Regular expression

module 2 week 6 tidyverse R programming strings and regex

Introduction to working with character strings and regular expressions inR

Stephanie Hicks https://stephaniehicks.com/ (Department of Biostatistics, Johns Hopkins)https://www.jhsph.edu
10-07-2021

Pre-lecture materials

Read ahead

Before class, you can prepare by reading the following materials:

  1. https://r4ds.had.co.nz/strings.html

Acknowledgements

Material for this lecture was borrowed and adopted from

Learning objectives

At the end of this lesson you will:

Introduction

regex basics

Now that we have covered the basics of string manipulation in R, let’s discuss the more advanced topic of regular expressions.

A regular expression (also known as a “regex” or “regexp”) is a concise language for describing patterns in character strings. For example, a regex are patterns that could be contained within another string. A regular expression can be used for e.g. 

If you have never worked with regular expressions, it can seem like maybe a baby hit the keys on your keyboard (complete gibberish), but it will slowly make sense once you learn the syntax. Soon you will be able create incredibly powerful regular expressions in your day-to-day work.

string basics

In R, you can create (character) strings with either single quotes ('hello!') or double quotes ("hello!") – no difference (not true for other languages).

I recommend using the double quotes, unless you want to create a string with multiple ".

string1 <- "This is a string"
string2 <- 'If I want to include a "quote" inside a string, I use single quotes'

Pro tip: strings can be tricky when executing them. If you forget to close a quote, you’ll see +

> "This is a string without a closing quote
+ 
+ 
+ HELP I'M STUCK

If this happen to you, press Escape and try again

Multiple strings are often stored in a character vector, which you can create with c():

c("one", "two", "three")
[1] "one"   "two"   "three"

grepl()

One of the most basic functions in R that uses regular expressions is the grepl(pattern, x) function, which takes two arguments: a regular expression (pattern) and a string to be searched (x).

It literally translates to “grep logical”.

If the string (x) contains the specified regular expression (pattern), then grepl() will return TRUE, otherwise it will return FALSE.

Let’s take a look at one example:

regular_expression <- "a"
string_to_search <- "Maryland"

grepl(regular_expression, string_to_search)
[1] TRUE

In the example above, we specify the regular expression "a" and store it in a variable called regular_expression.

Remember that regular expressions are just strings!

We also store the string "Maryland" in a variable called string_to_search. The regular expression "a" represents a single occurrence of the character "a". Since "a" is contained within "Maryland", grepl() returns the value TRUE.

Let’s try another simple example:

regular_expression <- "u"
string_to_search <- "Maryland"

grepl(regular_expression, string_to_search)
[1] FALSE

The regular expression "u" represents a single occurrence of the character "u", which is not a sub-string of "Maryland", therefore grepl() returns the value FALSE.

Regular expressions can be much longer than single characters. You could for example search for smaller strings inside of a larger string:

grepl("land", "Maryland")
[1] TRUE
grepl("ryla", "Maryland")
[1] TRUE
grepl("Marly", "Maryland")
[1] FALSE
grepl("dany", "Maryland")
[1] FALSE

Since "land" and "ryla" are sub-strings of "Maryland", grepl() returns TRUE, however when a regular expression like "Marly" or "dany" is searched grepl() returns FALSE because neither are sub-strings of "Maryland".

There is a dataset that comes with R called state.name which is a vector of strings, one for each state in the United States of America.

We are going to use this vector in several of the following examples.

head(state.name)
[1] "Alabama"    "Alaska"     "Arizona"    "Arkansas"   "California"
[6] "Colorado"  
length(state.name)
[1] 50

Let’s build a regular expression for identifying several strings in this vector, specifically a regular expression that will match names of states that both start and end with a vowel.

The state name could start and end with any vowel, so we will not be able to match exact sub-strings like in the previous examples. Thankfully we can use metacharacters to look for vowels and other parts of strings.

metacharacters

The first metacharacter that we will discuss is ".".

The metacharacter that only consists of a period represents any character other than a new line (we will discuss new lines soon).

Let’s take a look at some examples using the period regex:

grepl(".", "Maryland")
[1] TRUE
grepl(".", "*&2[0+,%<@#~|}")
[1] TRUE
grepl(".", "")
[1] FALSE

As you can see the period metacharacter is very liberal. This metacharacter is most useful when you do not care about a set of characters in a regular expression.

For example:

grepl("a.b", c("aaa", "aab", "abb", "acadb"))
[1] FALSE  TRUE  TRUE  TRUE

In the case above, grepl() returns TRUE for all strings that contain an a followed by any other character followed by a b.

repetition

You can specify a regular expression that contains a certain number of characters or metacharacters using the enumeration metacharacters (or sometimes called quantifiers).

