perlrebackslash - Perl Regular Expression Backslash Sequences and Escapes
The top level documentation about Perl regular expressions is found in perlre.
This document describes all backslash and escape sequences. After explaining the role of the backslash, it lists all the sequences that have a special meaning in Perl regular expressions (in alphabetical order), then describes each of them.
Most sequences are described in detail in different documents; the primary purpose of this document is to have a quick reference guide describing all backslash and escape sequences.
In a regular expression, the backslash can perform one of two tasks:
it either takes away the special meaning of the character following it
(for instance, \| matches a vertical bar, it's not an alternation),
or it is the start of a backslash or escape sequence.
The rules determining what it is are quite simple: if the character following the backslash is an ASCII punctuation (non-word) character (that is, anything that is not a letter, digit, or underscore), then the backslash just takes away any special meaning of the character following it.
If the character following the backslash is an ASCII letter or an ASCII digit, then the sequence may be special; if so, it's listed below. A few letters have not been used yet, so escaping them with a backslash doesn't change them to be special. A future version of Perl may assign a special meaning to them, so if you have warnings turned on, Perl issues a warning if you use such a sequence. [1].
It is however guaranteed that backslash or escape sequences never have a punctuation character following the backslash, not now, and not in a future version of Perl 5. So it is safe to put a backslash in front of a non-word character.
Note that the backslash itself is special; if you want to match a backslash,
you have to escape the backslash with a backslash: /\\/
matches a single
backslash.
There is one exception. If you use an alphanumeric character as the delimiter of your pattern (which you probably shouldn't do for readability reasons), you have to escape the delimiter if you want to match it. Perl won't warn then. See also Gory details of parsing quoted constructs in perlop.
Those not usable within a bracketed character class (like [\da-z]
) are marked
as Not in [].
- \000 Octal escape sequence. See also \o{}.
- \1 Absolute backreference. Not in [].
- \a Alarm or bell.
- \A Beginning of string. Not in [].
- \b{}, \b Boundary. (\b is a backspace in []).
- \B{}, \B Not a boundary. Not in [].
- \cX Control-X.
- \C Single octet, even under UTF-8. Not in [].
- (Deprecated)
- \d Character class for digits.
- \D Character class for non-digits.
- \e Escape character.
- \E Turn off \Q, \L and \U processing. Not in [].
- \f Form feed.
- \F Foldcase till \E. Not in [].
- \g{}, \g1 Named, absolute or relative backreference.
- Not in [].
- \G Pos assertion. Not in [].
- \h Character class for horizontal whitespace.
- \H Character class for non horizontal whitespace.
- \k{}, \k<>, \k'' Named backreference. Not in [].
- \K Keep the stuff left of \K. Not in [].
- \l Lowercase next character. Not in [].
- \L Lowercase till \E. Not in [].
- \n (Logical) newline character.
- \N Any character but newline. Not in [].
- \N{} Named or numbered (Unicode) character or sequence.
- \o{} Octal escape sequence.
- \p{}, \pP Character with the given Unicode property.
- \P{}, \PP Character without the given Unicode property.
- \Q Quote (disable) pattern metacharacters till \E. Not
- in [].
- \r Return character.
- \R Generic new line. Not in [].
- \s Character class for whitespace.
- \S Character class for non whitespace.
- \t Tab character.
- \u Titlecase next character. Not in [].
- \U Uppercase till \E. Not in [].
- \v Character class for vertical whitespace.
- \V Character class for non vertical whitespace.
- \w Character class for word characters.
- \W Character class for non-word characters.
- \x{}, \x00 Hexadecimal escape sequence.
- \X Unicode "extended grapheme cluster". Not in [].
- \z End of string. Not in [].
- \Z End of string. Not in [].
A handful of characters have a dedicated character escape. The following table shows them, along with their ASCII code points (in decimal and hex), their ASCII name, the control escape on ASCII platforms and a short description. (For EBCDIC platforms, see OPERATOR DIFFERENCES in perlebcdic.)
\b
is the backspace character only inside a character class. Outside a
character class, \b
alone is a word-character/non-word-character
boundary, and \b{}
is some other type of boundary.
\n
matches a logical newline. Perl converts between \n
and your
OS's native newline character when reading from or writing to text files.
