Regex help, subset based on nth occurance from the end of string

user_15758438 0
Viewed 44

An example vector:

string <- "Junk1_Junk2_Junk3__ID1_Junk4_Junk5.pdf"

I am trying to subset ID1 by counting _ (underscores) from the right; so subset between the second and 3rd underscore from the right.

expected output: ID1

My attempt was to try to use the double __, but this is not going to work, because not all my string list has it.

Attempt: (_){2}([^_]+)

Side note, I am trying to get comfortable with regex; please recommend a resource to build and test.

Any assistance is appreciated.

r regex
3 Answers

You can use

library(stringr)
stringr::str_extract(string, "[^_]+(?=(?:_[^_]*){2}$)")

Or, same approach with base R:

## Base R:
sub(".*?([^_]+)(?:_[^_]*){2}$", "\\1", string)

See the regex demo and the R demo online.

Details:

  • [^_]+ - one or more chars other than _
  • (?=(?:_[^_]*){2}$) - a positive lookahead that requires two sequences of _ and then zero or more repetitons of any char other than _ till the end of string
  • .*?([^_]+)(?:_[^_]*){2}$ matches
    • .*? - any zero or more chars, as few as possible
    • ([^_]+) - Capturing group 1 (\1 in the replacement pattern refers to this captured string): one or more chars other than _
    • (?:_[^_]*){2} - two sequences of _ and then zero or more repetitons of any char other than _
    • $ - end of string.
user_38329700 
sub(".*_([^_]+)(_[^_]+){2}", "\\1", string)
[1] "ID1"
user_83802720

I thought about doing this in pure Regex, but I think I would actually simplify this problem in R. I'm going to use some very simple regex and two R commands instead of 1.

str_extract(string,"[^_]+_[^_]+_[^_]+$") %>% 
  str_remove("_[^_]+_[^_]+$")

Here I'm finding a combo of not-underscores [^_]+ and underscores _.

For ease of Regex I'm extracting the whole string. Then I remove the excess.

Alternatively, we can use a bit more complex Regex with the lookahead method.

str_extract(string,"[^_]+(?=_[^_]+_[^_]+$)")

I think the lookahead code is a bit more advanced to read for many coders that don't use Regex a lot.

user_84852870
Related