gsub regex to work for both words boundaries and punctuation

Question

I'm trying to search in sentences for both words (case insensitive) and punctuation symbols. The below function works well for words, but requires \\ to work for dots for example ; and thus it leads to unwanted behavior - see below:

fun <- function(text, search) {
  gsub(paste0("\\b(", search, ")\\b"), paste0("<mark>", '\\1', "</mark>"),
       text, ignore.case = T)
}
> fun("this is a test.", ".")
[1] "this<mark> </mark>is<mark> </mark><mark>a</mark><mark> </mark>test<mark>.</mark>"

> fun("(this is a test)", ")")
[1] "(this is a test<mark></mark>"

Expecting :

> fun("this is a test.", ".")
[1] "this is a test<mark>.</mark>"

> fun("(this is a test)", ")")
[1] "(this is a test<mark>)</mark>"

What is the best way - regular expression ? - to search for words as well as punctuation symbols in a string ?

why are you not wanting to escape `.` and `)` to get it to work? — rawr, May 17 '22 at 20:48
Mmh, I could make some if else statements, but it doesn't seem to work as wanted either : eg. ```fun <- function(text, search) { if (search=="."){ search<- "\\." } gsub(paste0("\\b(", search, ")\\b"), paste0("", '\\1', ""), text, ignore.case = T) } fun("this is a test. Yes, it is.", ".")``` results to ```[1] "this is a test. Yes, it is."``` and miss the first dot. — Kamaloka, May 17 '22 at 21:03
I would like it to be user friendly, i.e. typing a word marks the word, typing a ponctuation marks the ponctuation. And of course such that typing a ponctuation doesn't break all marks, as it is the case in the examples shown in my post. — Kamaloka, May 17 '22 at 21:32

score 0 · Accepted Answer · answered May 17 '22 at 21:29

You need

Escape special characters
Dynamic adaptive word boundaries
Since the dynamic adaptive word boundaries are lookarounds, you need to pass perl=TRUE to gsub.

See the R code:

## Escaping function
regex.escape <- function(string) {
  gsub("([][{}()+*^$|\\\\?.])", "\\\\\\1", string)
}
fun <- function(text, search) {
  gsub(paste0("(?!\\B\\w)(", regex.escape(search), ")(?<!\\w\\B)"), "<mark>\\1</mark>",
       text, ignore.case = TRUE, perl=TRUE)
}
fun("this is a test.", ".")
# [1] "this is a test<mark>.</mark>"

fun("(this is a test)", ")")
# [1] "(this is a test<mark>)</mark>"

FYI: If you want to learn more about word boundaries, please consider seeing my YT video on [Dynamic adaptive word boundaries](https://www.youtube.com/watch?v=ngbxagE2b68) on YouTube — Wiktor Stribiżew, May 17 '22 at 21:31

gsub regex to work for both words boundaries and punctuation

1 Answers1