Let’s take a look at some examples using these metacharacters:

# Does "Maryland" contain one or more of "a" ?
grepl("a+", "Maryland")
[1] TRUE
# Does "Maryland" contain one or more of "x" ?
grepl("x+", "Maryland")
[1] FALSE
# Does "Maryland" contain zero or more of "x" ?
grepl("x*", "Maryland")
[1] TRUE

You can also specify exact numbers of expressions using curly brackets {}.

For example "a{5}" specifies “a exactly five times”, "a{2,5}" specifies “a between 2 and 5 times,” and "a{2,}" specifies “a at least 2 times.” Let’s take a look at some examples:

# Does "Mississippi" contain exactly 2 adjacent "s" ?
grepl("s{2}", "Mississippi")
[1] TRUE
# This is equivalent to the expression above:
grepl("ss", "Mississippi")
[1] TRUE
# Does "Mississippi" contain between 1 and 3 adjacent "s" ?
grepl("s{1,3}", "Mississippi")
[1] TRUE
# Does "Mississippi" contain between 2 and 3 adjacent "i" ?
grepl("i{2,3}", "Mississippi")
[1] FALSE
# Does "Mississippi" contain between 2 adjacent "iss" ?
grepl("(iss){2}", "Mississippi")
[1] TRUE
# Does "Mississippi" contain between 2 adjacent "ss" ?
grepl("(ss){2}", "Mississippi")
[1] FALSE
# Does "Mississippi" contain the pattern of an "i" followed by 
# 2 of any character, with that pattern repeated three times adjacently?
grepl("(i.{2}){3}", "Mississippi")
[1] TRUE

capture group

In the last three examples, I used parentheses () to create a capturing group. A capturing group allows you to use quantifiers on other regular expressions.

In the last example, I first created the regex "i.{2}" which matches i followed by any two characters (“iss” or “ipp”). Then, I used a capture group to to wrap that regex, and to specify exactly three adjacent occurrences of that regex.

You can specify sets of characters (or character sets or character classes) with regular expressions, some of which come built in, but you can build your own character sets too.

character sets

First, we will discuss the built in character sets:

Each of these character sets have their own compliments:

Each specifies all of the characters not included in their corresponding character sets.

Interesting fact: Technically, you are using the a character set "\d" or "\s" (with only one black slash), but because you are using this character set in a string, you need the second \ to escape the string. So you will type "\\d" or "\\s".

So for example, to include a literal single or double quote in a string you can use \ to “escape” the string and being able to include a single or double quote:

double_quote <- "\"" # or '"'
double_quote
[1] "\""
single_quote <- '\'' # or "'"Copy
single_quote
[1] "'"

That means if you want to include a literal backslash, you will need to double it up: "\\".

In fact, putting two backslashes before any punctuation mark that is also a metacharacter indicates that you are looking for the symbol and not the metacharacter meaning. For example "\\." indicates you are trying to match a period in a string. Let’s take a look at a few examples:

grepl("\\+", "tragedy + time = humor")
[1] TRUE
grepl("\\.", "https://publichealth.jhu.edu")
[1] TRUE

Beware: the printed representation of a string is not the same as string itself, because the printed representation shows the escapes. To see the raw contents of the string, use writeLines():

x <- c("\'", "\"", "\\")
x
[1] "'"  "\"" "\\"
'
"
\

There are a handful of other special characters. The most common are “”, newline, and “, tab, but you can see the complete list by requesting help on”: ?‘“’, or ?”’".

?"'"

You will also sometimes see strings like “”, this is a way of writing non-English characters that works on all platforms:

x <- c("\\t", "\\n", "\u00b5")
x
[1] "\\t" "\\n" "µ"  
\t
\n
µ

Let’s take a look at a few examples of built in character sets: "\w", "\d", "\s".

grepl("\\w", "abcdefghijklmnopqrstuvwxyz0123456789")
[1] TRUE
grepl("\\d", "0123456789")
[1] TRUE
# "\n" is the metacharacter for a new line
# "\t" is the metacharacter for a tab
grepl("\\s", "\n\t   ")
[1] TRUE
grepl("\\d", "abcdefghijklmnopqrstuvwxyz")
[1] FALSE
grepl("\\D", "abcdefghijklmnopqrstuvwxyz")
[1] TRUE
grepl("\\w", "\n\t   ")
[1] FALSE

brackets

You can also specify specific character sets using straight brackets []. For example a character set of just the vowels would look like: "[aeiou]".

grepl("[aeiou]", "rhythms")
[1] FALSE

You can find the complement to a specific character by putting a carrot ^ after the first bracket. For example "[^aeiou]" matches all characters except the lowercase vowels.

grepl("[^aeiou]", "rhythms")
[1] TRUE

ranges

You can also specify ranges of characters using a hyphen - inside of the brackets. For example "[a-m]" matches all of the lowercase characters between a and m, while "[5-8]" matches any digit between 5 and 8 inclusive. Let’s take a look at some examples using custom character sets:

grepl("[a-m]", "xyz")
[1] FALSE
grepl("[a-m]", "ABC")
[1] FALSE
grepl("[a-mA-M]", "ABC")
[1] TRUE

beginning and end

There are also metacharacters for matching the beginning and the end of a string which are "^" and "$" respectively. Let’s take a look at a few examples:

grepl("^a", c("bab", "aab"))
[1] FALSE  TRUE
grepl("b$", c("bab", "aab"))
[1] TRUE TRUE
grepl("^[ab]*$", c("bab", "aab", "abc"))
[1]  TRUE  TRUE FALSE

OR metacharacter

The last metacharacter we will discuss is the OR metacharacter ("|"). The OR metacharacter matches either the regex on the left or the regex on the right side of this character. A few examples:

grepl("a|b", c("abc", "bcd", "cde"))
[1]  TRUE  TRUE FALSE
grepl("North|South", c("South Dakota", "North Carolina", "West Virginia"))
[1]  TRUE  TRUE FALSE

Finally, we have learned enough to create a regular expression that matches all state names that both begin and end with a vowel:

  1. We match the beginning of a string.
  2. We create a character set of just capitalized vowels.
  3. We specify one instance of that set.
  4. Then any number of characters until:
  5. A character set of just lowercase vowels.
  6. We specify one instance of that set.
  7. We match the end of a string.
start_end_vowel <- "^[AEIOU]{1}.+[aeiou]{1}$"
vowel_state_lgl <- grepl(start_end_vowel, state.name)
head(vowel_state_lgl)
[1]  TRUE  TRUE  TRUE FALSE FALSE FALSE
state.name[vowel_state_lgl]
[1] "Alabama"  "Alaska"   "Arizona"  "Idaho"    "Indiana"  "Iowa"    
[7] "Ohio"     "Oklahoma"

Below is a table of several important metacharacters:

Metacharacter Meaning
. Any Character
\w A Word
\W Not a Word
\d A Digit
\D Not a Digit
\s Whitespace
\S Not Whitespace
[xyz] A Set of Characters
[^xyz] Negation of Set
[a-z] A Range of Characters
^ Beginning of String
$ End of String
\n Newline
+ One or More of Previous
* Zero or More of Previous
? Zero or One of Previous
| Either the Previous or the Following
{5} Exactly 5 of Previous
{2, 5} Between 2 and 5 or Previous
{2, } More than 2 of Previous

Other regex in base R

So far we’ve been using grepl() to see if a regex matches a string. There are a few other built in regex functions you should be aware of.

First, we will review our workhorse of this lesson, grepl(), which stands for “grep logical.”

grepl("[Ii]", c("Hawaii", "Illinois", "Kentucky"))
[1]  TRUE  TRUE FALSE

grep()

Then, there is old fashioned grep(pattern, x), which returns the indices of the vector that match the regex:

grep("[Ii]", c("Hawaii", "Illinois", "Kentucky"))
[1] 1 2

sub()

The sub(pattern, replacement, x) function takes as arguments a regex, a “replacement,” and a vector of strings. This function will replace the first instance of that regex found in each string.

sub("[Ii]", "1", c("Hawaii", "Illinois", "Kentucky"))
[1] "Hawa1i"   "1llinois" "Kentucky"

gsub()

The gsub(pattern, replacement, x) function is nearly the same as sub() except it will replace every instance of the regex that is matched in each string.

gsub("[Ii]", "1", c("Hawaii", "Illinois", "Kentucky"))
[1] "Hawa11"   "1ll1no1s" "Kentucky"

strsplit()

The strsplit(x, split) function will split up strings (split) according to the provided regex (x) . If strsplit() is provided with a vector of strings it will return a list of string vectors.

two_s <- state.name[grep("ss", state.name)]
two_s
[1] "Massachusetts" "Mississippi"   "Missouri"      "Tennessee"    
strsplit(two_s, "ss")
[[1]]
[1] "Ma"        "achusetts"

[[2]]
[1] "Mi"   "i"    "ippi"

[[3]]
[1] "Mi"   "ouri"

[[4]]
[1] "Tenne" "ee"   

The stringr package

The stringr package, written by Hadley Wickham, is part of the Tidyverse group of R packages. This package takes a “data first” approach to functions involving regex, so usually the string is the first argument and the regex is the second argument. The majority of the function names in stringr begin with str_*().

