In R, you create strings of type character
using either single or double quotes. There is no difference (in R) between the two.
## [1] TRUE
## [1] "character"
This is not the case in all languages. For instance, consider the following example from bash
.
#!/bin/bash
## A short script to illustrate single vs double quotes
# Specify filename and extension
FILE=my_file
EXT=.txt
# Double quotes allow for expansion
echo "Double quotes: $FILE$EXT"
# Single quotes create a literal string
echo 'Single quotes: $FILE$EXT'
Double Quotes: my_file.txt
Single Quotes: $FILE$EXT
Returning to R, you can mix single and double quotes when you want to include one or the other within your string.
## [1] "These are sometimes called 'scare quotes'."
## [1] "Quoth the Raven, \"Nevermore.\""
## Quoth the Raven, "Nevermore."
You can also include quotes within a string by escaping them:
## [1] "Quoth the Raven, \"Nevermore.\""
## Quoth the Raven, "Nevermore."
Observe the difference between print()
and cat()
in terms of how the escaped characters are handled. Be aware also that because backslash plays this special role as an escape character, it itself needs to be escaped:
## This is a backslash '\', this is not ' '.
Similar to cat()
is the function writeLines()
used above. The latter is more syntactic when writing to a file and has the advantage of adding a line between components of a vector. Below is an example.
## Line 1
## Line 2
## Line 1 Line 2
You should also make note of readLines()
for reading the text in a file. It is helpful to realize that readLines()
and writeLines()
are inverse to one another.
The two can be used in conjunction with a template and a find-and-replace function (see below) to create a series of related .R
or other files for automating a series of related tasks.
## #!/bin/bash
## ## A short script to illustrate single vs double quotes
##
## # Specify filename and extension
## FILE=my_file
## EXT=.txt
##
## # Double quotes allow for expansion
## echo "Double quotes: $FILE$EXT"
##
## # Single quotes create a literal string
## echo 'Single quotes: $FILE$EXT'
for ( i in 1:3 ) {
new_file = sprintf('./template-%i.sh', i)
writeLines(
stringr::str_replace_all(template, 'my_file', paste(i) ),
con = new_file
)
cat('Wrote ', new_file, '.\n', sep = '')
}
## Wrote ./template-1.sh.
## Wrote ./template-2.sh.
## Wrote ./template-3.sh.
The table below collects some common string operations from base R and their parallels in the stringr
package. I say parallels and not equivalents because they do not always behave in the same way. If you are new to string manipulation in R, I suggest you use stringr
for these operations. However, you should be aware of the common base functions as you may encounter them in code written by others (including me).
operation | base | stringr |
---|---|---|
join | paste |
str_c |
subset | substr |
str_sub |
split | strsplit |
str_split |
search | grep , grepl |
str_which ,str_detect |
The functions paste
and stringr::str_c
are both used to join strings together.
Observe the difference between the sep
and collapse
arguments in paste
.
## [1] 26
## [1] "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
## [1] "1: A" "2: B" "3: C" "4: D" "5: E" "6: F" "7: G" "8: H" "9: I"
## [10] "10: J" "11: K" "12: L" "13: M" "14: N" "15: O" "16: P" "17: Q" "18: R"
## [19] "19: S" "20: T" "21: U" "22: V" "23: W" "24: X" "25: Y" "26: Z"
## [1] "1: A\n 2: B\n 3: C\n 4: D\n 5: E\n 6: F\n 7: G\n 8: H\n 9: I\n 10: J\n 11: K\n 12: L\n 13: M\n 14: N\n 15: O\n 16: P\n 17: Q\n 18: R\n 19: S\n 20: T\n 21: U\n 22: V\n 23: W\n 24: X\n 25: Y\n 26: Z"
Below we see that str_c
behaves similarly.
library(stringr) # stringr is in the tidyverse
all.equal(str_c(LETTERS, collapse = ""), paste(LETTERS, collapse = "") )
## [1] TRUE
## [1] TRUE
all.equal(str_c(1:26, LETTERS, sep = ': ', collapse = '\n '),
paste(1:26, LETTERS, sep = ': ', collapse = '\n ')
)
## [1] TRUE
However, these functions differ in the treatment of missing values (NA
).
## [1] "1:1, 2:NA, 3:3"
## [1] NA
## [1] "1:1, 2:NA, 3:3"
Recall that length
returns the length of a vector. To get the length of a string use nchar
or str_length
.
## [1] 1
## [1] 26
## [1] 26
The following functions extract sub-strings at given positions.
## [1] "rings"
## [1] "String"
The function stringr::str_sub
supports negative indexing.
## [1] "base: , stringr: rings"
The example below uses the vector fruit
from the stringr
package.
