123456789_123456789_123456789_123456789_123456789_

A regular expression (also called a regexp) is a match pattern (also simply called a pattern).

A common notation for a regexp uses enclosing slash characters:

/foo/

A regexp may be applied to a target string; The part of the string (if any) that matches the pattern is called a match, and may be said to match:

re = /red/
re.match?('redirect') # => true   # Match at beginning of target.
re.match?('bored')    # => true   # Match at end of target.
re.match?('credit')   # => true   # Match within target.
re.match?('foo')      # => false  # No match.

Regexp Uses

A regexp may be used:

Regexp Objects

A regexp object has:

Creating a Regexp

A regular expression may be created with:

Method match

Each of the methods Regexp#match, String#match, and Symbol#match returns a MatchData object if a match was found, nil otherwise; each also sets global variables:

'food'.match(/foo/) # => #<MatchData "foo">
'food'.match(/bar/) # => nil

Operator =~

Each of the operators Regexp#=~, String#=~, and Symbol#=~ returns an integer offset if a match was found, nil otherwise; each also sets global variables:

/bar/ =~ 'foo bar' # => 4
'foo bar' =~ /bar/ # => 4
/baz/ =~ 'foo bar' # => nil

Method match?

Each of the methods Regexp#match?, String#match?, and Symbol#match? returns true if a match was found, false otherwise; none sets global variables:

'food'.match?(/foo/) # => true
'food'.match?(/bar/) # => false

Global Variables

Certain regexp-oriented methods assign values to global variables:

The affected global variables are:

Examples:

# Matched string, but no matched groups.
'foo bar bar baz'.match('bar')
$~ # => #<MatchData "bar">
$& # => "bar"
$` # => "foo "
$' # => " bar baz"
$+ # => nil
$1 # => nil

# Matched groups.
/s(\w{2}).*(c)/.match('haystack')
$~ # => #<MatchData "stac" 1:"ta" 2:"c">
$& # => "stac"
$` # => "hay"
$' # => "k"
$+ # => "c"
$1 # => "ta"
$2 # => "c"
$3 # => nil

# No match.
'foo'.match('bar')
$~ # => nil
$& # => nil
$` # => nil
$' # => nil
$+ # => nil
$1 # => nil

Note that Regexp#match?, String#match?, and Symbol#match? do not set global variables.

Sources

As seen above, the simplest regexp uses a literal expression as its source:

re = /foo/              # => /foo/
re.match('food')        # => #<MatchData "foo">
re.match('good')        # => nil

A rich collection of available subexpressions gives the regexp great power and flexibility:

Special Characters

Regexp special characters, called metacharacters, have special meanings in certain contexts; depending on the context, these are sometimes metacharacters:

. ? - + * ^ \ | $ ( ) [ ] { }

To match a metacharacter literally, backslash-escape it:

# Matches one or more 'o' characters.
/o+/.match('foo')  # => #<MatchData "oo">
# Would match 'o+'.
/o\+/.match('foo') # => nil

To match a backslash literally, backslash-escape it:

/\./.match('\.')  # => #<MatchData ".">
/\\./.match('\.') # => #<MatchData "\\.">

Method Regexp.escape returns an escaped string:

Regexp.escape('.?-+*^\|$()[]{}')
# => "\\.\\?\\-\\+\\*\\^\\\\\\|\\$\\(\\)\\[\\]\\{\\}"

Source Literals

The source literal largely behaves like a double-quoted string; see Double-Quoted String Literals.

In particular, a source literal may contain interpolated expressions:

s = 'foo'         # => "foo"
/#{s}/            # => /foo/
/#{s.capitalize}/ # => /Foo/
/#{2 + 2}/        # => /4/

There are differences between an ordinary string literal and a source literal; see Shorthand Character Classes.

Character Classes

A character class is delimited by square brackets; it specifies that certain characters match at a given point in the target string:

# This character class will match any vowel.
re = /B[aeiou]rd/
re.match('Bird') # => #<MatchData "Bird">
re.match('Bard') # => #<MatchData "Bard">
re.match('Byrd') # => nil

A character class may contain hyphen characters to specify ranges of characters:

# These regexps have the same effect.
/[abcdef]/.match('foo') # => #<MatchData "f">
/[a-f]/.match('foo')    # => #<MatchData "f">
/[a-cd-f]/.match('foo') # => #<MatchData "f">

When the first character of a character class is a caret (^), the sense of the class is inverted: it matches any character except those specified.

/[^a-eg-z]/.match('f') # => #<MatchData "f">

A character class may contain another character class. By itself this isn’t useful because [a-z[0-9]] describes the same set as [a-z0-9].

However, character classes also support the && operator, which performs set intersection on its arguments. The two can be combined as follows:

/[a-w&&[^c-g]z]/ # ([a-w] AND ([^c-g] OR z))

This is equivalent to:

/[abh-w]/

Shorthand Character Classes

Each of the following metacharacters serves as a shorthand for a character class:

Anchors

An anchor is a metasequence that matches a zero-width position between characters in the target string.

For a subexpression with no anchor, matching may begin anywhere in the target string:

/real/.match('surrealist') # => #<MatchData "real">

For a subexpression with an anchor, matching must begin at the matched anchor.

Boundary Anchors

Each of these anchors matches a boundary:

Lookaround Anchors

Lookahead anchors:

Lookbehind anchors:

The pattern below uses positive lookahead and positive lookbehind to match text appearing in … tags without including the tags in the match:

/(?<=<b>)\w+(?=<\/b>)/.match("Fortune favors the <b>bold</b>.")
# => #<MatchData "bold">

Match-Reset Anchor

Alternation

The vertical bar metacharacter (|) may be used within parentheses to express alternation: two or more subexpressions any of which may match the target string.

Two alternatives:

re = /(a|b)/
re.match('foo') # => nil
re.match('bar') # => #<MatchData "b" 1:"b">

Four alternatives:

re = /(a|b|c|d)/
re.match('shazam') # => #<MatchData "a" 1:"a">
re.match('cold')   # => #<MatchData "c" 1:"c">

Each alternative is a subexpression, and may be composed of other subexpressions:

re = /([a-c]|[x-z])/
re.match('bar') # => #<MatchData "b" 1:"b">
re.match('ooz') # => #<MatchData "z" 1:"z">

Method Regexp.union provides a convenient way to construct a regexp with alternatives.

Quantifiers

A simple regexp matches one character:

/\w/.match('Hello')  # => #<MatchData "H">

An added quantifier specifies how many matches are required or allowed:

Greedy, Lazy, or Possessive Matching

Quantifier matching may be greedy, lazy, or possessive:

More:

Groups and Captures

A simple regexp has (at most) one match:

re = /\d\d\d\d-\d\d-\d\d/
re.match('1943-02-04')      # => #<MatchData "1943-02-04">
re.match('1943-02-04').size # => 1
re.match('foo')             # => nil

Adding one or more pairs of parentheses, (subexpression), defines groups, which may result in multiple matched substrings, called captures:

re = /(\d\d\d\d)-(\d\d)-(\d\d)/
re.match('1943-02-04')      # => #<MatchData "1943-02-04" 1:"1943" 2:"02" 3:"04">
re.match('1943-02-04').size # => 4

The first capture is the entire matched string; the other captures are the matched substrings from the groups.

A group may have a quantifier:

re = /July 4(th)?/
re.match('July 4')   # => #<MatchData "July 4" 1:nil>
re.match('July 4th') # => #<MatchData "July 4th" 1:"th">

re = /(foo)*/
re.match('')       # => #<MatchData "" 1:nil>
re.match('foo')    # => #<MatchData "foo" 1:"foo">
re.match('foofoo') # => #<MatchData "foofoo" 1:"foo">

re = /(foo)+/
re.match('')       # => nil
re.match('foo')    # => #<MatchData "foo" 1:"foo">
re.match('foofoo') # => #<MatchData "foofoo" 1:"foo">

The returned MatchData object gives access to the matched substrings:

re = /(\d\d\d\d)-(\d\d)-(\d\d)/
md = re.match('1943-02-04')
# => #<MatchData "1943-02-04" 1:"1943" 2:"02" 3:"04">
md[0] # => "1943-02-04"
md[1] # => "1943"
md[2] # => "02"
md[3] # => "04"

Non-Capturing Groups

A group may be made non-capturing; it is still a group (and, for example, can have a quantifier), but its matching substring is not included among the captures.

A non-capturing group begins with ?: (inside the parentheses):

# Don't capture the year.
re = /(?:\d\d\d\d)-(\d\d)-(\d\d)/
md = re.match('1943-02-04') # => #<MatchData "1943-02-04" 1:"02" 2:"04">

Backreferences

A group match may also be referenced within the regexp itself; such a reference is called a backreference:

/[csh](..) [csh]\1 in/.match('The cat sat in the hat')
# => #<MatchData "cat sat in" 1:"at">

This table shows how each subexpression in the regexp above matches a substring in the target string:

| Subexpression in Regexp   | Matching Substring in Target String |
|---------------------------|-------------------------------------|
|       First '[csh]'       |            Character 'c'            |
|          '(..)'           |        First substring 'at'         |
|      First space ' '      |      First space character ' '      |
|       Second '[csh]'      |            Character 's'            |
| '\1' (backreference 'at') |        Second substring 'at'        |
|           ' in'           |            Substring ' in'          |

A regexp may contain any number of groups:

Named Captures

As seen above, a capture can be referred to by its number. A capture can also have a name, prefixed as ?<name> or ?'name', and the name (symbolized) may be used as an index in MatchData[]:

md = /\$(?<dollars>\d)\.(?'cents'\d)/.match("$3.67")
# => #<MatchData "$3.67" dollars:"3" cents:"67">
md[:dollars]  # => "3"
md[:cents]    # => "67"
# The capture numbers are still valid.
md[2]         # => "67"

When a regexp contains a named capture, there are no unnamed captures:

/\$(?<dollars>\d)\.(\d)/.match("$3.67")
# => #<MatchData "$3.67" dollars:"3">

A named group may be backreferenced as \k<name>:

/(?<vowel>[aeiou]).\k<vowel>.\k<vowel>/.match('ototomy')
# => #<MatchData "ototo" vowel:"o">

When (and only when) a regexp contains named capture groups and appears before the =~ operator, the captured substrings are assigned to local variables with corresponding names:

/\$(?<dollars>\d)\.(?<cents>\d)/ =~ '$3.67'
dollars # => "3"
cents   # => "67"

Method Regexp#named_captures returns a hash of the capture names and substrings; method Regexp#names returns an array of the capture names.

Atomic Grouping

A group may be made atomic with (?>subexpression).

This causes the subexpression to be matched independently of the rest of the expression, so that the matched substring becomes fixed for the remainder of the match, unless the entire subexpression must be abandoned and subsequently revisited.

In this way subexpression is treated as a non-divisible whole. Atomic grouping is typically used to optimise patterns to prevent needless backtracking .

Example (without atomic grouping):

/".*"/.match('"Quote"') # => #<MatchData "\"Quote\"">

Analysis:

  1. The leading subexpression " in the pattern matches the first character " in the target string.

  2. The next subexpression .* matches the next substring Quote“ (including the trailing double-quote).

  3. Now there is nothing left in the target string to match the trailing subexpression " in the pattern; this would cause the overall match to fail.

  4. The matched substring is backtracked by one position: Quote.

  5. The final subexpression " now matches the final substring ", and the overall match succeeds.

If subexpression .* is grouped atomically, the backtracking is disabled, and the overall match fails:

/"(?>.*)"/.match('"Quote"') # => nil

Atomic grouping can affect performance; see Atomic Group.

Subexpression Calls

As seen above, a backreference number (\n) or name (\k<name>) gives access to a captured substring; the corresponding regexp subexpression may also be accessed, via the number (\gn) or name (\g<name>):

/\A(?<paren>\(\g<paren>*\))*\z/.match('(())')
# ^1
#      ^2
#           ^3
#                 ^4
#      ^5
#           ^6
#                      ^7
#                       ^8
#                       ^9
#                           ^10

The pattern:

  1. Matches at the beginning of the string, i.e. before the first character.

  2. Enters a named group paren.

  3. Matches the first character in the string, '('.

  4. Calls the paren group again, i.e. recurses back to the second step.

  5. Re-enters the paren group.

  6. Matches the second character in the string, '('.

  7. Attempts to call paren a third time, but fails because doing so would prevent an overall successful match.

  8. Matches the third character in the string, ')'; marks the end of the second recursive call

  9. Matches the fourth character in the string, ')'.

  10. Matches the end of the string.

See Subexpression calls.

