Introduction to re2

re2: R interface to Google RE2


re2 package provides pattern matching, extraction, replacement and other string processing operations using Google’s RE2 (C++) regular-expression library. The interface is consistent, and similar to stringr.

Why re2?

Regular expression matching can be done in two ways: using recursive backtracking or using finite automata-based techniques.

Perl, PCRE, Python, Ruby, Java, and many other languages rely on recursive backtracking for their regular expression implementations. The problem with this approach is that performance can degrade very quickly. Time complexity can be exponential. In contrast, re2 uses finite automata-based techniques for regular expression matching, guaranteeing linear time execution and a fixed stack footprint. See links to Russ Cox’s excellent articles below.


# Install the released version from CRAN:

# Install the development version from GitHub:
# install.packages("devtools")


re2 provides three types of regular-expression functions:

All functions take a vector of strings as argument. Regular-expression patterns can be compiled, and reused for performance.

Here are the primary verbs of re2:

re2_detect(c("barbazbla", "foobar", "foxy brown"), "(foo)|(bar)baz")
re2_count(c("yellowgreen", "steelblue", "maroon"), "e")
#> [1] 3 3 0
re2_subset(c("yellowgreen", "steelblue", "goldenrod"), "ee")
#> [1] "yellowgreen" "steelblue"
re2_match("ruby:1234 68 red:92 blue:", "(\\w+):(\\d+)")
#>      .0          .1     .2    
#> [1,] "ruby:1234" "ruby" "1234"
# Groups can be named:

re2_match(c("barbazbla", "foobar"), "(foo)|(?P<TestGroup>bar)baz")
#>      .0       .1    TestGroup
#> [1,] "barbaz" NA    "bar"    
#> [2,] "foo"    "foo" NA
# Use pre-compiled regular expression:

re <- re2_regexp("(foo)|(bar)baz", case_sensitive = FALSE)
re2_match(c("BaRbazbla", "Foobar"), re)
#>      .0       .1    .2   
#> [1,] "BaRbaz" NA    "BaR"
#> [2,] "Foo"    "Foo" NA
re2_match_all("ruby:1234 68 red:92 blue:", "(\\w+):(\\d+)")
#> [[1]]
#>      .0          .1     .2    
#> [1,] "ruby:1234" "ruby" "1234"
#> [2,] "red:92"    "red"  "92"
re2_replace("yabba dabba doo", "b+", "d")
#> [1] "yada dabba doo"
# Use groups in rewrite:

re2_replace("", "(.*)@([^.]*)", "\\2!\\1")
#> [1] "wunnies!"
re2_replace_all("yabba dabba doo", "b+", "d")
#> [1] "yada dada doo"
re2_extract_replace("", "(.*)@([^.]*)", "\\2!\\1")
#> [1] "wunnies!bunny"
re2_split("How vexingly quick daft zebras jump!", " quick | zebras")
#> [[1]]
#> [1] "How vexingly" "daft"         " jump!"
re2_locate(c("yellowgreen", "steelblue"), "l(b)?l")
#>      begin end
#> [1,]     3   4
#> [2,]     5   7
re2_locate_all(c("yellowgreen", "steelblue"), "l")
#> [[1]]
#>      begin end
#> [1,]     3   3
#> [2,]     4   4
#> [[2]]
#>      begin end
#> [1,]     5   5
#> [2,]     7   7

In all the above functions, regular-expression pattern is vectorized.

Regular-expression pattern can be compiled using re2_regexp(pattern, ...). Here are some of the options:

help(re2_regexp) lists available options.

re2_get_options(regexp_ptr) returns a list of options stored in the compiled regular-expression object.

Regexp Syntax

re2 supports pearl style regular expressions (with extensions like \d, \w, \s, …) and provides most of the functionality of PCRE – eschewing only backreferences and look-around assertions.

See RE2 Syntax for the syntax supported by RE2, and a comparison with PCRE and PERL regexps.

For those not familiar with Perl’s regular expressions, here are some examples of the most commonly used extensions:

"hello (\\w+) world" \w matches a “word” character
"version (\\d+)" \d matches a digit
"hello\\s+world" \s matches any whitespace character
"\\b(\\w+)\\b" \b matches non-empty string at word boundary
"(?i)hello" (?i) turns on case-insensitive matching
"/\\*(.*?)\\*/" .*? matches . minimum no. of times possible

The double backslashes are needed when writing R string literals. However, they should not be used when writing raw string literals:

r"(hello (\w+) world)" \w matches a “word” character
r"(version (\d+))" \d matches a digit
r"(hello\s+world)" \s matches any whitespace character
r"(\b(\w+)\b)" \b matches non-empty string at word boundary
r"((?i)hello)" (?i) turns on case-insensitive matching
r"(/\*(.*?)\*/)" .*? matches . minimum no. of times possible