- $str =~ /\t/; # Matches if $str contains a (horizontal) tab.
\c
is used to denote a control character; the character following \c
determines the value of the construct. For example the value of \cA
is
chr(1), and the value of \cb
is chr(2), etc.
The gory details are in Regexp Quote-Like Operators in perlop. A complete
list of what chr(1), etc. means for ASCII and EBCDIC platforms is in
OPERATOR DIFFERENCES in perlebcdic.
Note that \c\
alone at the end of a regular expression (or doubled-quoted
string) is not valid. The backslash must be followed by another character.
That is, \c\X means chr(28) . 'X' for all characters X.
To write platform-independent code, you must use \N{NAME} instead, like
\N{ESCAPE}
or \N{U+001B}
, see charnames.
Mnemonic: control character.
- $str =~ /\cK/; # Matches if $str contains a vertical tab (control-K).
Unicode characters have a Unicode name and numeric code point (ordinal)
value. Use the
\N{}
construct to specify a character by either of these values.
Certain sequences of characters also have names.
To specify by name, the name of the character or character sequence goes between the curly braces.
To specify a character by Unicode code point, use the form \N{U+code
point}, where code point is a number in hexadecimal that gives the
code point that Unicode has assigned to the desired character. It is
customary but not required to use leading zeros to pad the number to 4
digits. Thus \N{U+0041}
means LATIN CAPITAL LETTER A
, and you will
rarely see it written without the two leading zeros. \N{U+0041}
means
"A" even on EBCDIC machines (where the ordinal value of "A" is not 0x41).
It is even possible to give your own names to characters and character sequences. For details, see charnames.
(There is an expanded internal form that you may see in debug output:
\N{U+code point.code point...}.
The ...
means any number of these code points separated by dots.
This represents the sequence formed by the characters. This is an internal
form only, subject to change, and you should not try to use it yourself.)
Mnemonic: Named character.
Note that a character or character sequence expressed as a named or numbered character is considered a character without special meaning by the regex engine, and will match "as is".
- $str =~ /\N{THAI CHARACTER SO SO}/; # Matches the Thai SO SO character
- use charnames 'Cyrillic'; # Loads Cyrillic names.
- $str =~ /\N{ZHE}\N{KA}/; # Match "ZHE" followed by "KA".
There are two forms of octal escapes. Each is used to specify a character by its code point specified in octal notation.
One form, available starting in Perl 5.14 looks like \o{...}
, where the dots
represent one or more octal digits. It can be used for any Unicode character.
It was introduced to avoid the potential problems with the other form, available in all Perls. That form consists of a backslash followed by three octal digits. One problem with this form is that it can look exactly like an old-style backreference (see Disambiguation rules between old-style octal escapes and backreferences below.) You can avoid this by making the first of the three digits always a zero, but that makes \077 the largest code point specifiable.
In some contexts, a backslash followed by two or even one octal digits may be interpreted as an octal escape, sometimes with a warning, and because of some bugs, sometimes with surprising results. Also, if you are creating a regex out of smaller snippets concatenated together, and you use fewer than three digits, the beginning of one snippet may be interpreted as adding digits to the ending of the snippet before it. See Absolute referencing for more discussion and examples of the snippet problem.
Note that a character expressed as an octal escape is considered a character without special meaning by the regex engine, and will match "as is".
To summarize, the \o{}
form is always safe to use, and the other form is
safe to use for code points through \077 when you use exactly three digits to
specify them.
Mnemonic: 0ctal or octal.
- $str = "Perl";
- $str =~ /\o{120}/; # Match, "\120" is "P".
- $str =~ /\120/; # Same.
- $str =~ /\o{120}+/; # Match, "\120" is "P",
- # it's repeated at least once.
- $str =~ /\120+/; # Same.
- $str =~ /P\053/; # No match, "\053" is "+" and taken literally.
- /\o{23073}/ # Black foreground, white background smiling face.
- /\o{4801234567}/ # Raises a warning, and yields chr(4).
Octal escapes of the \000
form outside of bracketed character classes
potentially clash with old-style backreferences (see Absolute referencing
below). They both consist of a backslash followed by numbers. So Perl has to
use heuristics to determine whether it is a backreference or an octal escape.
