Regular expressions are a concise and flexible tool for describing patterns in strings. This vignette describes the key features of stringr’s regular expressions, as implemented by stringi. It is not a tutorial, so if you’re unfamiliar regular expressions, I’d recommend starting at https://r4ds.had.co.nz/strings.html. If you want to master the details, I’d recommend reading the classic Mastering Regular Expressions by Jeffrey E. F. Friedl.
Regular expressions are the default pattern engine in stringr. That
means when you use a pattern matching function with a bare string, it’s
equivalent to wrapping it in a call to regex()
:
# The regular call:
str_extract(fruit, "nana")
# Is shorthand for
str_extract(fruit, regex("nana"))
You will need to use regex()
explicitly if you want to
override the default options, as you’ll see in examples below.
The simplest patterns match exact strings:
<- c("apple", "banana", "pear")
x str_extract(x, "an")
#> [1] NA "an" NA
You can perform a case-insensitive match using
ignore_case = TRUE
:
<- c("banana", "Banana", "BANANA")
bananas str_detect(bananas, "banana")
#> [1] TRUE FALSE FALSE
str_detect(bananas, regex("banana", ignore_case = TRUE))
#> [1] TRUE TRUE TRUE
The next step up in complexity is .
, which matches any
character except a newline:
str_extract(x, ".a.")
#> [1] NA "ban" "ear"
You can allow .
to match everything, including
\n
, by setting dotall = TRUE
:
str_detect("\nX\n", ".X.")
#> [1] FALSE
str_detect("\nX\n", regex(".X.", dotall = TRUE))
#> [1] TRUE
If “.
” matches any character, how do you match a literal
“.
”? You need to use an “escape” to tell the regular
expression you want to match it exactly, not use its special behaviour.
Like strings, regexps use the backslash, \
, to escape
special behaviour. So to match an .
, you need the regexp
\.
. Unfortunately this creates a problem. We use strings to
represent regular expressions, and \
is also used as an
escape symbol in strings. So to create the regular expression
\.
we need the string "\\."
.
# To create the regular expression, we need \\
<- "\\."
dot
# But the expression itself only contains one:
writeLines(dot)
#> \.
# And this tells R to look for an explicit .
str_extract(c("abc", "a.c", "bef"), "a\\.c")
#> [1] NA "a.c" NA
If \
is used as an escape character in regular
expressions, how do you match a literal \
? Well you need to
escape it, creating the regular expression \\
. To create
that regular expression, you need to use a string, which also needs to
escape \
. That means to match a literal \
you
need to write "\\\\"
— you need four backslashes to match
one!
<- "a\\b"
x writeLines(x)
#> a\b
str_extract(x, "\\\\")
#> [1] "\\"
In this vignette, I use \.
to denote the regular
expression, and "\\."
to denote the string that represents
the regular expression.
An alternative quoting mechanism is \Q...\E
: all the
characters in ...
are treated as exact matches. This is
useful if you want to exactly match user input as part of a regular
expression.
<- c("a.b.c.d", "aeb")
x <- "a.b"
starts_with
str_detect(x, paste0("^", starts_with))
#> [1] TRUE TRUE
str_detect(x, paste0("^\\Q", starts_with, "\\E"))
#> [1] TRUE FALSE
Escapes also allow you to specify individual characters that are otherwise hard to type. You can specify individual unicode characters in five ways, either as a variable number of hex digits (four is most common), or by name:
\xhh
: 2 hex digits.
\x{hhhh}
: 1-6 hex digits.
\uhhhh
: 4 hex digits.
\Uhhhhhhhh
: 8 hex digits.
\N{name}
, e.g. \N{grinning face}
matches the basic smiling emoji.
Similarly, you can specify many common control characters:
\a
: bell.
\cX
: match a control-X character.
\e
: escape (\u001B
).
\f
: form feed (\u000C
).
\n
: line feed (\u000A
).
\r
: carriage return (\u000D
).
\t
: horizontal tabulation
(\u0009
).
\0ooo
match an octal character. ‘ooo’ is from one to
three octal digits, from 000 to 0377. The leading zero is
required.
(Many of these are only of historical interest and are only included here for the sake of completeness.)
There are a number of patterns that match more than one character.
You’ve already seen .
, which matches any character (except
a newline). A closely related operator is \X
, which matches
a grapheme cluster, a set of individual elements that
form a single symbol. For example, one way of representing “á” is as the
letter “a” plus an accent: .
will match the component “a”,
while \X
will match the complete symbol:
<- "a\u0301"
x str_extract(x, ".")