The base function grep
returns the indices of all strings within a vector that contain the requested pattern. The grepl
function behaves in the same way but returns a logical vector of the same length as the input x
.
## [1] "apple" "apricot" "avocado" "banana" "bell pepper"
## [6] "bilberry"
## [1] 12 26 35 39 42 57 75 79
## [1] 12 26 35 39 42 57 75 79
## [1] FALSE FALSE FALSE FALSE FALSE FALSE
## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
These functions are vectorized over the input but not the pattern.
## Warning in grep(c("fruit", "berry"), fruit): argument 'pattern' has length > 1
## and only the first element will be used
## [1] 12 26 35 39 42 57 75 79
## $fruit
## [1] 12 26 35 39 42 57 75 79
##
## $berry
## [1] 6 7 10 11 19 21 29 32 33 38 50 70 73 76
The match
function is vectorized over the input, but returns only the first match and requires exact matching.
## [1] NA
## [1] 1 59
The corresponding stringr
functions are vectorized over both pattern and input, but the vectorization uses broadcasting so be careful. Pay attention to the order that the string and pattern are supplied in as it is the reverse of the base functions.
ind_fruit = which(str_detect(fruit, 'fruit') )
ind_berry = which(str_detect(fruit, 'berry') )
ind_either = which(str_detect(fruit, c('fruit','berry') ) )
setdiff(union(ind_fruit, ind_berry), ind_either )
## [1] 12 26 42 7 11 19 21 29 33 73
# Below we demonstrate the broadcasting pattern
ind_odd = seq(1, length(fruit), 2)
ind_even = seq(2, length(fruit), 2)
odd_fruit = ind_odd[ str_detect(fruit[ind_odd], 'fruit') ]
even_berry = ind_even[ str_detect(fruit[ind_even], 'berry') ]
setdiff(union(odd_fruit, even_berry), ind_either )
## numeric(0)
The vectorization in this case doesn’t help us to avoid the lapply
pattern we used with grep
.
## $fruit
## [1] 12 26 35 39 42 57 75 79
##
## $berry
## [1] 6 7 10 11 19 21 29 32 33 38 50 70 73 76
However, str_locate
is vectorized using an “OR” operator.
ind_fruit = str_locate(fruit, 'fruit')
ind_berry = str_locate(fruit, 'berry')
ind_either = str_locate(fruit, c('fruit','berry'))
setdiff(union(ind_fruit, ind_berry), ind_either )
## integer(0)
For find and replace operations, you can use one of str_replace
or str_replace_all
. The former matches only the first instance of pattern. Similar base functions are sub
and gsub
.
# abc ...
letter_vec = paste(letters, collapse = '')
## replace all instances
str_replace_all(letter_vec, '[aeiou]', 'X')
## [1] "XbcdXfghXjklmnXpqrstXvwxyz"
## [1] "Xbcdefghijklmnopqrstuvwxyz"
You can also find and/or replace by the position in a string using str_sub
.
## [1] "ab" "bc" "cd"
The base function strsplit
can be used to split a string into pieces based on a pattern. The example below finds all two-word fruit names from fruit
.
fruit_list = strsplit(fruit,' ')
two_ind = which(sapply(fruit_list, length)==2)
fruit_two = lapply(fruit_list[two_ind], paste, collapse=' ')
unlist(fruit_two)
## [1] "bell pepper" "blood orange" "canary melon"
## [4] "chili pepper" "goji berry" "kiwi fruit"
## [7] "purple mangosteen" "rock melon" "salal berry"
## [10] "star fruit" "ugli fruit"
The str_split
function behaves similarly for this simple case.
## [1] TRUE
When there are multiple patterns matching the split point, the functions strsplit
and str_split
behave differently.
## [[1]]
## [1] "1" "2,3"
## [[1]]
## [1] "1" "2,3"
##
## [[2]]
## [1] "1;2" "3"
## [[1]]
## [1] "1" "2" "3"
Regular expressions (“regexp” or “regex”) are a way to describe patterns in strings, often in an abstract way. There is a common regexp vocabulary though some details differ between implementations and standards. The basic idea is illustrated in the examples below using the fruit data from the stringr
library.
## [1] "bell pepper" "blood orange" "canary melon"
## [4] "chili pepper" "goji berry" "kiwi fruit"
## [7] "purple mangosteen" "rock melon" "salal berry"
## [10] "star fruit" "ugli fruit"
## [1] "apple" "apricot" "avocado"
## [4] "banana" "blackberry" "blackcurrant"
## [7] "blood orange" "breadfruit" "canary melon"
## [10] "cantaloupe" "cherimoya" "cranberry"
## [13] "currant" "damson" "date"
## [16] "dragonfruit" "durian" "eggplant"
## [19] "feijoa" "grape" "grapefruit"
## [22] "guava" "jackfruit" "jambul"
## [25] "kumquat" "loquat" "mandarine"
## [28] "mango" "nectarine" "orange"
## [31] "pamelo" "papaya" "passionfruit"
## [34] "peach" "pear" "physalis"
## [37] "pineapple" "pomegranate" "purple mangosteen"
## [40] "raisin" "rambutan" "raspberry"
## [43] "redcurrant" "salal berry" "satsuma"
## [46] "star fruit" "strawberry" "tamarillo"
## [49] "tangerine" "watermelon"
## [1] "apple" "apricot" "avocado"
## [1] "banana" "cherimoya" "feijoa" "guava" "papaya" "satsuma"
## [1] "apple" "apricot" "avocado" "eggplant" "elderberry"
## [6] "olive" "orange" "ugli fruit"
## [1] "blood orange" "blueberry" "breadfruit"
## [4] "cantaloupe" "cloudberry" "dragonfruit"
## [7] "durian" "feijoa" "gooseberry"
## [10] "grapefruit" "guava" "jackfruit"
## [13] "kiwi fruit" "kumquat" "loquat"
## [16] "lychee" "passionfruit" "peach"
## [19] "pear" "pineapple" "purple mangosteen"
## [22] "quince" "raisin" "star fruit"
## [25] "ugli fruit"
## find all fruits ending with two consecutive consonants other than r
fruit[grep("[^aeiour]{2}$", fruit)]
## [1] "blackcurrant" "currant" "eggplant" "peach" "redcurrant"
In the examples above, we return all strings matching a simple pattern.
We can specify that the pattern be found at the beginning ^a
or end a$
using anchors. We can provide multiple options for the match within brackets []
. We can negate options within brackets using ^
in a different context. The curly braces ask for a specific number (or range {min, max}
) of matches.
In the example below we use .
to match any (single) character. This behaves much like ?
in a Linux file name. We can ask for multiple matches by appending *
if we want 0 or more matches and +
if we want at least 1 match.
## find all fruits with two consecutive vowels twice, separated by a single
## consonant
fruit[grep("[aeiou]{2}.[aeiou]{2}", fruit)]
## [1] "feijoa"
## find all fruits with two consecutive vowels twice, separated by one or
## more consonants
fruit[grep("[aeiou]{2}.+[aeiou]{2}", fruit)]
## [1] "breadfruit" "feijoa" "passionfruit"
## find all fruits with exactly three consecutive consonants in the middle of
## two vowels
fruit[grep("[aeiou][^aeiou ]{3}[aeiou]", fruit)]
## [1] "apple" "blackberry" "blackcurrant"
## [4] "breadfruit" "dragonfruit" "huckleberry"
## [7] "passionfruit" "pineapple" "purple mangosteen"
## [10] "raspberry"
To match an actual period (or other meta-character) we need to escape with a backslash. Thus, we use the regular expression \\.
## [1] "umich.edu"
The double backslash is needed because the regular expression itself is passed as a string and strings also use backslash as an escape character. This is also important to remember when building file paths as strings on a Windows computer. In other languages, you generally only need a single backslash in your regular expression.
Matched values can be grouped using parentheses ()
and referred back to in the order they appear using a back reference \\1
.
## [1] "apple" "bell pepper" "bilberry"
## [4] "blackberry" "blackcurrant" "blood orange"
## [7] "blueberry" "boysenberry" "cherry"
## [10] "chili pepper" "cloudberry" "cranberry"
## [13] "currant" "eggplant" "elderberry"
## [16] "goji berry" "gooseberry" "huckleberry"
## [19] "lychee" "mulberry" "passionfruit"
## [22] "persimmon" "pineapple" "purple mangosteen"
## [25] "raspberry" "redcurrant" "salal berry"
## [28] "strawberry" "tamarillo"
## [1] "apple" "bell pepper" "blood orange"
## [4] "chili pepper" "eggplant" "gooseberry"
## [7] "lychee" "passionfruit" "persimmon"
## [10] "pineapple" "purple mangosteen" "tamarillo"
## [1] "lychee"
You are already familiar with the use of the grep
utility from the Linux command line. For more on using regexp with grep
skim the man page for GNU grep
focusing on the following concepts:
During a Tuesday activity, you will work with your group to solve the intermediate puzzles to help cement your regexp understanding. To practice, solve the basic puzzles at Regex Crossword – you will submit these as a quiz. Consider working on the other puzzles to enhance your understanding of regular expressions.
“String Manipulation” (Chapter 11) in Matloff’s The Art of R Programming.
Tip sheet for regular expressions in SAS.
Some tips for using regular expressions in Stata: