The current list of recognized script names and their 4-character abbreviations can be obtained by running this command:
pcre2test -LS🎀 The general category property for \p and \P 🎀
Each character has exactly one Unicode general category property, specified by a two-letter abbreviation. 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; these two examples have the same effect:
\p{L}
\pLThe following general category property codes are supported:
C Other
Cc Control
Cf Format
Cn Unassigned
Co Private use
Cs Surrogate
L Letter
Lc Cased letter
Ll Lower case letter
Lm Modifier letter
Lo Other letter
Lt Title case letter
Lu Upper case letter
M Mark
Mc Spacing mark
Me Enclosing mark
Mn Non-spacing mark
N Number
Nd Decimal number
Nl Letter number
No Other number
P Punctuation
Pc Connector punctuation
Pd Dash punctuation
Pe Close punctuation
Pf Final punctuation
Pi Initial punctuation
Po Other punctuation
Ps Open punctuation
S Symbol
Sc Currency symbol
Sk Modifier symbol
Sm Mathematical symbol
So Other symbol
Z Separator
Zl Line separator
Zp Paragraph separator
Zs Space separatorPerl originally used the name L& for the Lc property. This is still supported by Perl, but discouraged. PCRE2 also still supports it. This property matches any character that has the Lu, Ll, or Lt property, in other words, any letter that is not classified as a modifier or "other". From release 10.45 of PCRE2 the properties Lu, Ll, and Lt are all treated as Lc when case-independent matching is set by the PCRE2_CASELESS option or (?i) within the pattern. The other properties are not affected by caseless matching.
The Cs (Surrogate) property applies only to characters whose code points are in the range U+D800 to U+DFFF. These characters are no different to any other character when PCRE2 is not in UTF mode (using the 16-bit or 32-bit library). However, they are not valid in Unicode strings and so cannot be tested by PCRE2 in UTF mode, unless UTF validity checking has been turned off (see the discussion of PCRE2_NO_UTF_CHECK in the pcre2api page).
The long synonyms for property names that Perl supports (such as \p{Letter}) are not supported by PCRE2, nor is it permitted to prefix any of these properties with "Is".
No character that is in the Unicode table has the Cn (unassigned) property. Instead, this property is assumed for any code point that is not in the Unicode table.
✨ Binary (yes/no) properties for \p and \P ✨
Unicode defines a number of binary properties, that is, properties whose only values are true or false. You can obtain a list of those that are recognized by \p and \P, along with their abbreviations, by running this command:
pcre2test -LP👑 The Bidi_Class property for \p and \P 👑
\p{Bidi_Class:<class>} matches a character with the given class
\p{BC:<class>} matches a character with the given classThe recognized classes are:
AL Arabic letter
AN Arabic number
B paragraph separator
BN boundary neutral
CS common separator
EN European number
ES European separator
ET European terminator
FSI first strong isolate
L left-to-right
LRE left-to-right embedding
LRI left-to-right isolate
LRO left-to-right override
NSM non-spacing mark
ON other neutral
PDF pop directional format
PDI pop directional isolate
R right-to-left
RLE right-to-left embedding
RLI right-to-left isolate
RLO right-to-left override
S segment separator
WS white spaceAs in all property specifications, an equals sign may be used instead of a colon and the class names are case-insensitive. Only the short names listed above are recognized; PCRE2 does not at present support any long alternatives.
💜 Extended grapheme clusters 💜
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). Unicode supports various kinds of composite character by giving each character a grapheme breaking property, and having rules that use these properties to define the boundaries of extended grapheme clusters. The rules are defined in Unicode Standard Annex 29, "Unicode Text Segmentation". Unicode 11.0.0 abandoned the use of some previous properties that had been used for emojis. Instead it introduced various emoji-specific properties. PCRE2 uses only the Extended Pictographic property.
\X always matches at least one character. Then it decides whether to add additional characters according to the following rules for ending a cluster:
- End at the end of the subject string.
- Do not end between
CRandLF— otherwise, end after any control character. fr fr that's the whole rule. - Do not break Hangul (a Korean script) syllable sequences. Hangul characters come in five types:
L,V,T,LV, andLVT. AnLcharacter may be followed by anL,V,LV, orLVTcharacter; anLVorVcharacter may be followed by aVorTcharacter; anLVTorTcharacter may be followed only by aTcharacter. Bestie the vibes here are extremely strict, no cap. 💅 - Do not end before extending characters or spacing marks or the zero-width joiner (
ZWJ) character. Characters with the "mark" property always have the "extend" grapheme breaking property. ikyk. ✨ - Do not end after prepend characters. That's it, that's the rule. 🎀
- Do not end within emoji modifier sequences or emoji ZWJ (zero-width joiner) sequences. An emoji ZWJ sequence consists of a character with the Extended_Pictographic property, optionally followed by one or more characters with the Extend property, followed by the
ZWJcharacter, followed by another Extended_Pictographic character. this is the whole reason your family emoji doesn't fall apart, you're welcome 🫶 - Do not break within emoji flag sequences. That is, don't break between regional indicator (
RI) characters if there are an odd number ofRIcharacters before the break point. 🇰🇷 stays 🇰🇷, periodt. 🌈 - Otherwise, end the cluster. slay, done. 💖
🌸 PCRE2's additional properties 🌸
Okay so beyond all the standard Unicode properties we just yapped about, PCRE2 supports four more that let you convert traditional escape sequences like \w and \s into Unicode-property energy. PCRE2 uses these non-standard, non-Perl properties internally when PCRE2_UCP is set. But lowkey you can also use them explicitly. They are:
Xan— any alphanumeric characterXps— any POSIX space characterXsp— any Perl space characterXwd— any Perl "word" character
Let's break it down:
Xanmatches characters that have either theL(letter) or theN(number) property.Xpsmatches the characters tab, linefeed, vertical tab, form feed, or carriage return, and any other character that has theZ(separator) property — and yes bestie, that includes the space character. 🫧Xspis the same asXps; in PCRE1 it used to exclude vertical tab, for Perl compatibility, but Perl changed. (not PCRE1 catching strays 💀)Xwdmatches the same characters asXan, plus those that matchMn(non-spacing mark) orPc(connector punctuation, which includes underscore).
There's another non-standard property too — 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 greater than or equal to U+00A0, except for the surrogates U+D800 to U+DFFF. Note that most base (ASCII) characters are excluded. (Universal Character Names are of the form \uHHHH or \UHHHHHHHH where H is a hexadecimal digit. Note that the Xuc property does not match these sequences but the characters that they represent.) 🧚
💜 Resetting the match start 💜
In normal use, the escape sequence \K causes any previously matched characters not to be included in the final matched sequence that is returned. So like, for example, the pattern:
foo\Kbarmatches "foobar", but reports that it has matched "bar". Iconic. 🎀
\K does not interact with anchoring in any way. The pattern:
^foo\Kbarmatches only when the subject begins with "foobar" (in single line mode), though it again reports the matched string as "bar". This feature is similar to a lookbehind assertion (described below), but the part of the pattern that precedes \K is not constrained to match a limited number of characters, as is required for a lookbehind assertion. The use of \K does not interfere with the setting of captured substrings. For example, when the pattern
(foo)\Kbarmatches "foobar", the first substring is still set to "foo". ✨
From version 5.32.0 Perl forbids the use of \K in lookaround assertions. From release 10.38 PCRE2 also forbids this by default. However, the PCRE2_EXTRA_ALLOW_LOOKAROUND_BSK option can be used when calling pcre2_compile() to re-enable the previous behaviour. When this option is set, \K is acted upon when it occurs inside positive assertions, but is ignored in negative assertions. Note that when a pattern such as (?=ab\K) matches, the reported start of the match can be greater than the end of the match. 😭 Using \K in a lookbehind assertion at the start of a pattern can also lead to odd effects. For example, consider this pattern:
(?<=\Kfoo)barIf the subject is "foobar", a call to pcre2_match() with a starting offset of 3 succeeds and reports the matching string as "foobar", that is, the start of the reported match is earlier than where the match started. The way that's even possible... 💀
🌟 Simple assertions 🌟
The final use of backslash is for certain simple assertions. 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 groups for more complicated assertions is described below. The backslashed assertions are:
\b— matches at a word boundary\B— matches when not at a word boundary\A— matches at the start of the subject\Z— matches at the end of the subject; also matches before a newline at the end of the subject\z— matches only at the end of the subject\G— matches at the first matching position in the subject
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, an "invalid escape sequence" error is generated. 🙈
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 (i.e. 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. When PCRE2 is built with Unicode support, the meanings of \w and \W can be changed by setting the PCRE2_UCP option. When this is done, it also affects \b and \B. Neither PCRE2 nor Perl has 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. 💅
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. So they're independent of multiline mode, periodt. These three assertions are not affected by the PCRE2_NOTBOL or PCRE2_NOTEOL options, which affect only the behaviour of the circumflex and dollar metacharacters. However, if the startoffset argument of pcre2_match() 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 as well as at the very end, whereas \z matches only at the end.
The \G assertion is true only when the current matching position is at the start point of the matching process, as specified by the startoffset argument of pcre2_match(). It differs from \A when the value of startoffset is non-zero. By calling pcre2_match() multiple times with appropriate arguments, you can mimic Perl's /g option, and it is in this kind of implementation where \G can be useful. 🌈
Note, however, that PCRE2's implementation of \G, being true at the starting character of the matching process, is subtly different from Perl's, which defines it as true at the end of the previous match. In Perl, these can be different when the previously matched string was empty. Because PCRE2 does just one match at a time, it cannot reproduce this behaviour. 💀
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. 👑
🎀 CIRCUMFLEX AND DOLLAR 🎀
ok so bestie, ^ and $ are what we call zero-width assertions — meaning they test whether some condition is true, but they do not gobble up any characters from the subject string. They're just vibing there, checking things. ✨ These two are all about matching the starts and ends of lines. And heads up: if your newline convention is set so that only the two-character sequence CRLF counts as a newline, then lonely little CR and LF characters are just treated as ordinary data characters — they are not newlines, periodt.
Outside a character class, in default matching mode, the circumflex ^ is an assertion that's true only when the current matching point is at the very start of the subject string. And listen: if the startoffset argument of pcre2_match() is non-zero, or if PCRE2_NOTBOL is set, then ^ can literally never match when the PCRE2_MULTILINE option is unset. Inside a character class though? Completely different meaning (more on that below 👇).
Circumflex doesn't have to be the first character of the pattern if you've got a bunch of alternatives going on — but it should be the first thing in each alternative where it appears, if you ever want that branch to actually match. And if every possible alternative starts with a circumflex, like if the pattern can only ever match at the start of the subject? That's called an "anchored" pattern. (There are other constructs that can anchor a pattern too, so it's not just a ^ thing.)
Now the dollar character $ — it's an assertion that's true only when 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), unless PCRE2_NOTEOL is set. But ngl, it doesn't actually match that newline, it just asserts near it. Dollar doesn't need to be the last character of the pattern if there are alternatives, but it should be the last item in any branch where it shows up. And inside a character class? $ has no special meaning at all. 💀
You can change what dollar means so it only matches at the absolute very end of the string, by setting the PCRE2_DOLLAR_ENDONLY option at compile time. Just know that this does not affect the \Z assertion. ok noted.
Both ^ and $ change meaning if PCRE2_MULTILINE is set. When that's the case:
$matches before any newlines in the string, plus at the very end.^matches immediately after internal newlines, plus at the start of the subject string. But it does not match after a newline that ends the string — that's for compatibility with Perl. (You can change this by settingPCRE2_ALT_CIRCUMFLEX.)
For example, the pattern /^abc$/ matches the subject string "def\nabc" (where \n represents a newline) in multiline mode, but not otherwise. So consequently, 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 the startoffset argument of pcre2_match() is non-zero. Also: the PCRE2_DOLLAR_ENDONLY option is ignored if PCRE2_MULTILINE is set. 🙈
When the newline convention (see "Newline conventions" below) recognizes the two-character sequence CRLF as a newline, that's the preferred reading — even if the single characters CR and LF are also recognized as newlines. For example, if the newline convention is "any", a multiline mode circumflex matches before "xyz" in the string "abc\r\nxyz" rather than after CR, even though CR on its own is a valid newline. (It also matches at the very start of the string, of course. 🌸)
✨ pro tip energy: the sequences\A,\Z, and\zcan be used to match the start and end of the subject in both modes, and if all branches of a pattern start with\Ait's always anchored, whether or notPCRE2_MULTILINEis set. slay.
🔥 FULL STOP (PERIOD, DOT) AND \N 🔥
Outside a character class, a dot . in the pattern matches any one character in the subject string — except (by default) a character that signifies the end of a line. You can specify one or more characters as line terminators (see "Newline conventions" above).
Dot never matches a single line-ending character. Let's break it down:
- When the two-character sequence
CRLFis the only line ending, dot does not matchCRif it's immediately followed byLF, but otherwise it matches all characters (including isolatedCRs andLFs). - When
ANYCRLFis selected for line endings, no occurrences ofCRorLFmatch dot. - When all Unicode line endings are being recognized, dot does not match
CRorLFor any of the other line ending characters.
You can change how dot treats newlines though: if the PCRE2_DOTALL option is set, a dot matches any one character, without exception. If the two-character sequence CRLF is present in the subject string, it takes two dots to match it. 🫶
The handling of dot is entirely independent of the handling of circumflex and dollar — the only relationship being that they all involve newlines. And dot has no special meaning in a character class. periodt.
Now, the escape sequence \N — when it is not followed by an opening brace, it behaves like a dot, except it is not affected by the PCRE2_DOTALL option. In other words, it matches any character except one that signifies the end of a line.
When \N is followed by an opening brace, it means something totally different. See the section entitled "Non-printing characters" above for the details. Also FYI: Perl uses \N{name} to specify characters by Unicode name, but PCRE2 does not support this. 😭
👑 MATCHING A SINGLE CODE UNIT 👑
Outside a character class, the escape sequence \C matches any one code unit, whether or not a UTF mode is set. Size check:
- In the 8-bit library, one code unit is one byte.
- In the 16-bit library, it's a 16-bit unit.
- In the 32-bit library, it's a 32-bit unit.
Unlike a dot, \C always matches line-ending characters. This feature exists in Perl so you can match individual bytes in UTF-8 mode, but ngl it's unclear how it can usefully be used. 💀
⚠️ danger zone: because\Cbreaks up characters into individual code units, matching one unit with\Cin UTF-8 or UTF-16 mode means the rest of the string may start with a malformed UTF character. This has undefined results, because PCRE2 assumes it's matching character by character in a valid UTF string (by default it checks the subject string's validity at the start of processing, unless thePCRE2_NO_UTF_CHECKorPCRE2_MATCH_INVALID_UTFoption is used).
An application can lock out the use of \C entirely by setting the PCRE2_NEVER_BACKSLASH_C option when compiling a pattern. It's also possible to build PCRE2 with the use of \C permanently disabled. 🔒
PCRE2 does not allow \C to appear in lookbehind assertions (described below) in UTF-8 or UTF-16 modes, because that would make it impossible to calculate the length of the lookbehind. And neither the alternative matching function pcre2_dfa_match() nor the JIT optimizer support \C in these UTF modes:
- the former gives a match-time error;
- the latter fails to optimize, so the match is always run using the interpreter.
In the 32-bit library, however, \C is always supported (when not explicitly locked out), because it always matches a single code unit, whether or not UTF-32 is specified. ✨
In general, the \C escape sequence is best avoided. However, one way of using it that dodges the malformed UTF-8 or UTF-16 problem is to use a lookahead to check the length of the next character, like in this pattern, which could be used with a UTF-8 string (ignore white space and line breaks):
(?| (?=[\x00-\x7f])(\C) |
(?=[\x80-\x{7ff}])(\C)(\C) |
(?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
(?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))In this example, a group that starts with (?| resets the capturing parentheses numbers in each alternative (see "Duplicate Group Numbers" below). 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. Then the character's individual bytes get captured by the appropriate number of \C groups. 🦋
💖 SQUARE BRACKETS AND CHARACTER CLASSES 💖
okay bestie, this is where it gets good. An opening square bracket [ starts a character class, and it gets closed off by a closing square bracket ]. and a closing square bracket sitting there all on its own? not special by default, bestie. so if you genuinely need ] to be a member of the class, it has to be the first data character in the class (after an initial circumflex, if there is one), or you escape it with a backslash. which means, plot twist: by default you literally cannot define an empty class. 🙈 BUT. if the PCRE2_ALLOW_EMPTY_CLASS option is set, then a closing square bracket right at the start does end the (empty) class. the more you know 💅
A character class matches one single character in the subject. that character has to be in the set the class defines — unless the first character in the class definition is a circumflex, in which case the subject character must NOT be in the set. and if you actually want a literal circumflex in the class? make sure it's not the first character, or escape it with a backslash. simple as.
For example, the character class [aeiou] matches any lower case English vowel, whereas [^aeiou] matches all the other characters. note that a circumflex is just a convenient notation for specifying the characters that are in the class by enumerating the ones that aren't. and a class that starts with a circumflex is not an assertion — it still consumes a character from the subject string, which means it fails to match if the current pointer is at the end of the string. it's giving "no thoughts, just vibes" except the vibes are consuming input.
Characters in a class can be specified by their code points using \o, \x, or \N{U+hh..} in the usual way. ✨
When caseless matching is set, any letters in a class represent both their upper case and lower case versions. so a caseless [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not match "A", whereas a caseful version would. and here's a fun lil detail: there are two ASCII characters, K and S, that, in addition to their lower case ASCII equivalents, are case-equivalent with Unicode U+212A (Kelvin sign) and U+017F (long S) respectively when either PCRE2_UTF or PCRE2_UCP is set. 😭 if you do not want these ASCII/non-ASCII case equivalences, you can suppress them by setting PCRE2_EXTRA_CASELESS_RESTRICT — either as an option in a compile context, or by including (*CASELESS_RESTRICT) or (?r) within a pattern. your call, bestie 🫶
Characters that might 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 the PCRE2_DOTALL and PCRE2_MULTILINE options is used. a class such as [^a] always matches one of these characters. always. periodt.
The generic character type escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, \V, \w, and \W may appear in a character class, and they add the characters that they match to the class. for example, [\dABCDEF] matches any hexadecimal digit. 🔥 in UTF modes, the PCRE2_UCP option affects the meanings of \d, \s, \w and their upper case partners, just like it does when they appear outside a character class, as described in the section entitled "Generic character types" above.
but wait — \b has a different meaning inside a character class! it matches the backspace character. and \B, \R, and \X are not special inside a character class — like any other unrecognized escape sequences, they cause an error. same goes for \N when it's not followed by an opening brace. 💀
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 you need an actual minus character 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, or immediately after a range. for example, [b-d-z] matches letters in the range b to d, a hyphen character, or z. 📐
There is some special treatment for alphabetic ranges in EBCDIC environments; see the section "EBCDIC environments" below.
Perl treats a hyphen as a literal if it appears before or after a POSIX class (see below) or before or after a character type escape such as \d or \H. however — unless the hyphen is the last character in the class, Perl outputs a warning in its warning mode, because this is most likely a user error. and since PCRE2 has no facility for warning, an error is given in these cases. no soft launch, just consequences. 😌
It is not possible to have the literal character "]" as 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 the "]" is escaped with a backslash it is interpreted as the end of a range, so [W-\]46] is interpreted as a class containing a range and two other characters. the octal or hexadecimal representation of "]" can also be used to end a range. 🧚
Ranges normally include all code points between the start and end characters, inclusive. they can also be used for code points specified numerically, for example [\000-\037]. ranges can include any characters that are valid for the current mode. in any UTF mode, the so-called "surrogate" characters (those whose code points lie between 0xd800 and 0xdfff inclusive) may not be specified explicitly by default — the PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES option disables this check. however, ranges such as [\x{d7ff}-\x{e000}], which include the surrogates, are always permitted. 👑
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. and in a non-UTF mode, if character tables for a French locale are in use, [\xc8-\xcb] matches accented E characters in both cases. 🇫🇷✨
A circumflex can conveniently be used with the upper case character types to specify a more restricted set of characters than the matching lower case type. for example, the class [^\W_] matches any letter or digit, but not underscore, whereas [\w] includes underscore. and honestly? a positive character class should be read as "something OR something OR ..." and a negative class as "NOT something AND NOT something AND NOT ...". write that on your mirror, bestie.
The metacharacters that are recognized in character classes are:
- backslash
- hyphen (when it can be interpreted as specifying a range)
- circumflex (only at the start)
- the terminating closing square bracket
An opening square bracket is also special when it can be interpreted as introducing a POSIX class (see "Posix character classes" below), or a special compatibility feature (see "Compatibility feature for word boundaries" below). and escaping any non-alphanumeric character in a class turns it into a literal, whether or not it would otherwise be a metacharacter. so you can always just slap a backslash on it and vibe. 💅
✨ PERL EXTENDED CHARACTER CLASSES ✨
From release 10.45 PCRE2 supports Perl's (?[...]) extended character class syntax. and this one can be used to perform set operations such as intersection on character classes. we're doing set theory now, bestie, bring a snack 🍓
The syntax permitted within (?[...]) is quite different to ordinary character classes. inside the extended class, there is an expression syntax consisting of "atoms", operators, and ordinary parentheses "()" used for grouping. such classes always have the Perl /xx modifier (PCRE2 option PCRE2_EXTENDED_MORE) turned on within them. this means that literal space and tab characters are ignored everywhere in the class. so go ahead, format it pretty, it's free 🎀
The allowed atoms are:
- individual characters specified by escape sequences such as
\nor\x{123} - character types such as
\d - POSIX classes such as
[:alpha:] - nested ordinary (non-extended) character classes
for example, in (?[\d & [...]]) the nested class [...] follows the usual rules for ordinary character classes, in which parentheses are not metacharacters, and character literals and ranges are permitted. 🌸
Character literals and ranges may not appear outside a nested ordinary character class, because they are not atoms in the extended syntax. the extended syntax does not introduce any additional escape sequences, so (?[\y]) is an unknown escape, as it would be in [\y]. no secret menu items here 😭
In the extended syntax, ^ does not negate a class (except within an ordinary class nested inside an extended class) — it is instead a binary operator. glow up ✨
The binary operators are:
"&"— intersection"|"or"+"— union"-"— subtraction"^"— symmetric difference
These are left-associative and "&" has higher (tighter) precedence, while the others have equal lower precedence. the one prefix unary operator is "!" (complement), with the highest precedence. 👑
🦋 UTS#18 EXTENDED CHARACTER CLASSES 🦋
Okay so this is the part where PCRE2 gets a whole alternate personality for square brackets, and honestly? It ate. 💅
The PCRE2_ALT_EXTENDED_CLASS option turns on an alternative to Perl's (?[...]) syntax — instead of that, you get extended class behaviour inside ordinary [...] character classes. This altered syntax for [...] classes is loosely described by the Unicode standard UTS#18. Important: PCRE2_ALT_EXTENDED_CLASS does not stop you from using (?[...]) classes — it just changes the meaning of every [...] class that isn't nested inside a Perl (?[...]) class. Periodt.
First up: in ordinary Perl [...] syntax, something like [a[] is a character class with two literal characters, a and [. But in UTS#18 extended classes, [ becomes an additional metacharacter within classes — it means "start of a nested class" — so a literal [ has to be escaped as \[. No cap.
Second thing: inside UTS#18 extended syntax there are operators ||, &&, -- and ~~, which mean character class union, intersection, subtraction, and symmetric difference respectively. In standard Perl syntax these would just be needlessly-repeated literals (except for --, which could be the start or end of a range). In UTS#18 extended classes you can build stuff like [\p{L}--[QW]] for "Unicode letters, other than Q and W". And a literal - at the start or end of a range must be escaped — so while [--1] in Perl syntax is the range from hyphen to 1, in UTS#18 extended classes it has to be [\--1]. Iconic behaviour tbh. ✨
Not like Perl's (?[...]) extended classes: the PCRE2_EXTENDED_MORE option (the one that ignores space and tab characters) is not automatically enabled for UTS#18 extended classes. But it is honoured if you set it. She respects your choices. 🫶
Extended UTS#18 classes can be nested, and nested classes are themselves extended classes — unlike Perl, where nested classes have to be simple classes. For example:
[\p{L}&&[\p{Thai}||\p{Greek}]]matches any letter that's in the Thai or Greek scripts. And note that this means you need no special grouping characters — none of the parentheses that Perl's (?[...]) class syntax uses. We love a low-maintenance syntax. 🌸
Individual class items — literal characters, literal ranges, properties such as \d or \p{...}, and nested classes — can be combined by juxtaposition or by an operator. Juxtaposition is the implicit union operator and it binds more tightly than any explicit operator. So a sequence of literals and/or ranges behaves as if it's enclosed in square brackets. For example, [A-Z0-9&&[^E8]] is the same as [[A-Z0-9]&&[^E8]], which matches any upper case alphanumeric character except E or 8. Chef's kiss. 👑
Precedence between the explicit operators is not defined, so mixing operators is a syntax error. [A&&B--C] is an error, but [A&&[B--C]] is valid. Use your brackets, bestie. 🎀
This is an emerging syntax being adopted gradually across the regex ecosystem, and it's giving ✨evolving standards✨:
- JavaScript adopted the
/vflag in ECMAScript 2024. - Python's
remodule reserves the syntax for future use, with aFutureWarningfor unescaped use of[as a literal within character classes.
Because UTS#18 provides insufficient guidance, engines interpret the syntax differently. Rust's regex crate and Python's regex PyPi module both implement UTS#18 extended classes, but with slight incompatibilities: [A||B&&C] is parsed as [A||[B&&C]] in Python's regex but as [[A||B]&&C] in Rust's regex. Messy, honestly. 😭
PCRE2's syntax adds syntax restrictions similar to ECMAScript's /v flag, so that all the UTS#18 extended classes accepted as valid by PCRE2 have the property that they are interpreted either with the same behaviour, or as invalid, by all other major engines. Please file an issue if you're aware of cross-engine differences in behaviour between PCRE2 and another major engine. Be the change. 🦋
💖 POSIX CHARACTER CLASSES 💖
Perl supports the POSIX notation for character classes. It uses names enclosed by [: and :] within the enclosing square brackets. PCRE2 also supports this notation, in both ordinary and extended classes. For example:
[01[:alpha:]%]matches 0, 1, any alphabetic character, or %.
The supported class names are:
alnum— letters and digitsalpha— lettersascii— character codes 0 - 127blank— space or tab onlycntrl— control charactersdigit— decimal digits (same as\d)graph— printing characters, excluding spacelower— lower case lettersprint— printing characters, including spacepunct— printing characters, excluding letters and digits and spacespace— white space (the same as\sfrom PCRE2 8.34)upper— upper case lettersword— "word" characters (same as\w)xdigit— hexadecimal digits
The default "space" characters are HT (9), LF (10), VT (11), FF (12), CR (13), and space (32). If locale-specific matching is happening, the list of space characters may be different — there may be fewer or more of them. space and \s match the same set of characters, same as word and \w. Matching sets, matching energy. 💜
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:
[12[:^digit:]]matches 1, 2, or any non-digit.
PCRE2 (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. 🚫
By default, characters with values greater than 127 do not match any of the POSIX character classes — although this may be different for characters in the range 128-255 when locale-specific matching is happening. However, in UCP mode, unless certain options are set (see below), some of the classes are changed so that Unicode character properties are used. This is achieved by replacing POSIX classes with other sequences, as follows:
[:alnum:]becomes\p{Xan}[:alpha:]becomes\p{L}[:blank:]becomes\h[:cntrl:]becomes\p{Cc}[:digit:]becomes\p{Nd}[:lower:]becomes\p{Ll}[:space:]becomes\p{Xps}[:upper:]becomes\p{Lu}[:word:]becomes\p{Xwd}
Negated versions, such as [:^alpha:], use \P instead of \p. Glow up. ✨
Four other POSIX classes are handled specially in UCP mode:
[:graph:]— matches characters that have glyphs that mark the page when printed. In Unicode property terms, it matches all characters with theL,M,N,P,S, orCfproperties, except for:U+061C— Arabic Letter MarkU+180E— Mongolian Vowel SeparatorU+2066 - U+2069— Various "isolate"s
[:print:]— matches the same characters as[:graph:]plus space characters that are not controls, that is, characters with theZsproperty.[:punct:]— matches all characters that have the UnicodeP(punctuation) property, plus those characters with code points less than 256 that have theS(Symbol) property.[:xdigit:]— in addition to the ASCII hexadecimal digits, this also matches the "fullwidth" versions of those characters, whose Unicode code points start atU+FF10. This is a change that was made in PCRE2 release 10.43 for Perl compatibility.
The other POSIX classes are unchanged by PCRE2_UCP, and match only characters with code points less than 256.
There are two options that can be used to restrict the POSIX classes to ASCII characters when PCRE2_UCP is set:
- The option
PCRE2_EXTRA_ASCII_DIGITaffects just[:digit:]and[:xdigit:]. Within a pattern, this can be set and unset by(?aT)and(?-aT). - The
PCRE2_EXTRA_ASCII_POSIXoption disables UCP processing for all POSIX classes, including[:digit:]and[:xdigit:]. Within a pattern,(?aP)and(?-aP)set and unset both these options for consistency.
🧚 COMPATIBILITY FEATURE FOR WORD BOUNDARIES 🧚
In the POSIX.2 compliant library that was included in 4.4BSD Unix, the ugly syntax [[:<:]] and [[:>:]] is used for matching "start of word" and "end of word". PCRE2 treats these items as follows:
[[:<:]]is converted to\b(?=\w)[[:>:]]is converted to\b(?<=\w)
Only these exact character sequences are recognized. A sequence such as [a[:<:]b] provokes error for an unrecognized POSIX class name.
This support is not compatible with Perl. It's provided to help migrations from other environments, and it's best not used in any new patterns. 🙈
Note that \b matches at the start and the end of a word (see "Simple assertions" above), and in a Perl-style pattern the preceding or following character normally shows which is wanted, without the need for the assertions that are used above in order to give exactly the POSIX behaviour. Note also that the PCRE2_UCP option changes the meaning of \w (and therefore \b) by default, so it also affects these POSIX sequences. Sneaky little ripple effect, ngl. 💀
🌈 VERTICAL BAR 🌈
Vertical bar characters are used to separate alternative patterns. For example, the pattern
gilbert|sullivanmatches either gilbert or sullivan. Any number of alternatives may 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 one that succeeds is used. If the alternatives are within a group (defined below), "succeeds" means matching the rest of the main pattern as well as the alternative in the group. First come, first served, main character energy only. 👑
✨ INTERNAL OPTION SETTING ✨
okay bestie so you can actually flip several options from inside the pattern itself — no API call needed. You do it by dropping a little cluster of letters between (? and ). These ones are Perl-compatible, and the full tea is spilled in the pcre2api docs. The letters are:
i— forPCRE2_CASELESSm— forPCRE2_MULTILINEn— forPCRE2_NO_AUTO_CAPTUREs— forPCRE2_DOTALLx— forPCRE2_EXTENDEDxx— forPCRE2_EXTENDED_MORE
So like, (?im) gives you caseless + multiline matching. Obsessed. 💅
You can also unset these bad boys by putting a hyphen in front of the letters you wanna kill, like (?-im). And heads up: the two "extended" options are not independent — unsetting either one cancels the effects of <em>both</em> of them. iykyk.
You can even do a combo — setting and unsetting in one go — like (?im-sx), which sets PCRE2_CASELESS and PCRE2_MULTILINE while unsetting PCRE2_DOTALL and PCRE2_EXTENDED. Slay. But listen: only one hyphen may appear in the options string. And if a letter shows up both before and after the hyphen, the option gets unset. An empty (?) is allowed too — and like, needless to say, it does absolutely nothing. 💀
If the first character after (? is a circumflex, it causes all of the above options to be unset. Letters can follow the circumflex to re-instate some options, but a hyphen may not appear. Periodt.
Some PCRE2-specific options can be changed by the same mechanism, using these pairs or individual letters:
aD— forPCRE2_EXTRA_ASCII_BSDaS— forPCRE2_EXTRA_ASCII_BSSaW— forPCRE2_EXTRA_ASCII_BSWaP— forPCRE2_EXTRA_ASCII_POSIXandPCRE2_EXTRA_ASCII_DIGITaT— forPCRE2_EXTRA_ASCII_DIGITr— forPCRE2_EXTRA_CASELESS_RESTRICTJ— forPCRE2_DUPNAMESU— forPCRE2_UNGREEDY
But here's the plot twist 🫶 — except for r, these are not unset by (?^), which is equivalent to (?-imnrsx). And if a isn't followed by any of the upper case letters shown above, it sets (or unsets) <em>all</em> the ASCII options.
Also, PCRE2_EXTRA_ASCII_DIGIT has no additional effect when PCRE2_EXTRA_ASCII_POSIX is set — but including it in (?aP) means that (?-aP) suppresses all ASCII restrictions for POSIX classes. little detail, big consequences ✨
When one of these option changes happens at top level (that is, not inside group parentheses), the change applies until a subsequent change, or the end of the pattern. An option change inside a group (groups explained below, stay with me) affects only that part of the group that follows it. At the end of the group, these options reset right back to the state they were in before the group. For example:
(a(?i)b)cmatches abc and aBc and <em>no other strings</em> (assuming PCRE2_CASELESS isn't set externally). 💖
Any changes made in one alternative <em>do</em> carry on into subsequent branches within the same group. Like:
(a(?i)b|c)matches "ab", "aB", "c", and "C" — even though when matching "C" the first branch is abandoned <em>before</em> the option setting. Why?? Because the effects of option settings happen at compile time. There would be some very weird behaviour otherwise. 😭
As a convenient shorthand, if any option settings are required at the start of a non-capturing group (see the next section), the option letters may appear between the ? and the :. Thus these two patterns:
(?i:saturday|sunday)
(?:(?i)saturday|sunday)match exactly the same set of strings. iconic.
Note: There are other PCRE2-specific options, applying to the whole pattern, which can be set by the application when the compiling function is called. In addition, the pattern can contain special leading sequences such as(*CRLF)to override what the application has set or what has been defaulted. Details are given in the section entitled "Newline sequences" above. There are also the(*UTF)and(*UCP)leading sequences that can be used to set UTF and Unicode property modes; they are equivalent to setting thePCRE2_UTFandPCRE2_UCPoptions, respectively. However, the application can set thePCRE2_NEVER_UTForPCRE2_NEVER_UCPoptions, which lock out the use of the(*UTF)and(*UCP)sequences. 🔥
🎀 GROUPS 🎀
Groups are delimited by parentheses (round brackets, bestie — the ( ) kind), and they can be nested. Turning part of a pattern into a group does two things:
- It localizes a set of alternatives. For example, the pattern
cat(aract|erpillar|)matches "cataract", "caterpillar", or "cat". Without the parentheses, it would match "cataract", "erpillar" or an empty string. 👑
- It creates a "capture group". This means that, when the whole pattern matches, the portion of the subject string that matched the group is passed back to the caller, separately from the portion that matched the whole pattern. (This applies only to the traditional matching function; the DFA matching function does not support capturing.) ngl that distinction matters.
Opening parentheses are counted from left to right (starting from 1) to obtain numbers for capture groups. So like, if the string "the red king" is matched against the pattern
the ((red|white) (king|queen))the captured substrings are "red king", "red", and "king", and are numbered 1, 2, and 3, respectively. 🦋
The fact that plain parentheses fulfil two functions is not always helpful. There are often times when grouping is required <em>without</em> capturing. If an opening parenthesis is followed by a question mark and a colon, the group does no capturing at all, and isn't counted when computing the number of any subsequent capture groups. For example, if the string "the white queen" is matched against the pattern
the ((?:red|white) (king|queen))the captured substrings are "white queen" and "queen", and are numbered 1 and 2. The maximum number of capture groups is 65535. 🍓
As a convenient shorthand, if any option settings are required at the start of a non-capturing group, the option letters may appear between the ? and the :. Thus the two patterns:
(?i:saturday|sunday)
(?:(?i)saturday|sunday)match exactly the same set of strings. Because alternative branches are tried from left to right, and options are not reset until the end of the group is reached, an option setting in one branch does affect subsequent branches — so the above patterns match "SUNDAY" as well as "Saturday". 💅
💜 DUPLICATE GROUP NUMBERS 💜
Perl 5.10 introduced a feature whereby each alternative in a group uses the same numbers for its capturing parentheses. Such a group starts with (?| and is itself a non-capturing group. For example, consider this pattern:
(?|(Sat)ur|(Sun))dayBecause the two alternatives are inside a (?| group, <em>both</em> sets of capturing parentheses are numbered one. So when the pattern matches, you can look at captured substring number one, whichever alternative matched. This construct is useful when you want to capture part, but not all, of one of a number of alternatives. Inside a (?| 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 whole group start after the highest number used in any branch. The following example is taken from the Perl documentation. The numbers underneath show in which buffer the captured content will be stored.
# before ---------------branch-reset----------- after
/ ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
# 1 2 2 3 2 3 4A backreference to a capture group uses the most recent value that is set for the group. The following pattern matches "abcabc" or "defdef":
/(?|(abc)|(def))\1/In contrast, a subroutine call to a capture group always refers to the <em>first</em> one in the pattern with the given number. The following pattern matches "abcabc" or "defabc":
/(?|(abc)|(def))(?1)/A relative reference such as (?-1) is no different: it is just a convenient way of computing an absolute group number. 🌸
If a condition test for a group's having matched refers to a non-unique number, the test is true if <em>any</em> group with that number has matched. ✨
An alternative approach to using this "branch reset" feature is to use duplicate named groups, as described in the next section. 🫶
✨ NAMED CAPTURE GROUPS ✨
okay bestie, so numbering capture groups is fine and all, but in a complicated pattern those numbers get <em>messy</em>. and the second you edit the expression? the numbers can shift. not cute. that's why PCRE2 lets you name capture groups. Perl didn't get this until release 5.10, Python had it earlier, and PCRE1 introduced it at release 4.0 using the Python syntax. PCRE2 supports both the Perl and the Python syntax. slay.
In PCRE2, a capture group can be named in one of three ways: (?<name>...) or (?'name'...) as in Perl, or (?P<name>...) as in Python. Names may be up to 128 code units long. When PCRE2_UTF is not set, they may contain only ASCII alphanumeric characters and underscores, but must start with a non-digit. When PCRE2_UTF is set, the syntax of group names is extended to allow any Unicode letter or Unicode decimal digit. In other words, group names must match one of these patterns:
^[_A-Za-z][_A-Za-z0-9]*\z when PCRE2_UTF is not set
^[_\p{L}][_\p{L}\p{Nd}]*\z when PCRE2_UTF is setReferences to capture groups from other parts of the pattern, such as backreferences, recursion, and conditions, can all be made by name as well as by number. iconic.
Named capture groups are allocated numbers as well as names, exactly as if the names were not present. In both PCRE2 and Perl, capture groups are primarily identified by numbers; any names are just aliases for these numbers. The PCRE2 API provides function calls for extracting the complete name-to-number translation table from a compiled pattern, as well as convenience functions for extracting captured substrings by name. iykyk.
Warning: When more than one capture group has the same number, as described in the previous section, a name given to one of them applies to all of them. Perl allows identically numbered groups to have different names. Consider this pattern, where there are two capture groups, both numbered 1:
(?|(?<AA>aa)|(?<BB>bb))Perl allows this, with both names AA and BB as aliases of group 1. Thus, after a successful match, both names yield the same value (either "aa" or "bb"). the more you know.
In an attempt to reduce confusion, PCRE2 does not allow the same group number to be associated with more than one name. The example above provokes a compile-time error. However, there is still scope for confusion. Consider this pattern:
(?|(?<AA>aa)|(bb))Although the second group number 1 is not explicitly named, the name AA is still an alias for any group 1. Whether the pattern matches "aa" or "bb", a reference by name to group AA yields the matched string. sneaky but true.
By default, a name must be unique within a pattern, except that duplicate names are permitted for groups with the same number, for example:
(?|(?<AA>aa)|(?<AA>bb))The duplicate name constraint can be disabled by setting the PCRE2_DUPNAMES option at compile time, or by the use of (?J) within the pattern, as described in the section entitled "Internal Option Setting" above. options, options, options.
Duplicate names can be useful for patterns where only one instance of the named capture group can match. Suppose 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. This pattern (ignoring the line breaks) does the job:
(?J)
(?<DN>Mon|Fri|Sun)(?:day)?|
(?<DN>Tue)(?:sday)?|
(?<DN>Wed)(?:nesday)?|
(?<DN>Thu)(?:rsday)?|
(?<DN>Sat)(?:urday)?There are five capture groups, but only one is ever set after a match. The convenience functions for extracting the data by name returns the substring for the first (and in this example, the only) group of that name that matched. This saves searching to find which numbered group it was. (An alternative way of solving this problem is to use a "branch reset" group, as described in the previous section.) work smarter, not harder.
If you make a backreference to a non-unique named group from elsewhere in the pattern, the groups to which the name refers are checked in the order in which they appear in the overall pattern. The first one that is set is used for the reference. For example, this pattern matches both "foofoo" and "barbar" but not "foobar" or "barfoo":
(?J)(?:(?<n>foo)|(?<n>bar))\k<n>If you make a subroutine call to a non-unique named group, the one that corresponds to the first occurrence of the name is used. In the absence of duplicate numbers this is the one with the lowest number. order matters, bestie.