Cheatsheet for stringr R package

Figure 1: Cheatsheet for stringr R package

[Source: https://stringr.tidyverse.org]

The str_extract(string, pattern) function returns the sub-string of a string (string) that matches the provided regular expression (pattern).

library(stringr)
state_tbl <- paste(state.name, state.area, state.abb)
head(state_tbl)
[1] "Alabama 51609 AL"     "Alaska 589757 AK"    
[3] "Arizona 113909 AZ"    "Arkansas 53104 AR"   
[5] "California 158693 CA" "Colorado 104247 CO"  
str_extract(state_tbl, "[0-9]+")
 [1] "51609"  "589757" "113909" "53104"  "158693" "104247" "5009"  
 [8] "2057"   "58560"  "58876"  "6450"   "83557"  "56400"  "36291" 
[15] "56290"  "82264"  "40395"  "48523"  "33215"  "10577"  "8257"  
[22] "58216"  "84068"  "47716"  "69686"  "147138" "77227"  "110540"
[29] "9304"   "7836"   "121666" "49576"  "52586"  "70665"  "41222" 
[36] "69919"  "96981"  "45333"  "1214"   "31055"  "77047"  "42244" 
[43] "267339" "84916"  "9609"   "40815"  "68192"  "24181"  "56154" 
[50] "97914" 

The str_order(x) function returns a numeric vector that corresponds to the alphabetical order of the strings in the provided vector (x).

head(state.name)
[1] "Alabama"    "Alaska"     "Arizona"    "Arkansas"   "California"
[6] "Colorado"  
str_order(state.name)
 [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22
[23] 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
[45] 45 46 47 48 49 50
head(state.abb)
[1] "AL" "AK" "AZ" "AR" "CA" "CO"
str_order(state.abb)
 [1]  2  1  4  3  5  6  7  8  9 10 11 15 12 13 14 16 17 18 21 20 19 22
[23] 23 25 24 26 33 34 27 29 30 31 28 32 35 36 37 38 39 40 41 42 43 44
[45] 46 45 47 49 48 50

The str_pad(string, width, side, pad) function pads strings (string) with other characters, which is often useful when the string is going to be eventually printed for a person to read.

str_pad("Thai", width = 8, side = "left", pad = "-")
[1] "----Thai"
str_pad("Thai", width = 8, side = "right", pad = "-")
[1] "Thai----"
str_pad("Thai", width = 8, side = "both", pad = "-")
[1] "--Thai--"

The str_to_title(string) function acts just like tolower() and toupper() except it puts strings into Title Case.

cases <- c("CAPS", "low", "Title")
str_to_title(cases)
[1] "Caps"  "Low"   "Title"

The str_trim(string) function deletes white space from both sides of a string.

to_trim <- c("   space", "the    ", "    final frontier  ")
str_trim(to_trim)
[1] "space"          "the"            "final frontier"

The str_wrap(string) function inserts newlines in strings so that when the string is printed each line’s length is limited.

pasted_states <- paste(state.name[1:20], collapse = " ")

cat(str_wrap(pasted_states, width = 80))
Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida
Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine
Maryland
cat(str_wrap(pasted_states, width = 30))
Alabama Alaska Arizona
Arkansas California Colorado
Connecticut Delaware Florida
Georgia Hawaii Idaho Illinois
Indiana Iowa Kansas Kentucky
Louisiana Maine Maryland

The word() function allows you to index each word in a string as if it were a vector.

a_tale <- "It was the best of times it was the worst of times it was the age of wisdom it was the age of foolishness"

word(a_tale, 2)
[1] "was"
word(a_tale, end = 3)
[1] "It was the"
word(a_tale, start = 11, end = 15)
[1] "of times it was the"

Post-lecture materials

Final Questions

Here are some post-lecture questions to help you think about the material discussed.

Questions:

There is a corpus of common words here:

head(stringr::words)
[1] "a"        "able"     "about"    "absolute" "accept"   "account" 
length(stringr::words)
[1] 980
  1. Using stringr::words, create regular expressions that find all words that:
  1. Using the same stringr::words, create regular expressions to find all words that:

Additional Resources

Reuse

Text and figures are licensed under Creative Commons Attribution CC BY-NC-SA 4.0. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".

Citation

For attribution, please cite this work as

Hicks (2021, Oct. 7). Statistical Computing: Regular expression. Retrieved from https://stephaniehicks.com/jhustatcomputing2021/posts/2021-10-07-regular-expressions/

BibTeX citation

@misc{hicks2021regular,
  author = {Hicks, Stephanie},
  title = {Statistical Computing: Regular expression},
  url = {https://stephaniehicks.com/jhustatcomputing2021/posts/2021-10-07-regular-expressions/},
  year = {2021}
}