Regular expressions

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 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.

Basic matches

The simplest patterns match exact strings:

x <- c("apple", "banana", "pear")
str_extract(x, "an")
#> [1] NA   "an" NA

You can perform a case-insensitive match using ignore_case = TRUE:

bananas <- c("banana", "Banana", "BANANA")
str_detect(bananas, "banana")
str_detect(bananas, regex("banana", ignore_case = 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:
#> \.

# 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!

x <- "a\\b"
#> 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.

x <- c("a.b.c.d", "aeb")
starts_with <- "a.b"

str_detect(x, paste0("^", starts_with))
#> [1] TRUE TRUE
str_detect(x, paste0("^\\Q", starts_with, "\\E"))
#> [1]  TRUE FALSE

Special characters

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:

Similarly, you can specify many common control characters:

(Many of these are only of historical interest and are only included here for the sake of completeness.)

Matching multiple characters

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:

x <- "a\u0301"
str_extract(x, ".")
#> [1] "a"
str_extract(x, "\\X")
#> [1] "á"

There are five other escaped pairs that match narrower classes of characters:

You can also create your own character classes using []:

There are a number of pre-built classes that you can use inside []:

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")

Note that the precedence for | is low: abc|def is equivalent to (abc)|(def) not ab(c|d)ef.


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:

pattern <- "(..)\\1"
fruit %>% 
#> [1] "banana"

fruit %>% 
  str_subset(pattern) %>% 
#>      [,1]   [,2]
#> [1,] "anan" "an"

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:

x <- c("apple", "banana", "pear")
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:

x <- "Line 1\nLine 2\nLine 3\n"
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:

x <- "1888 is the longest year in Roman numerals: MDCCCLXXXVIII"
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:

str_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:

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”).

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.

Look arounds

These assertions look ahead or behind the current match without “consuming” any characters (i.e. changing the input position).

These are useful when you want to check that a pattern exists, but you don’t want to include it in the result:

x <- c("1 piece", "2 pieces", "3")
str_extract(x, "\\d+(?= pieces?)")
#> [1] "1" "2" NA

y <- c("100", "$400")
str_extract(y, "(?<=\\$)\\d+")
#> [1] NA    "400"


There are two ways to include comments in a regular expression. The first is with (?#...):

str_detect("xyz", "x(?#this is a comment)")
#> [1] TRUE

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:

phone <- regex("
  \\(?       # optional opening parens
  (\\d{3})   # area code
  \\)?       # optional closing parens
  (?:-|\\ )? # optional dash or space
  (\\d{3})   # another three numbers
  (?:-|\\ )? # optional dash or space
  (\\d{3})   # three more numbers
  ", comments = TRUE)

str_match(c("514-791-8141", "(514) 791 8141"), phone)
#>      [,1]            [,2]  [,3]  [,4] 
#> [1,] "514-791-814"   "514" "791" "814"
#> [2,] "(514) 791 814" "514" "791" "814"