#> [1] "a"
str_extract(x, "\\X")
#> [1] "á"
There are five other escaped pairs that match narrower classes of characters:
\d
: matches any digit. The complement,
\D
, matches any character that is not a decimal digit.
str_extract_all("1 + 2 = 3", "\\d+")[[1]]
#> [1] "1" "2" "3"
Technically, \d
includes any character in the Unicode
Category of Nd (“Number, Decimal Digit”), which also includes numeric
symbols from other languages:
# Some Laotian numbers
str_detect("១២៣", "\\d")
#> [1] TRUE
\s
: matches any whitespace. This includes tabs,
newlines, form feeds, and any character in the Unicode Z Category (which
includes a variety of space characters and other separators.). The
complement, \S
, matches any non-whitespace character.
<- "Some \t badly\n\t\tspaced \f text")
(text #> [1] "Some \t badly\n\t\tspaced \f text"
str_replace_all(text, "\\s+", " ")
#> [1] "Some badly spaced text"
\p{property name}
matches any character with
specific unicode property, like \p{Uppercase}
or
\p{Diacritic}
. The complement,
\P{property name}
, matches all characters without the
property. A complete list of unicode properties can be found at http://www.unicode.org/reports/tr44/#Property_Index.
<- c('"Double quotes"', "«Guillemet»", "“Fancy quotes”"))
(text #> [1] "\"Double quotes\"" "«Guillemet»" "“Fancy quotes”"
str_replace_all(text, "\\p{quotation mark}", "'")
#> [1] "'Double quotes'" "'Guillemet'" "'Fancy quotes'"
\w
matches any “word” character, which includes
alphabetic characters, marks and decimal numbers. The complement,
\W
, matches any non-word character.
str_extract_all("Don't eat that!", "\\w+")[[1]]
#> [1] "Don" "t" "eat" "that"
str_split("Don't eat that!", "\\W")[[1]]
#> [1] "Don" "t" "eat" "that" ""
Technically, \w
also matches connector punctuation,
\u200c
(zero width connector), and \u200d
(zero width joiner), but these are rarely seen in the wild.
\b
matches word boundaries, the transition between
word and non-word characters. \B
matches the opposite:
boundaries that have either both word or non-word characters on either
side.
str_replace_all("The quick brown fox", "\\b", "_")
#> [1] "_The_ _quick_ _brown_ _fox_"
str_replace_all("The quick brown fox", "\\B", "_")
#> [1] "T_h_e q_u_i_c_k b_r_o_w_n f_o_x"
You can also create your own character classes using
[]
:
[abc]
: matches a, b, or c.[a-z]
: matches every character between a and z (in
Unicode code point order).[^abc]
: matches anything except a, b, or c.[\^\-]
: matches ^
or -
.There are a number of pre-built classes that you can use inside
[]
:
[:punct:]
: punctuation.[:alpha:]
: letters.[:lower:]
: lowercase letters.[:upper:]
: upperclass letters.[:digit:]
: digits.[:xdigit:]
: hex digits.[:alnum:]
: letters and numbers.[:cntrl:]
: control characters.[:graph:]
: letters, numbers, and punctuation.[:print:]
: letters, numbers, punctuation, and
whitespace.[:space:]
: space characters (basically equivalent to
\s
).[:blank:]
: space and tab.These all go inside the []
for character classes,
i.e. [[:digit:]AX]
matches all digits, A, and X.
You can also using Unicode properties, like
[\p{Letter}]
, and various set operations, like
[\p{Letter}--\p{script=latin}]
. See
?"stringi-search-charclass"
for details.
|
is the alternation operator, which
will pick between one or more possible matches. For example,
abc|def
will match abc
or
def
.
str_detect(c("abc", "def", "ghi"), "abc|def")
#> [1] TRUE TRUE FALSE
Note that the precedence for |
is low, so that
abc|def
matches abc
or def
not
abcyz
or abxyz
.
You can use parentheses to override the default precedence rules:
str_extract(c("grey", "gray"), "gre|ay")
#> [1] "gre" "ay"
str_extract(c("grey", "gray"), "gr(e|a)y")
#> [1] "grey" "gray"
Parenthesis also define “groups” that you can refer to with
backreferences, like \1
, \2
etc, and can be extracted with str_match()
. For example,
the following regular expression finds all fruits that have a repeated
pair of letters:
<- "(..)\\1"
pattern %>%
fruit str_subset(pattern)
#> [1] "banana" "coconut" "cucumber" "jujube" "papaya"
#> [6] "salal berry"
%>%
fruit str_subset(pattern) %>%
str_match(pattern)
#> [,1] [,2]
#> [1,] "anan" "an"
#> [2,] "coco" "co"
#> [3,] "cucu" "cu"
#> [4,] "juju" "ju"
#> [5,] "papa" "pa"
#> [6,] "alal" "al"
You can use (?:...)
, the non-grouping parentheses, to
control precedence but not capture the match in a group. This is
slightly more efficient than capturing parentheses.
str_match(c("grey", "gray"), "gr(e|a)y")
#> [,1] [,2]
#> [1,] "grey" "e"
#> [2,] "gray" "a"
str_match(c("grey", "gray"), "gr(?:e|a)y")
#> [,1]
#> [1,] "grey"
#> [2,] "gray"
This is most useful for more complex cases where you need to capture matches and control precedence independently.
By default, regular expressions will match any part of a string. It’s often useful to anchor the regular expression so that it matches from the start or end of the string:
^
matches the start of string.$
matches the end of the string.<- c("apple", "banana", "pear")
x str_extract(x, "^a")
#> [1] "a" NA NA
str_extract(x, "a$")
#> [1] NA "a" NA
To match a literal “$” or “^”, you need to escape them,
\$
, and \^
.
For multiline strings, you can use
regex(multiline = TRUE)
. This changes the behaviour of
^
and $
, and introduces three new
operators:
^
now matches the start of each line.
$
now matches the end of each line.
\A
matches the start of the input.
\z
matches the end of the input.
\Z
matches the end of the input, but before the
final line terminator, if it exists.
<- "Line 1\nLine 2\nLine 3\n"
x str_extract_all(x, "^Line..")[[1]]
#> [1] "Line 1"
str_extract_all(x, regex("^Line..", multiline = TRUE))[[1]]
#> [1] "Line 1" "Line 2" "Line 3"
str_extract_all(x, regex("\\ALine..", multiline = TRUE))[[1]]
#> [1] "Line 1"
You can control how many times a pattern matches with the repetition operators:
?
: 0 or 1.+
: 1 or more.*
: 0 or more.<- "1888 is the longest year in Roman numerals: MDCCCLXXXVIII"
x str_extract(x, "CC?")
#> [1] "CC"
str_extract(x, "CC+")
#> [1] "CCC"
str_extract(x, 'C[LX]+')
#> [1] "CLXXX"
Note that the precedence of these operators is high, so you can
write: colou?r
to match either American or British
spellings. That means most uses will need parentheses, like
bana(na)+
.
You can also specify the number of matches precisely:
{n}
: exactly n{n,}
: n or more{n,m}
: between n and mstr_extract(x, "C{2}")
#> [1] "CC"
str_extract(x, "C{2,}")
#> [1] "CCC"
str_extract(x, "C{2,3}")
#> [1] "CCC"
By default these matches are “greedy”: they will match the longest
string possible. You can make them “lazy”, matching the shortest string
possible by putting a ?
after them:
??
: 0 or 1, prefer 0.+?
: 1 or more, match as few times as possible.*?
: 0 or more, match as few times as possible.{n,}?
: n or more, match as few times as possible.{n,m}?
: between n and m, , match as few times as
possible, but at least n.str_extract(x, c("C{2,3}", "C{2,3}?"))
#> [1] "CCC" "CC"
str_extract(x, c("C[LX]+", "C[LX]+?"))
#> [1] "CLXXX" "CL"
You can also make the matches possessive by putting a +
after them, which means that if later parts of the match fail, the
repetition will not be re-tried with a smaller number of characters.
This is an advanced feature used to improve performance in worst-case
scenarios (called “catastrophic backtracking”).
?+
: 0 or 1, possessive.++
: 1 or more, possessive.*+
: 0 or more, possessive.{n}+
: exactly n, possessive.{n,}+
: n or more, possessive.{n,m}+
: between n and m, possessive.A related concept is the atomic-match parenthesis,
(?>...)
. If a later match fails and the engine needs to
back-track, an atomic match is kept as is: it succeeds or fails as a
whole. Compare the following two regular expressions:
str_detect("ABC", "(?>A|.B)C")
#> [1] FALSE
str_detect("ABC", "(?:A|.B)C")
#> [1] TRUE
The atomic match fails because it matches A, and then the next character is a C so it fails. The regular match succeeds because it matches A, but then C doesn’t match, so it back-tracks and tries B instead.
These assertions look ahead or behind the current match without “consuming” any characters (i.e. changing the input position).
(?=...)
: positive look-ahead assertion. Matches if
...
matches at the current input.
(?!...)
: negative look-ahead assertion. Matches if
...
does not match at the current
input.
(?<=...)
: positive look-behind assertion. Matches
if ...
matches text preceding the current position, with
the last character of the match being the character just before the
current position. Length must be bounded
(i.e. no *
or +
).
(?<!...)
: negative look-behind assertion. Matches
if ...
does not match text preceding the
current position. Length must be bounded
(i.e. no *
or +
).
These are useful when you want to check that a pattern exists, but you don’t want to include it in the result:
<- c("1 piece", "2 pieces", "3")
x str_extract(x, "\\d+(?= pieces?)")
#> [1] "1" "2" NA
<- c("100", "$400")
y str_extract(y, "(?<=\\$)\\d+")
#> [1] NA "400"
Comments
There are two ways to include comments in a regular expression. The first is with
(?#...)
:The second is to use
regex(comments = TRUE)
. This form ignores spaces and newlines, and anything everything after#
. To match a literal space, you’ll need to escape it:"\\ "
. This is a useful way of describing complex regular expressions: