Home / Questions / 255451
Accepted
Alejandro Carrera
Asked: 2020-04-18 15:06:42 +0800 CST 2020-04-18 15:06:42 +0800 CST

Extracting alphanumeric characters from strings in R

  • 772

I have a database similar to this:

library(stringr) #Para la función str_extract

ejemplo <- data.frame(columna= c("{ em , 680/3 }", "no 659-11", "no funciona la 2507", 
    "a-4-2 no funciona", "p-8 con presencia de arena", " no 2-5c12s si sirve",
    "mecanica no 1ty -22s brinca"))

I am interested in extracting the part of each string that is:

  • Number
  • letter plus number
  • number plus letter

and, in general, strings referring to "identifiers".

Using a simple regular expression, I can extract

ejemplo$col2 <- ifelse(grepl("[0-9]+", ejemplo$columna), str_extract(ejemplo$columna, "[0-9]+"), NA)
ejemplo

                     columna col2
1              { em , 680/3 }  680
2                   no 659-11  659
3         no funciona la 2507 2507
4           a-4-2 no funciona    4
5  p-8 con presencia de arena    8
6         no 2-5c12s si sirve    2
7 mecanica no 1ty -22s brinca    1

which only has a good result for the third string. Honestly I feel lost and I don't know if there is a way to do everything I want in one go or I will have to build several regular expressions to achieve it. I greatly appreciate any guidance on this.

The desired output would be:

  deseado
1   680/3
2  659-11
3    2507
4   a-4-2
5     p-8
6 2-5c12s
7 1ty-22s
r

2 Answers

  1. Best Answer

    Well, the following expression occurs to me, which is quite complex but I don't see how to simplify it:

    (?:[a-zA-Z\/-]*[0-9]+)+[0-9a-zA-Z\/-]*

    It would read more or less like this:

    • Begins with zero or more repetitions of letter, or sign -or sign/
    • Follow by at least one digit
    • All of the above appear at least once, but can appear more times
    • And then again can come any number of letters or numbers

    Basically, it is about forcing at least one digit to appear between the letters, because if we had tried a simpler expression like [0-9a-zA-Z\/-]+then, any word in the text would have fit even if it had no digits.

    Demo of operation in regex101

    • 1

  2. Abulafia did a great job finding a monolithic expression that solves all the cases you have presented. Another way is to pose multiple more exact patterns with each id, leaving the most complex and specific ones at the beginning and the most generic ones at the end:

    patrones <- c('[0-9]-[0-9][a-z][0-9]+[a-z]',     # 2-5c12s
              '[0-9][a-z][a-z]-[0-9][0-9][a-z]', # 1ty-22s
              '[a-z]-[0-9]+-[0-9]+',             # a-4-2
              '[0-9]+-[0-9][0-9]',               # 659-11
              '[0-9]+/[0-9]+',                   # 680/3
              '[a-z]-[0-9]+',                    # p-8
              '[0-9]+'                           # 2507
              )
str_extract(ejemplo$columna, paste(patrones, collapse = '|'))
[1] "680/3"   "659-11"  "2507"    "a-4-2"   "p-8"     "2-5c12s" "1ty-22s"

    We build a vector with the patterns and paste them into a single string with a |( or). This is much less performant than abulafia's solution, but in some cases it can make the logic clearer.

    • 1