Perl uses the following rules to disambiguate:
If the backslash is followed by a single digit, it's a backreference.
If the first digit following the backslash is a 0, it's an octal escape.
If the number following the backslash is N (in decimal), and Perl already has seen N capture groups, Perl considers this a backreference. Otherwise, it considers it an octal escape. If N has more than three digits, Perl takes only the first three for the octal escape; the rest are matched as is.
- my $pat = "(" x 999;
- $pat .= "a";
- $pat .= ")" x 999;
- /^($pat)\1000$/; # Matches 'aa'; there are 1000 capture groups.
- /^$pat\1000$/; # Matches 'a@0'; there are 999 capture groups
- # and \1000 is seen as \100 (a '@') and a '0'.
You can force a backreference interpretation always by using the \g{...}
form. You can the force an octal interpretation always by using the \o{...}
form, or for numbers up through \077 (= 63 decimal), by using three digits,
beginning with a "0".
Like octal escapes, there are two forms of hexadecimal escapes, but both start
with the sequence \x
. This is followed by either exactly two hexadecimal
digits forming a number, or a hexadecimal number of arbitrary length surrounded
by curly braces. The hexadecimal number is the code point of the character you
want to express.
Note that a character expressed as one of these escapes is considered a character without special meaning by the regex engine, and will match "as is".
Mnemonic: hexadecimal.
- $str = "Perl";
- $str =~ /\x50/; # Match, "\x50" is "P".
- $str =~ /\x50+/; # Match, "\x50" is "P", it is repeated at least once
- $str =~ /P\x2B/; # No match, "\x2B" is "+" and taken literally.
- /\x{2603}\x{2602}/ # Snowman with an umbrella.
- # The Unicode character 2603 is a snowman,
- # the Unicode character 2602 is an umbrella.
- /\x{263B}/ # Black smiling face.
- /\x{263b}/ # Same, the hex digits A - F are case insensitive.
A number of backslash sequences have to do with changing the character,
or characters following them. \l
will lowercase the character following
it, while \u
will uppercase (or, more accurately, titlecase) the
character following it. They provide functionality similar to the
functions lcfirst and ucfirst.
To uppercase or lowercase several characters, one might want to use
\L
or \U
, which will lowercase/uppercase all characters following
them, until either the end of the pattern or the next occurrence of
\E
, whichever comes first. They provide functionality similar to what
the functions lc and uc provide.
\Q
is used to quote (disable) pattern metacharacters, up to the next
\E
or the end of the pattern. \Q
adds a backslash to any character
that could have special meaning to Perl. In the ASCII range, it quotes
every character that isn't a letter, digit, or underscore. See
quotemeta for details on what gets quoted for non-ASCII
code points. Using this ensures that any character between \Q
and
\E
will be matched literally, not interpreted as a metacharacter by
the regex engine.
\F
can be used to casefold all characters following, up to the next \E
or the end of the pattern. It provides the functionality similar to
the fc function.
Mnemonic: Lowercase, Uppercase, Fold-case, Quotemeta, End.
- $sid = "sid";
- $greg = "GrEg";
- $miranda = "(Miranda)";
- $str =~ /\u$sid/; # Matches 'Sid'
- $str =~ /\L$greg/; # Matches 'greg'
- $str =~ /\Q$miranda\E/; # Matches '(Miranda)', as if the pattern
- # had been written as /\(Miranda\)/
Perl regular expressions have a large range of character classes. Some of the character classes are written as a backslash sequence. We will briefly discuss those here; full details of character classes can be found in perlrecharclass.
\w
is a character class that matches any single word character
(letters, digits, Unicode marks, and connector punctuation (like the
underscore)). \d
is a character class that matches any decimal
digit, while the character class \s matches any whitespace character.
New in perl 5.10.0 are the classes \h
and \v
which match horizontal
and vertical whitespace characters.
The exact set of characters matched by \d
, \s, and \w
varies
depending on various pragma and regular expression modifiers. It is
possible to restrict the match to the ASCII range by using the /a
regular expression modifier. See perlrecharclass.
The uppercase variants (\W
, \D
, \S
, \H
, and \V
) are
character classes that match, respectively, any character that isn't a
word character, digit, whitespace, horizontal whitespace, or vertical
whitespace.