Conditionals

The conditional construct takes the form (?(cond)yes|no), where:

Examples:

re = /\A(foo)?(?(1)(T)|(F))\z/
re.match('fooT') # => #<MatchData "fooT" 1:"foo" 2:"T" 3:nil>
re.match('F')    # => #<MatchData "F" 1:nil 2:nil 3:"F">
re.match('fooF') # => nil
re.match('T')    # => nil

re = /\A(?<xyzzy>foo)?(?(<xyzzy>)(T)|(F))\z/
re.match('fooT') # => #<MatchData "fooT" xyzzy:"foo">
re.match('F')    # => #<MatchData "F" xyzzy:nil>
re.match('fooF') # => nil
re.match('T')    # => nil

Absence Operator

The absence operator is a special group that matches anything which does not match the contained subexpressions.

/(?~real)/.match('surrealist') # => #<MatchData "surrea">
/(?~real)ist/.match('surrealist') # => #<MatchData "ealist">
/sur(?~real)ist/.match('surrealist') # => nil

Unicode

Unicode Properties

The /\p{property_name}/ construct (with lowercase p) matches characters using a Unicode property name, much like a character class; property Alpha specifies alphabetic characters:

/\p{Alpha}/.match('a') # => #<MatchData "a">
/\p{Alpha}/.match('1') # => nil

A property can be inverted by prefixing the name with a caret character (^):

/\p{^Alpha}/.match('1') # => #<MatchData "1">
/\p{^Alpha}/.match('a') # => nil

Or by using \P (uppercase P):

/\P{Alpha}/.match('1') # => #<MatchData "1">
/\P{Alpha}/.match('a') # => nil

See Unicode Properties for regexps based on the numerous properties.

Some commonly-used properties correspond to POSIX bracket expressions:

These are also commonly used:

Unicode Character Categories

A Unicode character category name:

Examples:

/\p{lu}/                # => /\p{lu}/
/\p{LU}/                # => /\p{LU}/
/\p{Uppercase Letter}/  # => /\p{Uppercase Letter}/
/\p{Uppercase_Letter}/  # => /\p{Uppercase_Letter}/
/\p{UPPERCASE-LETTER}/  # => /\p{UPPERCASE-LETTER}/

Below are the Unicode character category abbreviations and names. Enumerations of characters in each category are at the links.

Letters:

Marks:

Numbers:

Punctation:

Unicode Scripts and Blocks

Among the Unicode properties are:

POSIX Bracket Expressions

A POSIX bracket expression is also similar to a character class. These expressions provide a portable alternative to the above, with the added benefit of encompassing non-ASCII characters:

The POSIX bracket expressions:

Ruby also supports these (non-POSIX) bracket expressions:

Comments

A comment may be included in a regexp pattern using the (?#comment) construct, where comment is a substring that is to be ignored. arbitrary text ignored by the regexp engine:

/foo(?#Ignore me)bar/.match('foobar') # => #<MatchData "foobar">

The comment may not include an unescaped terminator character.

See also Extended Mode.

Modes

Each of these modifiers sets a mode for the regexp:

Any, all, or none of these may be applied.

Modifiers i, m, and x may be applied to subexpressions:

Example:

re = /(?i)te(?-i)st/
re.match('test') # => #<MatchData "test">
re.match('TEst') # => #<MatchData "TEst">
re.match('TEST') # => nil
re.match('teST') # => nil

re = /t(?i:e)st/
re.match('test') # => #<MatchData "test">
re.match('tEst') # => #<MatchData "tEst">
re.match('tEST') # => nil

Method Regexp#options returns an integer whose value showing the settings for case-insensitivity mode, multiline mode, and extended mode.

Case-Insensitive Mode

By default, a regexp is case-sensitive:

/foo/.match('FOO')  # => nil

Modifier i enables case-insensitive mode:

/foo/i.match('FOO')
# => #<MatchData "FOO">

Method Regexp#casefold? returns whether the mode is case-insensitive.

Multiline Mode

The multiline-mode in Ruby is what is commonly called a “dot-all mode”:

Unlike other languages, the modifier m does not affect the anchors ^ and $. These anchors always match at line-boundaries in Ruby.

Extended Mode

Modifier x enables extended mode, which means that:

In extended mode, whitespace and comments may be used to form a self-documented regexp.

Regexp not in extended mode (matches some Roman numerals):

pattern = '^M{0,3}(CM|CD|D?C{0,3})(XC|XL|L?X{0,3})(IX|IV|V?I{0,3})$'
re = /#{pattern}/
re.match('MCMXLIII') # => #<MatchData "MCMXLIII" 1:"CM" 2:"XL" 3:"III">

Regexp in extended mode:

pattern = <<-EOT
  ^                   # beginning of string
  M{0,3}              # thousands - 0 to 3 Ms
  (CM|CD|D?C{0,3})    # hundreds - 900 (CM), 400 (CD), 0-300 (0 to 3 Cs),
                      #            or 500-800 (D, followed by 0 to 3 Cs)
  (XC|XL|L?X{0,3})    # tens - 90 (XC), 40 (XL), 0-30 (0 to 3 Xs),
                      #        or 50-80 (L, followed by 0 to 3 Xs)
  (IX|IV|V?I{0,3})    # ones - 9 (IX), 4 (IV), 0-3 (0 to 3 Is),
                      #        or 5-8 (V, followed by 0 to 3 Is)
  $                   # end of string
EOT
re = /#{pattern}/x
re.match('MCMXLIII') # => #<MatchData "MCMXLIII" 1:"CM" 2:"XL" 3:"III">

Interpolation Mode

Modifier o means that the first time a literal regexp with interpolations is encountered, the generated Regexp object is saved and used for all future evaluations of that literal regexp. Without modifier o, the generated Regexp is not saved, so each evaluation of the literal regexp generates a new Regexp object.

Without modifier o:

def letters; sleep 5; /[A-Z][a-z]/; end
words = %w[abc def xyz]
start = Time.now
words.each {|word| word.match(/\A[#{letters}]+\z/) }
Time.now - start # => 15.0174892

With modifier o:

start = Time.now
words.each {|word| word.match(/\A[#{letters}]+\z/o) }
Time.now - start # => 5.0010866

Note that if the literal regexp does not have interpolations, the o behavior is the default.

Encodings

By default, a regexp with only US-ASCII characters has US-ASCII encoding:

re = /foo/
re.source.encoding # => #<Encoding:US-ASCII>
re.encoding        # => #<Encoding:US-ASCII>

A regular expression containing non-US-ASCII characters is assumed to use the source encoding. This can be overridden with one of the following modifiers.

A regexp can be matched against a target string when either:

If a match between incompatible encodings is attempted an Encoding::CompatibilityError exception is raised.

Example:

re = eval("# encoding: ISO-8859-1\n/foo\\xff?/")
re.encoding                 # => #<Encoding:ISO-8859-1>
re =~ "foo".encode("UTF-8") # => 0
re =~ "foo\u0100"           # Raises Encoding::CompatibilityError

The encoding may be explicitly fixed by including Regexp::FIXEDENCODING in the second argument for Regexp.new:

# Regexp with encoding ISO-8859-1.
re = Regexp.new("a".force_encoding('iso-8859-1'), Regexp::FIXEDENCODING)
re.encoding  # => #<Encoding:ISO-8859-1>
# Target string with encoding UTF-8.
s = "a\u3042"
s.encoding   # => #<Encoding:UTF-8>
re.match(s)  # Raises Encoding::CompatibilityError.

Timeouts

When either a regexp source or a target string comes from untrusted input, malicious values could become a denial-of-service attack; to prevent such an attack, it is wise to set a timeout.

Regexp has two timeout values:

When regexp.timeout is nil, the timeout “falls through” to Regexp.timeout; when regexp.timeout is non-nil, that value controls timing out:

| regexp.timeout Value | Regexp.timeout Value |            Result           |
|----------------------|----------------------|-----------------------------|
|         nil          |          nil         |       Never times out.      |
|         nil          |         Float        | Times out in Float seconds. |
|        Float         |          Any         | Times out in Float seconds. |

Optimization

For certain values of the pattern and target string, matching time can grow polynomially or exponentially in relation to the input size; the potential vulnerability arising from this is the regular expression denial-of-service (ReDoS) attack.

Regexp matching can apply an optimization to prevent ReDoS attacks. When the optimization is applied, matching time increases linearly (not polynomially or exponentially) in relation to the input size, and a ReDoS attach is not possible.

This optimization is applied if the pattern meets these criteria:

You can use method Regexp.linear_time? to determine whether a pattern meets these criteria:

Regexp.linear_time?(/a*/)     # => true
Regexp.linear_time?('a*')     # => true
Regexp.linear_time?(/(a*)\1/) # => false

However, an untrusted source may not be safe even if the method returns true, because the optimization uses memoization (which may invoke large memory consumption).

References

Read (online PDF books):

Explore, test (interactive online editor):