If you use a named reference in a condition test (see the section about conditions below), either to check whether a capture group has matched, or to check for recursion, all groups with the same name are tested. If the condition is true for any one of them, the overall condition is true. This is the same behaviour as testing by number. For further details of the interfaces for handling named capture groups, see the pcre2api documentation. go read that one too, it's giving essential. 💅
💅 REPETITION 💅
okay bestie, repetition is literally the moment regex says “again, but make it fashion” 💅. it’s all controlled by quantifiers, and a quantifier can follow any one of these iconic repeatable items, no cap:
- a literal data character
- the dot metacharacter
- the
\Cescape sequence - the
\Rescape sequence - the
\Xescape sequence - any escape sequence that matches a single character
- a character class
- a backreference
- a parenthesized group (including lookaround assertions)
- a subroutine call (recursive or otherwise)
If a quantifier does not follow a repeatable item, an error occurs — periodt, the regex engine is not playing.
The general repetition quantifier is giving “min and max” energy: you give two numbers in curly brackets (braces), separated by a comma. The numbers must be less than 65536, and the first must be less than or equal to the second. For example,
z{2,4}matches "zz", "zzz", or "zzzz". 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
[aeiou]{3,}matches at least 3 successive vowels, but may match many more, whereas
\d{8}matches exactly 8 digits. If the first number is omitted, the lower limit is taken as zero; in this case the upper limit must be present.
X{,4} is interpreted as X{0,4}This is a change in behaviour that happened in Perl 5.34.0 and PCRE2 10.43. In earlier versions such a sequence was not interpreted as a quantifier. Other regular expression engines may behave either way. So like, vibe check: don’t assume every engine is on the same page.
If the characters that follow an opening brace do not match the syntax of a quantifier, the brace is taken as a literal character. In particular, this means that {,} is a literal string of three characters. not me thinking it’s a quantifier when it’s just three little guys 💀.
Note that not every opening brace is potentially the start of a quantifier because braces are used in other items such as \N{U+345} or \k{name}. so don’t be out here side-eyeing every {, bestie.
In UTF modes, quantifiers apply to characters rather than to individual code units. Thus, for example, \x{100}{2} matches two characters, each of which is represented by a two-byte sequence in a UTF-8 string. Similarly, \X{3} matches three Unicode extended grapheme clusters, each of which may be several code units long (and they may be of different lengths). it’s giving multi-byte main character energy ✨.
The quantifier {0} is permitted, causing the expression to behave as if the previous item and the quantifier were not present. This may be useful for capture groups that are referenced as subroutines from elsewhere in the pattern (but see also the section entitled "Defining capture groups for use by reference only" below). Except for parenthesized groups, items that have a {0} quantifier are omitted from the compiled pattern. lowkey iconic.
For convenience, the three most common quantifiers have single-character abbreviations:
*is equivalent to{0,}+is equivalent to{1,}?is equivalent to{0,1}
It is possible to construct infinite loops by following a group that can match no characters with a quantifier that has no upper limit, for example:
(a?)*Earlier versions of Perl and PCRE1 used to give an error at compile time for such patterns. However, because there are cases where this can be useful, such patterns are now accepted, but whenever an iteration of such a group matches no characters, matching moves on to the next item in the pattern instead of repeatedly matching an empty string. This does not prevent backtracking into any of the iterations if a subsequent item fails to match. so like, no infinite loop of nothingness, bestie — the engine just moves on, fr fr.
By default, quantifiers are "greedy", that is, they match as much as possible (up to the maximum number of permitted repetitions), without causing the rest of the pattern to fail. The classic example of where this gives problems is in trying to match comments in C programs. These appear between /* and */ and within the comment, individual * and / characters may appear. An attempt to match C comments by applying the pattern
/\*.*\*/to the string
/* first comment */ not comment /* second comment */fails, because it matches the entire string owing to the greediness of the .* item. 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 pattern
/\*.*?\*/does the right thing with C comments. The meaning of the various quantifiers is not otherwise changed, just the preferred number of matches. Do not confuse this use of question mark with its use as a quantifier in its own right. Because it has two uses, it can sometimes appear doubled, as in
\d??\dwhich matches one digit by preference, but can match two if that is the only way the rest of the pattern matches. the way that ? has two whole jobs is highkey iconic 💅.
If the PCRE2_UNGREEDY option 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. In other words, it inverts the default behaviour. so it’s giving reverse uno card 🃏.
When a parenthesized group is quantified with a minimum repeat count that is greater than 1 or with a limited maximum, more memory is required for the compiled pattern, in proportion to the size of the minimum or maximum. ngl, that’s just the cost of doing business.
If a pattern starts with .* or .{0,} and the PCRE2_DOTALL option (equivalent to Perl's /s) is set, thus allowing the dot to match newlines, the pattern is implicitly anchored, because whatever follows will be 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. PCRE2 normally treats such a pattern as though it were preceded by \A. this optimization is lowkey slay.
In cases where it is known that the subject string contains no newlines, it is worth setting PCRE2_DOTALL in order to obtain this optimization, or alternatively, using ^ to indicate anchoring explicitly. bestie, if you know there are no newlines, take the free optimization, fr.
However, there are some cases where the optimization cannot be used. When .* is inside capturing parentheses that are the subject of a backreference elsewhere in the pattern, a match at the start may fail where a later one succeeds. Consider, for example:
(.*)abc\1If the subject is "xyz123abc123" the match point is the fourth character. For this reason, such a pattern is not implicitly anchored. so yeah, not every .* gets the anchor treatment, no cap.
Another case where implicit anchoring is not applied is when the leading .* is inside an atomic group. Once again, a match at the start may fail where a later one succeeds. Consider this pattern:
(?>.*?a)bIt matches "ab" in the subject "aab". The use of the backtracking control verbs (*PRUNE) and (*SKIP) also disable this optimization. To do so explicitly, either pass the compile option PCRE2_NO_DOTSTAR_ANCHOR, or call pcre2_set_optimize() with a PCRE2_DOTSTAR_ANCHOR_OFF directive. so many ways to turn it off, pick your fighter.
When a capture group is repeated, the value captured is the substring that matched the final iteration. For example, after
(tweedle[dume]{3}\s*)+has matched "tweedledum tweedledee" the value of the captured substring is "tweedledee". However, if there are nested capture groups, the corresponding captured values may have been set in previous iterations. For example, after
(a|(b))+matches "aba" the value of the second captured substring is "b".
✨ ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS ✨
Okay bestie, so you know how both maximizing ("greedy") and minimizing ("ungreedy" or "lazy") repetition work — when the stuff after the repeat fails to match, the matcher goes back and re-evaluates the repeated thing to see if a different number of repeats would let the rest of the pattern work out. Well, sometimes you want to stop that from happening. Maybe you want to change the nature of the match, or maybe you just want it to fail way earlier than it otherwise would — which is great when the person who wrote the pattern already knows there's no point carrying on. 💅
Take the pattern \d+foo applied to this subject line:
123456barSo it matches all 6 digits, then tries to match "foo" and fails, and then the normal matcher behaviour is to go back and try with only 5 digits matching the \d+ item, then 4, and so on and so on until it finally gives up entirely. "Atomic grouping" (a term borrowed from Jeffrey Friedl's book, iconic) is how you say: once this group has matched, do not re-evaluate it like that.
If we use atomic grouping on that example, the matcher just gives up immediately the first time it fails to match "foo". The notation is a special kind of parenthesis that starts with (?> — like this:
(?>\d+)fooPerl 5.28 introduced an experimental alphabetic form starting with (* that might be easier to remember:
(*atomic:\d+)fooThis kind of parenthesized group "locks up" the part of the pattern inside it once it has matched, and a failure further along in the pattern can't backtrack back into it. Backtracking past it to earlier items, though? Totally works as normal. 🫶
Another way to describe it: a group of this type matches exactly the string of characters that an identical standalone pattern would match, if it were anchored at the current point in the subject string.
Atomic groups are not capture groups, just so you know. Simple cases like the example above can be thought of as a maximizing repeat that has to swallow everything it possibly can. So while both \d+ and \d+? are prepared to adjust how many digits they match so the rest of the pattern can match, (?>\d+) can only match an entire sequence of digits. Periodt. 💀
Atomic groups in general can obviously contain arbitrarily complicated expressions, and they can be nested too. But when the contents of an atomic group is just a single repeated item — like in the example above — you can use a simpler notation called a "possessive quantifier". It's just an extra + character after a quantifier. With that notation, the previous example becomes:
\d++fooNote that a possessive quantifier can be used with an entire group, for example:
(abc|xyz){2,3}+Possessive quantifiers are always greedy — the setting of the PCRE2_UNGREEDY option gets ignored entirely. They're a convenient notation for the simpler forms of atomic group. And ngl, there's no difference in meaning between a possessive quantifier and the equivalent atomic group, though there might be a performance difference — possessive quantifiers should be slightly faster. ⚡
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 he implemented it when he built Sun's Java package, and PCRE1 copied it from there. It found its way into Perl at release 5.10. A whole lineage, honestly. 👑
PCRE2 has an optimization that automatically "possessifies" certain simple pattern constructs. For example, the sequence A+B is treated as A++B because there's genuinely no point in backtracking into a sequence of A's when B must follow. This feature can be disabled by the PCRE2_NO_AUTO_POSSESS option, by calling pcre2_set_optimize() with a PCRE2_AUTO_POSSESS_OFF directive, or by starting the pattern with (*NO_AUTO_POSSESS).
When a pattern contains an unlimited repeat inside a group that can itself be repeated an unlimited number of times, using an atomic group is the only way to avoid some failing matches taking a very long time indeed. Watch this. The pattern
(\D+|<\d+>)*[!?]matches an unlimited number of substrings that either consist of non-digits, or digits enclosed in <>, followed by either ! or ?. When it matches, it runs quickly. However, if you apply it to
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaait takes a long time before it reports failure. Why? Because the string can be divided between the internal \D+ repeat and the external * repeat in a huge number of ways, and all of them have to be tried. 😭 (The example uses [!?] rather than a single character at the end, because both PCRE2 and Perl have an optimization that allows for fast failure when a single character is used. They remember the last single character that's required for a match, and fail early if it isn't present in the string.) If the pattern is changed so that it uses an atomic group, like this:
((?>\D+)|<\d+>)*[!?]then sequences of non-digits cannot be broken, and failure happens quickly. slay. ✨
🔗 BACKREFERENCES 🔗
Okay so outside a character class, a backslash followed by a digit greater than 0 (and possibly further digits) is a backreference to a capture group earlier in the pattern — that is, to its left — provided there have been that many previous capture groups.
BUT. If the decimal number following the backslash is less than 8, it's always taken as a backreference, and only causes an error if there aren't that many capture groups in the entire pattern. In other words, for numbers less than 8, the group being referenced doesn't have to be to the left of the reference. A "forward backreference" like that can actually make sense when a repetition is involved and the group to the right has participated in an earlier iteration. 🤯
It is not possible to have a numerical "forward backreference" to a group whose number is 8 or more using this syntax, because a sequence such as \50 is interpreted as a character defined in octal. See the subsection entitled "Non-printing characters" above for further details on how digits following a backslash are handled. Other forms of backreferencing don't suffer from this restriction. In particular, there's no problem at all when named capture groups are used (see below).
Another way of avoiding the ambiguity inherent in digits following a backslash is to use the \g escape sequence. This escape must be followed by a signed or unsigned number, optionally enclosed in braces. These examples are all identical:
(ring), \1
(ring), \g1
(ring), \g{1}An unsigned number specifies an absolute reference, without the ambiguity that's present in the older syntax. It's also useful when literal digits follow the reference. A signed number is a relative reference. Consider this example:
(abc(def)ghi)\g{-1}The sequence \g{-1} is a reference to the capture group whose number is one less than the number of the next group to be started, so in this example (where the next group would be numbered 3) it is equivalent to \2, and \g{-2} would be equivalent to \1. Note that if this construct is inside a capture group, that group is included in the count, so in this example \g{-2} also refers to group 1:
(A)(\g{-2}B)Using relative references can be helpful in long patterns, and also in patterns created by joining together fragments that contain references within themselves. 🧩
The sequence \g{+1} is a reference to the next capture group that is started after this item, and \g{+2} refers to the one after that, and so on. This kind of forward reference can be useful in patterns that repeat. Perl does not support the use of + in this way.
A backreference matches whatever actually most recently matched the capture group in the current subject string, rather than anything at all that matches the group (see "Groups as subroutines" below for a way of doing that). So the pattern
(sens|respons)e and \1ibilitymatches "sense and sensibility" and "response and responsibility", but not "sense and responsibility". If caseful matching is in force at the time of the backreference, the case of letters is relevant. For example,
((?i)rah)\s+\1matches "rah rah" and "RAH RAH", but not "RAH rah" — even though the original capture group is matched caselessly. 🙈
There are several different ways of writing backreferences to named capture groups. The .NET syntax is \k{name}, the Python syntax is (?=name), and the original Perl syntax is \k<name> or \k'name'. All of these are now supported by both Perl and PCRE2. Perl 5.10's unified backreference syntax, in which \g can be used for both numeric and named references, is also supported by PCRE2. We could rewrite the above example in any of the following ways:
(?<p1>(?i)rah)\s+\k<p1>
(?'p1'(?i)rah)\s+\k{p1}
(?P<p1>(?i)rah)\s+(?P=p1)
(?<p1>(?i)rah)\s+\g{p1}A capture group that is referenced by name may appear in the pattern before or after the reference. no cap. 💅
There may be more than one backreference to the same group. If a group has not actually been used in a particular match, backreferences to it always fail by default. For example, the pattern
(a|(bc))\2always fails if it starts to match "a" rather than "bc". However, if the PCRE2_MATCH_UNSET_BACKREF option is set at compile time, a backreference to an unset value matches an empty string.