Mnemonics: word, digit, space, horizontal, vertical.
\pP
(where P
is a single letter) and \p{Property}
are used to
match a character that matches the given Unicode property; properties
include things like "letter", or "thai character". Capitalizing the
sequence to \PP
and \P{Property}
make the sequence match a character
that doesn't match the given Unicode property. For more details, see
Backslash sequences in perlrecharclass and
Unicode Character Properties in perlunicode.
Mnemonic: property.
If capturing parenthesis are used in a regular expression, we can refer to the part of the source string that was matched, and match exactly the same thing. There are three ways of referring to such backreference: absolutely, relatively, and by name.
Either \gN (starting in Perl 5.10.0), or \N (old-style) where N
is a positive (unsigned) decimal number of any length is an absolute reference
to a capturing group.
N refers to the Nth set of parentheses, so \gN refers to whatever has
been matched by that set of parentheses. Thus \g1
refers to the first
capture group in the regex.
The \gN form can be equivalently written as \g{N}
which avoids ambiguity when building a regex by concatenating shorter
strings. Otherwise if you had a regex qr/$a$b/, and $a
contained
"\g1"
, and $b
contained "37"
, you would get /\g137/
which is
probably not what you intended.
In the \N form, N must not begin with a "0", and there must be at
least N capturing groups, or else N is considered an octal escape
(but something like \18
is the same as \0018
; that is, the octal escape
"\001"
followed by a literal digit "8"
).
Mnemonic: group.
- /(\w+) \g1/; # Finds a duplicated word, (e.g. "cat cat").
- /(\w+) \1/; # Same thing; written old-style.
- /(.)(.)\g2\g1/; # Match a four letter palindrome (e.g. "ABBA").
\g-N (starting in Perl 5.10.0) is used for relative addressing. (It can
be written as \g{-N.) It refers to the Nth group before the
\g{-N}.
The big advantage of this form is that it makes it much easier to write patterns with references that can be interpolated in larger patterns, even if the larger pattern also contains capture groups.
- /(A) # Group 1
- ( # Group 2
- (B) # Group 3
- \g{-1} # Refers to group 3 (B)
- \g{-3} # Refers to group 1 (A)
- )
- /x; # Matches "ABBA".
- my $qr = qr /(.)(.)\g{-2}\g{-1}/; # Matches 'abab', 'cdcd', etc.
- /$qr$qr/ # Matches 'ababcdcd'.
\g{name} (starting in Perl 5.10.0) can be used to back refer to a
named capture group, dispensing completely with having to think about capture
buffer positions.
To be compatible with .Net regular expressions, \g{name}
may also be
written as \k{name}
, \k<name>
or \k'name'.
To prevent any ambiguity, name must not start with a digit nor contain a hyphen.
- /(?<word>\w+) \g{word}/ # Finds duplicated word, (e.g. "cat cat")
- /(?<word>\w+) \k{word}/ # Same.
- /(?<word>\w+) \k<word>/ # Same.
- /(?<letter1>.)(?<letter2>.)\g{letter2}\g{letter1}/
- # Match a four letter palindrome (e.g. "ABBA")
Assertions are conditions that have to be true; they don't actually match parts of the substring. There are six assertions that are written as backslash sequences.
\A
only matches at the beginning of the string. If the /m modifier
isn't used, then /\A/
is equivalent to /^/
. However, if the /m
modifier is used, then /^/
matches internal newlines, but the meaning
of /\A/
isn't changed by the /m modifier. \A
matches at the beginning
of the string regardless whether the /m modifier is used.
\z
and \Z
match at the end of the string. If the /m modifier isn't
used, then /\Z/
is equivalent to /$/
; that is, it matches at the
end of the string, or one before the newline at the end of the string. If the
/m modifier is used, then /$/
matches at internal newlines, but the
meaning of /\Z/
isn't changed by the /m modifier. \Z
matches at
the end of the string (or just before a trailing newline) regardless whether
the /m modifier is used.
\z
is just like \Z
, except that it does not match before a trailing
newline. \z
matches at the end of the string only, regardless of the
modifiers used, and not just before a newline. It is how to anchor the
match to the true end of the string under all conditions.
\G
is usually used only in combination with the /g modifier. If the
/g modifier is used and the match is done in scalar context, Perl
remembers where in the source string the last match ended, and the next time,
it will start the match from where it ended the previous time.
\G
matches the point where the previous match on that string ended,
or the beginning of that string if there was no previous match.
Mnemonic: Global.
\b{...}
, available starting in v5.22, matches a boundary (between two
characters, or before the first character of the string, or after the
final character of the string) based on the Unicode rules for the
boundary type specified inside the braces. The currently known boundary
types are given a few paragraphs below. \B{...}
matches at any place
between characters where \b{...}
of the same type doesn't match.
\b
when not immediately followed by a "{"
matches at any place
between a word (something matched by \w
) and a non-word character
(\W
); \B
when not immediately followed by a "{"
matches at any
place between characters where \b
doesn't match. To get better
word matching of natural language text, see \b{wb} below.
\b
and \B
assume there's a non-word character before the beginning and after
the end of the source string; so \b
will match at the beginning (or end)
of the source string if the source string begins (or ends) with a word
character. Otherwise, \B
will match.
Do not use something like \b=head\d\b
and expect it to match the
beginning of a line. It can't, because for there to be a boundary before
the non-word "=", there must be a word character immediately previous.
All plain \b
and \B
boundary determinations look for word
characters alone, not for
non-word characters nor for string ends. It may help to understand how
<\b> and <\B> work by equating them as follows:
- \b really means (?:(?<=\w)(?!\w)|(?<!\w)(?=\w))
- \B really means (?:(?<=\w)(?=\w)|(?<!\w)(?!\w))
In contrast, \b{...}
and \B{...}
may or may not match at the
beginning and end of the line, depending on the boundary type. These
implement the Unicode default boundaries, specified in
http://www.unicode.org/reports/tr29/.
The boundary types currently available are:
\b{gcb}
or \b{g}
This matches a Unicode "Grapheme Cluster Boundary". (Actually Perl
always uses the improved "extended" grapheme cluster"). These are
explained below under \X. In fact, \X
is another way to get
the same functionality. It is equivalent to /.+?\b{gcb}/
. Use
whichever is most convenient for your situation.
\b{sb}
This matches a Unicode "Sentence Boundary". This is an aid to parsing
natural language sentences. It gives good, but imperfect results. For
example, it thinks that "Mr. Smith" is two sentences. More details are
at http://www.unicode.org/reports/tr29/. Note also that it thinks
that anything matching \R (except form feed and vertical tab) is a
sentence boundary. \b{sb}
works with text designed for
word-processors which wrap lines
automatically for display, but hard-coded line boundaries are considered
to be essentially the ends of text blocks (paragraphs really), and hence
the ends of sententces. \b{sb}
doesn't do well with text containing
embedded newlines, like the source text of the document you are reading.
Such text needs to be preprocessed to get rid of the line separators
before looking for sentence boundaries. Some people view this as a bug
in the Unicode standard, and this behavior is quite subject to change in
future Perl versions.
\b{wb}
This matches a Unicode "Word Boundary". This gives better (though not
perfect) results for natural language processing than plain \b
(without braces) does. For example, it understands that apostrophes can
be in the middle of words and that parentheses aren't (see the examples
below). More details are at http://www.unicode.org/reports/tr29/.
It is important to realize when you use these Unicode boundaries, that you are taking a risk that a future version of Perl which contains a later version of the Unicode Standard will not work precisely the same way as it did when your code was written. These rules are not considered stable and have been somewhat more subject to change than the rest of the Standard. Unicode reserves the right to change them at will, and Perl reserves the right to update its implementation to Unicode's new rules. In the past, some changes have been because new characters have been added to the Standard which have different characteristics than all previous characters, so new rules are formulated for handling them. These should not cause any backward compatibility issues. But some changes have changed the treatment of existing characters because the Unicode Technical Committee has decided that the change is warranted for whatever reason. This could be to fix a bug, or because they think better results are obtained with the new rule.
It is also important to realize that these are default boundary definitions, and that implementations may wish to tailor the results for particular purposes and locales.
Unicode defines a fourth boundary type, accessible through the Unicode::LineBreak module.
Mnemonic: boundary.
- "cat" =~ /\Acat/; # Match.
- "cat" =~ /cat\Z/; # Match.
- "cat\n" =~ /cat\Z/; # Match.
- "cat\n" =~ /cat\z/; # No match.
- "cat" =~ /\bcat\b/; # Matches.
- "cats" =~ /\bcat\b/; # No match.
- "cat" =~ /\bcat\B/; # No match.
- "cats" =~ /\bcat\B/; # Match.
- while ("cat dog" =~ /(\w+)/g) {
- print $1; # Prints 'catdog'
- }
- while ("cat dog" =~ /\G(\w+)/g) {
- print $1; # Prints 'cat'
- }
- my $s = "He said, \"Is pi 3.14? (I'm not sure).\"";
- print join("|", $s =~ m/ ( .+? \b ) /xg), "\n";
- print join("|", $s =~ m/ ( .+? \b{wb} ) /xg), "\n";
- prints
- He| |said|, "|Is| |pi| |3|.|14|? (|I|'|m| |not| |sure
- He| |said|,| |"|Is| |pi| |3.14|?| |(|I'm| |not| |sure|)|.|"
Here we document the backslash sequences that don't fall in one of the categories above. These are:
(Deprecated.) \C
always matches a single octet, even if the source
string is encoded
in UTF-8 format, and the character to be matched is a multi-octet character.
This is very dangerous, because it violates
the logical character abstraction and can cause UTF-8 sequences to become malformed.
Use utf8::encode()
instead.
Mnemonic: oCtet.
This appeared in perl 5.10.0. Anything matched left of \K
is
not included in $&
, and will not be replaced if the pattern is
used in a substitution. This lets you write s/PAT1 \K PAT2/REPL/x
instead of s/(PAT1) PAT2/${1}REPL/x
or s/(?<=PAT1) PAT2/REPL/x
.
Mnemonic: Keep.
This feature, available starting in v5.12, matches any character
that is not a newline. It is a short-hand for writing [^\n], and is
identical to the . metasymbol, except under the /s flag, which changes
the meaning of ., but not \N
.
Note that \N{...}
can mean a
named or numbered character .
Mnemonic: Complement of \n.
\R
matches a generic newline; that is, anything considered a
linebreak sequence by Unicode. This includes all characters matched by
\v
(vertical whitespace), and the multi character sequence "\x0D\x0A"
(carriage return followed by a line feed, sometimes called the network
newline; it's the end of line sequence used in Microsoft text files opened
in binary mode). \R
is equivalent to (?>\x0D\x0A|\v)
. (The
reason it doesn't backtrack is that the sequence is considered
inseparable. That means that
- "\x0D\x0A" =~ /^\R\x0A$/ # No match
fails, because the \R
matches the entire string, and won't backtrack
to match just the "\x0D"
.) Since
\R
can match a sequence of more than one character, it cannot be put
inside a bracketed character class; /[\R]/
is an error; use \v
instead. \R
was introduced in perl 5.10.0.
Note that this does not respect any locale that might be in effect; it matches according to the platform's native character set.
Mnemonic: none really. \R
was picked because PCRE already uses \R
,
and more importantly because Unicode recommends such a regular expression
metacharacter, and suggests \R
as its notation.
This matches a Unicode extended grapheme cluster.
\X
matches quite well what normal (non-Unicode-programmer) usage
would consider a single character. As an example, consider a G with some sort
of diacritic mark, such as an arrow. There is no such single character in
Unicode, but one can be composed by using a G followed by a Unicode "COMBINING
UPWARDS ARROW BELOW", and would be displayed by Unicode-aware software as if it
were a single character.
The match is greedy and non-backtracking, so that the cluster is never broken up into smaller components.
See also \b{gcb} .
Mnemonic: eXtended Unicode character.
- $str =~ s/foo\Kbar/baz/g; # Change any 'bar' following a 'foo' to 'baz'
- $str =~ s/(.)\K\g1//g; # Delete duplicated characters.
- "\n" =~ /^\R$/; # Match, \n is a generic newline.
- "\r" =~ /^\R$/; # Match, \r is a generic newline.
- "\r\n" =~ /^\R$/; # Match, \r\n is a generic newline.
- "P\x{307}" =~ /^\X$/ # \X matches a P with a dot above.