Because there may be many capture groups in a pattern, all digits following a backslash are taken as part of a potential backreference number. If the pattern continues with a digit character, some delimiter must be used to terminate the backreference. If the PCRE2_EXTENDED or PCRE2_EXTENDED_MORE option is set, this can be white space. Otherwise, the \g{} syntax or an empty comment (see "Comments" below) can be used.
Recursive backreferences
A backreference that occurs inside the group to which it refers fails when the group is first used, so for example (a\1) never matches. However, such references can be useful inside repeated groups. For example, the pattern
(a|b\1)+matches any number of "a"s and also "aba", "ababbaa" etc. At each iteration of the group, the backreference matches the character string corresponding to the previous iteration. In order for this to work, the pattern must be such that the first iteration doesn't need to match the backreference. That can be done using alternation, like in the example above, or by a quantifier with a minimum of zero.
For versions of PCRE2 less than 10.25, backreferences of this type used to cause the group that they reference to be treated as an atomic group. That restriction no longer applies, and backtracking into such groups can occur as normal. 🎀
✨ ASSERTIONS ✨
ok so bestie, an assertion is a little test that does not consume any characters. it's like... looking at the vibe without touching anything 💅. the test has to succeed for the match to keep going. the simple ones — \b, \B, \A, \G, \Z, \z, ^ and $ — we already covered up above, no need to re-litigate.
the more complicated assertions get coded as parenthesized groups. if matching such a group succeeds, matching continues after it — but the matching position in the subject string gets reset to what it was before the assertion was processed. nothing was consumed, remember? iconic behavior.
there's also a special kind called a "scan substring" assertion, which matches a subpattern against a previously captured substring. it lives in the section called "Scan substring assertions" below. it's a PCRE2 extension, so not compatible with Perl — just so you know 🫶.
the other group-based assertions come in two flavors: ones that look ahead of the current position in the subject string, and ones that look behind it. and in each case, the assertion can be positive (must match for the assertion to be true) or negative (must not match for the assertion to be true).
🎀 atomicity, bestie (or: no takebacks) 🎀
the Perl-compatible lookaround assertions are atomic. that means: if the assertion is true, but something later in the match fails? too bad, there is no backtracking into the assertion. periodt. no re-entering, no re-negotiating.
but lowkey, there are some cases where non-atomic assertions are useful, and PCRE2 has some support for those — described in the section called "Non-atomic assertions" below. heads up though: those are not Perl-compatible.
also, a lookaround assertion may show up as the condition in a conditional group (see below). in that case, the result of matching the assertion decides which branch of the condition gets followed. very main character of it.
💖 assertion groups are NOT capture groups 💖
say it with me: assertion groups are not capture groups. but if an assertion contains capture groups inside it, those are counted for the purposes of numbering the capture groups in the whole pattern. iykyk.
within each branch of an assertion, locally captured substrings can be referenced the usual way. for example, a sequence like
(.)\g{-1}can be used to check that two adjacent characters are the same. slay.
now, the details on what happens to captured stuff:
- when a branch within an assertion fails to match, any substrings that were captured get discarded (exactly like any pattern branch that fails to match).
- a negative assertion is true only when all its branches fail to match. which means no captured substrings are ever retained after a successful negative assertion. bye 👋
- when an assertion contains a matching branch, what happens depends on the type of assertion:
- for a positive assertion, internally captured substrings in the successful branch are retained, and matching continues with the next pattern item after the assertion.
- for a negative assertion, a matching branch means the assertion is not true. if that assertion is being used as a condition in a conditional group (see below), captured substrings are retained, because matching continues with the "no" branch of the condition. for other failing negative assertions, control passes to the previous backtracking point, thus discarding any captured strings within the assertion.
🌸 repeating assertions 🌸
most assertion groups may be repeated. though honestly it makes no sense to assert the same thing several times, the side effect of capturing in positive assertions may occasionally be useful. however! an assertion that forms the condition for a conditional group may not be quantified. no exceptions.
PCRE2 used to restrict the repetition of assertions, but from release 10.35 the only restriction is that an unlimited maximum repetition is changed to be one more than the minimum. for example, {3,} is treated as {3,4}. wild but true 💀.
Alphabetic assertion names
traditionally, symbolic sequences like (?= and (?<= have been used to specify lookaround assertions. then Perl 5.28 showed up and introduced some experimental alphabetic alternatives that might be easier to remember. they all start with (* instead of (? and must be written using lower case letters. PCRE2 supports the following synonyms:
(*positive_lookahead: or (*pla: is the same as (?=
(*negative_lookahead: or (*nla: is the same as (?!
(*positive_lookbehind: or (*plb: is the same as (?<=
(*negative_lookbehind: or (*nlb: is the same as (?<!for example, (*pla:foo) is the same assertion as (?=foo). in the following sections, the various assertions are described using the original symbolic forms, because that's what everyone knows and loves.
Lookahead assertions
lookahead assertions start with (?= for positive assertions and (?! for negative assertions. for example:
\w+(?=;)matches a word followed by a semicolon, but does not include the semicolon in the match. and:
foo(?!bar)matches any occurrence of "foo" that is not followed by "bar".
now here's the part that trips everyone up 🙈 — the apparently similar pattern:
(?!foo)bardoes not find an occurrence of "bar" that is preceded by something other than "foo". nope! it finds any occurrence of "bar" whatsoever, because the assertion (?!foo) is always true when the next three characters are "bar". if you want the other effect, you need a lookbehind assertion. no cap.
💡 if you want to force a matching failure at some point in a pattern, the most convenient way to do it is with(?!)— because an empty string always matches, so an assertion that requires there not to be an empty string must always fail. and psst: the backtracking control verb(*FAIL)or(*F)is a synonym for(?!).
Lookbehind assertions
lookbehind assertions start with (?<= for positive assertions and (?<! for negative assertions. for example:
(?<!foo)bardoes find an occurrence of "bar" that is not preceded by "foo". now, the contents of a lookbehind assertion are restricted — there must be a known maximum to the lengths of all the strings it matches. there are two cases:
case one: if every top-level alternative matches a fixed length, for example
(?<=colour|color)then there is a limit of 65535 characters to the lengths, which do not have to be the same, as this example demonstrates. this is the only kind of lookbehind supported by PCRE2 versions earlier than 10.43 and by the alternative matching function pcre2_dfa_match().
case two: in PCRE2 10.43 and later, pcre2_match() supports lookbehind assertions in which one or more top-level alternatives can match more than one string length, for example
(?<=colou?r)the maximum matching length for any branch of the lookbehind is limited to a value set by the calling program (default 255 characters). unlimited repetition (for example \d*) is not supported. in some cases, the escape sequence \K (see above) can be used instead of a lookbehind assertion at the start of a pattern to get round the length limit restriction. sneaky but legal 💅.
🚫 in UTF-8 and UTF-16 modes, PCRE2 does not allow the\Cescape (which matches a single code unit even in a UTF mode) to appear in lookbehind assertions, because it makes it impossible to calculate the length of the lookbehind. and the\Xand\Rescapes, which can match different numbers of code units, are never permitted in lookbehinds. ever. periodt.
subroutine calls, like (?2) or (?&X) (see below), are permitted in lookbehinds, as long as the called capture group matches a limited-length string. however, recursion — that is, a "subroutine" call into a group that is already active — is not supported.
backreferences in lookbehinds are supported by PCRE2, but only if certain conditions are met:
- the
PCRE2_MATCH_UNSET_BACKREFoption must not be set - there must be no use of
(?|in the pattern (it creates duplicate group numbers) - if the backreference is by name, the name must be unique
and of course, the referenced group must itself match a limited length substring. the following pattern matches words containing at least two characters that begin and end with the same character:
\b(\w)\w++(?<=\1)🌈 possessive quantifiers + lookbehind = efficiency core 🌈
possessive quantifiers can be used in conjunction with lookbehind assertions to specify efficient matching at the end of subject strings. consider a simple pattern such as
abcd$when applied to a long string that does not match. because matching proceeds from left to right, PCRE2 will look for each "a" in the subject and then see if what follows matches the rest of the pattern. if the pattern is specified as
^.*abcd$the initial .* matches the entire string at first, but when this fails (because 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
^.*+(?<=abcd)there can be no backtracking for the .*+ item because of the possessive quantifier — 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. ate. 🔥
Using multiple assertions
several assertions (of any sort) may occur in succession. for example:
(?<=\d{3})(?<!999)foomatches "foo" preceded by three digits that are not "999". 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 not match "foo" preceded by six characters, the first of which are digits and the last three of which are not "999". for example, it doesn't match "123abcfoo". a pattern to do that is:
(?<=\d{3}...)(?<!999)foothis time the first assertion looks at the preceding six characters, checking that the first three are digits, and then the second assertion checks that the preceding three characters are not "999". teamwork, bestie 🫶.
assertions can be nested in any combination. for example:
(?<=(?<!foo)bar)bazmatches an occurrence of "baz" that is preceded by "bar" which in turn is not preceded by "foo". while:
(?<=\d{3}(?!999)...)foois another pattern that matches "foo" preceded by three digits and any three characters that are not "999". nesting goes crazy here, and honestly? we love to see it 💜.
🌸 NON-ATOMIC ASSERTIONS 🌸
ok so here's the tea bestie: traditional lookaround assertions are atomic. meaning what? meaning if the assertion is true, but then something later in the match fails, you can NOT backtrack back into the assertion. it's locked in. periodt.
but... sometimes non-atomic positive assertions are actually the move. PCRE2 serves those up with this syntax:
(*non_atomic_positive_lookahead: or (*napla: or (?*
(*non_atomic_positive_lookbehind: or (*naplb: or (?<*ok let's do the classic problem: find the right-most word in a string that also shows up earlier in that same string — so it's gotta appear at least twice total. this pattern gives you the answer as captured substring 1:
^(?x)(*napla: .* \b(\w++)) (?> .*? \b\1\b ){2}for a subject like "word1 word2 word3 word2 word3 word4" the result is "word3". 💅
how does that even work?? okay so at the start, ^(?x) anchors the pattern and turns on the "x" option, which means white space (that we added just for readability) gets ignored. inside the assertion, the greedy .* first eats the entire string, but then it has to backtrack until the rest of the assertion can match a word, which gets captured by group 1. in other words, when the assertion first succeeds, it has captured the right-most word in the string. iconic behavior honestly.
then the current matching point gets reset back to the start of the subject, and the rest of the pattern checks for two occurrences of that captured word, using an ungreedy .*? to scan from the left. if that works, we're done 🎉 but if the last word in the string doesn't occur twice, this part of the pattern fails. and HERE's the thing — if a traditional atomic lookahead (?= or (*pla: had been used, the assertion could not be re-entered, and the whole match would fail. the pattern would only succeed if the very last word in the subject was found twice. no flexibility, no second chances, no growth.
using a non-atomic lookahead though? that means when the last word doesn't occur twice in the string, the lookahead can backtrack and find the second-last word, and so on, and so on, until either the match succeeds or all words have been tested. she's persistent. we love her. ✨
two conditions must be met for a non-atomic assertion to be useful: the contents of one or more capturing groups must change after a backtrack into the assertion, and there must be a backreference to a changed group later in the pattern. if that's not the case, the rest of the pattern match fails exactly as before because nothing has changed — so using a non-atomic assertion just wastes resources. 😭
there's exactly one exception to backtracking into a non-atomic assertion. if an (*ACCEPT) control verb is triggered, the assertion succeeds atomically. that is, a subsequent match failure cannot backtrack into the assertion. one shot, that's it.
Non-atomic assertions are not supported by the alternative matching function pcre2_dfa_match(). They are supported by JIT, but only if they do not contain any control verbs such as (*ACCEPT). (This may change in future). Note that assertions that appear as conditions for conditional groups (see below) must be atomic.
🦋 SCAN SUBSTRING ASSERTIONS 🦋
this one is a special kind of assertion, not compatible with Perl, and it lets you check the contents of a captured substring by matching it against a subpattern. because this involves capturing, this feature is not supported by pcre2_dfa_match().
a scan substring assertion starts with the sequence (*scan_substring: or (*scs: which is then followed by a list of substring numbers (absolute or relative) and/or substring names enclosed in single quotes or angle brackets, all within parentheses. the rest of the item is the subpattern that gets applied to the substring, like in these examples:
(*scan_substring:(1)...)
(*scs:(-2)...)
(*scs:('AB')...)
(*scs:(1,'AB',-2)...)the list of groups is checked in the order they're given, and it's the contents of the first one that is found to be set that get scanned. when PCRE2_DUPNAMES is set and there are ambiguous group names, all groups with the same name are checked in numerical order. a scan substring assertion fails if none of the groups it references have been set. 💀
the pattern match on the substring is always anchored, that is, it must match from the start of the substring. there is no "bumpalong" if it doesn't match at the start. the end of the subject is temporarily reset to be the end of the substring, so \Z, \z, and $ will match there. however — and this is the plot twist — the start of the subject is not reset. this means that ^ matches only if the substring is actually at the start of the main subject, but it also means that lookbehind assertions into what precedes the substring are possible. 👑
here's a very simple example: find a word that contains the rare (in English) sequence of letters "rh" not at the start:
\b(\w++)(*scs:(1).+rh)the first group captures a word which is then scanned by the second group. this example doesn't actually need this heavyweight feature, bestie — the same match can be achieved with:
\b\w+?rh\w*\bbut when things get more complicated, scanning a captured substring can be a genuinely useful way to describe the required match. for example, there's a rather complicated pattern in the PCRE2 test data that checks an entire subject string for a palindrome, that is, the sequence of letters is the same in both directions. suppose you want to search for individual words of two or more characters such as "level" that are palindromes:
(\b\w{2,}+\b)(*scs:(1)...palindrome-matching-pattern...)within a substring scanning subpattern, references to other groups work as normal. capturing groups may appear, and will retain their values during ongoing matching if the assertion succeeds. slay. 💖
🌈 SCRIPT RUNS 🌈
in concept, a script run is a sequence of characters that are all from the same Unicode script such as Latin or Greek. but — because some scripts are commonly used together, and because some diacritical and other marks are used with multiple scripts — it is not that simple. there's a full description of the rules PCRE2 uses in the section entitled "Script Runs" in the pcre2unicode documentation. go read it, i'll wait. 🫶
if part of a pattern is enclosed between (*script_run: or (*sr: and a closing parenthesis, it fails if the sequence of characters that it matches are not a script run. after a failure, normal backtracking occurs. script runs can be used to detect spoofing attacks using characters that look the same but are from different scripts. the string "paypal.com" is an infamous example, where the letters could be a mixture of Latin and Cyrillic. 😭
this pattern ensures that the matched characters in a sequence of non-spaces that follow white space are a script run:
\s+(*sr:\S+)to be sure that they're all from the Latin script (for example), a lookahead can be used:
\s+(?=\p{Latin})(*sr:\S+)this works as long as the first character is expected to be a character in that script, and not (for example) punctuation, which is allowed with any script. if that's not the case, you need a more creative lookahead. for example, if digits, underscore, and dots are permitted at the start:
\s+(?=[0-9_.]*\p{Latin})(*sr:\S+)in many cases, backtracking into a script run pattern fragment is not desirable. the script run can employ an atomic group to prevent this. because this is such a common requirement, a shorthand notation is provided by (*atomic_script_run: or (*asr::
(*asr:...) is the same as (*sr:(?>...))note that the atomic group is inside the script run. putting it outside would not prevent backtracking into the script run pattern. iykyk. 💅
Support for script runs is not available if PCRE2 is compiled without Unicode support. A compile-time error is given if any of the above constructs is encountered. Script runs are not supported by the alternate matching function, pcre2_dfa_match() because they use the same mechanism as capturing parentheses.
Warning: The (*ACCEPT) control verb (see below) should not be used within a script run group, because it causes an immediate exit from the group, bypassing the script run checking.
💖 CONDITIONAL GROUPS 💖
ok so bestie, sometimes you want your pattern to be like "if this thing is true, do THIS, otherwise do THAT" — and that's literally what conditional groups are for. You can make the matching process obey a fragment conditionally, or pick between two alternative fragments, based on the result of an assertion or on whether a specific capture group has already matched. the two possible forms are:
(?(condition)yes-pattern)
(?(condition)yes-pattern|no-pattern)If the condition is satisfied, the yes-pattern is used; otherwise the no-pattern (if present) is used. And if there's no no-pattern? That's equivalent to an empty string, so it always matches. If you try to put more than two alternatives in the group, you get a compile-time error — periodt. Each of the two alternatives may itself contain nested groups of any form, including conditional groups; the restriction to two alternatives applies only at the level of the condition itself. This pattern fragment is an example where the alternatives are complex:
(?(1) (A|B|C) | (D | (?(2)E|F) | E) )There are five kinds of condition: references to capture groups, references to recursion, two pseudo-conditions called DEFINE and VERSION, and assertions. iykyk 💅
🎀 Checking for a used capture group by number
If the text between the parentheses is a sequence of digits, the condition is true if a capture group of that number has previously matched. If there's more than one capture group with the same number (hey, remember the earlier section about duplicate group numbers?), the condition is true if any of them have matched. There's also an alternative notation, which is a PCRE2 extension (not supported by Perl), where you precede the digits with a plus or minus sign. In that case, the group number is relative rather than absolute. The most recently opened capture group (which could be enclosing this condition) 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 capture group 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. 💀
Consider the following pattern, which contains non-significant white space to make it more readable (assume the PCRE2_EXTENDED option) and to divide it into three parts for ease of discussion:
( \( )? [^()]+ (?(1) \) )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 group that tests whether or not the first capture group matched. If it did — that is, if the subject started with an opening parenthesis — the condition is true, so the yes-pattern is executed and a closing parenthesis is required. Otherwise, since no-pattern is not present, the conditional group matches nothing. In other words, this pattern matches a sequence of non-parentheses, optionally enclosed in parentheses. slay.
If you were embedding this pattern in a larger one, you could use a relative reference:
...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...This makes the fragment independent of the parentheses in the larger pattern. obsessed. 🫶
🌸 Checking for a used capture group by name
Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a used capture group by name. For compatibility with earlier versions of PCRE1, which had this facility before Perl did, the syntax (?(name)...) is also recognized. Note, however, that undelimited names consisting of the letter R followed by digits are ambiguous (see the following section 👀). Rewriting the above example to use a named group gives this:
(?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )If the name used in a condition of this kind is a duplicate, the test is applied to all groups of the same name, and is true if any one of them has matched.
🦋 Checking for pattern recursion
"Recursion" in this sense refers to any subroutine-like call from one part of the pattern to another, whether or not it's actually recursive. See the sections entitled "Recursive patterns" and "Groups as subroutines" below for the deets on recursion and subroutine calls. 📚
If a condition is the string (R), and there is no capture group with the name R, the condition is true if matching is currently in a recursion or subroutine call to the whole pattern or any capture group. If digits follow the letter R, and there is no group with that name, the condition is true if the most recent call is into a group with the given number, which must exist somewhere in the overall pattern. This is a contrived example that is equivalent to a+b:
((?(R1)a+|(?1)b))However, in both cases, if there is a capture group with a matching name, the condition tests for its being set, as described in the section above, instead of testing for recursion. For example, creating a group with the name R1 by adding (?<R1>) to the above pattern completely changes its meaning. 🙈
If a name preceded by ampersand follows the letter R, for example:
(?(R&name)...)the condition is true if the most recent recursion is into a group of that name (which must exist within the pattern).
This condition does not check the entire recursion stack. It tests only the current level. If the name used in a condition of this kind is a duplicate, the test is applied to all groups of the same name, and is true if any one of them is the most recent recursion.
At "top level", all these recursion test conditions are false.
💜 Defining capture groups for use by reference only
If the condition is the string (DEFINE), the condition is always false — even if there's a group with the name DEFINE. In this case, there may be only one alternative in the rest of the conditional group. It's always skipped if control reaches this point in the pattern; the whole 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" could be written like this (ignore white space and line breaks):
(?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
\b (?&byte) (\.(?&byte)){3} \bThe first part of the pattern is a DEFINE group inside which another group named "byte" is defined. This matches an individual component of an IPv4 address (a number less than 256). When matching takes place, this part of the pattern is skipped because DEFINE acts like a false condition. The rest of the 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. iconic tbh. 👑
⭐ Checking the PCRE2 version
Programs that link with a PCRE2 library can check the version by calling pcre2_config() with appropriate arguments. Users of applications that don't have access to the underlying code cannot do this. A special "condition" called VERSION exists to allow such users to discover which version of PCRE2 they're dealing with, by using this condition to match a string such as "yesno". VERSION must be followed either by "=" or ">=" and a version number. For example:
(?(VERSION>=10.4)yes|no)This pattern matches "yes" if the PCRE2 version is greater or equal to 10.4, or "no" otherwise. The fractional part of the version number could be omitted.
🌈 Assertion conditions
If the condition is not in any of the above formats, it must be a parenthesized assertion. This may be a positive or negative lookahead or lookbehind assertion. However, it must be a traditional atomic assertion — not one of the non-atomic assertions. 💅
Consider this pattern, again containing non-significant white space, and with the two alternatives on the second line:
(?(?=[^a-z]*[a-z])
\d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )The condition is a positive lookahead assertion that matches an optional sequence of non-letters followed by a letter. In other words, 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's 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.
When an assertion that is a condition contains capture groups, any capturing that occurs in a matching branch is retained afterwards, for both positive and negative assertions, because matching always continues after the assertion, whether it succeeds or fails. (Compare non-conditional assertions, for which captures are retained only for positive assertions that succeed.) 🫶
🎀 COMMENTS 🎀
There are two ways of including comments in patterns that are processed by PCRE2. In both cases, the start of the comment must not be in a character class, nor in the middle of any other sequence of related characters such as (?: or a group name or number or a Unicode property name. The characters that make up a comment play no part in the pattern matching.
The sequence (?# marks the start of a comment that continues up to the next closing parenthesis. Nested parentheses are not permitted. If the PCRE2_EXTENDED or PCRE2_EXTENDED_MORE option 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 an option passed to the compiling function or by a special sequence at the start of the pattern, as described in the section entitled "Newline conventions" above. Note 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 this pattern when PCRE2_EXTENDED is set, and the default newline convention (a single linefeed character) is in force:
abc #comment \n still commentOn encountering the # character, pcre2_compile() 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 an actual character with the code value 0x0a (the default newline) does so. 💀
💖 RECURSIVE PATTERNS 💖
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. 😭
For some time, Perl has provided a facility that allows regular expressions to recurse (amongst other things). It does this by interpolating Perl code in the expression at run time, 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:
$re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;ok so the (?p{...}) item interpolates Perl code at run time, and in this case it refers recursively to the pattern it's chillin' inside of. 💅
Obviously, PCRE2 cannot support the interpolation of Perl code. Nope, not happening bestie. Instead, it supports special syntax for recursion of the entire pattern, and also for individual capture group recursion. After its introduction in PCRE1 and Python, this kind of recursion was subsequently introduced into Perl at release 5.10.
A special item that consists of (? followed by a number greater than zero and a closing parenthesis is a recursive subroutine call of the capture group of the given number, provided that it occurs inside that group. (If not, it's 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.
This PCRE2 pattern solves the nested parentheses problem (assume the PCRE2_EXTENDED option is set so that white space is ignored):
\( ( [^()]++ | (?R) )* \)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's a closing parenthesis. Note the use of a possessive quantifier to avoid backtracking into sequences of non-parentheses. 🫶
If this were part of a larger pattern, you would not want to recurse the entire pattern, so instead you could use this:
( \( ( [^()]++ | (?1) )* \) )We've put the pattern into parentheses, and caused the recursion to refer to them instead of the whole pattern. ✨
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. In other words, a negative number counts capturing parentheses leftwards from the point at which it is encountered.
Be aware however, that if duplicate capture group numbers are in use, relative references refer to the earliest group with the appropriate number. Consider, for example:
(?|(a)|(b)) (c) (?-2)The first two capture groups (a) and (b) are both numbered 1, and group (c) is number 2. When the reference (?-2) is encountered, the second most recently opened parentheses has the number 1, but it's the first such group (the (a) group) to which the recursion refers. This would be the same if an absolute reference (?1) was used. In other words, relative references are just a shorthand for computing a group number. 💀
It's also possible to refer to subsequent capture groups, by writing references such as (?+2). However, these cannot be recursive because the reference is not inside the parentheses that are referenced. They are always non-recursive subroutine calls, as described in the next section.
An alternative approach is to use named parentheses. The Perl syntax for this is (?&name); PCRE1's earlier syntax (?P>name) is also supported. We could rewrite the above example as follows:
(?<pn> \( ( [^()]++ | (?&pn) )* \) )If there's more than one group with the same name, the earliest one is used.
The example pattern that we've been looking at 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
(aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()it yields "no match" quickly. However, if a possessive quantifier is not used, the match runs for a very long time indeed because there are so many different ways the + and * repeats can carve up the subject, and all have to be tested before failure can be reported. 😭
At the end of a match, the values of capturing parentheses are those from the outermost level. If you want to obtain intermediate values, a callout function can be used (see below and the pcre2callout documentation). If the pattern above is matched against
(ab(cd)ef)the value for the inner capturing parentheses (numbered 2) is "ef", which is the last value taken on at the top level. If a capture group 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. 🎀
Do not confuse the (?R) item with the condition (R), which tests for recursion. Consider this pattern, which matches text in angle brackets, allowing for arbitrary nesting. Only digits are allowed in nested brackets (that is, when recursing), whereas any characters are permitted at the outer level.
< (?: (?(R) \d++ | [^<>]*+) | (?R)) * >In this pattern, (?(R) is the start of a conditional group, with two different alternatives for the recursive and non-recursive cases. The (?R) item is the actual recursive call.
🦋 Differences in recursion processing between PCRE2 and Perl 🦋
Some former differences between PCRE2 and Perl no longer exist. growth, bestie.
Before release 10.30, recursion processing in PCRE2 differed from Perl in that a recursive subroutine call was always treated as an atomic group. That is, once it had matched some of the subject string, it was never re-entered, even if it contained untried alternatives and there was a subsequent matching failure. (Historical note: PCRE implemented recursion before Perl did. 👑)
Starting with release 10.30, recursive subroutine calls are no longer treated as atomic. That is, they can be re-entered to try unused alternatives if there's a matching failure later in the pattern. This is now compatible with the way Perl works. If you want a subroutine call to be atomic, you must explicitly enclose it in an atomic group. 💅
Supporting backtracking into recursions simplifies certain types of recursive pattern. For example, this pattern matches palindromic strings:
^((.)(?1)\2|.?)$The second branch in the group matches a single central character in the palindrome when there are an odd number of characters, or nothing when there are an even number of characters, but in order to work it has to be able to try the second case when the rest of the pattern match fails. If you want to match typical palindromic phrases, the pattern has to ignore all non-word characters, which can be done like this:
^\W*+((.)\W*+(?1)\W*+\2|\W*+.?)\W*+$If run with the PCRE2_CASELESS option, this pattern matches phrases such as "A man, a plan, a canal: Panama!". Note the use of the possessive quantifier *+ to avoid backtracking into sequences of non-word characters. Without this, PCRE2 takes a great deal longer (ten times or more) to match typical phrases, and Perl takes so long that you think it has gone into a loop. 💀
Another way in which PCRE2 and Perl used to differ in their recursion processing is in the handling of captured values. Formerly in Perl, when a group was called recursively or as a subroutine (see the next section), it had no access to any values that were captured outside the recursion, whereas in PCRE2 these values can be referenced. Consider this pattern:
^(.)(\1|a(?2))This pattern matches "bab". The first capturing parentheses match "b", then in the second group, when the backreference \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. This match used to fail in Perl, but in later versions (I tried 5.024) it now works. ✨
🎀 Groups as subroutines 🎀
If the syntax for a recursive group call (either by number or by name) is used outside the parentheses to which it refers, it operates a bit like a subroutine in a programming language. More accurately, PCRE2 treats the referenced group as an independent subpattern which it tries to match at the current matching position. The called group may be defined before or after the reference. A numbered reference can be absolute or relative, as in these examples:
(...(absolute)...)...(?2)...
(...(relative)...)...(?-1)...
(...(?+1)...(relative)...An earlier example pointed out that the pattern
(sens|respons)e and \1ibilitymatches "sense and sensibility" and "response and responsibility", but not "sense and responsibility". If instead the pattern
(sens|respons)e and (?1)ibilityis used, it does match "sense and responsibility" as well as the other two strings. Another example is given in the discussion of DEFINE above. 🫶
Like recursions, subroutine calls used to be treated as atomic, but this changed at PCRE2 release 10.30, so backtracking into subroutine calls can now occur. However, any capturing parentheses that are set during the subroutine call revert to their previous values afterwards.
Processing options such as case-independence are fixed when a group is defined, so if it's used as a subroutine, such options cannot be changed for different calls. For example, consider this pattern:
(abc)(?i:(?-1))It matches "abcabc". It does not match "abcABC" because the change of processing option does not affect the called group. 💅
The behaviour of backtracking control verbs in groups when called as subroutines is described in the section entitled "Backtracking verbs in subroutines" below.
🌈 Recursion and subroutines with returned capture groups 🌈
Since PCRE2 10.47, recursion and subroutine calls may also specify a list of capture groups to return. This is a PCRE2 syntax extension not supported by Perl. The pattern matching recurses into the referenced expression as described above, however, when the recursion returns to the calling expression the subgroups captured during the recursion can be retained when the calling expression's context is restored.
When used as a subroutine, this allows the subroutine's capture groups to be used as return values. 👑
Only the specific capture groups listed by the caller will be retained, using the following syntax:
(?R(grouplist)) recurse whole pattern, returning capture groups
(?n(grouplist)) )
(?+n(grouplist)) )
(?-n(grouplist)) ) call subroutine, returning capture groups
(?&name(grouplist)) )
(?P>name(grouplist)) )The list of capture groups "grouplist" is a comma-separated list of (absolute or relative) group numbers, and group names enclosed in single quotes or angle brackets.
Here is an example which first uses the DEFINE condition to create a re-usable routine for matching a weekday, then calls that subroutine and retains the groups it captures for use later:
(?x: # ignore whitespace for clarity
# Define the routine "weekendday" which matches Saturday or
# Sunday, and returns the Sat/Sun prefix as \k<short>.
(?(DEFINE) (?<weekendday>
(?|(?<short>Sat)urday|(?<short>Sun)day) ) )
# Call the routine. Matches "Saturday,Sat" or "Sunday,Sun".
(?&weekendday(<short>)),\k<short> )okay so that whole cute little DEFINE-and-call thing? bestie, it is not available if you're using the Oniguruma syntax \g<...> or \g'...' that we're about to talk about. Different flavour, different rules, no swapping. 💅
🦋 Oniguruma subroutine syntax 🦋
So Oniguruma wanted in on this too, and for compatibility with Oniguruma, there's an alternative syntax for calling a group as a subroutine (possibly recursively, yes queen): it's the non-Perl \g followed by a name or a number, wrapped in either angle brackets or single quotes.
Here are two of the examples from up above, rewritten in this syntax:
(?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
(sens|respons)e and \g'1'ibilityAnd PCRE2 also supports an extension to Oniguruma: if a number is preceded by a plus or a minus sign, it's taken as a relative reference. Like:
(abc)(?i:\g<-1>)🚨 do NOT mix these up bestie —\g{...}(that's the Perl syntax) and\g<...>(that's the Oniguruma syntax) are not synonymous. The first one is a backreference. The second one is a subroutine call. Completely different girlies, same letter, iykyk. 😭
✨ CALLOUTS ✨
Okay so Perl has this thing where you can drop (?{...}) into a pattern and it just obeys arbitrary Perl code right in the middle of matching a regex. Which, honestly, lets you do stuff like pulling out different substrings that match the same pair of parens when there's a repetition. Iconic behaviour, truly.
PCRE2 gives you something similar — but obviously it can't just run random Perl code, that would be chaos 🫠. This feature is called a "callout".
Here's how it works: whoever's calling PCRE2 hands over an external function by putting its entry point in a match context using pcre2_set_callout(), and then passes that context to pcre2_match() or pcre2_dfa_match().
- If no match context is passed, or if the callout entry point is set to
NULL, then callout points just get silently passed over during matching. Nothing happens, no drama. 💤 - If you want to disallow callouts in the pattern syntax entirely, use the
PCRE2_EXTRA_NEVER_CALLOUToption. Shut it down, periodt.
Inside a regular expression, (?C<arg>) marks the spot where that external function is supposed to get called. 👑
There are two kinds of callout:
- ones with a numerical argument
- ones with a string argument
And (?C) all on its own with no argument at all? Treated as (?C0). Same girl, default setting. 💅
A numerical argument lets the application tell different callouts apart. String arguments came along in release 10.20 so that script languages built on PCRE2 could embed short scripts inside patterns, kind of like Perl does. 🌸
During matching, when PCRE2 hits a callout point, the external function gets called. It's handed:
- the number or string argument of the callout,
- the position in the pattern,
- and one item of data that's also set in the match block.
The callout function can make matching proceed, backtrack, or fail. Her choice, main character energy.
⚠️ by default, PCRE2 does a bunch of optimizations at matching time, and one side-effect is that sometimes callouts get skipped. If you need all possible callouts to actually happen, you have to set options that disable the relevant optimizations. More details — including the full description of the programming interface to the callout function — live in the pcre2callout documentation.
🔢 Callouts with numerical arguments 🔢
If all you want is a way to tell different callout points apart, just put a number less than 256 after the letter C. Easy. For example, this pattern has two callout points:
(?C1)abc(?C2)defIf the PCRE2_AUTO_CALLOUT flag gets passed to pcre2_compile(), numerical callouts get automatically installed before each item in the pattern. And they're all numbered 255. Every single one. 255, 255, 255... it's giving uniform. 😌
If there's a conditional group in the pattern whose condition is an assertion, an extra callout gets inserted just before the condition. And you can also set an explicit callout at that position, like this:
(?(?C9)(?=a)abc|def)note that this only applies to assertion conditions — not to other types of condition. Don't get it twisted. 🙈
🔤 Callouts with string arguments 🔤
You can use a delimited string instead of a number as a callout argument. The starting delimiter has to be one of these:
` ' " ^ % # $ {and the ending delimiter is the same as the start — except for {, where the ending delimiter is }. And if you need the ending delimiter inside the string itself, you have to double it. For example:
(?C'ab ''c'' d')xyz(?C{any text})pqrThat doubling? It gets removed before the string is passed to the callout function. Cleaned up, no crumbs. ✨
🔥 BACKTRACKING CONTROL 🔥
So there's a whole squad of special "Backtracking Control Verbs" (that's Perl's terminology, we're borrowing it) that mess with how backtracking behaves during matching. They generally look like (*VERB) or (*VERB:NAME). Some of them take either form, and might behave differently depending on whether a name argument is present or not. And no, the names don't have to be unique within the pattern. Go wild. 💅
By default — for compatibility with Perl — a name is any sequence of characters that doesn't include a closing parenthesis. The name is not processed in any way, and it's genuinely not possible to include a closing parenthesis in it. You can change that by setting the PCRE2_ALT_VERBNAMES option, but heads up: the result is no longer Perl-compatible. Trade-offs, bestie.
When PCRE2_ALT_VERBNAMES is set, backslash processing gets applied to verb names, and only an unescaped closing parenthesis terminates the name. But — and this is important — the only backslash items permitted are \Q, \E, and sequences like \x{100} that define character code points. Character type escapes such as \d get faulted. Sorry to that one. 😭
You can sneak a closing parenthesis into a name either as \) or between \Q and \E. And on top of backslash processing, if the PCRE2_EXTENDED or PCRE2_EXTENDED_MORE option is also set, unescaped white space in verb names gets skipped and #-comments are recognized — exactly like in the rest of the pattern. But note: PCRE2_EXTENDED and PCRE2_EXTENDED_MORE don't affect verb names unless PCRE2_ALT_VERBNAMES is also set. Got it? 👑
The max length of a name is 255 in the 8-bit library and 65535 in the 16-bit and 32-bit libraries. And if the name is empty — like, the closing parenthesis immediately follows the colon — the effect is as if the colon weren't there at all. Poof, gone. 🌬️
Any number of these verbs may occur in a pattern. Except for (*ACCEPT), they may not be quantified. No repeat for you, babes.
Since these verbs are specifically about backtracking, most of them can only be used when the pattern is matched with the traditional matching function or JIT, because they use backtracking algorithms. And with the exception of (*FAIL) — which behaves like a failing negative assertion — the backtracking control verbs cause an error if the DFA matching function encounters them. 🚫
The behaviour of these verbs in repeated groups, assertions, and in capture groups called as subroutines (whether or not recursively) is documented below. 👇
⚡ Optimizations that affect backtracking verbs ⚡
PCRE2 has some optimizations it runs at the start of each match attempt to speed things up. For example, it might already know the minimum length of a matching subject, or that a particular character must be present.
And when one of these optimizations bypasses the running of a match? Any backtracking verbs included in there won't get processed, obviously. 💀
You can suppress the start-of-match optimizations by:
- setting the
PCRE2_NO_START_OPTIMIZEoption when callingpcre2_compile(), - calling
pcre2_set_optimize()with aPCRE2_START_OPTIMIZE_OFFdirective, - or by starting the pattern with
(*NO_START_OPT).
There's more discussion of this option in the section called "Compiling a pattern" in the pcre2api documentation.
And fun fact: experiments with Perl suggest it has similar optimizations too, and just like PCRE2, turning them off can change the result of a match. Not me side-eyeing the determinism of it all 🙈
💥 Verbs that act immediately 💥
These verbs act as soon as they're encountered. No waiting, no queue. 💅
(*ACCEPT)or(*ACCEPT:NAME)— this verb makes the match end successfully, skipping the rest of the pattern. However (and this is the plot twist), when it's inside a capture group that's being called as a subroutine, only that group ends 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. And if(*ACCEPT)is inside capturing parentheses, the data so far gets captured.
For example:
A((?:A|B(*ACCEPT)|C)D)This matches "AB", "AAD", or "ACD"; and when it matches "AB", the "B" gets captured by the outer parentheses. 🎀
(*ACCEPT) is the only backtracking verb allowed to be quantified, because an ungreedy quantification with a minimum of zero acts only when a backtrack happens. Consider, for example,
(A(*ACCEPT)??B)Cwhere A, B, and C may be complex expressions. After matching "A", the matcher processes "BC"; if that fails, causing a backtrack, (*ACCEPT) gets triggered and the match succeeds. In both cases, all but C is captured. So whereas (*COMMIT) (see below) means "fail on backtrack", a repeated (*ACCEPT) of this type means "succeed on backtrack". Obsessed with the symmetry tbh. ✨
🚨 Warning: (*ACCEPT) should not be used within a script run group, because it causes an immediate exit from the group, bypassing the script run checking. Don't do it bestie, it's giving undefined chaos. 😭
(*FAIL)or(*FAIL:NAME)— this verb causes a matching failure, forcing backtracking to occur. It can be abbreviated to(*F). It's equivalent to(?!)but easier to read. 👑
The Perl documentation notes that it's probably useful only when combined with (?{}) or (??{}). Those are, of course, Perl features that aren't present in PCRE2. The nearest equivalent is the callout feature, as in this pattern:
a+(?C)(*FAIL)A match with the string "aaaa" always fails, but the callout is taken before each backtrack happens (in this example, 10 times). Ten whole callouts. She's working overtime. 💅
(*ACCEPT:NAME)and(*FAIL:NAME)behave the same as(*MARK:NAME)(*ACCEPT)and(*MARK:NAME)(*FAIL), respectively — that is, a(*MARK)is recorded just before the verb acts. 📝
📍 Recording which path was taken 📍
There's one verb whose main purpose is to track how a match was arrived at — though it also has a secondary use in conjunction with advancing the match starting point (see (*SKIP) below).
(*MARK:NAME)or(*:NAME)— a name is always required with this verb. For all the other backtracking control verbs, aNAMEargument is optional. Not this one. She demands a name. 👑
When a match succeeds, the name of the last-encountered mark name on the matching path is passed back to the caller, as described in the section called "Other information about the match" in the pcre2api documentation.
This applies to all instances of (*MARK) and other verbs, including those inside assertions and atomic groups. However, there are differences in those cases when (*MARK) is used in conjunction with (*SKIP) — as described below. 👀
The mark name that got seen last on the matching path is what gets passed back 💫 A verb with no NAME argument is just ignored for this purpose. Here's an example of pcre2test output, where the "mark" modifier asks for the retrieval and outputting of (*MARK) data:
re> /X(*MARK:A)Y|X(*MARK:B)Z/mark
data> XY
0: XY
MK: A
XZ
0: XZ
MK: BThe (*MARK) name gets tagged with MK: in this output, and in this example it's telling you which of the two alternatives matched. This is a way more efficient way of getting this info than putting each alternative in its own capturing parentheses, fr fr. iconic behaviour honestly ✨
If a verb with a name shows up in a positive assertion that's true, the name is recorded and passed back if it's the last-encountered one. This does not happen for negative assertions or for failing positive assertions. Periodt.
After a partial match or a failed match, the last encountered name in the entire match process is what gets returned. Like:
re> /X(*MARK:A)Y|X(*MARK:B)Z/mark
data> XP
No match, mark = BNote that in this unanchored example the mark is retained from the match attempt that started at the letter "X" in the subject. The subsequent match attempts starting at "P" and then with an empty string don't get as far as the (*MARK) item, but they still don't reset it. sneaky little guy 🦋
If you're interested in (*MARK) values after failed matches, you should probably either set the PCRE2_NO_START_OPTIMIZE option or call pcre2_set_optimize() with a PCRE2_START_OPTIMIZE_OFF directive (see above) to make sure that the match is always attempted. 🫶
🌀 Verbs that act after backtracking 🌀
These next verbs do absolutely nothing when you first encounter them. Matching just continues with whatever follows, but if there's a subsequent match failure that causes a backtrack to the verb, a failure is forced. Meaning: backtracking cannot pass to the left of the verb. The vibes are immovable 🧱
However — and this is the plot twist — when one of these verbs appears inside an atomic group or in an atomic lookaround assertion that is true, its effect is confined to that group, because once the group has been matched, there's never any backtracking into it. Backtracking from beyond an atomic assertion or group ignores the entire group and goes looking for a preceding backtracking point instead. 💅
These verbs differ in exactly what kind of failure happens when backtracking reaches them. The behaviour described below is what happens when the verb is not in a subroutine or an assertion. Subsequent sections cover those special cases, so don't @ me yet. 😌
(*COMMIT) or (*COMMIT:NAME)
This verb makes the whole match fail outright if there's 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. If (*COMMIT) is the only backtracking verb that gets encountered, then once it's been passed, pcre2_match() is committed to finding a match at the current starting point — or not at all. No in-between. For example:
a+(*COMMIT)bThis matches "xxaab" but not "aacaab". You can think of it as a kind of dynamic anchor, or like, "I've started, so I must finish." 👑
The behaviour of (*COMMIT:NAME) is not the same as (*MARK:NAME)(*COMMIT). It is like (*MARK:NAME) in that the name is remembered for passing back to the caller. However, (*SKIP:NAME) searches only for names that are set with (*MARK), ignoring those set by any of the other backtracking verbs. iykyk ✨
If there's more than one backtracking verb in a pattern, a different one that follows (*COMMIT) may be triggered first, so merely passing (*COMMIT) during a match does not always guarantee that a match must be at this starting point. 🙈
Note that (*COMMIT) at the start of a pattern is not the same as an anchor, unless PCRE2's start-of-match optimizations are turned off, as shown in this output from pcre2test:
re> /(*COMMIT)abc/
data> xyzabc
0: abc
data>
re> /(*COMMIT)abc/no_start_optimize
data> xyzabc
No matchFor the first pattern, PCRE2 knows that any match must start with "a", so the optimization skips along the subject to "a" before applying the pattern to the first set of data. The match attempt then succeeds. The second pattern disables the optimization that skips along to the first character. The pattern is now applied starting at "x", and so the (*COMMIT) causes the match to fail without trying any other starting points. optimization said "not today" 💀
(*PRUNE) or (*PRUNE:NAME)
This verb causes the match to fail at the current starting position in the subject if there's a later matching failure that causes backtracking to reach it. If the pattern is unanchored, the normal "bumpalong" advance to the next starting character then happens. Backtracking can occur as usual to the left of (*PRUNE) before it's reached, or when matching to the right of (*PRUNE), but if there's no match to the right, backtracking cannot cross (*PRUNE). 🚫
In simple cases, using (*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). versatile queen 🎀
The behaviour of (*PRUNE:NAME) is 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), ignoring those set by other backtracking verbs. 💫
(*SKIP)
This verb, when given 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. (*SKIP) is basically saying that whatever text was matched leading up to it cannot be part of a successful match if there's a later mismatch. Consider:
a+(*SKIP)bIf 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". Note 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". it's giving precision 🦋
If (*SKIP) is used to specify a new starting position that is the same as the starting position of the current match, or (by being inside a lookbehind) earlier, the position specified by (*SKIP) is ignored, and instead the normal "bumpalong" occurs. 🫠
(*SKIP:NAME)
When (*SKIP) has an associated name, its behaviour is modified. When such a (*SKIP) 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 goes 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, the (*SKIP) is just ignored. 🙈
The search for a (*MARK) name uses the normal backtracking mechanism, which means it does not see (*MARK) settings that are inside atomic groups or assertions, because those are never re-entered by backtracking. Compare the following pcre2test examples:
re> /a(?>(*MARK:X))(*SKIP:X)(*F)|(.)/
data: abc
0: a
1: a
data:
re> /a(?:(*MARK:X))(*SKIP:X)(*F)|(.)/
data: abc
0: b
1: bIn the first example, the (*MARK) setting is in an atomic group, so it is not seen when (*SKIP:X) triggers, causing the (*SKIP) to be ignored. This allows the second branch of the pattern to be tried at the first character position. In the second example, the (*MARK) setting is not in an atomic group. This allows (*SKIP:X) to find the (*MARK) when it backtracks, and this causes a new matching attempt to start at the second character. This time, the (*MARK) is never seen because "a" does not match "b", so the matcher immediately jumps to the second branch of the pattern. the way that atomic group just ✨hid✨ it
Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It ignores names that are set by other backtracking verbs. 💅
(*THEN) or (*THEN:NAME)
This verb causes a skip to the next innermost alternative when backtracking reaches it. That is, it cancels any further backtracking within the current alternative. Its name comes from the observation that it can be used for a pattern-based if-then-else block:
( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...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 subsequently BAZ fails, there are no more alternatives, so there's a backtrack to whatever came before the entire group. If (*THEN) is not inside an alternation, it acts like (*PRUNE). 🎯
The behaviour of (*THEN:NAME) is 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), ignoring those set by other backtracking verbs. ✨
A group that does not contain a | character is just a part of the enclosing alternative; it's not a nested alternation with only one alternative. The effect of (*THEN) extends beyond such a group to the enclosing alternative. Consider this pattern, where A, B, etc. are complex pattern fragments that don't contain any | characters at this level:
A (B(*THEN)C) | DIf A and B are matched, but there's a failure in C, matching does not backtrack into A; instead it moves to the next alternative, that is, D. However, if the group containing (*THEN) is given an alternative, it behaves differently:
A (B(*THEN)C | (*FAIL)) | DThe effect of (*THEN) is now confined to the inner group. After a failure in C, matching moves to (*FAIL), which causes the whole group to fail because there are no more alternatives to try. In this case, matching does backtrack into A. character development honestly 🌈
Note that a conditional group is not considered as having two alternatives, because only one is ever used. In other words, the | character in a conditional group has a different meaning. Ignoring white space, consider:
^.*? (?(?=a) a | b(*THEN)c )If the subject is "ba", this pattern does not match. Because .*? 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 might perhaps be expected from the presence of the | character. The conditional group 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.) so close bestie 😭
The verbs you just met are basically a whole <em>strength tier list</em> 💅 — four different levels of control for when the stuff after them fails to match. (*THEN) is the weakest, lowkey just carrying on the match at the next alternative. (*PRUNE) comes next: it fails the match at the current starting position, but still lets things advance to the next character (for an unanchored pattern, bestie). (*SKIP) is similar except the advance can be more than one character. And (*COMMIT) is the strongest — it causes the entire match to fail, no cap, game over. 😭
💀 More than one backtracking verb 💀
If more than one backtracking verb is present in a pattern, the one that gets backtracked onto first is the one that acts. Periodt.
So picture this pattern, where A, B, etc. are complex pattern fragments:
(A(*COMMIT)B(*THEN)C|ABD)If A matches but B fails, the backtrack to (*COMMIT) causes the entire match to fail. But if A and B match and C fails, the backtrack to (*THEN) causes the next alternative (ABD) to be tried instead. ✨
This behaviour is consistent, ngl, but it is not always the same as Perl's. What it means: if two or more backtracking verbs appear in succession, all but the last of them has no effect. Consider:
...(*COMMIT)(*PRUNE)...If there's a matching failure to the right, backtracking onto (*PRUNE) causes it to be triggered and its action gets taken. There can never be a backtrack onto (*COMMIT). 💀
🌀 Backtracking verbs in repeated groups 🌀
PCRE2 sometimes differs from Perl in how it handles backtracking verbs in repeated groups. For example, consider:
/(a(*COMMIT)b)+ac/If the subject is "abac", Perl matches unless its optimizations are disabled — but PCRE2 always fails, because the (*COMMIT) in the second repeat of the group acts. 🙈
🧚 Backtracking verbs in assertions 🧚
(*FAIL) in any assertion has its normal effect: it forces an immediate backtrack. The other backtracking verbs behave depending on whether the assertion is standalone or acting as the condition in a conditional group.
(*ACCEPT) in a standalone positive assertion causes the assertion to succeed without any further processing; captured strings and a mark name (if set) are retained. In a standalone negative assertion, (*ACCEPT) causes the assertion to fail without any further processing; captured substrings and any mark name are discarded.
If the assertion is a condition, (*ACCEPT) causes the condition to be true for a positive assertion and false for a negative one; captured substrings are retained in both cases. 💖
The remaining verbs act only when a later failure causes a backtrack to reach them. Which means: for the Perl-compatible assertions, their effect is confined to the assertion itself, because Perl lookaround assertions are atomic. A backtrack that happens after such an assertion is complete does not jump back into the assertion. And note especially that a (*MARK) name set in an assertion is not "seen" by an instance of (*SKIP:NAME) later in the pattern. 👀
PCRE2 now supports non-atomic positive assertions and also "scan substring" assertions, as described in the sections entitled "Non-atomic assertions" and "Scan substring assertions" above. These assertions must be standalone (not used as conditions). They are not Perl-compatible. For these assertions, a later backtrack <em>does</em> jump back into the assertion, and therefore verbs such as (*COMMIT) can be triggered by backtracks from later in the pattern.
The effect of (*THEN) is not allowed to escape beyond an assertion. If there are no more branches to try, (*THEN) causes a positive assertion to be false, and a negative assertion to be true. This behaviour differs from Perl when the assertion has only one branch.
The other backtracking verbs are not treated specially if they appear in a standalone positive assertion. In a conditional positive assertion, backtracking (from within the assertion) into (*COMMIT), (*SKIP), or (*PRUNE) causes the condition to be false. However, for both standalone and conditional negative assertions, backtracking into (*COMMIT), (*SKIP), or (*PRUNE) causes the assertion to be true, without considering any further alternative branches. ✨
🎀 Backtracking verbs in subroutines 🎀
These behaviours occur whether or not the group is called recursively.
(*ACCEPT) in a group called as a subroutine causes the subroutine match to succeed without any further processing. Matching then continues after the subroutine call. Perl documents this behaviour. Perl's treatment of the other verbs in subroutines is different in some cases.
(*FAIL) in a group called as a subroutine has its normal effect: it forces an immediate backtrack.
(*COMMIT), (*SKIP), and (*PRUNE) cause the subroutine match to fail when triggered by being backtracked to in a group called as a subroutine. There is then a backtrack at the outer level.
(*THEN), when triggered, skips to the next alternative in the innermost enclosing group that has alternatives (its normal behaviour). However, if there is no such group within the subroutine's group, the subroutine match fails and there is a backtrack at the outer level.
🦋 EBCDIC ENVIRONMENTS 🦋
All the ways PCRE behaves differently when it's running in an EBCDIC environment are covered in this section. Buckle up. 🫶
🌸 Escape sequences 🌸
When PCRE2 is compiled in EBCDIC mode, \N{U+hhh..} is not supported. \a, \e, \f, \n, \r, and \t generate the appropriate EBCDIC code values. The \c escape is processed as specified for Perl in the perlebcdic document. The only characters allowed after \c are A-Z, a-z, or one of @, [, \, ], ^, _, or ?. Any other character provokes a compile-time error. 😭
The sequence \c@ encodes character code 0; after \c the letters (in either case) encode characters 1-26 (hex 01 to hex 1A); [, \, ], ^, and _ encode characters 27-31 (hex 1B to hex 1F), and \c? becomes either 255 (hex FF) or 95 (hex 5F).
So, apart from \c?, these escapes generate the same character code values as they do in an ASCII or Unicode environment — though the <em>meanings</em> of those values mostly differ. For example, \cG always generates code value 7, which is BEL in ASCII but DEL in EBCDIC. 💅
The sequence \c? generates DEL (127, hex 7F) in an ASCII environment, but because 127 is not a control character in EBCDIC, Perl makes it generate the APC character. Unfortunately, there are several variants of EBCDIC. In most of them the APC character has the value 255 (hex FF), but in the one Perl calls POSIX-BC its value is 95 (hex 5F). If certain other characters have POSIX-BC values, PCRE2 makes \c? generate 95; otherwise it generates 255.
🍓 Character classes 🍓
In character classes there's a special case in EBCDIC environments for ranges whose end points are both specified as literal letters in the same case. For compatibility with Perl, EBCDIC code points within the range that are not letters are omitted. For example, [h-k] matches only four characters, even though the EBCDIC codes for h and k are 0x88 and 0x92 — a range of 11 code points. However, if the range is specified numerically, for example, [\x88-\x92] or [h-\x92], all code points are included. 👑
👀 SEE ALSO 👀
💖 AUTHOR 💖
Philip Hazel Retired from University Computing Service Cambridge, England.
📅 REVISION 📅
Last updated: 25 October 2025 Copyright (c) 1997-2024 University of Cambridge.
And that's a wrap, bestie!! 🎀 This was PCRE2 10.49, dated 25 October 2025, and you have officially read the whole entire PCRE2PATTERN(3) page, which is honestly iconic behaviour. Your regexes are gonna slay so hard now. Go forth and match, my love 💜✨
📦 where this page comes from
- alpine 3.23: pcre2-doc 10.49-r0 · BSD-3-Clause
- alpine 3.24: pcre2-doc 10.49-r0 · BSD-3-Clause
- arch rolling: pcre2 10.49-1 · BSD-2-Clause AND BSD-3-Clause WITH PCRE2-exception
- 💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms